From 5fddf26cb8bf474b6649bd727bb6493f0c2c48ba Mon Sep 17 00:00:00 2001 From: Andy Belle-Isle Date: Sun, 1 Sep 2019 19:02:29 -0400 Subject: [PATCH] Added SOIL for OpenGL texture loading --- lib/libSOIL.a | Bin 0 -> 121960 bytes lib/soil/soil/SOIL.c | 2024 ++++++++++++ lib/soil/soil/SOIL.h | 433 +++ lib/soil/soil/image_DXT.c | 632 ++++ lib/soil/soil/image_DXT.h | 123 + lib/soil/soil/image_helper.c | 435 +++ lib/soil/soil/image_helper.h | 115 + lib/soil/soil/original/stb_image-1.09.c | 3632 +++++++++++++++++++++ lib/soil/soil/original/stb_image-1.16.c | 3821 +++++++++++++++++++++++ lib/soil/soil/stb_image_aug.c | 3682 ++++++++++++++++++++++ lib/soil/soil/stb_image_aug.h | 354 +++ lib/soil/soil/stbi_DDS_aug.h | 21 + lib/soil/soil/stbi_DDS_aug_c.h | 511 +++ lib/soil/soil/test_SOIL.cpp | 379 +++ 14 files changed, 16162 insertions(+) create mode 100644 lib/libSOIL.a create mode 100644 lib/soil/soil/SOIL.c create mode 100644 lib/soil/soil/SOIL.h create mode 100644 lib/soil/soil/image_DXT.c create mode 100644 lib/soil/soil/image_DXT.h create mode 100644 lib/soil/soil/image_helper.c create mode 100644 lib/soil/soil/image_helper.h create mode 100644 lib/soil/soil/original/stb_image-1.09.c create mode 100644 lib/soil/soil/original/stb_image-1.16.c create mode 100644 lib/soil/soil/stb_image_aug.c create mode 100644 lib/soil/soil/stb_image_aug.h create mode 100644 lib/soil/soil/stbi_DDS_aug.h create mode 100644 lib/soil/soil/stbi_DDS_aug_c.h create mode 100644 lib/soil/soil/test_SOIL.cpp diff --git a/lib/libSOIL.a b/lib/libSOIL.a new file mode 100644 index 0000000000000000000000000000000000000000..45ae424e4a41221574ec0457a2f18b116867cd73 GIT binary patch literal 121960 zcmeFa4R}=5wKqN|Gmrsd&Y)4#N=vNAHldMbG-^ho=8!XR#!i$9C=@hcJ|x!&7-kR^ zA#`S7PKN{B*52BCds}a>z4hMOVq1&&J(Gk4@GB&WAbtcyoFQNWC>g#o?{DpM&g5i> zt@pm~|NlJi^SlF@v(MUluf6x$Yp=ccT5IpQynJC(&7zyeey?Eor^NKCh0Wgu1y`53 zuGW$)7;CXumRKy~Hhs0f$!9H=zxQ{G-D3HFe~bp)eNS5cKmMy(vUqu2ldt-o248j4 z!o~O2S(@&>r`orq`r7NRWx=wtvg-P#CHFHx&64|<`h9iPHA@>+Y?llC%!5Z)ntOlaA1E zreC#m_t)LOr0F43l&@}?uR0?)gPBu4+}LnWb!}bElG-{Dun{a#eXk#UvaI1@z0~tl zXl0n;_xS7UK^{g|Gko!qMRi0t9rfQxV8*a?@yP8r;2Ij^?flkLOh)6=AamRA0i$K4 zOM^1gQGaOVQj(E+XE@xP-*@lA%yQD8%ydKLP?C|#84fq+U%IR|vz#<2Gu==*lw_oG zhQrPI8RHXknSl+<7W}JwU9#fnNo)VhvLs4Zl+vjG7xPjDifapA0p~dDm_0YDu1J| zdYP}O0YiD|l7_{Y(JVEK>T2$*uE&fw-CJGP)U>2&nPui}9bqvPDaL8jvOceLAaM=v$~`#k`8if@)@Yz+cz&Q1z`dZkttIvvBFcdm0)W zd=KeRU)>5MtMOqL$MknzbdCR>I+_a4ivSClfy2(G;iL_yI$Olf+3iS`7QrQ5$GMz2dodZUKsq0DG!Drp%yhFYgJMnhnV z9#)as1|o;nOSJkp4=?L~Q~m!&is9v+pD?4WjFt_nR=U*TI4u@!5rFQPyo4>uzib_T zU1=#>e!EIa%8YgRYPWY<7RO;Zu6;PmO_rfI18T86GjS+f!GaZ*T+0O(aVw=(c8SU< zNr`s%sJEun6)Ca1M>!_)n@>qxhm!10a`7U)gLzl zfuyvk&rkpnO21Up5=+_xhnu;6Wj8{xL3^xc6n8v!+CHXJ*@*{7U$Fz@#&ovU*H7jr z29ecb;U`)=gg!7?&Ax0@r8T& z!USJ-DRALgp zdHV@|PLBabILX4uhsZ6rWyxHhTiNO6H}67Re>#qkZ8zRp+)4{Dbs|*^CB=|3o{q!I z9^(Z?&*@|@r~<;l+0@t#Do1fc)7R^0<0->EEUl^sg zHZ;7CAV=tnsL=@|bKhleNHG6Bw+`y0wk=DLaD6g zUGqGT6nx!Q|s};_FZ_o0-@`ZbV326M!qX}N20F`!j$?8f09XGd85M&T0@A(v?1Gt}` zyHY3wwmAO!FbqDx-zUgU_v*oJ5n${LSM#nDi7g$M85HxefPwwPkO6iL83 z#Vv?(4}mX{11&Z=a5~2)g3BDBdRypRJidSG->3xW>y&jeIr%JJC#QIOR|F(nBB4q1NM9!NBENtyxh z)&mi)OZ13W$V2Ry7G*o(-7cxkI^ujK7TZ6%C8ZC;@=60A6*L;D~@nyvlA_^$Phi0D1L_l@jIDE0tL+ix*J^NQOea zLLsFQyvmy_%Gm{W1fO|@5;0;MPy8!>*LETCFXV)6Ja{Ea9SJ7sByRBUMUs>J_jWJm z!EO3@y1RwsGaH_W-B(r9CM%!ey;dkCgg`c5aNzw$sxvS`o=uX3ycY`a>UMSkf&raWkG~zty9$_?rp0Ocpqoa64x@PP z>qu=LbUW!C!=)8*G^*RRaUt?@ODtuTYf^!vRm9K;?9UFgWO=!E`SFMptmy8s+ol9g zSNIhYWpmP^%s&QVYypX})=Zud?-Vr&igkdz=E!!mOjcG34#g)-P-Y3WidUFrVve9d z?BI-w9e7wQp_BZEN>D@D9qQsYxDZmsl&HLg>hbF);Z+W$_~7LT1WskCl~(muEB|Xr zm>D~j6-#De7LWBfxM__ z8Q_hoblG5_CmGr&q55{O`jXI$X#TcNS(Stp#H=N1r43D5grGMvUs!V$a`1dr@YNk#7gMUJHG6V=L; zS1rC>R&5ntlsn7q+%Z+n`&Xm~GkFD!Tg*H~-Cm_uC{*5))n5t$MBSp6emBsPS2g4argAO{!Bh%w^HHlss+3PLjYvhTH6P&xNL9;k z{0#)JHv*=VBidjY8ikk3Yuadld6j-+Ng-CZI#Y=;AcNa^viEgXIg`Y$-!#*!-dPC= zT|x_Gt7#-+B?B?db}QMyNrE_ElhQUpQg+B&M*~X2kLqWa84FVT>D3{t)3bZ?h_0e* z7Z`IVns4*CT6`DsTO6*Ps}l6C>;lC9SvZ#jm{hdGNm4QU7&>U1ll;RQfvsaK|r ze!+o&_<|kZ9n+{1kdAk6*XARBi-qzlyCAEiBCIG#YSZeVI2!25j+6&6>n5$mZCT)Bu1fhd(8FOgFH&BEyaTx#MRl6hB=T${h`a^B!TM%~ zYGY%axC;B6cKm&azFkT29YLX)p7Qu<{`ff@X5tx7A< zL71RMGS}fL+6GRs6Hk>WdtR(S(pZm;=v~NU;)+d>hh>n5wL&fWtV}FwgFGxZ0+M3iykEqy#Aw(7As{?SprS(qd!74awfxU=jZc^Z!T2MHj0IGRT`gCM z_aGpDiw9)v77R*6YGQKD z8?u#FC-^G(zM7X^#l>lS@LM1nT5RE|XZbBQXLu?Eob7?|RLC~Tm6Ewr=$=-os53k@ z<;7sx<>rowfwR`tw|HuHF)>M^PQ$aF@d*vAunc7Dg?~eZ!!i)A7k0l{! zYCp-%Cn$`vA(v73v#fB0nY^ab4S7w)Bh1S0fUtI;^$yXs!@tF=ob)Q%nv%q4iHyQ$ z$f}q$#yJzD7>iL(c5U~6=2ec9%q8|=VvRCEWbqT50kMeumb}s7S)84v4D^|KjXMZs zTJwC4IOPOQ_lFUJ1Yz>yLxeFe(_*O`L9f!`j8@(OjB$MEBfM!e!#cR| zd6Kgv-m@KctcteD%Dc`ehI+vX`S%i50Tbk{(CQVCgkZuQZOcLO`2_%gCW0pWp-#J& zqA_mzWR-0|k!Ym_!$&U)tG*XFl0J0~W zh}&{2y(elQl0f0YijgfAAiZX|Xilh(B>w z@~Vn`yr<&nZQr>b?>8d?e-#yzQ>hBe^vcQ`3JRy+I=iUAHQ7121d(}^6(8>5+b!G* z2X|RcwtW*?4{}*1uo#QwyZV!0l!rdIYY&!A(!)R0p8}pt+T!^lB%Vdgb*#-2bMpgM z>*}nrxehVce)l-k3Q^7YqYtf}EmLzH8?vV6=Bw7JxdrQOVs7EuQPXov0;8qerCGO+ z%`E{$%q>6)q@0>-m&RF0tIe!+ni-xXdRndCb3!|O}Z8g zUjguViEHpFp(n};WXZV|)&dkxEf&X7nM?IFvu$`mUG*~EBa}Hk_n9o~H*F{r^+3zd zE3*P9qB_3c(-ZhqYQuUfsI=B5qF>lmSd!vIPYU+ z`{MJ!7AyXZ^sxfdNFOU(x1Aqjwa}XG{EyXQKOYE))iUNhI9s!gfagM%a`bsMVww}A z&n!V31EVcD86{k<<1xmM5&dl)PGiOV(Z)xf0>`>02m3r9dN2okKXlcECaS*hxh9p!^SFL1Na*W&!C=ulip!$ECO~Zr{Rxk zjA3Ep5AdsWc!9JNnP|Yzn&8Ar2K)wHl+lA^h5^4BQ7BmEmX|5LWFo8Y28v+sBaKBS9cf`7w==YT%p5d1O|e6>DKGx1!dk5j78cvJe9ba-a^Mj|dg zndM$inRWO#@if|XhY4=3&v;!9Wa9s^2|m$;C$7V(o)k3je_=|0xe3l+gcp1W|F`gG zq!+#fUTlJ2VMQ8^hht^f}G2kWIt3}qI27HtL zK0gEAufGQ~;Kn?$Jp*pYePgOJ@EdY}hG{%foFU(R8E`|sw`IT$`EHrK?4kR83-7_x z*Q7lc8KI`S#)Xr~bZ?o=yfG%fRKhhI{CF1d>7-7aQ;v8JCT=gUX4HV$0$D& zuTj47tUx9T>%%gQ!lBon#*q0#w4#sktU{Rb8{>fD=u@ia&%|q#Ydr5jS|iLd>R;2o zBK_{u^AqozKeCZOBf{uEBW*nbBlW*bLJ=Qhyi)&@2siQ@>s6|IMiGEu|IF-*7Rw5~ z14x`xl```a&dfdn$P)b;ePh3vL7a?NHV2H9zl?-2K7Q?y{?5#=EhrY%H7>1dB6pj= z=WBCa*>#%JO~Hb(U)9%UhPH`db*Rw4nXk=|?a3z`Azeu+NJ+|(Vm#f-8)D!T$8VU3 z7wEjo5X__a9ip;RR?hIhijk(}v8X-9ht8l~Op$UN0rD9ji(8aA4=4k0E|>(X>UK}f zVJ|cwJ!vx^IiNrZ5~n&z>MC*&nFQ0DZsy}Rw?Mmh8o>Fm!Ql^qx8%o3%h0b*9rV@= zs1H+?yIEO)o>iv}GP5BWq~5(GFMOryZ_5hdy-3@E60r{p@t zS9yi^b&Wl@AMHT9=6A|9P^+%Iz~WI$bG>TLIMQ}@%G^gV1tI!9ncK^30?^HIKuj(I zI}y%?!u~b%2tB8Vi@RMP@Efnd!W3%1I7eFNevl|Gb1Q`&sE*6{jjupKH;<{CbhixHQ06(K zLoaFF=a6lax;-gD?;j;DyA)g`D&53rIS49mgm&;7vJe7SG3DJ!jhiIpE&j0=r1IwAV3((SF8FTC!x&WCAia|cpCyDN$BXOQ$M4X~8~UIHIO0sMr^Cp^uoT_}!HNC0g` z8#cVn%3mow1?WrD)to%HGr9>naj5mV!=cowi!`11)vyM_B+2!|Ob5j|Sb^K?b|qK7 z^9ylcurh4B$P;McB-bqA$<;qWt)r3^qX4WxO(-i`i6*TSp2Yj>cpgMc+W62!@Qxn) z6k?u0_;Wnp$Fm>LJ$S~EayuV-?e8%XwdPrgkf9aNB(5dQXilOU0jfV#!1-W?8h5h#t)5eaz0}QGYqD?$)|uB z&OjLs#xTyaCP|a0e>cDRxUwB$(Oa{3pcB?7e)BQ0sKu?`0%LstEvk>_oGH)1E*&on z6Ha%i%lCEt9X3B2h8VVa;Tg>4dz5?2%Y7)ilB>Hl#qMZoFX?vI3)DjBc5^h{t{u8v zBEor!a_|S}kXn(uCPm7!t@Op;+F^EYq_Ox-ZZY;u#Qsl>prUgz} z`QSR_hp`K!Pl*g!EDqzFcb#p>*N0pQ;g-m6xs!wW=T58Yvu71;N6WxOvB4}Ud}r3K z*pqVF`q0(uUW(C zbv@AvyoZ`w(S+)(KznnbrCH4xU9@Y(1sYL64@jx)QaBsFuFf4I7hRDnu1QjJ+;BmW zlvq(e$mHBn)VX4uLAaO_3&XDyq?5wcpxtTzPZ;f*kqh+80sW%l3mO4IKh$G7>amZg zIC2$sippzJc%rDZQWd(X3TN~xyeBDpN2o%J5`#psYSpm1UE5dwJWWl__m$rax0O{7 ziL=zHDVj@Xd)3-?*|NICN{)mM_(nVmmk_3qfTIx^LDaj>1v;ARWqxs#EgCRj0Y1+E z5ELYzza27WG0d7UbI-+J^-ee}%B~Om`{lqko7eS;|7n=29hgFyFAkLnA3F4HkjP?T zMr-9ADm)*3{i~MHQGMoVLdU|{kPV7@gpccvO3I1j{CJo_Azwmwv(EoOe3{7dGF~k3 zB=M;8gEj-d-WHhNRti?u2sDPZ10b7loSzF%0jdMDg-M5xHnCL z3u-ZZlKb{y`nYwQ!AEe8nzTouRYMv%gy7rwdtD6A65xdZCt5fLrqhdIn*cz}1YlJz zj;0}=K$}erOTwVX^*$f^l%M@f`T9DRJi*cP~v?eVY-=!eM6%L(Q6oq z7)9p5>C8l@VN?l{a%bYZ^%N~PW43AHpf#CEuX*-P>^eQ2)c+-2mv%iR?M*JR7EqtfHhoN*+ZWEm2eZIWzk0!9O#;0uT^@L^*MPa4MrkCFn5!aDyv zfN*BrDV~~Y#;tc zDtEu0-l5w@*f;|0m=5nDgA_h&Jm8sy_Kk!`w;Ae6EM_(=Vdg;k8p(F3n+^ZRN94b% z4e)>eBzBHjaNb!Ob?ll+HKAR;1rF(r4>@ zKycDq8t{uva9RNw@Hcfhp)=dsey8&R;W69zUeV!{-fUyKN#_GfZ?>r|Gr`Sx*6Q#~ zJpXA*Z*K2t6WrXczvz5Mc+57S7j<|h9de0S_+-}eHzv5bo`2Qhnf2UbN^h>`VH4b3 zPbaHcTeqQ~*Bz>PLz1j9RU`kIx1Aoc{H`lWOY3M_Apg-fYLJt`51$yM$ z8E`{xeVzgD(fNrq4fHYa7;A7=MH_)1L8F^BcCXOq-LDKR;cP7NGG$ALDrw!c?cs{44bOXVS}Pzw!LO zp1%Z7^Czt5&%|rA-+0n|ZiLPG|M@G@uR+hRcT*b06Gr}w2&3Lc!de7~78M4-c@fPt;kMgc~ zhevrceRN?C>Ia(uj+)jA6~mOwGGd^Ng}?9uItEli7N!F7hc6qlec<4k$iI4;`{bn>BN=q?eb zo7^~S(1WVdGpawI(4WohiF&wEg$nvAa4R;Uev>EAk>v@rLrvXg^TM80z=s;JrKN)d z#pJt=<1zEKisQsWP^dtQs1^B!t}60~J+dr6GERzA=17t6+i+;9A^M--n1+)N97v@a zQa6K(q=rXqqriP!yVS5OXA;Ig*Ac07NNOHe6cb|ya$@^)l%0~&sjkO-%vp+(MaQSA zFTotik!Q3$Qb}jraw0SHpw(=McYyh=l{S&vB{dv)uK+JIa~h)max9gau1>JK8}@aA z5gWQvyK!2lVMpsJdPINKlS)b49_S=-2*ECL?^ddLR-l8W0B^ftFUoOquZe2u&2H{v zk&CZsq0%wKahBNNc%1TFGe`=%-;M$Dbug7&Qb=&us<6GQhnsX5lkNyftafl5mvtp;mrxo5nx+B!Jg4Y_wbl^a`OH}hLVz>$i zY5#6gsVIt+)>v{pbmdIBwfsV77Y+jFzCcZ45TI-k!?SS~xV6Abhb3g3mdNKr&*|R( zG+xr)|9t4*h}qV((fHDoJ>uQ4A!<4`=&YtWFhEFoH4TTZHh1ET-$fehxf{BO_rM6m ze4;ubOXON99mdw?*fcfABBHY3IZ27S0k-2(gIG_F8)y5#48R3tIoLr&Wf6P`ufg{T|j z&XdHJ29L&Ra7bKd!w%w7k?T_PN<{7*P=Bn*9p~PmgLs9N>SNkx5{s0ZAt&brZ3ggYB-g8B@_Atj=7VyUqrT2g=lu9k{+ zkJ6(=AeBS7FFXzu9!ONmi1f z9ji*zhf>ZRq`QOl!j=m|BZS))+t}%<#ATuGSyI?$_bA(C*cm;pqs+vyBIrT6Jw`s)57!R)U>&fE>otE=R99qiJ7TFUnBXhIEA6Ry zsEUN%3CkRcYvpk5#ttd$wUem_aF22t?UQhD1aiTv&L7Q*YDEgh`)p2&uh{`J-dLPK za)+X;$X0>V@KKK0@3#Ao$q>D6F6MQ8=zqtnbhd6%lDkq|bj_!;XIm^+{?~tD0>?!r zu}|q3?wU_ue;vSUpspB`#F6(N8(MH>f4{}D<^YeW8wc2T#SYkU@SK0HWz2T1exa04 zaz-JJAiS2_V||XTv^vto(`cLti}l$9DSkPEwpgDnkg`%RE7q6g>o3I}P#HO=jjHei844VbM7H41-Gc@N9-Oi&p(}}#rR33&$ z9vQ|rBa*_Nns=b8CYAX?=zYIIdE8FY9oGsuFetDUo1`wl;U!FpOpR=p)#6R`S|FIGy^+N3D++u9+N z@dtCwD3^Q5jiVr(mo0)TC6dn)+y)AD+Y93IH9E7l1=`An3*{iu25SG}&O{w}g#x{Q zB#e44l~qAj*rWFa-KITa|2mBR%|VwHpnqkIfOH?{vp)Ww*2g*M>in=<@uV{+^hJKXMi)!oe)NErttaUayXCO5!hHCLzbSQjmv`*LhdK2V~Z3^O} zdWVul<8GbUdgpqTBRLwFQPp)xZ!Ebvn7Pe)$V^qR3G5vnv;6HV9&=|3=UuQK(}-d?Lgv}SYnd3lIvaH z`$}K;r>=J&4ayo7=G~<;CCttqbp0qxf5~-@xIoz5UO{1QU<|H7LFqW|5 zJ|yzby)=VRC|lPl)`9A`%fr~he2 z@OgKcWvF;3F%mjeR#)4_JJ2>sy*-uQyF8A{58JYY0f3)D5q1eW5ln}f2|RxY`eJAT^7W^PQJLIpD- z!YQf(Rlt>6nmkcDyL&{JEy8aMf~>SYkOQY2{`e4Y$;v09#$PWHUm5%vDGw;KLKdYJ zjPO_mRj1yqtgom=fIE$;XuMnb44l<{NObMy*Ix|ebf%d*RkT;>=DNB+7L{K%7a+0D zjSl}Gh>xiR@WKK_(9S@7iF&{HTSMcAM&&f6^l1-s;8#ajUg5UjMe)|u!$H*jamIGNg0Um z9!RB>gj95@J9%RYRahOy{Vg!@O$@ zN0f8W9ZoE|y2zpG4N0AwLZ%MfI`)|Y30tX311Dw0CU_L=znMJN_>ZoO+bMGiFLyu; zB)`V5i-2?GNU0##ls+SK?WsdrI!vEAfzP-V*X#U*<*F0!U_k6%TC7sI=^Q{0oI_xc z0)q&o0)r|4ZY{qEK_qc~bTZs7y594z!*O2L9CB9hDm^AbvvEQ0Yq+_ik6W?LqwF4{ zAByF-{D7kvDjQq%tcI?C0txOrS;g%})Zl$x-@v_5TXu`#J8&=^qAGkx3Y?Piw>7@FxmXc*Gi*L`&28u zG{`rq1s+9BXAROYVx@>!M5C_62h5ZY{z|u-5AOhs+OErqbi2j&X&gg(e&bW-#B`xa`j2oB)GbiW8H86>hoo;eAMHp7dm z1;_TA2Jv?ee>lGXKnhQE2AceC;xIP3^_xt1_5y(YfH^gLi*zjC^$0^zLHVeEwT#pkE!eO%4I+I!agix~_Ic&dSQ*wP5?8r!^BeJcferee1GJe#=Jlq7wq1xxKOdXBAD{u5jbu2UBCJ} zfzzBXH*h-3KVHNfWV67hL+r2#)W>XRqUR5sw)y)g1vCrIH_yAvqo*N|of2acLvj`? zJ|=b6^o0S$H9u_d(?XKp;zsNbw+GH-t;R$!n>@IqTqiX1SI7{ z%Q4WX6jn98N)_azZaB)&J%&gO{0wCp6`;rkFuPK2IM^OIgWD=D(^2bfLKD*bW7>K- zObZ1xmb?*12eI2)1A>#&n{t9K+sl@fcY0^0^OqO+rj$GR&_hTRabu zM`E4rz^!PJEO>ckdzJlc1;FkuFs>_ryB-xm-lycUSAI1g{4wac>Z$TlK6IrC=d*yC zaY~V^GZ9K*pODSEK~nlNBg3`l$HkHkqcUP_FrO-^WhX=lx~UTTVb-PxXZ&}{rm_+j zXmV4I%*KXsDz0X6+Y{3ntQk$glr#kI{o~O1H^URHOgQQ3(Isi!nZFp*jKi2?(o?n# z+SL&T6>F5v-=v>Vl60OU_zP46Ua;l2^IPuV)LEB8VVD)3!QnC;enSXxlKO*ltjFN| zon^=yEORb_tnglvQdh2q-t81AYou__3;?A|;5VMs)IHJ+$jlbDY<*pwHD(a27B7~u zP_96ClA8jC+(EU}!*AhLZ*CTFgmWD5-^GU4!p0ZeaKo2EA0e`P?q0yM+0xqrIpvK^ zci@1D|2uMIdcHhmdO_0#Sk{~9r=nox&U{>V#PDEAf(3sEBS9dU@@GL`A84aHEj4+B zEBM&D#UB3+ItSuMX^lUQQ0!yeh|*@~_6%QcZyHM9l1@*$li0`b-5w!5t_eHS`@Nz1 zMFKt7(d9gJ;VcVA0tW%WU{SO{GJo~7NZ<%yq+RvDMORv8qX4F0_1Tq{Uby^MK;nE9 z?+1GzNjCG}?fq>y5J#>KmKP((LYghyGl-jWwPmc>eK4|-3S(_hPPmS(bVWi`og&<< z)~%=%*B#g8>=m|u6kTo0FJRgTT>f-Gl^mQdmL#kN>1;@ZW$AsvbW@f_AKYOFx<`J_ zns%CB<%zsX<*2s|;MU&wm9(VA#&10t|%*tii-x>z{r)z-vfQ9s&wm(^e zf1%x~{5bZitmZ|>m0kREk5$qZue(pYqWn*;cl|$13E2`>D~{H>*ToR^;F z4-YQ6qiMHF{c8%V?b%-C=*;*`6r!~Dy7YQ@px-F)KnapBD~I`?Vs42xpBd|Of`6|H z-v!~n4f@$Hd_V9ipT_f9Ru3s_NU7My1~0N9ga57eEL_(UPm8`Gyc$yA`zGOd8>;I- za;$HB8qGTlP49>7KBPF%q;Z(ochk^WB|3x@II?Kiq=*|i*9`7M-+*a}hc}X|!6=pb(4)G7<0Mtyj_`%ayJt))9a@(q0XZ;CpO|4)P z14qYKH5}XoHl+&>L(fnZu+FvLqOQE6Fn;VPi4Q)w2W_}jee#MztUOeAs!H|QxX{s6 zm$CIH!{5fmgM=S5WU9(_$bVQYI!cY&coSID8KvN3@babgD&eOKFSUIqA0iKH@XjKG z*f8*{8T^1~6x@Vw&A#TZIBy&JoX$6TeP6m_5j=w8H|bTTl>t?G5!_}HGAg02Euc}7 zI(F1YE`3GTgK|`M|pE7OAIT9o}AbzTi^gc1}7uGPWWgN zyb@(uoJVC`X^ZsTE0^Y{`0t?R!C@YJ%2R$^M@apM zYHT8~5-m^pa40TH(@6!JJK_%_B9NTw8;j$9y1kAT?h6>W%8&bO`{I8AQ`7Si{kcqk z{)_(nDV`wFfxf`$Lf_axGRHS6kZkn-fl>LcVe@l58;A*43Tlz327TkGSc5+Rjstw< z0H_*>bAe>mgFTe?3O&zd`m+^@2tH1Kx9ZO`fI~t2G{yC4;U_5kQ-oC+b{JNn!2SVW zBybqV5bg>b7(fRY<&%7k-;13^S)U&mewXm;<5D`3yN0?C!Y z3aMyT?$P|s^ea_0agwl~%gSyP_y%XA3($1zjX7+#tglduxy#o06 zRq3cjWEx+udKCZ=L$+4yGkajxQP?*$Xr)5$C9L0X>_-CsEn=X-DzK*E2~Wy zLpaEEVfTSCoq?ELnVzTI<_Mh1U)9}8L^I~iKhf-c{tnbM84$nERE}xHc0?X@xZYg( zoya43OmGyv*?mA+=Ag-2smz0>3vF;*o}8zT2U3H@Z$phE%N(T5>g~|vgL!=vSnc5X z4KJXTQpBByA*L;ip0icnNaj4>6|^WK4PjB0gMThq$#jI>aY=FK5y5C>x=F&?_Nk9k zsVM;&3HkNpd-?(`Z1nMg%bu|ecLjes0nw|D1qL0yMT}31j>$EKEXxbTM=6!CihmA!QjQoqGD=cv z@>qh$kidO)1Dna6a( z^{`47R4JdeOTxD-q#Sqjmay>xpB2(I1>#m@#O107vct2AQOI#9$6>x#zMm($-d*)I zi1*15O{+U;ER{ix0{56+$PjQK?3gc*W1pd9ofOxmQ{c0EMx ziaAf#E55OViTm(o%$rc?a!6EdvT1RK5!R^J2c8M#zEq<67L$ zXnIP_Vr4M7yc81`)qB6GJ@I)2taK#mv4|)BM*C!dB*;wy|CoU*7`%kRYZ!cv4p(+7 z9D3nwVkbCNiD`7`kj9>+MkW3WB^k2hDx7*SIQuKuC)k|1=13S4j&7rsWq7*GkRhh$ zNLY}U50fz~kjNR0_aCPv!$zYgsO76Q67pq+rhf5px)@`z1ILLk+~ zk`tXBMIE@5BPU0B1f2H)6^w1drThjT#2d7U#YDH45vCV`Uhpi`pZWTew#eiSQLfyF zQmlMcoZmbJ*B-ECqG6x;0woh}?5pWAS+ZBt?!gK0jWCsAGxTi)v6Na!QFQyegY<<< znFRnVGyGgf(6@;iRQ4I8s~C1?#f#HnE?$5zfMf^D@j+wuQ9PF*?QUzz6WC^}a((8H z5;po7gB%R(!Br=Xn=`4Nmv z5tTDaa$t`XnP~%Upd-bM;lzR?aL7uW;=mDoe#3Uw8GF3S$8P1UN9i!+16)h6Am4Bw zqN|j1Ze^hNc5VF(zo43Sw_0z-1<31{hV$6-0a2aC$&a%)X?>HZ-T^1RpB9MAy zj*%W)DnAZ)t0_2L{MZ91{IRBraJiL|WwDZ}s0*-ILs8eDC?EETkw@W0H=T2J@WEfA zCw1z$AT6k5&B2|tY#+jh1dLhgwsn2%3EqAbA@Fu$BE<~T z>FaXH?L0{L0@H{~Yd87OFRAa=w9#N4Zc8KVa|!1*cwy0iSi1||MbJvE%=i@jd5!-3 zuKv6NPvRRWKYRLUl!nT&7zw1x_~6@+B(yD7r@@z~55w8PhfX1qoc*x14K-5yBeuX` z0Uvw=VH!abQ7oQ9k1@=!@Jf7yOrm<{dx#hscXADW1ZpCsCC-u?qIb-mJsYpaAY_A2 zj>QX#I#G0f!UsvHlZuF=(Khield%86 zE>#gO9qVyL{re@Xh3#ZRcXRK!i*|_^(tn^e4hf%7)K}<=`vkZe+u-$6j!g`16u{i{ z_zVve4)sBaQ5t7@hj=UC6PgZw_rRAo`ct1v*rMOb1k^d zKW^hXujKJT>|s0CiTMg^V28xvPNOZ!c+0z5VT2jKq!9nJ_6{uxy*I{xgLkSvYERYI zzzZVFY;^s_^B}HV6GJ>y6eKTUj3<{8Wsd}DiVlT{l$0-EKfwx?K^OuxiY zLM~$B<#6hKv-=Qyn!8Vhx+GqH6YqgD*{d&7aG4U;F+00Yb9=enfm7M5jx*|eYF>x4 zsKANg(p>mf0K>puhJ#+Z;m?^eVY!&FNVFa@I2Lq~yRH{Pi#4h02tnd*TIj;fj+R1GUIpD7Kwp28#Sjl47#C)3$B1fRN9qY1AB?S-APnXZpZw1CY#@Q zCoM4d4)quI<7qOVNYE6b%@?e{Z^0{;tvGN%8=Epwd9(LIW1sp3lL}L=e-tWUu+|6h z3lGusLrTX9S@BUrFI+r{^{DE%;!=eP>;@V!a4ySt6C;=ZIk<`iwqg2&wLEMqbx*l| z@lH36!lh!tM>tV)Sc4Txp+1aXv2a(S8%Jnfl4lpSYDalco^6 z9e6-t$77FAL0LD>oH4bG5A|!D8&Bde1>13C(;`=M<>rHv@WPf9uV7W9`F;^WUqfP$ z9J^fyzwQ#s0N^12@gxR}zn2bAi2sUV#!7peUX;oJ=-$tDz|)l4D3a`g!{%VZ{bO?q zvay_(Vv#w;ypd8|f)r$5!Gdw)El2>dn9pyfUwtblNed%ihv8*TKkPvJnnhUHk<}!4 zgO)yz(&tlEIqjDvu111kRl!IE5wNYsY>3}93fbsqP73IkA1{w&86R3g5q}oKS_0PRW9t^;cT+!RJr>6pOaz<8z1})+WakaTC0HlG8XwPmx}{! zkW3J0)K7Bgs4pAX^WbekrPwhy=0l84Q0#I9KSWp3ivwNF)Y7ylJF4?DG>Njwpa%Pl zCXckmIu?R@x?Z@HVg5TWfA1?AJW!cGIQ;@Oqv57!=ljr~fHCnYzO;TUW#hU9qN6b0{2^p30_rWC+IVr6NA+z) zTXuStQ$yv2nVX^PI_Ad@fp&rHKi9~f_;=P<20c45>Cn%i6(%BpXUqW&tWg^M*>8%T z^G510^kx^%lq0JPJVj?^>@B!MN~i1iqd8hVxu~mu{2{ASq3Rp)Rxzb171uSYqDg2|s1KW~-(VHr@4T`3s46GI;t1$MaE zEa=s=Pqd>F5TVsajLe2s8gr`Kp}e!}u&Chm6OIJXhP_Axuh=*p!l8Ikc{DF@09Wfh z1CE8a8h(YwMggdR+L4rgHw7O)NclYKU8OIO+%P5_42x5{G}YZdB9}*fnY4B<6d;G5 zB)(UVe*p2i`4L2-o&6F_f)I<^**)5fgx`QL@kHGDo+<9arYq5i5*kX+0cvPHueO7vSveQ`Q}NT4M|L7?$s06@Rg=`GsMMEp_IQ&gAZccj{T@hfX2 z=O`KNV)-p)8UDDt-+Q>Bt5@dxmR zjsh=OWA!9YajJkpt zsVX6z!+#=%sJakin5atcl%1tk-b1_D0Q7KoS@6d+D5Ft*bPhFT;{i7G`{Gw01knLY z>;eEp4+0IkKk^`^EZFjPFPCKXVYKW~v<%}UDh7IZ*)_;Yhxx_xp}){r3bYlP3=AwEHPAveWrmndbL}|Xz?xPc$^H=>Dg|{9?HMfu78ZVKaH?p zDE+8(`oxBzeInJrpH@+<7gsr;5X?(l$Oyyg@4ppS))M0^#H35K#XWl1Cd6;jf}*mGsbH$# zGg1lhnd$JH_@zd&_ISAlj+g4s;&igIctJW`9zTdF90lSk;`m+!#WkNKQB9OTpN=IL z_;RqNp?L!f&OST+E@~Ouk-NcBq_!w$N^VkcL4TPdqu07gc){uEBckdiwW#k#N$J6+ z?aDqNfvL-P6)iMZn3T4cAOk6F5kFjMOLWF66v$4J=EjbNX&HuL`IWTA0wM4_dT!qR znQ}clz($KSNr@{wevVSJQdqy9Hl@ zwz0zc053K$6IK@F8|I@BOqT9cEN(@9w&Nz)0xVU_asL$I{eM5e5Yju;UMxn>M2cNQWh8kZus;H$JYt z!)kSi;YwxsuOkd_^hKAed6!{@*pT`sq`EI!#U%-9zc2L-3bjFbgLw<2!Xg;I3aeIf zIIPER69^K65kY08m9$@DAul7P0&UecqZo8{F%f}gsuK_hw0^l4NibH669)i*D1zKU zJ#GG{NPXzgf9R9uOBMW{o>QF2esDY@n~+zD%g*G!bv=liN{ z0`cg#jeS~LkM1pc4`U<`dh{77(xc0+R{#FMwk!{K#>>TI?u_I*x)N?SE5UAPDy9?~ zAu{9Jgl1M{?L7xUdpFuoR9#NpM*-?bv=hBb{YZfa5fE4}hHhY~K_NyO)i#e&1}8;9 zn<26!sx_hp(Jkn8pg{Urf(?DDoxY*7570c&Q67Y-XD1}SOQ_j?67x!77rLiCmdvJ; z+=o_!%ZrJHp|SuM2XDYQ1INa*kjjbkvQapZ&RJ#Wu^n%O{fnFrgKH@bIDG*%y#&q# zxKKf6_+f#7YdH=r#I$McnExna^3Y|-LX}-k<1sZdJra|VNaGhqqjXXt7XECB3kVCf z1ClY(dzxwiIe;H(6tE~Id60M#b13Tzo$r`Gm+mq9Hk(ystS|7-J#tOEaLxRqF6fMN}kKXog%iVJi} z423ikgv~b;_%FlJ3hf>SO-e#GW3kk{B30n~toerT@*z4F+59j!hY$Ubg|aXR$iIWc zB7QXWVXO0`RtgnBH~*KgQZSGOG#;yOTp3TOWlT3Vft^L*-7%a*9CU*{JFFlzsC0t* z4M!T!JISu>3B>I;4y?wwST_|?m3A8&uA|ujB@oRDke{@Tnxsd-P5@;)uKS^4=g>Vp zg3>PIou)p3${;_mCJ%v(C>eTeRJxQp?!cRDoM3Gcf7b}M#a~Ad=PjWj=R@R3gI~o( zXb^((sb)SzKd^wjIP4JrcO)Fs>Y5<%p*o}^d18k=nNKOuU3N0@;0_5Y(haKZmYW7A+-BY&4_NNFS1os7QkS{yx2QSS zc~xNz21k$*Q*D?(a~^Pt;pu1T=7V05`#_pfEUf&ArV$&2rvo{*MDlPIz^du&E+Z-c zDgyekaD=dD!)bo1Hd&Q$NEG{jO|J#LDqF&zCZ$OH%^j&!754=$p!Oev$)tcox0E7V z0f$)xa;5~A)Id^^RMGCcNK9e_zQPh{ZKkH*{{~5seV&>xjOMD}CpTZ2!#ah`vfOc! zMRqTDQdV;|)Ncy2V}!D6r~iw@m1qN@o}+O#C~Hq_qT;{@BMhd*A0gY(4bl|wF0Vd; zH>_1~UHehwER?1+LgcBx1`R<4YCDr8jo{S}*)=+ww~-l1uG#0U>FM>L@=_SaHj*}t zgfv_$+%LY2p0|YO3yYom;y-yE!vt>Rzy)g?Mm+@H5^@Ln#db4AjHQuRIsH8Gq?h|R z{@~}70*%1RQXqMumOmc92iX#LpviPZvM_O--gx8A34_9h07=t8H-P^Ma+CxKL1}1?dYz33WkQ2r@$MVL46XL3>2B2AG!HmohPeo`)c%#(ot+ zI)?lcV!kh~(%G5Te0Wks(LnPz901dIuI~ zh|nQ%NeqHz8x=xVe&7LRhSre;D}>B7i_Cw4@KN@l(*+_0lZhBploGoaNYXHur0m|K|}NffJqa5bWo8*-vcDrG|~4I5<>LN#{wusMH1^eO!6)t?~s2? z;8Afa7-EulbWms{d4~&fAddb!@=gHGVe*b$n)8oC0*_Xk5K#hti-)#P2+16X0U?Ex zk^>+_Vh=>qCS5MQgTeMUmTqNT(f3N@cf7E5KSs%T>7lA08ek^Nnr)OW7Ij&rV73&s8@QuUg^h#C=gTQ zI@$9{p@%)67W&!qS-}DXe9=oVG7odN#@3Y@$7CO?mIrGQT9p5h3Phuu*%OOxHm?p) zTd}O7yF(C~gXYTATs*z3*&A9@DRx>Bvp*KowW)ga%RPvekczYwcDRKE$?+I6W1C#_ ztgsYeT7F`4jazUez)eL@`?VH?M9=@g4Paq(V0YG=vvd<+$_fq7f4(SP76!`pUrW$(=p<+bEVi>(h~fU;@o9P*P<3sE);VMnTKhiI z$Jca_KHZ0FnZU2WwuP`&G~;p!nAis5x1%iJ(of92>WQo(U6y~bCo&BOE2rff=O2A- zOs_d?`9S%<^=FebPuGWbXw>4g9O6U2CMtoKXkkGVGqc4eM%mt94CTk*Ekx}h4b1D6 z-rrG?zs97NI7|KUvlg0*aUc`tF0^*XuLg+ph!|hyoXq{6GT(!EQJH3gCw*p;t8^4M zvm~QQAm!+w_hY0>2pl?xZY~SpI*-tCl#lxefyxfC;8FjH?z5BcgT^Fg_o%_qWuUU* zZy#up<&LS_S$m*ykGo#uLw`nokGeLBT$s{Gq%4w^ZN|NCqVm2wve@o+9eH#KYAves z1ulLkRl*IsZjaCI0(;Q?B-Esv2@Y>in=;$R2j>Ec>n8Y(Gbqr*p3U&Jky&SB;#-Jdz2J`Av=ue2zztlHX*O|Sz}Y1Zd?C4_O=YNW8|tf_`(~FX zL9vCJFA!btthy>J0G?%l zR)|@;rwQU$RE}bPOFV`WBIBH0@ZpgFh9(%!eRPW;e)$Rt%zVrX9ac&{@CB|0s>IoB zK+fPV8vB%-oNX!Pk@v-c&y~uz>1y50xJD6P;Yp=(PoM+t(i-JR{|)~^!mLyt!Gy-N zxO{L9R4X6^FLzXaPEDo#QlO_Fi4MY3mETxHOa}RBH;?t(p`|cJHO6Q8Uige+X^t+S zBdb*kn-sxm)3CZl#e1)1<2(LSOoha_^N<$km_Q-X!QOTGqN5`a{L|J9zK1&bx5edG ziJdT>;-9^O4w`|kSl16OCfo74fDQPjOt=VgtNn}eB_Tm zgWMXP|3FBD<&75Abk_oI&v15;Z=Qv*Mogv~o&JP~DPQv+m1~Yeo2HhY#Nk%_faQR) zp%a}+KMuEXG>XB=nEkx6;cdX^5c3IYpYqt|&uN%}prUIt)s^4)FdMO{_#~7KLWFi| zt&D$Nt5y6$s+01+xce6PsH$u4nPk8yF%vavRFqN28Y1s@NnX zfuJEtlL;cC1Sf%{vxOe5BOYDC`vM$XYHV9U8zj(y%R0WqsOh#9 zF1~Sk@eQ{pC?q>jTI)zh$+i`=a{wI%>lnIAgI9O9PZxnaL58V(TB&&$J06Di zJGO&gsrfujgr$0)(Y229EPPBL{D7!^QrSmcDCEzb>*`Vcp(;hd{0P3Jr200}a7pzq zNcqlnsDa~Kk=g$FPttO4ul*nNNzpWvwRB5e5ShLINbUR67HM>}%Gv1p>wAL3if8cS zul^9T&J5Jj(O-b>rr{;wmBVTMak%JBiXU}t**`8xePRH_eJ|ff?+ba2u6rmV-ok&# z$)ISXWK&Hf8;);f4pPex1be{&@)(87W!lsBP)65sQapSz+yPx7rB|Je;Xcr#BL;tr zi-q5CuDl)OKy$>%22e^P=gRrSgbQ6{v&iEp{kwx1dAr0AB=>hB2_uIw%`aA;8SG0# z){|MQj#jo%YY;&O z!m^oAf;F;e)mQq(W`Rh4TmL-Ok1s|Xk#lVR!U1PHEzt2lPParQpx2_1pbDR8-Vdb} z4!{?`y%twa-^LQtfi8J+Mq99x?dDwS>nU)C*W(2$R=^3>8efs9nb(q-usl&HBh!$N zzT*H^Z=flTQ~^SM8VJTZ2pPE(lMK9WHx!;0_7)5tI6${9^ka`H!3BoY@Bn*e!f-vX z3kNC;jU|()*$VfU8%K=~EA!u>7oTM%S%xkimqT zj>6?0x-|EPdXBoC3df zDKr`|2T+_~Dif_{Zo`{!Ih)e_Xxd{~i}nEIY~wTcV9^^o0y+P1wi%LV`ff%IgS>Pt zdDZv~WA1}%M@B&*a~PRr7-bfeK_a8Ptuo3V|C0>8BWmd43Jhw^cAfuG^G`H)4)YZ1h4J zF-#$ZR7yubG>YkeKJyDS?ZiBDbUa{j%w51EcO!SO9%=kP_hFIYbSalpw>qtigRs{E_zpbSyhs$R-1zfI`vdPZA_4c0$#)j{vEOYDxtPC#LuvSN zGCZ1yhsi))?(O)CzrtdzaqdH92S7WH&`7 zV78>8x5S*BbFt5M^J;&!&rurA8Xmwh)BfdiyQ#W?%m_K*n}jC{AX-hgTxEoANyA9% zGo%mnT6@IK5W!KjPZg#UZOF=`WpkY~Q#xnbJ7?Pb_GgWYhc~|iFDl!K#>G<{Ez`j> zvI{94o3BAZV8E7^uu6{4N_~pmOu^Lavk!Kb`>2|PqI~!bd*e7Z|B3w3LofiNYdEzu z{igUSgL8GrORUksNNoouG?T+MhesCE+~#QiQ=n_LyG_JtwtErnb+M`h7UoA?4Lf_! z8_XN*pCsxpSg0YUtt}jOU`Ox4QM`5jumd}LzZ})+gIN^YJ)Nz#&e9apgWvvApm49V z{Tsy94^N0j=|yOnreBP!^ny@Nn-S`5n?Deli;n@klR3X5-SC}( z-(U{2{W<$~d}P3De-fL}V*4;`ewB&;Xp>$#{6Me+FZUiE6}c8SbGs=D?aT2T z+o@5zPNh$JW@q3;RMdtye;ApE$Etk78uJLP0wq0GK88GaKjzfK-V4ZS+>A^X*@#SO zgdP{3`>*f!`$PTq zlCH_I=hJb8L=3(C7#{QG$ zc4KlKoJr7yT@cbjU%G-n@K_DwL*oNkm{PK?+zjE2gl_TlPK5!*g9>-@ddHl?+?QSY);R!bNeT}ZFqTRvc*w#b~UnB?` z%+;S!Z$HU|N{}wIA0z0yFrwl#f{}mB+@+K=nSOJW%~UG$b~TK}ilFv!Uxa3DW>zhJlDJ zY$D({N5XsW#c-ae!BP?HqCZHvKAK9DMmIS~v|JPww!Y1noq^XpRL{pI{T~KLkk8Kc zkKSma^r6p{ivZQ<%D)HokqEMHN#DN@l>?WM0@JbOwtZUe!ukk&pJz~RT8ZsGFhx6j zx)3>hJ^If-_@x6Lw%nZ+7c-h!LEg8G2&vus&^n=HbgvqPXkFe&6>w}mNgs&bgFzyD z>o`hpdQIyU`yHf(&=dlvIT?QRdnM@2zF2N5k{{* zX5eK22r6`NI9BC@BeiAPCuX2Ujt()Z2U8G#1`$e?j$wTNpM3uXtLdkF-wvimmy;$` zQ3nM231|su`*XN%TY~60J7&NQ8vD1V@cf8A?v}y~x)dYyzWu-_UOReg46`xqz!zAg zwfsVaI3&YGR-hQ{p6DwNlsVha0fln~=1tfpdZ+qIob9UtNM*I3hI-+|a>EO~y^o|5 zG6WqRuLm!M?WJL{U(TfP7>ql8=v>i_4jlsndhbj++rsNoZ+~t@Sk-=k`AUL6}c_3qNb#iqXj6XhXDE9&m>!fbfIu zEAZP5MWgEI==1(jtTQRWg69v&C>-1mG*k_32)3l_28_6WD#RpEw{Qmd9_hR!!4EXzd`_IIbEiWS1aWJ=GD6<|z)xtJ!a zx1WtfgOV?XVa<{0Hxe)ALTfW%e4}XL%OWh$payMmIy(R8O*m9E;2RzA9!0T@D~bz5 z&_95A(zu9o-P_?A=-+tBTllWC<3)r`+3>-m@f0&)z2O_83m=hvMbC}@0H(8JMj2W- zrREnjYCv9X%z@YZ)rZ7TdN-n+gF4vBCE=xZEFU2E=f_Ns;po_1+DbD|F(dTJR!q_t zg$mWnsS{@;jU?@)xuUD~K@93=!;U*E{)}sOXj7{lT_yOO)!-hP?)T%7%h`S&v>+>mJm-8Yj> zy!U&&)9DyXX>JrYrY+qkC?iWiMs|bAyxybm#T*AVvGZV?*DXOz^yrC9fThX$IjY!p zUrIbp>l*<~x*c;_*l|I?{t2yu8;O25cvKpZ8;LZZfhO5(G{C||qqaKEK+Sn33_z%c z<}PvN#oR@MV7dP4tlRzePY2(Pg=g&Oiv}rn#E;L>DBL4wVUC`+%ZWONf!3p3b3^QA zQJ*jtMI8wlZpKv4!_-bfc0{3Af(?2mQ6{+*7FCJ;-tklJY^U#>2m#2@>Ztwc34|Sm zkU&g4!KNY@d5V?kkZxqDxeeo%spwF{f|BziPeL_O`3j%&I;TGCv+wldvoYSz^a4M| z9WkzOAL$QYw!?Pf1xUM*mr%>+Qe5E+Qkc4QhWUlhd}H)ehWVuV!syS;w|c)CmG{=@ zx6BvtF|pB48|DkxWQf8dUilR0%Gv&l+J*yc$LFquq&q zM$95dcFM+dtsL6qBJ?h@8FmIPh}58!6ieseGfC1I;~JE{qjY&r9+JBI5~S|d$muAr z5Cn2d)SBkN{G}n06;VDJYCrkMc~bIE#!Oz=kJy~81dErI{k$UV=M!l^kw=l)yj`Pw z=4(c1AB;5>>_tS#R!eDHTBI+G*A=r5Y)A9@E#Hsb!{g*3smJ8m_w77nq-5oGX62$+ zf>tAy6u#8DXRD)Nu*7~)23u-J$ouTAj~L;}@hXpW%E(4zx(N44blL`2#qmD5oAS$d z(&uxLXOU!R-yu3usBh>R-?%tfVn2-W?Q&FI4ub^gT}i7n>%>GV+S4;Gi@hxz6m|zK zx==Wd=W)+ltHE(pXZ^x&KSYdk2jBH}o$NDrNI$SvsCu@0j34=G7HrsuiF{O_M_$8} z#R0Ms?1+l-=OOt&Q21nPA12a>M0*}vr3ls0Zex_?r-8A^^5rtidz|e>BoahGFk&(q z_3+rgL_~p&p_aAB>Da&YyvRIcZl)_d!1=!f z3t^{prrQFTltbf+ASp3-&}Y-^#mN7|u~B3GQMkhQh>-tB(cVRGjQ$ke8p!`&nQv1@ ze|z-XGXH-@`JY^Z$p3Fq{(p<|Ke~!PLvU}$@_*MTLmBIpA_rm~rO+s>)rW6*bdx^! z3w=I$%UMF7S@;mdm32Y7y@nT)iMT5%t8~yb;Cx3+M`a=2+h0p#WE^j!9nLeVX2zwEAxa(2LEisF8BIjdn zDTS5GRYaP{$5rz6@H^<$t9M_7Ud*f%AKDwv+o*Pgt8`!2xmbQN52MN-mvl8!zwp-_ zz$OvEF$e(Dq|r{=NpEz#7o3IRWH&kqB7T~Swj|`B_&YaX6lNp75z}j@`G{C7@}Pq< zC_aLL&rUq(SSGK(q3c_5js43V)F7abesp4!(85ElkEvdT60@R!YPPZmLr8;4aW1HB zk;UljqGnVk+D+Sv(;$-Vz$a)st+)!^rIG#k107`1Vw<0Dl+j<^ojIL?vM%T$u;#Cg3T@e29VNAJjcC3J$^!5F?Akz(8<>mZQi- zZ~qI}(<9X`upg;i5cxaehy-=Ehrl5|Q9Y@C``a>2@lBhdX)4l@(v;GX(o;*v?UC`| zN3oH{&l|*ai;iyMWR}t%Wrpc+mtrsYle;FT8Rk*_PSwAO4LLB2Z@-K(GXf@;ae%YP zO#pb(=(fpH&JZmK9nfYrMfu5w8qNNdo!I>^NPj#S`7vr=*gs)1_VqYm?(0kWIXdRA zPnR_fMcoL>&nJShqb^TW_AFdXEvGBlOpUb(xj52p8p7uo#{8 z)Lm=7AMMPibjj?yoWhKCJjZzU8v#K3Uvo-r8Qy2bctg-uu2UBeW1u_RbqTK4FPgIz`LUH80bP z;sayJ+hx@xDkEeQ1wHjRVVd7os#Iegc~mYh$+hxlQZD%Ayj>VdT+kREUu|=*y{Zp5>oEs+yU?wyvm)TN{<|I+0kwd_6DAAH0Rd*DK zH0G|#zog~EZ9U(~UXs3l52njnlEZ3mN%9Ywp_Og4u6B8}u7=e$m~|m5YU3!d_==jQ z;O~>2$4$cE`iyF?U}b`6vh~Z|m!j>NeH^?9+|u*X3MC zH#RIn6WX}0^twr9c|BLcu@4SjXEVyMfa?*u?zf9_m-6e$-yfOFB|)R1b;fB2j!4R4AOYJ4Q-J@f#4+WU_*!@k+M zVM+?ERzv6DUb~$8M7w67q5mnC$)krE+X$4q)6xHR0={t!Z|Uuv?LN+n-nQdKriL$c zpfpbd!O65kVmg+}XLSbLCxYT6Y*p!8MRP7N*!Cex;Kyl@9@no=3K!KmH_&1lObT^w z*o9$2eBgN(n4E@L-|M#GV$`6&_v14?ooJeX`iIyh{!AmBemPMhh>-^J1FoMVYDIVo z2&emdBWDs3y#P5*2k{w-;8bie`G^aNDD08eGvf1}e(X7l+1H_Wj>7gXG<5p{w#iSW z9?t6R=)A*h#c5xM{q*@xzWXNZPr+wT08k$&2rwS(0M+5nVk~O%wEoUwsMk^fq5502 z`94qx>4o$GU62~^-)rwP>`!B2x_y_i^DChcW9K160-FrZcAr9uKNT&rk%)vXLW{vf zDv0=CU_|`FIPs}Q6(x*<4S(SfzQh(au{5y`i%iThsndH@Y62p= zuYwhL!*0WCf3?(n7oX-kr{rhulHLzozR~X*JNNnU&GO!T_?l|DJ7r239oY7#O3j_v zMMskD1KCbvX9R=02EKFSMR2JI`uhR%snYpRN}7nvJNl`e`@Ht2zzBRCkOxa=Y0)JS z(4x!HJ9hSAGx%^;9rlcePKSjZr_lB|^Lr!(K19&FGh*K^pJEF0(3j<-qXOaL)F-HU zjCw6C&8eGzi`m}60p4o>98jyJS`=(Ej9G6{pgJt z{gBikB{CRYXb-lor&k@=XUOw)XwcJo57|>%tYKUQdlDbORes=+XPWaF1HPqqFQ&<1!;w!7`xd-nx6A5` zw`-&u{)HS6^|#ZQ1Q~X^x3k85tOpJzL%NlbZoKq0&8neEBiJNu!|M-|_W-RE*Iq;? z6NYvT9`qo)_aadj(WYF-d1$LO=enJg{}JCaL3li{Er{+3;jAfPN0WGpSHx4yn8T~r z2x`0q>XR8i5%yCYFkg|(zz7bForf%+kLW*W$%ynH>C%6oLp zPzr17;W0e0*%kE|oWrC2&-mOZx}Y)DY1sz_$W_!UT7Y2%+T0Ays)yTPm0l#JN6WA6 zSU^NG0sUcoIhf|qm|LmlmDaa?9cFverzE^HT=)nBh%kOszQoz_JT4*BtaFZnl~1HU z&JJ4Q6MD>j7XkqBABo~Y_Zr-J_HXsd@26n$AT2ZBzd6&{zLE&Ay#Pvu*;pP5ua)M& zGyHn-V+-v?*LFJgGDlX$6d#Wb>z8^Edc7UHIzc!uc=dpezB`2(mDskgx9{k}od&jy z_MVBYLmljFyf!n6-tifQ^gc%H#7Z`j0fSw(N9d23vM1=V zjZ7-`XLXA;Og@Y*!7K+*rekpi?J==#qrdtf!tjffz(#oZcA6N0FE#g~@j!O$?a#z) zA*@ZxceZZ=Z%i6QzBBh>lqM6a*1&688q5oOPxQ^i+(M)4DhG|TJuTw9aI+T^NL*qv z6SmccQba&R)RpeW&`e>#UBB=iqw7L!01JjM^u6#A<}FR#gt>@QKSKcs_Qh$mBKk4^ zFKKW8{--ep=cc9h;VIA4c*OD{y00L1mCEDSj_(n!)8S#iwm) zx8_$cmFt#sN~@ti3?5)wfVmChvuIT}rPvR7I&(YyDFJ(rQMhwy1Z&-wU|S|6L4nT{ zs-!JDJU*~Xqxw?lOGofZ8ogW509OO!z}mQ-&h{$Y_$Unx#2HO3EGFo?a|u<*kf2!B z#AzzNs}QxNN64TaqCa9l3fjboJDB(%!eTm`a}^E!!;h2?`?~GWpvSp-E6sYnrN$r8J=I9KH(;bglSyOoJA{@!r_L&Q{+psUc1d_n_D zHfQ^LAS5Ui4BQW=0mFBN$x&oGR~1p}!|su|hx;D<1GFwbeMxW1uL#q&{D;L`eXHZ2 zLl)#!u%U!A!etI0QVd!9Dc|Ugg-?kT@;ayN!UN2ea4z2pmd5B!G^u0(65nO?zMJ9= z7m1pJSYN$(5E%vA<-w=Jq7V<&o}<}QW9bV#)dTQksrl~ScO2wW?XPmb!iFq+XBJ9r z8hyl+=2vLHyIy>5=b*3f>7~7Dg;$odgVy~ZY%c~dJLoflm{28`U1Am{wv@RWj{{v} z+`VVOisc8`(aNjwmGJrUgE2#m7CJ5 z?|1~Hr4T8f!gWV(gyeDWP+E&L?tAP0xObv*Mto$)!tE~qoL)Dv;EI|6=mYtO{B|ln zorEL}^k>KK`45NZ#gef4g?U{*hY>DK>7wcGHYBfxVXEp4MXl zz|Y^IG@LCyB|TD%QaC$Kqez6}J|&IyRdKIM4fJn`jmIS6|I%F;1-%>f)+En;crLz7 zkmPym0X(O!te6&CN5+rYW9Q?GK%nMtC64E>v9>I7f*@-rWhRqn3+8)cDh!ItRut0b z8eJzDU9G6$ywmq63XsT7*sMtWBlYJC0_s`&pRD)cxS^LW`sopE^-oOSygq~N7rPco z`~3pFDv1Y=u)W7MA`v674zVbg=Py9eITT;G(ppp{PLlD368hFQl5xcKl42>@=3HK& zL~Aml>j}~26we*z)^`KuwfFx`v=ezAgkLWE>R&6Ar1wuED?UeAQGUOLJAeN5^;7?h zK40;NMNTuW^pPxT$O*f(n{QzMSEv>%U{SH4&n>!Wf2+(mM^Pt-0}H4lI#<{jN@5S& zC9MWRe(k$Y)H1SA6GbxyUmomft)(okOd6Y^yhH;Q$kpXo{caARAQZ1+A$mODzSrKQ z)QR6@XfCQCNTEflIxoI5(kAVF82vo!N`3;$C}e*}MHoETPWZc#l8bP8 z!sD?$!FaLPCAui6mID8VZr(#o+U4jaO{NW&rA6rTQ|wQ_n{xk%-nj@B&8j}|9TQ~> zmXt>3Qpq+noWAYEv_vpXrry?DNpY>GD0`5}1Q&O_c+CTSd?(}h<8KfYtx|}}Pqs*& z9xViX*>FJE7mUlAmaq10q|Fe>1@3P}^>)1I4?t@GEs?IwzQ4=|U)lwDNdCRbC(Bcn ziRE;cZ)OCrk(wHA@R?thh%YapGhU_=;z|QA_{O$e;Am7tn!l#`q!#elZE>R+(e-Ji zSWFM60sn{DPc!}1AI-qdW!UF8*{}e@n`xNn6+*^>BYc-OCy4pM>pEV?=gzv$z()n( z3G?0NtKpUx^tAsTt3DIOAUYmjfj#;t0ZF2)A#)3azfCe;BKEaQRzw_eeoH?l zyrB?Q3sptqB-*kQiJ1Fs^9%mR^8*?wY`8j#Z~l<@LVsx^i9L`<|=XRpiPm zjIuAAT5OwBRpXjeT4uxP3f|aM>uP8Wx>{PBni`vfwKW&H^3I<#KbZ9C^nlM+*Rq(x z$;&N>GA4#94rMdEE+MpyuE5MyQ`;1*bIorNK@|{Wt2+OZhSmiYxNIR_xwf+}osN}! z#?;BaC?mgjIy0IwZF;FuK5_gwlC7e-zP6%y-ke(7%!a7kE+U@W*u0=BXsc#r0uV7KSH5f9I9Gl?{TCPcI4EMcqZ1e+*nU zCI((RnuNwZ@wuk@TG%mJEGs@J?ANxDao) zbPoOhzn>eL=Lc)&F0{>A&{R?1SXHB9Sl!qFvkF#36O!8&x4OBu3Xhu_=1GsOo|#CHV~;uYn8kI-bqlH*D%jB|ehQ|{(V||$Z{~5w+tQ9Y zdiXKJMmW+_k4?$2WA!zT#dBKc*Vk0=YXTB~AW}J^l7y)Qa(j&4voRIH}i=4L~0PU#t*t2oEaH2J2_u#CwedhzLg^cP~=|%mL-1@2Xq<)pY)UP#QdROUH z>63J+bV&LmJ&;w9E>6ICGS1U*j>361&hv1N$9XZ%$v9`=tj0-p7Q%Ti&P_P~jx&WK zq969O%%f9Mk2x;eb;|K09K+I2IQ7h&+|g&9mUZm#qcTn$dG`173JS-a=RW-;=gDWB zGiLln7w4b8qX3wZkuxd7H7O$-?@Y=_E2gxeN_9O>Iw&3KAgqlpa1_WN8%)r)ny-IF zzttI;p_G)0jI=46h<-gko5sI{ZRYFCHH_M#bWD{$wlVnqfUm3Qw*ZvV}>ul@8eCadkOn znUOtNB!Qr-`3`hs_bZw1X66UGzX?IvD9zXMvs!uvy|1LWFQ>3Vj)hT`Oc{Zyt?fmo zHKX<+?Vx|+>&n~5o8=JbI7 zSn0T&(gTJO8cr!{Nr)x!0NKh~=G9MWoYd~v7p<|U7AMOm370mH6 z%`aR&!Y|CRBqQ4^3J4L`0xG0&Y)=+f$PrD$HhbzB6xX4$O(EIH20NKwBQb^UR;Cz~ zzXPIdx=d;sWoa}nVhkIkJ(QjnT?-qEmCfFa%GAAyi{>dKa`C!@3?D}c{Sjeb#(t~> z0S82+OGeC$`dUhx0)FQ*(y1l_JyQRK^Zem0MuA>ON~ zp2_)AloO)dZbH69aThJGr;I?JMF9?4yo0bTqr7KJCdPnn7O zoivZAy~y+&U>?^H5AaX9LIyzgkqh}&6!Uv1Epn^}8C*3@Ry?OYZE9o8L7l`qj z&CjY5@+(V$cOAm7o8J-b8%iyoys+xSIY@ zx=@u4rLl)@`Psh`mrmx9C@%N$vqW*};b$vJu2}x_Q@x=iqb#LD&wi+FkZ=E--%(|N z2Ntt#3rHPQMM z*_}MR+{}7)*c@j0{xS|8{QkY-!yH>r@vuyIqIA+lwO7E#My=UV# zpP$vzGgOsOgD%d<@nz&r$|#tW;qhhoCuNjP%9v$jR8GpMo0QQsDPz$jc%?}hopvLm z+g_6KOI)wP^;-MnjP>^7jE8XFV=qS;MrFfl<}c(zy9=!^_#olVa!W)IlT;9sR1ikS zB655Ng76{;Vug_~wEvmcKa+2>j;DxWOpg9&K8Vzn%g@fEXE+G+GA8l{?makXF*btR zYn(2Ywsd(O~R^Akk@PO<+ zekaWDtfGE~1pNYKsS6UwBI|lNCx8)-J<003vGV$|jDeKY%M#c3Mg2VhK8u*&rK}&l zZ_QFi^}*`>%%DZfjoD*p(|y93Ot`iOB;ZEtwb z1WGSP%`*Wk;$6lS=9j*me+jA-RS5$?s*(@7i+Ndnq6g0&;%EQUGKcyxzGfa)|A$9O z56VIg+b5#8*-DgZqo+fURMD%Q2&V>G{6X$2MdKs-o`I3PguA$KQE!qTB0QJz0^HM4 zgp=?l#(yPH*&j@JCievhABSgT4}@1TemT=OF}|Mhsf^QL7txE(dR$jAPCuz))|oU5zRKJ(KchaGo*OWW>#&=dbM#zj9fum_nB ze#D{pKaM-%ukO|Flfa43o?|4v=wrs^cE;DUK%zex7)?F1LkWvG>A7kFl=Q$#6L<2@ z0Dg#nE)$A=WPYBil&R5^p%MSxj4A%v2#@&mu-_AX$e^J%5aBAHi@spsV}KK%9_FL; z?PK~qC&>q@-}g$!*U%sm4)Q~ET%-6i?v}8vig6cZTpS9Y$9NgzqHh<^7c;(t@mlc~ z`qhDxoE}aPFXI7}6BI8G4f5d-eY^O*jp@bu5ZtqQVMA+4zvEP*j>)s0aWOEa;SVvM zdzv7Rj;T=kQ@Cm{Nc6vg-aFUbaO$VE`?INbnJWJx2$sgNf z#w!^YeV}+Y132*!ySn3^>OFMSGkx3H@?P|P;__CeujFvqOl`jqI43UX-s1$jpXohJ z&*o^`$ozMlEBTAQP+aa}ddin{aJIKihGCQ5%JL;XU;fxm00H4WjEjCtJfrj^yt_~m zHZuK}jN2}ffaqt$r32|rdZ^@dU&8b>ZbE)ZIWiaHXEN@BTy(fNe!0wN58te@MP2#%Y8;%OoKB266eAqMspg(Z`3&XMmI5Je+{a zGIlflnt765^b_H7ub@B5Mhms*V71ylXS|njF7|B0z>xGqewPj|@@;1@p2_r7?$U7~ zaFXA3qr4Y=O1S*#2#K$`P2!>+jLS=yzWYvzi#nRTg#YJA@xA&S6D?g$h z-y-m#d7kR<5B?su4a6Z1;3N=H5Po0z=w%34E@o6F}^?t zas2?}vn}}lVVun*`c$m9gC4d<5qTk2$^pN}f^QalhKu@z{^%*@zf1>ly@T;bE%?if z@3P=;F#kRa{w~wMXTkfJUgZn*_#?(WoL?0FIpdWU{E)zhiFq4}R{V;LIm~9|Kb-ON zEqu}$_ge7d8K=cR`jO2z)l>B2492NGq91O?ueadeXS~6J%X;>3Tc-uTK+q45ma{61 zBIa|mg?^I2kB-)F6+Jde#4*hF3k&^J#_zP?a;$Z@jrPIUkE@yfJ_~)Nz=ucYMW}b? zGJTJQzMgTazvxFZ<8N8;MT}eJxrOm8i=6F@=UDKcFz&J7-2xvTorj^zlqzPB>4?pC zJLpwiMSt=@5_|$&RU-OQ)Ke1Se@nurHwpfH68xnk_!~)Z+R#3cJbRPiA11-SNP>Tz z1W!q#pQDrD&LsFrz>h~cyQW&EiYmvy&-i-AA%-|EV!VfOu145)pxi&s_UWLD>A+GP z)0o~>BkxtYU(2}1g5Sh=CF82({wd>47W{XNw=rJK;rB4UgK@6D*q&$HQ!5D-zK`+s zjE9-eVc?0<*8w{uy?N$J!evZ2IZ z$v?oj3imF3rRv1 zZa(8ZjH_^`GVW@Z^eWw}8DG!1%BMFeK8*jGlJ^-m7Lod zAF$wi7|&fU8T^v@e8%_=##OwIJx1y$cZDRJ#q{Sg-pzO|;}y&`v<72#%ab*v) z7~jFTvWEqXyG+UFJmybx04Tl67+2|dC*$iG7vl+d=g*AybV&x2na}f#XL3S`aRNO1 zP~od2p%@Fm<0k7l8O=-YEe^PZK8z z-@%4Z&GgqQK6gvtCB}n{ySSpP(tRc4m5l#{>F;EGfN_;BHtSlc`8Cx+o&t|W8dVFL zXw^{T4HdLhM!Q{DF)kjc8-mQMZK!RYUv0zcp5|Iwb5fgkfvpZ}GiW}&t-8Lps-d;1 z;<~2VdA7Qm=89lrMSV@Pt!`X>MN4%RLY*(J=T$9OP=#weBN9;B>c`bd5^*I6tn>RT zD$1t&%FBOPF=gh|$>k-}rdCvpoiKiU0TwyMty4%umwzEvXpuxrYg20;5)of8{z9uF zpqY8K!2fC!5XTB_#O+G-aCYa41Rnro|TvC;<%DzIkAmMDn)f&ylS zbxndDen~!D2OF=iZLnd*P&F0|RV=7#zP`2@D-{;cpSQSb9uYG(7V8{>n9Mx~%Qk|w zEkRpwUX{G4scGR0v673EKDO*3kwgH7|O?b9`T5f!i1WckPp>TlxQW2iekdx3v(}k!kelZqGWR6$Kv|=b8IZZ zoT`>u=x%ORYkjbyW`0Yss-ZdtNFp2I6-c_~D)pMY5xC%xF#P;{VF?Xm=eIOhAss0B z#PS(+Ay($8r_$iY!v0!T6)-wzM4~Jb|(Vq!ff61$dZnrr7FL)1!Rh9!v?$)=?@uaM5!iVFWOX=83(r`1r6 zmWm3^+Js>X*3k!&XY^LLO6?IjK$$J(25al! z-cX96xODtO@rK4aqyV%hSs#Hw=RkQ`WPOq6Vp>qRWTWne>VTfaR3M@mh4i?qvO=m# z&_y9Do$zH*l0?!eH$7AyRTOFAL?JyK0%PH*JIODmg?L%Z+WXK5?Z)XNawF=U}8ynRl_{xN#@O|Kt)GtXu<^(n5nFkcxRR8am_>+nOuum z0eIZ_imLjix~jxa^CukXDe4$ls;s}6NU1SXpsTr7dlR4(Hukn2U+pa&Cf&qdh^4c_ zV@i!t@%vPzm;JhOjofP7V{;90oQM7aIvhBEj#J^DB=|2hoSHU@p4JZ1LEVa}IEfz} znP^ATp>V#nP3H@u`vp$LXRbz1V*?6r(eP_Ee5r=(^kV-q&{4QLzEz{w`RvkgoqmvU zt9~Y+{Y!@xFJhd+{Si(j|6~oPcf5R0?Rq*0)Ajj*Mo;=we1@UDNr%#>ytCQ9r{TK% zKVY2r>v~?6g#JE_UblyhN$@8%K6-rBe0h>nkMHwnSJFXv1x^*O-5Rdbf5ro7R{BF4 zuG0^55_ie}RQv|*;?{PYhF1bo;bv<%^)V>?bjFD{HPsbdRy{Y8uuJ$02R{CbU^`j(>jiMYW*^mpP^ zcyAK?%_O+(B;qoZ&%`A7LyW6*d`!M>+pOW0z?3{sYq*|{J2d{f9loUDt2I8mdHxW^ z_g5M|>=fMKQ1TQJ0gh8B@o-r2KU(k&oR1#X@JhU=MB?G7RD(!UH+e+1^@@) zci~j@ZzRDNpH9?LPTg-T({SRY{OTu+6ZUJIivI;?;0A|^??@uRQEkEBB&UmGorcfD zz2dV$!}V}CYq&1w(;BYR?`51YJstPB0N_w^zQOWbe`XBV`8R90&i@}8PD$~AUB3T` zhF1bo@j9sCbsGK+z0Ed#3=EKoZtKpTnSN!WWe7=VNm~ktAv49wKivMip zf3q7P2eMqEBRRQO}UBj>~Rt&y*1)ga-OB(dbr;4 zG5RKreyWD+^a~g#OfNTs8ojQcn>1YKb5edR+#hM-p043K{ke=2rrXI_jh^gtvt(<# zNW&|EDLoh({$mZF!Z`8Q?ekuZUXRxU8m`NE$^vn$O z1tQoX;r^X*vWI4zN)I1uxGw*Q3nd?-Z_(&a)Nq~tYz+@;^yh21PJg+Ew`%k=HC(5! z*6@WI{Q?cw>Cx3F4wC-{jeeqr>-38XWAwTmF4b_IehcG--L3H-(CBqO2Q*ygbHPQi zaLI0Gr^@@)8cy`5a=Ysf7JMY*&sp%VkCJ@e)$mgRy}^FF{9^f>^doS(z_Hwd&t`ne zMERU#xd*4RpBES>GQAvmO~dv0-gJrNL-d3{!1CXd1b@PU)3`Dn5erV^%5)4^aHWT? zOXYJ4m*Pv`MWEvs8eWNe6|cKBe36FV%Q(rd`-im}zFMO{q~Z5y_yz+vIEc^fI90gM zYq;(=zDR-}KZ&>u<@1y`hW{LwivN{UVz@5n?-?ik+^Eq%r{Rk=d~&hmL-aRk_+=VS zaZ~k&yBH_zKQ;OkKW=bXai50k`Z?{g7`+~^g&MBYcQa0MF2Skv@CyysJpO4`_V!booM~*VE*=ym!}WCe4-MDD{RiV#{a;op z`CIW;4cGNNa#~CtUC(E0xE}6r87E9n#|JceU2l(RxXx#*#z)tiw=5>lEx1(a-p)Aj z(CJUQB1XSdqd$#tqW_JCpENy2uk)Y6IMLs!(f=$7{eLH+e@Vmd)%ct_LxyYR{}AJ3 z54t?tG1UUQQy41$o;g#}6Q;LMix?+9Z8#O5$r`TnS*YR5H2Rx0 zT&KT7!*A8-*J!v-|GO(ego7|WzV@pGEb93oTq-%!G+gKN4&#LBcK+aO$;XP{OiILI z#S5w>PWt%~P8F_S!}V|%GfvpgHTv5$dY#Yh8m{xXHwnH;!+)poPp!cX4w94jsD4|w zhPMM(^5<%JDsV2wZC(x6>B}^n=p8uesIlM@ML(qlSK)4AoZi>d_g#&?1J4xw2O9lv zG(3$u#c&Xx2XHF-6Eyr@4Zl~z^>BA<_l%jo^g^#FZT;I zTn~4(#z*J>D-9=JivQn|@UhJ!X@=VEat)8OLk-vE`ESN4UR1ww+ledM?DOR>;USz# z&kHqNm-A!B30tGl4{G$+Y4~uu!C}>Bi-s@9a~1B*8m@=CigCj3(dYx$;|2%e{|Bey zGe^U9{cO{4-EUl7Ph6y&q@TaBe%dW~ji9pG9%h`v)$L)6Mo%Iud7jsBdPk-EJ=9pm zq0*&I+}Uh9HGBmy#b>vM>wI=oBN>O{LuCLRD_i9+(cX7*=!{5{R|5C&IHG1No&h~Ra!z(rV!y3L)!&9jeiG$?V`5(hL=~*vdvNT-R z!$TS$JsqD+g1?poPhCvX$nJsk_E z@`}UC|Ar*^&lo4Yb>LL;uh-~*qv3C8xL&>tr^+r4Yq;~0;43v;FL(Z1!+#BaN}d-r zT#xU#TSTx!!kw+*I{zgauJix3hU@&lN`fE1REBHSTQ1|I2VHNoHC)%*at+tR-IfGD zw=Hpe0~)T!w_d~b_=YuH54S%FoJonS7^8%?)ypbl#mQJ1wVSea4|k4z8W4EUk%s8{c95Z8 zwTNSF68dM8(0{Da>w0rjByl)!>U#U3hU@aTGH%t+ts4Drz)R)dpK18B8vbhy*Y&nn z!}V~_rOFTvl2gYgF>aM-wnneZ)1l#dxc{y3S%ol^{D0GM3R~4rHfwxzdCsLu2@b2A z?Md*b87KK~!>PhOhbj#?2=B(J@Fk2Be?1-V(r`T;|EA%(oTpJG1BX>k2UR8*e+|DX z+`J_CU#Sv+!^&p_DKPk~EncfggLV>Mi-?_r#<2Q>P8s?6aa zyceh9b1CB#E+whLr)W66qw?dU8XsMsyEJ-oEsD<@8vSoH{Gf)vt>LM6;06bUsr!x7 z7$-S(zj3aH>vAs8_~?4>&~QE6&on-|9u8@^&fk6~Nh#&j^>7yBB&V*2F&e$D2fv2D zgLjnPW@xyshwC-|x*l%Oa9z${YWS((uiD+m{|YxaNKV4lJcJevC)cawxmCk;{+}>T z*wY&QpoL!9bJ|@X!a?Ec_I!qh6E8(SS;KYuvnZ3}AU=9}jn;5o&R)g|)6?q{4gWKq zsra_sjT;=+c!f3m?|820Q|}SvL*zL|!*zL%W85mwRY~w>jgM|u-5R}aSC479Za?p9 zxUPrdHBx@7JbjnOa3WFh{d08;*V|1JmFYrh(|3G~j;;u!bL0th7wZ>P_(M#eE+0qf z>zO{&f;TaqWx<1tXIpT!-`_|JuJ$-|S@1UI6EN6}dcQHQ3g0EygxfZ;e@q7!u zn(+b){!7M-Ecgeem0A`3o?@g)|#i19WH?qR&$ zf?wJw0j@Si?cDWaiLbQKU&w-YTku~qzS@E}H%Q=@7W^Q~bC(5wkL6io!E0DC_gU~v z#@AZ#4J{IQz=B`S_<9SDHBI7p(1MFeD8L@F;IDJIE(^}LHdaSWpTp!6tNwEtx9X>f zajQIPpTg~ya4UyP;0+6|_9+~&;Mf9694@>3RerCD@p=oc_7S9g;pkBGYG1(jEV$Yi zFrNrDdbQ7An*~?<_U*9XYTv%B)ObE>f4&?GuJ-5iSnz?9<@0!DH{-PL3LT1%+E?#$3$FIh%Vv92^lJaSG7GNu$y;N=)&6z^ z7F_LPNBel_@~D06R$6eikKJPyTpZyD9y2mr1&AEBy?K(|!(gD0;O|%;y$d?GuxBOgz2XCuXDtchyL`Tnp}D+-t!r z8K?am=uq-BF+7HG(B0gNT?~BKRt9@UVSa7v}-gXPF z_P5JQkLR!UiOI3xYM&Ud1y}pTEVAHgpP1DaTMIn_Qf zUJI`FiK(>UYM+?Z7F_KUv)+QMePTYh;A)?k%#3(>)IKp|EO>64q^q>xYZzZ*!3P-c zvEXgXB%e1dcn{+TEx6hzW{gw5r}V7$iSb+Tfm8Vj!WiCJmE*E9WE3$FHw*=)hpJ~85Bi3m#RN9_|clIz6^ zSNp^iSa7vZ%r7mt+9&2c3$FHw$>#Qo;;;6J$+zGxQ|iHQ!PP!7vn{yVCuWTWSNp^~ zXu;J!F{-_&6`&{xP#HxY|Evtp!*6 zYkqFQ)xI+2CtCP(f593H?&5a%V-~!UakYL*$*K00v5jrHaY3+Z4t}w(fcULbSIzi9 z@7N&r%dm~DsR~xv#?EPJv5jqRq#Gj78;iXV^jmC}Al{K!^jURF5F0BnA#Om1twk2p zHh?j{%v(DaU#~o0Y?Z+8*VbXxZqw?c2jMI!zCkCP!C_1cBhIwYWE75F|8ypw($opb>ylG{K{+j z_X2#io{n(^1-S*{s`8knzQB|ey9|f;*=?5$;ZL}~p8m?iBYr$6zB6t9&`l&5ZGw;Kb*J^+jqzxI z7q0yB(LekCkA}Q&n4g+2ckH_5D)Xt1CvUmJYxerWp255xBXltJhKou&o(xX(cMJv# zC|*YBkUe;%uVYs*50@#yYy4puM;X<;9!Q1AI!Af&?PiucuLoDSFEzJ|Xd;$R`oj}t zl*CW}lcWdry#Fove)Rr1O?l2yHsdPeN@J!`K63_X&ezf7TvdY9@|n;0I-Yc{nuu%j zg^pd$&JnmW%>4j6DFI8&Zx9b>`^AEIqZx3g`^;@V=enJf!$s4*W;t|PJt$KGQT3Yd zc+Ep)T|xIq<6%3>dGl@Kp`%WrWZmL5zY=ubu5x$we^~^y&WU*RS^F?ebSYIk)?=uy>dZ@$j1)of~rfp&ol+G_Hz{@pLX4 z^mOJQ6DZu?`mPb$m{MxrJvE%3?YF<`D|~V3wh}WUh4PyF1LnuZLzF|LyynNkI4FE& z8`oTG%r>qyeq>Z|dc-FcnNMC%Jdw1|Z=w5^$8)L?YJ@C16_Z*G$@gSDKoO)x;R*+l=Bw6SP@l27TS{(lg^e>eXl$=160h)@4Z zlkNWht@s~Ng3X!)+ho4_r}Z;)##LT(dx<$H zaVKF+(+O$H8swZfYES z0VBmF5zr;_soB>x+ns4R*KKjG+YJ%2k(vGG+fp0mYXS2xjJ3qvDuDf7wk-?Q^?J!j4OP9fP-=M~X?C+a2>8{?UiL&JF3ucsu>0W!l z{t7%tseOm9@afh^NU%0%$F&ZdjV{xIbtG2kV2ZQ-7*fGON9&d3&3fNWL0NElXrDb` z4)pFzDG3K}cL@LO+^{IsxdG))kHfj4?67m)^Uigzgd%O7bzeI-_`eD5YwK(xR6yVO z3%9mDE8Y`&>007;z?EhhW{-I=wB7DKdaFMibbFk>ZC#JK*Fa0=Cg|27h|S$=Q=RSf z{cz(6%Jrdd?6({T<3)Pl@Yx5wh2Jba_$WOx=vlf5$Nsue_~z0t@yyxr4!v9Wrn6%= zE)cG#@FQpVUNWIRuUYR-dzjw6B3Sh z>&~GFp}0U%>s-DB{+!arMY#~ZH}b) zdCj0Z)7)y9TSM>LaXm)N+UX_(9TPxsUa$FyxzBtHWf_&+?SBUcDKX0QcB%rG!6E}9 zr}UYxLJ@wj&Y2F-g($8C!Jw=qA+%goV$8ZaJlSr9Hrr8ZgTmZ|G7~X0UUhyR+S3{VS1-nGi}O1tjp0 zw7IPNsOBRLct_Z$x9~%++2<_XQV3e-$}=&yP}pLrc>pGv6?ql_`A51DwoMipCgtKa zvejPk`z3Uc7P$(bo$W@vPhUG?9pd_={3JG%r{AUvZkr5bk@}Y zNwv;A|t%+iY}alsKO5K$J^)aAI9%uMFQ_Jt5qgbuljNP#GMob4yYWDpM{b3wMO z#_bq{c92t%IkEV7eWRd&s_ksK&gGKa}&#w9{99v{%y_uun~os?vW?q_7m)T)v0 zl=5(7J>np^M#~$mwc$dw#KF}Pm_lAp=y1xyOA}T@vbAc+TM(bTuvQJ(6sr@NpBd)M z23jvrLpb2h@`uZjuf>;JC=;d&Y0&8KlUJmU1;jNI%1jS^We-jYeU%cNNs;iGd;D;u zW~6RBDmHagyAgk&!hFdB5`e-mH;dcImBKN3;9WjZzEg4T{}bvxiSXNji??HSu1E&> z2#ni%aqY8NaT@cZMu?_lX^FYt-?2aF^P4aE(UzQ8Vm{+XJ99#*xup~p{_{)C=c(O! zj-T6ERNE_ycRG1J{U=F#wBml4AXNfKawTvqsytK))M|htANo-Q6?wa4U5_;HMXipM z@9~>k{Zte6qfH2vDX$4FF*QnTpy{GHS&E{NML;Q$p2{9&;VdZLO9@jq68Wp{MEXNH z&D*uo-Gfx5kg1f0&MT3cv;F20w_hd{9J`-+Ew3ly=JaA2AOM=`{l7+e z%26QK3$N=3N0%QgrLRZLR9Mb$shE?kbzM&9@(p;PWHrp);2I7nGe{#IE)s>$*!99f(!0dsE>S!nC|gnHRXKAzJM zdWp)C{ZGnb43#CBUZ~e8p)CF+vYhfy%JLU@Jd0k?n-pg+GL3>aCX!>tE;3kYZe)ML zd1|UTb%;1C-Rm_QmzEukA8r1hqBTQBebyi-CU zDi1M9%-%>5;ueo=7=q+d0T=hy?+Kx-B+?jGEepQ~m6c5SJ-nf%(!Qh;Tp@OYP(cDx>(5-?|@HEDijn9usm=K}CHnW!d9 zhffYZfRsTK%^rY9%MMQC6k!Jg2Y4d;F`<(vO#TmUH|qbMlm4$Rk^c*jH`+nshrRA{ z;r{~eGPE#Bv|N%XCxC>86GU5xJYl9t1`;fXq@r3k^hq4x6A@>Sq&wSRP2v;lk)O(m zwdhPF_5}&NBej{xNDqQ50&0RP>W}0dNt#@erl9p~SjN{#QPB@1<$^@z@RMdVJ*mnI z|0p8#vDf@asetkU{HKbyHKH9o$QjP|!^vEi6LD72L$4lc6(VC$By$ruREngEoHT9Z zhj=HJ3sZ~9o}OOLJSfdr+0){vv6f7Xz@rGGAm zeistOTtaV%@<0jlMXeaQgfTvd^TWid^i=KxRLco=Z2I} zpSS3Qh3MxL(xd&BhM;WhQDN3J*itaWe>(5&ZF$BY+GzI|e%ku5h(DZl>zJ_P8hejX zxH)(ch6ZZvJ%yWsxrtf#- zWuo;JsD8;9-EVI77WUojBR^XB=FQ_us&|_C?vm>50dq%S^sa#Y6Tkfsm)Fh>`F2ld zehQ6b6mD8N(3S7@*yvZU34%b^kyMS`R2>y_PLg@#O<*!fzW80L(+g@|Awl`fZ)kCyX;%+n*tZ3m&V^!@A(x#HO9@?CP0gU+Z zs6}!0gSxyPgj+)a37fnE32J^3I=pn@Vifyb7j_jp(D3tiUDMPxEzR52nC|Ubn%T8< zq_^wxfEQAy1;VG(%igXT<=(EA`hY2Jp_Htyo6155m$arWJM8t&a0Msj?Wa*eR8OLv z-+pxKLN$xbfjG-mJ{K>@LA(B~G9N5^=O`FY-u`7@(r@c}&;D)R;94W}rK|NfD14T^ zZvz~>1KqcK& zqAGNQ7YXTt{f(q-DF)w!QfWkudJ3qU1((nS$eit;#wrw1xg#SG1oY2I7sEfJM3Y%r zNFbvrgAVPrBSo^24z|#s{gfVbX4CJJkY?nv-ZTvF!GYv@%?L6H`nR{^&eO5q+3^y@ zrlH1;nUDk=W>NbZzd8KtqF)1H*gUr2v9Qc6uywB3MU?_{D}d5V35j6F;C>7uo`tFt z)u?2f4{-$mPlG2l=zM9XaKL63x&|I^!<*9N!muWg<`8KC7}+%5bC zx(~@wpfWMrUFJ1EL)Op7JtBp?M1Ev;zlZWQH4zJ9J$&Gn12KiMdc$>g)*u^+5!zz+ z7S_9SoVSigC=>@f4Tj{RGc&*WK`KT4e#o0S8@RlDQ z#@v^;A6`uK@VJrBb7j|#G0XU&7c4wv176vyEc!ux=KDVLYYaLsEsoHw1Lj#NG||s551C0C{QAg!y>7>akfyA)J+iSyx-YAg-S(K*ddBH)(>dD4N+Fn?0rET>k8y?CnNM8 z^5Bb<5_ng3Zc#TVKb8RX0Wu(Zz68nu(!M706v}8})Km)?l8qj~Q!m($@mDW#kHp|L zL`?Ic-SR!s8*(PP@yU)bCgj}El;X8-^mOKS`cwS&gGS-b8~R9VoZ3kIOlVB1i{8%1 z%N`Lma^8_)u^h|x78y&Sn>%{@{%>#R16@Z|=kZC>*3huif<4-mDnlty5KEG_2o+0S z{xnIOHl+#l53Db*$xHIKXG4x>En_x@rME7FPGL$33nhvaLl$Kn0cR zqPsv9x7n^zL#-S@ul=8YWBdiI{oo0-r2?(cW+@BW#YdnYrqcOI+5?9_}8 zCQKr2uYKx`(oBneIBMdY@k5w8t4i#-z6#ZNHtI=m=vd{U=cHEPqz8$HiF!C&#?<<# z+M~Tq+zyV0XCK=@?W;!bGvlWJf{ICA`IwAxLjRu5lSo4I&Jz_C2j#JYryz{3MLe^o zt>OWs6;~g>Gu3(gRgB6>{hE;a^)1YGpniQ1Gj!-<qrKzCeB5Qj7H?G0dBl-VhKob{1W4RpKi#{#^Vw)pYt1|HZlH9=~zF2gy8;12I6MQC>$Qnct-9@(Li5o z*)ufHrooLMeNd!rA7AL9{!q2GS6bODsd-R^3A&c?6Xw*7IUpJlm1!DGM;Yr<|BUQ6 z?XHw<;UqeS5snAv<5G&wpnIhYopt<^h5N~4H9tFf67!H%+oqaBMo&xKuiWYN3kH<< ztN8rNB_$4;Z^lP4l!odfGvwnhwT~Z>!GQ;x#~(?&e5hf@%`)@QG5(wOJ*%cYV6I{J z7Or2HmO7;Fpgph4cmT?wFdon)?Wp|q?6-3F+w0lK$L=<%MmxyA*;jk~i)P#bx6X36 zhw{Qy=!6v694Rk5Y1X{w`jeP6G$W4(%aqm0s%=wnVw#7>7NIQW?3uC#7s@B{r|g-s z6&v3e+hV$p@n=?!pTIPUOlR3qZ0ty5H>q2<0i|{7#2NNFAYmRyXdcJ>!BKMG;3mc?6syYbK_l5VXFPG)p}FMb82m|DdZlV!MhLO;<7g2+Df$)XFmr`8npQ z`4Mr@cgCN@3HR7%;AzugjJ;pHgUV;;OJw##CQByw6rU3}mn2Er>ekGkxu@7D9SQXm z^3c^yFm3PPW*z@%VZSsn6N^pl#ns`l!$1C`{m8Ax%4@~B*w4H}($A4=7vEt9vf5=n?%5R?Z9OyxgN%_r@e#7L`O|kD}LwkwnDu%j|()9e?uO)xC9pj(ni0J>NX*Bt;PF{VypkT z<*;B;p}h{U!=t6472^o{I={QFR5!#e97EGe3=p+k^+j>q_uz|09*JG zK9V~&GLjw57ix^HZ;|ACIO@d5DqA}uEn<_JAXsK9Sd`C3=F`Kee0s8IBsJt+(dG<3 z;fUI)dhks;QFXahjfF&_k-LxNsxII2{MO%3I>&?W07aaWq_#=TAP}dMDy`qROd&R zE~vxThix#=4q>xm$C{vG`^<_nPJKi5UYv7<&ruokpi*-8V z*3Dsb&GfLkt)>exte(L?a$Ihocj@svWjz$mOo!ELLpb&CvHMv)CL>NkV>t7+%EoZ^ ztyPU-v}a0VIDhxlL|A*vv_yEx&NnoNt1H)s>srGe$gFF_UUqtT363Yi`H)9E#BU6% zTWi*UH%;QpV|b!wO*wH}?lN*NVX9pjAFWd+`IjJhbnjWXMnHEsa|?X50{@$0@~`rr zD0A$ydTfVxO#FUOk8ROomF-a4R3k+d`~>W2J^pz~i1A%p!FMgNL72p*h*4UxnjCepw?=F)yU5Max;T)zT^>kyH)DN z9;p|*ryv8jNWIuO4OQZO)5F^4aDIX+z>1p6)^Odbu%{&)YzaqJhEVtr)Xc0BjTn%a zf}hu`Ye@Qs%2uysHI;)%ZwqoD-jR%Mav2qGIJjQmdm#CLfGIEf0Cd{CxbKmDTq@@4 zZ?m|}J{f2Nv+EKMOUNhy<9L{bDBQ%eeu(my{mNq$G zkH3t0j@Pw+Z$;%tkbueEOLI-tj!HPO6HdGW>qa=Bayd4d0c4#JkWh&g;X?qARdfJW z&8Prf0kJlvYe4b8l}OHmnznPXZ`0NyZ5&z9{?WAYW@+QnF@?nv^BV7}4w`hMRLFfHZ1Q|-MTdID!wGw{O{sG;=EzQjV1gPX1HatH6Kt?o}>zLBatn!W(1c^>9_-a$A&t)@*9SF5*C@3FJz{%uEn}PPrztaO6eh zOAEwq{8ss($6r&v+~ZZIs!a`U@_4o3lgnKmWRZ04_vBAk{#B2^S-Cv($KkEYZN7^~ z&Q@M&)qE?T9wN+7l)O~ zQMpQjjEjqZ7I|C-c{ylCV>G zE6KAMyC-&HgldeLyMwqH_7yEDm2%J!Vjw z4w!K{`2Ucn-^}5G|JdW-R6Wmn{1LN1EqK+lE2gS~>15GvJ`Q=;ck#jhjN!irKdsQK z7XK&yG}t>Gl`T>w5HsN%<+ZwhnXVUum$Q4h%8z*R^0;3)`7J7cugc%4>lPO+2;CkK{mAm@_cOIJxIpb%QH}`aic?R~& z`RBYCzCiWdpnBw5Q-VC`Z}jVl&fNQf_bcD8+}!tpUk6^!uDt8+w8B}mSWo`J>DP>! z`$On?G)B)4WB4;M{MRx3e`5G4$WOWSpBcmD|18SMUlzmH$ME48{=OK#FNS{(T*~V) zJu%z!5dPmOM*ase{Adh+LG=gn5)A}%e~R@fC_iIY-#ujR^}ye*d_>29dQgugEO+wq z(i#MFzk~Hg87e;5XsQkHgI6|nG}X(`Lit%NKX^M+ zN7Kdf^G^A}`?BhvhiNGc{f`ahlX#gG#_BN}lEJ*fXwW;D?%kBc zcrz3w)1#x=(VSlktTw4E{t>PpMX3^MYLv{V*@|YRwvA@;>12K&#k(6fj|4@*B&rmO z)LJySI1#7QLBYV#8cYJL&TLN)<$``p(^+ZRveuGj(A-w)R4$MDDJ}@MjHO4fom7bj zhFa74F5?{}YZQMf7+bTgo-%dY9`I`Sz|cAvSUsBUP4xB2Je|&OykY!V^2m3~C zlOG+}m=0jDMO0zkm)V-|_v<}vT3Y4llw4y=xN2a zlfO~9$jfV!9p6Qq({qCOY?6P4ILpteum>dl?IeGmaVhG z&hkfz%lNdDpCHciZ^U)A1UbxnE^*nE{7G1%T=u)LIQ$+O*Vk{7$W8*C+c)_*<8&n5jg zlRWF+OY*G$PLI2E-bZ@a?je#to9upvsC%%#7f2my3FSnd7U%w*G<^B@sVf{5aRUql_B|Yw) z=)(JmUrzFA;#bD-tucIe4F9Zh$=7+L|6bypultFg4*A_!B>a%{uf@Le&k5pOug=m7 zBgt<+$uA(zesk|=7kT!>29jrfT@3%582;rL?%vTZb~*j;_zM68KmF^Ki$A%Zr%0aj zwV62U-%Y#+x|}~hN1XF}nD_w6^Sj>X6Ms$RC7oQpSA+lvlCKOFXLld*4aDyv&iVQm z;(tx@-zUC__%p;4#E&cY)AIb za$N*JzZWQ%e6ih&iL>2d(!+M&OPuZAMV#~dWzx@fzd@Ys{*Lsp-D>I75F}r0_afqK zw^6yD-&Mrf?k%K;?d~JacE3iP?LJKUIlo7Uvt9R%G{0WCcj5c_UH3-g2IB{Ak4fc{ zU)Fz^*q+a}^%UJnsxyU&t-w);GBw!2UY6~VWgAkKC-5of!*l>7PJN1W{*Aw66_e?gq> zo-Ucubh6!xl>2tyMV#$kPkPwyhl#V@2Z*!XBcxw8U41)Bob8?|gB=LIpWi{8?e-97 zyIJLaey=0Wb`Oyr&hO*I+3u^vxjmkY(cdeBFPg6bteqdeu3XZ2A@N6vb2?uoJy((Z z?6+VCLG-)>i_?D|an@5yoZI{5#92P8T*_r7>0hJsSblkp6KB8Otz6P6U89>H{|fO1 zr2pH*IX%B1{ToQWQwDJmME^xtoFCppob}vAyq4s@K%C`I(}_IMUq|w@iL?Br%Edox zN&ae*uP5F|ob}|0FC_W*5@-2c#5tX}6KDClw?Qd_U%qRJ^Y!qTh;KnW=jVr&SAs1f z{y1^AJ52_I5XA1)Se%|k%0EU|Y840yKjkS~SiG)t(?~e>NdRTsM48J3W-=|#Cc`+7e_b_oz{{?432tm@%epsSh z^t=;`(~}_1dXmJYooaIkHhbLZxktI!-AMZHC(ixVW2EO2lK(g2oSv?^5JC|B?9aX! zJ{H4w#_)YH{0ZXRZ(n$}NsdX+2sWJ`))VLR(U9^=$ntqGN4$~r+(UY}fBPzNuAde2 zis?y^o*Lz1m-U}c@|>^hNuJL~HxXw&50M@&-ye}YpI?4Tob^l-AtUz@h;MTDe1X~INQCS^sxTJB+uvNhl#VEXGss2z; zS~?2^vCI9*Eaj4Z?oZAj&UzY25BvE_l4n0(L!9+|ob+&iau;#-^G`|7Ql!=S;V5y| z|6+`u-;+G|8W%+7UzWz<(gZz-$vrFI;5&mlhT@yB$0 z$N#3Wdk*nbwd3?WruR!1dfYvS_>j-%+VstvX4g)Sd*1K{kDEyth(79Z_dMX!9*?x2 zTJ(le{qA|dKl`^3&6g7On}+72iTFC=0(m9E(EsBX5N$hOo)jLmJ=r-whgnm$ql8+oj*_-;9lN`Cq5`ci*s-{GW=& z#drCa^!T=cg6d1;7v-n(n=`Q8fW?o03YcGZK*50~`NjJiyoTxL|1`*yi+`_7o+0eB KtHQ4z|9=6r@Uw&f literal 0 HcmV?d00001 diff --git a/lib/soil/soil/SOIL.c b/lib/soil/soil/SOIL.c new file mode 100644 index 0000000..44d62db --- /dev/null +++ b/lib/soil/soil/SOIL.c @@ -0,0 +1,2024 @@ +/* + Jonathan Dummer + 2007-07-26-10.36 + + Simple OpenGL Image Library + + Public Domain + using Sean Barret's stb_image as a base + + Thanks to: + * Sean Barret - for the awesome stb_image + * Dan Venkitachalam - for finding some non-compliant DDS files, and patching some explicit casts + * everybody at gamedev.net +*/ + +#define SOIL_CHECK_FOR_GL_ERRORS 0 + +#ifdef WIN32 + #define WIN32_LEAN_AND_MEAN + #include + #include + #include +#elif defined(__APPLE__) || defined(__APPLE_CC__) + /* I can't test this Apple stuff! */ + #include + #include + #define APIENTRY +#else + #include + #include +#endif + +#include "SOIL.h" +#include "stb_image_aug.h" +#include "image_helper.h" +#include "image_DXT.h" + +#include +#include + +/* error reporting */ +char *result_string_pointer = "SOIL initialized"; + +/* for loading cube maps */ +enum{ + SOIL_CAPABILITY_UNKNOWN = -1, + SOIL_CAPABILITY_NONE = 0, + SOIL_CAPABILITY_PRESENT = 1 +}; +static int has_cubemap_capability = SOIL_CAPABILITY_UNKNOWN; +int query_cubemap_capability( void ); +#define SOIL_TEXTURE_WRAP_R 0x8072 +#define SOIL_CLAMP_TO_EDGE 0x812F +#define SOIL_NORMAL_MAP 0x8511 +#define SOIL_REFLECTION_MAP 0x8512 +#define SOIL_TEXTURE_CUBE_MAP 0x8513 +#define SOIL_TEXTURE_BINDING_CUBE_MAP 0x8514 +#define SOIL_TEXTURE_CUBE_MAP_POSITIVE_X 0x8515 +#define SOIL_TEXTURE_CUBE_MAP_NEGATIVE_X 0x8516 +#define SOIL_TEXTURE_CUBE_MAP_POSITIVE_Y 0x8517 +#define SOIL_TEXTURE_CUBE_MAP_NEGATIVE_Y 0x8518 +#define SOIL_TEXTURE_CUBE_MAP_POSITIVE_Z 0x8519 +#define SOIL_TEXTURE_CUBE_MAP_NEGATIVE_Z 0x851A +#define SOIL_PROXY_TEXTURE_CUBE_MAP 0x851B +#define SOIL_MAX_CUBE_MAP_TEXTURE_SIZE 0x851C +/* for non-power-of-two texture */ +static int has_NPOT_capability = SOIL_CAPABILITY_UNKNOWN; +int query_NPOT_capability( void ); +/* for texture rectangles */ +static int has_tex_rectangle_capability = SOIL_CAPABILITY_UNKNOWN; +int query_tex_rectangle_capability( void ); +#define SOIL_TEXTURE_RECTANGLE_ARB 0x84F5 +#define SOIL_MAX_RECTANGLE_TEXTURE_SIZE_ARB 0x84F8 +/* for using DXT compression */ +static int has_DXT_capability = SOIL_CAPABILITY_UNKNOWN; +int query_DXT_capability( void ); +#define SOIL_RGB_S3TC_DXT1 0x83F0 +#define SOIL_RGBA_S3TC_DXT1 0x83F1 +#define SOIL_RGBA_S3TC_DXT3 0x83F2 +#define SOIL_RGBA_S3TC_DXT5 0x83F3 +typedef void (APIENTRY * P_SOIL_GLCOMPRESSEDTEXIMAGE2DPROC) (GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLint border, GLsizei imageSize, const GLvoid * data); +P_SOIL_GLCOMPRESSEDTEXIMAGE2DPROC soilGlCompressedTexImage2D = NULL; +unsigned int SOIL_direct_load_DDS( + const char *filename, + unsigned int reuse_texture_ID, + int flags, + int loading_as_cubemap ); +unsigned int SOIL_direct_load_DDS_from_memory( + const unsigned char *const buffer, + int buffer_length, + unsigned int reuse_texture_ID, + int flags, + int loading_as_cubemap ); +/* other functions */ +unsigned int + SOIL_internal_create_OGL_texture + ( + const unsigned char *const data, + int width, int height, int channels, + unsigned int reuse_texture_ID, + unsigned int flags, + unsigned int opengl_texture_type, + unsigned int opengl_texture_target, + unsigned int texture_check_size_enum + ); + +/* and the code magic begins here [8^) */ +unsigned int + SOIL_load_OGL_texture + ( + const char *filename, + int force_channels, + unsigned int reuse_texture_ID, + unsigned int flags + ) +{ + /* variables */ + unsigned char* img; + int width, height, channels; + unsigned int tex_id; + /* does the user want direct uploading of the image as a DDS file? */ + if( flags & SOIL_FLAG_DDS_LOAD_DIRECT ) + { + /* 1st try direct loading of the image as a DDS file + note: direct uploading will only load what is in the + DDS file, no MIPmaps will be generated, the image will + not be flipped, etc. */ + tex_id = SOIL_direct_load_DDS( filename, reuse_texture_ID, flags, 0 ); + if( tex_id ) + { + /* hey, it worked!! */ + return tex_id; + } + } + /* try to load the image */ + img = SOIL_load_image( filename, &width, &height, &channels, force_channels ); + /* channels holds the original number of channels, which may have been forced */ + if( (force_channels >= 1) && (force_channels <= 4) ) + { + channels = force_channels; + } + if( NULL == img ) + { + /* image loading failed */ + result_string_pointer = stbi_failure_reason(); + return 0; + } + /* OK, make it a texture! */ + tex_id = SOIL_internal_create_OGL_texture( + img, width, height, channels, + reuse_texture_ID, flags, + GL_TEXTURE_2D, GL_TEXTURE_2D, + GL_MAX_TEXTURE_SIZE ); + /* and nuke the image data */ + SOIL_free_image_data( img ); + /* and return the handle, such as it is */ + return tex_id; +} + +unsigned int + SOIL_load_OGL_HDR_texture + ( + const char *filename, + int fake_HDR_format, + int rescale_to_max, + unsigned int reuse_texture_ID, + unsigned int flags + ) +{ + /* variables */ + unsigned char* img; + int width, height, channels; + unsigned int tex_id; + /* no direct uploading of the image as a DDS file */ + /* error check */ + if( (fake_HDR_format != SOIL_HDR_RGBE) && + (fake_HDR_format != SOIL_HDR_RGBdivA) && + (fake_HDR_format != SOIL_HDR_RGBdivA2) ) + { + result_string_pointer = "Invalid fake HDR format specified"; + return 0; + } + /* try to load the image (only the HDR type) */ + img = stbi_hdr_load_rgbe( filename, &width, &height, &channels, 4 ); + /* channels holds the original number of channels, which may have been forced */ + if( NULL == img ) + { + /* image loading failed */ + result_string_pointer = stbi_failure_reason(); + return 0; + } + /* the load worked, do I need to convert it? */ + if( fake_HDR_format == SOIL_HDR_RGBdivA ) + { + RGBE_to_RGBdivA( img, width, height, rescale_to_max ); + } else if( fake_HDR_format == SOIL_HDR_RGBdivA2 ) + { + RGBE_to_RGBdivA2( img, width, height, rescale_to_max ); + } + /* OK, make it a texture! */ + tex_id = SOIL_internal_create_OGL_texture( + img, width, height, channels, + reuse_texture_ID, flags, + GL_TEXTURE_2D, GL_TEXTURE_2D, + GL_MAX_TEXTURE_SIZE ); + /* and nuke the image data */ + SOIL_free_image_data( img ); + /* and return the handle, such as it is */ + return tex_id; +} + +unsigned int + SOIL_load_OGL_texture_from_memory + ( + const unsigned char *const buffer, + int buffer_length, + int force_channels, + unsigned int reuse_texture_ID, + unsigned int flags + ) +{ + /* variables */ + unsigned char* img; + int width, height, channels; + unsigned int tex_id; + /* does the user want direct uploading of the image as a DDS file? */ + if( flags & SOIL_FLAG_DDS_LOAD_DIRECT ) + { + /* 1st try direct loading of the image as a DDS file + note: direct uploading will only load what is in the + DDS file, no MIPmaps will be generated, the image will + not be flipped, etc. */ + tex_id = SOIL_direct_load_DDS_from_memory( + buffer, buffer_length, + reuse_texture_ID, flags, 0 ); + if( tex_id ) + { + /* hey, it worked!! */ + return tex_id; + } + } + /* try to load the image */ + img = SOIL_load_image_from_memory( + buffer, buffer_length, + &width, &height, &channels, + force_channels ); + /* channels holds the original number of channels, which may have been forced */ + if( (force_channels >= 1) && (force_channels <= 4) ) + { + channels = force_channels; + } + if( NULL == img ) + { + /* image loading failed */ + result_string_pointer = stbi_failure_reason(); + return 0; + } + /* OK, make it a texture! */ + tex_id = SOIL_internal_create_OGL_texture( + img, width, height, channels, + reuse_texture_ID, flags, + GL_TEXTURE_2D, GL_TEXTURE_2D, + GL_MAX_TEXTURE_SIZE ); + /* and nuke the image data */ + SOIL_free_image_data( img ); + /* and return the handle, such as it is */ + return tex_id; +} + +unsigned int + SOIL_load_OGL_cubemap + ( + const char *x_pos_file, + const char *x_neg_file, + const char *y_pos_file, + const char *y_neg_file, + const char *z_pos_file, + const char *z_neg_file, + int force_channels, + unsigned int reuse_texture_ID, + unsigned int flags + ) +{ + /* variables */ + unsigned char* img; + int width, height, channels; + unsigned int tex_id; + /* error checking */ + if( (x_pos_file == NULL) || + (x_neg_file == NULL) || + (y_pos_file == NULL) || + (y_neg_file == NULL) || + (z_pos_file == NULL) || + (z_neg_file == NULL) ) + { + result_string_pointer = "Invalid cube map files list"; + return 0; + } + /* capability checking */ + if( query_cubemap_capability() != SOIL_CAPABILITY_PRESENT ) + { + result_string_pointer = "No cube map capability present"; + return 0; + } + /* 1st face: try to load the image */ + img = SOIL_load_image( x_pos_file, &width, &height, &channels, force_channels ); + /* channels holds the original number of channels, which may have been forced */ + if( (force_channels >= 1) && (force_channels <= 4) ) + { + channels = force_channels; + } + if( NULL == img ) + { + /* image loading failed */ + result_string_pointer = stbi_failure_reason(); + return 0; + } + /* upload the texture, and create a texture ID if necessary */ + tex_id = SOIL_internal_create_OGL_texture( + img, width, height, channels, + reuse_texture_ID, flags, + SOIL_TEXTURE_CUBE_MAP, SOIL_TEXTURE_CUBE_MAP_POSITIVE_X, + SOIL_MAX_CUBE_MAP_TEXTURE_SIZE ); + /* and nuke the image data */ + SOIL_free_image_data( img ); + /* continue? */ + if( tex_id != 0 ) + { + /* 1st face: try to load the image */ + img = SOIL_load_image( x_neg_file, &width, &height, &channels, force_channels ); + /* channels holds the original number of channels, which may have been forced */ + if( (force_channels >= 1) && (force_channels <= 4) ) + { + channels = force_channels; + } + if( NULL == img ) + { + /* image loading failed */ + result_string_pointer = stbi_failure_reason(); + return 0; + } + /* upload the texture, but reuse the assigned texture ID */ + tex_id = SOIL_internal_create_OGL_texture( + img, width, height, channels, + tex_id, flags, + SOIL_TEXTURE_CUBE_MAP, SOIL_TEXTURE_CUBE_MAP_NEGATIVE_X, + SOIL_MAX_CUBE_MAP_TEXTURE_SIZE ); + /* and nuke the image data */ + SOIL_free_image_data( img ); + } + /* continue? */ + if( tex_id != 0 ) + { + /* 1st face: try to load the image */ + img = SOIL_load_image( y_pos_file, &width, &height, &channels, force_channels ); + /* channels holds the original number of channels, which may have been forced */ + if( (force_channels >= 1) && (force_channels <= 4) ) + { + channels = force_channels; + } + if( NULL == img ) + { + /* image loading failed */ + result_string_pointer = stbi_failure_reason(); + return 0; + } + /* upload the texture, but reuse the assigned texture ID */ + tex_id = SOIL_internal_create_OGL_texture( + img, width, height, channels, + tex_id, flags, + SOIL_TEXTURE_CUBE_MAP, SOIL_TEXTURE_CUBE_MAP_POSITIVE_Y, + SOIL_MAX_CUBE_MAP_TEXTURE_SIZE ); + /* and nuke the image data */ + SOIL_free_image_data( img ); + } + /* continue? */ + if( tex_id != 0 ) + { + /* 1st face: try to load the image */ + img = SOIL_load_image( y_neg_file, &width, &height, &channels, force_channels ); + /* channels holds the original number of channels, which may have been forced */ + if( (force_channels >= 1) && (force_channels <= 4) ) + { + channels = force_channels; + } + if( NULL == img ) + { + /* image loading failed */ + result_string_pointer = stbi_failure_reason(); + return 0; + } + /* upload the texture, but reuse the assigned texture ID */ + tex_id = SOIL_internal_create_OGL_texture( + img, width, height, channels, + tex_id, flags, + SOIL_TEXTURE_CUBE_MAP, SOIL_TEXTURE_CUBE_MAP_NEGATIVE_Y, + SOIL_MAX_CUBE_MAP_TEXTURE_SIZE ); + /* and nuke the image data */ + SOIL_free_image_data( img ); + } + /* continue? */ + if( tex_id != 0 ) + { + /* 1st face: try to load the image */ + img = SOIL_load_image( z_pos_file, &width, &height, &channels, force_channels ); + /* channels holds the original number of channels, which may have been forced */ + if( (force_channels >= 1) && (force_channels <= 4) ) + { + channels = force_channels; + } + if( NULL == img ) + { + /* image loading failed */ + result_string_pointer = stbi_failure_reason(); + return 0; + } + /* upload the texture, but reuse the assigned texture ID */ + tex_id = SOIL_internal_create_OGL_texture( + img, width, height, channels, + tex_id, flags, + SOIL_TEXTURE_CUBE_MAP, SOIL_TEXTURE_CUBE_MAP_POSITIVE_Z, + SOIL_MAX_CUBE_MAP_TEXTURE_SIZE ); + /* and nuke the image data */ + SOIL_free_image_data( img ); + } + /* continue? */ + if( tex_id != 0 ) + { + /* 1st face: try to load the image */ + img = SOIL_load_image( z_neg_file, &width, &height, &channels, force_channels ); + /* channels holds the original number of channels, which may have been forced */ + if( (force_channels >= 1) && (force_channels <= 4) ) + { + channels = force_channels; + } + if( NULL == img ) + { + /* image loading failed */ + result_string_pointer = stbi_failure_reason(); + return 0; + } + /* upload the texture, but reuse the assigned texture ID */ + tex_id = SOIL_internal_create_OGL_texture( + img, width, height, channels, + tex_id, flags, + SOIL_TEXTURE_CUBE_MAP, SOIL_TEXTURE_CUBE_MAP_NEGATIVE_Z, + SOIL_MAX_CUBE_MAP_TEXTURE_SIZE ); + /* and nuke the image data */ + SOIL_free_image_data( img ); + } + /* and return the handle, such as it is */ + return tex_id; +} + +unsigned int + SOIL_load_OGL_cubemap_from_memory + ( + const unsigned char *const x_pos_buffer, + int x_pos_buffer_length, + const unsigned char *const x_neg_buffer, + int x_neg_buffer_length, + const unsigned char *const y_pos_buffer, + int y_pos_buffer_length, + const unsigned char *const y_neg_buffer, + int y_neg_buffer_length, + const unsigned char *const z_pos_buffer, + int z_pos_buffer_length, + const unsigned char *const z_neg_buffer, + int z_neg_buffer_length, + int force_channels, + unsigned int reuse_texture_ID, + unsigned int flags + ) +{ + /* variables */ + unsigned char* img; + int width, height, channels; + unsigned int tex_id; + /* error checking */ + if( (x_pos_buffer == NULL) || + (x_neg_buffer == NULL) || + (y_pos_buffer == NULL) || + (y_neg_buffer == NULL) || + (z_pos_buffer == NULL) || + (z_neg_buffer == NULL) ) + { + result_string_pointer = "Invalid cube map buffers list"; + return 0; + } + /* capability checking */ + if( query_cubemap_capability() != SOIL_CAPABILITY_PRESENT ) + { + result_string_pointer = "No cube map capability present"; + return 0; + } + /* 1st face: try to load the image */ + img = SOIL_load_image_from_memory( + x_pos_buffer, x_pos_buffer_length, + &width, &height, &channels, force_channels ); + /* channels holds the original number of channels, which may have been forced */ + if( (force_channels >= 1) && (force_channels <= 4) ) + { + channels = force_channels; + } + if( NULL == img ) + { + /* image loading failed */ + result_string_pointer = stbi_failure_reason(); + return 0; + } + /* upload the texture, and create a texture ID if necessary */ + tex_id = SOIL_internal_create_OGL_texture( + img, width, height, channels, + reuse_texture_ID, flags, + SOIL_TEXTURE_CUBE_MAP, SOIL_TEXTURE_CUBE_MAP_POSITIVE_X, + SOIL_MAX_CUBE_MAP_TEXTURE_SIZE ); + /* and nuke the image data */ + SOIL_free_image_data( img ); + /* continue? */ + if( tex_id != 0 ) + { + /* 1st face: try to load the image */ + img = SOIL_load_image_from_memory( + x_neg_buffer, x_neg_buffer_length, + &width, &height, &channels, force_channels ); + /* channels holds the original number of channels, which may have been forced */ + if( (force_channels >= 1) && (force_channels <= 4) ) + { + channels = force_channels; + } + if( NULL == img ) + { + /* image loading failed */ + result_string_pointer = stbi_failure_reason(); + return 0; + } + /* upload the texture, but reuse the assigned texture ID */ + tex_id = SOIL_internal_create_OGL_texture( + img, width, height, channels, + tex_id, flags, + SOIL_TEXTURE_CUBE_MAP, SOIL_TEXTURE_CUBE_MAP_NEGATIVE_X, + SOIL_MAX_CUBE_MAP_TEXTURE_SIZE ); + /* and nuke the image data */ + SOIL_free_image_data( img ); + } + /* continue? */ + if( tex_id != 0 ) + { + /* 1st face: try to load the image */ + img = SOIL_load_image_from_memory( + y_pos_buffer, y_pos_buffer_length, + &width, &height, &channels, force_channels ); + /* channels holds the original number of channels, which may have been forced */ + if( (force_channels >= 1) && (force_channels <= 4) ) + { + channels = force_channels; + } + if( NULL == img ) + { + /* image loading failed */ + result_string_pointer = stbi_failure_reason(); + return 0; + } + /* upload the texture, but reuse the assigned texture ID */ + tex_id = SOIL_internal_create_OGL_texture( + img, width, height, channels, + tex_id, flags, + SOIL_TEXTURE_CUBE_MAP, SOIL_TEXTURE_CUBE_MAP_POSITIVE_Y, + SOIL_MAX_CUBE_MAP_TEXTURE_SIZE ); + /* and nuke the image data */ + SOIL_free_image_data( img ); + } + /* continue? */ + if( tex_id != 0 ) + { + /* 1st face: try to load the image */ + img = SOIL_load_image_from_memory( + y_neg_buffer, y_neg_buffer_length, + &width, &height, &channels, force_channels ); + /* channels holds the original number of channels, which may have been forced */ + if( (force_channels >= 1) && (force_channels <= 4) ) + { + channels = force_channels; + } + if( NULL == img ) + { + /* image loading failed */ + result_string_pointer = stbi_failure_reason(); + return 0; + } + /* upload the texture, but reuse the assigned texture ID */ + tex_id = SOIL_internal_create_OGL_texture( + img, width, height, channels, + tex_id, flags, + SOIL_TEXTURE_CUBE_MAP, SOIL_TEXTURE_CUBE_MAP_NEGATIVE_Y, + SOIL_MAX_CUBE_MAP_TEXTURE_SIZE ); + /* and nuke the image data */ + SOIL_free_image_data( img ); + } + /* continue? */ + if( tex_id != 0 ) + { + /* 1st face: try to load the image */ + img = SOIL_load_image_from_memory( + z_pos_buffer, z_pos_buffer_length, + &width, &height, &channels, force_channels ); + /* channels holds the original number of channels, which may have been forced */ + if( (force_channels >= 1) && (force_channels <= 4) ) + { + channels = force_channels; + } + if( NULL == img ) + { + /* image loading failed */ + result_string_pointer = stbi_failure_reason(); + return 0; + } + /* upload the texture, but reuse the assigned texture ID */ + tex_id = SOIL_internal_create_OGL_texture( + img, width, height, channels, + tex_id, flags, + SOIL_TEXTURE_CUBE_MAP, SOIL_TEXTURE_CUBE_MAP_POSITIVE_Z, + SOIL_MAX_CUBE_MAP_TEXTURE_SIZE ); + /* and nuke the image data */ + SOIL_free_image_data( img ); + } + /* continue? */ + if( tex_id != 0 ) + { + /* 1st face: try to load the image */ + img = SOIL_load_image_from_memory( + z_neg_buffer, z_neg_buffer_length, + &width, &height, &channels, force_channels ); + /* channels holds the original number of channels, which may have been forced */ + if( (force_channels >= 1) && (force_channels <= 4) ) + { + channels = force_channels; + } + if( NULL == img ) + { + /* image loading failed */ + result_string_pointer = stbi_failure_reason(); + return 0; + } + /* upload the texture, but reuse the assigned texture ID */ + tex_id = SOIL_internal_create_OGL_texture( + img, width, height, channels, + tex_id, flags, + SOIL_TEXTURE_CUBE_MAP, SOIL_TEXTURE_CUBE_MAP_NEGATIVE_Z, + SOIL_MAX_CUBE_MAP_TEXTURE_SIZE ); + /* and nuke the image data */ + SOIL_free_image_data( img ); + } + /* and return the handle, such as it is */ + return tex_id; +} + +unsigned int + SOIL_load_OGL_single_cubemap + ( + const char *filename, + const char face_order[6], + int force_channels, + unsigned int reuse_texture_ID, + unsigned int flags + ) +{ + /* variables */ + unsigned char* img; + int width, height, channels, i; + unsigned int tex_id = 0; + /* error checking */ + if( filename == NULL ) + { + result_string_pointer = "Invalid single cube map file name"; + return 0; + } + /* does the user want direct uploading of the image as a DDS file? */ + if( flags & SOIL_FLAG_DDS_LOAD_DIRECT ) + { + /* 1st try direct loading of the image as a DDS file + note: direct uploading will only load what is in the + DDS file, no MIPmaps will be generated, the image will + not be flipped, etc. */ + tex_id = SOIL_direct_load_DDS( filename, reuse_texture_ID, flags, 1 ); + if( tex_id ) + { + /* hey, it worked!! */ + return tex_id; + } + } + /* face order checking */ + for( i = 0; i < 6; ++i ) + { + if( (face_order[i] != 'N') && + (face_order[i] != 'S') && + (face_order[i] != 'W') && + (face_order[i] != 'E') && + (face_order[i] != 'U') && + (face_order[i] != 'D') ) + { + result_string_pointer = "Invalid single cube map face order"; + return 0; + }; + } + /* capability checking */ + if( query_cubemap_capability() != SOIL_CAPABILITY_PRESENT ) + { + result_string_pointer = "No cube map capability present"; + return 0; + } + /* 1st off, try to load the full image */ + img = SOIL_load_image( filename, &width, &height, &channels, force_channels ); + /* channels holds the original number of channels, which may have been forced */ + if( (force_channels >= 1) && (force_channels <= 4) ) + { + channels = force_channels; + } + if( NULL == img ) + { + /* image loading failed */ + result_string_pointer = stbi_failure_reason(); + return 0; + } + /* now, does this image have the right dimensions? */ + if( (width != 6*height) && + (6*width != height) ) + { + SOIL_free_image_data( img ); + result_string_pointer = "Single cubemap image must have a 6:1 ratio"; + return 0; + } + /* try the image split and create */ + tex_id = SOIL_create_OGL_single_cubemap( + img, width, height, channels, + face_order, reuse_texture_ID, flags + ); + /* nuke the temporary image data and return the texture handle */ + SOIL_free_image_data( img ); + return tex_id; +} + +unsigned int + SOIL_load_OGL_single_cubemap_from_memory + ( + const unsigned char *const buffer, + int buffer_length, + const char face_order[6], + int force_channels, + unsigned int reuse_texture_ID, + unsigned int flags + ) +{ + /* variables */ + unsigned char* img; + int width, height, channels, i; + unsigned int tex_id = 0; + /* error checking */ + if( buffer == NULL ) + { + result_string_pointer = "Invalid single cube map buffer"; + return 0; + } + /* does the user want direct uploading of the image as a DDS file? */ + if( flags & SOIL_FLAG_DDS_LOAD_DIRECT ) + { + /* 1st try direct loading of the image as a DDS file + note: direct uploading will only load what is in the + DDS file, no MIPmaps will be generated, the image will + not be flipped, etc. */ + tex_id = SOIL_direct_load_DDS_from_memory( + buffer, buffer_length, + reuse_texture_ID, flags, 1 ); + if( tex_id ) + { + /* hey, it worked!! */ + return tex_id; + } + } + /* face order checking */ + for( i = 0; i < 6; ++i ) + { + if( (face_order[i] != 'N') && + (face_order[i] != 'S') && + (face_order[i] != 'W') && + (face_order[i] != 'E') && + (face_order[i] != 'U') && + (face_order[i] != 'D') ) + { + result_string_pointer = "Invalid single cube map face order"; + return 0; + }; + } + /* capability checking */ + if( query_cubemap_capability() != SOIL_CAPABILITY_PRESENT ) + { + result_string_pointer = "No cube map capability present"; + return 0; + } + /* 1st off, try to load the full image */ + img = SOIL_load_image_from_memory( + buffer, buffer_length, + &width, &height, &channels, + force_channels ); + /* channels holds the original number of channels, which may have been forced */ + if( (force_channels >= 1) && (force_channels <= 4) ) + { + channels = force_channels; + } + if( NULL == img ) + { + /* image loading failed */ + result_string_pointer = stbi_failure_reason(); + return 0; + } + /* now, does this image have the right dimensions? */ + if( (width != 6*height) && + (6*width != height) ) + { + SOIL_free_image_data( img ); + result_string_pointer = "Single cubemap image must have a 6:1 ratio"; + return 0; + } + /* try the image split and create */ + tex_id = SOIL_create_OGL_single_cubemap( + img, width, height, channels, + face_order, reuse_texture_ID, flags + ); + /* nuke the temporary image data and return the texture handle */ + SOIL_free_image_data( img ); + return tex_id; +} + +unsigned int + SOIL_create_OGL_single_cubemap + ( + const unsigned char *const data, + int width, int height, int channels, + const char face_order[6], + unsigned int reuse_texture_ID, + unsigned int flags + ) +{ + /* variables */ + unsigned char* sub_img; + int dw, dh, sz, i; + unsigned int tex_id; + /* error checking */ + if( data == NULL ) + { + result_string_pointer = "Invalid single cube map image data"; + return 0; + } + /* face order checking */ + for( i = 0; i < 6; ++i ) + { + if( (face_order[i] != 'N') && + (face_order[i] != 'S') && + (face_order[i] != 'W') && + (face_order[i] != 'E') && + (face_order[i] != 'U') && + (face_order[i] != 'D') ) + { + result_string_pointer = "Invalid single cube map face order"; + return 0; + }; + } + /* capability checking */ + if( query_cubemap_capability() != SOIL_CAPABILITY_PRESENT ) + { + result_string_pointer = "No cube map capability present"; + return 0; + } + /* now, does this image have the right dimensions? */ + if( (width != 6*height) && + (6*width != height) ) + { + result_string_pointer = "Single cubemap image must have a 6:1 ratio"; + return 0; + } + /* which way am I stepping? */ + if( width > height ) + { + dw = height; + dh = 0; + } else + { + dw = 0; + dh = width; + } + sz = dw+dh; + sub_img = (unsigned char *)malloc( sz*sz*channels ); + /* do the splitting and uploading */ + tex_id = reuse_texture_ID; + for( i = 0; i < 6; ++i ) + { + int x, y, idx = 0; + unsigned int cubemap_target = 0; + /* copy in the sub-image */ + for( y = i*dh; y < i*dh+sz; ++y ) + { + for( x = i*dw*channels; x < (i*dw+sz)*channels; ++x ) + { + sub_img[idx++] = data[y*width*channels+x]; + } + } + /* what is my texture target? + remember, this coordinate system is + LHS if viewed from inside the cube! */ + switch( face_order[i] ) + { + case 'N': + cubemap_target = SOIL_TEXTURE_CUBE_MAP_POSITIVE_Z; + break; + case 'S': + cubemap_target = SOIL_TEXTURE_CUBE_MAP_NEGATIVE_Z; + break; + case 'W': + cubemap_target = SOIL_TEXTURE_CUBE_MAP_NEGATIVE_X; + break; + case 'E': + cubemap_target = SOIL_TEXTURE_CUBE_MAP_POSITIVE_X; + break; + case 'U': + cubemap_target = SOIL_TEXTURE_CUBE_MAP_POSITIVE_Y; + break; + case 'D': + cubemap_target = SOIL_TEXTURE_CUBE_MAP_NEGATIVE_Y; + break; + } + /* upload it as a texture */ + tex_id = SOIL_internal_create_OGL_texture( + sub_img, sz, sz, channels, + tex_id, flags, + SOIL_TEXTURE_CUBE_MAP, + cubemap_target, + SOIL_MAX_CUBE_MAP_TEXTURE_SIZE ); + } + /* and nuke the image and sub-image data */ + SOIL_free_image_data( sub_img ); + /* and return the handle, such as it is */ + return tex_id; +} + +unsigned int + SOIL_create_OGL_texture + ( + const unsigned char *const data, + int width, int height, int channels, + unsigned int reuse_texture_ID, + unsigned int flags + ) +{ + /* wrapper function for 2D textures */ + return SOIL_internal_create_OGL_texture( + data, width, height, channels, + reuse_texture_ID, flags, + GL_TEXTURE_2D, GL_TEXTURE_2D, + GL_MAX_TEXTURE_SIZE ); +} + +#if SOIL_CHECK_FOR_GL_ERRORS +void check_for_GL_errors( const char *calling_location ) +{ + /* check for errors */ + GLenum err_code = glGetError(); + while( GL_NO_ERROR != err_code ) + { + printf( "OpenGL Error @ %s: %i", calling_location, err_code ); + err_code = glGetError(); + } +} +#else +void check_for_GL_errors( const char *calling_location ) +{ + /* no check for errors */ +} +#endif + +unsigned int + SOIL_internal_create_OGL_texture + ( + const unsigned char *const data, + int width, int height, int channels, + unsigned int reuse_texture_ID, + unsigned int flags, + unsigned int opengl_texture_type, + unsigned int opengl_texture_target, + unsigned int texture_check_size_enum + ) +{ + /* variables */ + unsigned char* img; + unsigned int tex_id; + unsigned int internal_texture_format = 0, original_texture_format = 0; + int DXT_mode = SOIL_CAPABILITY_UNKNOWN; + int max_supported_size; + /* If the user wants to use the texture rectangle I kill a few flags */ + if( flags & SOIL_FLAG_TEXTURE_RECTANGLE ) + { + /* well, the user asked for it, can we do that? */ + if( query_tex_rectangle_capability() == SOIL_CAPABILITY_PRESENT ) + { + /* only allow this if the user in _NOT_ trying to do a cubemap! */ + if( opengl_texture_type == GL_TEXTURE_2D ) + { + /* clean out the flags that cannot be used with texture rectangles */ + flags &= ~( + SOIL_FLAG_POWER_OF_TWO | SOIL_FLAG_MIPMAPS | + SOIL_FLAG_TEXTURE_REPEATS + ); + /* and change my target */ + opengl_texture_target = SOIL_TEXTURE_RECTANGLE_ARB; + opengl_texture_type = SOIL_TEXTURE_RECTANGLE_ARB; + } else + { + /* not allowed for any other uses (yes, I'm looking at you, cubemaps!) */ + flags &= ~SOIL_FLAG_TEXTURE_RECTANGLE; + } + + } else + { + /* can't do it, and that is a breakable offense (uv coords use pixels instead of [0,1]!) */ + result_string_pointer = "Texture Rectangle extension unsupported"; + return 0; + } + } + /* create a copy the image data */ + img = (unsigned char*)malloc( width*height*channels ); + memcpy( img, data, width*height*channels ); + /* does the user want me to invert the image? */ + if( flags & SOIL_FLAG_INVERT_Y ) + { + int i, j; + for( j = 0; j*2 < height; ++j ) + { + int index1 = j * width * channels; + int index2 = (height - 1 - j) * width * channels; + for( i = width * channels; i > 0; --i ) + { + unsigned char temp = img[index1]; + img[index1] = img[index2]; + img[index2] = temp; + ++index1; + ++index2; + } + } + } + /* does the user want me to scale the colors into the NTSC safe RGB range? */ + if( flags & SOIL_FLAG_NTSC_SAFE_RGB ) + { + scale_image_RGB_to_NTSC_safe( img, width, height, channels ); + } + /* does the user want me to convert from straight to pre-multiplied alpha? + (and do we even _have_ alpha?) */ + if( flags & SOIL_FLAG_MULTIPLY_ALPHA ) + { + int i; + switch( channels ) + { + case 2: + for( i = 0; i < 2*width*height; i += 2 ) + { + img[i] = (img[i] * img[i+1] + 128) >> 8; + } + break; + case 4: + for( i = 0; i < 4*width*height; i += 4 ) + { + img[i+0] = (img[i+0] * img[i+3] + 128) >> 8; + img[i+1] = (img[i+1] * img[i+3] + 128) >> 8; + img[i+2] = (img[i+2] * img[i+3] + 128) >> 8; + } + break; + default: + /* no other number of channels contains alpha data */ + break; + } + } + /* if the user can't support NPOT textures, make sure we force the POT option */ + if( (query_NPOT_capability() == SOIL_CAPABILITY_NONE) && + !(flags & SOIL_FLAG_TEXTURE_RECTANGLE) ) + { + /* add in the POT flag */ + flags |= SOIL_FLAG_POWER_OF_TWO; + } + /* how large of a texture can this OpenGL implementation handle? */ + /* texture_check_size_enum will be GL_MAX_TEXTURE_SIZE or SOIL_MAX_CUBE_MAP_TEXTURE_SIZE */ + glGetIntegerv( texture_check_size_enum, &max_supported_size ); + /* do I need to make it a power of 2? */ + if( + (flags & SOIL_FLAG_POWER_OF_TWO) || /* user asked for it */ + (flags & SOIL_FLAG_MIPMAPS) || /* need it for the MIP-maps */ + (width > max_supported_size) || /* it's too big, (make sure it's */ + (height > max_supported_size) ) /* 2^n for later down-sampling) */ + { + int new_width = 1; + int new_height = 1; + while( new_width < width ) + { + new_width *= 2; + } + while( new_height < height ) + { + new_height *= 2; + } + /* still? */ + if( (new_width != width) || (new_height != height) ) + { + /* yep, resize */ + unsigned char *resampled = (unsigned char*)malloc( channels*new_width*new_height ); + up_scale_image( + img, width, height, channels, + resampled, new_width, new_height ); + /* OJO this is for debug only! */ + /* + SOIL_save_image( "\\showme.bmp", SOIL_SAVE_TYPE_BMP, + new_width, new_height, channels, + resampled ); + */ + /* nuke the old guy, then point it at the new guy */ + SOIL_free_image_data( img ); + img = resampled; + width = new_width; + height = new_height; + } + } + /* now, if it is too large... */ + if( (width > max_supported_size) || (height > max_supported_size) ) + { + /* I've already made it a power of two, so simply use the MIPmapping + code to reduce its size to the allowable maximum. */ + unsigned char *resampled; + int reduce_block_x = 1, reduce_block_y = 1; + int new_width, new_height; + if( width > max_supported_size ) + { + reduce_block_x = width / max_supported_size; + } + if( height > max_supported_size ) + { + reduce_block_y = height / max_supported_size; + } + new_width = width / reduce_block_x; + new_height = height / reduce_block_y; + resampled = (unsigned char*)malloc( channels*new_width*new_height ); + /* perform the actual reduction */ + mipmap_image( img, width, height, channels, + resampled, reduce_block_x, reduce_block_y ); + /* nuke the old guy, then point it at the new guy */ + SOIL_free_image_data( img ); + img = resampled; + width = new_width; + height = new_height; + } + /* does the user want us to use YCoCg color space? */ + if( flags & SOIL_FLAG_CoCg_Y ) + { + /* this will only work with RGB and RGBA images */ + convert_RGB_to_YCoCg( img, width, height, channels ); + /* + save_image_as_DDS( "CoCg_Y.dds", width, height, channels, img ); + */ + } + /* create the OpenGL texture ID handle + (note: allowing a forced texture ID lets me reload a texture) */ + tex_id = reuse_texture_ID; + if( tex_id == 0 ) + { + glGenTextures( 1, &tex_id ); + } + check_for_GL_errors( "glGenTextures" ); + /* Note: sometimes glGenTextures fails (usually no OpenGL context) */ + if( tex_id ) + { + /* and what type am I using as the internal texture format? */ + switch( channels ) + { + case 1: + original_texture_format = GL_LUMINANCE; + break; + case 2: + original_texture_format = GL_LUMINANCE_ALPHA; + break; + case 3: + original_texture_format = GL_RGB; + break; + case 4: + original_texture_format = GL_RGBA; + break; + } + internal_texture_format = original_texture_format; + /* does the user want me to, and can I, save as DXT? */ + if( flags & SOIL_FLAG_COMPRESS_TO_DXT ) + { + DXT_mode = query_DXT_capability(); + if( DXT_mode == SOIL_CAPABILITY_PRESENT ) + { + /* I can use DXT, whether I compress it or OpenGL does */ + if( (channels & 1) == 1 ) + { + /* 1 or 3 channels = DXT1 */ + internal_texture_format = SOIL_RGB_S3TC_DXT1; + } else + { + /* 2 or 4 channels = DXT5 */ + internal_texture_format = SOIL_RGBA_S3TC_DXT5; + } + } + } + /* bind an OpenGL texture ID */ + glBindTexture( opengl_texture_type, tex_id ); + check_for_GL_errors( "glBindTexture" ); + /* upload the main image */ + if( DXT_mode == SOIL_CAPABILITY_PRESENT ) + { + /* user wants me to do the DXT conversion! */ + int DDS_size; + unsigned char *DDS_data = NULL; + if( (channels & 1) == 1 ) + { + /* RGB, use DXT1 */ + DDS_data = convert_image_to_DXT1( img, width, height, channels, &DDS_size ); + } else + { + /* RGBA, use DXT5 */ + DDS_data = convert_image_to_DXT5( img, width, height, channels, &DDS_size ); + } + if( DDS_data ) + { + soilGlCompressedTexImage2D( + opengl_texture_target, 0, + internal_texture_format, width, height, 0, + DDS_size, DDS_data ); + check_for_GL_errors( "glCompressedTexImage2D" ); + SOIL_free_image_data( DDS_data ); + /* printf( "Internal DXT compressor\n" ); */ + } else + { + /* my compression failed, try the OpenGL driver's version */ + glTexImage2D( + opengl_texture_target, 0, + internal_texture_format, width, height, 0, + original_texture_format, GL_UNSIGNED_BYTE, img ); + check_for_GL_errors( "glTexImage2D" ); + /* printf( "OpenGL DXT compressor\n" ); */ + } + } else + { + /* user want OpenGL to do all the work! */ + glTexImage2D( + opengl_texture_target, 0, + internal_texture_format, width, height, 0, + original_texture_format, GL_UNSIGNED_BYTE, img ); + check_for_GL_errors( "glTexImage2D" ); + /*printf( "OpenGL DXT compressor\n" ); */ + } + /* are any MIPmaps desired? */ + if( flags & SOIL_FLAG_MIPMAPS ) + { + int MIPlevel = 1; + int MIPwidth = (width+1) / 2; + int MIPheight = (height+1) / 2; + unsigned char *resampled = (unsigned char*)malloc( channels*MIPwidth*MIPheight ); + while( ((1< 0; --i ) + { + unsigned char temp = pixel_data[index1]; + pixel_data[index1] = pixel_data[index2]; + pixel_data[index2] = temp; + ++index1; + ++index2; + } + } + + /* save the image */ + save_result = SOIL_save_image( filename, image_type, width, height, 3, pixel_data); + + /* And free the memory */ + SOIL_free_image_data( pixel_data ); + return save_result; +} + +unsigned char* + SOIL_load_image + ( + const char *filename, + int *width, int *height, int *channels, + int force_channels + ) +{ + unsigned char *result = stbi_load( filename, + width, height, channels, force_channels ); + if( result == NULL ) + { + result_string_pointer = stbi_failure_reason(); + } else + { + result_string_pointer = "Image loaded"; + } + return result; +} + +unsigned char* + SOIL_load_image_from_memory + ( + const unsigned char *const buffer, + int buffer_length, + int *width, int *height, int *channels, + int force_channels + ) +{ + unsigned char *result = stbi_load_from_memory( + buffer, buffer_length, + width, height, channels, + force_channels ); + if( result == NULL ) + { + result_string_pointer = stbi_failure_reason(); + } else + { + result_string_pointer = "Image loaded from memory"; + } + return result; +} + +int + SOIL_save_image + ( + const char *filename, + int image_type, + int width, int height, int channels, + const unsigned char *const data + ) +{ + int save_result; + + /* error check */ + if( (width < 1) || (height < 1) || + (channels < 1) || (channels > 4) || + (data == NULL) || + (filename == NULL) ) + { + return 0; + } + if( image_type == SOIL_SAVE_TYPE_BMP ) + { + save_result = stbi_write_bmp( filename, + width, height, channels, (void*)data ); + } else + if( image_type == SOIL_SAVE_TYPE_TGA ) + { + save_result = stbi_write_tga( filename, + width, height, channels, (void*)data ); + } else + if( image_type == SOIL_SAVE_TYPE_DDS ) + { + save_result = save_image_as_DDS( filename, + width, height, channels, (const unsigned char *const)data ); + } else + { + save_result = 0; + } + if( save_result == 0 ) + { + result_string_pointer = "Saving the image failed"; + } else + { + result_string_pointer = "Image saved"; + } + return save_result; +} + +void + SOIL_free_image_data + ( + unsigned char *img_data + ) +{ + free( (void*)img_data ); +} + +const char* + SOIL_last_result + ( + void + ) +{ + return result_string_pointer; +} + +unsigned int SOIL_direct_load_DDS_from_memory( + const unsigned char *const buffer, + int buffer_length, + unsigned int reuse_texture_ID, + int flags, + int loading_as_cubemap ) +{ + /* variables */ + DDS_header header; + unsigned int buffer_index = 0; + unsigned int tex_ID = 0; + /* file reading variables */ + unsigned int S3TC_type = 0; + unsigned char *DDS_data; + unsigned int DDS_main_size; + unsigned int DDS_full_size; + unsigned int width, height; + int mipmaps, cubemap, uncompressed, block_size = 16; + unsigned int flag; + unsigned int cf_target, ogl_target_start, ogl_target_end; + unsigned int opengl_texture_type; + int i; + /* 1st off, does the filename even exist? */ + if( NULL == buffer ) + { + /* we can't do it! */ + result_string_pointer = "NULL buffer"; + return 0; + } + if( buffer_length < sizeof( DDS_header ) ) + { + /* we can't do it! */ + result_string_pointer = "DDS file was too small to contain the DDS header"; + return 0; + } + /* try reading in the header */ + memcpy ( (void*)(&header), (const void *)buffer, sizeof( DDS_header ) ); + buffer_index = sizeof( DDS_header ); + /* guilty until proven innocent */ + result_string_pointer = "Failed to read a known DDS header"; + /* validate the header (warning, "goto"'s ahead, shield your eyes!!) */ + flag = ('D'<<0)|('D'<<8)|('S'<<16)|(' '<<24); + if( header.dwMagic != flag ) {goto quick_exit;} + if( header.dwSize != 124 ) {goto quick_exit;} + /* I need all of these */ + flag = DDSD_CAPS | DDSD_HEIGHT | DDSD_WIDTH | DDSD_PIXELFORMAT; + if( (header.dwFlags & flag) != flag ) {goto quick_exit;} + /* According to the MSDN spec, the dwFlags should contain + DDSD_LINEARSIZE if it's compressed, or DDSD_PITCH if + uncompressed. Some DDS writers do not conform to the + spec, so I need to make my reader more tolerant */ + /* I need one of these */ + flag = DDPF_FOURCC | DDPF_RGB; + if( (header.sPixelFormat.dwFlags & flag) == 0 ) {goto quick_exit;} + if( header.sPixelFormat.dwSize != 32 ) {goto quick_exit;} + if( (header.sCaps.dwCaps1 & DDSCAPS_TEXTURE) == 0 ) {goto quick_exit;} + /* make sure it is a type we can upload */ + if( (header.sPixelFormat.dwFlags & DDPF_FOURCC) && + !( + (header.sPixelFormat.dwFourCC == (('D'<<0)|('X'<<8)|('T'<<16)|('1'<<24))) || + (header.sPixelFormat.dwFourCC == (('D'<<0)|('X'<<8)|('T'<<16)|('3'<<24))) || + (header.sPixelFormat.dwFourCC == (('D'<<0)|('X'<<8)|('T'<<16)|('5'<<24))) + ) ) + { + goto quick_exit; + } + /* OK, validated the header, let's load the image data */ + result_string_pointer = "DDS header loaded and validated"; + width = header.dwWidth; + height = header.dwHeight; + uncompressed = 1 - (header.sPixelFormat.dwFlags & DDPF_FOURCC) / DDPF_FOURCC; + cubemap = (header.sCaps.dwCaps2 & DDSCAPS2_CUBEMAP) / DDSCAPS2_CUBEMAP; + if( uncompressed ) + { + S3TC_type = GL_RGB; + block_size = 3; + if( header.sPixelFormat.dwFlags & DDPF_ALPHAPIXELS ) + { + S3TC_type = GL_RGBA; + block_size = 4; + } + DDS_main_size = width * height * block_size; + } else + { + /* can we even handle direct uploading to OpenGL DXT compressed images? */ + if( query_DXT_capability() != SOIL_CAPABILITY_PRESENT ) + { + /* we can't do it! */ + result_string_pointer = "Direct upload of S3TC images not supported by the OpenGL driver"; + return 0; + } + /* well, we know it is DXT1/3/5, because we checked above */ + switch( (header.sPixelFormat.dwFourCC >> 24) - '0' ) + { + case 1: + S3TC_type = SOIL_RGBA_S3TC_DXT1; + block_size = 8; + break; + case 3: + S3TC_type = SOIL_RGBA_S3TC_DXT3; + block_size = 16; + break; + case 5: + S3TC_type = SOIL_RGBA_S3TC_DXT5; + block_size = 16; + break; + } + DDS_main_size = ((width+3)>>2)*((height+3)>>2)*block_size; + } + if( cubemap ) + { + /* does the user want a cubemap? */ + if( !loading_as_cubemap ) + { + /* we can't do it! */ + result_string_pointer = "DDS image was a cubemap"; + return 0; + } + /* can we even handle cubemaps with the OpenGL driver? */ + if( query_cubemap_capability() != SOIL_CAPABILITY_PRESENT ) + { + /* we can't do it! */ + result_string_pointer = "Direct upload of cubemap images not supported by the OpenGL driver"; + return 0; + } + ogl_target_start = SOIL_TEXTURE_CUBE_MAP_POSITIVE_X; + ogl_target_end = SOIL_TEXTURE_CUBE_MAP_NEGATIVE_Z; + opengl_texture_type = SOIL_TEXTURE_CUBE_MAP; + } else + { + /* does the user want a non-cubemap? */ + if( loading_as_cubemap ) + { + /* we can't do it! */ + result_string_pointer = "DDS image was not a cubemap"; + return 0; + } + ogl_target_start = GL_TEXTURE_2D; + ogl_target_end = GL_TEXTURE_2D; + opengl_texture_type = GL_TEXTURE_2D; + } + if( (header.sCaps.dwCaps1 & DDSCAPS_MIPMAP) && (header.dwMipMapCount > 1) ) + { + int shift_offset; + mipmaps = header.dwMipMapCount - 1; + DDS_full_size = DDS_main_size; + if( uncompressed ) + { + /* uncompressed DDS, simple MIPmap size calculation */ + shift_offset = 0; + } else + { + /* compressed DDS, MIPmap size calculation is block based */ + shift_offset = 2; + } + for( i = 1; i <= mipmaps; ++ i ) + { + int w, h; + w = width >> (shift_offset + i); + h = height >> (shift_offset + i); + if( w < 1 ) + { + w = 1; + } + if( h < 1 ) + { + h = 1; + } + DDS_full_size += w*h*block_size; + } + } else + { + mipmaps = 0; + DDS_full_size = DDS_main_size; + } + DDS_data = (unsigned char*)malloc( DDS_full_size ); + /* got the image data RAM, create or use an existing OpenGL texture handle */ + tex_ID = reuse_texture_ID; + if( tex_ID == 0 ) + { + glGenTextures( 1, &tex_ID ); + } + /* bind an OpenGL texture ID */ + glBindTexture( opengl_texture_type, tex_ID ); + /* do this for each face of the cubemap! */ + for( cf_target = ogl_target_start; cf_target <= ogl_target_end; ++cf_target ) + { + if( buffer_index + DDS_full_size <= buffer_length ) + { + unsigned int byte_offset = DDS_main_size; + memcpy( (void*)DDS_data, (const void*)(&buffer[buffer_index]), DDS_full_size ); + buffer_index += DDS_full_size; + /* upload the main chunk */ + if( uncompressed ) + { + /* and remember, DXT uncompressed uses BGR(A), + so swap to RGB(A) for ALL MIPmap levels */ + for( i = 0; i < DDS_full_size; i += block_size ) + { + unsigned char temp = DDS_data[i]; + DDS_data[i] = DDS_data[i+2]; + DDS_data[i+2] = temp; + } + glTexImage2D( + cf_target, 0, + S3TC_type, width, height, 0, + S3TC_type, GL_UNSIGNED_BYTE, DDS_data ); + } else + { + soilGlCompressedTexImage2D( + cf_target, 0, + S3TC_type, width, height, 0, + DDS_main_size, DDS_data ); + } + /* upload the mipmaps, if we have them */ + for( i = 1; i <= mipmaps; ++i ) + { + int w, h, mip_size; + w = width >> i; + h = height >> i; + if( w < 1 ) + { + w = 1; + } + if( h < 1 ) + { + h = 1; + } + /* upload this mipmap */ + if( uncompressed ) + { + mip_size = w*h*block_size; + glTexImage2D( + cf_target, i, + S3TC_type, w, h, 0, + S3TC_type, GL_UNSIGNED_BYTE, &DDS_data[byte_offset] ); + } else + { + mip_size = ((w+3)/4)*((h+3)/4)*block_size; + soilGlCompressedTexImage2D( + cf_target, i, + S3TC_type, w, h, 0, + mip_size, &DDS_data[byte_offset] ); + } + /* and move to the next mipmap */ + byte_offset += mip_size; + } + /* it worked! */ + result_string_pointer = "DDS file loaded"; + } else + { + glDeleteTextures( 1, & tex_ID ); + tex_ID = 0; + cf_target = ogl_target_end + 1; + result_string_pointer = "DDS file was too small for expected image data"; + } + }/* end reading each face */ + SOIL_free_image_data( DDS_data ); + if( tex_ID ) + { + /* did I have MIPmaps? */ + if( mipmaps > 0 ) + { + /* instruct OpenGL to use the MIPmaps */ + glTexParameteri( opengl_texture_type, GL_TEXTURE_MAG_FILTER, GL_LINEAR ); + glTexParameteri( opengl_texture_type, GL_TEXTURE_MIN_FILTER, GL_LINEAR_MIPMAP_LINEAR ); + } else + { + /* instruct OpenGL _NOT_ to use the MIPmaps */ + glTexParameteri( opengl_texture_type, GL_TEXTURE_MAG_FILTER, GL_LINEAR ); + glTexParameteri( opengl_texture_type, GL_TEXTURE_MIN_FILTER, GL_LINEAR ); + } + /* does the user want clamping, or wrapping? */ + if( flags & SOIL_FLAG_TEXTURE_REPEATS ) + { + glTexParameteri( opengl_texture_type, GL_TEXTURE_WRAP_S, GL_REPEAT ); + glTexParameteri( opengl_texture_type, GL_TEXTURE_WRAP_T, GL_REPEAT ); + glTexParameteri( opengl_texture_type, SOIL_TEXTURE_WRAP_R, GL_REPEAT ); + } else + { + /* unsigned int clamp_mode = SOIL_CLAMP_TO_EDGE; */ + unsigned int clamp_mode = GL_CLAMP; + glTexParameteri( opengl_texture_type, GL_TEXTURE_WRAP_S, clamp_mode ); + glTexParameteri( opengl_texture_type, GL_TEXTURE_WRAP_T, clamp_mode ); + glTexParameteri( opengl_texture_type, SOIL_TEXTURE_WRAP_R, clamp_mode ); + } + } + +quick_exit: + /* report success or failure */ + return tex_ID; +} + +unsigned int SOIL_direct_load_DDS( + const char *filename, + unsigned int reuse_texture_ID, + int flags, + int loading_as_cubemap ) +{ + FILE *f; + unsigned char *buffer; + size_t buffer_length, bytes_read; + unsigned int tex_ID = 0; + /* error checks */ + if( NULL == filename ) + { + result_string_pointer = "NULL filename"; + return 0; + } + f = fopen( filename, "rb" ); + if( NULL == f ) + { + /* the file doesn't seem to exist (or be open-able) */ + result_string_pointer = "Can not find DDS file"; + return 0; + } + fseek( f, 0, SEEK_END ); + buffer_length = ftell( f ); + fseek( f, 0, SEEK_SET ); + buffer = (unsigned char *) malloc( buffer_length ); + if( NULL == buffer ) + { + result_string_pointer = "malloc failed"; + fclose( f ); + return 0; + } + bytes_read = fread( (void*)buffer, 1, buffer_length, f ); + fclose( f ); + if( bytes_read < buffer_length ) + { + /* huh? */ + buffer_length = bytes_read; + } + /* now try to do the loading */ + tex_ID = SOIL_direct_load_DDS_from_memory( + (const unsigned char *const)buffer, buffer_length, + reuse_texture_ID, flags, loading_as_cubemap ); + SOIL_free_image_data( buffer ); + return tex_ID; +} + +int query_NPOT_capability( void ) +{ + /* check for the capability */ + if( has_NPOT_capability == SOIL_CAPABILITY_UNKNOWN ) + { + /* we haven't yet checked for the capability, do so */ + if( + (NULL == strstr( (char const*)glGetString( GL_EXTENSIONS ), + "GL_ARB_texture_non_power_of_two" ) ) + ) + { + /* not there, flag the failure */ + has_NPOT_capability = SOIL_CAPABILITY_NONE; + } else + { + /* it's there! */ + has_NPOT_capability = SOIL_CAPABILITY_PRESENT; + } + } + /* let the user know if we can do non-power-of-two textures or not */ + return has_NPOT_capability; +} + +int query_tex_rectangle_capability( void ) +{ + /* check for the capability */ + if( has_tex_rectangle_capability == SOIL_CAPABILITY_UNKNOWN ) + { + /* we haven't yet checked for the capability, do so */ + if( + (NULL == strstr( (char const*)glGetString( GL_EXTENSIONS ), + "GL_ARB_texture_rectangle" ) ) + && + (NULL == strstr( (char const*)glGetString( GL_EXTENSIONS ), + "GL_EXT_texture_rectangle" ) ) + && + (NULL == strstr( (char const*)glGetString( GL_EXTENSIONS ), + "GL_NV_texture_rectangle" ) ) + ) + { + /* not there, flag the failure */ + has_tex_rectangle_capability = SOIL_CAPABILITY_NONE; + } else + { + /* it's there! */ + has_tex_rectangle_capability = SOIL_CAPABILITY_PRESENT; + } + } + /* let the user know if we can do texture rectangles or not */ + return has_tex_rectangle_capability; +} + +int query_cubemap_capability( void ) +{ + /* check for the capability */ + if( has_cubemap_capability == SOIL_CAPABILITY_UNKNOWN ) + { + /* we haven't yet checked for the capability, do so */ + if( + (NULL == strstr( (char const*)glGetString( GL_EXTENSIONS ), + "GL_ARB_texture_cube_map" ) ) + && + (NULL == strstr( (char const*)glGetString( GL_EXTENSIONS ), + "GL_EXT_texture_cube_map" ) ) + ) + { + /* not there, flag the failure */ + has_cubemap_capability = SOIL_CAPABILITY_NONE; + } else + { + /* it's there! */ + has_cubemap_capability = SOIL_CAPABILITY_PRESENT; + } + } + /* let the user know if we can do cubemaps or not */ + return has_cubemap_capability; +} + +int query_DXT_capability( void ) +{ + /* check for the capability */ + if( has_DXT_capability == SOIL_CAPABILITY_UNKNOWN ) + { + /* we haven't yet checked for the capability, do so */ + if( NULL == strstr( + (char const*)glGetString( GL_EXTENSIONS ), + "GL_EXT_texture_compression_s3tc" ) ) + { + /* not there, flag the failure */ + has_DXT_capability = SOIL_CAPABILITY_NONE; + } else + { + /* and find the address of the extension function */ + P_SOIL_GLCOMPRESSEDTEXIMAGE2DPROC ext_addr = NULL; + #ifdef WIN32 + ext_addr = (P_SOIL_GLCOMPRESSEDTEXIMAGE2DPROC) + wglGetProcAddress + ( + "glCompressedTexImage2DARB" + ); + #elif defined(__APPLE__) || defined(__APPLE_CC__) + /* I can't test this Apple stuff! */ + CFBundleRef bundle; + CFURLRef bundleURL = + CFURLCreateWithFileSystemPath( + kCFAllocatorDefault, + CFSTR("/System/Library/Frameworks/OpenGL.framework"), + kCFURLPOSIXPathStyle, + true ); + CFStringRef extensionName = + CFStringCreateWithCString( + kCFAllocatorDefault, + "glCompressedTexImage2DARB", + kCFStringEncodingASCII ); + bundle = CFBundleCreate( kCFAllocatorDefault, bundleURL ); + assert( bundle != NULL ); + ext_addr = (P_SOIL_GLCOMPRESSEDTEXIMAGE2DPROC) + CFBundleGetFunctionPointerForName + ( + bundle, extensionName + ); + CFRelease( bundleURL ); + CFRelease( extensionName ); + CFRelease( bundle ); + #else + ext_addr = (P_SOIL_GLCOMPRESSEDTEXIMAGE2DPROC) + glXGetProcAddressARB + ( + (const GLubyte *)"glCompressedTexImage2DARB" + ); + #endif + /* Flag it so no checks needed later */ + if( NULL == ext_addr ) + { + /* hmm, not good!! This should not happen, but does on my + laptop's VIA chipset. The GL_EXT_texture_compression_s3tc + spec requires that ARB_texture_compression be present too. + this means I can upload and have the OpenGL drive do the + conversion, but I can't use my own routines or load DDS files + from disk and upload them directly [8^( */ + has_DXT_capability = SOIL_CAPABILITY_NONE; + } else + { + /* all's well! */ + soilGlCompressedTexImage2D = ext_addr; + has_DXT_capability = SOIL_CAPABILITY_PRESENT; + } + } + } + /* let the user know if we can do DXT or not */ + return has_DXT_capability; +} diff --git a/lib/soil/soil/SOIL.h b/lib/soil/soil/SOIL.h new file mode 100644 index 0000000..57f375b --- /dev/null +++ b/lib/soil/soil/SOIL.h @@ -0,0 +1,433 @@ +/** + @mainpage SOIL + + Jonathan Dummer + 2007-07-26-10.36 + + Simple OpenGL Image Library + + A tiny c library for uploading images as + textures into OpenGL. Also saving and + loading of images is supported. + + I'm using Sean's Tool Box image loader as a base: + http://www.nothings.org/ + + I'm upgrading it to load TGA and DDS files, and a direct + path for loading DDS files straight into OpenGL textures, + when applicable. + + Image Formats: + - BMP load & save + - TGA load & save + - DDS load & save + - PNG load + - JPG load + + OpenGL Texture Features: + - resample to power-of-two sizes + - MIPmap generation + - compressed texture S3TC formats (if supported) + - can pre-multiply alpha for you, for better compositing + - can flip image about the y-axis (except pre-compressed DDS files) + + Thanks to: + * Sean Barret - for the awesome stb_image + * Dan Venkitachalam - for finding some non-compliant DDS files, and patching some explicit casts + * everybody at gamedev.net +**/ + +#ifndef HEADER_SIMPLE_OPENGL_IMAGE_LIBRARY +#define HEADER_SIMPLE_OPENGL_IMAGE_LIBRARY + +#ifdef __cplusplus +extern "C" { +#endif + +/** + The format of images that may be loaded (force_channels). + SOIL_LOAD_AUTO leaves the image in whatever format it was found. + SOIL_LOAD_L forces the image to load as Luminous (greyscale) + SOIL_LOAD_LA forces the image to load as Luminous with Alpha + SOIL_LOAD_RGB forces the image to load as Red Green Blue + SOIL_LOAD_RGBA forces the image to load as Red Green Blue Alpha +**/ +enum +{ + SOIL_LOAD_AUTO = 0, + SOIL_LOAD_L = 1, + SOIL_LOAD_LA = 2, + SOIL_LOAD_RGB = 3, + SOIL_LOAD_RGBA = 4 +}; + +/** + Passed in as reuse_texture_ID, will cause SOIL to + register a new texture ID using glGenTextures(). + If the value passed into reuse_texture_ID > 0 then + SOIL will just re-use that texture ID (great for + reloading image assets in-game!) +**/ +enum +{ + SOIL_CREATE_NEW_ID = 0 +}; + +/** + flags you can pass into SOIL_load_OGL_texture() + and SOIL_create_OGL_texture(). + (note that if SOIL_FLAG_DDS_LOAD_DIRECT is used + the rest of the flags with the exception of + SOIL_FLAG_TEXTURE_REPEATS will be ignored while + loading already-compressed DDS files.) + + SOIL_FLAG_POWER_OF_TWO: force the image to be POT + SOIL_FLAG_MIPMAPS: generate mipmaps for the texture + SOIL_FLAG_TEXTURE_REPEATS: otherwise will clamp + SOIL_FLAG_MULTIPLY_ALPHA: for using (GL_ONE,GL_ONE_MINUS_SRC_ALPHA) blending + SOIL_FLAG_INVERT_Y: flip the image vertically + SOIL_FLAG_COMPRESS_TO_DXT: if the card can display them, will convert RGB to DXT1, RGBA to DXT5 + SOIL_FLAG_DDS_LOAD_DIRECT: will load DDS files directly without _ANY_ additional processing + SOIL_FLAG_NTSC_SAFE_RGB: clamps RGB components to the range [16,235] + SOIL_FLAG_CoCg_Y: Google YCoCg; RGB=>CoYCg, RGBA=>CoCgAY + SOIL_FLAG_TEXTURE_RECTANGE: uses ARB_texture_rectangle ; pixel indexed & no repeat or MIPmaps or cubemaps +**/ +enum +{ + SOIL_FLAG_POWER_OF_TWO = 1, + SOIL_FLAG_MIPMAPS = 2, + SOIL_FLAG_TEXTURE_REPEATS = 4, + SOIL_FLAG_MULTIPLY_ALPHA = 8, + SOIL_FLAG_INVERT_Y = 16, + SOIL_FLAG_COMPRESS_TO_DXT = 32, + SOIL_FLAG_DDS_LOAD_DIRECT = 64, + SOIL_FLAG_NTSC_SAFE_RGB = 128, + SOIL_FLAG_CoCg_Y = 256, + SOIL_FLAG_TEXTURE_RECTANGLE = 512 +}; + +/** + The types of images that may be saved. + (TGA supports uncompressed RGB / RGBA) + (BMP supports uncompressed RGB) + (DDS supports DXT1 and DXT5) +**/ +enum +{ + SOIL_SAVE_TYPE_TGA = 0, + SOIL_SAVE_TYPE_BMP = 1, + SOIL_SAVE_TYPE_DDS = 2 +}; + +/** + Defines the order of faces in a DDS cubemap. + I recommend that you use the same order in single + image cubemap files, so they will be interchangeable + with DDS cubemaps when using SOIL. +**/ +#define SOIL_DDS_CUBEMAP_FACE_ORDER "EWUDNS" + +/** + The types of internal fake HDR representations + + SOIL_HDR_RGBE: RGB * pow( 2.0, A - 128.0 ) + SOIL_HDR_RGBdivA: RGB / A + SOIL_HDR_RGBdivA2: RGB / (A*A) +**/ +enum +{ + SOIL_HDR_RGBE = 0, + SOIL_HDR_RGBdivA = 1, + SOIL_HDR_RGBdivA2 = 2 +}; + +/** + Loads an image from disk into an OpenGL texture. + \param filename the name of the file to upload as a texture + \param force_channels 0-image format, 1-luminous, 2-luminous/alpha, 3-RGB, 4-RGBA + \param reuse_texture_ID 0-generate a new texture ID, otherwise reuse the texture ID (overwriting the old texture) + \param flags can be any of SOIL_FLAG_POWER_OF_TWO | SOIL_FLAG_MIPMAPS | SOIL_FLAG_TEXTURE_REPEATS | SOIL_FLAG_MULTIPLY_ALPHA | SOIL_FLAG_INVERT_Y | SOIL_FLAG_COMPRESS_TO_DXT | SOIL_FLAG_DDS_LOAD_DIRECT + \return 0-failed, otherwise returns the OpenGL texture handle +**/ +unsigned int + SOIL_load_OGL_texture + ( + const char *filename, + int force_channels, + unsigned int reuse_texture_ID, + unsigned int flags + ); + +/** + Loads 6 images from disk into an OpenGL cubemap texture. + \param x_pos_file the name of the file to upload as the +x cube face + \param x_neg_file the name of the file to upload as the -x cube face + \param y_pos_file the name of the file to upload as the +y cube face + \param y_neg_file the name of the file to upload as the -y cube face + \param z_pos_file the name of the file to upload as the +z cube face + \param z_neg_file the name of the file to upload as the -z cube face + \param force_channels 0-image format, 1-luminous, 2-luminous/alpha, 3-RGB, 4-RGBA + \param reuse_texture_ID 0-generate a new texture ID, otherwise reuse the texture ID (overwriting the old texture) + \param flags can be any of SOIL_FLAG_POWER_OF_TWO | SOIL_FLAG_MIPMAPS | SOIL_FLAG_TEXTURE_REPEATS | SOIL_FLAG_MULTIPLY_ALPHA | SOIL_FLAG_INVERT_Y | SOIL_FLAG_COMPRESS_TO_DXT | SOIL_FLAG_DDS_LOAD_DIRECT + \return 0-failed, otherwise returns the OpenGL texture handle +**/ +unsigned int + SOIL_load_OGL_cubemap + ( + const char *x_pos_file, + const char *x_neg_file, + const char *y_pos_file, + const char *y_neg_file, + const char *z_pos_file, + const char *z_neg_file, + int force_channels, + unsigned int reuse_texture_ID, + unsigned int flags + ); + +/** + Loads 1 image from disk and splits it into an OpenGL cubemap texture. + \param filename the name of the file to upload as a texture + \param face_order the order of the faces in the file, any combination of NSWEUD, for North, South, Up, etc. + \param force_channels 0-image format, 1-luminous, 2-luminous/alpha, 3-RGB, 4-RGBA + \param reuse_texture_ID 0-generate a new texture ID, otherwise reuse the texture ID (overwriting the old texture) + \param flags can be any of SOIL_FLAG_POWER_OF_TWO | SOIL_FLAG_MIPMAPS | SOIL_FLAG_TEXTURE_REPEATS | SOIL_FLAG_MULTIPLY_ALPHA | SOIL_FLAG_INVERT_Y | SOIL_FLAG_COMPRESS_TO_DXT | SOIL_FLAG_DDS_LOAD_DIRECT + \return 0-failed, otherwise returns the OpenGL texture handle +**/ +unsigned int + SOIL_load_OGL_single_cubemap + ( + const char *filename, + const char face_order[6], + int force_channels, + unsigned int reuse_texture_ID, + unsigned int flags + ); + +/** + Loads an HDR image from disk into an OpenGL texture. + \param filename the name of the file to upload as a texture + \param fake_HDR_format SOIL_HDR_RGBE, SOIL_HDR_RGBdivA, SOIL_HDR_RGBdivA2 + \param reuse_texture_ID 0-generate a new texture ID, otherwise reuse the texture ID (overwriting the old texture) + \param flags can be any of SOIL_FLAG_POWER_OF_TWO | SOIL_FLAG_MIPMAPS | SOIL_FLAG_TEXTURE_REPEATS | SOIL_FLAG_MULTIPLY_ALPHA | SOIL_FLAG_INVERT_Y | SOIL_FLAG_COMPRESS_TO_DXT + \return 0-failed, otherwise returns the OpenGL texture handle +**/ +unsigned int + SOIL_load_OGL_HDR_texture + ( + const char *filename, + int fake_HDR_format, + int rescale_to_max, + unsigned int reuse_texture_ID, + unsigned int flags + ); + +/** + Loads an image from RAM into an OpenGL texture. + \param buffer the image data in RAM just as if it were still in a file + \param buffer_length the size of the buffer in bytes + \param force_channels 0-image format, 1-luminous, 2-luminous/alpha, 3-RGB, 4-RGBA + \param reuse_texture_ID 0-generate a new texture ID, otherwise reuse the texture ID (overwriting the old texture) + \param flags can be any of SOIL_FLAG_POWER_OF_TWO | SOIL_FLAG_MIPMAPS | SOIL_FLAG_TEXTURE_REPEATS | SOIL_FLAG_MULTIPLY_ALPHA | SOIL_FLAG_INVERT_Y | SOIL_FLAG_COMPRESS_TO_DXT | SOIL_FLAG_DDS_LOAD_DIRECT + \return 0-failed, otherwise returns the OpenGL texture handle +**/ +unsigned int + SOIL_load_OGL_texture_from_memory + ( + const unsigned char *const buffer, + int buffer_length, + int force_channels, + unsigned int reuse_texture_ID, + unsigned int flags + ); + +/** + Loads 6 images from memory into an OpenGL cubemap texture. + \param x_pos_buffer the image data in RAM to upload as the +x cube face + \param x_pos_buffer_length the size of the above buffer + \param x_neg_buffer the image data in RAM to upload as the +x cube face + \param x_neg_buffer_length the size of the above buffer + \param y_pos_buffer the image data in RAM to upload as the +x cube face + \param y_pos_buffer_length the size of the above buffer + \param y_neg_buffer the image data in RAM to upload as the +x cube face + \param y_neg_buffer_length the size of the above buffer + \param z_pos_buffer the image data in RAM to upload as the +x cube face + \param z_pos_buffer_length the size of the above buffer + \param z_neg_buffer the image data in RAM to upload as the +x cube face + \param z_neg_buffer_length the size of the above buffer + \param force_channels 0-image format, 1-luminous, 2-luminous/alpha, 3-RGB, 4-RGBA + \param reuse_texture_ID 0-generate a new texture ID, otherwise reuse the texture ID (overwriting the old texture) + \param flags can be any of SOIL_FLAG_POWER_OF_TWO | SOIL_FLAG_MIPMAPS | SOIL_FLAG_TEXTURE_REPEATS | SOIL_FLAG_MULTIPLY_ALPHA | SOIL_FLAG_INVERT_Y | SOIL_FLAG_COMPRESS_TO_DXT | SOIL_FLAG_DDS_LOAD_DIRECT + \return 0-failed, otherwise returns the OpenGL texture handle +**/ +unsigned int + SOIL_load_OGL_cubemap_from_memory + ( + const unsigned char *const x_pos_buffer, + int x_pos_buffer_length, + const unsigned char *const x_neg_buffer, + int x_neg_buffer_length, + const unsigned char *const y_pos_buffer, + int y_pos_buffer_length, + const unsigned char *const y_neg_buffer, + int y_neg_buffer_length, + const unsigned char *const z_pos_buffer, + int z_pos_buffer_length, + const unsigned char *const z_neg_buffer, + int z_neg_buffer_length, + int force_channels, + unsigned int reuse_texture_ID, + unsigned int flags + ); + +/** + Loads 1 image from RAM and splits it into an OpenGL cubemap texture. + \param buffer the image data in RAM just as if it were still in a file + \param buffer_length the size of the buffer in bytes + \param face_order the order of the faces in the file, any combination of NSWEUD, for North, South, Up, etc. + \param force_channels 0-image format, 1-luminous, 2-luminous/alpha, 3-RGB, 4-RGBA + \param reuse_texture_ID 0-generate a new texture ID, otherwise reuse the texture ID (overwriting the old texture) + \param flags can be any of SOIL_FLAG_POWER_OF_TWO | SOIL_FLAG_MIPMAPS | SOIL_FLAG_TEXTURE_REPEATS | SOIL_FLAG_MULTIPLY_ALPHA | SOIL_FLAG_INVERT_Y | SOIL_FLAG_COMPRESS_TO_DXT | SOIL_FLAG_DDS_LOAD_DIRECT + \return 0-failed, otherwise returns the OpenGL texture handle +**/ +unsigned int + SOIL_load_OGL_single_cubemap_from_memory + ( + const unsigned char *const buffer, + int buffer_length, + const char face_order[6], + int force_channels, + unsigned int reuse_texture_ID, + unsigned int flags + ); + +/** + Creates a 2D OpenGL texture from raw image data. Note that the raw data is + _NOT_ freed after the upload (so the user can load various versions). + \param data the raw data to be uploaded as an OpenGL texture + \param width the width of the image in pixels + \param height the height of the image in pixels + \param channels the number of channels: 1-luminous, 2-luminous/alpha, 3-RGB, 4-RGBA + \param reuse_texture_ID 0-generate a new texture ID, otherwise reuse the texture ID (overwriting the old texture) + \param flags can be any of SOIL_FLAG_POWER_OF_TWO | SOIL_FLAG_MIPMAPS | SOIL_FLAG_TEXTURE_REPEATS | SOIL_FLAG_MULTIPLY_ALPHA | SOIL_FLAG_INVERT_Y | SOIL_FLAG_COMPRESS_TO_DXT + \return 0-failed, otherwise returns the OpenGL texture handle +**/ +unsigned int + SOIL_create_OGL_texture + ( + const unsigned char *const data, + int width, int height, int channels, + unsigned int reuse_texture_ID, + unsigned int flags + ); + +/** + Creates an OpenGL cubemap texture by splitting up 1 image into 6 parts. + \param data the raw data to be uploaded as an OpenGL texture + \param width the width of the image in pixels + \param height the height of the image in pixels + \param channels the number of channels: 1-luminous, 2-luminous/alpha, 3-RGB, 4-RGBA + \param face_order the order of the faces in the file, and combination of NSWEUD, for North, South, Up, etc. + \param reuse_texture_ID 0-generate a new texture ID, otherwise reuse the texture ID (overwriting the old texture) + \param flags can be any of SOIL_FLAG_POWER_OF_TWO | SOIL_FLAG_MIPMAPS | SOIL_FLAG_TEXTURE_REPEATS | SOIL_FLAG_MULTIPLY_ALPHA | SOIL_FLAG_INVERT_Y | SOIL_FLAG_COMPRESS_TO_DXT | SOIL_FLAG_DDS_LOAD_DIRECT + \return 0-failed, otherwise returns the OpenGL texture handle +**/ +unsigned int + SOIL_create_OGL_single_cubemap + ( + const unsigned char *const data, + int width, int height, int channels, + const char face_order[6], + unsigned int reuse_texture_ID, + unsigned int flags + ); + +/** + Captures the OpenGL window (RGB) and saves it to disk + \return 0 if it failed, otherwise returns 1 +**/ +int + SOIL_save_screenshot + ( + const char *filename, + int image_type, + int x, int y, + int width, int height + ); + +/** + Loads an image from disk into an array of unsigned chars. + Note that *channels return the original channel count of the + image. If force_channels was other than SOIL_LOAD_AUTO, + the resulting image has force_channels, but *channels may be + different (if the original image had a different channel + count). + \return 0 if failed, otherwise returns 1 +**/ +unsigned char* + SOIL_load_image + ( + const char *filename, + int *width, int *height, int *channels, + int force_channels + ); + +/** + Loads an image from memory into an array of unsigned chars. + Note that *channels return the original channel count of the + image. If force_channels was other than SOIL_LOAD_AUTO, + the resulting image has force_channels, but *channels may be + different (if the original image had a different channel + count). + \return 0 if failed, otherwise returns 1 +**/ +unsigned char* + SOIL_load_image_from_memory + ( + const unsigned char *const buffer, + int buffer_length, + int *width, int *height, int *channels, + int force_channels + ); + +/** + Saves an image from an array of unsigned chars (RGBA) to disk + \return 0 if failed, otherwise returns 1 +**/ +int + SOIL_save_image + ( + const char *filename, + int image_type, + int width, int height, int channels, + const unsigned char *const data + ); + +/** + Frees the image data (note, this is just C's "free()"...this function is + present mostly so C++ programmers don't forget to use "free()" and call + "delete []" instead [8^) +**/ +void + SOIL_free_image_data + ( + unsigned char *img_data + ); + +/** + This function resturn a pointer to a string describing the last thing + that happened inside SOIL. It can be used to determine why an image + failed to load. +**/ +const char* + SOIL_last_result + ( + void + ); + + +#ifdef __cplusplus +} +#endif + +#endif /* HEADER_SIMPLE_OPENGL_IMAGE_LIBRARY */ diff --git a/lib/soil/soil/image_DXT.c b/lib/soil/soil/image_DXT.c new file mode 100644 index 0000000..eb90be6 --- /dev/null +++ b/lib/soil/soil/image_DXT.c @@ -0,0 +1,632 @@ +/* + Jonathan Dummer + 2007-07-31-10.32 + + simple DXT compression / decompression code + + public domain +*/ + +#include "image_DXT.h" +#include +#include +#include +#include + +/* set this =1 if you want to use the covarince matrix method... + which is better than my method of using standard deviations + overall, except on the infintesimal chance that the power + method fails for finding the largest eigenvector */ +#define USE_COV_MAT 1 + +/********* Function Prototypes *********/ +/* + Takes a 4x4 block of pixels and compresses it into 8 bytes + in DXT1 format (color only, no alpha). Speed is valued + over prettyness, at least for now. +*/ +void compress_DDS_color_block( + int channels, + const unsigned char *const uncompressed, + unsigned char compressed[8] ); +/* + Takes a 4x4 block of pixels and compresses the alpha + component it into 8 bytes for use in DXT5 DDS files. + Speed is valued over prettyness, at least for now. +*/ +void compress_DDS_alpha_block( + const unsigned char *const uncompressed, + unsigned char compressed[8] ); + +/********* Actual Exposed Functions *********/ +int + save_image_as_DDS + ( + const char *filename, + int width, int height, int channels, + const unsigned char *const data + ) +{ + /* variables */ + FILE *fout; + unsigned char *DDS_data; + DDS_header header; + int DDS_size; + /* error check */ + if( (NULL == filename) || + (width < 1) || (height < 1) || + (channels < 1) || (channels > 4) || + (data == NULL ) ) + { + return 0; + } + /* Convert the image */ + if( (channels & 1) == 1 ) + { + /* no alpha, just use DXT1 */ + DDS_data = convert_image_to_DXT1( data, width, height, channels, &DDS_size ); + } else + { + /* has alpha, so use DXT5 */ + DDS_data = convert_image_to_DXT5( data, width, height, channels, &DDS_size ); + } + /* save it */ + memset( &header, 0, sizeof( DDS_header ) ); + header.dwMagic = ('D' << 0) | ('D' << 8) | ('S' << 16) | (' ' << 24); + header.dwSize = 124; + header.dwFlags = DDSD_CAPS | DDSD_HEIGHT | DDSD_WIDTH | DDSD_PIXELFORMAT | DDSD_LINEARSIZE; + header.dwWidth = width; + header.dwHeight = height; + header.dwPitchOrLinearSize = DDS_size; + header.sPixelFormat.dwSize = 32; + header.sPixelFormat.dwFlags = DDPF_FOURCC; + if( (channels & 1) == 1 ) + { + header.sPixelFormat.dwFourCC = ('D' << 0) | ('X' << 8) | ('T' << 16) | ('1' << 24); + } else + { + header.sPixelFormat.dwFourCC = ('D' << 0) | ('X' << 8) | ('T' << 16) | ('5' << 24); + } + header.sCaps.dwCaps1 = DDSCAPS_TEXTURE; + /* write it out */ + fout = fopen( filename, "wb"); + fwrite( &header, sizeof( DDS_header ), 1, fout ); + fwrite( DDS_data, 1, DDS_size, fout ); + fclose( fout ); + /* done */ + free( DDS_data ); + return 1; +} + +unsigned char* convert_image_to_DXT1( + const unsigned char *const uncompressed, + int width, int height, int channels, + int *out_size ) +{ + unsigned char *compressed; + int i, j, x, y; + unsigned char ublock[16*3]; + unsigned char cblock[8]; + int index = 0, chan_step = 1; + int block_count = 0; + /* error check */ + *out_size = 0; + if( (width < 1) || (height < 1) || + (NULL == uncompressed) || + (channels < 1) || (channels > 4) ) + { + return NULL; + } + /* for channels == 1 or 2, I do not step forward for R,G,B values */ + if( channels < 3 ) + { + chan_step = 0; + } + /* get the RAM for the compressed image + (8 bytes per 4x4 pixel block) */ + *out_size = ((width+3) >> 2) * ((height+3) >> 2) * 8; + compressed = (unsigned char*)malloc( *out_size ); + /* go through each block */ + for( j = 0; j < height; j += 4 ) + { + for( i = 0; i < width; i += 4 ) + { + /* copy this block into a new one */ + int idx = 0; + int mx = 4, my = 4; + if( j+4 >= height ) + { + my = height - j; + } + if( i+4 >= width ) + { + mx = width - i; + } + for( y = 0; y < my; ++y ) + { + for( x = 0; x < mx; ++x ) + { + ublock[idx++] = uncompressed[(j+y)*width*channels+(i+x)*channels]; + ublock[idx++] = uncompressed[(j+y)*width*channels+(i+x)*channels+chan_step]; + ublock[idx++] = uncompressed[(j+y)*width*channels+(i+x)*channels+chan_step+chan_step]; + } + for( x = mx; x < 4; ++x ) + { + ublock[idx++] = ublock[0]; + ublock[idx++] = ublock[1]; + ublock[idx++] = ublock[2]; + } + } + for( y = my; y < 4; ++y ) + { + for( x = 0; x < 4; ++x ) + { + ublock[idx++] = ublock[0]; + ublock[idx++] = ublock[1]; + ublock[idx++] = ublock[2]; + } + } + /* compress the block */ + ++block_count; + compress_DDS_color_block( 3, ublock, cblock ); + /* copy the data from the block into the main block */ + for( x = 0; x < 8; ++x ) + { + compressed[index++] = cblock[x]; + } + } + } + return compressed; +} + +unsigned char* convert_image_to_DXT5( + const unsigned char *const uncompressed, + int width, int height, int channels, + int *out_size ) +{ + unsigned char *compressed; + int i, j, x, y; + unsigned char ublock[16*4]; + unsigned char cblock[8]; + int index = 0, chan_step = 1; + int block_count = 0, has_alpha; + /* error check */ + *out_size = 0; + if( (width < 1) || (height < 1) || + (NULL == uncompressed) || + (channels < 1) || ( channels > 4) ) + { + return NULL; + } + /* for channels == 1 or 2, I do not step forward for R,G,B vales */ + if( channels < 3 ) + { + chan_step = 0; + } + /* # channels = 1 or 3 have no alpha, 2 & 4 do have alpha */ + has_alpha = 1 - (channels & 1); + /* get the RAM for the compressed image + (16 bytes per 4x4 pixel block) */ + *out_size = ((width+3) >> 2) * ((height+3) >> 2) * 16; + compressed = (unsigned char*)malloc( *out_size ); + /* go through each block */ + for( j = 0; j < height; j += 4 ) + { + for( i = 0; i < width; i += 4 ) + { + /* local variables, and my block counter */ + int idx = 0; + int mx = 4, my = 4; + if( j+4 >= height ) + { + my = height - j; + } + if( i+4 >= width ) + { + mx = width - i; + } + for( y = 0; y < my; ++y ) + { + for( x = 0; x < mx; ++x ) + { + ublock[idx++] = uncompressed[(j+y)*width*channels+(i+x)*channels]; + ublock[idx++] = uncompressed[(j+y)*width*channels+(i+x)*channels+chan_step]; + ublock[idx++] = uncompressed[(j+y)*width*channels+(i+x)*channels+chan_step+chan_step]; + ublock[idx++] = + has_alpha * uncompressed[(j+y)*width*channels+(i+x)*channels+channels-1] + + (1-has_alpha)*255; + } + for( x = mx; x < 4; ++x ) + { + ublock[idx++] = ublock[0]; + ublock[idx++] = ublock[1]; + ublock[idx++] = ublock[2]; + ublock[idx++] = ublock[3]; + } + } + for( y = my; y < 4; ++y ) + { + for( x = 0; x < 4; ++x ) + { + ublock[idx++] = ublock[0]; + ublock[idx++] = ublock[1]; + ublock[idx++] = ublock[2]; + ublock[idx++] = ublock[3]; + } + } + /* now compress the alpha block */ + compress_DDS_alpha_block( ublock, cblock ); + /* copy the data from the compressed alpha block into the main buffer */ + for( x = 0; x < 8; ++x ) + { + compressed[index++] = cblock[x]; + } + /* then compress the color block */ + ++block_count; + compress_DDS_color_block( 4, ublock, cblock ); + /* copy the data from the compressed color block into the main buffer */ + for( x = 0; x < 8; ++x ) + { + compressed[index++] = cblock[x]; + } + } + } + return compressed; +} + +/********* Helper Functions *********/ +int convert_bit_range( int c, int from_bits, int to_bits ) +{ + int b = (1 << (from_bits - 1)) + c * ((1 << to_bits) - 1); + return (b + (b >> from_bits)) >> from_bits; +} + +int rgb_to_565( int r, int g, int b ) +{ + return + (convert_bit_range( r, 8, 5 ) << 11) | + (convert_bit_range( g, 8, 6 ) << 05) | + (convert_bit_range( b, 8, 5 ) << 00); +} + +void rgb_888_from_565( unsigned int c, int *r, int *g, int *b ) +{ + *r = convert_bit_range( (c >> 11) & 31, 5, 8 ); + *g = convert_bit_range( (c >> 05) & 63, 6, 8 ); + *b = convert_bit_range( (c >> 00) & 31, 5, 8 ); +} + +void compute_color_line_STDEV( + const unsigned char *const uncompressed, + int channels, + float point[3], float direction[3] ) +{ + const float inv_16 = 1.0f / 16.0f; + int i; + float sum_r = 0.0f, sum_g = 0.0f, sum_b = 0.0f; + float sum_rr = 0.0f, sum_gg = 0.0f, sum_bb = 0.0f; + float sum_rg = 0.0f, sum_rb = 0.0f, sum_gb = 0.0f; + /* calculate all data needed for the covariance matrix + ( to compare with _rygdxt code) */ + for( i = 0; i < 16*channels; i += channels ) + { + sum_r += uncompressed[i+0]; + sum_rr += uncompressed[i+0] * uncompressed[i+0]; + sum_g += uncompressed[i+1]; + sum_gg += uncompressed[i+1] * uncompressed[i+1]; + sum_b += uncompressed[i+2]; + sum_bb += uncompressed[i+2] * uncompressed[i+2]; + sum_rg += uncompressed[i+0] * uncompressed[i+1]; + sum_rb += uncompressed[i+0] * uncompressed[i+2]; + sum_gb += uncompressed[i+1] * uncompressed[i+2]; + } + /* convert the sums to averages */ + sum_r *= inv_16; + sum_g *= inv_16; + sum_b *= inv_16; + /* and convert the squares to the squares of the value - avg_value */ + sum_rr -= 16.0f * sum_r * sum_r; + sum_gg -= 16.0f * sum_g * sum_g; + sum_bb -= 16.0f * sum_b * sum_b; + sum_rg -= 16.0f * sum_r * sum_g; + sum_rb -= 16.0f * sum_r * sum_b; + sum_gb -= 16.0f * sum_g * sum_b; + /* the point on the color line is the average */ + point[0] = sum_r; + point[1] = sum_g; + point[2] = sum_b; + #if USE_COV_MAT + /* + The following idea was from ryg. + (https://mollyrocket.com/forums/viewtopic.php?t=392) + The method worked great (less RMSE than mine) most of + the time, but had some issues handling some simple + boundary cases, like full green next to full red, + which would generate a covariance matrix like this: + + | 1 -1 0 | + | -1 1 0 | + | 0 0 0 | + + For a given starting vector, the power method can + generate all zeros! So no starting with {1,1,1} + as I was doing! This kind of error is still a + slight posibillity, but will be very rare. + */ + /* use the covariance matrix directly + (1st iteration, don't use all 1.0 values!) */ + sum_r = 1.0f; + sum_g = 2.718281828f; + sum_b = 3.141592654f; + direction[0] = sum_r*sum_rr + sum_g*sum_rg + sum_b*sum_rb; + direction[1] = sum_r*sum_rg + sum_g*sum_gg + sum_b*sum_gb; + direction[2] = sum_r*sum_rb + sum_g*sum_gb + sum_b*sum_bb; + /* 2nd iteration, use results from the 1st guy */ + sum_r = direction[0]; + sum_g = direction[1]; + sum_b = direction[2]; + direction[0] = sum_r*sum_rr + sum_g*sum_rg + sum_b*sum_rb; + direction[1] = sum_r*sum_rg + sum_g*sum_gg + sum_b*sum_gb; + direction[2] = sum_r*sum_rb + sum_g*sum_gb + sum_b*sum_bb; + /* 3rd iteration, use results from the 2nd guy */ + sum_r = direction[0]; + sum_g = direction[1]; + sum_b = direction[2]; + direction[0] = sum_r*sum_rr + sum_g*sum_rg + sum_b*sum_rb; + direction[1] = sum_r*sum_rg + sum_g*sum_gg + sum_b*sum_gb; + direction[2] = sum_r*sum_rb + sum_g*sum_gb + sum_b*sum_bb; + #else + /* use my standard deviation method + (very robust, a tiny bit slower and less accurate) */ + direction[0] = sqrt( sum_rr ); + direction[1] = sqrt( sum_gg ); + direction[2] = sqrt( sum_bb ); + /* which has a greater component */ + if( sum_gg > sum_rr ) + { + /* green has greater component, so base the other signs off of green */ + if( sum_rg < 0.0f ) + { + direction[0] = -direction[0]; + } + if( sum_gb < 0.0f ) + { + direction[2] = -direction[2]; + } + } else + { + /* red has a greater component */ + if( sum_rg < 0.0f ) + { + direction[1] = -direction[1]; + } + if( sum_rb < 0.0f ) + { + direction[2] = -direction[2]; + } + } + #endif +} + +void LSE_master_colors_max_min( + int *cmax, int *cmin, + int channels, + const unsigned char *const uncompressed ) +{ + int i, j; + /* the master colors */ + int c0[3], c1[3]; + /* used for fitting the line */ + float sum_x[] = { 0.0f, 0.0f, 0.0f }; + float sum_x2[] = { 0.0f, 0.0f, 0.0f }; + float dot_max = 1.0f, dot_min = -1.0f; + float vec_len2 = 0.0f; + float dot; + /* error check */ + if( (channels < 3) || (channels > 4) ) + { + return; + } + compute_color_line_STDEV( uncompressed, channels, sum_x, sum_x2 ); + vec_len2 = 1.0f / ( 0.00001f + + sum_x2[0]*sum_x2[0] + sum_x2[1]*sum_x2[1] + sum_x2[2]*sum_x2[2] ); + /* finding the max and min vector values */ + dot_max = + ( + sum_x2[0] * uncompressed[0] + + sum_x2[1] * uncompressed[1] + + sum_x2[2] * uncompressed[2] + ); + dot_min = dot_max; + for( i = 1; i < 16; ++i ) + { + dot = + ( + sum_x2[0] * uncompressed[i*channels+0] + + sum_x2[1] * uncompressed[i*channels+1] + + sum_x2[2] * uncompressed[i*channels+2] + ); + if( dot < dot_min ) + { + dot_min = dot; + } else if( dot > dot_max ) + { + dot_max = dot; + } + } + /* and the offset (from the average location) */ + dot = sum_x2[0]*sum_x[0] + sum_x2[1]*sum_x[1] + sum_x2[2]*sum_x[2]; + dot_min -= dot; + dot_max -= dot; + /* post multiply by the scaling factor */ + dot_min *= vec_len2; + dot_max *= vec_len2; + /* OK, build the master colors */ + for( i = 0; i < 3; ++i ) + { + /* color 0 */ + c0[i] = (int)(0.5f + sum_x[i] + dot_max * sum_x2[i]); + if( c0[i] < 0 ) + { + c0[i] = 0; + } else if( c0[i] > 255 ) + { + c0[i] = 255; + } + /* color 1 */ + c1[i] = (int)(0.5f + sum_x[i] + dot_min * sum_x2[i]); + if( c1[i] < 0 ) + { + c1[i] = 0; + } else if( c1[i] > 255 ) + { + c1[i] = 255; + } + } + /* down_sample (with rounding?) */ + i = rgb_to_565( c0[0], c0[1], c0[2] ); + j = rgb_to_565( c1[0], c1[1], c1[2] ); + if( i > j ) + { + *cmax = i; + *cmin = j; + } else + { + *cmax = j; + *cmin = i; + } +} + +void + compress_DDS_color_block + ( + int channels, + const unsigned char *const uncompressed, + unsigned char compressed[8] + ) +{ + /* variables */ + int i; + int next_bit; + int enc_c0, enc_c1; + int c0[4], c1[4]; + float color_line[] = { 0.0f, 0.0f, 0.0f, 0.0f }; + float vec_len2 = 0.0f, dot_offset = 0.0f; + /* stupid order */ + int swizzle4[] = { 0, 2, 3, 1 }; + /* get the master colors */ + LSE_master_colors_max_min( &enc_c0, &enc_c1, channels, uncompressed ); + /* store the 565 color 0 and color 1 */ + compressed[0] = (enc_c0 >> 0) & 255; + compressed[1] = (enc_c0 >> 8) & 255; + compressed[2] = (enc_c1 >> 0) & 255; + compressed[3] = (enc_c1 >> 8) & 255; + /* zero out the compressed data */ + compressed[4] = 0; + compressed[5] = 0; + compressed[6] = 0; + compressed[7] = 0; + /* reconstitute the master color vectors */ + rgb_888_from_565( enc_c0, &c0[0], &c0[1], &c0[2] ); + rgb_888_from_565( enc_c1, &c1[0], &c1[1], &c1[2] ); + /* the new vector */ + vec_len2 = 0.0f; + for( i = 0; i < 3; ++i ) + { + color_line[i] = (float)(c1[i] - c0[i]); + vec_len2 += color_line[i] * color_line[i]; + } + if( vec_len2 > 0.0f ) + { + vec_len2 = 1.0f / vec_len2; + } + /* pre-proform the scaling */ + color_line[0] *= vec_len2; + color_line[1] *= vec_len2; + color_line[2] *= vec_len2; + /* compute the offset (constant) portion of the dot product */ + dot_offset = color_line[0]*c0[0] + color_line[1]*c0[1] + color_line[2]*c0[2]; + /* store the rest of the bits */ + next_bit = 8*4; + for( i = 0; i < 16; ++i ) + { + /* find the dot product of this color, to place it on the line + (should be [-1,1]) */ + int next_value = 0; + float dot_product = + color_line[0] * uncompressed[i*channels+0] + + color_line[1] * uncompressed[i*channels+1] + + color_line[2] * uncompressed[i*channels+2] - + dot_offset; + /* map to [0,3] */ + next_value = (int)( dot_product * 3.0f + 0.5f ); + if( next_value > 3 ) + { + next_value = 3; + } else if( next_value < 0 ) + { + next_value = 0; + } + /* OK, store this value */ + compressed[next_bit >> 3] |= swizzle4[ next_value ] << (next_bit & 7); + next_bit += 2; + } + /* done compressing to DXT1 */ +} + +void + compress_DDS_alpha_block + ( + const unsigned char *const uncompressed, + unsigned char compressed[8] + ) +{ + /* variables */ + int i; + int next_bit; + int a0, a1; + float scale_me; + /* stupid order */ + int swizzle8[] = { 1, 7, 6, 5, 4, 3, 2, 0 }; + /* get the alpha limits (a0 > a1) */ + a0 = a1 = uncompressed[3]; + for( i = 4+3; i < 16*4; i += 4 ) + { + if( uncompressed[i] > a0 ) + { + a0 = uncompressed[i]; + } else if( uncompressed[i] < a1 ) + { + a1 = uncompressed[i]; + } + } + /* store those limits, and zero the rest of the compressed dataset */ + compressed[0] = a0; + compressed[1] = a1; + /* zero out the compressed data */ + compressed[2] = 0; + compressed[3] = 0; + compressed[4] = 0; + compressed[5] = 0; + compressed[6] = 0; + compressed[7] = 0; + /* store the all of the alpha values */ + next_bit = 8*2; + scale_me = 7.9999f / (a0 - a1); + for( i = 3; i < 16*4; i += 4 ) + { + /* convert this alpha value to a 3 bit number */ + int svalue; + int value = (int)((uncompressed[i] - a1) * scale_me); + svalue = swizzle8[ value&7 ]; + /* OK, store this value, start with the 1st byte */ + compressed[next_bit >> 3] |= svalue << (next_bit & 7); + if( (next_bit & 7) > 5 ) + { + /* spans 2 bytes, fill in the start of the 2nd byte */ + compressed[1 + (next_bit >> 3)] |= svalue >> (8 - (next_bit & 7) ); + } + next_bit += 3; + } + /* done compressing to DXT1 */ +} diff --git a/lib/soil/soil/image_DXT.h b/lib/soil/soil/image_DXT.h new file mode 100644 index 0000000..ce77164 --- /dev/null +++ b/lib/soil/soil/image_DXT.h @@ -0,0 +1,123 @@ +/* + Jonathan Dummer + 2007-07-31-10.32 + + simple DXT compression / decompression code + + public domain +*/ + +#ifndef HEADER_IMAGE_DXT +#define HEADER_IMAGE_DXT + +/** + Converts an image from an array of unsigned chars (RGB or RGBA) to + DXT1 or DXT5, then saves the converted image to disk. + \return 0 if failed, otherwise returns 1 +**/ +int +save_image_as_DDS +( + const char *filename, + int width, int height, int channels, + const unsigned char *const data +); + +/** + take an image and convert it to DXT1 (no alpha) +**/ +unsigned char* +convert_image_to_DXT1 +( + const unsigned char *const uncompressed, + int width, int height, int channels, + int *out_size +); + +/** + take an image and convert it to DXT5 (with alpha) +**/ +unsigned char* +convert_image_to_DXT5 +( + const unsigned char *const uncompressed, + int width, int height, int channels, + int *out_size +); + +/** A bunch of DirectDraw Surface structures and flags **/ +typedef struct +{ + unsigned int dwMagic; + unsigned int dwSize; + unsigned int dwFlags; + unsigned int dwHeight; + unsigned int dwWidth; + unsigned int dwPitchOrLinearSize; + unsigned int dwDepth; + unsigned int dwMipMapCount; + unsigned int dwReserved1[ 11 ]; + + /* DDPIXELFORMAT */ + struct + { + unsigned int dwSize; + unsigned int dwFlags; + unsigned int dwFourCC; + unsigned int dwRGBBitCount; + unsigned int dwRBitMask; + unsigned int dwGBitMask; + unsigned int dwBBitMask; + unsigned int dwAlphaBitMask; + } + sPixelFormat; + + /* DDCAPS2 */ + struct + { + unsigned int dwCaps1; + unsigned int dwCaps2; + unsigned int dwDDSX; + unsigned int dwReserved; + } + sCaps; + unsigned int dwReserved2; +} +DDS_header ; + +/* the following constants were copied directly off the MSDN website */ + +/* The dwFlags member of the original DDSURFACEDESC2 structure + can be set to one or more of the following values. */ +#define DDSD_CAPS 0x00000001 +#define DDSD_HEIGHT 0x00000002 +#define DDSD_WIDTH 0x00000004 +#define DDSD_PITCH 0x00000008 +#define DDSD_PIXELFORMAT 0x00001000 +#define DDSD_MIPMAPCOUNT 0x00020000 +#define DDSD_LINEARSIZE 0x00080000 +#define DDSD_DEPTH 0x00800000 + +/* DirectDraw Pixel Format */ +#define DDPF_ALPHAPIXELS 0x00000001 +#define DDPF_FOURCC 0x00000004 +#define DDPF_RGB 0x00000040 + +/* The dwCaps1 member of the DDSCAPS2 structure can be + set to one or more of the following values. */ +#define DDSCAPS_COMPLEX 0x00000008 +#define DDSCAPS_TEXTURE 0x00001000 +#define DDSCAPS_MIPMAP 0x00400000 + +/* The dwCaps2 member of the DDSCAPS2 structure can be + set to one or more of the following values. */ +#define DDSCAPS2_CUBEMAP 0x00000200 +#define DDSCAPS2_CUBEMAP_POSITIVEX 0x00000400 +#define DDSCAPS2_CUBEMAP_NEGATIVEX 0x00000800 +#define DDSCAPS2_CUBEMAP_POSITIVEY 0x00001000 +#define DDSCAPS2_CUBEMAP_NEGATIVEY 0x00002000 +#define DDSCAPS2_CUBEMAP_POSITIVEZ 0x00004000 +#define DDSCAPS2_CUBEMAP_NEGATIVEZ 0x00008000 +#define DDSCAPS2_VOLUME 0x00200000 + +#endif /* HEADER_IMAGE_DXT */ diff --git a/lib/soil/soil/image_helper.c b/lib/soil/soil/image_helper.c new file mode 100644 index 0000000..445f6bb --- /dev/null +++ b/lib/soil/soil/image_helper.c @@ -0,0 +1,435 @@ +/* + Jonathan Dummer + + image helper functions + + MIT license +*/ + +#include "image_helper.h" +#include +#include + +/* Upscaling the image uses simple bilinear interpolation */ +int + up_scale_image + ( + const unsigned char* const orig, + int width, int height, int channels, + unsigned char* resampled, + int resampled_width, int resampled_height + ) +{ + float dx, dy; + int x, y, c; + + /* error(s) check */ + if ( (width < 1) || (height < 1) || + (resampled_width < 2) || (resampled_height < 2) || + (channels < 1) || + (NULL == orig) || (NULL == resampled) ) + { + /* signify badness */ + return 0; + } + /* + for each given pixel in the new map, find the exact location + from the original map which would contribute to this guy + */ + dx = (width - 1.0f) / (resampled_width - 1.0f); + dy = (height - 1.0f) / (resampled_height - 1.0f); + for ( y = 0; y < resampled_height; ++y ) + { + /* find the base y index and fractional offset from that */ + float sampley = y * dy; + int inty = (int)sampley; + /* if( inty < 0 ) { inty = 0; } else */ + if( inty > height - 2 ) { inty = height - 2; } + sampley -= inty; + for ( x = 0; x < resampled_width; ++x ) + { + float samplex = x * dx; + int intx = (int)samplex; + int base_index; + /* find the base x index and fractional offset from that */ + /* if( intx < 0 ) { intx = 0; } else */ + if( intx > width - 2 ) { intx = width - 2; } + samplex -= intx; + /* base index into the original image */ + base_index = (inty * width + intx) * channels; + for ( c = 0; c < channels; ++c ) + { + /* do the sampling */ + float value = 0.5f; + value += orig[base_index] + *(1.0f-samplex)*(1.0f-sampley); + value += orig[base_index+channels] + *(samplex)*(1.0f-sampley); + value += orig[base_index+width*channels] + *(1.0f-samplex)*(sampley); + value += orig[base_index+width*channels+channels] + *(samplex)*(sampley); + /* move to the next channel */ + ++base_index; + /* save the new value */ + resampled[y*resampled_width*channels+x*channels+c] = + (unsigned char)(value); + } + } + } + /* done */ + return 1; +} + +int + mipmap_image + ( + const unsigned char* const orig, + int width, int height, int channels, + unsigned char* resampled, + int block_size_x, int block_size_y + ) +{ + int mip_width, mip_height; + int i, j, c; + + /* error check */ + if( (width < 1) || (height < 1) || + (channels < 1) || (orig == NULL) || + (resampled == NULL) || + (block_size_x < 1) || (block_size_y < 1) ) + { + /* nothing to do */ + return 0; + } + mip_width = width / block_size_x; + mip_height = height / block_size_y; + if( mip_width < 1 ) + { + mip_width = 1; + } + if( mip_height < 1 ) + { + mip_height = 1; + } + for( j = 0; j < mip_height; ++j ) + { + for( i = 0; i < mip_width; ++i ) + { + for( c = 0; c < channels; ++c ) + { + const int index = (j*block_size_y)*width*channels + (i*block_size_x)*channels + c; + int sum_value; + int u,v; + int u_block = block_size_x; + int v_block = block_size_y; + int block_area; + /* do a bit of checking so we don't over-run the boundaries + (necessary for non-square textures!) */ + if( block_size_x * (i+1) > width ) + { + u_block = width - i*block_size_y; + } + if( block_size_y * (j+1) > height ) + { + v_block = height - j*block_size_y; + } + block_area = u_block*v_block; + /* for this pixel, see what the average + of all the values in the block are. + note: start the sum at the rounding value, not at 0 */ + sum_value = block_area >> 1; + for( v = 0; v < v_block; ++v ) + for( u = 0; u < u_block; ++u ) + { + sum_value += orig[index + v*width*channels + u*channels]; + } + resampled[j*mip_width*channels + i*channels + c] = sum_value / block_area; + } + } + } + return 1; +} + +int + scale_image_RGB_to_NTSC_safe + ( + unsigned char* orig, + int width, int height, int channels + ) +{ + const float scale_lo = 16.0f - 0.499f; + const float scale_hi = 235.0f + 0.499f; + int i, j; + int nc = channels; + unsigned char scale_LUT[256]; + /* error check */ + if( (width < 1) || (height < 1) || + (channels < 1) || (orig == NULL) ) + { + /* nothing to do */ + return 0; + } + /* set up the scaling Look Up Table */ + for( i = 0; i < 256; ++i ) + { + scale_LUT[i] = (unsigned char)((scale_hi - scale_lo) * i / 255.0f + scale_lo); + } + /* for channels = 2 or 4, ignore the alpha component */ + nc -= 1 - (channels & 1); + /* OK, go through the image and scale any non-alpha components */ + for( i = 0; i < width*height*channels; i += channels ) + { + for( j = 0; j < nc; ++j ) + { + orig[i+j] = scale_LUT[orig[i+j]]; + } + } + return 1; +} + +unsigned char clamp_byte( int x ) { return ( (x) < 0 ? (0) : ( (x) > 255 ? 255 : (x) ) ); } + +/* + This function takes the RGB components of the image + and converts them into YCoCg. 3 components will be + re-ordered to CoYCg (for optimum DXT1 compression), + while 4 components will be ordered CoCgAY (for DXT5 + compression). +*/ +int + convert_RGB_to_YCoCg + ( + unsigned char* orig, + int width, int height, int channels + ) +{ + int i; + /* error check */ + if( (width < 1) || (height < 1) || + (channels < 3) || (channels > 4) || + (orig == NULL) ) + { + /* nothing to do */ + return -1; + } + /* do the conversion */ + if( channels == 3 ) + { + for( i = 0; i < width*height*3; i += 3 ) + { + int r = orig[i+0]; + int g = (orig[i+1] + 1) >> 1; + int b = orig[i+2]; + int tmp = (2 + r + b) >> 2; + /* Co */ + orig[i+0] = clamp_byte( 128 + ((r - b + 1) >> 1) ); + /* Y */ + orig[i+1] = clamp_byte( g + tmp ); + /* Cg */ + orig[i+2] = clamp_byte( 128 + g - tmp ); + } + } else + { + for( i = 0; i < width*height*4; i += 4 ) + { + int r = orig[i+0]; + int g = (orig[i+1] + 1) >> 1; + int b = orig[i+2]; + unsigned char a = orig[i+3]; + int tmp = (2 + r + b) >> 2; + /* Co */ + orig[i+0] = clamp_byte( 128 + ((r - b + 1) >> 1) ); + /* Cg */ + orig[i+1] = clamp_byte( 128 + g - tmp ); + /* Alpha */ + orig[i+2] = a; + /* Y */ + orig[i+3] = clamp_byte( g + tmp ); + } + } + /* done */ + return 0; +} + +/* + This function takes the YCoCg components of the image + and converts them into RGB. See above. +*/ +int + convert_YCoCg_to_RGB + ( + unsigned char* orig, + int width, int height, int channels + ) +{ + int i; + /* error check */ + if( (width < 1) || (height < 1) || + (channels < 3) || (channels > 4) || + (orig == NULL) ) + { + /* nothing to do */ + return -1; + } + /* do the conversion */ + if( channels == 3 ) + { + for( i = 0; i < width*height*3; i += 3 ) + { + int co = orig[i+0] - 128; + int y = orig[i+1]; + int cg = orig[i+2] - 128; + /* R */ + orig[i+0] = clamp_byte( y + co - cg ); + /* G */ + orig[i+1] = clamp_byte( y + cg ); + /* B */ + orig[i+2] = clamp_byte( y - co - cg ); + } + } else + { + for( i = 0; i < width*height*4; i += 4 ) + { + int co = orig[i+0] - 128; + int cg = orig[i+1] - 128; + unsigned char a = orig[i+2]; + int y = orig[i+3]; + /* R */ + orig[i+0] = clamp_byte( y + co - cg ); + /* G */ + orig[i+1] = clamp_byte( y + cg ); + /* B */ + orig[i+2] = clamp_byte( y - co - cg ); + /* A */ + orig[i+3] = a; + } + } + /* done */ + return 0; +} + +float +find_max_RGBE +( + unsigned char *image, + int width, int height +) +{ + float max_val = 0.0f; + unsigned char *img = image; + int i, j; + for( i = width * height; i > 0; --i ) + { + /* float scale = powf( 2.0f, img[3] - 128.0f ) / 255.0f; */ + float scale = ldexp( 1.0f / 255.0f, (int)(img[3]) - 128 ); + for( j = 0; j < 3; ++j ) + { + if( img[j] * scale > max_val ) + { + max_val = img[j] * scale; + } + } + /* next pixel */ + img += 4; + } + return max_val; +} + +int +RGBE_to_RGBdivA +( + unsigned char *image, + int width, int height, + int rescale_to_max +) +{ + /* local variables */ + int i, iv; + unsigned char *img = image; + float scale = 1.0f; + /* error check */ + if( (!image) || (width < 1) || (height < 1) ) + { + return 0; + } + /* convert (note: no negative numbers, but 0.0 is possible) */ + if( rescale_to_max ) + { + scale = 255.0f / find_max_RGBE( image, width, height ); + } + for( i = width * height; i > 0; --i ) + { + /* decode this pixel, and find the max */ + float r,g,b,e, m; + /* e = scale * powf( 2.0f, img[3] - 128.0f ) / 255.0f; */ + e = scale * ldexp( 1.0f / 255.0f, (int)(img[3]) - 128 ); + r = e * img[0]; + g = e * img[1]; + b = e * img[2]; + m = (r > g) ? r : g; + m = (b > m) ? b : m; + /* and encode it into RGBdivA */ + iv = (m != 0.0f) ? (int)(255.0f / m) : 1.0f; + iv = (iv < 1) ? 1 : iv; + img[3] = (iv > 255) ? 255 : iv; + iv = (int)(img[3] * r + 0.5f); + img[0] = (iv > 255) ? 255 : iv; + iv = (int)(img[3] * g + 0.5f); + img[1] = (iv > 255) ? 255 : iv; + iv = (int)(img[3] * b + 0.5f); + img[2] = (iv > 255) ? 255 : iv; + /* and on to the next pixel */ + img += 4; + } + return 1; +} + +int +RGBE_to_RGBdivA2 +( + unsigned char *image, + int width, int height, + int rescale_to_max +) +{ + /* local variables */ + int i, iv; + unsigned char *img = image; + float scale = 1.0f; + /* error check */ + if( (!image) || (width < 1) || (height < 1) ) + { + return 0; + } + /* convert (note: no negative numbers, but 0.0 is possible) */ + if( rescale_to_max ) + { + scale = 255.0f * 255.0f / find_max_RGBE( image, width, height ); + } + for( i = width * height; i > 0; --i ) + { + /* decode this pixel, and find the max */ + float r,g,b,e, m; + /* e = scale * powf( 2.0f, img[3] - 128.0f ) / 255.0f; */ + e = scale * ldexp( 1.0f / 255.0f, (int)(img[3]) - 128 ); + r = e * img[0]; + g = e * img[1]; + b = e * img[2]; + m = (r > g) ? r : g; + m = (b > m) ? b : m; + /* and encode it into RGBdivA */ + iv = (m != 0.0f) ? (int)sqrtf( 255.0f * 255.0f / m ) : 1.0f; + iv = (iv < 1) ? 1 : iv; + img[3] = (iv > 255) ? 255 : iv; + iv = (int)(img[3] * img[3] * r / 255.0f + 0.5f); + img[0] = (iv > 255) ? 255 : iv; + iv = (int)(img[3] * img[3] * g / 255.0f + 0.5f); + img[1] = (iv > 255) ? 255 : iv; + iv = (int)(img[3] * img[3] * b / 255.0f + 0.5f); + img[2] = (iv > 255) ? 255 : iv; + /* and on to the next pixel */ + img += 4; + } + return 1; +} diff --git a/lib/soil/soil/image_helper.h b/lib/soil/soil/image_helper.h new file mode 100644 index 0000000..abb257c --- /dev/null +++ b/lib/soil/soil/image_helper.h @@ -0,0 +1,115 @@ +/* + Jonathan Dummer + + Image helper functions + + MIT license +*/ + +#ifndef HEADER_IMAGE_HELPER +#define HEADER_IMAGE_HELPER + +#ifdef __cplusplus +extern "C" { +#endif + +/** + This function upscales an image. + Not to be used to create MIPmaps, + but to make it square, + or to make it a power-of-two sized. +**/ +int + up_scale_image + ( + const unsigned char* const orig, + int width, int height, int channels, + unsigned char* resampled, + int resampled_width, int resampled_height + ); + +/** + This function downscales an image. + Used for creating MIPmaps, + the incoming image should be a + power-of-two sized. +**/ +int + mipmap_image + ( + const unsigned char* const orig, + int width, int height, int channels, + unsigned char* resampled, + int block_size_x, int block_size_y + ); + +/** + This function takes the RGB components of the image + and scales each channel from [0,255] to [16,235]. + This makes the colors "Safe" for display on NTSC + displays. Note that this is _NOT_ a good idea for + loading images like normal- or height-maps! +**/ +int + scale_image_RGB_to_NTSC_safe + ( + unsigned char* orig, + int width, int height, int channels + ); + +/** + This function takes the RGB components of the image + and converts them into YCoCg. 3 components will be + re-ordered to CoYCg (for optimum DXT1 compression), + while 4 components will be ordered CoCgAY (for DXT5 + compression). +**/ +int + convert_RGB_to_YCoCg + ( + unsigned char* orig, + int width, int height, int channels + ); + +/** + This function takes the YCoCg components of the image + and converts them into RGB. See above. +**/ +int + convert_YCoCg_to_RGB + ( + unsigned char* orig, + int width, int height, int channels + ); + +/** + Converts an HDR image from an array + of unsigned chars (RGBE) to RGBdivA + \return 0 if failed, otherwise returns 1 +**/ +int + RGBE_to_RGBdivA + ( + unsigned char *image, + int width, int height, + int rescale_to_max + ); + +/** + Converts an HDR image from an array + of unsigned chars (RGBE) to RGBdivA2 + \return 0 if failed, otherwise returns 1 +**/ +int + RGBE_to_RGBdivA2 + ( + unsigned char *image, + int width, int height, + int rescale_to_max + ); + +#ifdef __cplusplus +} +#endif + +#endif /* HEADER_IMAGE_HELPER */ diff --git a/lib/soil/soil/original/stb_image-1.09.c b/lib/soil/soil/original/stb_image-1.09.c new file mode 100644 index 0000000..a30c7da --- /dev/null +++ b/lib/soil/soil/original/stb_image-1.09.c @@ -0,0 +1,3632 @@ +/* stbi-1.09 - public domain JPEG/PNG reader - http://nothings.org/stb_image.c + when you control the images you're loading + + QUICK NOTES: + Primarily of interest to game developers and other people who can + avoid problematic images and only need the trivial interface + + JPEG baseline (no JPEG progressive, no oddball channel decimations) + PNG non-interlaced + BMP non-1bpp, non-RLE + TGA (not sure what subset, if a subset) + PSD (composited view only, no extra channels) + HDR (radiance rgbE format) + writes BMP,TGA (define STBI_NO_WRITE to remove code) + decoded from memory or through stdio FILE (define STBI_NO_STDIO to remove code) + + TODO: + stbi_info_* + + history: + 1.09 Fix format-conversion for PSD code (bad global variables!) + 1.08 Thatcher Ulrich's PSD code integrated by Nicolas Schulz + 1.07 attempt to fix C++ warning/errors again + 1.06 attempt to fix C++ warning/errors again + 1.05 fix TGA loading to return correct *comp and use good luminance calc + 1.04 default float alpha is 1, not 255; use 'void *' for stbi_image_free + 1.03 bugfixes to STBI_NO_STDIO, STBI_NO_HDR + 1.02 support for (subset of) HDR files, float interface for preferred access to them + 1.01 fix bug: possible bug in handling right-side up bmps... not sure + fix bug: the stbi_bmp_load() and stbi_tga_load() functions didn't work at all + 1.00 interface to zlib that skips zlib header + 0.99 correct handling of alpha in palette + 0.98 TGA loader by lonesock; dynamically add loaders (untested) + 0.97 jpeg errors on too large a file; also catch another malloc failure + 0.96 fix detection of invalid v value - particleman@mollyrocket forum + 0.95 during header scan, seek to markers in case of padding + 0.94 STBI_NO_STDIO to disable stdio usage; rename all #defines the same + 0.93 handle jpegtran output; verbose errors + 0.92 read 4,8,16,24,32-bit BMP files of several formats + 0.91 output 24-bit Windows 3.0 BMP files + 0.90 fix a few more warnings; bump version number to approach 1.0 + 0.61 bugfixes due to Marc LeBlanc, Christopher Lloyd + 0.60 fix compiling as c++ + 0.59 fix warnings: merge Dave Moore's -Wall fixes + 0.58 fix bug: zlib uncompressed mode len/nlen was wrong endian + 0.57 fix bug: jpg last huffman symbol before marker was >9 bits but less + than 16 available + 0.56 fix bug: zlib uncompressed mode len vs. nlen + 0.55 fix bug: restart_interval not initialized to 0 + 0.54 allow NULL for 'int *comp' + 0.53 fix bug in png 3->4; speedup png decoding + 0.52 png handles req_comp=3,4 directly; minor cleanup; jpeg comments + 0.51 obey req_comp requests, 1-component jpegs return as 1-component, + on 'test' only check type, not whether we support this variant +*/ + + +//// begin header file //////////////////////////////////////////////////// +// +// Limitations: +// - no progressive/interlaced support (jpeg, png) +// - 8-bit samples only (jpeg, png) +// - not threadsafe +// - channel subsampling of at most 2 in each dimension (jpeg) +// - no delayed line count (jpeg) -- IJG doesn't support either +// +// Basic usage (see HDR discussion below): +// int x,y,n; +// unsigned char *data = stbi_load(filename, &x, &y, &n, 0); +// // ... process data if not NULL ... +// // ... x = width, y = height, n = # 8-bit components per pixel ... +// // ... replace '0' with '1'..'4' to force that many components per pixel +// stbi_image_free(data) +// +// Standard parameters: +// int *x -- outputs image width in pixels +// int *y -- outputs image height in pixels +// int *comp -- outputs # of image components in image file +// int req_comp -- if non-zero, # of image components requested in result +// +// The return value from an image loader is an 'unsigned char *' which points +// to the pixel data. The pixel data consists of *y scanlines of *x pixels, +// with each pixel consisting of N interleaved 8-bit components; the first +// pixel pointed to is top-left-most in the image. There is no padding between +// image scanlines or between pixels, regardless of format. The number of +// components N is 'req_comp' if req_comp is non-zero, or *comp otherwise. +// If req_comp is non-zero, *comp has the number of components that _would_ +// have been output otherwise. E.g. if you set req_comp to 4, you will always +// get RGBA output, but you can check *comp to easily see if it's opaque. +// +// An output image with N components has the following components interleaved +// in this order in each pixel: +// +// N=#comp components +// 1 grey +// 2 grey, alpha +// 3 red, green, blue +// 4 red, green, blue, alpha +// +// If image loading fails for any reason, the return value will be NULL, +// and *x, *y, *comp will be unchanged. The function stbi_failure_reason() +// can be queried for an extremely brief, end-user unfriendly explanation +// of why the load failed. Define STBI_NO_FAILURE_STRINGS to avoid +// compiling these strings at all, and STBI_FAILURE_USERMSG to get slightly +// more user-friendly ones. +// +// Paletted PNG and BMP images are automatically depalettized. +// +// +// =========================================================================== +// +// HDR image support (disable by defining STBI_NO_HDR) +// +// stb_image now supports loading HDR images in general, and currently +// the Radiance .HDR file format, although the support is provided +// generically. You can still load any file through the existing interface; +// if you attempt to load an HDR file, it will be automatically remapped to +// LDR, assuming gamma 2.2 and an arbitrary scale factor defaulting to 1; +// both of these constants can be reconfigured through this interface: +// +// stbi_hdr_to_ldr_gamma(2.2f); +// stbi_hdr_to_ldr_scale(1.0f); +// +// (note, do not use _inverse_ constants; stbi_image will invert them +// appropriately). +// +// Additionally, there is a new, parallel interface for loading files as +// (linear) floats to preserve the full dynamic range: +// +// float *data = stbi_loadf(filename, &x, &y, &n, 0); +// +// If you load LDR images through this interface, those images will +// be promoted to floating point values, run through the inverse of +// constants corresponding to the above: +// +// stbi_ldr_to_hdr_scale(1.0f); +// stbi_ldr_to_hdr_gamma(2.2f); +// +// Finally, given a filename (or an open file or memory block--see header +// file for details) containing image data, you can query for the "most +// appropriate" interface to use (that is, whether the image is HDR or +// not), using: +// +// stbi_is_hdr(char *filename); + + +#ifndef STBI_NO_STDIO +#include +#endif + +#ifndef STBI_NO_HDR +#include // ldexp +#include // strcmp +#endif + +enum +{ + STBI_default = 0, // only used for req_comp + + STBI_grey = 1, + STBI_grey_alpha = 2, + STBI_rgb = 3, + STBI_rgb_alpha = 4, +}; + +typedef unsigned char stbi_uc; + +#ifdef __cplusplus +extern "C" { +#endif + +// WRITING API + +#if !defined(STBI_NO_WRITE) && !defined(STBI_NO_STDIO) +// write a BMP/TGA file given tightly packed 'comp' channels (no padding, nor bmp-stride-padding) +// (you must include the appropriate extension in the filename). +// returns TRUE on success, FALSE if couldn't open file, error writing file +extern int stbi_write_bmp (char *filename, int x, int y, int comp, void *data); +extern int stbi_write_tga (char *filename, int x, int y, int comp, void *data); +#endif + +// PRIMARY API - works on images of any type + +// load image by filename, open file, or memory buffer +#ifndef STBI_NO_STDIO +extern stbi_uc *stbi_load (char *filename, int *x, int *y, int *comp, int req_comp); +extern stbi_uc *stbi_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +extern int stbi_info_from_file (FILE *f, int *x, int *y, int *comp); +#endif +extern stbi_uc *stbi_load_from_memory(stbi_uc *buffer, int len, int *x, int *y, int *comp, int req_comp); +// for stbi_load_from_file, file pointer is left pointing immediately after image + +#ifndef STBI_NO_HDR +#ifndef STBI_NO_STDIO +extern float *stbi_loadf (char *filename, int *x, int *y, int *comp, int req_comp); +extern float *stbi_loadf_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +#endif +extern float *stbi_loadf_from_memory(stbi_uc *buffer, int len, int *x, int *y, int *comp, int req_comp); + +extern void stbi_hdr_to_ldr_gamma(float gamma); +extern void stbi_hdr_to_ldr_scale(float scale); + +extern void stbi_ldr_to_hdr_gamma(float gamma); +extern void stbi_ldr_to_hdr_scale(float scale); + +#endif // STBI_NO_HDR + +// get a VERY brief reason for failure +extern char *stbi_failure_reason (void); + +// free the loaded image -- this is just free() +extern void stbi_image_free (void *retval_from_stbi_load); + +// get image dimensions & components without fully decoding +extern int stbi_info_from_memory(stbi_uc *buffer, int len, int *x, int *y, int *comp); +extern int stbi_is_hdr_from_memory(stbi_uc *buffer, int len); +#ifndef STBI_NO_STDIO +extern int stbi_info (char *filename, int *x, int *y, int *comp); +extern int stbi_is_hdr (char *filename); +extern int stbi_is_hdr_from_file(FILE *f); +#endif + +// ZLIB client - used by PNG, available for other purposes + +extern char *stbi_zlib_decode_malloc_guesssize(int initial_size, int *outlen); +extern char *stbi_zlib_decode_malloc(char *buffer, int len, int *outlen); +extern int stbi_zlib_decode_buffer(char *obuffer, int olen, char *ibuffer, int ilen); + +extern char *stbi_zlib_decode_noheader_malloc(char *buffer, int len, int *outlen); +extern int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, char *ibuffer, int ilen); + + +// TYPE-SPECIFIC ACCESS + +// is it a jpeg? +extern int stbi_jpeg_test_memory (stbi_uc *buffer, int len); +extern stbi_uc *stbi_jpeg_load_from_memory(stbi_uc *buffer, int len, int *x, int *y, int *comp, int req_comp); +extern int stbi_jpeg_info_from_memory(stbi_uc *buffer, int len, int *x, int *y, int *comp); + +#ifndef STBI_NO_STDIO +extern stbi_uc *stbi_jpeg_load (char *filename, int *x, int *y, int *comp, int req_comp); +extern int stbi_jpeg_test_file (FILE *f); +extern stbi_uc *stbi_jpeg_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); + +extern int stbi_jpeg_info (char *filename, int *x, int *y, int *comp); +extern int stbi_jpeg_info_from_file (FILE *f, int *x, int *y, int *comp); +#endif + +extern int stbi_jpeg_dc_only; // only decode DC component + +// is it a png? +extern int stbi_png_test_memory (stbi_uc *buffer, int len); +extern stbi_uc *stbi_png_load_from_memory (stbi_uc *buffer, int len, int *x, int *y, int *comp, int req_comp); +extern int stbi_png_info_from_memory (stbi_uc *buffer, int len, int *x, int *y, int *comp); + +#ifndef STBI_NO_STDIO +extern stbi_uc *stbi_png_load (char *filename, int *x, int *y, int *comp, int req_comp); +extern int stbi_png_info (char *filename, int *x, int *y, int *comp); +extern int stbi_png_test_file (FILE *f); +extern stbi_uc *stbi_png_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +extern int stbi_png_info_from_file (FILE *f, int *x, int *y, int *comp); +#endif + +// is it a bmp? +extern int stbi_bmp_test_memory (stbi_uc *buffer, int len); + +extern stbi_uc *stbi_bmp_load (char *filename, int *x, int *y, int *comp, int req_comp); +extern stbi_uc *stbi_bmp_load_from_memory (stbi_uc *buffer, int len, int *x, int *y, int *comp, int req_comp); +#ifndef STBI_NO_STDIO +extern int stbi_bmp_test_file (FILE *f); +extern stbi_uc *stbi_bmp_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +#endif + +// is it a tga? +extern int stbi_tga_test_memory (stbi_uc *buffer, int len); + +extern stbi_uc *stbi_tga_load (char *filename, int *x, int *y, int *comp, int req_comp); +extern stbi_uc *stbi_tga_load_from_memory (stbi_uc *buffer, int len, int *x, int *y, int *comp, int req_comp); +#ifndef STBI_NO_STDIO +extern int stbi_tga_test_file (FILE *f); +extern stbi_uc *stbi_tga_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +#endif + +// is it a psd? +extern int stbi_psd_test_memory (stbi_uc *buffer, int len); + +extern stbi_uc *stbi_psd_load (char *filename, int *x, int *y, int *comp, int req_comp); +extern stbi_uc *stbi_psd_load_from_memory (stbi_uc *buffer, int len, int *x, int *y, int *comp, int req_comp); +#ifndef STBI_NO_STDIO +extern int stbi_psd_test_file (FILE *f); +extern stbi_uc *stbi_psd_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +#endif + +// is it an hdr? +extern int stbi_hdr_test_memory (stbi_uc *buffer, int len); + +extern float * stbi_hdr_load (char *filename, int *x, int *y, int *comp, int req_comp); +extern float * stbi_hdr_load_from_memory (stbi_uc *buffer, int len, int *x, int *y, int *comp, int req_comp); +#ifndef STBI_NO_STDIO +extern int stbi_hdr_test_file (FILE *f); +extern float * stbi_hdr_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +#endif + +// define new loaders +typedef struct +{ + int (*test_memory)(stbi_uc *buffer, int len); + stbi_uc * (*load_from_memory)(stbi_uc *buffer, int len, int *x, int *y, int *comp, int req_comp); + #ifndef STBI_NO_STDIO + int (*test_file)(FILE *f); + stbi_uc * (*load_from_file)(FILE *f, int *x, int *y, int *comp, int req_comp); + #endif +} stbi_loader; + +// register a loader by filling out the above structure (you must defined ALL functions) +// returns 1 if added or already added, 0 if not added (too many loaders) +extern int stbi_register_loader(stbi_loader *loader); + +#ifdef __cplusplus +} +#endif + +// +// +//// end header file ///////////////////////////////////////////////////// + +#ifndef STBI_NO_STDIO +#include +#endif +#include +#include +#include +#include + +#ifndef _MSC_VER +#define __forceinline +#endif + +// implementation: +typedef unsigned char uint8; +typedef unsigned short uint16; +typedef signed short int16; +typedef unsigned int uint32; +typedef signed int int32; +typedef unsigned int uint; + +// should produce compiler error if size is wrong +typedef unsigned char validate_uint32[sizeof(uint32)==4]; + +#if defined(STBI_NO_STDIO) && !defined(STBI_NO_WRITE) +#define STBI_NO_WRITE +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// Generic API that works on all image types +// + +static char *failure_reason; + +char *stbi_failure_reason(void) +{ + return failure_reason; +} + +static int e(char *str) +{ + failure_reason = str; + return 0; +} + +#ifdef STBI_NO_FAILURE_STRINGS + #define e(x,y) 0 +#elif defined(STBI_FAILURE_USERMSG) + #define e(x,y) e(y) +#else + #define e(x,y) e(x) +#endif + +#define epf(x,y) ((float *) (e(x,y)?NULL:NULL)) +#define epuc(x,y) ((unsigned char *) (e(x,y)?NULL:NULL)) + +void stbi_image_free(void *retval_from_stbi_load) +{ + free(retval_from_stbi_load); +} + +#define MAX_LOADERS 32 +stbi_loader *loaders[MAX_LOADERS]; +static int max_loaders = 0; + +int stbi_register_loader(stbi_loader *loader) +{ + int i; + for (i=0; i < MAX_LOADERS; ++i) { + // already present? + if (loaders[i] == loader) + return 1; + // end of the list? + if (loaders[i] == NULL) { + loaders[i] = loader; + max_loaders = i+1; + return 1; + } + } + // no room for it + return 0; +} + +#ifndef STBI_NO_HDR +static float *ldr_to_hdr(stbi_uc *data, int x, int y, int comp); +static stbi_uc *hdr_to_ldr(float *data, int x, int y, int comp); +#endif + +#ifndef STBI_NO_STDIO +unsigned char *stbi_load(char *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = fopen(filename, "rb"); + unsigned char *result; + if (!f) return epuc("can't fopen", "Unable to open file"); + result = stbi_load_from_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} + +unsigned char *stbi_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + int i; + if (stbi_jpeg_test_file(f)) + return stbi_jpeg_load_from_file(f,x,y,comp,req_comp); + if (stbi_png_test_file(f)) + return stbi_png_load_from_file(f,x,y,comp,req_comp); + if (stbi_bmp_test_file(f)) + return stbi_bmp_load_from_file(f,x,y,comp,req_comp); + if (stbi_psd_test_file(f)) + return stbi_psd_load_from_file(f,x,y,comp,req_comp); + #ifndef STBI_NO_HDR + if (stbi_hdr_test_file(f)) { + float *hdr = stbi_hdr_load_from_file(f, x,y,comp,req_comp); + return hdr_to_ldr(hdr, *x, *y, req_comp ? req_comp : *comp); + } + #endif + for (i=0; i < max_loaders; ++i) + if (loaders[i]->test_file(f)) + return loaders[i]->load_from_file(f,x,y,comp,req_comp); + // test tga last because it's a crappy test! + if (stbi_tga_test_file(f)) + return stbi_tga_load_from_file(f,x,y,comp,req_comp); + return epuc("unknown image type", "Image not of any known type, or corrupt"); +} +#endif + +unsigned char *stbi_load_from_memory(stbi_uc *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + int i; + if (stbi_jpeg_test_memory(buffer,len)) + return stbi_jpeg_load_from_memory(buffer,len,x,y,comp,req_comp); + if (stbi_png_test_memory(buffer,len)) + return stbi_png_load_from_memory(buffer,len,x,y,comp,req_comp); + if (stbi_bmp_test_memory(buffer,len)) + return stbi_bmp_load_from_memory(buffer,len,x,y,comp,req_comp); + if (stbi_psd_test_memory(buffer,len)) + return stbi_psd_load_from_memory(buffer,len,x,y,comp,req_comp); + #ifndef STBI_NO_HDR + if (stbi_hdr_test_memory(buffer, len)) { + float *hdr = stbi_hdr_load_from_memory(buffer, len,x,y,comp,req_comp); + return hdr_to_ldr(hdr, *x, *y, req_comp ? req_comp : *comp); + } + #endif + for (i=0; i < max_loaders; ++i) + if (loaders[i]->test_memory(buffer,len)) + return loaders[i]->load_from_memory(buffer,len,x,y,comp,req_comp); + // test tga last because it's a crappy test! + if (stbi_tga_test_memory(buffer,len)) + return stbi_tga_load_from_memory(buffer,len,x,y,comp,req_comp); + return epuc("unknown image type", "Image not of any known type, or corrupt"); +} + +#ifndef STBI_NO_HDR + +#ifndef STBI_NO_STDIO +float *stbi_loadf(char *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = fopen(filename, "rb"); + float *result; + if (!f) return epf("can't fopen", "Unable to open file"); + result = stbi_loadf_from_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} + +float *stbi_loadf_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *data; + #ifndef STBI_NO_HDR + if (stbi_hdr_test_file(f)) + return stbi_hdr_load_from_file(f,x,y,comp,req_comp); + #endif + data = stbi_load_from_file(f, x, y, comp, req_comp); + if (data) + return ldr_to_hdr(data, *x, *y, req_comp ? req_comp : *comp); + return epf("unknown image type", "Image not of any known type, or corrupt"); +} +#endif + +float *stbi_loadf_from_memory(stbi_uc *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi_uc *data; + #ifndef STBI_NO_HDR + if (stbi_hdr_test_memory(buffer, len)) + return stbi_hdr_load_from_memory(buffer, len,x,y,comp,req_comp); + #endif + data = stbi_load_from_memory(buffer, len, x, y, comp, req_comp); + if (data) + return ldr_to_hdr(data, *x, *y, req_comp ? req_comp : *comp); + return epf("unknown image type", "Image not of any known type, or corrupt"); +} +#endif + +// these is-hdr-or-not is defined independent of whether STBI_NO_HDR is +// defined, for API simplicity; if STBI_NO_HDR is defined, it always +// reports false! + +extern int stbi_is_hdr_from_memory(stbi_uc *buffer, int len) +{ + #ifndef STBI_NO_HDR + return stbi_hdr_test_memory(buffer, len); + #else + return 0; + #endif +} + +#ifndef STBI_NO_STDIO +extern int stbi_is_hdr (char *filename) +{ + FILE *f = fopen(filename, "rb"); + int result=0; + if (f) { + result = stbi_is_hdr_from_file(f); + fclose(f); + } + return result; +} + +extern int stbi_is_hdr_from_file(FILE *f) +{ + #ifndef STBI_NO_HDR + return stbi_hdr_test_file(f); + #else + return 0; + #endif +} + +#endif + +// @TODO: get image dimensions & components without fully decoding +#ifndef STBI_NO_STDIO +extern int stbi_info (char *filename, int *x, int *y, int *comp); +extern int stbi_info_from_file (FILE *f, int *x, int *y, int *comp); +#endif +extern int stbi_info_from_memory(stbi_uc *buffer, int len, int *x, int *y, int *comp); + +#ifndef STBI_NO_HDR +static float h2l_gamma_i=1.0f/2.2f, h2l_scale_i=1.0f; +static float l2h_gamma=2.2f, l2h_scale=1.0f; + +void stbi_hdr_to_ldr_gamma(float gamma) { h2l_gamma_i = 1/gamma; } +void stbi_hdr_to_ldr_scale(float scale) { h2l_scale_i = 1/scale; } + +void stbi_ldr_to_hdr_gamma(float gamma) { l2h_gamma = gamma; } +void stbi_ldr_to_hdr_scale(float scale) { l2h_scale = scale; } +#endif + + +////////////////////////////////////////////////////////////////////////////// +// +// Common code used by all image loaders +// + +// image width, height, # components +static uint32 img_x, img_y; +static int img_n, img_out_n; + +enum +{ + SCAN_load=0, + SCAN_type, + SCAN_header, +}; + +// An API for reading either from memory or file. +#ifndef STBI_NO_STDIO +static FILE *img_file; +#endif +static uint8 *img_buffer, *img_buffer_end; + +#ifndef STBI_NO_STDIO +static void start_file(FILE *f) +{ + img_file = f; +} +#endif + +static void start_mem(uint8 *buffer, int len) +{ +#ifndef STBI_NO_STDIO + img_file = NULL; +#endif + img_buffer = buffer; + img_buffer_end = buffer+len; +} + +static int get8(void) +{ +#ifndef STBI_NO_STDIO + if (img_file) { + int c = fgetc(img_file); + return c == EOF ? 0 : c; + } +#endif + if (img_buffer < img_buffer_end) + return *img_buffer++; + return 0; +} + +static int at_eof(void) +{ +#ifndef STBI_NO_STDIO + if (img_file) + return feof(img_file); +#endif + return img_buffer >= img_buffer_end; +} + +static uint8 get8u(void) +{ + return (uint8) get8(); +} + +static void skip(int n) +{ +#ifndef STBI_NO_STDIO + if (img_file) + fseek(img_file, n, SEEK_CUR); + else +#endif + img_buffer += n; +} + +static int get16(void) +{ + int z = get8(); + return (z << 8) + get8(); +} + +static uint32 get32(void) +{ + uint32 z = get16(); + return (z << 16) + get16(); +} + +static int get16le(void) +{ + int z = get8(); + return z + (get8() << 8); +} + +static uint32 get32le(void) +{ + uint32 z = get16le(); + return z + (get16le() << 16); +} + +static void getn(stbi_uc *buffer, int n) +{ +#ifndef STBI_NO_STDIO + if (img_file) { + fread(buffer, 1, n, img_file); + return; + } +#endif + memcpy(buffer, img_buffer, n); + img_buffer += n; +} + +////////////////////////////////////////////////////////////////////////////// +// +// generic converter from built-in img_n to req_comp +// individual types do this automatically as much as possible (e.g. jpeg +// does all cases internally since it needs to colorspace convert anyway, +// and it never has alpha, so very few cases ). png can automatically +// interleave an alpha=255 channel, but falls back to this for other cases +// +// assume data buffer is malloced, so malloc a new one and free that one +// only failure mode is malloc failing + +static uint8 compute_y(int r, int g, int b) +{ + return (uint8) (((r*77) + (g*150) + (29*b)) >> 8); +} + +static unsigned char *convert_format(unsigned char *data, int img_n, int req_comp) +{ + uint i,j; + unsigned char *good; + + if (req_comp == img_n) return data; + assert(req_comp >= 1 && req_comp <= 4); + + good = (unsigned char *) malloc(req_comp * img_x * img_y); + if (good == NULL) { + free(data); + return epuc("outofmem", "Out of memory"); + } + + for (j=0; j < img_y; ++j) { + unsigned char *src = data + j * img_x * img_n ; + unsigned char *dest = good + j * img_x * req_comp; + + #define COMBO(a,b) ((a)*8+(b)) + #define CASE(a,b) case COMBO(a,b): for(i=0; i < img_x; ++i, src += a, dest += b) + + // convert source image with img_n components to one with req_comp components; + // avoid switch per pixel, so use switch per scanline and massive macros + switch(COMBO(img_n, req_comp)) { + CASE(1,2) dest[0]=src[0], dest[1]=255; break; + CASE(1,3) dest[0]=dest[1]=dest[2]=src[0]; break; + CASE(1,4) dest[0]=dest[1]=dest[2]=src[0], dest[3]=255; break; + CASE(2,1) dest[0]=src[0]; break; + CASE(2,3) dest[0]=dest[1]=dest[2]=src[0]; break; + CASE(2,4) dest[0]=dest[1]=dest[2]=src[0], dest[3]=src[1]; break; + CASE(3,4) dest[0]=src[0],dest[1]=src[1],dest[2]=src[2],dest[3]=255; break; + CASE(3,1) dest[0]=compute_y(src[0],src[1],src[2]); break; + CASE(3,2) dest[0]=compute_y(src[0],src[1],src[2]), dest[1] = 255; break; + CASE(4,1) dest[0]=compute_y(src[0],src[1],src[2]); break; + CASE(4,2) dest[0]=compute_y(src[0],src[1],src[2]), dest[1] = src[3]; break; + CASE(4,3) dest[0]=src[0],dest[1]=src[1],dest[2]=src[2]; break; + default: assert(0); + } + #undef CASE + } + + free(data); + img_out_n = req_comp; + return good; +} + +#ifndef STBI_NO_HDR +static float *ldr_to_hdr(stbi_uc *data, int x, int y, int comp) +{ + int i,k,n; + float *output = (float *) malloc(x * y * comp * sizeof(float)); + if (output == NULL) { free(data); return epf("outofmem", "Out of memory"); } + // compute number of non-alpha components + if (comp & 1) n = comp; else n = comp-1; + for (i=0; i < x*y; ++i) { + for (k=0; k < n; ++k) { + output[i*comp + k] = (float) pow(data[i*comp+k]/255.0f, l2h_gamma) * l2h_scale; + } + if (k < comp) output[i*comp + k] = data[i*comp+k]/255.0f; + } + free(data); + return output; +} + +#define float2int(x) ((int) (x)) +static stbi_uc *hdr_to_ldr(float *data, int x, int y, int comp) +{ + int i,k,n; + stbi_uc *output = (stbi_uc *) malloc(x * y * comp); + if (output == NULL) { free(data); return epuc("outofmem", "Out of memory"); } + // compute number of non-alpha components + if (comp & 1) n = comp; else n = comp-1; + for (i=0; i < x*y; ++i) { + for (k=0; k < n; ++k) { + float z = (float) pow(data[i*comp+k]*h2l_scale_i, h2l_gamma_i) * 255 + 0.5f; + if (z < 0) z = 0; + if (z > 255) z = 255; + output[i*comp + k] = float2int(z); + } + if (k < comp) { + float z = data[i*comp+k] * 255 + 0.5f; + if (z < 0) z = 0; + if (z > 255) z = 255; + output[i*comp + k] = float2int(z); + } + } + free(data); + return output; +} +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// "baseline" JPEG/JFIF decoder (not actually fully baseline implementation) +// +// simple implementation +// - channel subsampling of at most 2 in each dimension +// - doesn't support delayed output of y-dimension +// - simple interface (only one output format: 8-bit interleaved RGB) +// - doesn't try to recover corrupt jpegs +// - doesn't allow partial loading, loading multiple at once +// - still fast on x86 (copying globals into locals doesn't help x86) +// - allocates lots of intermediate memory (full size of all components) +// - non-interleaved case requires this anyway +// - allows good upsampling (see next) +// high-quality +// - upsampled channels are bilinearly interpolated, even across blocks +// - quality integer IDCT derived from IJG's 'slow' +// performance +// - fast huffman; reasonable integer IDCT +// - uses a lot of intermediate memory, could cache poorly +// - load http://nothings.org/remote/anemones.jpg 3 times on 2.8Ghz P4 +// stb_jpeg: 1.34 seconds (MSVC6, default release build) +// stb_jpeg: 1.06 seconds (MSVC6, processor = Pentium Pro) +// IJL11.dll: 1.08 seconds (compiled by intel) +// IJG 1998: 0.98 seconds (MSVC6, makefile provided by IJG) +// IJG 1998: 0.95 seconds (MSVC6, makefile + proc=PPro) + +int stbi_jpeg_dc_only; + +// huffman decoding acceleration +#define FAST_BITS 9 // larger handles more cases; smaller stomps less cache + +typedef struct +{ + uint8 fast[1 << FAST_BITS]; + // weirdly, repacking this into AoS is a 10% speed loss, instead of a win + uint16 code[256]; + uint8 values[256]; + uint8 size[257]; + unsigned int maxcode[18]; + int delta[17]; // old 'firstsymbol' - old 'firstcode' +} huffman; + +static huffman huff_dc[4]; // baseline is 2 tables, extended is 4 +static huffman huff_ac[4]; +static uint8 dequant[4][64]; + +static int build_huffman(huffman *h, int *count) +{ + int i,j,k=0,code; + // build size list for each symbol (from JPEG spec) + for (i=0; i < 16; ++i) + for (j=0; j < count[i]; ++j) + h->size[k++] = (uint8) (i+1); + h->size[k] = 0; + + // compute actual symbols (from jpeg spec) + code = 0; + k = 0; + for(j=1; j <= 16; ++j) { + // compute delta to add to code to compute symbol id + h->delta[j] = k - code; + if (h->size[k] == j) { + while (h->size[k] == j) + h->code[k++] = (uint16) (code++); + if (code-1 >= (1 << j)) return e("bad code lengths","Corrupt JPEG"); + } + // compute largest code + 1 for this size, preshifted as needed later + h->maxcode[j] = code << (16-j); + code <<= 1; + } + h->maxcode[j] = 0xffffffff; + + // build non-spec acceleration table; 255 is flag for not-accelerated + memset(h->fast, 255, 1 << FAST_BITS); + for (i=0; i < k; ++i) { + int s = h->size[i]; + if (s <= FAST_BITS) { + int c = h->code[i] << (FAST_BITS-s); + int m = 1 << (FAST_BITS-s); + for (j=0; j < m; ++j) { + h->fast[c+j] = (uint8) i; + } + } + } + return 1; +} + +// sizes for components, interleaved MCUs +static int img_h_max, img_v_max; +static int img_mcu_x, img_mcu_y; +static int img_mcu_w, img_mcu_h; + +// definition of jpeg image component +static struct +{ + int id; + int h,v; + int tq; + int hd,ha; + int dc_pred; + + int x,y,w2,h2; + uint8 *data; +} img_comp[4]; + +static unsigned long code_buffer; // jpeg entropy-coded buffer +static int code_bits; // number of valid bits +static unsigned char marker; // marker seen while filling entropy buffer +static int nomore; // flag if we saw a marker so must stop + +static void grow_buffer_unsafe(void) +{ + do { + int b = nomore ? 0 : get8(); + if (b == 0xff) { + int c = get8(); + if (c != 0) { + marker = (unsigned char) c; + nomore = 1; + return; + } + } + code_buffer = (code_buffer << 8) | b; + code_bits += 8; + } while (code_bits <= 24); +} + +// (1 << n) - 1 +static unsigned long bmask[17]={0,1,3,7,15,31,63,127,255,511,1023,2047,4095,8191,16383,32767,65535}; + +// decode a jpeg huffman value from the bitstream +__forceinline static int decode(huffman *h) +{ + unsigned int temp; + int c,k; + + if (code_bits < 16) grow_buffer_unsafe(); + + // look at the top FAST_BITS and determine what symbol ID it is, + // if the code is <= FAST_BITS + c = (code_buffer >> (code_bits - FAST_BITS)) & ((1 << FAST_BITS)-1); + k = h->fast[c]; + if (k < 255) { + if (h->size[k] > code_bits) + return -1; + code_bits -= h->size[k]; + return h->values[k]; + } + + // naive test is to shift the code_buffer down so k bits are + // valid, then test against maxcode. To speed this up, we've + // preshifted maxcode left so that it has (16-k) 0s at the + // end; in other words, regardless of the number of bits, it + // wants to be compared against something shifted to have 16; + // that way we don't need to shift inside the loop. + if (code_bits < 16) + temp = (code_buffer << (16 - code_bits)) & 0xffff; + else + temp = (code_buffer >> (code_bits - 16)) & 0xffff; + for (k=FAST_BITS+1 ; ; ++k) + if (temp < h->maxcode[k]) + break; + if (k == 17) { + // error! code not found + code_bits -= 16; + return -1; + } + + if (k > code_bits) + return -1; + + // convert the huffman code to the symbol id + c = ((code_buffer >> (code_bits - k)) & bmask[k]) + h->delta[k]; + assert((((code_buffer) >> (code_bits - h->size[c])) & bmask[h->size[c]]) == h->code[c]); + + // convert the id to a symbol + code_bits -= k; + return h->values[c]; +} + +// combined JPEG 'receive' and JPEG 'extend', since baseline +// always extends everything it receives. +__forceinline static int extend_receive(int n) +{ + unsigned int m = 1 << (n-1); + unsigned int k; + if (code_bits < n) grow_buffer_unsafe(); + k = (code_buffer >> (code_bits - n)) & bmask[n]; + code_bits -= n; + // the following test is probably a random branch that won't + // predict well. I tried to table accelerate it but failed. + // maybe it's compiling as a conditional move? + if (k < m) + return (-1 << n) + k + 1; + else + return k; +} + +// given a value that's at position X in the zigzag stream, +// where does it appear in the 8x8 matrix coded as row-major? +static uint8 dezigzag[64+15] = +{ + 0, 1, 8, 16, 9, 2, 3, 10, + 17, 24, 32, 25, 18, 11, 4, 5, + 12, 19, 26, 33, 40, 48, 41, 34, + 27, 20, 13, 6, 7, 14, 21, 28, + 35, 42, 49, 56, 57, 50, 43, 36, + 29, 22, 15, 23, 30, 37, 44, 51, + 58, 59, 52, 45, 38, 31, 39, 46, + 53, 60, 61, 54, 47, 55, 62, 63, + // let corrupt input sample past end + 63, 63, 63, 63, 63, 63, 63, 63, + 63, 63, 63, 63, 63, 63, 63 +}; + +// decode one 64-entry block-- +static int decode_block(short data[64], huffman *hdc, huffman *hac, int b) +{ + int diff,dc,k; + int t = decode(hdc); + if (t < 0) return e("bad huffman code","Corrupt JPEG"); + + // 0 all the ac values now so we can do it 32-bits at a time + memset(data,0,64*sizeof(data[0])); + + diff = t ? extend_receive(t) : 0; + dc = img_comp[b].dc_pred + diff; + img_comp[b].dc_pred = dc; + data[0] = (short) dc; + + // decode AC components, see JPEG spec + k = 1; + do { + int r,s; + int rs = decode(hac); + if (rs < 0) return e("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (rs != 0xf0) break; // end block + k += 16; + } else { + k += r; + // decode into unzigzag'd location + data[dezigzag[k++]] = (short) extend_receive(s); + } + } while (k < 64); + return 1; +} + +// take a -128..127 value and clamp it and convert to 0..255 +__forceinline static uint8 clamp(int x) +{ + x += 128; + // trick to use a single test to catch both cases + if ((unsigned int) x > 255) { + if (x < 0) return 0; + if (x > 255) return 255; + } + return (uint8) x; +} + +#define f2f(x) (int) (((x) * 4096 + 0.5)) +#define fsh(x) ((x) << 12) + +// derived from jidctint -- DCT_ISLOW +#define IDCT_1D(s0,s1,s2,s3,s4,s5,s6,s7) \ + int t0,t1,t2,t3,p1,p2,p3,p4,p5,x0,x1,x2,x3; \ + p2 = s2; \ + p3 = s6; \ + p1 = (p2+p3) * f2f(0.5411961f); \ + t2 = p1 + p3*f2f(-1.847759065f); \ + t3 = p1 + p2*f2f( 0.765366865f); \ + p2 = s0; \ + p3 = s4; \ + t0 = fsh(p2+p3); \ + t1 = fsh(p2-p3); \ + x0 = t0+t3; \ + x3 = t0-t3; \ + x1 = t1+t2; \ + x2 = t1-t2; \ + t0 = s7; \ + t1 = s5; \ + t2 = s3; \ + t3 = s1; \ + p3 = t0+t2; \ + p4 = t1+t3; \ + p1 = t0+t3; \ + p2 = t1+t2; \ + p5 = (p3+p4)*f2f( 1.175875602f); \ + t0 = t0*f2f( 0.298631336f); \ + t1 = t1*f2f( 2.053119869f); \ + t2 = t2*f2f( 3.072711026f); \ + t3 = t3*f2f( 1.501321110f); \ + p1 = p5 + p1*f2f(-0.899976223f); \ + p2 = p5 + p2*f2f(-2.562915447f); \ + p3 = p3*f2f(-1.961570560f); \ + p4 = p4*f2f(-0.390180644f); \ + t3 += p1+p4; \ + t2 += p2+p3; \ + t1 += p2+p4; \ + t0 += p1+p3; + +// .344 seconds on 3*anemones.jpg +static void idct_block(uint8 *out, int out_stride, short data[64], uint8 *dequantize) +{ + int i,val[64],*v=val; + uint8 *o,*dq = dequantize; + short *d = data; + + if (stbi_jpeg_dc_only) { + // ok, I don't really know why this is right, but it seems to be: + int z = 128 + ((d[0] * dq[0]) >> 3); + for (i=0; i < 8; ++i) { + out[0] = out[1] = out[2] = out[3] = out[4] = out[5] = out[6] = out[7] = z; + out += out_stride; + } + return; + } + + // columns + for (i=0; i < 8; ++i,++d,++dq, ++v) { + // if all zeroes, shortcut -- this avoids dequantizing 0s and IDCTing + if (d[ 8]==0 && d[16]==0 && d[24]==0 && d[32]==0 + && d[40]==0 && d[48]==0 && d[56]==0) { + // no shortcut 0 seconds + // (1|2|3|4|5|6|7)==0 0 seconds + // all separate -0.047 seconds + // 1 && 2|3 && 4|5 && 6|7: -0.047 seconds + int dcterm = d[0] * dq[0] << 2; + v[0] = v[8] = v[16] = v[24] = v[32] = v[40] = v[48] = v[56] = dcterm; + } else { + IDCT_1D(d[ 0]*dq[ 0],d[ 8]*dq[ 8],d[16]*dq[16],d[24]*dq[24], + d[32]*dq[32],d[40]*dq[40],d[48]*dq[48],d[56]*dq[56]) + // constants scaled things up by 1<<12; let's bring them back + // down, but keep 2 extra bits of precision + x0 += 512; x1 += 512; x2 += 512; x3 += 512; + v[ 0] = (x0+t3) >> 10; + v[56] = (x0-t3) >> 10; + v[ 8] = (x1+t2) >> 10; + v[48] = (x1-t2) >> 10; + v[16] = (x2+t1) >> 10; + v[40] = (x2-t1) >> 10; + v[24] = (x3+t0) >> 10; + v[32] = (x3-t0) >> 10; + } + } + + for (i=0, v=val, o=out; i < 8; ++i,v+=8,o+=out_stride) { + // no fast case since the first 1D IDCT spread components out + IDCT_1D(v[0],v[1],v[2],v[3],v[4],v[5],v[6],v[7]) + // constants scaled things up by 1<<12, plus we had 1<<2 from first + // loop, plus horizontal and vertical each scale by sqrt(8) so together + // we've got an extra 1<<3, so 1<<17 total we need to remove. + x0 += 65536; x1 += 65536; x2 += 65536; x3 += 65536; + o[0] = clamp((x0+t3) >> 17); + o[7] = clamp((x0-t3) >> 17); + o[1] = clamp((x1+t2) >> 17); + o[6] = clamp((x1-t2) >> 17); + o[2] = clamp((x2+t1) >> 17); + o[5] = clamp((x2-t1) >> 17); + o[3] = clamp((x3+t0) >> 17); + o[4] = clamp((x3-t0) >> 17); + } +} + +#define MARKER_none 0xff +// if there's a pending marker from the entropy stream, return that +// otherwise, fetch from the stream and get a marker. if there's no +// marker, return 0xff, which is never a valid marker value +static uint8 get_marker(void) +{ + uint8 x; + if (marker != MARKER_none) { x = marker; marker = MARKER_none; return x; } + x = get8u(); + if (x != 0xff) return MARKER_none; + while (x == 0xff) + x = get8u(); + return x; +} + +// in each scan, we'll have scan_n components, and the order +// of the components is specified by order[] +static int scan_n, order[4]; +static int restart_interval, todo; +#define RESTART(x) ((x) >= 0xd0 && (x) <= 0xd7) + +// after a restart interval, reset the entropy decoder and +// the dc prediction +static void reset(void) +{ + code_bits = 0; + code_buffer = 0; + nomore = 0; + img_comp[0].dc_pred = img_comp[1].dc_pred = img_comp[2].dc_pred = 0; + marker = MARKER_none; + todo = restart_interval ? restart_interval : 0x7fffffff; + // no more than 1<<31 MCUs if no restart_interal? that's plenty safe, + // since we don't even allow 1<<30 pixels +} + +static int parse_entropy_coded_data(void) +{ + reset(); + if (scan_n == 1) { + int i,j; + short data[64]; + int n = order[0]; + // non-interleaved data, we just need to process one block at a time, + // in trivial scanline order + // number of blocks to do just depends on how many actual "pixels" this + // component has, independent of interleaved MCU blocking and such + int w = (img_comp[n].x+7) >> 3; + int h = (img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + if (!decode_block(data, huff_dc+img_comp[n].hd, huff_ac+img_comp[n].ha, n)) return 0; + idct_block(img_comp[n].data+img_comp[n].w2*j*8+i*8, img_comp[n].w2, data, dequant[img_comp[n].tq]); + // every data block is an MCU, so countdown the restart interval + if (--todo <= 0) { + if (code_bits < 24) grow_buffer_unsafe(); + // if it's NOT a restart, then just bail, so we get corrupt data + // rather than no data + if (!RESTART(marker)) return 1; + reset(); + } + } + } + } else { // interleaved! + int i,j,k,x,y; + short data[64]; + for (j=0; j < img_mcu_y; ++j) { + for (i=0; i < img_mcu_x; ++i) { + // scan an interleaved mcu... process scan_n components in order + for (k=0; k < scan_n; ++k) { + int n = order[k]; + // scan out an mcu's worth of this component; that's just determined + // by the basic H and V specified for the component + for (y=0; y < img_comp[n].v; ++y) { + for (x=0; x < img_comp[n].h; ++x) { + int x2 = (i*img_comp[n].h + x)*8; + int y2 = (j*img_comp[n].v + y)*8; + if (!decode_block(data, huff_dc+img_comp[n].hd, huff_ac+img_comp[n].ha, n)) return 0; + idct_block(img_comp[n].data+img_comp[n].w2*y2+x2, img_comp[n].w2, data, dequant[img_comp[n].tq]); + } + } + } + // after all interleaved components, that's an interleaved MCU, + // so now count down the restart interval + if (--todo <= 0) { + if (code_bits < 24) grow_buffer_unsafe(); + // if it's NOT a restart, then just bail, so we get corrupt data + // rather than no data + if (!RESTART(marker)) return 1; + reset(); + } + } + } + } + return 1; +} + +static int process_marker(int m) +{ + int L; + switch (m) { + case MARKER_none: // no marker found + return e("expected marker","Corrupt JPEG"); + + case 0xC2: // SOF - progressive + return e("progressive jpeg","JPEG format not supported (progressive)"); + + case 0xDD: // DRI - specify restart interval + if (get16() != 4) return e("bad DRI len","Corrupt JPEG"); + restart_interval = get16(); + return 1; + + case 0xDB: // DQT - define quantization table + L = get16()-2; + while (L > 0) { + int z = get8(); + int p = z >> 4; + int t = z & 15,i; + if (p != 0) return e("bad DQT type","Corrupt JPEG"); + if (t > 3) return e("bad DQT table","Corrupt JPEG"); + for (i=0; i < 64; ++i) + dequant[t][dezigzag[i]] = get8u(); + L -= 65; + } + return L==0; + + case 0xC4: // DHT - define huffman table + L = get16()-2; + while (L > 0) { + uint8 *v; + int sizes[16],i,m=0; + int z = get8(); + int tc = z >> 4; + int th = z & 15; + if (tc > 1 || th > 3) return e("bad DHT header","Corrupt JPEG"); + for (i=0; i < 16; ++i) { + sizes[i] = get8(); + m += sizes[i]; + } + L -= 17; + if (tc == 0) { + if (!build_huffman(huff_dc+th, sizes)) return 0; + v = huff_dc[th].values; + } else { + if (!build_huffman(huff_ac+th, sizes)) return 0; + v = huff_ac[th].values; + } + for (i=0; i < m; ++i) + v[i] = get8u(); + L -= m; + } + return L==0; + } + // check for comment block or APP blocks + if ((m >= 0xE0 && m <= 0xEF) || m == 0xFE) { + skip(get16()-2); + return 1; + } + return 0; +} + +// after we see SOS +static int process_scan_header(void) +{ + int i; + int Ls = get16(); + scan_n = get8(); + if (scan_n < 1 || scan_n > 4 || scan_n > (int) img_n) return e("bad SOS component count","Corrupt JPEG"); + if (Ls != 6+2*scan_n) return e("bad SOS len","Corrupt JPEG"); + for (i=0; i < scan_n; ++i) { + int id = get8(), which; + int z = get8(); + for (which = 0; which < img_n; ++which) + if (img_comp[which].id == id) + break; + if (which == img_n) return 0; + img_comp[which].hd = z >> 4; if (img_comp[which].hd > 3) return e("bad DC huff","Corrupt JPEG"); + img_comp[which].ha = z & 15; if (img_comp[which].ha > 3) return e("bad AC huff","Corrupt JPEG"); + order[i] = which; + } + if (get8() != 0) return e("bad SOS","Corrupt JPEG"); + get8(); // should be 63, but might be 0 + if (get8() != 0) return e("bad SOS","Corrupt JPEG"); + + return 1; +} + +static int process_frame_header(int scan) +{ + int Lf,p,i,z, h_max=1,v_max=1; + Lf = get16(); if (Lf < 11) return e("bad SOF len","Corrupt JPEG"); // JPEG + p = get8(); if (p != 8) return e("only 8-bit","JPEG format not supported: 8-bit only"); // JPEG baseline + img_y = get16(); if (img_y == 0) return e("no header height", "JPEG format not supported: delayed height"); // Legal, but we don't handle it--but neither does IJG + img_x = get16(); if (img_x == 0) return e("0 width","Corrupt JPEG"); // JPEG requires + img_n = get8(); + if (img_n != 3 && img_n != 1) return e("bad component count","Corrupt JPEG"); // JFIF requires + + if (Lf != 8+3*img_n) return e("bad SOF len","Corrupt JPEG"); + + for (i=0; i < img_n; ++i) { + img_comp[i].id = get8(); + if (img_comp[i].id != i+1) // JFIF requires + if (img_comp[i].id != i) // jpegtran outputs non-JFIF-compliant files! + return e("bad component ID","Corrupt JPEG"); + z = get8(); + img_comp[i].h = (z >> 4); if (!img_comp[i].h || img_comp[i].h > 4) return e("bad H","Corrupt JPEG"); + img_comp[i].v = z & 15; if (!img_comp[i].v || img_comp[i].v > 4) return e("bad V","Corrupt JPEG"); + img_comp[i].tq = get8(); if (img_comp[i].tq > 3) return e("bad TQ","Corrupt JPEG"); + } + + if (scan != SCAN_load) return 1; + + if ((1 << 30) / img_x / img_n < img_y) return e("too large", "Image too large to decode"); + + for (i=0; i < img_n; ++i) { + if (img_comp[i].h > h_max) h_max = img_comp[i].h; + if (img_comp[i].v > v_max) v_max = img_comp[i].v; + } + + // compute interleaved mcu info + img_h_max = h_max; + img_v_max = v_max; + img_mcu_w = h_max * 8; + img_mcu_h = v_max * 8; + img_mcu_x = (img_x + img_mcu_w-1) / img_mcu_w; + img_mcu_y = (img_y + img_mcu_h-1) / img_mcu_h; + + for (i=0; i < img_n; ++i) { + // number of effective pixels (e.g. for non-interleaved MCU) + img_comp[i].x = (img_x * img_comp[i].h + h_max-1) / h_max; + img_comp[i].y = (img_y * img_comp[i].v + v_max-1) / v_max; + // to simplify generation, we'll allocate enough memory to decode + // the bogus oversized data from using interleaved MCUs and their + // big blocks (e.g. a 16x16 iMCU on an image of width 33); we won't + // discard the extra data until colorspace conversion + img_comp[i].w2 = img_mcu_x * img_comp[i].h * 8; + img_comp[i].h2 = img_mcu_y * img_comp[i].v * 8; + img_comp[i].data = (uint8 *) malloc(img_comp[i].w2 * img_comp[i].h2); + if (img_comp[i].data == NULL) { + for(--i; i >= 0; --i) + free(img_comp[i].data); + return e("outofmem", "Out of memory"); + } + } + + return 1; +} + +// use comparisons since in some cases we handle more than one case (e.g. SOF) +#define DNL(x) ((x) == 0xdc) +#define SOI(x) ((x) == 0xd8) +#define EOI(x) ((x) == 0xd9) +#define SOF(x) ((x) == 0xc0 || (x) == 0xc1) +#define SOS(x) ((x) == 0xda) + +static int decode_jpeg_header(int scan) +{ + int m; + marker = MARKER_none; // initialize cached marker to empty + m = get_marker(); + if (!SOI(m)) return e("no SOI","Corrupt JPEG"); + if (scan == SCAN_type) return 1; + m = get_marker(); + while (!SOF(m)) { + if (!process_marker(m)) return 0; + m = get_marker(); + while (m == MARKER_none) { + // some files have extra padding after their blocks, so ok, we'll scan + if (at_eof()) return e("no SOF", "Corrupt JPEG"); + m = get_marker(); + } + } + if (!process_frame_header(scan)) return 0; + return 1; +} + +static int decode_jpeg_image(void) +{ + int m; + restart_interval = 0; + if (!decode_jpeg_header(SCAN_load)) return 0; + m = get_marker(); + while (!EOI(m)) { + if (SOS(m)) { + if (!process_scan_header()) return 0; + if (!parse_entropy_coded_data()) return 0; + } else { + if (!process_marker(m)) return 0; + } + m = get_marker(); + } + return 1; +} + +// static jfif-centered resampling with cross-block smoothing +// here by cross-block smoothing what I mean is that the resampling +// is bilerp and crosses blocks; I dunno what IJG means + +#define div4(x) ((uint8) ((x) >> 2)) + +static void resample_v_2(uint8 *out1, uint8 *input, int w, int h, int s) +{ + // need to generate two samples vertically for every one in input + uint8 *above; + uint8 *below; + uint8 *source; + uint8 *out2; + int i,j; + source = input; + out2 = out1+w; + for (j=0; j < h; ++j) { + above = source; + source = input + j*s; + below = source + s; if (j == h-1) below = source; + for (i=0; i < w; ++i) { + int n = source[i]*3; + out1[i] = div4(above[i] + n); + out2[i] = div4(below[i] + n); + } + out1 += w*2; + out2 += w*2; + } +} + +static void resample_h_2(uint8 *out, uint8 *input, int w, int h, int s) +{ + // need to generate two samples horizontally for every one in input + int i,j; + if (w == 1) { + for (j=0; j < h; ++j) + out[j*2+0] = out[j*2+1] = input[j*s]; + return; + } + for (j=0; j < h; ++j) { + out[0] = input[0]; + out[1] = div4(input[0]*3 + input[1]); + for (i=1; i < w-1; ++i) { + int n = input[i]*3; + out[i*2-2] = div4(input[i-1] + n); + out[i*2-1] = div4(input[i+1] + n); + } + out[w*2-2] = div4(input[w-2]*3 + input[w-1]); + out[w*2-1] = input[w-1]; + out += w*2; + input += s; + } +} + +// .172 seconds on 3*anemones.jpg +static void resample_hv_2(uint8 *out, uint8 *input, int w, int h, int s) +{ + // need to generate 2x2 samples for every one in input + int i,j; + int os = w*2; + // generate edge samples... @TODO lerp them! + for (i=0; i < w; ++i) { + out[i*2+0] = out[i*2+1] = input[i]; + out[i*2+(2*h-1)*os+0] = out[i*2+(2*h-1)*os+1] = input[i+(h-1)*w]; + } + for (j=0; j < h; ++j) { + out[j*os*2+0] = out[j*os*2+os+0] = input[j*w]; + out[j*os*2+os-1] = out[j*os*2+os+os-1] = input[j*w+i-1]; + } + // now generate interior samples; i & j point to top left of input + for (j=0; j < h-1; ++j) { + uint8 *in1 = input+j*s; + uint8 *in2 = in1 + s; + uint8 *out1 = out + (j*2+1)*os + 1; + uint8 *out2 = out1 + os; + for (i=0; i < w-1; ++i) { + int p00 = in1[0], p01=in1[1], p10=in2[0], p11=in2[1]; + int p00_3 = p00*3, p01_3 = p01*3, p10_3 = p10*3, p11_3 = p11*3; + + #define div16(x) ((uint8) ((x) >> 4)) + + out1[0] = div16(p00*9 + p01_3 + p10_3 + p11); + out1[1] = div16(p01*9 + p00_3 + p01_3 + p10); + out2[0] = div16(p10*9 + p11_3 + p00_3 + p01); + out2[1] = div16(p11*9 + p10_3 + p01_3 + p00); + out1 += 2; + out2 += 2; + ++in1; + ++in2; + } + } +} + +#define float2fixed(x) ((int) ((x) * 65536 + 0.5)) + +// 0.38 seconds on 3*anemones.jpg (0.25 with processor = Pro) +// VC6 without processor=Pro is generating multiple LEAs per multiply! +static void YCbCr_to_RGB_row(uint8 *out, uint8 *y, uint8 *pcb, uint8 *pcr, int count, int step) +{ + int i; + for (i=0; i < count; ++i) { + int y_fixed = (y[i] << 16) + 32768; // rounding + int r,g,b; + int cr = pcr[i] - 128; + int cb = pcb[i] - 128; + r = y_fixed + cr*float2fixed(1.40200f); + g = y_fixed - cr*float2fixed(0.71414f) - cb*float2fixed(0.34414f); + b = y_fixed + cb*float2fixed(1.77200f); + r >>= 16; + g >>= 16; + b >>= 16; + if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; } + if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; } + if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; } + out[0] = (uint8)r; + out[1] = (uint8)g; + out[2] = (uint8)b; + if (step == 4) out[3] = 255; + out += step; + } +} + +// clean up the temporary component buffers +static void cleanup_jpeg(void) +{ + int i; + for (i=0; i < img_n; ++i) { + if (img_comp[i].data) { + free(img_comp[i].data); + img_comp[i].data = NULL; + } + } +} + +static uint8 *load_jpeg_image(int *out_x, int *out_y, int *comp, int req_comp) +{ + int i, n; + // validate req_comp + if (req_comp < 0 || req_comp > 4) return epuc("bad req_comp", "Internal error"); + + // load a jpeg image from whichever source + if (!decode_jpeg_image()) { cleanup_jpeg(); return NULL; } + + // determine actual number of components to generate + n = req_comp ? req_comp : img_n; + + // resample components to full size... memory wasteful, but this + // lets us bilerp across blocks while upsampling + for (i=0; i < img_n; ++i) { + // if we're outputting fewer than 3 components, we're grey not RGB; + // in that case, don't bother upsampling Cb or Cr + if (n < 3 && i) continue; + + // check if the component scale is less than max; if so it needs upsampling + if (img_comp[i].h != img_h_max || img_comp[i].v != img_v_max) { + int stride = img_x; + // allocate final size; make sure it's big enough for upsampling off + // the edges with upsample up to 4x4 (although we only support 2x2 + // currently) + uint8 *new_data = (uint8 *) malloc((img_x+3)*(img_y+3)); + if (new_data == NULL) { + cleanup_jpeg(); + return epuc("outofmem", "Out of memory (image too large?)"); + } + if (img_comp[i].h*2 == img_h_max && img_comp[i].v*2 == img_v_max) { + int tx = (img_x+1)>>1; + resample_hv_2(new_data, img_comp[i].data, tx,(img_y+1)>>1, img_comp[i].w2); + stride = tx*2; + } else if (img_comp[i].h == img_h_max && img_comp[i].v*2 == img_v_max) { + resample_v_2(new_data, img_comp[i].data, img_x,(img_y+1)>>1, img_comp[i].w2); + } else if (img_comp[i].h*2 == img_h_max && img_comp[i].v == img_v_max) { + int tx = (img_x+1)>>1; + resample_h_2(new_data, img_comp[i].data, tx,img_y, img_comp[i].w2); + stride = tx*2; + } else { + // @TODO resample uncommon sampling pattern with nearest neighbor + free(new_data); + cleanup_jpeg(); + return epuc("uncommon H or V", "JPEG not supported: atypical downsampling mode"); + } + img_comp[i].w2 = stride; + free(img_comp[i].data); + img_comp[i].data = new_data; + } + } + + // now convert components to output image + { + uint32 i,j; + uint8 *output = (uint8 *) malloc(n * img_x * img_y + 1); + if (n >= 3) { // output STBI_rgb_* + for (j=0; j < img_y; ++j) { + uint8 *y = img_comp[0].data + j*img_comp[0].w2; + uint8 *out = output + n * img_x * j; + if (img_n == 3) { + uint8 *cb = img_comp[1].data + j*img_comp[1].w2; + uint8 *cr = img_comp[2].data + j*img_comp[2].w2; + YCbCr_to_RGB_row(out, y, cb, cr, img_x, n); + } else { + for (i=0; i < img_x; ++i) { + out[0] = out[1] = out[2] = y[i]; + out[3] = 255; // not used if n == 3 + out += n; + } + } + } + } else { // output STBI_grey_* + for (j=0; j < img_y; ++j) { + uint8 *y = img_comp[0].data + j*img_comp[0].w2; + uint8 *out = output + n * img_x * j; + if (n == 1) + for (i=0; i < img_x; ++i) *out++ = *y++; + else + for (i=0; i < img_x; ++i) *out++ = *y++, *out++ = 255; + } + } + cleanup_jpeg(); + *out_x = img_x; + *out_y = img_y; + if (comp) *comp = img_n; // report original components, not output + return output; + } +} + +#ifndef STBI_NO_STDIO +unsigned char *stbi_jpeg_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + start_file(f); + return load_jpeg_image(x,y,comp,req_comp); +} + +unsigned char *stbi_jpeg_load(char *filename, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *data; + FILE *f = fopen(filename, "rb"); + if (!f) return NULL; + data = stbi_jpeg_load_from_file(f,x,y,comp,req_comp); + fclose(f); + return data; +} +#endif + +unsigned char *stbi_jpeg_load_from_memory(stbi_uc *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + start_mem(buffer,len); + return load_jpeg_image(x,y,comp,req_comp); +} + +#ifndef STBI_NO_STDIO +int stbi_jpeg_test_file(FILE *f) +{ + int n,r; + n = ftell(f); + start_file(f); + r = decode_jpeg_header(SCAN_type); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +int stbi_jpeg_test_memory(unsigned char *buffer, int len) +{ + start_mem(buffer,len); + return decode_jpeg_header(SCAN_type); +} + +// @TODO: +#ifndef STBI_NO_STDIO +extern int stbi_jpeg_info (char *filename, int *x, int *y, int *comp); +extern int stbi_jpeg_info_from_file (FILE *f, int *x, int *y, int *comp); +#endif +extern int stbi_jpeg_info_from_memory(stbi_uc *buffer, int len, int *x, int *y, int *comp); + +// public domain zlib decode v0.2 Sean Barrett 2006-11-18 +// simple implementation +// - all input must be provided in an upfront buffer +// - all output is written to a single output buffer (can malloc/realloc) +// performance +// - fast huffman + +// fast-way is faster to check than jpeg huffman, but slow way is slower +#define ZFAST_BITS 9 // accelerate all cases in default tables +#define ZFAST_MASK ((1 << ZFAST_BITS) - 1) + +// zlib-style huffman encoding +// (jpegs packs from left, zlib from right, so can't share code) +typedef struct +{ + uint16 fast[1 << ZFAST_BITS]; + uint16 firstcode[16]; + int maxcode[17]; + uint16 firstsymbol[16]; + uint8 size[288]; + uint16 value[288]; +} zhuffman; + +__forceinline static int bitreverse16(int n) +{ + n = ((n & 0xAAAA) >> 1) | ((n & 0x5555) << 1); + n = ((n & 0xCCCC) >> 2) | ((n & 0x3333) << 2); + n = ((n & 0xF0F0) >> 4) | ((n & 0x0F0F) << 4); + n = ((n & 0xFF00) >> 8) | ((n & 0x00FF) << 8); + return n; +} + +__forceinline static int bit_reverse(int v, int bits) +{ + assert(bits <= 16); + // to bit reverse n bits, reverse 16 and shift + // e.g. 11 bits, bit reverse and shift away 5 + return bitreverse16(v) >> (16-bits); +} + +static int zbuild_huffman(zhuffman *z, uint8 *sizelist, int num) +{ + int i,k=0; + int code, next_code[16], sizes[17]; + + // DEFLATE spec for generating codes + memset(sizes, 0, sizeof(sizes)); + memset(z->fast, 255, sizeof(z->fast)); + for (i=0; i < num; ++i) + ++sizes[sizelist[i]]; + sizes[0] = 0; + for (i=1; i < 16; ++i) + assert(sizes[i] <= (1 << i)); + code = 0; + for (i=1; i < 16; ++i) { + next_code[i] = code; + z->firstcode[i] = (uint16) code; + z->firstsymbol[i] = (uint16) k; + code = (code + sizes[i]); + if (sizes[i]) + if (code-1 >= (1 << i)) return e("bad codelengths","Corrupt JPEG"); + z->maxcode[i] = code << (16-i); // preshift for inner loop + code <<= 1; + k += sizes[i]; + } + z->maxcode[16] = 0x10000; // sentinel + for (i=0; i < num; ++i) { + int s = sizelist[i]; + if (s) { + int c = next_code[s] - z->firstcode[s] + z->firstsymbol[s]; + z->size[c] = (uint8)s; + z->value[c] = (uint16)i; + if (s <= ZFAST_BITS) { + int k = bit_reverse(next_code[s],s); + while (k < (1 << ZFAST_BITS)) { + z->fast[k] = (uint16) c; + k += (1 << s); + } + } + ++next_code[s]; + } + } + return 1; +} + +// zlib-from-memory implementation for PNG reading +// because PNG allows splitting the zlib stream arbitrarily, +// and it's annoying structurally to have PNG call ZLIB call PNG, +// we require PNG read all the IDATs and combine them into a single +// memory buffer + +static uint8 *zbuffer, *zbuffer_end; + +__forceinline static int zget8(void) +{ + if (zbuffer >= zbuffer_end) return 0; + return *zbuffer++; +} + +//static unsigned long code_buffer; +static int num_bits; + +static void fill_bits(void) +{ + do { + assert(code_buffer < (1U << num_bits)); + code_buffer |= zget8() << num_bits; + num_bits += 8; + } while (num_bits <= 24); +} + +__forceinline static unsigned int zreceive(int n) +{ + unsigned int k; + if (num_bits < n) fill_bits(); + k = code_buffer & ((1 << n) - 1); + code_buffer >>= n; + num_bits -= n; + return k; +} + +__forceinline static int zhuffman_decode(zhuffman *z) +{ + int b,s,k; + if (num_bits < 16) fill_bits(); + b = z->fast[code_buffer & ZFAST_MASK]; + if (b < 0xffff) { + s = z->size[b]; + code_buffer >>= s; + num_bits -= s; + return z->value[b]; + } + + // not resolved by fast table, so compute it the slow way + // use jpeg approach, which requires MSbits at top + k = bit_reverse(code_buffer, 16); + for (s=ZFAST_BITS+1; ; ++s) + if (k < z->maxcode[s]) + break; + if (s == 16) return -1; // invalid code! + // code size is s, so: + b = (k >> (16-s)) - z->firstcode[s] + z->firstsymbol[s]; + assert(z->size[b] == s); + code_buffer >>= s; + num_bits -= s; + return z->value[b]; +} + +static char *zout; +static char *zout_start; +static char *zout_end; +static int z_expandable; + +static int expand(int n) // need to make room for n bytes +{ + char *q; + int cur, limit; + if (!z_expandable) return e("output buffer limit","Corrupt PNG"); + cur = (int) (zout - zout_start); + limit = (int) (zout_end - zout_start); + while (cur + n > limit) + limit *= 2; + q = (char *) realloc(zout_start, limit); + if (q == NULL) return e("outofmem", "Out of memory"); + zout_start = q; + zout = q + cur; + zout_end = q + limit; + return 1; +} + +static zhuffman z_length, z_distance; + +static int length_base[31] = { + 3,4,5,6,7,8,9,10,11,13, + 15,17,19,23,27,31,35,43,51,59, + 67,83,99,115,131,163,195,227,258,0,0 }; + +static int length_extra[31]= +{ 0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0,0,0 }; + +static int dist_base[32] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193, +257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577,0,0}; + +static int dist_extra[32] = +{ 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13}; + +static int parse_huffman_block(void) +{ + for(;;) { + int z = zhuffman_decode(&z_length); + if (z < 256) { + if (z < 0) return e("bad huffman code","Corrupt PNG"); // error in huffman codes + if (zout >= zout_end) if (!expand(1)) return 0; + *zout++ = (char) z; + } else { + uint8 *p; + int len,dist; + if (z == 256) return 1; + z -= 257; + len = length_base[z]; + if (length_extra[z]) len += zreceive(length_extra[z]); + z = zhuffman_decode(&z_distance); + if (z < 0) return e("bad huffman code","Corrupt PNG"); + dist = dist_base[z]; + if (dist_extra[z]) dist += zreceive(dist_extra[z]); + if (zout - zout_start < dist) return e("bad dist","Corrupt PNG"); + if (zout + len > zout_end) if (!expand(len)) return 0; + p = (uint8 *) (zout - dist); + while (len--) + *zout++ = *p++; + } + } +} + +static int compute_huffman_codes(void) +{ + static uint8 length_dezigzag[19] = { 16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15 }; + static zhuffman z_codelength; // static just to save stack space + uint8 lencodes[286+32+137];//padding for maximum single op + uint8 codelength_sizes[19]; + int i,n; + + int hlit = zreceive(5) + 257; + int hdist = zreceive(5) + 1; + int hclen = zreceive(4) + 4; + + memset(codelength_sizes, 0, sizeof(codelength_sizes)); + for (i=0; i < hclen; ++i) { + int s = zreceive(3); + codelength_sizes[length_dezigzag[i]] = (uint8) s; + } + if (!zbuild_huffman(&z_codelength, codelength_sizes, 19)) return 0; + + n = 0; + while (n < hlit + hdist) { + int c = zhuffman_decode(&z_codelength); + assert(c >= 0 && c < 19); + if (c < 16) + lencodes[n++] = (uint8) c; + else if (c == 16) { + c = zreceive(2)+3; + memset(lencodes+n, lencodes[n-1], c); + n += c; + } else if (c == 17) { + c = zreceive(3)+3; + memset(lencodes+n, 0, c); + n += c; + } else { + assert(c == 18); + c = zreceive(7)+11; + memset(lencodes+n, 0, c); + n += c; + } + } + if (n != hlit+hdist) return e("bad codelengths","Corrupt PNG"); + if (!zbuild_huffman(&z_length, lencodes, hlit)) return 0; + if (!zbuild_huffman(&z_distance, lencodes+hlit, hdist)) return 0; + return 1; +} + +static int parse_uncompressed_block(void) +{ + uint8 header[4]; + int len,nlen,k; + if (num_bits & 7) + zreceive(num_bits & 7); // discard + // drain the bit-packed data into header + k = 0; + while (num_bits > 0) { + header[k++] = (uint8) (code_buffer & 255); // wtf this warns? + code_buffer >>= 8; + num_bits -= 8; + } + assert(num_bits == 0); + // now fill header the normal way + while (k < 4) + header[k++] = (uint8) zget8(); + len = header[1] * 256 + header[0]; + nlen = header[3] * 256 + header[2]; + if (nlen != (len ^ 0xffff)) return e("zlib corrupt","Corrupt PNG"); + if (zbuffer + len > zbuffer_end) return e("read past buffer","Corrupt PNG"); + if (zout + len > zout_end) + if (!expand(len)) return 0; + memcpy(zout, zbuffer, len); + zbuffer += len; + zout += len; + return 1; +} + +static int parse_zlib_header(void) +{ + int cmf = zget8(); + int cm = cmf & 15; + /* int cinfo = cmf >> 4; */ + int flg = zget8(); + if ((cmf*256+flg) % 31 != 0) return e("bad zlib header","Corrupt PNG"); // zlib spec + if (flg & 32) return e("no preset dict","Corrupt PNG"); // preset dictionary not allowed in png + if (cm != 8) return e("bad compression","Corrupt PNG"); // DEFLATE required for png + // window = 1 << (8 + cinfo)... but who cares, we fully buffer output + return 1; +} + +static uint8 default_length[288], default_distance[32]; +static void init_defaults(void) +{ + int i; // use <= to match clearly with spec + for (i=0; i <= 143; ++i) default_length[i] = 8; + for ( ; i <= 255; ++i) default_length[i] = 9; + for ( ; i <= 279; ++i) default_length[i] = 7; + for ( ; i <= 287; ++i) default_length[i] = 8; + + for (i=0; i <= 31; ++i) default_distance[i] = 5; +} + +static int parse_zlib(int parse_header) +{ + int final, type; + if (parse_header) + if (!parse_zlib_header()) return 0; + num_bits = 0; + code_buffer = 0; + do { + final = zreceive(1); + type = zreceive(2); + if (type == 0) { + if (!parse_uncompressed_block()) return 0; + } else if (type == 3) { + return 0; + } else { + if (type == 1) { + // use fixed code lengths + if (!default_length[0]) init_defaults(); + if (!zbuild_huffman(&z_length , default_length , 288)) return 0; + if (!zbuild_huffman(&z_distance, default_distance, 32)) return 0; + } else { + if (!compute_huffman_codes()) return 0; + } + if (!parse_huffman_block()) return 0; + } + } while (!final); + return 1; +} + +static int do_zlib(char *obuf, int olen, int exp, int parse_header) +{ + zout_start = obuf; + zout = obuf; + zout_end = obuf + olen; + z_expandable = exp; + + return parse_zlib(parse_header); +} + +char *stbi_zlib_decode_malloc_guesssize(int initial_size, int *outlen) +{ + char *p = (char *) malloc(initial_size); + if (p == NULL) return NULL; + if (do_zlib(p, initial_size, 1, 1)) { + *outlen = (int) (zout - zout_start); + return zout_start; + } else { + free(zout_start); + return NULL; + } +} + +char *stbi_zlib_decode_malloc(char *buffer, int len, int *outlen) +{ + zbuffer = (uint8 *) buffer; + zbuffer_end = (uint8 *) buffer+len; + return stbi_zlib_decode_malloc_guesssize(16384, outlen); +} + +int stbi_zlib_decode_buffer(char *obuffer, int olen, char *ibuffer, int ilen) +{ + zbuffer = (uint8 *) ibuffer; + zbuffer_end = (uint8 *) ibuffer + ilen; + if (do_zlib(obuffer, olen, 0, 1)) + return (int) (zout - zout_start); + else + return -1; +} + +char *stbi_zlib_decode_noheader_malloc(char *buffer, int len, int *outlen) +{ + char *p = (char *) malloc(16384); + if (p == NULL) return NULL; + zbuffer = (uint8 *) buffer; + zbuffer_end = (uint8 *) buffer+len; + if (do_zlib(p, 16384, 1, 0)) { + *outlen = (int) (zout - zout_start); + return zout_start; + } else { + free(zout_start); + return NULL; + } +} + +int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, char *ibuffer, int ilen) +{ + zbuffer = (uint8 *) ibuffer; + zbuffer_end = (uint8 *) ibuffer + ilen; + if (do_zlib(obuffer, olen, 0, 0)) + return (int) (zout - zout_start); + else + return -1; +} + +// public domain "baseline" PNG decoder v0.10 Sean Barrett 2006-11-18 +// simple implementation +// - only 8-bit samples +// - no CRC checking +// - allocates lots of intermediate memory +// - avoids problem of streaming data between subsystems +// - avoids explicit window management +// performance +// - uses stb_zlib, a PD zlib implementation with fast huffman decoding + + +typedef struct +{ + unsigned long length; + unsigned long type; +} chunk; + +#define PNG_TYPE(a,b,c,d) (((a) << 24) + ((b) << 16) + ((c) << 8) + (d)) + +static chunk get_chunk_header(void) +{ + chunk c; + c.length = get32(); + c.type = get32(); + return c; +} + +static int check_png_header(void) +{ + static uint8 png_sig[8] = { 137,80,78,71,13,10,26,10 }; + int i; + for (i=0; i < 8; ++i) + if (get8() != png_sig[i]) return e("bad png sig","Not a PNG"); + return 1; +} + +static uint8 *idata, *expanded, *out; + +enum { + F_none=0, F_sub=1, F_up=2, F_avg=3, F_paeth=4, + F_avg_first, F_paeth_first, +}; + +static uint8 first_row_filter[5] = +{ + F_none, F_sub, F_none, F_avg_first, F_paeth_first +}; + +static int paeth(int a, int b, int c) +{ + int p = a + b - c; + int pa = abs(p-a); + int pb = abs(p-b); + int pc = abs(p-c); + if (pa <= pb && pa <= pc) return a; + if (pb <= pc) return b; + return c; +} + +// create the png data from post-deflated data +static int create_png_image(uint8 *raw, uint32 raw_len, int out_n) +{ + uint32 i,j,stride = img_x*out_n; + int k; + assert(out_n == img_n || out_n == img_n+1); + out = (uint8 *) malloc(img_x * img_y * out_n); + if (!out) return e("outofmem", "Out of memory"); + if (raw_len != (img_n * img_x + 1) * img_y) return e("not enough pixels","Corrupt PNG"); + for (j=0; j < img_y; ++j) { + uint8 *cur = out + stride*j; + uint8 *prior = cur - stride; + int filter = *raw++; + if (filter > 4) return e("invalid filter","Corrupt PNG"); + // if first row, use special filter that doesn't sample previous row + if (j == 0) filter = first_row_filter[filter]; + // handle first pixel explicitly + for (k=0; k < img_n; ++k) { + switch(filter) { + case F_none : cur[k] = raw[k]; break; + case F_sub : cur[k] = raw[k]; break; + case F_up : cur[k] = raw[k] + prior[k]; break; + case F_avg : cur[k] = raw[k] + (prior[k]>>1); break; + case F_paeth : cur[k] = (uint8) (raw[k] + paeth(0,prior[k],0)); break; + case F_avg_first : cur[k] = raw[k]; break; + case F_paeth_first: cur[k] = raw[k]; break; + } + } + if (img_n != out_n) cur[img_n] = 255; + raw += img_n; + cur += out_n; + prior += out_n; + // this is a little gross, so that we don't switch per-pixel or per-component + if (img_n == out_n) { + #define CASE(f) \ + case f: \ + for (i=1; i < img_x; ++i, raw+=img_n,cur+=img_n,prior+=img_n) \ + for (k=0; k < img_n; ++k) + switch(filter) { + CASE(F_none) cur[k] = raw[k]; break; + CASE(F_sub) cur[k] = raw[k] + cur[k-img_n]; break; + CASE(F_up) cur[k] = raw[k] + prior[k]; break; + CASE(F_avg) cur[k] = raw[k] + ((prior[k] + cur[k-img_n])>>1); break; + CASE(F_paeth) cur[k] = (uint8) (raw[k] + paeth(cur[k-img_n],prior[k],prior[k-img_n])); break; + CASE(F_avg_first) cur[k] = raw[k] + (cur[k-img_n] >> 1); break; + CASE(F_paeth_first) cur[k] = (uint8) (raw[k] + paeth(cur[k-img_n],0,0)); break; + } + #undef CASE + } else { + assert(img_n+1 == out_n); + #define CASE(f) \ + case f: \ + for (i=1; i < img_x; ++i, cur[img_n]=255,raw+=img_n,cur+=out_n,prior+=out_n) \ + for (k=0; k < img_n; ++k) + switch(filter) { + CASE(F_none) cur[k] = raw[k]; break; + CASE(F_sub) cur[k] = raw[k] + cur[k-out_n]; break; + CASE(F_up) cur[k] = raw[k] + prior[k]; break; + CASE(F_avg) cur[k] = raw[k] + ((prior[k] + cur[k-out_n])>>1); break; + CASE(F_paeth) cur[k] = (uint8) (raw[k] + paeth(cur[k-out_n],prior[k],prior[k-out_n])); break; + CASE(F_avg_first) cur[k] = raw[k] + (cur[k-out_n] >> 1); break; + CASE(F_paeth_first) cur[k] = (uint8) (raw[k] + paeth(cur[k-out_n],0,0)); break; + } + #undef CASE + } + } + return 1; +} + +static int compute_transparency(uint8 tc[3], int out_n) +{ + uint32 i, pixel_count = img_x * img_y; + uint8 *p = out; + + // compute color-based transparency, assuming we've + // already got 255 as the alpha value in the output + assert(out_n == 2 || out_n == 4); + + p = out; + if (out_n == 2) { + for (i=0; i < pixel_count; ++i) { + p[1] = (p[0] == tc[0] ? 0 : 255); + p += 2; + } + } else { + for (i=0; i < pixel_count; ++i) { + if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2]) + p[3] = 0; + p += 4; + } + } + return 1; +} + +static int expand_palette(uint8 *palette, int len, int pal_img_n) +{ + uint32 i, pixel_count = img_x * img_y; + uint8 *p, *temp_out, *orig = out; + + p = (uint8 *) malloc(pixel_count * pal_img_n); + if (p == NULL) return e("outofmem", "Out of memory"); + + // between here and free(out) below, exitting would leak + temp_out = p; + + if (pal_img_n == 3) { + for (i=0; i < pixel_count; ++i) { + int n = orig[i]*4; + p[0] = palette[n ]; + p[1] = palette[n+1]; + p[2] = palette[n+2]; + p += 3; + } + } else { + for (i=0; i < pixel_count; ++i) { + int n = orig[i]*4; + p[0] = palette[n ]; + p[1] = palette[n+1]; + p[2] = palette[n+2]; + p[3] = palette[n+3]; + p += 4; + } + } + free(out); + out = temp_out; + return 1; +} + +static int parse_png_file(int scan, int req_comp) +{ + uint8 palette[1024], pal_img_n=0; + uint8 has_trans=0, tc[3]; + uint32 ioff=0, idata_limit=0, i, pal_len=0; + int first=1,k; + + if (!check_png_header()) return 0; + + if (scan == SCAN_type) return 1; + + for(;;first=0) { + chunk c = get_chunk_header(); + if (first && c.type != PNG_TYPE('I','H','D','R')) + return e("first not IHDR","Corrupt PNG"); + switch (c.type) { + case PNG_TYPE('I','H','D','R'): { + int depth,color,interlace,comp,filter; + if (!first) return e("multiple IHDR","Corrupt PNG"); + if (c.length != 13) return e("bad IHDR len","Corrupt PNG"); + img_x = get32(); if (img_x > (1 << 24)) return e("too large","Very large image (corrupt?)"); + img_y = get32(); if (img_y > (1 << 24)) return e("too large","Very large image (corrupt?)"); + depth = get8(); if (depth != 8) return e("8bit only","PNG not supported: 8-bit only"); + color = get8(); if (color > 6) return e("bad ctype","Corrupt PNG"); + if (color == 3) pal_img_n = 3; else if (color & 1) return e("bad ctype","Corrupt PNG"); + comp = get8(); if (comp) return e("bad comp method","Corrupt PNG"); + filter= get8(); if (filter) return e("bad filter method","Corrupt PNG"); + interlace = get8(); if (interlace) return e("interlaced","PNG not supported: interlaced mode"); + if (!img_x || !img_y) return e("0-pixel image","Corrupt PNG"); + if (!pal_img_n) { + img_n = (color & 2 ? 3 : 1) + (color & 4 ? 1 : 0); + if ((1 << 30) / img_x / img_n < img_y) return e("too large", "Image too large to decode"); + if (scan == SCAN_header) return 1; + } else { + // if paletted, then pal_n is our final components, and + // img_n is # components to decompress/filter. + img_n = 1; + if ((1 << 30) / img_x / 4 < img_y) return e("too large","Corrupt PNG"); + // if SCAN_header, have to scan to see if we have a tRNS + } + break; + } + + case PNG_TYPE('P','L','T','E'): { + if (c.length > 256*3) return e("invalid PLTE","Corrupt PNG"); + pal_len = c.length / 3; + if (pal_len * 3 != c.length) return e("invalid PLTE","Corrupt PNG"); + for (i=0; i < pal_len; ++i) { + palette[i*4+0] = get8u(); + palette[i*4+1] = get8u(); + palette[i*4+2] = get8u(); + palette[i*4+3] = 255; + } + break; + } + + case PNG_TYPE('t','R','N','S'): { + if (idata) return e("tRNS after IDAT","Corrupt PNG"); + if (pal_img_n) { + if (scan == SCAN_header) { img_n = 4; return 1; } + if (pal_len == 0) return e("tRNS before PLTE","Corrupt PNG"); + if (c.length > pal_len) return e("bad tRNS len","Corrupt PNG"); + pal_img_n = 4; + for (i=0; i < c.length; ++i) + palette[i*4+3] = get8u(); + } else { + if (!(img_n & 1)) return e("tRNS with alpha","Corrupt PNG"); + if (c.length != (uint32) img_n*2) return e("bad tRNS len","Corrupt PNG"); + has_trans = 1; + for (k=0; k < img_n; ++k) + tc[k] = (uint8) get16(); // non 8-bit images will be larger + } + break; + } + + case PNG_TYPE('I','D','A','T'): { + if (pal_img_n && !pal_len) return e("no PLTE","Corrupt PNG"); + if (scan == SCAN_header) { img_n = pal_img_n; return 1; } + if (ioff + c.length > idata_limit) { + uint8 *p; + if (idata_limit == 0) idata_limit = c.length > 4096 ? c.length : 4096; + while (ioff + c.length > idata_limit) + idata_limit *= 2; + p = (uint8 *) realloc(idata, idata_limit); if (p == NULL) return e("outofmem", "Out of memory"); + idata = p; + } + #ifndef STBI_NO_STDIO + if (img_file) + { + if (fread(idata+ioff,1,c.length,img_file) != c.length) return e("outofdata","Corrupt PNG"); + } + else + #endif + { + memcpy(idata+ioff, img_buffer, c.length); + img_buffer += c.length; + } + ioff += c.length; + break; + } + + case PNG_TYPE('I','E','N','D'): { + uint32 raw_len; + if (scan != SCAN_load) return 1; + if (idata == NULL) return e("no IDAT","Corrupt PNG"); + expanded = (uint8 *) stbi_zlib_decode_malloc((char *) idata, ioff, (int *) &raw_len); + if (expanded == NULL) return 0; // zlib should set error + free(idata); idata = NULL; + if ((req_comp == img_n+1 && req_comp != 3 && !pal_img_n) || has_trans) + img_out_n = img_n+1; + else + img_out_n = img_n; + if (!create_png_image(expanded, raw_len, img_out_n)) return 0; + if (has_trans) + if (!compute_transparency(tc, img_out_n)) return 0; + if (pal_img_n) { + // pal_img_n == 3 or 4 + img_n = pal_img_n; // record the actual colors we had + img_out_n = pal_img_n; + if (req_comp >= 3) img_out_n = req_comp; + if (!expand_palette(palette, pal_len, img_out_n)) + return 0; + } + free(expanded); expanded = NULL; + return 1; + } + + default: + // if critical, fail + if ((c.type & (1 << 29)) == 0) { + #ifndef STBI_NO_FAILURE_STRINGS + static char invalid_chunk[] = "XXXX chunk not known"; + invalid_chunk[0] = (uint8) (c.type >> 24); + invalid_chunk[1] = (uint8) (c.type >> 16); + invalid_chunk[2] = (uint8) (c.type >> 8); + invalid_chunk[3] = (uint8) (c.type >> 0); + #endif + return e(invalid_chunk, "PNG not supported: unknown chunk type"); + } + skip(c.length); + break; + } + // end of chunk, read and skip CRC + get8(); get8(); get8(); get8(); + } +} + +static unsigned char *do_png(int *x, int *y, int *n, int req_comp) +{ + unsigned char *result=NULL; + if (req_comp < 0 || req_comp > 4) return epuc("bad req_comp", "Internal error"); + if (parse_png_file(SCAN_load, req_comp)) { + result = out; + out = NULL; + if (req_comp && req_comp != img_out_n) { + result = convert_format(result, img_out_n, req_comp); + if (result == NULL) return result; + } + *x = img_x; + *y = img_y; + if (n) *n = img_n; + } + free(out); out = NULL; + free(expanded); expanded = NULL; + free(idata); idata = NULL; + + return result; +} + +#ifndef STBI_NO_STDIO +unsigned char *stbi_png_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + start_file(f); + return do_png(x,y,comp,req_comp); +} + +unsigned char *stbi_png_load(char *filename, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *data; + FILE *f = fopen(filename, "rb"); + if (!f) return NULL; + data = stbi_png_load_from_file(f,x,y,comp,req_comp); + fclose(f); + return data; +} +#endif + +unsigned char *stbi_png_load_from_memory(unsigned char *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + start_mem(buffer,len); + return do_png(x,y,comp,req_comp); +} + +#ifndef STBI_NO_STDIO +int stbi_png_test_file(FILE *f) +{ + int n,r; + n = ftell(f); + start_file(f); + r = parse_png_file(SCAN_type,STBI_default); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +int stbi_png_test_memory(unsigned char *buffer, int len) +{ + start_mem(buffer, len); + return parse_png_file(SCAN_type,STBI_default); +} + +// TODO: load header from png +#ifndef STBI_NO_STDIO +extern int stbi_png_info (char *filename, int *x, int *y, int *comp); +extern int stbi_png_info_from_file (FILE *f, int *x, int *y, int *comp); +#endif +extern int stbi_png_info_from_memory (stbi_uc *buffer, int len, int *x, int *y, int *comp); + +// Microsoft/Windows BMP image + +static int bmp_test(void) +{ + int sz; + if (get8() != 'B') return 0; + if (get8() != 'M') return 0; + get32le(); // discard filesize + get16le(); // discard reserved + get16le(); // discard reserved + get32le(); // discard data offset + sz = get32le(); + if (sz == 12 || sz == 40 || sz == 56 || sz == 108) return 1; + return 0; +} + +#ifndef STBI_NO_STDIO +int stbi_bmp_test_file (FILE *f) +{ + int r,n = ftell(f); + start_file(f); + r = bmp_test(); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +int stbi_bmp_test_memory (stbi_uc *buffer, int len) +{ + start_mem(buffer, len); + return bmp_test(); +} + +// returns 0..31 for the highest set bit +static int high_bit(unsigned int z) +{ + int n=0; + if (z == 0) return -1; + if (z >= 0x10000) n += 16, z >>= 16; + if (z >= 0x00100) n += 8, z >>= 8; + if (z >= 0x00010) n += 4, z >>= 4; + if (z >= 0x00004) n += 2, z >>= 2; + if (z >= 0x00002) n += 1, z >>= 1; + return n; +} + +static int bitcount(unsigned int a) +{ + a = (a & 0x55555555) + ((a >> 1) & 0x55555555); // max 2 + a = (a & 0x33333333) + ((a >> 2) & 0x33333333); // max 4 + a = (a + (a >> 4)) & 0x0f0f0f0f; // max 8 per 4, now 8 bits + a = (a + (a >> 8)); // max 16 per 8 bits + a = (a + (a >> 16)); // max 32 per 8 bits + return a & 0xff; +} + +static int shiftsigned(int v, int shift, int bits) +{ + int result; + int z=0; + + if (shift < 0) v <<= -shift; + else v >>= shift; + result = v; + + z = bits; + while (z < 8) { + result += v >> z; + z += bits; + } + return result; +} + +static stbi_uc *bmp_load(int *x, int *y, int *comp, int req_comp) +{ + unsigned int mr=0,mg=0,mb=0,ma=0; + stbi_uc pal[256][4]; + int psize=0,i,j,compress=0,width; + int bpp, flip_vertically, pad, target, offset, hsz; + if (get8() != 'B' || get8() != 'M') return epuc("not BMP", "Corrupt BMP"); + get32le(); // discard filesize + get16le(); // discard reserved + get16le(); // discard reserved + offset = get32le(); + hsz = get32le(); + if (hsz != 12 && hsz != 40 && hsz != 56 && hsz != 108) return epuc("unknown BMP", "BMP type not supported: unknown"); + failure_reason = "bad BMP"; + if (hsz == 12) { + img_x = get16le(); + img_y = get16le(); + } else { + img_x = get32le(); + img_y = get32le(); + } + if (get16le() != 1) return 0; + bpp = get16le(); + if (bpp == 1) return epuc("monochrome", "BMP type not supported: 1-bit"); + flip_vertically = ((int) img_y) > 0; + img_y = abs((int) img_y); + if (hsz == 12) { + if (bpp < 24) + psize = (offset - 14 - 24) / 3; + } else { + compress = get32le(); + if (compress == 1 || compress == 2) return epuc("BMP RLE", "BMP type not supported: RLE"); + get32le(); // discard sizeof + get32le(); // discard hres + get32le(); // discard vres + get32le(); // discard colorsused + get32le(); // discard max important + if (hsz == 40 || hsz == 56) { + if (hsz == 56) { + get32le(); + get32le(); + get32le(); + get32le(); + } + if (bpp == 16 || bpp == 32) { + mr = mg = mb = 0; + if (compress == 0) { + if (bpp == 32) { + mr = 0xff << 16; + mg = 0xff << 8; + mb = 0xff << 0; + } else { + mr = 31 << 10; + mg = 31 << 5; + mb = 31 << 0; + } + } else if (compress == 3) { + mr = get32le(); + mg = get32le(); + mb = get32le(); + // not documented, but generated by photoshop and handled by mspaint + if (mr == mg && mg == mb) { + // ?!?!? + return NULL; + } + } else + return NULL; + } + } else { + assert(hsz == 108); + mr = get32le(); + mg = get32le(); + mb = get32le(); + ma = get32le(); + get32le(); // discard color space + for (i=0; i < 12; ++i) + get32le(); // discard color space parameters + } + if (bpp < 16) + psize = (offset - 14 - hsz) >> 2; + } + img_n = ma ? 4 : 3; + if (req_comp && req_comp >= 3) // we can directly decode 3 or 4 + target = req_comp; + else + target = img_n; // if they want monochrome, we'll post-convert + out = (stbi_uc *) malloc(target * img_x * img_y); + if (!out) return epuc("outofmem", "Out of memory"); + if (bpp < 16) { + int z=0; + if (psize == 0 || psize > 256) return epuc("invalid", "Corrupt BMP"); + for (i=0; i < psize; ++i) { + pal[i][2] = get8(); + pal[i][1] = get8(); + pal[i][0] = get8(); + if (hsz != 12) get8(); + pal[i][3] = 255; + } + skip(offset - 14 - hsz - psize * (hsz == 12 ? 3 : 4)); + if (bpp == 4) width = (img_x + 1) >> 1; + else if (bpp == 8) width = img_x; + else return epuc("bad bpp", "Corrupt BMP"); + pad = (-width)&3; + for (j=0; j < (int) img_y; ++j) { + for (i=0; i < (int) img_x; i += 2) { + int v=get8(),v2=0; + if (bpp == 4) { + v2 = v & 15; + v >>= 4; + } + out[z++] = pal[v][0]; + out[z++] = pal[v][1]; + out[z++] = pal[v][2]; + if (target == 4) out[z++] = 255; + if (i+1 == (int) img_x) break; + v = (bpp == 8) ? get8() : v2; + out[z++] = pal[v][0]; + out[z++] = pal[v][1]; + out[z++] = pal[v][2]; + if (target == 4) out[z++] = 255; + } + skip(pad); + } + } else { + int rshift=0,gshift=0,bshift=0,ashift=0,rcount=0,gcount=0,bcount=0,acount=0; + int z = 0; + int easy=0; + skip(offset - 14 - hsz); + if (bpp == 24) width = 3 * img_x; + else if (bpp == 16) width = 2*img_x; + else /* bpp = 32 and pad = 0 */ width=0; + pad = (-width) & 3; + if (bpp == 24) { + easy = 1; + } else if (bpp == 32) { + if (mb == 0xff && mg == 0xff00 && mr == 0xff000000 && ma == 0xff000000) + easy = 2; + } + if (!easy) { + if (!mr || !mg || !mb) return epuc("bad masks", "Corrupt BMP"); + // right shift amt to put high bit in position #7 + rshift = high_bit(mr)-7; rcount = bitcount(mr); + gshift = high_bit(mg)-7; gcount = bitcount(mr); + bshift = high_bit(mb)-7; bcount = bitcount(mr); + ashift = high_bit(ma)-7; acount = bitcount(mr); + } + for (j=0; j < (int) img_y; ++j) { + if (easy) { + for (i=0; i < (int) img_x; ++i) { + int a; + out[z+2] = get8(); + out[z+1] = get8(); + out[z+0] = get8(); + z += 3; + a = (easy == 2 ? get8() : 255); + if (target == 4) out[z++] = a; + } + } else { + for (i=0; i < (int) img_x; ++i) { + unsigned long v = (bpp == 16 ? get16le() : get32le()); + int a; + out[z++] = shiftsigned(v & mr, rshift, rcount); + out[z++] = shiftsigned(v & mg, gshift, gcount); + out[z++] = shiftsigned(v & mb, bshift, bcount); + a = (ma ? shiftsigned(v & ma, ashift, acount) : 255); + if (target == 4) out[z++] = a; + } + } + skip(pad); + } + } + if (flip_vertically) { + stbi_uc t; + for (j=0; j < (int) img_y>>1; ++j) { + stbi_uc *p1 = out + j *img_x*target; + stbi_uc *p2 = out + (img_y-1-j)*img_x*target; + for (i=0; i < (int) img_x*target; ++i) { + t = p1[i], p1[i] = p2[i], p2[i] = t; + } + } + } + + if (req_comp && req_comp != target) { + out = convert_format(out, target, req_comp); + if (out == NULL) return out; // convert_format frees input on failure + } + + *x = img_x; + *y = img_y; + if (comp) *comp = target; + return out; +} + +#ifndef STBI_NO_STDIO +stbi_uc *stbi_bmp_load (char *filename, int *x, int *y, int *comp, int req_comp) +{ + stbi_uc *data; + FILE *f = fopen(filename, "rb"); + if (!f) return NULL; + data = stbi_bmp_load_from_file(f, x,y,comp,req_comp); + fclose(f); + return data; +} + +stbi_uc *stbi_bmp_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp) +{ + start_file(f); + return bmp_load(x,y,comp,req_comp); +} +#endif + +stbi_uc *stbi_bmp_load_from_memory (stbi_uc *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + start_mem(buffer, len); + return bmp_load(x,y,comp,req_comp); +} + +// Targa Truevision - TGA +// by Jonathan Dummer + +static int tga_test(void) +{ + int sz; + get8u(); // discard Offset + sz = get8u(); // color type + if( sz > 1 ) return 0; // only RGB or indexed allowed + sz = get8u(); // image type + if( (sz != 1) && (sz != 2) && (sz != 3) && (sz != 9) && (sz != 10) && (sz != 11) ) return 0; // only RGB or grey allowed, +/- RLE + get16(); // discard palette start + get16(); // discard palette length + get8(); // discard bits per palette color entry + get16(); // discard x origin + get16(); // discard y origin + if( get16() < 1 ) return 0; // test width + if( get16() < 1 ) return 0; // test height + sz = get8(); // bits per pixel + if( (sz != 8) && (sz != 16) && (sz != 24) && (sz != 32) ) return 0; // only RGB or RGBA or grey allowed + return 1; // seems to have passed everything +} + +#ifndef STBI_NO_STDIO +int stbi_tga_test_file (FILE *f) +{ + int r,n = ftell(f); + start_file(f); + r = tga_test(); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +int stbi_tga_test_memory (stbi_uc *buffer, int len) +{ + start_mem(buffer, len); + return tga_test(); +} + +static stbi_uc *tga_load(int *x, int *y, int *comp, int req_comp) +{ + // read in the TGA header stuff + int tga_offset = get8u(); + int tga_indexed = get8u(); + int tga_image_type = get8u(); + int tga_is_RLE = 0; + int tga_palette_start = get16le(); + int tga_palette_len = get16le(); + int tga_palette_bits = get8u(); + int tga_x_origin = get16le(); + int tga_y_origin = get16le(); + int tga_width = get16le(); + int tga_height = get16le(); + int tga_bits_per_pixel = get8u(); + int tga_inverted = get8u(); + // image data + unsigned char *tga_data; + unsigned char *tga_palette = NULL; + int i, j; + unsigned char raw_data[4]; + unsigned char trans_data[4]; + int RLE_count = 0; + int RLE_repeating = 0; + int read_next_pixel = 1; + // do a tiny bit of precessing + if( tga_image_type >= 8 ) + { + tga_image_type -= 8; + tga_is_RLE = 1; + } + /* int tga_alpha_bits = tga_inverted & 15; */ + tga_inverted = 1 - ((tga_inverted >> 5) & 1); + + // error check + if( //(tga_indexed) || + (tga_width < 1) || (tga_height < 1) || + (tga_image_type < 1) || (tga_image_type > 3) || + ((tga_bits_per_pixel != 8) && (tga_bits_per_pixel != 16) && + (tga_bits_per_pixel != 24) && (tga_bits_per_pixel != 32)) + ) + { + return NULL; + } + + // If I'm paletted, then I'll use the number of bits from the palette + if( tga_indexed ) + { + tga_bits_per_pixel = tga_palette_bits; + } + + // tga info + *x = tga_width; + *y = tga_height; + if( (req_comp < 1) || (req_comp > 4) ) + { + // just use whatever the file was + req_comp = tga_bits_per_pixel / 8; + *comp = req_comp; + } else + { + // force a new number of components + *comp = tga_bits_per_pixel/8; + } + tga_data = (unsigned char*)malloc( tga_width * tga_height * req_comp ); + + // skip to the data's starting position (offset usually = 0) + skip( tga_offset ); + // do I need to load a palette? + if( tga_indexed ) + { + // any data to skip? (offset usually = 0) + skip( tga_palette_start ); + // load the palette + tga_palette = (unsigned char*)malloc( tga_palette_len * tga_palette_bits / 8 ); + getn( tga_palette, tga_palette_len * tga_palette_bits / 8 ); + } + // load the data + for( i = 0; i < tga_width * tga_height; ++i ) + { + // if I'm in RLE mode, do I need to get a RLE chunk? + if( tga_is_RLE ) + { + if( RLE_count == 0 ) + { + // yep, get the next byte as a RLE command + int RLE_cmd = get8u(); + RLE_count = 1 + (RLE_cmd & 127); + RLE_repeating = RLE_cmd >> 7; + read_next_pixel = 1; + } else if( !RLE_repeating ) + { + read_next_pixel = 1; + } + } else + { + read_next_pixel = 1; + } + // OK, if I need to read a pixel, do it now + if( read_next_pixel ) + { + // load however much data we did have + if( tga_indexed ) + { + // read in 1 byte, then perform the lookup + int pal_idx = get8u(); + if( pal_idx >= tga_palette_len ) + { + // invalid index + pal_idx = 0; + } + pal_idx *= tga_bits_per_pixel / 8; + for( j = 0; j*8 < tga_bits_per_pixel; ++j ) + { + raw_data[j] = tga_palette[pal_idx+j]; + } + } else + { + // read in the data raw + for( j = 0; j*8 < tga_bits_per_pixel; ++j ) + { + raw_data[j] = get8u(); + } + } + // convert raw to the intermediate format + switch( tga_bits_per_pixel ) + { + case 8: + // Luminous => RGBA + trans_data[0] = raw_data[0]; + trans_data[1] = raw_data[0]; + trans_data[2] = raw_data[0]; + trans_data[3] = 255; + break; + case 16: + // Luminous,Alpha => RGBA + trans_data[0] = raw_data[0]; + trans_data[1] = raw_data[0]; + trans_data[2] = raw_data[0]; + trans_data[3] = raw_data[1]; + break; + case 24: + // BGR => RGBA + trans_data[0] = raw_data[2]; + trans_data[1] = raw_data[1]; + trans_data[2] = raw_data[0]; + trans_data[3] = 255; + break; + case 32: + // BGRA => RGBA + trans_data[0] = raw_data[2]; + trans_data[1] = raw_data[1]; + trans_data[2] = raw_data[0]; + trans_data[3] = raw_data[3]; + break; + } + // clear the reading flag for the next pixel + read_next_pixel = 0; + } // end of reading a pixel + // convert to final format + switch( req_comp ) + { + case 1: + // RGBA => Luminance + tga_data[i*req_comp+0] = compute_y(trans_data[0],trans_data[1],trans_data[2]); + break; + case 2: + // RGBA => Luminance,Alpha + tga_data[i*req_comp+0] = compute_y(trans_data[0],trans_data[1],trans_data[2]); + tga_data[i*req_comp+1] = trans_data[3]; + break; + case 3: + // RGBA => RGB + tga_data[i*req_comp+0] = trans_data[0]; + tga_data[i*req_comp+1] = trans_data[1]; + tga_data[i*req_comp+2] = trans_data[2]; + break; + case 4: + // RGBA => RGBA + tga_data[i*req_comp+0] = trans_data[0]; + tga_data[i*req_comp+1] = trans_data[1]; + tga_data[i*req_comp+2] = trans_data[2]; + tga_data[i*req_comp+3] = trans_data[3]; + break; + } + // in case we're in RLE mode, keep counting down + --RLE_count; + } + // do I need to invert the image? + if( tga_inverted ) + { + for( j = 0; j*2 < tga_height; ++j ) + { + int index1 = j * tga_width * req_comp; + int index2 = (tga_height - 1 - j) * tga_width * req_comp; + for( i = tga_width * req_comp; i > 0; --i ) + { + unsigned char temp = tga_data[index1]; + tga_data[index1] = tga_data[index2]; + tga_data[index2] = temp; + ++index1; + ++index2; + } + } + } + // clear my palette, if I had one + if( tga_palette != NULL ) + { + free( tga_palette ); + } + // the things I do to get rid of an error message, and yet keep + // Microsoft's C compilers happy... [8^( + tga_palette_start = tga_palette_len = tga_palette_bits = + tga_x_origin = tga_y_origin = 0; + // OK, done + return tga_data; +} + +#ifndef STBI_NO_STDIO +stbi_uc *stbi_tga_load (char *filename, int *x, int *y, int *comp, int req_comp) +{ + stbi_uc *data; + FILE *f = fopen(filename, "rb"); + if (!f) return NULL; + data = stbi_tga_load_from_file(f, x,y,comp,req_comp); + fclose(f); + return data; +} + +stbi_uc *stbi_tga_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp) +{ + start_file(f); + return tga_load(x,y,comp,req_comp); +} +#endif + +stbi_uc *stbi_tga_load_from_memory (stbi_uc *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + start_mem(buffer, len); + return tga_load(x,y,comp,req_comp); +} + + +// ************************************************************************************************* +// Photoshop PSD loader -- PD by Thatcher Ulrich, integration by Nicholas Schulz, tweaked by STB + +static int psd_test(void) +{ + if (get32() != 0x38425053) return 0; // "8BPS" + else return 1; +} + +#ifndef STBI_NO_STDIO +int stbi_psd_test_file(FILE *f) +{ + int r,n = ftell(f); + start_file(f); + r = psd_test(); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +int stbi_psd_test_memory(stbi_uc *buffer, int len) +{ + start_mem(buffer, len); + return psd_test(); +} + +static stbi_uc *psd_load(int *x, int *y, int *comp, int req_comp) +{ + int pixelCount; + int channelCount, compression; + int channel, i, count, len; + int w,h; + + // Check identifier + if (get32() != 0x38425053) // "8BPS" + return epuc("not PSD", "Corrupt PSD image"); + + // Check file type version. + if (get16() != 1) + return epuc("wrong version", "Unsupported version of PSD image"); + + // Skip 6 reserved bytes. + skip( 6 ); + + // Read the number of channels (R, G, B, A, etc). + channelCount = get16(); + if (channelCount < 0 || channelCount > 16) + return epuc("wrong channel count", "Unsupported number of channels in PSD image"); + + // Read the rows and columns of the image. + h = get32(); + w = get32(); + + // Make sure the depth is 8 bits. + if (get16() != 8) + return epuc("unsupported bit depth", "PSD bit depth is not 8 bit"); + + // Make sure the color mode is RGB. + // Valid options are: + // 0: Bitmap + // 1: Grayscale + // 2: Indexed color + // 3: RGB color + // 4: CMYK color + // 7: Multichannel + // 8: Duotone + // 9: Lab color + if (get16() != 3) + return epuc("wrong color format", "PSD is not in RGB color format"); + + // Skip the Mode Data. (It's the palette for indexed color; other info for other modes.) + skip(get32() ); + + // Skip the image resources. (resolution, pen tool paths, etc) + skip( get32() ); + + // Skip the reserved data. + skip( get32() ); + + // Find out if the data is compressed. + // Known values: + // 0: no compression + // 1: RLE compressed + compression = get16(); + if (compression > 1) + return epuc("unknown compression type", "PSD has an unknown compression format"); + + // Create the destination image. + out = (stbi_uc *) malloc(4 * w*h); + if (!out) return epuc("outofmem", "Out of memory"); + pixelCount = w*h; + + // Initialize the data to zero. + //memset( out, 0, pixelCount * 4 ); + + // Finally, the image data. + if (compression) { + // RLE as used by .PSD and .TIFF + // Loop until you get the number of unpacked bytes you are expecting: + // Read the next source byte into n. + // If n is between 0 and 127 inclusive, copy the next n+1 bytes literally. + // Else if n is between -127 and -1 inclusive, copy the next byte -n+1 times. + // Else if n is 128, noop. + // Endloop + + // The RLE-compressed data is preceeded by a 2-byte data count for each row in the data, + // which we're going to just skip. + skip( h * channelCount * 2 ); + + // Read the RLE data by channel. + for (channel = 0; channel < 4; channel++) { + uint8 *p; + + p = out+channel; + if (channel >= channelCount) { + // Fill this channel with default data. + for (i = 0; i < pixelCount; i++) *p = (channel == 3 ? 255 : 0), p += 4; + } else { + // Read the RLE data. + count = 0; + while (count < pixelCount) { + len = get8(); + if (len == 128) { + // No-op. + } else if (len < 128) { + // Copy next len+1 bytes literally. + len++; + count += len; + while (len) { + *p = get8(); + p += 4; + len--; + } + } else if (len > 128) { + uint32 val; + // Next -len+1 bytes in the dest are replicated from next source byte. + // (Interpret len as a negative 8-bit int.) + len ^= 0x0FF; + len += 2; + val = get8(); + count += len; + while (len) { + *p = val; + p += 4; + len--; + } + } + } + } + } + + } else { + // We're at the raw image data. It's each channel in order (Red, Green, Blue, Alpha, ...) + // where each channel consists of an 8-bit value for each pixel in the image. + + // Read the data by channel. + for (channel = 0; channel < 4; channel++) { + uint8 *p; + + p = out + channel; + if (channel > channelCount) { + // Fill this channel with default data. + for (i = 0; i < pixelCount; i++) *p = channel == 3 ? 255 : 0, p += 4; + } else { + // Read the data. + count = 0; + for (i = 0; i < pixelCount; i++) + *p = get8(), p += 4; + } + } + } + + if (req_comp && req_comp != 4) { + img_x = w; + img_y = h; + out = convert_format(out, 4, req_comp); + if (out == NULL) return out; // convert_format frees input on failure + } + + if (comp) *comp = channelCount; + *y = h; + *x = w; + + return out; +} + +#ifndef STBI_NO_STDIO +stbi_uc *stbi_psd_load(char *filename, int *x, int *y, int *comp, int req_comp) +{ + stbi_uc *data; + FILE *f = fopen(filename, "rb"); + if (!f) return NULL; + data = stbi_psd_load_from_file(f, x,y,comp,req_comp); + fclose(f); + return data; +} + +stbi_uc *stbi_psd_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + start_file(f); + return psd_load(x,y,comp,req_comp); +} +#endif + +stbi_uc *stbi_psd_load_from_memory (stbi_uc *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + start_mem(buffer, len); + return psd_load(x,y,comp,req_comp); +} + + +// ************************************************************************************************* +// Radiance RGBE HDR loader +// originally by Nicolas Schulz +#ifndef STBI_NO_HDR +static int hdr_test(void) +{ + char *signature = "#?RADIANCE\n"; + int i; + for (i=0; signature[i]; ++i) + if (get8() != signature[i]) + return 0; + return 1; +} + +int stbi_hdr_test_memory(stbi_uc *buffer, int len) +{ + start_mem(buffer, len); + return hdr_test(); +} + +#ifndef STBI_NO_STDIO +int stbi_hdr_test_file(FILE *f) +{ + int r,n = ftell(f); + start_file(f); + r = hdr_test(); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +#define HDR_BUFLEN 1024 +static char *hdr_gettoken(char *buffer) +{ + int len=0; + char *s = buffer, c = '\0'; + + c = get8(); + + while (!at_eof() && c != '\n') { + buffer[len++] = c; + if (len == HDR_BUFLEN-1) { + // flush to end of line + while (!at_eof() && get8() != '\n') + ; + break; + } + c = get8(); + } + + buffer[len] = 0; + return buffer; +} + +static void hdr_convert(float *output, stbi_uc *input, int req_comp) +{ + if( input[3] != 0 ) { + float f1; + // Exponent + f1 = (float) ldexp(1.0f, input[3] - (int)(128 + 8)); + if (req_comp <= 2) + output[0] = (input[0] + input[1] + input[2]) * f1 / 3; + else { + output[0] = input[0] * f1; + output[1] = input[1] * f1; + output[2] = input[2] * f1; + } + if (req_comp == 2) output[1] = 1; + if (req_comp == 4) output[3] = 1; + } else { + switch (req_comp) { + case 4: output[3] = 1; /* fallthrough */ + case 3: output[0] = output[1] = output[2] = 0; + break; + case 2: output[1] = 1; /* fallthrough */ + case 1: output[0] = 0; + break; + } + } +} + + +static float *hdr_load(int *x, int *y, int *comp, int req_comp) +{ + char buffer[HDR_BUFLEN]; + char *token; + int valid = 0; + int width, height; + stbi_uc *scanline; + float *hdr_data; + int len; + unsigned char count, value; + int i, j, k, c1,c2, z; + + + // Check identifier + if (strcmp(hdr_gettoken(buffer), "#?RADIANCE") != 0) + return epf("not HDR", "Corrupt HDR image"); + + // Parse header + while(1) { + token = hdr_gettoken(buffer); + if (token[0] == 0) break; + if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1; + } + + if (!valid) return epf("unsupported format", "Unsupported HDR format"); + + // Parse width and height + // can't use sscanf() if we're not using stdio! + token = hdr_gettoken(buffer); + if (strncmp(token, "-Y ", 3)) return epf("unsupported data layout", "Unsupported HDR format"); + token += 3; + height = strtol(token, &token, 10); + while (*token == ' ') ++token; + if (strncmp(token, "+X ", 3)) return epf("unsupported data layout", "Unsupported HDR format"); + token += 3; + width = strtol(token, NULL, 10); + + *x = width; + *y = height; + + *comp = 3; + if (req_comp == 0) req_comp = 3; + + // Read data + hdr_data = (float *) malloc(height * width * req_comp * sizeof(float)); + + // Load image data + // image data is stored as some number of sca + if( width < 8 || width >= 32768) { + // Read flat data + for (j=0; j < height; ++j) { + for (i=0; i < width; ++i) { + stbi_uc rgbe[4]; + main_decode_loop: + getn(rgbe, 4); + hdr_convert(hdr_data + j * width * req_comp + i * req_comp, rgbe, req_comp); + } + } + } else { + // Read RLE-encoded data + scanline = NULL; + + for (j = 0; j < height; ++j) { + c1 = get8(); + c2 = get8(); + len = get8(); + if (c1 != 2 || c2 != 2 || (len & 0x80)) { + // not run-length encoded, so we have to actually use THIS data as a decoded + // pixel (note this can't be a valid pixel--one of RGB must be >= 128) + stbi_uc rgbe[4] = { c1,c2,len, get8() }; + hdr_convert(hdr_data, rgbe, req_comp); + i = 1; + j = 0; + free(scanline); + goto main_decode_loop; // yes, this is fucking insane; blame the fucking insane format + } + len <<= 8; + len |= get8(); + if (len != width) { free(hdr_data); free(scanline); return epf("invalid decoded scanline length", "corrupt HDR"); } + if (scanline == NULL) scanline = (stbi_uc *) malloc(width * 4); + + for (k = 0; k < 4; ++k) { + i = 0; + while (i < width) { + count = get8(); + if (count > 128) { + // Run + value = get8(); + count -= 128; + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = value; + } else { + // Dump + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = get8(); + } + } + } + for (i=0; i < width; ++i) + hdr_convert(hdr_data+(j*width + i)*req_comp, scanline + i*4, req_comp); + } + free(scanline); + } + + return hdr_data; +} + +#ifndef STBI_NO_STDIO +float *stbi_hdr_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + start_file(f); + return hdr_load(x,y,comp,req_comp); +} +#endif + +float *stbi_hdr_load_from_memory(stbi_uc *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + start_mem(buffer, len); + return hdr_load(x,y,comp,req_comp); +} + +#endif // STBI_NO_HDR + +/////////////////////// write image /////////////////////// + +#ifndef STBI_NO_WRITE + +static void write8(FILE *f, int x) { uint8 z = (uint8) x; fwrite(&z,1,1,f); } + +static void writefv(FILE *f, char *fmt, va_list v) +{ + while (*fmt) { + switch (*fmt++) { + case ' ': break; + case '1': { uint8 x = va_arg(v, int); write8(f,x); break; } + case '2': { int16 x = va_arg(v, int); write8(f,x); write8(f,x>>8); break; } + case '4': { int32 x = va_arg(v, int); write8(f,x); write8(f,x>>8); write8(f,x>>16); write8(f,x>>24); break; } + default: + assert(0); + va_end(v); + return; + } + } +} + +static void writef(FILE *f, char *fmt, ...) +{ + va_list v; + va_start(v, fmt); + writefv(f,fmt,v); + va_end(v); +} + +static void write_pixels(FILE *f, int rgb_dir, int vdir, int x, int y, int comp, void *data, int write_alpha, int scanline_pad) +{ + uint8 bg[3] = { 255, 0, 255}, px[3]; + uint32 zero = 0; + int i,j,k, j_end; + + if (vdir < 0) + j_end = -1, j = y-1; + else + j_end = y, j = 0; + + for (; j != j_end; j += vdir) { + for (i=0; i < x; ++i) { + uint8 *d = (uint8 *) data + (j*x+i)*comp; + if (write_alpha < 0) + fwrite(&d[comp-1], 1, 1, f); + switch (comp) { + case 1: + case 2: writef(f, "111", d[0],d[0],d[0]); + break; + case 4: + if (!write_alpha) { + for (k=0; k < 3; ++k) + px[k] = bg[k] + ((d[k] - bg[k]) * d[3])/255; + writef(f, "111", px[1-rgb_dir],px[1],px[1+rgb_dir]); + break; + } + /* FALLTHROUGH */ + case 3: + writef(f, "111", d[1-rgb_dir],d[1],d[1+rgb_dir]); + break; + } + if (write_alpha > 0) + fwrite(&d[comp-1], 1, 1, f); + } + fwrite(&zero,scanline_pad,1,f); + } +} + +static int outfile(char *filename, int rgb_dir, int vdir, int x, int y, int comp, void *data, int alpha, int pad, char *fmt, ...) +{ + FILE *f = fopen(filename, "wb"); + if (f) { + va_list v; + va_start(v, fmt); + writefv(f, fmt, v); + va_end(v); + write_pixels(f,rgb_dir,vdir,x,y,comp,data,alpha,pad); + fclose(f); + } + return f != NULL; +} + +int stbi_write_bmp(char *filename, int x, int y, int comp, void *data) +{ + int pad = (-x*3) & 3; + return outfile(filename,-1,-1,x,y,comp,data,0,pad, + "11 4 22 4" "4 44 22 444444", + 'B', 'M', 14+40+(x*3+pad)*y, 0,0, 14+40, // file header + 40, x,y, 1,24, 0,0,0,0,0,0); // bitmap header +} + +int stbi_write_tga(char *filename, int x, int y, int comp, void *data) +{ + int has_alpha = !(comp & 1); + return outfile(filename, -1,-1, x, y, comp, data, has_alpha, 0, + "111 221 2222 11", 0,0,2, 0,0,0, 0,0,x,y, 24+8*has_alpha, 8*has_alpha); +} + +// any other image formats that do interleaved rgb data? +// PNG: requires adler32,crc32 -- significant amount of code +// PSD: no, channels output separately +// TIFF: no, stripwise-interleaved... i think + +#endif // STBI_NO_WRITE diff --git a/lib/soil/soil/original/stb_image-1.16.c b/lib/soil/soil/original/stb_image-1.16.c new file mode 100644 index 0000000..029b58e --- /dev/null +++ b/lib/soil/soil/original/stb_image-1.16.c @@ -0,0 +1,3821 @@ +/* stbi-1.16 - public domain JPEG/PNG reader - http://nothings.org/stb_image.c + when you control the images you're loading + + QUICK NOTES: + Primarily of interest to game developers and other people who can + avoid problematic images and only need the trivial interface + + JPEG baseline (no JPEG progressive, no oddball channel decimations) + PNG non-interlaced + BMP non-1bpp, non-RLE + TGA (not sure what subset, if a subset) + PSD (composited view only, no extra channels) + HDR (radiance rgbE format) + writes BMP,TGA (define STBI_NO_WRITE to remove code) + decoded from memory or through stdio FILE (define STBI_NO_STDIO to remove code) + supports installable dequantizing-IDCT, YCbCr-to-RGB conversion (define STBI_SIMD) + + TODO: + stbi_info_* + + history: + 1.16 major bugfix - convert_format converted one too many pixels + 1.15 initialize some fields for thread safety + 1.14 fix threadsafe conversion bug; header-file-only version (#define STBI_HEADER_FILE_ONLY before including) + 1.13 threadsafe + 1.12 const qualifiers in the API + 1.11 Support installable IDCT, colorspace conversion routines + 1.10 Fixes for 64-bit (don't use "unsigned long") + optimized upsampling by Fabian "ryg" Giesen + 1.09 Fix format-conversion for PSD code (bad global variables!) + 1.08 Thatcher Ulrich's PSD code integrated by Nicolas Schulz + 1.07 attempt to fix C++ warning/errors again + 1.06 attempt to fix C++ warning/errors again + 1.05 fix TGA loading to return correct *comp and use good luminance calc + 1.04 default float alpha is 1, not 255; use 'void *' for stbi_image_free + 1.03 bugfixes to STBI_NO_STDIO, STBI_NO_HDR + 1.02 support for (subset of) HDR files, float interface for preferred access to them + 1.01 fix bug: possible bug in handling right-side up bmps... not sure + fix bug: the stbi_bmp_load() and stbi_tga_load() functions didn't work at all + 1.00 interface to zlib that skips zlib header + 0.99 correct handling of alpha in palette + 0.98 TGA loader by lonesock; dynamically add loaders (untested) + 0.97 jpeg errors on too large a file; also catch another malloc failure + 0.96 fix detection of invalid v value - particleman@mollyrocket forum + 0.95 during header scan, seek to markers in case of padding + 0.94 STBI_NO_STDIO to disable stdio usage; rename all #defines the same + 0.93 handle jpegtran output; verbose errors + 0.92 read 4,8,16,24,32-bit BMP files of several formats + 0.91 output 24-bit Windows 3.0 BMP files + 0.90 fix a few more warnings; bump version number to approach 1.0 + 0.61 bugfixes due to Marc LeBlanc, Christopher Lloyd + 0.60 fix compiling as c++ + 0.59 fix warnings: merge Dave Moore's -Wall fixes + 0.58 fix bug: zlib uncompressed mode len/nlen was wrong endian + 0.57 fix bug: jpg last huffman symbol before marker was >9 bits but less + than 16 available + 0.56 fix bug: zlib uncompressed mode len vs. nlen + 0.55 fix bug: restart_interval not initialized to 0 + 0.54 allow NULL for 'int *comp' + 0.53 fix bug in png 3->4; speedup png decoding + 0.52 png handles req_comp=3,4 directly; minor cleanup; jpeg comments + 0.51 obey req_comp requests, 1-component jpegs return as 1-component, + on 'test' only check type, not whether we support this variant +*/ + + +#ifndef STBI_INCLUDE_STB_IMAGE_H +#define STBI_INCLUDE_STB_IMAGE_H + +//// begin header file //////////////////////////////////////////////////// +// +// Limitations: +// - no progressive/interlaced support (jpeg, png) +// - 8-bit samples only (jpeg, png) +// - not threadsafe +// - channel subsampling of at most 2 in each dimension (jpeg) +// - no delayed line count (jpeg) -- IJG doesn't support either +// +// Basic usage (see HDR discussion below): +// int x,y,n; +// unsigned char *data = stbi_load(filename, &x, &y, &n, 0); +// // ... process data if not NULL ... +// // ... x = width, y = height, n = # 8-bit components per pixel ... +// // ... replace '0' with '1'..'4' to force that many components per pixel +// stbi_image_free(data) +// +// Standard parameters: +// int *x -- outputs image width in pixels +// int *y -- outputs image height in pixels +// int *comp -- outputs # of image components in image file +// int req_comp -- if non-zero, # of image components requested in result +// +// The return value from an image loader is an 'unsigned char *' which points +// to the pixel data. The pixel data consists of *y scanlines of *x pixels, +// with each pixel consisting of N interleaved 8-bit components; the first +// pixel pointed to is top-left-most in the image. There is no padding between +// image scanlines or between pixels, regardless of format. The number of +// components N is 'req_comp' if req_comp is non-zero, or *comp otherwise. +// If req_comp is non-zero, *comp has the number of components that _would_ +// have been output otherwise. E.g. if you set req_comp to 4, you will always +// get RGBA output, but you can check *comp to easily see if it's opaque. +// +// An output image with N components has the following components interleaved +// in this order in each pixel: +// +// N=#comp components +// 1 grey +// 2 grey, alpha +// 3 red, green, blue +// 4 red, green, blue, alpha +// +// If image loading fails for any reason, the return value will be NULL, +// and *x, *y, *comp will be unchanged. The function stbi_failure_reason() +// can be queried for an extremely brief, end-user unfriendly explanation +// of why the load failed. Define STBI_NO_FAILURE_STRINGS to avoid +// compiling these strings at all, and STBI_FAILURE_USERMSG to get slightly +// more user-friendly ones. +// +// Paletted PNG and BMP images are automatically depalettized. +// +// +// =========================================================================== +// +// HDR image support (disable by defining STBI_NO_HDR) +// +// stb_image now supports loading HDR images in general, and currently +// the Radiance .HDR file format, although the support is provided +// generically. You can still load any file through the existing interface; +// if you attempt to load an HDR file, it will be automatically remapped to +// LDR, assuming gamma 2.2 and an arbitrary scale factor defaulting to 1; +// both of these constants can be reconfigured through this interface: +// +// stbi_hdr_to_ldr_gamma(2.2f); +// stbi_hdr_to_ldr_scale(1.0f); +// +// (note, do not use _inverse_ constants; stbi_image will invert them +// appropriately). +// +// Additionally, there is a new, parallel interface for loading files as +// (linear) floats to preserve the full dynamic range: +// +// float *data = stbi_loadf(filename, &x, &y, &n, 0); +// +// If you load LDR images through this interface, those images will +// be promoted to floating point values, run through the inverse of +// constants corresponding to the above: +// +// stbi_ldr_to_hdr_scale(1.0f); +// stbi_ldr_to_hdr_gamma(2.2f); +// +// Finally, given a filename (or an open file or memory block--see header +// file for details) containing image data, you can query for the "most +// appropriate" interface to use (that is, whether the image is HDR or +// not), using: +// +// stbi_is_hdr(char *filename); + +#ifndef STBI_NO_STDIO +#include +#endif + +#define STBI_VERSION 1 + +enum +{ + STBI_default = 0, // only used for req_comp + + STBI_grey = 1, + STBI_grey_alpha = 2, + STBI_rgb = 3, + STBI_rgb_alpha = 4, +}; + +typedef unsigned char stbi_uc; + +#ifdef __cplusplus +extern "C" { +#endif + +// WRITING API + +#if !defined(STBI_NO_WRITE) && !defined(STBI_NO_STDIO) +// write a BMP/TGA file given tightly packed 'comp' channels (no padding, nor bmp-stride-padding) +// (you must include the appropriate extension in the filename). +// returns TRUE on success, FALSE if couldn't open file, error writing file +extern int stbi_write_bmp (char const *filename, int x, int y, int comp, void *data); +extern int stbi_write_tga (char const *filename, int x, int y, int comp, void *data); +#endif + +// PRIMARY API - works on images of any type + +// load image by filename, open file, or memory buffer +#ifndef STBI_NO_STDIO +extern stbi_uc *stbi_load (char const *filename, int *x, int *y, int *comp, int req_comp); +extern stbi_uc *stbi_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +extern int stbi_info_from_file (FILE *f, int *x, int *y, int *comp); +#endif +extern stbi_uc *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); +// for stbi_load_from_file, file pointer is left pointing immediately after image + +#ifndef STBI_NO_HDR +#ifndef STBI_NO_STDIO +extern float *stbi_loadf (char const *filename, int *x, int *y, int *comp, int req_comp); +extern float *stbi_loadf_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +#endif +extern float *stbi_loadf_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); + +extern void stbi_hdr_to_ldr_gamma(float gamma); +extern void stbi_hdr_to_ldr_scale(float scale); + +extern void stbi_ldr_to_hdr_gamma(float gamma); +extern void stbi_ldr_to_hdr_scale(float scale); + +#endif // STBI_NO_HDR + +// get a VERY brief reason for failure +// NOT THREADSAFE +extern char *stbi_failure_reason (void); + +// free the loaded image -- this is just free() +extern void stbi_image_free (void *retval_from_stbi_load); + +// get image dimensions & components without fully decoding +extern int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp); +extern int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len); +#ifndef STBI_NO_STDIO +extern int stbi_info (char const *filename, int *x, int *y, int *comp); +extern int stbi_is_hdr (char const *filename); +extern int stbi_is_hdr_from_file(FILE *f); +#endif + +// ZLIB client - used by PNG, available for other purposes + +extern char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen); +extern char *stbi_zlib_decode_malloc(const char *buffer, int len, int *outlen); +extern int stbi_zlib_decode_buffer(char *obuffer, int olen, const char *ibuffer, int ilen); + +extern char *stbi_zlib_decode_noheader_malloc(const char *buffer, int len, int *outlen); +extern int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen); + +// TYPE-SPECIFIC ACCESS + +// is it a jpeg? +extern int stbi_jpeg_test_memory (stbi_uc const *buffer, int len); +extern stbi_uc *stbi_jpeg_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); +extern int stbi_jpeg_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp); + +#ifndef STBI_NO_STDIO +extern stbi_uc *stbi_jpeg_load (char const *filename, int *x, int *y, int *comp, int req_comp); +extern int stbi_jpeg_test_file (FILE *f); +extern stbi_uc *stbi_jpeg_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); + +extern int stbi_jpeg_info (char const *filename, int *x, int *y, int *comp); +extern int stbi_jpeg_info_from_file (FILE *f, int *x, int *y, int *comp); +#endif + +// is it a png? +extern int stbi_png_test_memory (stbi_uc const *buffer, int len); +extern stbi_uc *stbi_png_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); +extern int stbi_png_info_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp); + +#ifndef STBI_NO_STDIO +extern stbi_uc *stbi_png_load (char const *filename, int *x, int *y, int *comp, int req_comp); +extern int stbi_png_info (char const *filename, int *x, int *y, int *comp); +extern int stbi_png_test_file (FILE *f); +extern stbi_uc *stbi_png_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +extern int stbi_png_info_from_file (FILE *f, int *x, int *y, int *comp); +#endif + +// is it a bmp? +extern int stbi_bmp_test_memory (stbi_uc const *buffer, int len); + +extern stbi_uc *stbi_bmp_load (char const *filename, int *x, int *y, int *comp, int req_comp); +extern stbi_uc *stbi_bmp_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); +#ifndef STBI_NO_STDIO +extern int stbi_bmp_test_file (FILE *f); +extern stbi_uc *stbi_bmp_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +#endif + +// is it a tga? +extern int stbi_tga_test_memory (stbi_uc const *buffer, int len); + +extern stbi_uc *stbi_tga_load (char const *filename, int *x, int *y, int *comp, int req_comp); +extern stbi_uc *stbi_tga_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); +#ifndef STBI_NO_STDIO +extern int stbi_tga_test_file (FILE *f); +extern stbi_uc *stbi_tga_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +#endif + +// is it a psd? +extern int stbi_psd_test_memory (stbi_uc const *buffer, int len); + +extern stbi_uc *stbi_psd_load (char const *filename, int *x, int *y, int *comp, int req_comp); +extern stbi_uc *stbi_psd_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); +#ifndef STBI_NO_STDIO +extern int stbi_psd_test_file (FILE *f); +extern stbi_uc *stbi_psd_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +#endif + +// is it an hdr? +extern int stbi_hdr_test_memory (stbi_uc const *buffer, int len); + +extern float * stbi_hdr_load (char const *filename, int *x, int *y, int *comp, int req_comp); +extern float * stbi_hdr_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); +#ifndef STBI_NO_STDIO +extern int stbi_hdr_test_file (FILE *f); +extern float * stbi_hdr_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +#endif + +// define new loaders +typedef struct +{ + int (*test_memory)(stbi_uc const *buffer, int len); + stbi_uc * (*load_from_memory)(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); + #ifndef STBI_NO_STDIO + int (*test_file)(FILE *f); + stbi_uc * (*load_from_file)(FILE *f, int *x, int *y, int *comp, int req_comp); + #endif +} stbi_loader; + +// register a loader by filling out the above structure (you must defined ALL functions) +// returns 1 if added or already added, 0 if not added (too many loaders) +// NOT THREADSAFE +extern int stbi_register_loader(stbi_loader *loader); + +// define faster low-level operations (typically SIMD support) +#if STBI_SIMD +typedef void (*stbi_idct_8x8)(uint8 *out, int out_stride, short data[64], unsigned short *dequantize); +// compute an integer IDCT on "input" +// input[x] = data[x] * dequantize[x] +// write results to 'out': 64 samples, each run of 8 spaced by 'out_stride' +// CLAMP results to 0..255 +typedef void (*stbi_YCbCr_to_RGB_run)(uint8 *output, uint8 const *y, uint8 const *cb, uint8 const *cr, int count, int step); +// compute a conversion from YCbCr to RGB +// 'count' pixels +// write pixels to 'output'; each pixel is 'step' bytes (either 3 or 4; if 4, write '255' as 4th), order R,G,B +// y: Y input channel +// cb: Cb input channel; scale/biased to be 0..255 +// cr: Cr input channel; scale/biased to be 0..255 + +extern void stbi_install_idct(stbi_idct_8x8 func); +extern void stbi_install_YCbCr_to_RGB(stbi_YCbCr_to_RGB_run func); +#endif // STBI_SIMD + +#ifdef __cplusplus +} +#endif + +// +// +//// end header file ///////////////////////////////////////////////////// +#endif // STBI_INCLUDE_STB_IMAGE_H + +#ifndef STBI_HEADER_FILE_ONLY + +#ifndef STBI_NO_HDR +#include // ldexp +#include // strcmp +#endif + +#ifndef STBI_NO_STDIO +#include +#endif +#include +#include +#include +#include + +#ifndef _MSC_VER + #ifdef __cplusplus + #define __forceinline inline + #else + #define __forceinline + #endif +#endif + + +// implementation: +typedef unsigned char uint8; +typedef unsigned short uint16; +typedef signed short int16; +typedef unsigned int uint32; +typedef signed int int32; +typedef unsigned int uint; + +// should produce compiler error if size is wrong +typedef unsigned char validate_uint32[sizeof(uint32)==4]; + +#if defined(STBI_NO_STDIO) && !defined(STBI_NO_WRITE) +#define STBI_NO_WRITE +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// Generic API that works on all image types +// + +// this is not threadsafe +static char *failure_reason; + +char *stbi_failure_reason(void) +{ + return failure_reason; +} + +static int e(char *str) +{ + failure_reason = str; + return 0; +} + +#ifdef STBI_NO_FAILURE_STRINGS + #define e(x,y) 0 +#elif defined(STBI_FAILURE_USERMSG) + #define e(x,y) e(y) +#else + #define e(x,y) e(x) +#endif + +#define epf(x,y) ((float *) (e(x,y)?NULL:NULL)) +#define epuc(x,y) ((unsigned char *) (e(x,y)?NULL:NULL)) + +void stbi_image_free(void *retval_from_stbi_load) +{ + free(retval_from_stbi_load); +} + +#define MAX_LOADERS 32 +stbi_loader *loaders[MAX_LOADERS]; +static int max_loaders = 0; + +int stbi_register_loader(stbi_loader *loader) +{ + int i; + for (i=0; i < MAX_LOADERS; ++i) { + // already present? + if (loaders[i] == loader) + return 1; + // end of the list? + if (loaders[i] == NULL) { + loaders[i] = loader; + max_loaders = i+1; + return 1; + } + } + // no room for it + return 0; +} + +#ifndef STBI_NO_HDR +static float *ldr_to_hdr(stbi_uc *data, int x, int y, int comp); +static stbi_uc *hdr_to_ldr(float *data, int x, int y, int comp); +#endif + +#ifndef STBI_NO_STDIO +unsigned char *stbi_load(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = fopen(filename, "rb"); + unsigned char *result; + if (!f) return epuc("can't fopen", "Unable to open file"); + result = stbi_load_from_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} + +unsigned char *stbi_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + int i; + if (stbi_jpeg_test_file(f)) + return stbi_jpeg_load_from_file(f,x,y,comp,req_comp); + if (stbi_png_test_file(f)) + return stbi_png_load_from_file(f,x,y,comp,req_comp); + if (stbi_bmp_test_file(f)) + return stbi_bmp_load_from_file(f,x,y,comp,req_comp); + if (stbi_psd_test_file(f)) + return stbi_psd_load_from_file(f,x,y,comp,req_comp); + #ifndef STBI_NO_HDR + if (stbi_hdr_test_file(f)) { + float *hdr = stbi_hdr_load_from_file(f, x,y,comp,req_comp); + return hdr_to_ldr(hdr, *x, *y, req_comp ? req_comp : *comp); + } + #endif + for (i=0; i < max_loaders; ++i) + if (loaders[i]->test_file(f)) + return loaders[i]->load_from_file(f,x,y,comp,req_comp); + // test tga last because it's a crappy test! + if (stbi_tga_test_file(f)) + return stbi_tga_load_from_file(f,x,y,comp,req_comp); + return epuc("unknown image type", "Image not of any known type, or corrupt"); +} +#endif + +unsigned char *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + int i; + if (stbi_jpeg_test_memory(buffer,len)) + return stbi_jpeg_load_from_memory(buffer,len,x,y,comp,req_comp); + if (stbi_png_test_memory(buffer,len)) + return stbi_png_load_from_memory(buffer,len,x,y,comp,req_comp); + if (stbi_bmp_test_memory(buffer,len)) + return stbi_bmp_load_from_memory(buffer,len,x,y,comp,req_comp); + if (stbi_psd_test_memory(buffer,len)) + return stbi_psd_load_from_memory(buffer,len,x,y,comp,req_comp); + #ifndef STBI_NO_HDR + if (stbi_hdr_test_memory(buffer, len)) { + float *hdr = stbi_hdr_load_from_memory(buffer, len,x,y,comp,req_comp); + return hdr_to_ldr(hdr, *x, *y, req_comp ? req_comp : *comp); + } + #endif + for (i=0; i < max_loaders; ++i) + if (loaders[i]->test_memory(buffer,len)) + return loaders[i]->load_from_memory(buffer,len,x,y,comp,req_comp); + // test tga last because it's a crappy test! + if (stbi_tga_test_memory(buffer,len)) + return stbi_tga_load_from_memory(buffer,len,x,y,comp,req_comp); + return epuc("unknown image type", "Image not of any known type, or corrupt"); +} + +#ifndef STBI_NO_HDR + +#ifndef STBI_NO_STDIO +float *stbi_loadf(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = fopen(filename, "rb"); + float *result; + if (!f) return epf("can't fopen", "Unable to open file"); + result = stbi_loadf_from_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} + +float *stbi_loadf_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *data; + #ifndef STBI_NO_HDR + if (stbi_hdr_test_file(f)) + return stbi_hdr_load_from_file(f,x,y,comp,req_comp); + #endif + data = stbi_load_from_file(f, x, y, comp, req_comp); + if (data) + return ldr_to_hdr(data, *x, *y, req_comp ? req_comp : *comp); + return epf("unknown image type", "Image not of any known type, or corrupt"); +} +#endif + +float *stbi_loadf_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi_uc *data; + #ifndef STBI_NO_HDR + if (stbi_hdr_test_memory(buffer, len)) + return stbi_hdr_load_from_memory(buffer, len,x,y,comp,req_comp); + #endif + data = stbi_load_from_memory(buffer, len, x, y, comp, req_comp); + if (data) + return ldr_to_hdr(data, *x, *y, req_comp ? req_comp : *comp); + return epf("unknown image type", "Image not of any known type, or corrupt"); +} +#endif + +// these is-hdr-or-not is defined independent of whether STBI_NO_HDR is +// defined, for API simplicity; if STBI_NO_HDR is defined, it always +// reports false! + +int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len) +{ + #ifndef STBI_NO_HDR + return stbi_hdr_test_memory(buffer, len); + #else + return 0; + #endif +} + +#ifndef STBI_NO_STDIO +extern int stbi_is_hdr (char const *filename) +{ + FILE *f = fopen(filename, "rb"); + int result=0; + if (f) { + result = stbi_is_hdr_from_file(f); + fclose(f); + } + return result; +} + +extern int stbi_is_hdr_from_file(FILE *f) +{ + #ifndef STBI_NO_HDR + return stbi_hdr_test_file(f); + #else + return 0; + #endif +} + +#endif + +// @TODO: get image dimensions & components without fully decoding +#ifndef STBI_NO_STDIO +extern int stbi_info (char const *filename, int *x, int *y, int *comp); +extern int stbi_info_from_file (FILE *f, int *x, int *y, int *comp); +#endif +extern int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp); + +#ifndef STBI_NO_HDR +static float h2l_gamma_i=1.0f/2.2f, h2l_scale_i=1.0f; +static float l2h_gamma=2.2f, l2h_scale=1.0f; + +void stbi_hdr_to_ldr_gamma(float gamma) { h2l_gamma_i = 1/gamma; } +void stbi_hdr_to_ldr_scale(float scale) { h2l_scale_i = 1/scale; } + +void stbi_ldr_to_hdr_gamma(float gamma) { l2h_gamma = gamma; } +void stbi_ldr_to_hdr_scale(float scale) { l2h_scale = scale; } +#endif + + +////////////////////////////////////////////////////////////////////////////// +// +// Common code used by all image loaders +// + +enum +{ + SCAN_load=0, + SCAN_type, + SCAN_header, +}; + +typedef struct +{ + uint32 img_x, img_y; + int img_n, img_out_n; + + #ifndef STBI_NO_STDIO + FILE *img_file; + #endif + uint8 *img_buffer, *img_buffer_end; +} stbi; + +#ifndef STBI_NO_STDIO +static void start_file(stbi *s, FILE *f) +{ + s->img_file = f; +} +#endif + +static void start_mem(stbi *s, uint8 const *buffer, int len) +{ +#ifndef STBI_NO_STDIO + s->img_file = NULL; +#endif + s->img_buffer = (uint8 *) buffer; + s->img_buffer_end = (uint8 *) buffer+len; +} + +__forceinline static int get8(stbi *s) +{ +#ifndef STBI_NO_STDIO + if (s->img_file) { + int c = fgetc(s->img_file); + return c == EOF ? 0 : c; + } +#endif + if (s->img_buffer < s->img_buffer_end) + return *s->img_buffer++; + return 0; +} + +__forceinline static int at_eof(stbi *s) +{ +#ifndef STBI_NO_STDIO + if (s->img_file) + return feof(s->img_file); +#endif + return s->img_buffer >= s->img_buffer_end; +} + +__forceinline static uint8 get8u(stbi *s) +{ + return (uint8) get8(s); +} + +static void skip(stbi *s, int n) +{ +#ifndef STBI_NO_STDIO + if (s->img_file) + fseek(s->img_file, n, SEEK_CUR); + else +#endif + s->img_buffer += n; +} + +static int get16(stbi *s) +{ + int z = get8(s); + return (z << 8) + get8(s); +} + +static uint32 get32(stbi *s) +{ + uint32 z = get16(s); + return (z << 16) + get16(s); +} + +static int get16le(stbi *s) +{ + int z = get8(s); + return z + (get8(s) << 8); +} + +static uint32 get32le(stbi *s) +{ + uint32 z = get16le(s); + return z + (get16le(s) << 16); +} + +static void getn(stbi *s, stbi_uc *buffer, int n) +{ +#ifndef STBI_NO_STDIO + if (s->img_file) { + fread(buffer, 1, n, s->img_file); + return; + } +#endif + memcpy(buffer, s->img_buffer, n); + s->img_buffer += n; +} + +////////////////////////////////////////////////////////////////////////////// +// +// generic converter from built-in img_n to req_comp +// individual types do this automatically as much as possible (e.g. jpeg +// does all cases internally since it needs to colorspace convert anyway, +// and it never has alpha, so very few cases ). png can automatically +// interleave an alpha=255 channel, but falls back to this for other cases +// +// assume data buffer is malloced, so malloc a new one and free that one +// only failure mode is malloc failing + +static uint8 compute_y(int r, int g, int b) +{ + return (uint8) (((r*77) + (g*150) + (29*b)) >> 8); +} + +static unsigned char *convert_format(unsigned char *data, int img_n, int req_comp, uint x, uint y) +{ + int i,j; + unsigned char *good; + + if (req_comp == img_n) return data; + assert(req_comp >= 1 && req_comp <= 4); + + good = (unsigned char *) malloc(req_comp * x * y); + if (good == NULL) { + free(data); + return epuc("outofmem", "Out of memory"); + } + + for (j=0; j < (int) y; ++j) { + unsigned char *src = data + j * x * img_n ; + unsigned char *dest = good + j * x * req_comp; + + #define COMBO(a,b) ((a)*8+(b)) + #define CASE(a,b) case COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b) + // convert source image with img_n components to one with req_comp components; + // avoid switch per pixel, so use switch per scanline and massive macros + switch(COMBO(img_n, req_comp)) { + CASE(1,2) dest[0]=src[0], dest[1]=255; break; + CASE(1,3) dest[0]=dest[1]=dest[2]=src[0]; break; + CASE(1,4) dest[0]=dest[1]=dest[2]=src[0], dest[3]=255; break; + CASE(2,1) dest[0]=src[0]; break; + CASE(2,3) dest[0]=dest[1]=dest[2]=src[0]; break; + CASE(2,4) dest[0]=dest[1]=dest[2]=src[0], dest[3]=src[1]; break; + CASE(3,4) dest[0]=src[0],dest[1]=src[1],dest[2]=src[2],dest[3]=255; break; + CASE(3,1) dest[0]=compute_y(src[0],src[1],src[2]); break; + CASE(3,2) dest[0]=compute_y(src[0],src[1],src[2]), dest[1] = 255; break; + CASE(4,1) dest[0]=compute_y(src[0],src[1],src[2]); break; + CASE(4,2) dest[0]=compute_y(src[0],src[1],src[2]), dest[1] = src[3]; break; + CASE(4,3) dest[0]=src[0],dest[1]=src[1],dest[2]=src[2]; break; + default: assert(0); + } + #undef CASE + } + + free(data); + return good; +} + +#ifndef STBI_NO_HDR +static float *ldr_to_hdr(stbi_uc *data, int x, int y, int comp) +{ + int i,k,n; + float *output = (float *) malloc(x * y * comp * sizeof(float)); + if (output == NULL) { free(data); return epf("outofmem", "Out of memory"); } + // compute number of non-alpha components + if (comp & 1) n = comp; else n = comp-1; + for (i=0; i < x*y; ++i) { + for (k=0; k < n; ++k) { + output[i*comp + k] = (float) pow(data[i*comp+k]/255.0f, l2h_gamma) * l2h_scale; + } + if (k < comp) output[i*comp + k] = data[i*comp+k]/255.0f; + } + free(data); + return output; +} + +#define float2int(x) ((int) (x)) +static stbi_uc *hdr_to_ldr(float *data, int x, int y, int comp) +{ + int i,k,n; + stbi_uc *output = (stbi_uc *) malloc(x * y * comp); + if (output == NULL) { free(data); return epuc("outofmem", "Out of memory"); } + // compute number of non-alpha components + if (comp & 1) n = comp; else n = comp-1; + for (i=0; i < x*y; ++i) { + for (k=0; k < n; ++k) { + float z = (float) pow(data[i*comp+k]*h2l_scale_i, h2l_gamma_i) * 255 + 0.5f; + if (z < 0) z = 0; + if (z > 255) z = 255; + output[i*comp + k] = float2int(z); + } + if (k < comp) { + float z = data[i*comp+k] * 255 + 0.5f; + if (z < 0) z = 0; + if (z > 255) z = 255; + output[i*comp + k] = float2int(z); + } + } + free(data); + return output; +} +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// "baseline" JPEG/JFIF decoder (not actually fully baseline implementation) +// +// simple implementation +// - channel subsampling of at most 2 in each dimension +// - doesn't support delayed output of y-dimension +// - simple interface (only one output format: 8-bit interleaved RGB) +// - doesn't try to recover corrupt jpegs +// - doesn't allow partial loading, loading multiple at once +// - still fast on x86 (copying globals into locals doesn't help x86) +// - allocates lots of intermediate memory (full size of all components) +// - non-interleaved case requires this anyway +// - allows good upsampling (see next) +// high-quality +// - upsampled channels are bilinearly interpolated, even across blocks +// - quality integer IDCT derived from IJG's 'slow' +// performance +// - fast huffman; reasonable integer IDCT +// - uses a lot of intermediate memory, could cache poorly +// - load http://nothings.org/remote/anemones.jpg 3 times on 2.8Ghz P4 +// stb_jpeg: 1.34 seconds (MSVC6, default release build) +// stb_jpeg: 1.06 seconds (MSVC6, processor = Pentium Pro) +// IJL11.dll: 1.08 seconds (compiled by intel) +// IJG 1998: 0.98 seconds (MSVC6, makefile provided by IJG) +// IJG 1998: 0.95 seconds (MSVC6, makefile + proc=PPro) + +// huffman decoding acceleration +#define FAST_BITS 9 // larger handles more cases; smaller stomps less cache + +typedef struct +{ + uint8 fast[1 << FAST_BITS]; + // weirdly, repacking this into AoS is a 10% speed loss, instead of a win + uint16 code[256]; + uint8 values[256]; + uint8 size[257]; + unsigned int maxcode[18]; + int delta[17]; // old 'firstsymbol' - old 'firstcode' +} huffman; + +typedef struct +{ + #if STBI_SIMD + unsigned short dequant2[4][64]; + #endif + stbi s; + huffman huff_dc[4]; + huffman huff_ac[4]; + uint8 dequant[4][64]; + +// sizes for components, interleaved MCUs + int img_h_max, img_v_max; + int img_mcu_x, img_mcu_y; + int img_mcu_w, img_mcu_h; + +// definition of jpeg image component + struct + { + int id; + int h,v; + int tq; + int hd,ha; + int dc_pred; + + int x,y,w2,h2; + uint8 *data; + void *raw_data; + uint8 *linebuf; + } img_comp[4]; + + uint32 code_buffer; // jpeg entropy-coded buffer + int code_bits; // number of valid bits + unsigned char marker; // marker seen while filling entropy buffer + int nomore; // flag if we saw a marker so must stop + + int scan_n, order[4]; + int restart_interval, todo; +} jpeg; + +static int build_huffman(huffman *h, int *count) +{ + int i,j,k=0,code; + // build size list for each symbol (from JPEG spec) + for (i=0; i < 16; ++i) + for (j=0; j < count[i]; ++j) + h->size[k++] = (uint8) (i+1); + h->size[k] = 0; + + // compute actual symbols (from jpeg spec) + code = 0; + k = 0; + for(j=1; j <= 16; ++j) { + // compute delta to add to code to compute symbol id + h->delta[j] = k - code; + if (h->size[k] == j) { + while (h->size[k] == j) + h->code[k++] = (uint16) (code++); + if (code-1 >= (1 << j)) return e("bad code lengths","Corrupt JPEG"); + } + // compute largest code + 1 for this size, preshifted as needed later + h->maxcode[j] = code << (16-j); + code <<= 1; + } + h->maxcode[j] = 0xffffffff; + + // build non-spec acceleration table; 255 is flag for not-accelerated + memset(h->fast, 255, 1 << FAST_BITS); + for (i=0; i < k; ++i) { + int s = h->size[i]; + if (s <= FAST_BITS) { + int c = h->code[i] << (FAST_BITS-s); + int m = 1 << (FAST_BITS-s); + for (j=0; j < m; ++j) { + h->fast[c+j] = (uint8) i; + } + } + } + return 1; +} + +static void grow_buffer_unsafe(jpeg *j) +{ + do { + int b = j->nomore ? 0 : get8(&j->s); + if (b == 0xff) { + int c = get8(&j->s); + if (c != 0) { + j->marker = (unsigned char) c; + j->nomore = 1; + return; + } + } + j->code_buffer = (j->code_buffer << 8) | b; + j->code_bits += 8; + } while (j->code_bits <= 24); +} + +// (1 << n) - 1 +static uint32 bmask[17]={0,1,3,7,15,31,63,127,255,511,1023,2047,4095,8191,16383,32767,65535}; + +// decode a jpeg huffman value from the bitstream +__forceinline static int decode(jpeg *j, huffman *h) +{ + unsigned int temp; + int c,k; + + if (j->code_bits < 16) grow_buffer_unsafe(j); + + // look at the top FAST_BITS and determine what symbol ID it is, + // if the code is <= FAST_BITS + c = (j->code_buffer >> (j->code_bits - FAST_BITS)) & ((1 << FAST_BITS)-1); + k = h->fast[c]; + if (k < 255) { + if (h->size[k] > j->code_bits) + return -1; + j->code_bits -= h->size[k]; + return h->values[k]; + } + + // naive test is to shift the code_buffer down so k bits are + // valid, then test against maxcode. To speed this up, we've + // preshifted maxcode left so that it has (16-k) 0s at the + // end; in other words, regardless of the number of bits, it + // wants to be compared against something shifted to have 16; + // that way we don't need to shift inside the loop. + if (j->code_bits < 16) + temp = (j->code_buffer << (16 - j->code_bits)) & 0xffff; + else + temp = (j->code_buffer >> (j->code_bits - 16)) & 0xffff; + for (k=FAST_BITS+1 ; ; ++k) + if (temp < h->maxcode[k]) + break; + if (k == 17) { + // error! code not found + j->code_bits -= 16; + return -1; + } + + if (k > j->code_bits) + return -1; + + // convert the huffman code to the symbol id + c = ((j->code_buffer >> (j->code_bits - k)) & bmask[k]) + h->delta[k]; + assert((((j->code_buffer) >> (j->code_bits - h->size[c])) & bmask[h->size[c]]) == h->code[c]); + + // convert the id to a symbol + j->code_bits -= k; + return h->values[c]; +} + +// combined JPEG 'receive' and JPEG 'extend', since baseline +// always extends everything it receives. +__forceinline static int extend_receive(jpeg *j, int n) +{ + unsigned int m = 1 << (n-1); + unsigned int k; + if (j->code_bits < n) grow_buffer_unsafe(j); + k = (j->code_buffer >> (j->code_bits - n)) & bmask[n]; + j->code_bits -= n; + // the following test is probably a random branch that won't + // predict well. I tried to table accelerate it but failed. + // maybe it's compiling as a conditional move? + if (k < m) + return (-1 << n) + k + 1; + else + return k; +} + +// given a value that's at position X in the zigzag stream, +// where does it appear in the 8x8 matrix coded as row-major? +static uint8 dezigzag[64+15] = +{ + 0, 1, 8, 16, 9, 2, 3, 10, + 17, 24, 32, 25, 18, 11, 4, 5, + 12, 19, 26, 33, 40, 48, 41, 34, + 27, 20, 13, 6, 7, 14, 21, 28, + 35, 42, 49, 56, 57, 50, 43, 36, + 29, 22, 15, 23, 30, 37, 44, 51, + 58, 59, 52, 45, 38, 31, 39, 46, + 53, 60, 61, 54, 47, 55, 62, 63, + // let corrupt input sample past end + 63, 63, 63, 63, 63, 63, 63, 63, + 63, 63, 63, 63, 63, 63, 63 +}; + +// decode one 64-entry block-- +static int decode_block(jpeg *j, short data[64], huffman *hdc, huffman *hac, int b) +{ + int diff,dc,k; + int t = decode(j, hdc); + if (t < 0) return e("bad huffman code","Corrupt JPEG"); + + // 0 all the ac values now so we can do it 32-bits at a time + memset(data,0,64*sizeof(data[0])); + + diff = t ? extend_receive(j, t) : 0; + dc = j->img_comp[b].dc_pred + diff; + j->img_comp[b].dc_pred = dc; + data[0] = (short) dc; + + // decode AC components, see JPEG spec + k = 1; + do { + int r,s; + int rs = decode(j, hac); + if (rs < 0) return e("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (rs != 0xf0) break; // end block + k += 16; + } else { + k += r; + // decode into unzigzag'd location + data[dezigzag[k++]] = (short) extend_receive(j,s); + } + } while (k < 64); + return 1; +} + +// take a -128..127 value and clamp it and convert to 0..255 +__forceinline static uint8 clamp(int x) +{ + x += 128; + // trick to use a single test to catch both cases + if ((unsigned int) x > 255) { + if (x < 0) return 0; + if (x > 255) return 255; + } + return (uint8) x; +} + +#define f2f(x) (int) (((x) * 4096 + 0.5)) +#define fsh(x) ((x) << 12) + +// derived from jidctint -- DCT_ISLOW +#define IDCT_1D(s0,s1,s2,s3,s4,s5,s6,s7) \ + int t0,t1,t2,t3,p1,p2,p3,p4,p5,x0,x1,x2,x3; \ + p2 = s2; \ + p3 = s6; \ + p1 = (p2+p3) * f2f(0.5411961f); \ + t2 = p1 + p3*f2f(-1.847759065f); \ + t3 = p1 + p2*f2f( 0.765366865f); \ + p2 = s0; \ + p3 = s4; \ + t0 = fsh(p2+p3); \ + t1 = fsh(p2-p3); \ + x0 = t0+t3; \ + x3 = t0-t3; \ + x1 = t1+t2; \ + x2 = t1-t2; \ + t0 = s7; \ + t1 = s5; \ + t2 = s3; \ + t3 = s1; \ + p3 = t0+t2; \ + p4 = t1+t3; \ + p1 = t0+t3; \ + p2 = t1+t2; \ + p5 = (p3+p4)*f2f( 1.175875602f); \ + t0 = t0*f2f( 0.298631336f); \ + t1 = t1*f2f( 2.053119869f); \ + t2 = t2*f2f( 3.072711026f); \ + t3 = t3*f2f( 1.501321110f); \ + p1 = p5 + p1*f2f(-0.899976223f); \ + p2 = p5 + p2*f2f(-2.562915447f); \ + p3 = p3*f2f(-1.961570560f); \ + p4 = p4*f2f(-0.390180644f); \ + t3 += p1+p4; \ + t2 += p2+p3; \ + t1 += p2+p4; \ + t0 += p1+p3; + +#if !STBI_SIMD +// .344 seconds on 3*anemones.jpg +static void idct_block(uint8 *out, int out_stride, short data[64], uint8 *dequantize) +{ + int i,val[64],*v=val; + uint8 *o,*dq = dequantize; + short *d = data; + + // columns + for (i=0; i < 8; ++i,++d,++dq, ++v) { + // if all zeroes, shortcut -- this avoids dequantizing 0s and IDCTing + if (d[ 8]==0 && d[16]==0 && d[24]==0 && d[32]==0 + && d[40]==0 && d[48]==0 && d[56]==0) { + // no shortcut 0 seconds + // (1|2|3|4|5|6|7)==0 0 seconds + // all separate -0.047 seconds + // 1 && 2|3 && 4|5 && 6|7: -0.047 seconds + int dcterm = d[0] * dq[0] << 2; + v[0] = v[8] = v[16] = v[24] = v[32] = v[40] = v[48] = v[56] = dcterm; + } else { + IDCT_1D(d[ 0]*dq[ 0],d[ 8]*dq[ 8],d[16]*dq[16],d[24]*dq[24], + d[32]*dq[32],d[40]*dq[40],d[48]*dq[48],d[56]*dq[56]) + // constants scaled things up by 1<<12; let's bring them back + // down, but keep 2 extra bits of precision + x0 += 512; x1 += 512; x2 += 512; x3 += 512; + v[ 0] = (x0+t3) >> 10; + v[56] = (x0-t3) >> 10; + v[ 8] = (x1+t2) >> 10; + v[48] = (x1-t2) >> 10; + v[16] = (x2+t1) >> 10; + v[40] = (x2-t1) >> 10; + v[24] = (x3+t0) >> 10; + v[32] = (x3-t0) >> 10; + } + } + + for (i=0, v=val, o=out; i < 8; ++i,v+=8,o+=out_stride) { + // no fast case since the first 1D IDCT spread components out + IDCT_1D(v[0],v[1],v[2],v[3],v[4],v[5],v[6],v[7]) + // constants scaled things up by 1<<12, plus we had 1<<2 from first + // loop, plus horizontal and vertical each scale by sqrt(8) so together + // we've got an extra 1<<3, so 1<<17 total we need to remove. + x0 += 65536; x1 += 65536; x2 += 65536; x3 += 65536; + o[0] = clamp((x0+t3) >> 17); + o[7] = clamp((x0-t3) >> 17); + o[1] = clamp((x1+t2) >> 17); + o[6] = clamp((x1-t2) >> 17); + o[2] = clamp((x2+t1) >> 17); + o[5] = clamp((x2-t1) >> 17); + o[3] = clamp((x3+t0) >> 17); + o[4] = clamp((x3-t0) >> 17); + } +} +#else +static void idct_block(uint8 *out, int out_stride, short data[64], unsigned short *dequantize) +{ + int i,val[64],*v=val; + uint8 *o; + unsigned short *dq = dequantize; + short *d = data; + + // columns + for (i=0; i < 8; ++i,++d,++dq, ++v) { + // if all zeroes, shortcut -- this avoids dequantizing 0s and IDCTing + if (d[ 8]==0 && d[16]==0 && d[24]==0 && d[32]==0 + && d[40]==0 && d[48]==0 && d[56]==0) { + // no shortcut 0 seconds + // (1|2|3|4|5|6|7)==0 0 seconds + // all separate -0.047 seconds + // 1 && 2|3 && 4|5 && 6|7: -0.047 seconds + int dcterm = d[0] * dq[0] << 2; + v[0] = v[8] = v[16] = v[24] = v[32] = v[40] = v[48] = v[56] = dcterm; + } else { + IDCT_1D(d[ 0]*dq[ 0],d[ 8]*dq[ 8],d[16]*dq[16],d[24]*dq[24], + d[32]*dq[32],d[40]*dq[40],d[48]*dq[48],d[56]*dq[56]) + // constants scaled things up by 1<<12; let's bring them back + // down, but keep 2 extra bits of precision + x0 += 512; x1 += 512; x2 += 512; x3 += 512; + v[ 0] = (x0+t3) >> 10; + v[56] = (x0-t3) >> 10; + v[ 8] = (x1+t2) >> 10; + v[48] = (x1-t2) >> 10; + v[16] = (x2+t1) >> 10; + v[40] = (x2-t1) >> 10; + v[24] = (x3+t0) >> 10; + v[32] = (x3-t0) >> 10; + } + } + + for (i=0, v=val, o=out; i < 8; ++i,v+=8,o+=out_stride) { + // no fast case since the first 1D IDCT spread components out + IDCT_1D(v[0],v[1],v[2],v[3],v[4],v[5],v[6],v[7]) + // constants scaled things up by 1<<12, plus we had 1<<2 from first + // loop, plus horizontal and vertical each scale by sqrt(8) so together + // we've got an extra 1<<3, so 1<<17 total we need to remove. + x0 += 65536; x1 += 65536; x2 += 65536; x3 += 65536; + o[0] = clamp((x0+t3) >> 17); + o[7] = clamp((x0-t3) >> 17); + o[1] = clamp((x1+t2) >> 17); + o[6] = clamp((x1-t2) >> 17); + o[2] = clamp((x2+t1) >> 17); + o[5] = clamp((x2-t1) >> 17); + o[3] = clamp((x3+t0) >> 17); + o[4] = clamp((x3-t0) >> 17); + } +} +static stbi_idct_8x8 stbi_idct_installed = idct_block; + +extern void stbi_install_idct(stbi_idct_8x8 func) +{ + stbi_idct_installed = func; +} +#endif + +#define MARKER_none 0xff +// if there's a pending marker from the entropy stream, return that +// otherwise, fetch from the stream and get a marker. if there's no +// marker, return 0xff, which is never a valid marker value +static uint8 get_marker(jpeg *j) +{ + uint8 x; + if (j->marker != MARKER_none) { x = j->marker; j->marker = MARKER_none; return x; } + x = get8u(&j->s); + if (x != 0xff) return MARKER_none; + while (x == 0xff) + x = get8u(&j->s); + return x; +} + +// in each scan, we'll have scan_n components, and the order +// of the components is specified by order[] +#define RESTART(x) ((x) >= 0xd0 && (x) <= 0xd7) + +// after a restart interval, reset the entropy decoder and +// the dc prediction +static void reset(jpeg *j) +{ + j->code_bits = 0; + j->code_buffer = 0; + j->nomore = 0; + j->img_comp[0].dc_pred = j->img_comp[1].dc_pred = j->img_comp[2].dc_pred = 0; + j->marker = MARKER_none; + j->todo = j->restart_interval ? j->restart_interval : 0x7fffffff; + // no more than 1<<31 MCUs if no restart_interal? that's plenty safe, + // since we don't even allow 1<<30 pixels +} + +static int parse_entropy_coded_data(jpeg *z) +{ + reset(z); + if (z->scan_n == 1) { + int i,j; + #if STBI_SIMD + __declspec(align(16)) + #endif + short data[64]; + int n = z->order[0]; + // non-interleaved data, we just need to process one block at a time, + // in trivial scanline order + // number of blocks to do just depends on how many actual "pixels" this + // component has, independent of interleaved MCU blocking and such + int w = (z->img_comp[n].x+7) >> 3; + int h = (z->img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + if (!decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+z->img_comp[n].ha, n)) return 0; + #if STBI_SIMD + stbi_idct_installed(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data, z->dequant2[z->img_comp[n].tq]); + #else + idct_block(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data, z->dequant[z->img_comp[n].tq]); + #endif + // every data block is an MCU, so countdown the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) grow_buffer_unsafe(z); + // if it's NOT a restart, then just bail, so we get corrupt data + // rather than no data + if (!RESTART(z->marker)) return 1; + reset(z); + } + } + } + } else { // interleaved! + int i,j,k,x,y; + short data[64]; + for (j=0; j < z->img_mcu_y; ++j) { + for (i=0; i < z->img_mcu_x; ++i) { + // scan an interleaved mcu... process scan_n components in order + for (k=0; k < z->scan_n; ++k) { + int n = z->order[k]; + // scan out an mcu's worth of this component; that's just determined + // by the basic H and V specified for the component + for (y=0; y < z->img_comp[n].v; ++y) { + for (x=0; x < z->img_comp[n].h; ++x) { + int x2 = (i*z->img_comp[n].h + x)*8; + int y2 = (j*z->img_comp[n].v + y)*8; + if (!decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+z->img_comp[n].ha, n)) return 0; + #if STBI_SIMD + stbi_idct_installed(z->img_comp[n].data+z->img_comp[n].w2*y2+x2, z->img_comp[n].w2, data, z->dequant2[z->img_comp[n].tq]); + #else + idct_block(z->img_comp[n].data+z->img_comp[n].w2*y2+x2, z->img_comp[n].w2, data, z->dequant[z->img_comp[n].tq]); + #endif + } + } + } + // after all interleaved components, that's an interleaved MCU, + // so now count down the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) grow_buffer_unsafe(z); + // if it's NOT a restart, then just bail, so we get corrupt data + // rather than no data + if (!RESTART(z->marker)) return 1; + reset(z); + } + } + } + } + return 1; +} + +static int process_marker(jpeg *z, int m) +{ + int L; + switch (m) { + case MARKER_none: // no marker found + return e("expected marker","Corrupt JPEG"); + + case 0xC2: // SOF - progressive + return e("progressive jpeg","JPEG format not supported (progressive)"); + + case 0xDD: // DRI - specify restart interval + if (get16(&z->s) != 4) return e("bad DRI len","Corrupt JPEG"); + z->restart_interval = get16(&z->s); + return 1; + + case 0xDB: // DQT - define quantization table + L = get16(&z->s)-2; + while (L > 0) { + int q = get8(&z->s); + int p = q >> 4; + int t = q & 15,i; + if (p != 0) return e("bad DQT type","Corrupt JPEG"); + if (t > 3) return e("bad DQT table","Corrupt JPEG"); + for (i=0; i < 64; ++i) + z->dequant[t][dezigzag[i]] = get8u(&z->s); + #if STBI_SIMD + for (i=0; i < 64; ++i) + z->dequant2[t][i] = dequant[t][i]; + #endif + L -= 65; + } + return L==0; + + case 0xC4: // DHT - define huffman table + L = get16(&z->s)-2; + while (L > 0) { + uint8 *v; + int sizes[16],i,m=0; + int q = get8(&z->s); + int tc = q >> 4; + int th = q & 15; + if (tc > 1 || th > 3) return e("bad DHT header","Corrupt JPEG"); + for (i=0; i < 16; ++i) { + sizes[i] = get8(&z->s); + m += sizes[i]; + } + L -= 17; + if (tc == 0) { + if (!build_huffman(z->huff_dc+th, sizes)) return 0; + v = z->huff_dc[th].values; + } else { + if (!build_huffman(z->huff_ac+th, sizes)) return 0; + v = z->huff_ac[th].values; + } + for (i=0; i < m; ++i) + v[i] = get8u(&z->s); + L -= m; + } + return L==0; + } + // check for comment block or APP blocks + if ((m >= 0xE0 && m <= 0xEF) || m == 0xFE) { + skip(&z->s, get16(&z->s)-2); + return 1; + } + return 0; +} + +// after we see SOS +static int process_scan_header(jpeg *z) +{ + int i; + int Ls = get16(&z->s); + z->scan_n = get8(&z->s); + if (z->scan_n < 1 || z->scan_n > 4 || z->scan_n > (int) z->s.img_n) return e("bad SOS component count","Corrupt JPEG"); + if (Ls != 6+2*z->scan_n) return e("bad SOS len","Corrupt JPEG"); + for (i=0; i < z->scan_n; ++i) { + int id = get8(&z->s), which; + int q = get8(&z->s); + for (which = 0; which < z->s.img_n; ++which) + if (z->img_comp[which].id == id) + break; + if (which == z->s.img_n) return 0; + z->img_comp[which].hd = q >> 4; if (z->img_comp[which].hd > 3) return e("bad DC huff","Corrupt JPEG"); + z->img_comp[which].ha = q & 15; if (z->img_comp[which].ha > 3) return e("bad AC huff","Corrupt JPEG"); + z->order[i] = which; + } + if (get8(&z->s) != 0) return e("bad SOS","Corrupt JPEG"); + get8(&z->s); // should be 63, but might be 0 + if (get8(&z->s) != 0) return e("bad SOS","Corrupt JPEG"); + + return 1; +} + +static int process_frame_header(jpeg *z, int scan) +{ + stbi *s = &z->s; + int Lf,p,i,q, h_max=1,v_max=1,c; + Lf = get16(s); if (Lf < 11) return e("bad SOF len","Corrupt JPEG"); // JPEG + p = get8(s); if (p != 8) return e("only 8-bit","JPEG format not supported: 8-bit only"); // JPEG baseline + s->img_y = get16(s); if (s->img_y == 0) return e("no header height", "JPEG format not supported: delayed height"); // Legal, but we don't handle it--but neither does IJG + s->img_x = get16(s); if (s->img_x == 0) return e("0 width","Corrupt JPEG"); // JPEG requires + c = get8(s); + if (c != 3 && c != 1) return e("bad component count","Corrupt JPEG"); // JFIF requires + s->img_n = c; + for (i=0; i < c; ++i) { + z->img_comp[i].data = NULL; + z->img_comp[i].linebuf = NULL; + } + + if (Lf != 8+3*s->img_n) return e("bad SOF len","Corrupt JPEG"); + + for (i=0; i < s->img_n; ++i) { + z->img_comp[i].id = get8(s); + if (z->img_comp[i].id != i+1) // JFIF requires + if (z->img_comp[i].id != i) // some version of jpegtran outputs non-JFIF-compliant files! + return e("bad component ID","Corrupt JPEG"); + q = get8(s); + z->img_comp[i].h = (q >> 4); if (!z->img_comp[i].h || z->img_comp[i].h > 4) return e("bad H","Corrupt JPEG"); + z->img_comp[i].v = q & 15; if (!z->img_comp[i].v || z->img_comp[i].v > 4) return e("bad V","Corrupt JPEG"); + z->img_comp[i].tq = get8(s); if (z->img_comp[i].tq > 3) return e("bad TQ","Corrupt JPEG"); + } + + if (scan != SCAN_load) return 1; + + if ((1 << 30) / s->img_x / s->img_n < s->img_y) return e("too large", "Image too large to decode"); + + for (i=0; i < s->img_n; ++i) { + if (z->img_comp[i].h > h_max) h_max = z->img_comp[i].h; + if (z->img_comp[i].v > v_max) v_max = z->img_comp[i].v; + } + + // compute interleaved mcu info + z->img_h_max = h_max; + z->img_v_max = v_max; + z->img_mcu_w = h_max * 8; + z->img_mcu_h = v_max * 8; + z->img_mcu_x = (s->img_x + z->img_mcu_w-1) / z->img_mcu_w; + z->img_mcu_y = (s->img_y + z->img_mcu_h-1) / z->img_mcu_h; + + for (i=0; i < s->img_n; ++i) { + // number of effective pixels (e.g. for non-interleaved MCU) + z->img_comp[i].x = (s->img_x * z->img_comp[i].h + h_max-1) / h_max; + z->img_comp[i].y = (s->img_y * z->img_comp[i].v + v_max-1) / v_max; + // to simplify generation, we'll allocate enough memory to decode + // the bogus oversized data from using interleaved MCUs and their + // big blocks (e.g. a 16x16 iMCU on an image of width 33); we won't + // discard the extra data until colorspace conversion + z->img_comp[i].w2 = z->img_mcu_x * z->img_comp[i].h * 8; + z->img_comp[i].h2 = z->img_mcu_y * z->img_comp[i].v * 8; + z->img_comp[i].raw_data = malloc(z->img_comp[i].w2 * z->img_comp[i].h2+15); + if (z->img_comp[i].raw_data == NULL) { + for(--i; i >= 0; --i) { + free(z->img_comp[i].raw_data); + z->img_comp[i].data = NULL; + } + return e("outofmem", "Out of memory"); + } + // align blocks for installable-idct using mmx/sse + z->img_comp[i].data = (uint8*) (((size_t) z->img_comp[i].raw_data + 15) & ~15); + z->img_comp[i].linebuf = NULL; + } + + return 1; +} + +// use comparisons since in some cases we handle more than one case (e.g. SOF) +#define DNL(x) ((x) == 0xdc) +#define SOI(x) ((x) == 0xd8) +#define EOI(x) ((x) == 0xd9) +#define SOF(x) ((x) == 0xc0 || (x) == 0xc1) +#define SOS(x) ((x) == 0xda) + +static int decode_jpeg_header(jpeg *z, int scan) +{ + int m; + z->marker = MARKER_none; // initialize cached marker to empty + m = get_marker(z); + if (!SOI(m)) return e("no SOI","Corrupt JPEG"); + if (scan == SCAN_type) return 1; + m = get_marker(z); + while (!SOF(m)) { + if (!process_marker(z,m)) return 0; + m = get_marker(z); + while (m == MARKER_none) { + // some files have extra padding after their blocks, so ok, we'll scan + if (at_eof(&z->s)) return e("no SOF", "Corrupt JPEG"); + m = get_marker(z); + } + } + if (!process_frame_header(z, scan)) return 0; + return 1; +} + +static int decode_jpeg_image(jpeg *j) +{ + int m; + j->restart_interval = 0; + if (!decode_jpeg_header(j, SCAN_load)) return 0; + m = get_marker(j); + while (!EOI(m)) { + if (SOS(m)) { + if (!process_scan_header(j)) return 0; + if (!parse_entropy_coded_data(j)) return 0; + } else { + if (!process_marker(j, m)) return 0; + } + m = get_marker(j); + } + return 1; +} + +// static jfif-centered resampling (across block boundaries) + +typedef uint8 *(*resample_row_func)(uint8 *out, uint8 *in0, uint8 *in1, + int w, int hs); + +#define div4(x) ((uint8) ((x) >> 2)) + +static uint8 *resample_row_1(uint8 *out, uint8 *in_near, uint8 *in_far, int w, int hs) +{ + return in_near; +} + +static uint8* resample_row_v_2(uint8 *out, uint8 *in_near, uint8 *in_far, int w, int hs) +{ + // need to generate two samples vertically for every one in input + int i; + for (i=0; i < w; ++i) + out[i] = div4(3*in_near[i] + in_far[i] + 2); + return out; +} + +static uint8* resample_row_h_2(uint8 *out, uint8 *in_near, uint8 *in_far, int w, int hs) +{ + // need to generate two samples horizontally for every one in input + int i; + uint8 *input = in_near; + if (w == 1) { + // if only one sample, can't do any interpolation + out[0] = out[1] = input[0]; + return out; + } + + out[0] = input[0]; + out[1] = div4(input[0]*3 + input[1] + 2); + for (i=1; i < w-1; ++i) { + int n = 3*input[i]+2; + out[i*2+0] = div4(n+input[i-1]); + out[i*2+1] = div4(n+input[i+1]); + } + out[i*2+0] = div4(input[w-2]*3 + input[w-1] + 2); + out[i*2+1] = input[w-1]; + return out; +} + +#define div16(x) ((uint8) ((x) >> 4)) + +static uint8 *resample_row_hv_2(uint8 *out, uint8 *in_near, uint8 *in_far, int w, int hs) +{ + // need to generate 2x2 samples for every one in input + int i,t0,t1; + if (w == 1) { + out[0] = out[1] = div4(3*in_near[0] + in_far[0] + 2); + return out; + } + + t1 = 3*in_near[0] + in_far[0]; + out[0] = div4(t1+2); + for (i=1; i < w; ++i) { + t0 = t1; + t1 = 3*in_near[i]+in_far[i]; + out[i*2-1] = div16(3*t0 + t1 + 8); + out[i*2 ] = div16(3*t1 + t0 + 8); + } + out[w*2-1] = div4(t1+2); + return out; +} + +static uint8 *resample_row_generic(uint8 *out, uint8 *in_near, uint8 *in_far, int w, int hs) +{ + // resample with nearest-neighbor + int i,j; + for (i=0; i < w; ++i) + for (j=0; j < hs; ++j) + out[i*hs+j] = in_near[i]; + return out; +} + +#define float2fixed(x) ((int) ((x) * 65536 + 0.5)) + +// 0.38 seconds on 3*anemones.jpg (0.25 with processor = Pro) +// VC6 without processor=Pro is generating multiple LEAs per multiply! +static void YCbCr_to_RGB_row(uint8 *out, uint8 *y, uint8 *pcb, uint8 *pcr, int count, int step) +{ + int i; + for (i=0; i < count; ++i) { + int y_fixed = (y[i] << 16) + 32768; // rounding + int r,g,b; + int cr = pcr[i] - 128; + int cb = pcb[i] - 128; + r = y_fixed + cr*float2fixed(1.40200f); + g = y_fixed - cr*float2fixed(0.71414f) - cb*float2fixed(0.34414f); + b = y_fixed + cb*float2fixed(1.77200f); + r >>= 16; + g >>= 16; + b >>= 16; + if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; } + if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; } + if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; } + out[0] = (uint8)r; + out[1] = (uint8)g; + out[2] = (uint8)b; + out[3] = 255; + out += step; + } +} + +#if STBI_SIMD +static stbi_YCbCr_to_RGB_run stbi_YCbCr_installed = YCbCr_to_RGB_row; + +void stbi_install_YCbCr_to_RGB(stbi_YCbCr_to_RGB_run func) +{ + stbi_YCbCr_installed = func; +} +#endif + + +// clean up the temporary component buffers +static void cleanup_jpeg(jpeg *j) +{ + int i; + for (i=0; i < j->s.img_n; ++i) { + if (j->img_comp[i].data) { + free(j->img_comp[i].raw_data); + j->img_comp[i].data = NULL; + } + if (j->img_comp[i].linebuf) { + free(j->img_comp[i].linebuf); + j->img_comp[i].linebuf = NULL; + } + } +} + +typedef struct +{ + resample_row_func resample; + uint8 *line0,*line1; + int hs,vs; // expansion factor in each axis + int w_lores; // horizontal pixels pre-expansion + int ystep; // how far through vertical expansion we are + int ypos; // which pre-expansion row we're on +} stbi_resample; + +static uint8 *load_jpeg_image(jpeg *z, int *out_x, int *out_y, int *comp, int req_comp) +{ + int n, decode_n; + // validate req_comp + if (req_comp < 0 || req_comp > 4) return epuc("bad req_comp", "Internal error"); + z->s.img_n = 0; + + // load a jpeg image from whichever source + if (!decode_jpeg_image(z)) { cleanup_jpeg(z); return NULL; } + + // determine actual number of components to generate + n = req_comp ? req_comp : z->s.img_n; + + if (z->s.img_n == 3 && n < 3) + decode_n = 1; + else + decode_n = z->s.img_n; + + // resample and color-convert + { + int k; + uint i,j; + uint8 *output; + uint8 *coutput[4]; + + stbi_resample res_comp[4]; + + for (k=0; k < decode_n; ++k) { + stbi_resample *r = &res_comp[k]; + + // allocate line buffer big enough for upsampling off the edges + // with upsample factor of 4 + z->img_comp[k].linebuf = (uint8 *) malloc(z->s.img_x + 3); + if (!z->img_comp[k].linebuf) { cleanup_jpeg(z); return epuc("outofmem", "Out of memory"); } + + r->hs = z->img_h_max / z->img_comp[k].h; + r->vs = z->img_v_max / z->img_comp[k].v; + r->ystep = r->vs >> 1; + r->w_lores = (z->s.img_x + r->hs-1) / r->hs; + r->ypos = 0; + r->line0 = r->line1 = z->img_comp[k].data; + + if (r->hs == 1 && r->vs == 1) r->resample = resample_row_1; + else if (r->hs == 1 && r->vs == 2) r->resample = resample_row_v_2; + else if (r->hs == 2 && r->vs == 1) r->resample = resample_row_h_2; + else if (r->hs == 2 && r->vs == 2) r->resample = resample_row_hv_2; + else r->resample = resample_row_generic; + } + + // can't error after this so, this is safe + output = (uint8 *) malloc(n * z->s.img_x * z->s.img_y + 1); + if (!output) { cleanup_jpeg(z); return epuc("outofmem", "Out of memory"); } + + // now go ahead and resample + for (j=0; j < z->s.img_y; ++j) { + uint8 *out = output + n * z->s.img_x * j; + for (k=0; k < decode_n; ++k) { + stbi_resample *r = &res_comp[k]; + int y_bot = r->ystep >= (r->vs >> 1); + coutput[k] = r->resample(z->img_comp[k].linebuf, + y_bot ? r->line1 : r->line0, + y_bot ? r->line0 : r->line1, + r->w_lores, r->hs); + if (++r->ystep >= r->vs) { + r->ystep = 0; + r->line0 = r->line1; + if (++r->ypos < z->img_comp[k].y) + r->line1 += z->img_comp[k].w2; + } + } + if (n >= 3) { + uint8 *y = coutput[0]; + if (z->s.img_n == 3) { + #if STBI_SIMD + stbi_YCbCr_installed(out, y, coutput[1], coutput[2], z->s.img_x, n); + #else + YCbCr_to_RGB_row(out, y, coutput[1], coutput[2], z->s.img_x, n); + #endif + } else + for (i=0; i < z->s.img_x; ++i) { + out[0] = out[1] = out[2] = y[i]; + out[3] = 255; // not used if n==3 + out += n; + } + } else { + uint8 *y = coutput[0]; + if (n == 1) + for (i=0; i < z->s.img_x; ++i) out[i] = y[i]; + else + for (i=0; i < z->s.img_x; ++i) *out++ = y[i], *out++ = 255; + } + } + cleanup_jpeg(z); + *out_x = z->s.img_x; + *out_y = z->s.img_y; + if (comp) *comp = z->s.img_n; // report original components, not output + return output; + } +} + +#ifndef STBI_NO_STDIO +unsigned char *stbi_jpeg_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + jpeg j; + start_file(&j.s, f); + return load_jpeg_image(&j, x,y,comp,req_comp); +} + +unsigned char *stbi_jpeg_load(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *data; + FILE *f = fopen(filename, "rb"); + if (!f) return NULL; + data = stbi_jpeg_load_from_file(f,x,y,comp,req_comp); + fclose(f); + return data; +} +#endif + +unsigned char *stbi_jpeg_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + jpeg j; + start_mem(&j.s, buffer,len); + return load_jpeg_image(&j, x,y,comp,req_comp); +} + +#ifndef STBI_NO_STDIO +int stbi_jpeg_test_file(FILE *f) +{ + int n,r; + jpeg j; + n = ftell(f); + start_file(&j.s, f); + r = decode_jpeg_header(&j, SCAN_type); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +int stbi_jpeg_test_memory(stbi_uc const *buffer, int len) +{ + jpeg j; + start_mem(&j.s, buffer,len); + return decode_jpeg_header(&j, SCAN_type); +} + +// @TODO: +#ifndef STBI_NO_STDIO +extern int stbi_jpeg_info (char const *filename, int *x, int *y, int *comp); +extern int stbi_jpeg_info_from_file (FILE *f, int *x, int *y, int *comp); +#endif +extern int stbi_jpeg_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp); + +// public domain zlib decode v0.2 Sean Barrett 2006-11-18 +// simple implementation +// - all input must be provided in an upfront buffer +// - all output is written to a single output buffer (can malloc/realloc) +// performance +// - fast huffman + +// fast-way is faster to check than jpeg huffman, but slow way is slower +#define ZFAST_BITS 9 // accelerate all cases in default tables +#define ZFAST_MASK ((1 << ZFAST_BITS) - 1) + +// zlib-style huffman encoding +// (jpegs packs from left, zlib from right, so can't share code) +typedef struct +{ + uint16 fast[1 << ZFAST_BITS]; + uint16 firstcode[16]; + int maxcode[17]; + uint16 firstsymbol[16]; + uint8 size[288]; + uint16 value[288]; +} zhuffman; + +__forceinline static int bitreverse16(int n) +{ + n = ((n & 0xAAAA) >> 1) | ((n & 0x5555) << 1); + n = ((n & 0xCCCC) >> 2) | ((n & 0x3333) << 2); + n = ((n & 0xF0F0) >> 4) | ((n & 0x0F0F) << 4); + n = ((n & 0xFF00) >> 8) | ((n & 0x00FF) << 8); + return n; +} + +__forceinline static int bit_reverse(int v, int bits) +{ + assert(bits <= 16); + // to bit reverse n bits, reverse 16 and shift + // e.g. 11 bits, bit reverse and shift away 5 + return bitreverse16(v) >> (16-bits); +} + +static int zbuild_huffman(zhuffman *z, uint8 *sizelist, int num) +{ + int i,k=0; + int code, next_code[16], sizes[17]; + + // DEFLATE spec for generating codes + memset(sizes, 0, sizeof(sizes)); + memset(z->fast, 255, sizeof(z->fast)); + for (i=0; i < num; ++i) + ++sizes[sizelist[i]]; + sizes[0] = 0; + for (i=1; i < 16; ++i) + assert(sizes[i] <= (1 << i)); + code = 0; + for (i=1; i < 16; ++i) { + next_code[i] = code; + z->firstcode[i] = (uint16) code; + z->firstsymbol[i] = (uint16) k; + code = (code + sizes[i]); + if (sizes[i]) + if (code-1 >= (1 << i)) return e("bad codelengths","Corrupt JPEG"); + z->maxcode[i] = code << (16-i); // preshift for inner loop + code <<= 1; + k += sizes[i]; + } + z->maxcode[16] = 0x10000; // sentinel + for (i=0; i < num; ++i) { + int s = sizelist[i]; + if (s) { + int c = next_code[s] - z->firstcode[s] + z->firstsymbol[s]; + z->size[c] = (uint8)s; + z->value[c] = (uint16)i; + if (s <= ZFAST_BITS) { + int k = bit_reverse(next_code[s],s); + while (k < (1 << ZFAST_BITS)) { + z->fast[k] = (uint16) c; + k += (1 << s); + } + } + ++next_code[s]; + } + } + return 1; +} + +// zlib-from-memory implementation for PNG reading +// because PNG allows splitting the zlib stream arbitrarily, +// and it's annoying structurally to have PNG call ZLIB call PNG, +// we require PNG read all the IDATs and combine them into a single +// memory buffer + +typedef struct +{ + uint8 *zbuffer, *zbuffer_end; + int num_bits; + uint32 code_buffer; + + char *zout; + char *zout_start; + char *zout_end; + int z_expandable; + + zhuffman z_length, z_distance; +} zbuf; + +__forceinline static int zget8(zbuf *z) +{ + if (z->zbuffer >= z->zbuffer_end) return 0; + return *z->zbuffer++; +} + +static void fill_bits(zbuf *z) +{ + do { + assert(z->code_buffer < (1U << z->num_bits)); + z->code_buffer |= zget8(z) << z->num_bits; + z->num_bits += 8; + } while (z->num_bits <= 24); +} + +__forceinline static unsigned int zreceive(zbuf *z, int n) +{ + unsigned int k; + if (z->num_bits < n) fill_bits(z); + k = z->code_buffer & ((1 << n) - 1); + z->code_buffer >>= n; + z->num_bits -= n; + return k; +} + +__forceinline static int zhuffman_decode(zbuf *a, zhuffman *z) +{ + int b,s,k; + if (a->num_bits < 16) fill_bits(a); + b = z->fast[a->code_buffer & ZFAST_MASK]; + if (b < 0xffff) { + s = z->size[b]; + a->code_buffer >>= s; + a->num_bits -= s; + return z->value[b]; + } + + // not resolved by fast table, so compute it the slow way + // use jpeg approach, which requires MSbits at top + k = bit_reverse(a->code_buffer, 16); + for (s=ZFAST_BITS+1; ; ++s) + if (k < z->maxcode[s]) + break; + if (s == 16) return -1; // invalid code! + // code size is s, so: + b = (k >> (16-s)) - z->firstcode[s] + z->firstsymbol[s]; + assert(z->size[b] == s); + a->code_buffer >>= s; + a->num_bits -= s; + return z->value[b]; +} + +static int expand(zbuf *z, int n) // need to make room for n bytes +{ + char *q; + int cur, limit; + if (!z->z_expandable) return e("output buffer limit","Corrupt PNG"); + cur = (int) (z->zout - z->zout_start); + limit = (int) (z->zout_end - z->zout_start); + while (cur + n > limit) + limit *= 2; + q = (char *) realloc(z->zout_start, limit); + if (q == NULL) return e("outofmem", "Out of memory"); + z->zout_start = q; + z->zout = q + cur; + z->zout_end = q + limit; + return 1; +} + +static int length_base[31] = { + 3,4,5,6,7,8,9,10,11,13, + 15,17,19,23,27,31,35,43,51,59, + 67,83,99,115,131,163,195,227,258,0,0 }; + +static int length_extra[31]= +{ 0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0,0,0 }; + +static int dist_base[32] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193, +257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577,0,0}; + +static int dist_extra[32] = +{ 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13}; + +static int parse_huffman_block(zbuf *a) +{ + for(;;) { + int z = zhuffman_decode(a, &a->z_length); + if (z < 256) { + if (z < 0) return e("bad huffman code","Corrupt PNG"); // error in huffman codes + if (a->zout >= a->zout_end) if (!expand(a, 1)) return 0; + *a->zout++ = (char) z; + } else { + uint8 *p; + int len,dist; + if (z == 256) return 1; + z -= 257; + len = length_base[z]; + if (length_extra[z]) len += zreceive(a, length_extra[z]); + z = zhuffman_decode(a, &a->z_distance); + if (z < 0) return e("bad huffman code","Corrupt PNG"); + dist = dist_base[z]; + if (dist_extra[z]) dist += zreceive(a, dist_extra[z]); + if (a->zout - a->zout_start < dist) return e("bad dist","Corrupt PNG"); + if (a->zout + len > a->zout_end) if (!expand(a, len)) return 0; + p = (uint8 *) (a->zout - dist); + while (len--) + *a->zout++ = *p++; + } + } +} + +static int compute_huffman_codes(zbuf *a) +{ + static uint8 length_dezigzag[19] = { 16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15 }; + static zhuffman z_codelength; // static just to save stack space + uint8 lencodes[286+32+137];//padding for maximum single op + uint8 codelength_sizes[19]; + int i,n; + + int hlit = zreceive(a,5) + 257; + int hdist = zreceive(a,5) + 1; + int hclen = zreceive(a,4) + 4; + + memset(codelength_sizes, 0, sizeof(codelength_sizes)); + for (i=0; i < hclen; ++i) { + int s = zreceive(a,3); + codelength_sizes[length_dezigzag[i]] = (uint8) s; + } + if (!zbuild_huffman(&z_codelength, codelength_sizes, 19)) return 0; + + n = 0; + while (n < hlit + hdist) { + int c = zhuffman_decode(a, &z_codelength); + assert(c >= 0 && c < 19); + if (c < 16) + lencodes[n++] = (uint8) c; + else if (c == 16) { + c = zreceive(a,2)+3; + memset(lencodes+n, lencodes[n-1], c); + n += c; + } else if (c == 17) { + c = zreceive(a,3)+3; + memset(lencodes+n, 0, c); + n += c; + } else { + assert(c == 18); + c = zreceive(a,7)+11; + memset(lencodes+n, 0, c); + n += c; + } + } + if (n != hlit+hdist) return e("bad codelengths","Corrupt PNG"); + if (!zbuild_huffman(&a->z_length, lencodes, hlit)) return 0; + if (!zbuild_huffman(&a->z_distance, lencodes+hlit, hdist)) return 0; + return 1; +} + +static int parse_uncompressed_block(zbuf *a) +{ + uint8 header[4]; + int len,nlen,k; + if (a->num_bits & 7) + zreceive(a, a->num_bits & 7); // discard + // drain the bit-packed data into header + k = 0; + while (a->num_bits > 0) { + header[k++] = (uint8) (a->code_buffer & 255); // wtf this warns? + a->code_buffer >>= 8; + a->num_bits -= 8; + } + assert(a->num_bits == 0); + // now fill header the normal way + while (k < 4) + header[k++] = (uint8) zget8(a); + len = header[1] * 256 + header[0]; + nlen = header[3] * 256 + header[2]; + if (nlen != (len ^ 0xffff)) return e("zlib corrupt","Corrupt PNG"); + if (a->zbuffer + len > a->zbuffer_end) return e("read past buffer","Corrupt PNG"); + if (a->zout + len > a->zout_end) + if (!expand(a, len)) return 0; + memcpy(a->zout, a->zbuffer, len); + a->zbuffer += len; + a->zout += len; + return 1; +} + +static int parse_zlib_header(zbuf *a) +{ + int cmf = zget8(a); + int cm = cmf & 15; + /* int cinfo = cmf >> 4; */ + int flg = zget8(a); + if ((cmf*256+flg) % 31 != 0) return e("bad zlib header","Corrupt PNG"); // zlib spec + if (flg & 32) return e("no preset dict","Corrupt PNG"); // preset dictionary not allowed in png + if (cm != 8) return e("bad compression","Corrupt PNG"); // DEFLATE required for png + // window = 1 << (8 + cinfo)... but who cares, we fully buffer output + return 1; +} + +// @TODO: should statically initialize these for optimal thread safety +static uint8 default_length[288], default_distance[32]; +static void init_defaults(void) +{ + int i; // use <= to match clearly with spec + for (i=0; i <= 143; ++i) default_length[i] = 8; + for ( ; i <= 255; ++i) default_length[i] = 9; + for ( ; i <= 279; ++i) default_length[i] = 7; + for ( ; i <= 287; ++i) default_length[i] = 8; + + for (i=0; i <= 31; ++i) default_distance[i] = 5; +} + +static int parse_zlib(zbuf *a, int parse_header) +{ + int final, type; + if (parse_header) + if (!parse_zlib_header(a)) return 0; + a->num_bits = 0; + a->code_buffer = 0; + do { + final = zreceive(a,1); + type = zreceive(a,2); + if (type == 0) { + if (!parse_uncompressed_block(a)) return 0; + } else if (type == 3) { + return 0; + } else { + if (type == 1) { + // use fixed code lengths + if (!default_distance[31]) init_defaults(); + if (!zbuild_huffman(&a->z_length , default_length , 288)) return 0; + if (!zbuild_huffman(&a->z_distance, default_distance, 32)) return 0; + } else { + if (!compute_huffman_codes(a)) return 0; + } + if (!parse_huffman_block(a)) return 0; + } + } while (!final); + return 1; +} + +static int do_zlib(zbuf *a, char *obuf, int olen, int exp, int parse_header) +{ + a->zout_start = obuf; + a->zout = obuf; + a->zout_end = obuf + olen; + a->z_expandable = exp; + + return parse_zlib(a, parse_header); +} + +char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen) +{ + zbuf a; + char *p = (char *) malloc(initial_size); + if (p == NULL) return NULL; + a.zbuffer = (uint8 *) buffer; + a.zbuffer_end = (uint8 *) buffer + len; + if (do_zlib(&a, p, initial_size, 1, 1)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + free(a.zout_start); + return NULL; + } +} + +char *stbi_zlib_decode_malloc(char const *buffer, int len, int *outlen) +{ + return stbi_zlib_decode_malloc_guesssize(buffer, len, 16384, outlen); +} + +int stbi_zlib_decode_buffer(char *obuffer, int olen, char const *ibuffer, int ilen) +{ + zbuf a; + a.zbuffer = (uint8 *) ibuffer; + a.zbuffer_end = (uint8 *) ibuffer + ilen; + if (do_zlib(&a, obuffer, olen, 0, 1)) + return (int) (a.zout - a.zout_start); + else + return -1; +} + +char *stbi_zlib_decode_noheader_malloc(char const *buffer, int len, int *outlen) +{ + zbuf a; + char *p = (char *) malloc(16384); + if (p == NULL) return NULL; + a.zbuffer = (uint8 *) buffer; + a.zbuffer_end = (uint8 *) buffer+len; + if (do_zlib(&a, p, 16384, 1, 0)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + free(a.zout_start); + return NULL; + } +} + +int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen) +{ + zbuf a; + a.zbuffer = (uint8 *) ibuffer; + a.zbuffer_end = (uint8 *) ibuffer + ilen; + if (do_zlib(&a, obuffer, olen, 0, 0)) + return (int) (a.zout - a.zout_start); + else + return -1; +} + +// public domain "baseline" PNG decoder v0.10 Sean Barrett 2006-11-18 +// simple implementation +// - only 8-bit samples +// - no CRC checking +// - allocates lots of intermediate memory +// - avoids problem of streaming data between subsystems +// - avoids explicit window management +// performance +// - uses stb_zlib, a PD zlib implementation with fast huffman decoding + + +typedef struct +{ + uint32 length; + uint32 type; +} chunk; + +#define PNG_TYPE(a,b,c,d) (((a) << 24) + ((b) << 16) + ((c) << 8) + (d)) + +static chunk get_chunk_header(stbi *s) +{ + chunk c; + c.length = get32(s); + c.type = get32(s); + return c; +} + +static int check_png_header(stbi *s) +{ + static uint8 png_sig[8] = { 137,80,78,71,13,10,26,10 }; + int i; + for (i=0; i < 8; ++i) + if (get8(s) != png_sig[i]) return e("bad png sig","Not a PNG"); + return 1; +} + +typedef struct +{ + stbi s; + uint8 *idata, *expanded, *out; +} png; + + +enum { + F_none=0, F_sub=1, F_up=2, F_avg=3, F_paeth=4, + F_avg_first, F_paeth_first, +}; + +static uint8 first_row_filter[5] = +{ + F_none, F_sub, F_none, F_avg_first, F_paeth_first +}; + +static int paeth(int a, int b, int c) +{ + int p = a + b - c; + int pa = abs(p-a); + int pb = abs(p-b); + int pc = abs(p-c); + if (pa <= pb && pa <= pc) return a; + if (pb <= pc) return b; + return c; +} + +// create the png data from post-deflated data +static int create_png_image(png *a, uint8 *raw, uint32 raw_len, int out_n) +{ + stbi *s = &a->s; + uint32 i,j,stride = s->img_x*out_n; + int k; + int img_n = s->img_n; // copy it into a local for later + assert(out_n == s->img_n || out_n == s->img_n+1); + a->out = (uint8 *) malloc(s->img_x * s->img_y * out_n); + if (!a->out) return e("outofmem", "Out of memory"); + if (raw_len != (img_n * s->img_x + 1) * s->img_y) return e("not enough pixels","Corrupt PNG"); + for (j=0; j < s->img_y; ++j) { + uint8 *cur = a->out + stride*j; + uint8 *prior = cur - stride; + int filter = *raw++; + if (filter > 4) return e("invalid filter","Corrupt PNG"); + // if first row, use special filter that doesn't sample previous row + if (j == 0) filter = first_row_filter[filter]; + // handle first pixel explicitly + for (k=0; k < img_n; ++k) { + switch(filter) { + case F_none : cur[k] = raw[k]; break; + case F_sub : cur[k] = raw[k]; break; + case F_up : cur[k] = raw[k] + prior[k]; break; + case F_avg : cur[k] = raw[k] + (prior[k]>>1); break; + case F_paeth : cur[k] = (uint8) (raw[k] + paeth(0,prior[k],0)); break; + case F_avg_first : cur[k] = raw[k]; break; + case F_paeth_first: cur[k] = raw[k]; break; + } + } + if (img_n != out_n) cur[img_n] = 255; + raw += img_n; + cur += out_n; + prior += out_n; + // this is a little gross, so that we don't switch per-pixel or per-component + if (img_n == out_n) { + #define CASE(f) \ + case f: \ + for (i=s->img_x-1; i >= 1; --i, raw+=img_n,cur+=img_n,prior+=img_n) \ + for (k=0; k < img_n; ++k) + switch(filter) { + CASE(F_none) cur[k] = raw[k]; break; + CASE(F_sub) cur[k] = raw[k] + cur[k-img_n]; break; + CASE(F_up) cur[k] = raw[k] + prior[k]; break; + CASE(F_avg) cur[k] = raw[k] + ((prior[k] + cur[k-img_n])>>1); break; + CASE(F_paeth) cur[k] = (uint8) (raw[k] + paeth(cur[k-img_n],prior[k],prior[k-img_n])); break; + CASE(F_avg_first) cur[k] = raw[k] + (cur[k-img_n] >> 1); break; + CASE(F_paeth_first) cur[k] = (uint8) (raw[k] + paeth(cur[k-img_n],0,0)); break; + } + #undef CASE + } else { + assert(img_n+1 == out_n); + #define CASE(f) \ + case f: \ + for (i=s->img_x-1; i >= 1; --i, cur[img_n]=255,raw+=img_n,cur+=out_n,prior+=out_n) \ + for (k=0; k < img_n; ++k) + switch(filter) { + CASE(F_none) cur[k] = raw[k]; break; + CASE(F_sub) cur[k] = raw[k] + cur[k-out_n]; break; + CASE(F_up) cur[k] = raw[k] + prior[k]; break; + CASE(F_avg) cur[k] = raw[k] + ((prior[k] + cur[k-out_n])>>1); break; + CASE(F_paeth) cur[k] = (uint8) (raw[k] + paeth(cur[k-out_n],prior[k],prior[k-out_n])); break; + CASE(F_avg_first) cur[k] = raw[k] + (cur[k-out_n] >> 1); break; + CASE(F_paeth_first) cur[k] = (uint8) (raw[k] + paeth(cur[k-out_n],0,0)); break; + } + #undef CASE + } + } + return 1; +} + +static int compute_transparency(png *z, uint8 tc[3], int out_n) +{ + stbi *s = &z->s; + uint32 i, pixel_count = s->img_x * s->img_y; + uint8 *p = z->out; + + // compute color-based transparency, assuming we've + // already got 255 as the alpha value in the output + assert(out_n == 2 || out_n == 4); + + if (out_n == 2) { + for (i=0; i < pixel_count; ++i) { + p[1] = (p[0] == tc[0] ? 0 : 255); + p += 2; + } + } else { + for (i=0; i < pixel_count; ++i) { + if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2]) + p[3] = 0; + p += 4; + } + } + return 1; +} + +static int expand_palette(png *a, uint8 *palette, int len, int pal_img_n) +{ + uint32 i, pixel_count = a->s.img_x * a->s.img_y; + uint8 *p, *temp_out, *orig = a->out; + + p = (uint8 *) malloc(pixel_count * pal_img_n); + if (p == NULL) return e("outofmem", "Out of memory"); + + // between here and free(out) below, exitting would leak + temp_out = p; + + if (pal_img_n == 3) { + for (i=0; i < pixel_count; ++i) { + int n = orig[i]*4; + p[0] = palette[n ]; + p[1] = palette[n+1]; + p[2] = palette[n+2]; + p += 3; + } + } else { + for (i=0; i < pixel_count; ++i) { + int n = orig[i]*4; + p[0] = palette[n ]; + p[1] = palette[n+1]; + p[2] = palette[n+2]; + p[3] = palette[n+3]; + p += 4; + } + } + free(a->out); + a->out = temp_out; + return 1; +} + +static int parse_png_file(png *z, int scan, int req_comp) +{ + uint8 palette[1024], pal_img_n=0; + uint8 has_trans=0, tc[3]; + uint32 ioff=0, idata_limit=0, i, pal_len=0; + int first=1,k; + stbi *s = &z->s; + + if (!check_png_header(s)) return 0; + + if (scan == SCAN_type) return 1; + + for(;;first=0) { + chunk c = get_chunk_header(s); + if (first && c.type != PNG_TYPE('I','H','D','R')) + return e("first not IHDR","Corrupt PNG"); + switch (c.type) { + case PNG_TYPE('I','H','D','R'): { + int depth,color,interlace,comp,filter; + if (!first) return e("multiple IHDR","Corrupt PNG"); + if (c.length != 13) return e("bad IHDR len","Corrupt PNG"); + s->img_x = get32(s); if (s->img_x > (1 << 24)) return e("too large","Very large image (corrupt?)"); + s->img_y = get32(s); if (s->img_y > (1 << 24)) return e("too large","Very large image (corrupt?)"); + depth = get8(s); if (depth != 8) return e("8bit only","PNG not supported: 8-bit only"); + color = get8(s); if (color > 6) return e("bad ctype","Corrupt PNG"); + if (color == 3) pal_img_n = 3; else if (color & 1) return e("bad ctype","Corrupt PNG"); + comp = get8(s); if (comp) return e("bad comp method","Corrupt PNG"); + filter= get8(s); if (filter) return e("bad filter method","Corrupt PNG"); + interlace = get8(s); if (interlace) return e("interlaced","PNG not supported: interlaced mode"); + if (!s->img_x || !s->img_y) return e("0-pixel image","Corrupt PNG"); + if (!pal_img_n) { + s->img_n = (color & 2 ? 3 : 1) + (color & 4 ? 1 : 0); + if ((1 << 30) / s->img_x / s->img_n < s->img_y) return e("too large", "Image too large to decode"); + if (scan == SCAN_header) return 1; + } else { + // if paletted, then pal_n is our final components, and + // img_n is # components to decompress/filter. + s->img_n = 1; + if ((1 << 30) / s->img_x / 4 < s->img_y) return e("too large","Corrupt PNG"); + // if SCAN_header, have to scan to see if we have a tRNS + } + break; + } + + case PNG_TYPE('P','L','T','E'): { + if (c.length > 256*3) return e("invalid PLTE","Corrupt PNG"); + pal_len = c.length / 3; + if (pal_len * 3 != c.length) return e("invalid PLTE","Corrupt PNG"); + for (i=0; i < pal_len; ++i) { + palette[i*4+0] = get8u(s); + palette[i*4+1] = get8u(s); + palette[i*4+2] = get8u(s); + palette[i*4+3] = 255; + } + break; + } + + case PNG_TYPE('t','R','N','S'): { + if (z->idata) return e("tRNS after IDAT","Corrupt PNG"); + if (pal_img_n) { + if (scan == SCAN_header) { s->img_n = 4; return 1; } + if (pal_len == 0) return e("tRNS before PLTE","Corrupt PNG"); + if (c.length > pal_len) return e("bad tRNS len","Corrupt PNG"); + pal_img_n = 4; + for (i=0; i < c.length; ++i) + palette[i*4+3] = get8u(s); + } else { + if (!(s->img_n & 1)) return e("tRNS with alpha","Corrupt PNG"); + if (c.length != (uint32) s->img_n*2) return e("bad tRNS len","Corrupt PNG"); + has_trans = 1; + for (k=0; k < s->img_n; ++k) + tc[k] = (uint8) get16(s); // non 8-bit images will be larger + } + break; + } + + case PNG_TYPE('I','D','A','T'): { + if (pal_img_n && !pal_len) return e("no PLTE","Corrupt PNG"); + if (scan == SCAN_header) { s->img_n = pal_img_n; return 1; } + if (ioff + c.length > idata_limit) { + uint8 *p; + if (idata_limit == 0) idata_limit = c.length > 4096 ? c.length : 4096; + while (ioff + c.length > idata_limit) + idata_limit *= 2; + p = (uint8 *) realloc(z->idata, idata_limit); if (p == NULL) return e("outofmem", "Out of memory"); + z->idata = p; + } + #ifndef STBI_NO_STDIO + if (s->img_file) + { + if (fread(z->idata+ioff,1,c.length,s->img_file) != c.length) return e("outofdata","Corrupt PNG"); + } + else + #endif + { + memcpy(z->idata+ioff, s->img_buffer, c.length); + s->img_buffer += c.length; + } + ioff += c.length; + break; + } + + case PNG_TYPE('I','E','N','D'): { + uint32 raw_len; + if (scan != SCAN_load) return 1; + if (z->idata == NULL) return e("no IDAT","Corrupt PNG"); + z->expanded = (uint8 *) stbi_zlib_decode_malloc((char *) z->idata, ioff, (int *) &raw_len); + if (z->expanded == NULL) return 0; // zlib should set error + free(z->idata); z->idata = NULL; + if ((req_comp == s->img_n+1 && req_comp != 3 && !pal_img_n) || has_trans) + s->img_out_n = s->img_n+1; + else + s->img_out_n = s->img_n; + if (!create_png_image(z, z->expanded, raw_len, s->img_out_n)) return 0; + if (has_trans) + if (!compute_transparency(z, tc, s->img_out_n)) return 0; + if (pal_img_n) { + // pal_img_n == 3 or 4 + s->img_n = pal_img_n; // record the actual colors we had + s->img_out_n = pal_img_n; + if (req_comp >= 3) s->img_out_n = req_comp; + if (!expand_palette(z, palette, pal_len, s->img_out_n)) + return 0; + } + free(z->expanded); z->expanded = NULL; + return 1; + } + + default: + // if critical, fail + if ((c.type & (1 << 29)) == 0) { + #ifndef STBI_NO_FAILURE_STRINGS + // not threadsafe + static char invalid_chunk[] = "XXXX chunk not known"; + invalid_chunk[0] = (uint8) (c.type >> 24); + invalid_chunk[1] = (uint8) (c.type >> 16); + invalid_chunk[2] = (uint8) (c.type >> 8); + invalid_chunk[3] = (uint8) (c.type >> 0); + #endif + return e(invalid_chunk, "PNG not supported: unknown chunk type"); + } + skip(s, c.length); + break; + } + // end of chunk, read and skip CRC + get32(s); + } +} + +static unsigned char *do_png(png *p, int *x, int *y, int *n, int req_comp) +{ + unsigned char *result=NULL; + p->expanded = NULL; + p->idata = NULL; + p->out = NULL; + if (req_comp < 0 || req_comp > 4) return epuc("bad req_comp", "Internal error"); + if (parse_png_file(p, SCAN_load, req_comp)) { + result = p->out; + p->out = NULL; + if (req_comp && req_comp != p->s.img_out_n) { + result = convert_format(result, p->s.img_out_n, req_comp, p->s.img_x, p->s.img_y); + p->s.img_out_n = req_comp; + if (result == NULL) return result; + } + *x = p->s.img_x; + *y = p->s.img_y; + if (n) *n = p->s.img_n; + } + free(p->out); p->out = NULL; + free(p->expanded); p->expanded = NULL; + free(p->idata); p->idata = NULL; + + return result; +} + +#ifndef STBI_NO_STDIO +unsigned char *stbi_png_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + png p; + start_file(&p.s, f); + return do_png(&p, x,y,comp,req_comp); +} + +unsigned char *stbi_png_load(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *data; + FILE *f = fopen(filename, "rb"); + if (!f) return NULL; + data = stbi_png_load_from_file(f,x,y,comp,req_comp); + fclose(f); + return data; +} +#endif + +unsigned char *stbi_png_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + png p; + start_mem(&p.s, buffer,len); + return do_png(&p, x,y,comp,req_comp); +} + +#ifndef STBI_NO_STDIO +int stbi_png_test_file(FILE *f) +{ + png p; + int n,r; + n = ftell(f); + start_file(&p.s, f); + r = parse_png_file(&p, SCAN_type,STBI_default); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +int stbi_png_test_memory(stbi_uc const *buffer, int len) +{ + png p; + start_mem(&p.s, buffer, len); + return parse_png_file(&p, SCAN_type,STBI_default); +} + +// TODO: load header from png +#ifndef STBI_NO_STDIO +extern int stbi_png_info (char const *filename, int *x, int *y, int *comp); +extern int stbi_png_info_from_file (FILE *f, int *x, int *y, int *comp); +#endif +extern int stbi_png_info_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp); + +// Microsoft/Windows BMP image + +static int bmp_test(stbi *s) +{ + int sz; + if (get8(s) != 'B') return 0; + if (get8(s) != 'M') return 0; + get32le(s); // discard filesize + get16le(s); // discard reserved + get16le(s); // discard reserved + get32le(s); // discard data offset + sz = get32le(s); + if (sz == 12 || sz == 40 || sz == 56 || sz == 108) return 1; + return 0; +} + +#ifndef STBI_NO_STDIO +int stbi_bmp_test_file (FILE *f) +{ + stbi s; + int r,n = ftell(f); + start_file(&s,f); + r = bmp_test(&s); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +int stbi_bmp_test_memory (stbi_uc const *buffer, int len) +{ + stbi s; + start_mem(&s, buffer, len); + return bmp_test(&s); +} + +// returns 0..31 for the highest set bit +static int high_bit(unsigned int z) +{ + int n=0; + if (z == 0) return -1; + if (z >= 0x10000) n += 16, z >>= 16; + if (z >= 0x00100) n += 8, z >>= 8; + if (z >= 0x00010) n += 4, z >>= 4; + if (z >= 0x00004) n += 2, z >>= 2; + if (z >= 0x00002) n += 1, z >>= 1; + return n; +} + +static int bitcount(unsigned int a) +{ + a = (a & 0x55555555) + ((a >> 1) & 0x55555555); // max 2 + a = (a & 0x33333333) + ((a >> 2) & 0x33333333); // max 4 + a = (a + (a >> 4)) & 0x0f0f0f0f; // max 8 per 4, now 8 bits + a = (a + (a >> 8)); // max 16 per 8 bits + a = (a + (a >> 16)); // max 32 per 8 bits + return a & 0xff; +} + +static int shiftsigned(int v, int shift, int bits) +{ + int result; + int z=0; + + if (shift < 0) v <<= -shift; + else v >>= shift; + result = v; + + z = bits; + while (z < 8) { + result += v >> z; + z += bits; + } + return result; +} + +static stbi_uc *bmp_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + uint8 *out; + unsigned int mr=0,mg=0,mb=0,ma=0; + stbi_uc pal[256][4]; + int psize=0,i,j,compress=0,width; + int bpp, flip_vertically, pad, target, offset, hsz; + if (get8(s) != 'B' || get8(s) != 'M') return epuc("not BMP", "Corrupt BMP"); + get32le(s); // discard filesize + get16le(s); // discard reserved + get16le(s); // discard reserved + offset = get32le(s); + hsz = get32le(s); + if (hsz != 12 && hsz != 40 && hsz != 56 && hsz != 108) return epuc("unknown BMP", "BMP type not supported: unknown"); + failure_reason = "bad BMP"; + if (hsz == 12) { + s->img_x = get16le(s); + s->img_y = get16le(s); + } else { + s->img_x = get32le(s); + s->img_y = get32le(s); + } + if (get16le(s) != 1) return 0; + bpp = get16le(s); + if (bpp == 1) return epuc("monochrome", "BMP type not supported: 1-bit"); + flip_vertically = ((int) s->img_y) > 0; + s->img_y = abs((int) s->img_y); + if (hsz == 12) { + if (bpp < 24) + psize = (offset - 14 - 24) / 3; + } else { + compress = get32le(s); + if (compress == 1 || compress == 2) return epuc("BMP RLE", "BMP type not supported: RLE"); + get32le(s); // discard sizeof + get32le(s); // discard hres + get32le(s); // discard vres + get32le(s); // discard colorsused + get32le(s); // discard max important + if (hsz == 40 || hsz == 56) { + if (hsz == 56) { + get32le(s); + get32le(s); + get32le(s); + get32le(s); + } + if (bpp == 16 || bpp == 32) { + mr = mg = mb = 0; + if (compress == 0) { + if (bpp == 32) { + mr = 0xff << 16; + mg = 0xff << 8; + mb = 0xff << 0; + } else { + mr = 31 << 10; + mg = 31 << 5; + mb = 31 << 0; + } + } else if (compress == 3) { + mr = get32le(s); + mg = get32le(s); + mb = get32le(s); + // not documented, but generated by photoshop and handled by mspaint + if (mr == mg && mg == mb) { + // ?!?!? + return NULL; + } + } else + return NULL; + } + } else { + assert(hsz == 108); + mr = get32le(s); + mg = get32le(s); + mb = get32le(s); + ma = get32le(s); + get32le(s); // discard color space + for (i=0; i < 12; ++i) + get32le(s); // discard color space parameters + } + if (bpp < 16) + psize = (offset - 14 - hsz) >> 2; + } + s->img_n = ma ? 4 : 3; + if (req_comp && req_comp >= 3) // we can directly decode 3 or 4 + target = req_comp; + else + target = s->img_n; // if they want monochrome, we'll post-convert + out = (stbi_uc *) malloc(target * s->img_x * s->img_y); + if (!out) return epuc("outofmem", "Out of memory"); + if (bpp < 16) { + int z=0; + if (psize == 0 || psize > 256) { free(out); return epuc("invalid", "Corrupt BMP"); } + for (i=0; i < psize; ++i) { + pal[i][2] = get8(s); + pal[i][1] = get8(s); + pal[i][0] = get8(s); + if (hsz != 12) get8(s); + pal[i][3] = 255; + } + skip(s, offset - 14 - hsz - psize * (hsz == 12 ? 3 : 4)); + if (bpp == 4) width = (s->img_x + 1) >> 1; + else if (bpp == 8) width = s->img_x; + else { free(out); return epuc("bad bpp", "Corrupt BMP"); } + pad = (-width)&3; + for (j=0; j < (int) s->img_y; ++j) { + for (i=0; i < (int) s->img_x; i += 2) { + int v=get8(s),v2=0; + if (bpp == 4) { + v2 = v & 15; + v >>= 4; + } + out[z++] = pal[v][0]; + out[z++] = pal[v][1]; + out[z++] = pal[v][2]; + if (target == 4) out[z++] = 255; + if (i+1 == (int) s->img_x) break; + v = (bpp == 8) ? get8(s) : v2; + out[z++] = pal[v][0]; + out[z++] = pal[v][1]; + out[z++] = pal[v][2]; + if (target == 4) out[z++] = 255; + } + skip(s, pad); + } + } else { + int rshift=0,gshift=0,bshift=0,ashift=0,rcount=0,gcount=0,bcount=0,acount=0; + int z = 0; + int easy=0; + skip(s, offset - 14 - hsz); + if (bpp == 24) width = 3 * s->img_x; + else if (bpp == 16) width = 2*s->img_x; + else /* bpp = 32 and pad = 0 */ width=0; + pad = (-width) & 3; + if (bpp == 24) { + easy = 1; + } else if (bpp == 32) { + if (mb == 0xff && mg == 0xff00 && mr == 0xff000000 && ma == 0xff000000) + easy = 2; + } + if (!easy) { + if (!mr || !mg || !mb) return epuc("bad masks", "Corrupt BMP"); + // right shift amt to put high bit in position #7 + rshift = high_bit(mr)-7; rcount = bitcount(mr); + gshift = high_bit(mg)-7; gcount = bitcount(mr); + bshift = high_bit(mb)-7; bcount = bitcount(mr); + ashift = high_bit(ma)-7; acount = bitcount(mr); + } + for (j=0; j < (int) s->img_y; ++j) { + if (easy) { + for (i=0; i < (int) s->img_x; ++i) { + int a; + out[z+2] = get8(s); + out[z+1] = get8(s); + out[z+0] = get8(s); + z += 3; + a = (easy == 2 ? get8(s) : 255); + if (target == 4) out[z++] = a; + } + } else { + for (i=0; i < (int) s->img_x; ++i) { + uint32 v = (bpp == 16 ? get16le(s) : get32le(s)); + int a; + out[z++] = shiftsigned(v & mr, rshift, rcount); + out[z++] = shiftsigned(v & mg, gshift, gcount); + out[z++] = shiftsigned(v & mb, bshift, bcount); + a = (ma ? shiftsigned(v & ma, ashift, acount) : 255); + if (target == 4) out[z++] = a; + } + } + skip(s, pad); + } + } + if (flip_vertically) { + stbi_uc t; + for (j=0; j < (int) s->img_y>>1; ++j) { + stbi_uc *p1 = out + j *s->img_x*target; + stbi_uc *p2 = out + (s->img_y-1-j)*s->img_x*target; + for (i=0; i < (int) s->img_x*target; ++i) { + t = p1[i], p1[i] = p2[i], p2[i] = t; + } + } + } + + if (req_comp && req_comp != target) { + out = convert_format(out, target, req_comp, s->img_x, s->img_y); + if (out == NULL) return out; // convert_format frees input on failure + } + + *x = s->img_x; + *y = s->img_y; + if (comp) *comp = target; + return out; +} + +#ifndef STBI_NO_STDIO +stbi_uc *stbi_bmp_load (char const *filename, int *x, int *y, int *comp, int req_comp) +{ + stbi_uc *data; + FILE *f = fopen(filename, "rb"); + if (!f) return NULL; + data = stbi_bmp_load_from_file(f, x,y,comp,req_comp); + fclose(f); + return data; +} + +stbi_uc *stbi_bmp_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_file(&s, f); + return bmp_load(&s, x,y,comp,req_comp); +} +#endif + +stbi_uc *stbi_bmp_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_mem(&s, buffer, len); + return bmp_load(&s, x,y,comp,req_comp); +} + +// Targa Truevision - TGA +// by Jonathan Dummer + +static int tga_test(stbi *s) +{ + int sz; + get8u(s); // discard Offset + sz = get8u(s); // color type + if( sz > 1 ) return 0; // only RGB or indexed allowed + sz = get8u(s); // image type + if( (sz != 1) && (sz != 2) && (sz != 3) && (sz != 9) && (sz != 10) && (sz != 11) ) return 0; // only RGB or grey allowed, +/- RLE + get16(s); // discard palette start + get16(s); // discard palette length + get8(s); // discard bits per palette color entry + get16(s); // discard x origin + get16(s); // discard y origin + if( get16(s) < 1 ) return 0; // test width + if( get16(s) < 1 ) return 0; // test height + sz = get8(s); // bits per pixel + if( (sz != 8) && (sz != 16) && (sz != 24) && (sz != 32) ) return 0; // only RGB or RGBA or grey allowed + return 1; // seems to have passed everything +} + +#ifndef STBI_NO_STDIO +int stbi_tga_test_file (FILE *f) +{ + stbi s; + int r,n = ftell(f); + start_file(&s, f); + r = tga_test(&s); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +int stbi_tga_test_memory (stbi_uc const *buffer, int len) +{ + stbi s; + start_mem(&s, buffer, len); + return tga_test(&s); +} + +static stbi_uc *tga_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + // read in the TGA header stuff + int tga_offset = get8u(s); + int tga_indexed = get8u(s); + int tga_image_type = get8u(s); + int tga_is_RLE = 0; + int tga_palette_start = get16le(s); + int tga_palette_len = get16le(s); + int tga_palette_bits = get8u(s); + int tga_x_origin = get16le(s); + int tga_y_origin = get16le(s); + int tga_width = get16le(s); + int tga_height = get16le(s); + int tga_bits_per_pixel = get8u(s); + int tga_inverted = get8u(s); + // image data + unsigned char *tga_data; + unsigned char *tga_palette = NULL; + int i, j; + unsigned char raw_data[4]; + unsigned char trans_data[4]; + int RLE_count = 0; + int RLE_repeating = 0; + int read_next_pixel = 1; + // do a tiny bit of precessing + if( tga_image_type >= 8 ) + { + tga_image_type -= 8; + tga_is_RLE = 1; + } + /* int tga_alpha_bits = tga_inverted & 15; */ + tga_inverted = 1 - ((tga_inverted >> 5) & 1); + + // error check + if( //(tga_indexed) || + (tga_width < 1) || (tga_height < 1) || + (tga_image_type < 1) || (tga_image_type > 3) || + ((tga_bits_per_pixel != 8) && (tga_bits_per_pixel != 16) && + (tga_bits_per_pixel != 24) && (tga_bits_per_pixel != 32)) + ) + { + return NULL; + } + + // If I'm paletted, then I'll use the number of bits from the palette + if( tga_indexed ) + { + tga_bits_per_pixel = tga_palette_bits; + } + + // tga info + *x = tga_width; + *y = tga_height; + if( (req_comp < 1) || (req_comp > 4) ) + { + // just use whatever the file was + req_comp = tga_bits_per_pixel / 8; + *comp = req_comp; + } else + { + // force a new number of components + *comp = tga_bits_per_pixel/8; + } + tga_data = (unsigned char*)malloc( tga_width * tga_height * req_comp ); + + // skip to the data's starting position (offset usually = 0) + skip(s, tga_offset ); + // do I need to load a palette? + if( tga_indexed ) + { + // any data to skip? (offset usually = 0) + skip(s, tga_palette_start ); + // load the palette + tga_palette = (unsigned char*)malloc( tga_palette_len * tga_palette_bits / 8 ); + getn(s, tga_palette, tga_palette_len * tga_palette_bits / 8 ); + } + // load the data + for( i = 0; i < tga_width * tga_height; ++i ) + { + // if I'm in RLE mode, do I need to get a RLE chunk? + if( tga_is_RLE ) + { + if( RLE_count == 0 ) + { + // yep, get the next byte as a RLE command + int RLE_cmd = get8u(s); + RLE_count = 1 + (RLE_cmd & 127); + RLE_repeating = RLE_cmd >> 7; + read_next_pixel = 1; + } else if( !RLE_repeating ) + { + read_next_pixel = 1; + } + } else + { + read_next_pixel = 1; + } + // OK, if I need to read a pixel, do it now + if( read_next_pixel ) + { + // load however much data we did have + if( tga_indexed ) + { + // read in 1 byte, then perform the lookup + int pal_idx = get8u(s); + if( pal_idx >= tga_palette_len ) + { + // invalid index + pal_idx = 0; + } + pal_idx *= tga_bits_per_pixel / 8; + for( j = 0; j*8 < tga_bits_per_pixel; ++j ) + { + raw_data[j] = tga_palette[pal_idx+j]; + } + } else + { + // read in the data raw + for( j = 0; j*8 < tga_bits_per_pixel; ++j ) + { + raw_data[j] = get8u(s); + } + } + // convert raw to the intermediate format + switch( tga_bits_per_pixel ) + { + case 8: + // Luminous => RGBA + trans_data[0] = raw_data[0]; + trans_data[1] = raw_data[0]; + trans_data[2] = raw_data[0]; + trans_data[3] = 255; + break; + case 16: + // Luminous,Alpha => RGBA + trans_data[0] = raw_data[0]; + trans_data[1] = raw_data[0]; + trans_data[2] = raw_data[0]; + trans_data[3] = raw_data[1]; + break; + case 24: + // BGR => RGBA + trans_data[0] = raw_data[2]; + trans_data[1] = raw_data[1]; + trans_data[2] = raw_data[0]; + trans_data[3] = 255; + break; + case 32: + // BGRA => RGBA + trans_data[0] = raw_data[2]; + trans_data[1] = raw_data[1]; + trans_data[2] = raw_data[0]; + trans_data[3] = raw_data[3]; + break; + } + // clear the reading flag for the next pixel + read_next_pixel = 0; + } // end of reading a pixel + // convert to final format + switch( req_comp ) + { + case 1: + // RGBA => Luminance + tga_data[i*req_comp+0] = compute_y(trans_data[0],trans_data[1],trans_data[2]); + break; + case 2: + // RGBA => Luminance,Alpha + tga_data[i*req_comp+0] = compute_y(trans_data[0],trans_data[1],trans_data[2]); + tga_data[i*req_comp+1] = trans_data[3]; + break; + case 3: + // RGBA => RGB + tga_data[i*req_comp+0] = trans_data[0]; + tga_data[i*req_comp+1] = trans_data[1]; + tga_data[i*req_comp+2] = trans_data[2]; + break; + case 4: + // RGBA => RGBA + tga_data[i*req_comp+0] = trans_data[0]; + tga_data[i*req_comp+1] = trans_data[1]; + tga_data[i*req_comp+2] = trans_data[2]; + tga_data[i*req_comp+3] = trans_data[3]; + break; + } + // in case we're in RLE mode, keep counting down + --RLE_count; + } + // do I need to invert the image? + if( tga_inverted ) + { + for( j = 0; j*2 < tga_height; ++j ) + { + int index1 = j * tga_width * req_comp; + int index2 = (tga_height - 1 - j) * tga_width * req_comp; + for( i = tga_width * req_comp; i > 0; --i ) + { + unsigned char temp = tga_data[index1]; + tga_data[index1] = tga_data[index2]; + tga_data[index2] = temp; + ++index1; + ++index2; + } + } + } + // clear my palette, if I had one + if( tga_palette != NULL ) + { + free( tga_palette ); + } + // the things I do to get rid of an error message, and yet keep + // Microsoft's C compilers happy... [8^( + tga_palette_start = tga_palette_len = tga_palette_bits = + tga_x_origin = tga_y_origin = 0; + // OK, done + return tga_data; +} + +#ifndef STBI_NO_STDIO +stbi_uc *stbi_tga_load (char const *filename, int *x, int *y, int *comp, int req_comp) +{ + stbi_uc *data; + FILE *f = fopen(filename, "rb"); + if (!f) return NULL; + data = stbi_tga_load_from_file(f, x,y,comp,req_comp); + fclose(f); + return data; +} + +stbi_uc *stbi_tga_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_file(&s, f); + return tga_load(&s, x,y,comp,req_comp); +} +#endif + +stbi_uc *stbi_tga_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_mem(&s, buffer, len); + return tga_load(&s, x,y,comp,req_comp); +} + + +// ************************************************************************************************* +// Photoshop PSD loader -- PD by Thatcher Ulrich, integration by Nicholas Schulz, tweaked by STB + +static int psd_test(stbi *s) +{ + if (get32(s) != 0x38425053) return 0; // "8BPS" + else return 1; +} + +#ifndef STBI_NO_STDIO +int stbi_psd_test_file(FILE *f) +{ + stbi s; + int r,n = ftell(f); + start_file(&s, f); + r = psd_test(&s); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +int stbi_psd_test_memory(stbi_uc const *buffer, int len) +{ + stbi s; + start_mem(&s, buffer, len); + return psd_test(&s); +} + +static stbi_uc *psd_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + int pixelCount; + int channelCount, compression; + int channel, i, count, len; + int w,h; + uint8 *out; + + // Check identifier + if (get32(s) != 0x38425053) // "8BPS" + return epuc("not PSD", "Corrupt PSD image"); + + // Check file type version. + if (get16(s) != 1) + return epuc("wrong version", "Unsupported version of PSD image"); + + // Skip 6 reserved bytes. + skip(s, 6 ); + + // Read the number of channels (R, G, B, A, etc). + channelCount = get16(s); + if (channelCount < 0 || channelCount > 16) + return epuc("wrong channel count", "Unsupported number of channels in PSD image"); + + // Read the rows and columns of the image. + h = get32(s); + w = get32(s); + + // Make sure the depth is 8 bits. + if (get16(s) != 8) + return epuc("unsupported bit depth", "PSD bit depth is not 8 bit"); + + // Make sure the color mode is RGB. + // Valid options are: + // 0: Bitmap + // 1: Grayscale + // 2: Indexed color + // 3: RGB color + // 4: CMYK color + // 7: Multichannel + // 8: Duotone + // 9: Lab color + if (get16(s) != 3) + return epuc("wrong color format", "PSD is not in RGB color format"); + + // Skip the Mode Data. (It's the palette for indexed color; other info for other modes.) + skip(s,get32(s) ); + + // Skip the image resources. (resolution, pen tool paths, etc) + skip(s, get32(s) ); + + // Skip the reserved data. + skip(s, get32(s) ); + + // Find out if the data is compressed. + // Known values: + // 0: no compression + // 1: RLE compressed + compression = get16(s); + if (compression > 1) + return epuc("bad compression", "PSD has an unknown compression format"); + + // Create the destination image. + out = (stbi_uc *) malloc(4 * w*h); + if (!out) return epuc("outofmem", "Out of memory"); + pixelCount = w*h; + + // Initialize the data to zero. + //memset( out, 0, pixelCount * 4 ); + + // Finally, the image data. + if (compression) { + // RLE as used by .PSD and .TIFF + // Loop until you get the number of unpacked bytes you are expecting: + // Read the next source byte into n. + // If n is between 0 and 127 inclusive, copy the next n+1 bytes literally. + // Else if n is between -127 and -1 inclusive, copy the next byte -n+1 times. + // Else if n is 128, noop. + // Endloop + + // The RLE-compressed data is preceeded by a 2-byte data count for each row in the data, + // which we're going to just skip. + skip(s, h * channelCount * 2 ); + + // Read the RLE data by channel. + for (channel = 0; channel < 4; channel++) { + uint8 *p; + + p = out+channel; + if (channel >= channelCount) { + // Fill this channel with default data. + for (i = 0; i < pixelCount; i++) *p = (channel == 3 ? 255 : 0), p += 4; + } else { + // Read the RLE data. + count = 0; + while (count < pixelCount) { + len = get8(s); + if (len == 128) { + // No-op. + } else if (len < 128) { + // Copy next len+1 bytes literally. + len++; + count += len; + while (len) { + *p = get8(s); + p += 4; + len--; + } + } else if (len > 128) { + uint32 val; + // Next -len+1 bytes in the dest are replicated from next source byte. + // (Interpret len as a negative 8-bit int.) + len ^= 0x0FF; + len += 2; + val = get8(s); + count += len; + while (len) { + *p = val; + p += 4; + len--; + } + } + } + } + } + + } else { + // We're at the raw image data. It's each channel in order (Red, Green, Blue, Alpha, ...) + // where each channel consists of an 8-bit value for each pixel in the image. + + // Read the data by channel. + for (channel = 0; channel < 4; channel++) { + uint8 *p; + + p = out + channel; + if (channel > channelCount) { + // Fill this channel with default data. + for (i = 0; i < pixelCount; i++) *p = channel == 3 ? 255 : 0, p += 4; + } else { + // Read the data. + count = 0; + for (i = 0; i < pixelCount; i++) + *p = get8(s), p += 4; + } + } + } + + if (req_comp && req_comp != 4) { + out = convert_format(out, 4, req_comp, w, h); + if (out == NULL) return out; // convert_format frees input on failure + } + + if (comp) *comp = channelCount; + *y = h; + *x = w; + + return out; +} + +#ifndef STBI_NO_STDIO +stbi_uc *stbi_psd_load(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + stbi_uc *data; + FILE *f = fopen(filename, "rb"); + if (!f) return NULL; + data = stbi_psd_load_from_file(f, x,y,comp,req_comp); + fclose(f); + return data; +} + +stbi_uc *stbi_psd_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_file(&s, f); + return psd_load(&s, x,y,comp,req_comp); +} +#endif + +stbi_uc *stbi_psd_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_mem(&s, buffer, len); + return psd_load(&s, x,y,comp,req_comp); +} + + +// ************************************************************************************************* +// Radiance RGBE HDR loader +// originally by Nicolas Schulz +#ifndef STBI_NO_HDR +static int hdr_test(stbi *s) +{ + char *signature = "#?RADIANCE\n"; + int i; + for (i=0; signature[i]; ++i) + if (get8(s) != signature[i]) + return 0; + return 1; +} + +int stbi_hdr_test_memory(stbi_uc const *buffer, int len) +{ + stbi s; + start_mem(&s, buffer, len); + return hdr_test(&s); +} + +#ifndef STBI_NO_STDIO +int stbi_hdr_test_file(FILE *f) +{ + stbi s; + int r,n = ftell(f); + start_file(&s, f); + r = hdr_test(&s); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +#define HDR_BUFLEN 1024 +static char *hdr_gettoken(stbi *z, char *buffer) +{ + int len=0; + char *s = buffer, c = '\0'; + + c = get8(z); + + while (!at_eof(z) && c != '\n') { + buffer[len++] = c; + if (len == HDR_BUFLEN-1) { + // flush to end of line + while (!at_eof(z) && get8(z) != '\n') + ; + break; + } + c = get8(z); + } + + buffer[len] = 0; + return buffer; +} + +static void hdr_convert(float *output, stbi_uc *input, int req_comp) +{ + if( input[3] != 0 ) { + float f1; + // Exponent + f1 = (float) ldexp(1.0f, input[3] - (int)(128 + 8)); + if (req_comp <= 2) + output[0] = (input[0] + input[1] + input[2]) * f1 / 3; + else { + output[0] = input[0] * f1; + output[1] = input[1] * f1; + output[2] = input[2] * f1; + } + if (req_comp == 2) output[1] = 1; + if (req_comp == 4) output[3] = 1; + } else { + switch (req_comp) { + case 4: output[3] = 1; /* fallthrough */ + case 3: output[0] = output[1] = output[2] = 0; + break; + case 2: output[1] = 1; /* fallthrough */ + case 1: output[0] = 0; + break; + } + } +} + + +static float *hdr_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + char buffer[HDR_BUFLEN]; + char *token; + int valid = 0; + int width, height; + stbi_uc *scanline; + float *hdr_data; + int len; + unsigned char count, value; + int i, j, k, c1,c2, z; + + + // Check identifier + if (strcmp(hdr_gettoken(s,buffer), "#?RADIANCE") != 0) + return epf("not HDR", "Corrupt HDR image"); + + // Parse header + while(1) { + token = hdr_gettoken(s,buffer); + if (token[0] == 0) break; + if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1; + } + + if (!valid) return epf("unsupported format", "Unsupported HDR format"); + + // Parse width and height + // can't use sscanf() if we're not using stdio! + token = hdr_gettoken(s,buffer); + if (strncmp(token, "-Y ", 3)) return epf("unsupported data layout", "Unsupported HDR format"); + token += 3; + height = strtol(token, &token, 10); + while (*token == ' ') ++token; + if (strncmp(token, "+X ", 3)) return epf("unsupported data layout", "Unsupported HDR format"); + token += 3; + width = strtol(token, NULL, 10); + + *x = width; + *y = height; + + *comp = 3; + if (req_comp == 0) req_comp = 3; + + // Read data + hdr_data = (float *) malloc(height * width * req_comp * sizeof(float)); + + // Load image data + // image data is stored as some number of sca + if( width < 8 || width >= 32768) { + // Read flat data + for (j=0; j < height; ++j) { + for (i=0; i < width; ++i) { + stbi_uc rgbe[4]; + main_decode_loop: + getn(s, rgbe, 4); + hdr_convert(hdr_data + j * width * req_comp + i * req_comp, rgbe, req_comp); + } + } + } else { + // Read RLE-encoded data + scanline = NULL; + + for (j = 0; j < height; ++j) { + c1 = get8(s); + c2 = get8(s); + len = get8(s); + if (c1 != 2 || c2 != 2 || (len & 0x80)) { + // not run-length encoded, so we have to actually use THIS data as a decoded + // pixel (note this can't be a valid pixel--one of RGB must be >= 128) + stbi_uc rgbe[4] = { c1,c2,len, get8(s) }; + hdr_convert(hdr_data, rgbe, req_comp); + i = 1; + j = 0; + free(scanline); + goto main_decode_loop; // yes, this is fucking insane; blame the fucking insane format + } + len <<= 8; + len |= get8(s); + if (len != width) { free(hdr_data); free(scanline); return epf("invalid decoded scanline length", "corrupt HDR"); } + if (scanline == NULL) scanline = (stbi_uc *) malloc(width * 4); + + for (k = 0; k < 4; ++k) { + i = 0; + while (i < width) { + count = get8(s); + if (count > 128) { + // Run + value = get8(s); + count -= 128; + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = value; + } else { + // Dump + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = get8(s); + } + } + } + for (i=0; i < width; ++i) + hdr_convert(hdr_data+(j*width + i)*req_comp, scanline + i*4, req_comp); + } + free(scanline); + } + + return hdr_data; +} + +#ifndef STBI_NO_STDIO +float *stbi_hdr_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_file(&s,f); + return hdr_load(&s,x,y,comp,req_comp); +} +#endif + +float *stbi_hdr_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_mem(&s,buffer, len); + return hdr_load(&s,x,y,comp,req_comp); +} + +#endif // STBI_NO_HDR + +/////////////////////// write image /////////////////////// + +#ifndef STBI_NO_WRITE + +static void write8(FILE *f, int x) { uint8 z = (uint8) x; fwrite(&z,1,1,f); } + +static void writefv(FILE *f, char *fmt, va_list v) +{ + while (*fmt) { + switch (*fmt++) { + case ' ': break; + case '1': { uint8 x = va_arg(v, int); write8(f,x); break; } + case '2': { int16 x = va_arg(v, int); write8(f,x); write8(f,x>>8); break; } + case '4': { int32 x = va_arg(v, int); write8(f,x); write8(f,x>>8); write8(f,x>>16); write8(f,x>>24); break; } + default: + assert(0); + va_end(v); + return; + } + } +} + +static void writef(FILE *f, char *fmt, ...) +{ + va_list v; + va_start(v, fmt); + writefv(f,fmt,v); + va_end(v); +} + +static void write_pixels(FILE *f, int rgb_dir, int vdir, int x, int y, int comp, void *data, int write_alpha, int scanline_pad) +{ + uint8 bg[3] = { 255, 0, 255}, px[3]; + uint32 zero = 0; + int i,j,k, j_end; + + if (vdir < 0) + j_end = -1, j = y-1; + else + j_end = y, j = 0; + + for (; j != j_end; j += vdir) { + for (i=0; i < x; ++i) { + uint8 *d = (uint8 *) data + (j*x+i)*comp; + if (write_alpha < 0) + fwrite(&d[comp-1], 1, 1, f); + switch (comp) { + case 1: + case 2: writef(f, "111", d[0],d[0],d[0]); + break; + case 4: + if (!write_alpha) { + for (k=0; k < 3; ++k) + px[k] = bg[k] + ((d[k] - bg[k]) * d[3])/255; + writef(f, "111", px[1-rgb_dir],px[1],px[1+rgb_dir]); + break; + } + /* FALLTHROUGH */ + case 3: + writef(f, "111", d[1-rgb_dir],d[1],d[1+rgb_dir]); + break; + } + if (write_alpha > 0) + fwrite(&d[comp-1], 1, 1, f); + } + fwrite(&zero,scanline_pad,1,f); + } +} + +static int outfile(char const *filename, int rgb_dir, int vdir, int x, int y, int comp, void *data, int alpha, int pad, char *fmt, ...) +{ + FILE *f = fopen(filename, "wb"); + if (f) { + va_list v; + va_start(v, fmt); + writefv(f, fmt, v); + va_end(v); + write_pixels(f,rgb_dir,vdir,x,y,comp,data,alpha,pad); + fclose(f); + } + return f != NULL; +} + +int stbi_write_bmp(char const *filename, int x, int y, int comp, void *data) +{ + int pad = (-x*3) & 3; + return outfile(filename,-1,-1,x,y,comp,data,0,pad, + "11 4 22 4" "4 44 22 444444", + 'B', 'M', 14+40+(x*3+pad)*y, 0,0, 14+40, // file header + 40, x,y, 1,24, 0,0,0,0,0,0); // bitmap header +} + +int stbi_write_tga(char const *filename, int x, int y, int comp, void *data) +{ + int has_alpha = !(comp & 1); + return outfile(filename, -1,-1, x, y, comp, data, has_alpha, 0, + "111 221 2222 11", 0,0,2, 0,0,0, 0,0,x,y, 24+8*has_alpha, 8*has_alpha); +} + +// any other image formats that do interleaved rgb data? +// PNG: requires adler32,crc32 -- significant amount of code +// PSD: no, channels output separately +// TIFF: no, stripwise-interleaved... i think + +#endif // STBI_NO_WRITE + +#endif // STBI_HEADER_FILE_ONLY + diff --git a/lib/soil/soil/stb_image_aug.c b/lib/soil/soil/stb_image_aug.c new file mode 100644 index 0000000..2fa7233 --- /dev/null +++ b/lib/soil/soil/stb_image_aug.c @@ -0,0 +1,3682 @@ +/* stbi-1.16 - public domain JPEG/PNG reader - http://nothings.org/stb_image.c + when you control the images you're loading + + QUICK NOTES: + Primarily of interest to game developers and other people who can + avoid problematic images and only need the trivial interface + + JPEG baseline (no JPEG progressive, no oddball channel decimations) + PNG non-interlaced + BMP non-1bpp, non-RLE + TGA (not sure what subset, if a subset) + PSD (composited view only, no extra channels) + HDR (radiance rgbE format) + writes BMP,TGA (define STBI_NO_WRITE to remove code) + decoded from memory or through stdio FILE (define STBI_NO_STDIO to remove code) + supports installable dequantizing-IDCT, YCbCr-to-RGB conversion (define STBI_SIMD) + + TODO: + stbi_info_* + + history: + 1.16 major bugfix - convert_format converted one too many pixels + 1.15 initialize some fields for thread safety + 1.14 fix threadsafe conversion bug; header-file-only version (#define STBI_HEADER_FILE_ONLY before including) + 1.13 threadsafe + 1.12 const qualifiers in the API + 1.11 Support installable IDCT, colorspace conversion routines + 1.10 Fixes for 64-bit (don't use "unsigned long") + optimized upsampling by Fabian "ryg" Giesen + 1.09 Fix format-conversion for PSD code (bad global variables!) + 1.08 Thatcher Ulrich's PSD code integrated by Nicolas Schulz + 1.07 attempt to fix C++ warning/errors again + 1.06 attempt to fix C++ warning/errors again + 1.05 fix TGA loading to return correct *comp and use good luminance calc + 1.04 default float alpha is 1, not 255; use 'void *' for stbi_image_free + 1.03 bugfixes to STBI_NO_STDIO, STBI_NO_HDR + 1.02 support for (subset of) HDR files, float interface for preferred access to them + 1.01 fix bug: possible bug in handling right-side up bmps... not sure + fix bug: the stbi_bmp_load() and stbi_tga_load() functions didn't work at all + 1.00 interface to zlib that skips zlib header + 0.99 correct handling of alpha in palette + 0.98 TGA loader by lonesock; dynamically add loaders (untested) + 0.97 jpeg errors on too large a file; also catch another malloc failure + 0.96 fix detection of invalid v value - particleman@mollyrocket forum + 0.95 during header scan, seek to markers in case of padding + 0.94 STBI_NO_STDIO to disable stdio usage; rename all #defines the same + 0.93 handle jpegtran output; verbose errors + 0.92 read 4,8,16,24,32-bit BMP files of several formats + 0.91 output 24-bit Windows 3.0 BMP files + 0.90 fix a few more warnings; bump version number to approach 1.0 + 0.61 bugfixes due to Marc LeBlanc, Christopher Lloyd + 0.60 fix compiling as c++ + 0.59 fix warnings: merge Dave Moore's -Wall fixes + 0.58 fix bug: zlib uncompressed mode len/nlen was wrong endian + 0.57 fix bug: jpg last huffman symbol before marker was >9 bits but less + than 16 available + 0.56 fix bug: zlib uncompressed mode len vs. nlen + 0.55 fix bug: restart_interval not initialized to 0 + 0.54 allow NULL for 'int *comp' + 0.53 fix bug in png 3->4; speedup png decoding + 0.52 png handles req_comp=3,4 directly; minor cleanup; jpeg comments + 0.51 obey req_comp requests, 1-component jpegs return as 1-component, + on 'test' only check type, not whether we support this variant +*/ + +#include "stb_image_aug.h" + +#ifndef STBI_NO_HDR +#include // ldexp +#include // strcmp +#endif + +#ifndef STBI_NO_STDIO +#include +#endif +#include +#include +#include +#include + +#ifndef _MSC_VER + #ifdef __cplusplus + #define __forceinline inline + #else + #define __forceinline + #endif +#endif + + +// implementation: +typedef unsigned char uint8; +typedef unsigned short uint16; +typedef signed short int16; +typedef unsigned int uint32; +typedef signed int int32; +typedef unsigned int uint; + +// should produce compiler error if size is wrong +typedef unsigned char validate_uint32[sizeof(uint32)==4]; + +#if defined(STBI_NO_STDIO) && !defined(STBI_NO_WRITE) +#define STBI_NO_WRITE +#endif + +#ifndef STBI_NO_DDS +#include "stbi_DDS_aug.h" +#endif + +// I (JLD) want full messages for SOIL +#define STBI_FAILURE_USERMSG 1 + +////////////////////////////////////////////////////////////////////////////// +// +// Generic API that works on all image types +// + +// this is not threadsafe +static char *failure_reason; + +char *stbi_failure_reason(void) +{ + return failure_reason; +} + +static int e(char *str) +{ + failure_reason = str; + return 0; +} + +#ifdef STBI_NO_FAILURE_STRINGS + #define e(x,y) 0 +#elif defined(STBI_FAILURE_USERMSG) + #define e(x,y) e(y) +#else + #define e(x,y) e(x) +#endif + +#define epf(x,y) ((float *) (e(x,y)?NULL:NULL)) +#define epuc(x,y) ((unsigned char *) (e(x,y)?NULL:NULL)) + +void stbi_image_free(void *retval_from_stbi_load) +{ + free(retval_from_stbi_load); +} + +#define MAX_LOADERS 32 +stbi_loader *loaders[MAX_LOADERS]; +static int max_loaders = 0; + +int stbi_register_loader(stbi_loader *loader) +{ + int i; + for (i=0; i < MAX_LOADERS; ++i) { + // already present? + if (loaders[i] == loader) + return 1; + // end of the list? + if (loaders[i] == NULL) { + loaders[i] = loader; + max_loaders = i+1; + return 1; + } + } + // no room for it + return 0; +} + +#ifndef STBI_NO_HDR +static float *ldr_to_hdr(stbi_uc *data, int x, int y, int comp); +static stbi_uc *hdr_to_ldr(float *data, int x, int y, int comp); +#endif + +#ifndef STBI_NO_STDIO +unsigned char *stbi_load(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = fopen(filename, "rb"); + unsigned char *result; + if (!f) return epuc("can't fopen", "Unable to open file"); + result = stbi_load_from_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} + +unsigned char *stbi_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + int i; + if (stbi_jpeg_test_file(f)) + return stbi_jpeg_load_from_file(f,x,y,comp,req_comp); + if (stbi_png_test_file(f)) + return stbi_png_load_from_file(f,x,y,comp,req_comp); + if (stbi_bmp_test_file(f)) + return stbi_bmp_load_from_file(f,x,y,comp,req_comp); + if (stbi_psd_test_file(f)) + return stbi_psd_load_from_file(f,x,y,comp,req_comp); + #ifndef STBI_NO_DDS + if (stbi_dds_test_file(f)) + return stbi_dds_load_from_file(f,x,y,comp,req_comp); + #endif + #ifndef STBI_NO_HDR + if (stbi_hdr_test_file(f)) { + float *hdr = stbi_hdr_load_from_file(f, x,y,comp,req_comp); + return hdr_to_ldr(hdr, *x, *y, req_comp ? req_comp : *comp); + } + #endif + for (i=0; i < max_loaders; ++i) + if (loaders[i]->test_file(f)) + return loaders[i]->load_from_file(f,x,y,comp,req_comp); + // test tga last because it's a crappy test! + if (stbi_tga_test_file(f)) + return stbi_tga_load_from_file(f,x,y,comp,req_comp); + return epuc("unknown image type", "Image not of any known type, or corrupt"); +} +#endif + +unsigned char *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + int i; + if (stbi_jpeg_test_memory(buffer,len)) + return stbi_jpeg_load_from_memory(buffer,len,x,y,comp,req_comp); + if (stbi_png_test_memory(buffer,len)) + return stbi_png_load_from_memory(buffer,len,x,y,comp,req_comp); + if (stbi_bmp_test_memory(buffer,len)) + return stbi_bmp_load_from_memory(buffer,len,x,y,comp,req_comp); + if (stbi_psd_test_memory(buffer,len)) + return stbi_psd_load_from_memory(buffer,len,x,y,comp,req_comp); + #ifndef STBI_NO_DDS + if (stbi_dds_test_memory(buffer,len)) + return stbi_dds_load_from_memory(buffer,len,x,y,comp,req_comp); + #endif + #ifndef STBI_NO_HDR + if (stbi_hdr_test_memory(buffer, len)) { + float *hdr = stbi_hdr_load_from_memory(buffer, len,x,y,comp,req_comp); + return hdr_to_ldr(hdr, *x, *y, req_comp ? req_comp : *comp); + } + #endif + for (i=0; i < max_loaders; ++i) + if (loaders[i]->test_memory(buffer,len)) + return loaders[i]->load_from_memory(buffer,len,x,y,comp,req_comp); + // test tga last because it's a crappy test! + if (stbi_tga_test_memory(buffer,len)) + return stbi_tga_load_from_memory(buffer,len,x,y,comp,req_comp); + return epuc("unknown image type", "Image not of any known type, or corrupt"); +} + +#ifndef STBI_NO_HDR + +#ifndef STBI_NO_STDIO +float *stbi_loadf(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = fopen(filename, "rb"); + float *result; + if (!f) return epf("can't fopen", "Unable to open file"); + result = stbi_loadf_from_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} + +float *stbi_loadf_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *data; + #ifndef STBI_NO_HDR + if (stbi_hdr_test_file(f)) + return stbi_hdr_load_from_file(f,x,y,comp,req_comp); + #endif + data = stbi_load_from_file(f, x, y, comp, req_comp); + if (data) + return ldr_to_hdr(data, *x, *y, req_comp ? req_comp : *comp); + return epf("unknown image type", "Image not of any known type, or corrupt"); +} +#endif + +float *stbi_loadf_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi_uc *data; + #ifndef STBI_NO_HDR + if (stbi_hdr_test_memory(buffer, len)) + return stbi_hdr_load_from_memory(buffer, len,x,y,comp,req_comp); + #endif + data = stbi_load_from_memory(buffer, len, x, y, comp, req_comp); + if (data) + return ldr_to_hdr(data, *x, *y, req_comp ? req_comp : *comp); + return epf("unknown image type", "Image not of any known type, or corrupt"); +} +#endif + +// these is-hdr-or-not is defined independent of whether STBI_NO_HDR is +// defined, for API simplicity; if STBI_NO_HDR is defined, it always +// reports false! + +int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len) +{ + #ifndef STBI_NO_HDR + return stbi_hdr_test_memory(buffer, len); + #else + return 0; + #endif +} + +#ifndef STBI_NO_STDIO +extern int stbi_is_hdr (char const *filename) +{ + FILE *f = fopen(filename, "rb"); + int result=0; + if (f) { + result = stbi_is_hdr_from_file(f); + fclose(f); + } + return result; +} + +extern int stbi_is_hdr_from_file(FILE *f) +{ + #ifndef STBI_NO_HDR + return stbi_hdr_test_file(f); + #else + return 0; + #endif +} + +#endif + +// @TODO: get image dimensions & components without fully decoding +#ifndef STBI_NO_STDIO +extern int stbi_info (char const *filename, int *x, int *y, int *comp); +extern int stbi_info_from_file (FILE *f, int *x, int *y, int *comp); +#endif +extern int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp); + +#ifndef STBI_NO_HDR +static float h2l_gamma_i=1.0f/2.2f, h2l_scale_i=1.0f; +static float l2h_gamma=2.2f, l2h_scale=1.0f; + +void stbi_hdr_to_ldr_gamma(float gamma) { h2l_gamma_i = 1/gamma; } +void stbi_hdr_to_ldr_scale(float scale) { h2l_scale_i = 1/scale; } + +void stbi_ldr_to_hdr_gamma(float gamma) { l2h_gamma = gamma; } +void stbi_ldr_to_hdr_scale(float scale) { l2h_scale = scale; } +#endif + + +////////////////////////////////////////////////////////////////////////////// +// +// Common code used by all image loaders +// + +enum +{ + SCAN_load=0, + SCAN_type, + SCAN_header, +}; + +typedef struct +{ + uint32 img_x, img_y; + int img_n, img_out_n; + + #ifndef STBI_NO_STDIO + FILE *img_file; + #endif + uint8 *img_buffer, *img_buffer_end; +} stbi; + +#ifndef STBI_NO_STDIO +static void start_file(stbi *s, FILE *f) +{ + s->img_file = f; +} +#endif + +static void start_mem(stbi *s, uint8 const *buffer, int len) +{ +#ifndef STBI_NO_STDIO + s->img_file = NULL; +#endif + s->img_buffer = (uint8 *) buffer; + s->img_buffer_end = (uint8 *) buffer+len; +} + +__forceinline static int get8(stbi *s) +{ +#ifndef STBI_NO_STDIO + if (s->img_file) { + int c = fgetc(s->img_file); + return c == EOF ? 0 : c; + } +#endif + if (s->img_buffer < s->img_buffer_end) + return *s->img_buffer++; + return 0; +} + +__forceinline static int at_eof(stbi *s) +{ +#ifndef STBI_NO_STDIO + if (s->img_file) + return feof(s->img_file); +#endif + return s->img_buffer >= s->img_buffer_end; +} + +__forceinline static uint8 get8u(stbi *s) +{ + return (uint8) get8(s); +} + +static void skip(stbi *s, int n) +{ +#ifndef STBI_NO_STDIO + if (s->img_file) + fseek(s->img_file, n, SEEK_CUR); + else +#endif + s->img_buffer += n; +} + +static int get16(stbi *s) +{ + int z = get8(s); + return (z << 8) + get8(s); +} + +static uint32 get32(stbi *s) +{ + uint32 z = get16(s); + return (z << 16) + get16(s); +} + +static int get16le(stbi *s) +{ + int z = get8(s); + return z + (get8(s) << 8); +} + +static uint32 get32le(stbi *s) +{ + uint32 z = get16le(s); + return z + (get16le(s) << 16); +} + +static void getn(stbi *s, stbi_uc *buffer, int n) +{ +#ifndef STBI_NO_STDIO + if (s->img_file) { + fread(buffer, 1, n, s->img_file); + return; + } +#endif + memcpy(buffer, s->img_buffer, n); + s->img_buffer += n; +} + +////////////////////////////////////////////////////////////////////////////// +// +// generic converter from built-in img_n to req_comp +// individual types do this automatically as much as possible (e.g. jpeg +// does all cases internally since it needs to colorspace convert anyway, +// and it never has alpha, so very few cases ). png can automatically +// interleave an alpha=255 channel, but falls back to this for other cases +// +// assume data buffer is malloced, so malloc a new one and free that one +// only failure mode is malloc failing + +static uint8 compute_y(int r, int g, int b) +{ + return (uint8) (((r*77) + (g*150) + (29*b)) >> 8); +} + +static unsigned char *convert_format(unsigned char *data, int img_n, int req_comp, uint x, uint y) +{ + int i,j; + unsigned char *good; + + if (req_comp == img_n) return data; + assert(req_comp >= 1 && req_comp <= 4); + + good = (unsigned char *) malloc(req_comp * x * y); + if (good == NULL) { + free(data); + return epuc("outofmem", "Out of memory"); + } + + for (j=0; j < (int) y; ++j) { + unsigned char *src = data + j * x * img_n ; + unsigned char *dest = good + j * x * req_comp; + + #define COMBO(a,b) ((a)*8+(b)) + #define CASE(a,b) case COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b) + // convert source image with img_n components to one with req_comp components; + // avoid switch per pixel, so use switch per scanline and massive macros + switch(COMBO(img_n, req_comp)) { + CASE(1,2) dest[0]=src[0], dest[1]=255; break; + CASE(1,3) dest[0]=dest[1]=dest[2]=src[0]; break; + CASE(1,4) dest[0]=dest[1]=dest[2]=src[0], dest[3]=255; break; + CASE(2,1) dest[0]=src[0]; break; + CASE(2,3) dest[0]=dest[1]=dest[2]=src[0]; break; + CASE(2,4) dest[0]=dest[1]=dest[2]=src[0], dest[3]=src[1]; break; + CASE(3,4) dest[0]=src[0],dest[1]=src[1],dest[2]=src[2],dest[3]=255; break; + CASE(3,1) dest[0]=compute_y(src[0],src[1],src[2]); break; + CASE(3,2) dest[0]=compute_y(src[0],src[1],src[2]), dest[1] = 255; break; + CASE(4,1) dest[0]=compute_y(src[0],src[1],src[2]); break; + CASE(4,2) dest[0]=compute_y(src[0],src[1],src[2]), dest[1] = src[3]; break; + CASE(4,3) dest[0]=src[0],dest[1]=src[1],dest[2]=src[2]; break; + default: assert(0); + } + #undef CASE + } + + free(data); + return good; +} + +#ifndef STBI_NO_HDR +static float *ldr_to_hdr(stbi_uc *data, int x, int y, int comp) +{ + int i,k,n; + float *output = (float *) malloc(x * y * comp * sizeof(float)); + if (output == NULL) { free(data); return epf("outofmem", "Out of memory"); } + // compute number of non-alpha components + if (comp & 1) n = comp; else n = comp-1; + for (i=0; i < x*y; ++i) { + for (k=0; k < n; ++k) { + output[i*comp + k] = (float) pow(data[i*comp+k]/255.0f, l2h_gamma) * l2h_scale; + } + if (k < comp) output[i*comp + k] = data[i*comp+k]/255.0f; + } + free(data); + return output; +} + +#define float2int(x) ((int) (x)) +static stbi_uc *hdr_to_ldr(float *data, int x, int y, int comp) +{ + int i,k,n; + stbi_uc *output = (stbi_uc *) malloc(x * y * comp); + if (output == NULL) { free(data); return epuc("outofmem", "Out of memory"); } + // compute number of non-alpha components + if (comp & 1) n = comp; else n = comp-1; + for (i=0; i < x*y; ++i) { + for (k=0; k < n; ++k) { + float z = (float) pow(data[i*comp+k]*h2l_scale_i, h2l_gamma_i) * 255 + 0.5f; + if (z < 0) z = 0; + if (z > 255) z = 255; + output[i*comp + k] = float2int(z); + } + if (k < comp) { + float z = data[i*comp+k] * 255 + 0.5f; + if (z < 0) z = 0; + if (z > 255) z = 255; + output[i*comp + k] = float2int(z); + } + } + free(data); + return output; +} +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// "baseline" JPEG/JFIF decoder (not actually fully baseline implementation) +// +// simple implementation +// - channel subsampling of at most 2 in each dimension +// - doesn't support delayed output of y-dimension +// - simple interface (only one output format: 8-bit interleaved RGB) +// - doesn't try to recover corrupt jpegs +// - doesn't allow partial loading, loading multiple at once +// - still fast on x86 (copying globals into locals doesn't help x86) +// - allocates lots of intermediate memory (full size of all components) +// - non-interleaved case requires this anyway +// - allows good upsampling (see next) +// high-quality +// - upsampled channels are bilinearly interpolated, even across blocks +// - quality integer IDCT derived from IJG's 'slow' +// performance +// - fast huffman; reasonable integer IDCT +// - uses a lot of intermediate memory, could cache poorly +// - load http://nothings.org/remote/anemones.jpg 3 times on 2.8Ghz P4 +// stb_jpeg: 1.34 seconds (MSVC6, default release build) +// stb_jpeg: 1.06 seconds (MSVC6, processor = Pentium Pro) +// IJL11.dll: 1.08 seconds (compiled by intel) +// IJG 1998: 0.98 seconds (MSVC6, makefile provided by IJG) +// IJG 1998: 0.95 seconds (MSVC6, makefile + proc=PPro) + +// huffman decoding acceleration +#define FAST_BITS 9 // larger handles more cases; smaller stomps less cache + +typedef struct +{ + uint8 fast[1 << FAST_BITS]; + // weirdly, repacking this into AoS is a 10% speed loss, instead of a win + uint16 code[256]; + uint8 values[256]; + uint8 size[257]; + unsigned int maxcode[18]; + int delta[17]; // old 'firstsymbol' - old 'firstcode' +} huffman; + +typedef struct +{ + #if STBI_SIMD + unsigned short dequant2[4][64]; + #endif + stbi s; + huffman huff_dc[4]; + huffman huff_ac[4]; + uint8 dequant[4][64]; + +// sizes for components, interleaved MCUs + int img_h_max, img_v_max; + int img_mcu_x, img_mcu_y; + int img_mcu_w, img_mcu_h; + +// definition of jpeg image component + struct + { + int id; + int h,v; + int tq; + int hd,ha; + int dc_pred; + + int x,y,w2,h2; + uint8 *data; + void *raw_data; + uint8 *linebuf; + } img_comp[4]; + + uint32 code_buffer; // jpeg entropy-coded buffer + int code_bits; // number of valid bits + unsigned char marker; // marker seen while filling entropy buffer + int nomore; // flag if we saw a marker so must stop + + int scan_n, order[4]; + int restart_interval, todo; +} jpeg; + +static int build_huffman(huffman *h, int *count) +{ + int i,j,k=0,code; + // build size list for each symbol (from JPEG spec) + for (i=0; i < 16; ++i) + for (j=0; j < count[i]; ++j) + h->size[k++] = (uint8) (i+1); + h->size[k] = 0; + + // compute actual symbols (from jpeg spec) + code = 0; + k = 0; + for(j=1; j <= 16; ++j) { + // compute delta to add to code to compute symbol id + h->delta[j] = k - code; + if (h->size[k] == j) { + while (h->size[k] == j) + h->code[k++] = (uint16) (code++); + if (code-1 >= (1 << j)) return e("bad code lengths","Corrupt JPEG"); + } + // compute largest code + 1 for this size, preshifted as needed later + h->maxcode[j] = code << (16-j); + code <<= 1; + } + h->maxcode[j] = 0xffffffff; + + // build non-spec acceleration table; 255 is flag for not-accelerated + memset(h->fast, 255, 1 << FAST_BITS); + for (i=0; i < k; ++i) { + int s = h->size[i]; + if (s <= FAST_BITS) { + int c = h->code[i] << (FAST_BITS-s); + int m = 1 << (FAST_BITS-s); + for (j=0; j < m; ++j) { + h->fast[c+j] = (uint8) i; + } + } + } + return 1; +} + +static void grow_buffer_unsafe(jpeg *j) +{ + do { + int b = j->nomore ? 0 : get8(&j->s); + if (b == 0xff) { + int c = get8(&j->s); + if (c != 0) { + j->marker = (unsigned char) c; + j->nomore = 1; + return; + } + } + j->code_buffer = (j->code_buffer << 8) | b; + j->code_bits += 8; + } while (j->code_bits <= 24); +} + +// (1 << n) - 1 +static uint32 bmask[17]={0,1,3,7,15,31,63,127,255,511,1023,2047,4095,8191,16383,32767,65535}; + +// decode a jpeg huffman value from the bitstream +__forceinline static int decode(jpeg *j, huffman *h) +{ + unsigned int temp; + int c,k; + + if (j->code_bits < 16) grow_buffer_unsafe(j); + + // look at the top FAST_BITS and determine what symbol ID it is, + // if the code is <= FAST_BITS + c = (j->code_buffer >> (j->code_bits - FAST_BITS)) & ((1 << FAST_BITS)-1); + k = h->fast[c]; + if (k < 255) { + if (h->size[k] > j->code_bits) + return -1; + j->code_bits -= h->size[k]; + return h->values[k]; + } + + // naive test is to shift the code_buffer down so k bits are + // valid, then test against maxcode. To speed this up, we've + // preshifted maxcode left so that it has (16-k) 0s at the + // end; in other words, regardless of the number of bits, it + // wants to be compared against something shifted to have 16; + // that way we don't need to shift inside the loop. + if (j->code_bits < 16) + temp = (j->code_buffer << (16 - j->code_bits)) & 0xffff; + else + temp = (j->code_buffer >> (j->code_bits - 16)) & 0xffff; + for (k=FAST_BITS+1 ; ; ++k) + if (temp < h->maxcode[k]) + break; + if (k == 17) { + // error! code not found + j->code_bits -= 16; + return -1; + } + + if (k > j->code_bits) + return -1; + + // convert the huffman code to the symbol id + c = ((j->code_buffer >> (j->code_bits - k)) & bmask[k]) + h->delta[k]; + assert((((j->code_buffer) >> (j->code_bits - h->size[c])) & bmask[h->size[c]]) == h->code[c]); + + // convert the id to a symbol + j->code_bits -= k; + return h->values[c]; +} + +// combined JPEG 'receive' and JPEG 'extend', since baseline +// always extends everything it receives. +__forceinline static int extend_receive(jpeg *j, int n) +{ + unsigned int m = 1 << (n-1); + unsigned int k; + if (j->code_bits < n) grow_buffer_unsafe(j); + k = (j->code_buffer >> (j->code_bits - n)) & bmask[n]; + j->code_bits -= n; + // the following test is probably a random branch that won't + // predict well. I tried to table accelerate it but failed. + // maybe it's compiling as a conditional move? + if (k < m) + return (-1 << n) + k + 1; + else + return k; +} + +// given a value that's at position X in the zigzag stream, +// where does it appear in the 8x8 matrix coded as row-major? +static uint8 dezigzag[64+15] = +{ + 0, 1, 8, 16, 9, 2, 3, 10, + 17, 24, 32, 25, 18, 11, 4, 5, + 12, 19, 26, 33, 40, 48, 41, 34, + 27, 20, 13, 6, 7, 14, 21, 28, + 35, 42, 49, 56, 57, 50, 43, 36, + 29, 22, 15, 23, 30, 37, 44, 51, + 58, 59, 52, 45, 38, 31, 39, 46, + 53, 60, 61, 54, 47, 55, 62, 63, + // let corrupt input sample past end + 63, 63, 63, 63, 63, 63, 63, 63, + 63, 63, 63, 63, 63, 63, 63 +}; + +// decode one 64-entry block-- +static int decode_block(jpeg *j, short data[64], huffman *hdc, huffman *hac, int b) +{ + int diff,dc,k; + int t = decode(j, hdc); + if (t < 0) return e("bad huffman code","Corrupt JPEG"); + + // 0 all the ac values now so we can do it 32-bits at a time + memset(data,0,64*sizeof(data[0])); + + diff = t ? extend_receive(j, t) : 0; + dc = j->img_comp[b].dc_pred + diff; + j->img_comp[b].dc_pred = dc; + data[0] = (short) dc; + + // decode AC components, see JPEG spec + k = 1; + do { + int r,s; + int rs = decode(j, hac); + if (rs < 0) return e("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (rs != 0xf0) break; // end block + k += 16; + } else { + k += r; + // decode into unzigzag'd location + data[dezigzag[k++]] = (short) extend_receive(j,s); + } + } while (k < 64); + return 1; +} + +// take a -128..127 value and clamp it and convert to 0..255 +__forceinline static uint8 clamp(int x) +{ + x += 128; + // trick to use a single test to catch both cases + if ((unsigned int) x > 255) { + if (x < 0) return 0; + if (x > 255) return 255; + } + return (uint8) x; +} + +#define f2f(x) (int) (((x) * 4096 + 0.5)) +#define fsh(x) ((x) << 12) + +// derived from jidctint -- DCT_ISLOW +#define IDCT_1D(s0,s1,s2,s3,s4,s5,s6,s7) \ + int t0,t1,t2,t3,p1,p2,p3,p4,p5,x0,x1,x2,x3; \ + p2 = s2; \ + p3 = s6; \ + p1 = (p2+p3) * f2f(0.5411961f); \ + t2 = p1 + p3*f2f(-1.847759065f); \ + t3 = p1 + p2*f2f( 0.765366865f); \ + p2 = s0; \ + p3 = s4; \ + t0 = fsh(p2+p3); \ + t1 = fsh(p2-p3); \ + x0 = t0+t3; \ + x3 = t0-t3; \ + x1 = t1+t2; \ + x2 = t1-t2; \ + t0 = s7; \ + t1 = s5; \ + t2 = s3; \ + t3 = s1; \ + p3 = t0+t2; \ + p4 = t1+t3; \ + p1 = t0+t3; \ + p2 = t1+t2; \ + p5 = (p3+p4)*f2f( 1.175875602f); \ + t0 = t0*f2f( 0.298631336f); \ + t1 = t1*f2f( 2.053119869f); \ + t2 = t2*f2f( 3.072711026f); \ + t3 = t3*f2f( 1.501321110f); \ + p1 = p5 + p1*f2f(-0.899976223f); \ + p2 = p5 + p2*f2f(-2.562915447f); \ + p3 = p3*f2f(-1.961570560f); \ + p4 = p4*f2f(-0.390180644f); \ + t3 += p1+p4; \ + t2 += p2+p3; \ + t1 += p2+p4; \ + t0 += p1+p3; + +#if !STBI_SIMD +// .344 seconds on 3*anemones.jpg +static void idct_block(uint8 *out, int out_stride, short data[64], uint8 *dequantize) +{ + int i,val[64],*v=val; + uint8 *o,*dq = dequantize; + short *d = data; + + // columns + for (i=0; i < 8; ++i,++d,++dq, ++v) { + // if all zeroes, shortcut -- this avoids dequantizing 0s and IDCTing + if (d[ 8]==0 && d[16]==0 && d[24]==0 && d[32]==0 + && d[40]==0 && d[48]==0 && d[56]==0) { + // no shortcut 0 seconds + // (1|2|3|4|5|6|7)==0 0 seconds + // all separate -0.047 seconds + // 1 && 2|3 && 4|5 && 6|7: -0.047 seconds + int dcterm = d[0] * dq[0] << 2; + v[0] = v[8] = v[16] = v[24] = v[32] = v[40] = v[48] = v[56] = dcterm; + } else { + IDCT_1D(d[ 0]*dq[ 0],d[ 8]*dq[ 8],d[16]*dq[16],d[24]*dq[24], + d[32]*dq[32],d[40]*dq[40],d[48]*dq[48],d[56]*dq[56]) + // constants scaled things up by 1<<12; let's bring them back + // down, but keep 2 extra bits of precision + x0 += 512; x1 += 512; x2 += 512; x3 += 512; + v[ 0] = (x0+t3) >> 10; + v[56] = (x0-t3) >> 10; + v[ 8] = (x1+t2) >> 10; + v[48] = (x1-t2) >> 10; + v[16] = (x2+t1) >> 10; + v[40] = (x2-t1) >> 10; + v[24] = (x3+t0) >> 10; + v[32] = (x3-t0) >> 10; + } + } + + for (i=0, v=val, o=out; i < 8; ++i,v+=8,o+=out_stride) { + // no fast case since the first 1D IDCT spread components out + IDCT_1D(v[0],v[1],v[2],v[3],v[4],v[5],v[6],v[7]) + // constants scaled things up by 1<<12, plus we had 1<<2 from first + // loop, plus horizontal and vertical each scale by sqrt(8) so together + // we've got an extra 1<<3, so 1<<17 total we need to remove. + x0 += 65536; x1 += 65536; x2 += 65536; x3 += 65536; + o[0] = clamp((x0+t3) >> 17); + o[7] = clamp((x0-t3) >> 17); + o[1] = clamp((x1+t2) >> 17); + o[6] = clamp((x1-t2) >> 17); + o[2] = clamp((x2+t1) >> 17); + o[5] = clamp((x2-t1) >> 17); + o[3] = clamp((x3+t0) >> 17); + o[4] = clamp((x3-t0) >> 17); + } +} +#else +static void idct_block(uint8 *out, int out_stride, short data[64], unsigned short *dequantize) +{ + int i,val[64],*v=val; + uint8 *o; + unsigned short *dq = dequantize; + short *d = data; + + // columns + for (i=0; i < 8; ++i,++d,++dq, ++v) { + // if all zeroes, shortcut -- this avoids dequantizing 0s and IDCTing + if (d[ 8]==0 && d[16]==0 && d[24]==0 && d[32]==0 + && d[40]==0 && d[48]==0 && d[56]==0) { + // no shortcut 0 seconds + // (1|2|3|4|5|6|7)==0 0 seconds + // all separate -0.047 seconds + // 1 && 2|3 && 4|5 && 6|7: -0.047 seconds + int dcterm = d[0] * dq[0] << 2; + v[0] = v[8] = v[16] = v[24] = v[32] = v[40] = v[48] = v[56] = dcterm; + } else { + IDCT_1D(d[ 0]*dq[ 0],d[ 8]*dq[ 8],d[16]*dq[16],d[24]*dq[24], + d[32]*dq[32],d[40]*dq[40],d[48]*dq[48],d[56]*dq[56]) + // constants scaled things up by 1<<12; let's bring them back + // down, but keep 2 extra bits of precision + x0 += 512; x1 += 512; x2 += 512; x3 += 512; + v[ 0] = (x0+t3) >> 10; + v[56] = (x0-t3) >> 10; + v[ 8] = (x1+t2) >> 10; + v[48] = (x1-t2) >> 10; + v[16] = (x2+t1) >> 10; + v[40] = (x2-t1) >> 10; + v[24] = (x3+t0) >> 10; + v[32] = (x3-t0) >> 10; + } + } + + for (i=0, v=val, o=out; i < 8; ++i,v+=8,o+=out_stride) { + // no fast case since the first 1D IDCT spread components out + IDCT_1D(v[0],v[1],v[2],v[3],v[4],v[5],v[6],v[7]) + // constants scaled things up by 1<<12, plus we had 1<<2 from first + // loop, plus horizontal and vertical each scale by sqrt(8) so together + // we've got an extra 1<<3, so 1<<17 total we need to remove. + x0 += 65536; x1 += 65536; x2 += 65536; x3 += 65536; + o[0] = clamp((x0+t3) >> 17); + o[7] = clamp((x0-t3) >> 17); + o[1] = clamp((x1+t2) >> 17); + o[6] = clamp((x1-t2) >> 17); + o[2] = clamp((x2+t1) >> 17); + o[5] = clamp((x2-t1) >> 17); + o[3] = clamp((x3+t0) >> 17); + o[4] = clamp((x3-t0) >> 17); + } +} +static stbi_idct_8x8 stbi_idct_installed = idct_block; + +extern void stbi_install_idct(stbi_idct_8x8 func) +{ + stbi_idct_installed = func; +} +#endif + +#define MARKER_none 0xff +// if there's a pending marker from the entropy stream, return that +// otherwise, fetch from the stream and get a marker. if there's no +// marker, return 0xff, which is never a valid marker value +static uint8 get_marker(jpeg *j) +{ + uint8 x; + if (j->marker != MARKER_none) { x = j->marker; j->marker = MARKER_none; return x; } + x = get8u(&j->s); + if (x != 0xff) return MARKER_none; + while (x == 0xff) + x = get8u(&j->s); + return x; +} + +// in each scan, we'll have scan_n components, and the order +// of the components is specified by order[] +#define RESTART(x) ((x) >= 0xd0 && (x) <= 0xd7) + +// after a restart interval, reset the entropy decoder and +// the dc prediction +static void reset(jpeg *j) +{ + j->code_bits = 0; + j->code_buffer = 0; + j->nomore = 0; + j->img_comp[0].dc_pred = j->img_comp[1].dc_pred = j->img_comp[2].dc_pred = 0; + j->marker = MARKER_none; + j->todo = j->restart_interval ? j->restart_interval : 0x7fffffff; + // no more than 1<<31 MCUs if no restart_interal? that's plenty safe, + // since we don't even allow 1<<30 pixels +} + +static int parse_entropy_coded_data(jpeg *z) +{ + reset(z); + if (z->scan_n == 1) { + int i,j; + #if STBI_SIMD + __declspec(align(16)) + #endif + short data[64]; + int n = z->order[0]; + // non-interleaved data, we just need to process one block at a time, + // in trivial scanline order + // number of blocks to do just depends on how many actual "pixels" this + // component has, independent of interleaved MCU blocking and such + int w = (z->img_comp[n].x+7) >> 3; + int h = (z->img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + if (!decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+z->img_comp[n].ha, n)) return 0; + #if STBI_SIMD + stbi_idct_installed(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data, z->dequant2[z->img_comp[n].tq]); + #else + idct_block(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data, z->dequant[z->img_comp[n].tq]); + #endif + // every data block is an MCU, so countdown the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) grow_buffer_unsafe(z); + // if it's NOT a restart, then just bail, so we get corrupt data + // rather than no data + if (!RESTART(z->marker)) return 1; + reset(z); + } + } + } + } else { // interleaved! + int i,j,k,x,y; + short data[64]; + for (j=0; j < z->img_mcu_y; ++j) { + for (i=0; i < z->img_mcu_x; ++i) { + // scan an interleaved mcu... process scan_n components in order + for (k=0; k < z->scan_n; ++k) { + int n = z->order[k]; + // scan out an mcu's worth of this component; that's just determined + // by the basic H and V specified for the component + for (y=0; y < z->img_comp[n].v; ++y) { + for (x=0; x < z->img_comp[n].h; ++x) { + int x2 = (i*z->img_comp[n].h + x)*8; + int y2 = (j*z->img_comp[n].v + y)*8; + if (!decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+z->img_comp[n].ha, n)) return 0; + #if STBI_SIMD + stbi_idct_installed(z->img_comp[n].data+z->img_comp[n].w2*y2+x2, z->img_comp[n].w2, data, z->dequant2[z->img_comp[n].tq]); + #else + idct_block(z->img_comp[n].data+z->img_comp[n].w2*y2+x2, z->img_comp[n].w2, data, z->dequant[z->img_comp[n].tq]); + #endif + } + } + } + // after all interleaved components, that's an interleaved MCU, + // so now count down the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) grow_buffer_unsafe(z); + // if it's NOT a restart, then just bail, so we get corrupt data + // rather than no data + if (!RESTART(z->marker)) return 1; + reset(z); + } + } + } + } + return 1; +} + +static int process_marker(jpeg *z, int m) +{ + int L; + switch (m) { + case MARKER_none: // no marker found + return e("expected marker","Corrupt JPEG"); + + case 0xC2: // SOF - progressive + return e("progressive jpeg","JPEG format not supported (progressive)"); + + case 0xDD: // DRI - specify restart interval + if (get16(&z->s) != 4) return e("bad DRI len","Corrupt JPEG"); + z->restart_interval = get16(&z->s); + return 1; + + case 0xDB: // DQT - define quantization table + L = get16(&z->s)-2; + while (L > 0) { + int q = get8(&z->s); + int p = q >> 4; + int t = q & 15,i; + if (p != 0) return e("bad DQT type","Corrupt JPEG"); + if (t > 3) return e("bad DQT table","Corrupt JPEG"); + for (i=0; i < 64; ++i) + z->dequant[t][dezigzag[i]] = get8u(&z->s); + #if STBI_SIMD + for (i=0; i < 64; ++i) + z->dequant2[t][i] = dequant[t][i]; + #endif + L -= 65; + } + return L==0; + + case 0xC4: // DHT - define huffman table + L = get16(&z->s)-2; + while (L > 0) { + uint8 *v; + int sizes[16],i,m=0; + int q = get8(&z->s); + int tc = q >> 4; + int th = q & 15; + if (tc > 1 || th > 3) return e("bad DHT header","Corrupt JPEG"); + for (i=0; i < 16; ++i) { + sizes[i] = get8(&z->s); + m += sizes[i]; + } + L -= 17; + if (tc == 0) { + if (!build_huffman(z->huff_dc+th, sizes)) return 0; + v = z->huff_dc[th].values; + } else { + if (!build_huffman(z->huff_ac+th, sizes)) return 0; + v = z->huff_ac[th].values; + } + for (i=0; i < m; ++i) + v[i] = get8u(&z->s); + L -= m; + } + return L==0; + } + // check for comment block or APP blocks + if ((m >= 0xE0 && m <= 0xEF) || m == 0xFE) { + skip(&z->s, get16(&z->s)-2); + return 1; + } + return 0; +} + +// after we see SOS +static int process_scan_header(jpeg *z) +{ + int i; + int Ls = get16(&z->s); + z->scan_n = get8(&z->s); + if (z->scan_n < 1 || z->scan_n > 4 || z->scan_n > (int) z->s.img_n) return e("bad SOS component count","Corrupt JPEG"); + if (Ls != 6+2*z->scan_n) return e("bad SOS len","Corrupt JPEG"); + for (i=0; i < z->scan_n; ++i) { + int id = get8(&z->s), which; + int q = get8(&z->s); + for (which = 0; which < z->s.img_n; ++which) + if (z->img_comp[which].id == id) + break; + if (which == z->s.img_n) return 0; + z->img_comp[which].hd = q >> 4; if (z->img_comp[which].hd > 3) return e("bad DC huff","Corrupt JPEG"); + z->img_comp[which].ha = q & 15; if (z->img_comp[which].ha > 3) return e("bad AC huff","Corrupt JPEG"); + z->order[i] = which; + } + if (get8(&z->s) != 0) return e("bad SOS","Corrupt JPEG"); + get8(&z->s); // should be 63, but might be 0 + if (get8(&z->s) != 0) return e("bad SOS","Corrupt JPEG"); + + return 1; +} + +static int process_frame_header(jpeg *z, int scan) +{ + stbi *s = &z->s; + int Lf,p,i,q, h_max=1,v_max=1,c; + Lf = get16(s); if (Lf < 11) return e("bad SOF len","Corrupt JPEG"); // JPEG + p = get8(s); if (p != 8) return e("only 8-bit","JPEG format not supported: 8-bit only"); // JPEG baseline + s->img_y = get16(s); if (s->img_y == 0) return e("no header height", "JPEG format not supported: delayed height"); // Legal, but we don't handle it--but neither does IJG + s->img_x = get16(s); if (s->img_x == 0) return e("0 width","Corrupt JPEG"); // JPEG requires + c = get8(s); + if (c != 3 && c != 1) return e("bad component count","Corrupt JPEG"); // JFIF requires + s->img_n = c; + for (i=0; i < c; ++i) { + z->img_comp[i].data = NULL; + z->img_comp[i].linebuf = NULL; + } + + if (Lf != 8+3*s->img_n) return e("bad SOF len","Corrupt JPEG"); + + for (i=0; i < s->img_n; ++i) { + z->img_comp[i].id = get8(s); + if (z->img_comp[i].id != i+1) // JFIF requires + if (z->img_comp[i].id != i) // some version of jpegtran outputs non-JFIF-compliant files! + return e("bad component ID","Corrupt JPEG"); + q = get8(s); + z->img_comp[i].h = (q >> 4); if (!z->img_comp[i].h || z->img_comp[i].h > 4) return e("bad H","Corrupt JPEG"); + z->img_comp[i].v = q & 15; if (!z->img_comp[i].v || z->img_comp[i].v > 4) return e("bad V","Corrupt JPEG"); + z->img_comp[i].tq = get8(s); if (z->img_comp[i].tq > 3) return e("bad TQ","Corrupt JPEG"); + } + + if (scan != SCAN_load) return 1; + + if ((1 << 30) / s->img_x / s->img_n < s->img_y) return e("too large", "Image too large to decode"); + + for (i=0; i < s->img_n; ++i) { + if (z->img_comp[i].h > h_max) h_max = z->img_comp[i].h; + if (z->img_comp[i].v > v_max) v_max = z->img_comp[i].v; + } + + // compute interleaved mcu info + z->img_h_max = h_max; + z->img_v_max = v_max; + z->img_mcu_w = h_max * 8; + z->img_mcu_h = v_max * 8; + z->img_mcu_x = (s->img_x + z->img_mcu_w-1) / z->img_mcu_w; + z->img_mcu_y = (s->img_y + z->img_mcu_h-1) / z->img_mcu_h; + + for (i=0; i < s->img_n; ++i) { + // number of effective pixels (e.g. for non-interleaved MCU) + z->img_comp[i].x = (s->img_x * z->img_comp[i].h + h_max-1) / h_max; + z->img_comp[i].y = (s->img_y * z->img_comp[i].v + v_max-1) / v_max; + // to simplify generation, we'll allocate enough memory to decode + // the bogus oversized data from using interleaved MCUs and their + // big blocks (e.g. a 16x16 iMCU on an image of width 33); we won't + // discard the extra data until colorspace conversion + z->img_comp[i].w2 = z->img_mcu_x * z->img_comp[i].h * 8; + z->img_comp[i].h2 = z->img_mcu_y * z->img_comp[i].v * 8; + z->img_comp[i].raw_data = malloc(z->img_comp[i].w2 * z->img_comp[i].h2+15); + if (z->img_comp[i].raw_data == NULL) { + for(--i; i >= 0; --i) { + free(z->img_comp[i].raw_data); + z->img_comp[i].data = NULL; + } + return e("outofmem", "Out of memory"); + } + // align blocks for installable-idct using mmx/sse + z->img_comp[i].data = (uint8*) (((size_t) z->img_comp[i].raw_data + 15) & ~15); + z->img_comp[i].linebuf = NULL; + } + + return 1; +} + +// use comparisons since in some cases we handle more than one case (e.g. SOF) +#define DNL(x) ((x) == 0xdc) +#define SOI(x) ((x) == 0xd8) +#define EOI(x) ((x) == 0xd9) +#define SOF(x) ((x) == 0xc0 || (x) == 0xc1) +#define SOS(x) ((x) == 0xda) + +static int decode_jpeg_header(jpeg *z, int scan) +{ + int m; + z->marker = MARKER_none; // initialize cached marker to empty + m = get_marker(z); + if (!SOI(m)) return e("no SOI","Corrupt JPEG"); + if (scan == SCAN_type) return 1; + m = get_marker(z); + while (!SOF(m)) { + if (!process_marker(z,m)) return 0; + m = get_marker(z); + while (m == MARKER_none) { + // some files have extra padding after their blocks, so ok, we'll scan + if (at_eof(&z->s)) return e("no SOF", "Corrupt JPEG"); + m = get_marker(z); + } + } + if (!process_frame_header(z, scan)) return 0; + return 1; +} + +static int decode_jpeg_image(jpeg *j) +{ + int m; + j->restart_interval = 0; + if (!decode_jpeg_header(j, SCAN_load)) return 0; + m = get_marker(j); + while (!EOI(m)) { + if (SOS(m)) { + if (!process_scan_header(j)) return 0; + if (!parse_entropy_coded_data(j)) return 0; + } else { + if (!process_marker(j, m)) return 0; + } + m = get_marker(j); + } + return 1; +} + +// static jfif-centered resampling (across block boundaries) + +typedef uint8 *(*resample_row_func)(uint8 *out, uint8 *in0, uint8 *in1, + int w, int hs); + +#define div4(x) ((uint8) ((x) >> 2)) + +static uint8 *resample_row_1(uint8 *out, uint8 *in_near, uint8 *in_far, int w, int hs) +{ + return in_near; +} + +static uint8* resample_row_v_2(uint8 *out, uint8 *in_near, uint8 *in_far, int w, int hs) +{ + // need to generate two samples vertically for every one in input + int i; + for (i=0; i < w; ++i) + out[i] = div4(3*in_near[i] + in_far[i] + 2); + return out; +} + +static uint8* resample_row_h_2(uint8 *out, uint8 *in_near, uint8 *in_far, int w, int hs) +{ + // need to generate two samples horizontally for every one in input + int i; + uint8 *input = in_near; + if (w == 1) { + // if only one sample, can't do any interpolation + out[0] = out[1] = input[0]; + return out; + } + + out[0] = input[0]; + out[1] = div4(input[0]*3 + input[1] + 2); + for (i=1; i < w-1; ++i) { + int n = 3*input[i]+2; + out[i*2+0] = div4(n+input[i-1]); + out[i*2+1] = div4(n+input[i+1]); + } + out[i*2+0] = div4(input[w-2]*3 + input[w-1] + 2); + out[i*2+1] = input[w-1]; + return out; +} + +#define div16(x) ((uint8) ((x) >> 4)) + +static uint8 *resample_row_hv_2(uint8 *out, uint8 *in_near, uint8 *in_far, int w, int hs) +{ + // need to generate 2x2 samples for every one in input + int i,t0,t1; + if (w == 1) { + out[0] = out[1] = div4(3*in_near[0] + in_far[0] + 2); + return out; + } + + t1 = 3*in_near[0] + in_far[0]; + out[0] = div4(t1+2); + for (i=1; i < w; ++i) { + t0 = t1; + t1 = 3*in_near[i]+in_far[i]; + out[i*2-1] = div16(3*t0 + t1 + 8); + out[i*2 ] = div16(3*t1 + t0 + 8); + } + out[w*2-1] = div4(t1+2); + return out; +} + +static uint8 *resample_row_generic(uint8 *out, uint8 *in_near, uint8 *in_far, int w, int hs) +{ + // resample with nearest-neighbor + int i,j; + for (i=0; i < w; ++i) + for (j=0; j < hs; ++j) + out[i*hs+j] = in_near[i]; + return out; +} + +#define float2fixed(x) ((int) ((x) * 65536 + 0.5)) + +// 0.38 seconds on 3*anemones.jpg (0.25 with processor = Pro) +// VC6 without processor=Pro is generating multiple LEAs per multiply! +static void YCbCr_to_RGB_row(uint8 *out, uint8 *y, uint8 *pcb, uint8 *pcr, int count, int step) +{ + int i; + for (i=0; i < count; ++i) { + int y_fixed = (y[i] << 16) + 32768; // rounding + int r,g,b; + int cr = pcr[i] - 128; + int cb = pcb[i] - 128; + r = y_fixed + cr*float2fixed(1.40200f); + g = y_fixed - cr*float2fixed(0.71414f) - cb*float2fixed(0.34414f); + b = y_fixed + cb*float2fixed(1.77200f); + r >>= 16; + g >>= 16; + b >>= 16; + if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; } + if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; } + if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; } + out[0] = (uint8)r; + out[1] = (uint8)g; + out[2] = (uint8)b; + out[3] = 255; + out += step; + } +} + +#if STBI_SIMD +static stbi_YCbCr_to_RGB_run stbi_YCbCr_installed = YCbCr_to_RGB_row; + +void stbi_install_YCbCr_to_RGB(stbi_YCbCr_to_RGB_run func) +{ + stbi_YCbCr_installed = func; +} +#endif + + +// clean up the temporary component buffers +static void cleanup_jpeg(jpeg *j) +{ + int i; + for (i=0; i < j->s.img_n; ++i) { + if (j->img_comp[i].data) { + free(j->img_comp[i].raw_data); + j->img_comp[i].data = NULL; + } + if (j->img_comp[i].linebuf) { + free(j->img_comp[i].linebuf); + j->img_comp[i].linebuf = NULL; + } + } +} + +typedef struct +{ + resample_row_func resample; + uint8 *line0,*line1; + int hs,vs; // expansion factor in each axis + int w_lores; // horizontal pixels pre-expansion + int ystep; // how far through vertical expansion we are + int ypos; // which pre-expansion row we're on +} stbi_resample; + +static uint8 *load_jpeg_image(jpeg *z, int *out_x, int *out_y, int *comp, int req_comp) +{ + int n, decode_n; + // validate req_comp + if (req_comp < 0 || req_comp > 4) return epuc("bad req_comp", "Internal error"); + z->s.img_n = 0; + + // load a jpeg image from whichever source + if (!decode_jpeg_image(z)) { cleanup_jpeg(z); return NULL; } + + // determine actual number of components to generate + n = req_comp ? req_comp : z->s.img_n; + + if (z->s.img_n == 3 && n < 3) + decode_n = 1; + else + decode_n = z->s.img_n; + + // resample and color-convert + { + int k; + uint i,j; + uint8 *output; + uint8 *coutput[4]; + + stbi_resample res_comp[4]; + + for (k=0; k < decode_n; ++k) { + stbi_resample *r = &res_comp[k]; + + // allocate line buffer big enough for upsampling off the edges + // with upsample factor of 4 + z->img_comp[k].linebuf = (uint8 *) malloc(z->s.img_x + 3); + if (!z->img_comp[k].linebuf) { cleanup_jpeg(z); return epuc("outofmem", "Out of memory"); } + + r->hs = z->img_h_max / z->img_comp[k].h; + r->vs = z->img_v_max / z->img_comp[k].v; + r->ystep = r->vs >> 1; + r->w_lores = (z->s.img_x + r->hs-1) / r->hs; + r->ypos = 0; + r->line0 = r->line1 = z->img_comp[k].data; + + if (r->hs == 1 && r->vs == 1) r->resample = resample_row_1; + else if (r->hs == 1 && r->vs == 2) r->resample = resample_row_v_2; + else if (r->hs == 2 && r->vs == 1) r->resample = resample_row_h_2; + else if (r->hs == 2 && r->vs == 2) r->resample = resample_row_hv_2; + else r->resample = resample_row_generic; + } + + // can't error after this so, this is safe + output = (uint8 *) malloc(n * z->s.img_x * z->s.img_y + 1); + if (!output) { cleanup_jpeg(z); return epuc("outofmem", "Out of memory"); } + + // now go ahead and resample + for (j=0; j < z->s.img_y; ++j) { + uint8 *out = output + n * z->s.img_x * j; + for (k=0; k < decode_n; ++k) { + stbi_resample *r = &res_comp[k]; + int y_bot = r->ystep >= (r->vs >> 1); + coutput[k] = r->resample(z->img_comp[k].linebuf, + y_bot ? r->line1 : r->line0, + y_bot ? r->line0 : r->line1, + r->w_lores, r->hs); + if (++r->ystep >= r->vs) { + r->ystep = 0; + r->line0 = r->line1; + if (++r->ypos < z->img_comp[k].y) + r->line1 += z->img_comp[k].w2; + } + } + if (n >= 3) { + uint8 *y = coutput[0]; + if (z->s.img_n == 3) { + #if STBI_SIMD + stbi_YCbCr_installed(out, y, coutput[1], coutput[2], z->s.img_x, n); + #else + YCbCr_to_RGB_row(out, y, coutput[1], coutput[2], z->s.img_x, n); + #endif + } else + for (i=0; i < z->s.img_x; ++i) { + out[0] = out[1] = out[2] = y[i]; + out[3] = 255; // not used if n==3 + out += n; + } + } else { + uint8 *y = coutput[0]; + if (n == 1) + for (i=0; i < z->s.img_x; ++i) out[i] = y[i]; + else + for (i=0; i < z->s.img_x; ++i) *out++ = y[i], *out++ = 255; + } + } + cleanup_jpeg(z); + *out_x = z->s.img_x; + *out_y = z->s.img_y; + if (comp) *comp = z->s.img_n; // report original components, not output + return output; + } +} + +#ifndef STBI_NO_STDIO +unsigned char *stbi_jpeg_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + jpeg j; + start_file(&j.s, f); + return load_jpeg_image(&j, x,y,comp,req_comp); +} + +unsigned char *stbi_jpeg_load(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *data; + FILE *f = fopen(filename, "rb"); + if (!f) return NULL; + data = stbi_jpeg_load_from_file(f,x,y,comp,req_comp); + fclose(f); + return data; +} +#endif + +unsigned char *stbi_jpeg_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + jpeg j; + start_mem(&j.s, buffer,len); + return load_jpeg_image(&j, x,y,comp,req_comp); +} + +#ifndef STBI_NO_STDIO +int stbi_jpeg_test_file(FILE *f) +{ + int n,r; + jpeg j; + n = ftell(f); + start_file(&j.s, f); + r = decode_jpeg_header(&j, SCAN_type); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +int stbi_jpeg_test_memory(stbi_uc const *buffer, int len) +{ + jpeg j; + start_mem(&j.s, buffer,len); + return decode_jpeg_header(&j, SCAN_type); +} + +// @TODO: +#ifndef STBI_NO_STDIO +extern int stbi_jpeg_info (char const *filename, int *x, int *y, int *comp); +extern int stbi_jpeg_info_from_file (FILE *f, int *x, int *y, int *comp); +#endif +extern int stbi_jpeg_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp); + +// public domain zlib decode v0.2 Sean Barrett 2006-11-18 +// simple implementation +// - all input must be provided in an upfront buffer +// - all output is written to a single output buffer (can malloc/realloc) +// performance +// - fast huffman + +// fast-way is faster to check than jpeg huffman, but slow way is slower +#define ZFAST_BITS 9 // accelerate all cases in default tables +#define ZFAST_MASK ((1 << ZFAST_BITS) - 1) + +// zlib-style huffman encoding +// (jpegs packs from left, zlib from right, so can't share code) +typedef struct +{ + uint16 fast[1 << ZFAST_BITS]; + uint16 firstcode[16]; + int maxcode[17]; + uint16 firstsymbol[16]; + uint8 size[288]; + uint16 value[288]; +} zhuffman; + +__forceinline static int bitreverse16(int n) +{ + n = ((n & 0xAAAA) >> 1) | ((n & 0x5555) << 1); + n = ((n & 0xCCCC) >> 2) | ((n & 0x3333) << 2); + n = ((n & 0xF0F0) >> 4) | ((n & 0x0F0F) << 4); + n = ((n & 0xFF00) >> 8) | ((n & 0x00FF) << 8); + return n; +} + +__forceinline static int bit_reverse(int v, int bits) +{ + assert(bits <= 16); + // to bit reverse n bits, reverse 16 and shift + // e.g. 11 bits, bit reverse and shift away 5 + return bitreverse16(v) >> (16-bits); +} + +static int zbuild_huffman(zhuffman *z, uint8 *sizelist, int num) +{ + int i,k=0; + int code, next_code[16], sizes[17]; + + // DEFLATE spec for generating codes + memset(sizes, 0, sizeof(sizes)); + memset(z->fast, 255, sizeof(z->fast)); + for (i=0; i < num; ++i) + ++sizes[sizelist[i]]; + sizes[0] = 0; + for (i=1; i < 16; ++i) + assert(sizes[i] <= (1 << i)); + code = 0; + for (i=1; i < 16; ++i) { + next_code[i] = code; + z->firstcode[i] = (uint16) code; + z->firstsymbol[i] = (uint16) k; + code = (code + sizes[i]); + if (sizes[i]) + if (code-1 >= (1 << i)) return e("bad codelengths","Corrupt JPEG"); + z->maxcode[i] = code << (16-i); // preshift for inner loop + code <<= 1; + k += sizes[i]; + } + z->maxcode[16] = 0x10000; // sentinel + for (i=0; i < num; ++i) { + int s = sizelist[i]; + if (s) { + int c = next_code[s] - z->firstcode[s] + z->firstsymbol[s]; + z->size[c] = (uint8)s; + z->value[c] = (uint16)i; + if (s <= ZFAST_BITS) { + int k = bit_reverse(next_code[s],s); + while (k < (1 << ZFAST_BITS)) { + z->fast[k] = (uint16) c; + k += (1 << s); + } + } + ++next_code[s]; + } + } + return 1; +} + +// zlib-from-memory implementation for PNG reading +// because PNG allows splitting the zlib stream arbitrarily, +// and it's annoying structurally to have PNG call ZLIB call PNG, +// we require PNG read all the IDATs and combine them into a single +// memory buffer + +typedef struct +{ + uint8 *zbuffer, *zbuffer_end; + int num_bits; + uint32 code_buffer; + + char *zout; + char *zout_start; + char *zout_end; + int z_expandable; + + zhuffman z_length, z_distance; +} zbuf; + +__forceinline static int zget8(zbuf *z) +{ + if (z->zbuffer >= z->zbuffer_end) return 0; + return *z->zbuffer++; +} + +static void fill_bits(zbuf *z) +{ + do { + assert(z->code_buffer < (1U << z->num_bits)); + z->code_buffer |= zget8(z) << z->num_bits; + z->num_bits += 8; + } while (z->num_bits <= 24); +} + +__forceinline static unsigned int zreceive(zbuf *z, int n) +{ + unsigned int k; + if (z->num_bits < n) fill_bits(z); + k = z->code_buffer & ((1 << n) - 1); + z->code_buffer >>= n; + z->num_bits -= n; + return k; +} + +__forceinline static int zhuffman_decode(zbuf *a, zhuffman *z) +{ + int b,s,k; + if (a->num_bits < 16) fill_bits(a); + b = z->fast[a->code_buffer & ZFAST_MASK]; + if (b < 0xffff) { + s = z->size[b]; + a->code_buffer >>= s; + a->num_bits -= s; + return z->value[b]; + } + + // not resolved by fast table, so compute it the slow way + // use jpeg approach, which requires MSbits at top + k = bit_reverse(a->code_buffer, 16); + for (s=ZFAST_BITS+1; ; ++s) + if (k < z->maxcode[s]) + break; + if (s == 16) return -1; // invalid code! + // code size is s, so: + b = (k >> (16-s)) - z->firstcode[s] + z->firstsymbol[s]; + assert(z->size[b] == s); + a->code_buffer >>= s; + a->num_bits -= s; + return z->value[b]; +} + +static int expand(zbuf *z, int n) // need to make room for n bytes +{ + char *q; + int cur, limit; + if (!z->z_expandable) return e("output buffer limit","Corrupt PNG"); + cur = (int) (z->zout - z->zout_start); + limit = (int) (z->zout_end - z->zout_start); + while (cur + n > limit) + limit *= 2; + q = (char *) realloc(z->zout_start, limit); + if (q == NULL) return e("outofmem", "Out of memory"); + z->zout_start = q; + z->zout = q + cur; + z->zout_end = q + limit; + return 1; +} + +static int length_base[31] = { + 3,4,5,6,7,8,9,10,11,13, + 15,17,19,23,27,31,35,43,51,59, + 67,83,99,115,131,163,195,227,258,0,0 }; + +static int length_extra[31]= +{ 0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0,0,0 }; + +static int dist_base[32] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193, +257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577,0,0}; + +static int dist_extra[32] = +{ 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13}; + +static int parse_huffman_block(zbuf *a) +{ + for(;;) { + int z = zhuffman_decode(a, &a->z_length); + if (z < 256) { + if (z < 0) return e("bad huffman code","Corrupt PNG"); // error in huffman codes + if (a->zout >= a->zout_end) if (!expand(a, 1)) return 0; + *a->zout++ = (char) z; + } else { + uint8 *p; + int len,dist; + if (z == 256) return 1; + z -= 257; + len = length_base[z]; + if (length_extra[z]) len += zreceive(a, length_extra[z]); + z = zhuffman_decode(a, &a->z_distance); + if (z < 0) return e("bad huffman code","Corrupt PNG"); + dist = dist_base[z]; + if (dist_extra[z]) dist += zreceive(a, dist_extra[z]); + if (a->zout - a->zout_start < dist) return e("bad dist","Corrupt PNG"); + if (a->zout + len > a->zout_end) if (!expand(a, len)) return 0; + p = (uint8 *) (a->zout - dist); + while (len--) + *a->zout++ = *p++; + } + } +} + +static int compute_huffman_codes(zbuf *a) +{ + static uint8 length_dezigzag[19] = { 16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15 }; + static zhuffman z_codelength; // static just to save stack space + uint8 lencodes[286+32+137];//padding for maximum single op + uint8 codelength_sizes[19]; + int i,n; + + int hlit = zreceive(a,5) + 257; + int hdist = zreceive(a,5) + 1; + int hclen = zreceive(a,4) + 4; + + memset(codelength_sizes, 0, sizeof(codelength_sizes)); + for (i=0; i < hclen; ++i) { + int s = zreceive(a,3); + codelength_sizes[length_dezigzag[i]] = (uint8) s; + } + if (!zbuild_huffman(&z_codelength, codelength_sizes, 19)) return 0; + + n = 0; + while (n < hlit + hdist) { + int c = zhuffman_decode(a, &z_codelength); + assert(c >= 0 && c < 19); + if (c < 16) + lencodes[n++] = (uint8) c; + else if (c == 16) { + c = zreceive(a,2)+3; + memset(lencodes+n, lencodes[n-1], c); + n += c; + } else if (c == 17) { + c = zreceive(a,3)+3; + memset(lencodes+n, 0, c); + n += c; + } else { + assert(c == 18); + c = zreceive(a,7)+11; + memset(lencodes+n, 0, c); + n += c; + } + } + if (n != hlit+hdist) return e("bad codelengths","Corrupt PNG"); + if (!zbuild_huffman(&a->z_length, lencodes, hlit)) return 0; + if (!zbuild_huffman(&a->z_distance, lencodes+hlit, hdist)) return 0; + return 1; +} + +static int parse_uncompressed_block(zbuf *a) +{ + uint8 header[4]; + int len,nlen,k; + if (a->num_bits & 7) + zreceive(a, a->num_bits & 7); // discard + // drain the bit-packed data into header + k = 0; + while (a->num_bits > 0) { + header[k++] = (uint8) (a->code_buffer & 255); // wtf this warns? + a->code_buffer >>= 8; + a->num_bits -= 8; + } + assert(a->num_bits == 0); + // now fill header the normal way + while (k < 4) + header[k++] = (uint8) zget8(a); + len = header[1] * 256 + header[0]; + nlen = header[3] * 256 + header[2]; + if (nlen != (len ^ 0xffff)) return e("zlib corrupt","Corrupt PNG"); + if (a->zbuffer + len > a->zbuffer_end) return e("read past buffer","Corrupt PNG"); + if (a->zout + len > a->zout_end) + if (!expand(a, len)) return 0; + memcpy(a->zout, a->zbuffer, len); + a->zbuffer += len; + a->zout += len; + return 1; +} + +static int parse_zlib_header(zbuf *a) +{ + int cmf = zget8(a); + int cm = cmf & 15; + /* int cinfo = cmf >> 4; */ + int flg = zget8(a); + if ((cmf*256+flg) % 31 != 0) return e("bad zlib header","Corrupt PNG"); // zlib spec + if (flg & 32) return e("no preset dict","Corrupt PNG"); // preset dictionary not allowed in png + if (cm != 8) return e("bad compression","Corrupt PNG"); // DEFLATE required for png + // window = 1 << (8 + cinfo)... but who cares, we fully buffer output + return 1; +} + +// @TODO: should statically initialize these for optimal thread safety +static uint8 default_length[288], default_distance[32]; +static void init_defaults(void) +{ + int i; // use <= to match clearly with spec + for (i=0; i <= 143; ++i) default_length[i] = 8; + for ( ; i <= 255; ++i) default_length[i] = 9; + for ( ; i <= 279; ++i) default_length[i] = 7; + for ( ; i <= 287; ++i) default_length[i] = 8; + + for (i=0; i <= 31; ++i) default_distance[i] = 5; +} + +static int parse_zlib(zbuf *a, int parse_header) +{ + int final, type; + if (parse_header) + if (!parse_zlib_header(a)) return 0; + a->num_bits = 0; + a->code_buffer = 0; + do { + final = zreceive(a,1); + type = zreceive(a,2); + if (type == 0) { + if (!parse_uncompressed_block(a)) return 0; + } else if (type == 3) { + return 0; + } else { + if (type == 1) { + // use fixed code lengths + if (!default_distance[31]) init_defaults(); + if (!zbuild_huffman(&a->z_length , default_length , 288)) return 0; + if (!zbuild_huffman(&a->z_distance, default_distance, 32)) return 0; + } else { + if (!compute_huffman_codes(a)) return 0; + } + if (!parse_huffman_block(a)) return 0; + } + } while (!final); + return 1; +} + +static int do_zlib(zbuf *a, char *obuf, int olen, int exp, int parse_header) +{ + a->zout_start = obuf; + a->zout = obuf; + a->zout_end = obuf + olen; + a->z_expandable = exp; + + return parse_zlib(a, parse_header); +} + +char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen) +{ + zbuf a; + char *p = (char *) malloc(initial_size); + if (p == NULL) return NULL; + a.zbuffer = (uint8 *) buffer; + a.zbuffer_end = (uint8 *) buffer + len; + if (do_zlib(&a, p, initial_size, 1, 1)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + free(a.zout_start); + return NULL; + } +} + +char *stbi_zlib_decode_malloc(char const *buffer, int len, int *outlen) +{ + return stbi_zlib_decode_malloc_guesssize(buffer, len, 16384, outlen); +} + +int stbi_zlib_decode_buffer(char *obuffer, int olen, char const *ibuffer, int ilen) +{ + zbuf a; + a.zbuffer = (uint8 *) ibuffer; + a.zbuffer_end = (uint8 *) ibuffer + ilen; + if (do_zlib(&a, obuffer, olen, 0, 1)) + return (int) (a.zout - a.zout_start); + else + return -1; +} + +char *stbi_zlib_decode_noheader_malloc(char const *buffer, int len, int *outlen) +{ + zbuf a; + char *p = (char *) malloc(16384); + if (p == NULL) return NULL; + a.zbuffer = (uint8 *) buffer; + a.zbuffer_end = (uint8 *) buffer+len; + if (do_zlib(&a, p, 16384, 1, 0)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + free(a.zout_start); + return NULL; + } +} + +int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen) +{ + zbuf a; + a.zbuffer = (uint8 *) ibuffer; + a.zbuffer_end = (uint8 *) ibuffer + ilen; + if (do_zlib(&a, obuffer, olen, 0, 0)) + return (int) (a.zout - a.zout_start); + else + return -1; +} + +// public domain "baseline" PNG decoder v0.10 Sean Barrett 2006-11-18 +// simple implementation +// - only 8-bit samples +// - no CRC checking +// - allocates lots of intermediate memory +// - avoids problem of streaming data between subsystems +// - avoids explicit window management +// performance +// - uses stb_zlib, a PD zlib implementation with fast huffman decoding + + +typedef struct +{ + uint32 length; + uint32 type; +} chunk; + +#define PNG_TYPE(a,b,c,d) (((a) << 24) + ((b) << 16) + ((c) << 8) + (d)) + +static chunk get_chunk_header(stbi *s) +{ + chunk c; + c.length = get32(s); + c.type = get32(s); + return c; +} + +static int check_png_header(stbi *s) +{ + static uint8 png_sig[8] = { 137,80,78,71,13,10,26,10 }; + int i; + for (i=0; i < 8; ++i) + if (get8(s) != png_sig[i]) return e("bad png sig","Not a PNG"); + return 1; +} + +typedef struct +{ + stbi s; + uint8 *idata, *expanded, *out; +} png; + + +enum { + F_none=0, F_sub=1, F_up=2, F_avg=3, F_paeth=4, + F_avg_first, F_paeth_first, +}; + +static uint8 first_row_filter[5] = +{ + F_none, F_sub, F_none, F_avg_first, F_paeth_first +}; + +static int paeth(int a, int b, int c) +{ + int p = a + b - c; + int pa = abs(p-a); + int pb = abs(p-b); + int pc = abs(p-c); + if (pa <= pb && pa <= pc) return a; + if (pb <= pc) return b; + return c; +} + +// create the png data from post-deflated data +static int create_png_image(png *a, uint8 *raw, uint32 raw_len, int out_n) +{ + stbi *s = &a->s; + uint32 i,j,stride = s->img_x*out_n; + int k; + int img_n = s->img_n; // copy it into a local for later + assert(out_n == s->img_n || out_n == s->img_n+1); + a->out = (uint8 *) malloc(s->img_x * s->img_y * out_n); + if (!a->out) return e("outofmem", "Out of memory"); + if (raw_len != (img_n * s->img_x + 1) * s->img_y) return e("not enough pixels","Corrupt PNG"); + for (j=0; j < s->img_y; ++j) { + uint8 *cur = a->out + stride*j; + uint8 *prior = cur - stride; + int filter = *raw++; + if (filter > 4) return e("invalid filter","Corrupt PNG"); + // if first row, use special filter that doesn't sample previous row + if (j == 0) filter = first_row_filter[filter]; + // handle first pixel explicitly + for (k=0; k < img_n; ++k) { + switch(filter) { + case F_none : cur[k] = raw[k]; break; + case F_sub : cur[k] = raw[k]; break; + case F_up : cur[k] = raw[k] + prior[k]; break; + case F_avg : cur[k] = raw[k] + (prior[k]>>1); break; + case F_paeth : cur[k] = (uint8) (raw[k] + paeth(0,prior[k],0)); break; + case F_avg_first : cur[k] = raw[k]; break; + case F_paeth_first: cur[k] = raw[k]; break; + } + } + if (img_n != out_n) cur[img_n] = 255; + raw += img_n; + cur += out_n; + prior += out_n; + // this is a little gross, so that we don't switch per-pixel or per-component + if (img_n == out_n) { + #define CASE(f) \ + case f: \ + for (i=s->img_x-1; i >= 1; --i, raw+=img_n,cur+=img_n,prior+=img_n) \ + for (k=0; k < img_n; ++k) + switch(filter) { + CASE(F_none) cur[k] = raw[k]; break; + CASE(F_sub) cur[k] = raw[k] + cur[k-img_n]; break; + CASE(F_up) cur[k] = raw[k] + prior[k]; break; + CASE(F_avg) cur[k] = raw[k] + ((prior[k] + cur[k-img_n])>>1); break; + CASE(F_paeth) cur[k] = (uint8) (raw[k] + paeth(cur[k-img_n],prior[k],prior[k-img_n])); break; + CASE(F_avg_first) cur[k] = raw[k] + (cur[k-img_n] >> 1); break; + CASE(F_paeth_first) cur[k] = (uint8) (raw[k] + paeth(cur[k-img_n],0,0)); break; + } + #undef CASE + } else { + assert(img_n+1 == out_n); + #define CASE(f) \ + case f: \ + for (i=s->img_x-1; i >= 1; --i, cur[img_n]=255,raw+=img_n,cur+=out_n,prior+=out_n) \ + for (k=0; k < img_n; ++k) + switch(filter) { + CASE(F_none) cur[k] = raw[k]; break; + CASE(F_sub) cur[k] = raw[k] + cur[k-out_n]; break; + CASE(F_up) cur[k] = raw[k] + prior[k]; break; + CASE(F_avg) cur[k] = raw[k] + ((prior[k] + cur[k-out_n])>>1); break; + CASE(F_paeth) cur[k] = (uint8) (raw[k] + paeth(cur[k-out_n],prior[k],prior[k-out_n])); break; + CASE(F_avg_first) cur[k] = raw[k] + (cur[k-out_n] >> 1); break; + CASE(F_paeth_first) cur[k] = (uint8) (raw[k] + paeth(cur[k-out_n],0,0)); break; + } + #undef CASE + } + } + return 1; +} + +static int compute_transparency(png *z, uint8 tc[3], int out_n) +{ + stbi *s = &z->s; + uint32 i, pixel_count = s->img_x * s->img_y; + uint8 *p = z->out; + + // compute color-based transparency, assuming we've + // already got 255 as the alpha value in the output + assert(out_n == 2 || out_n == 4); + + if (out_n == 2) { + for (i=0; i < pixel_count; ++i) { + p[1] = (p[0] == tc[0] ? 0 : 255); + p += 2; + } + } else { + for (i=0; i < pixel_count; ++i) { + if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2]) + p[3] = 0; + p += 4; + } + } + return 1; +} + +static int expand_palette(png *a, uint8 *palette, int len, int pal_img_n) +{ + uint32 i, pixel_count = a->s.img_x * a->s.img_y; + uint8 *p, *temp_out, *orig = a->out; + + p = (uint8 *) malloc(pixel_count * pal_img_n); + if (p == NULL) return e("outofmem", "Out of memory"); + + // between here and free(out) below, exitting would leak + temp_out = p; + + if (pal_img_n == 3) { + for (i=0; i < pixel_count; ++i) { + int n = orig[i]*4; + p[0] = palette[n ]; + p[1] = palette[n+1]; + p[2] = palette[n+2]; + p += 3; + } + } else { + for (i=0; i < pixel_count; ++i) { + int n = orig[i]*4; + p[0] = palette[n ]; + p[1] = palette[n+1]; + p[2] = palette[n+2]; + p[3] = palette[n+3]; + p += 4; + } + } + free(a->out); + a->out = temp_out; + return 1; +} + +static int parse_png_file(png *z, int scan, int req_comp) +{ + uint8 palette[1024], pal_img_n=0; + uint8 has_trans=0, tc[3]; + uint32 ioff=0, idata_limit=0, i, pal_len=0; + int first=1,k; + stbi *s = &z->s; + + if (!check_png_header(s)) return 0; + + if (scan == SCAN_type) return 1; + + for(;;first=0) { + chunk c = get_chunk_header(s); + if (first && c.type != PNG_TYPE('I','H','D','R')) + return e("first not IHDR","Corrupt PNG"); + switch (c.type) { + case PNG_TYPE('I','H','D','R'): { + int depth,color,interlace,comp,filter; + if (!first) return e("multiple IHDR","Corrupt PNG"); + if (c.length != 13) return e("bad IHDR len","Corrupt PNG"); + s->img_x = get32(s); if (s->img_x > (1 << 24)) return e("too large","Very large image (corrupt?)"); + s->img_y = get32(s); if (s->img_y > (1 << 24)) return e("too large","Very large image (corrupt?)"); + depth = get8(s); if (depth != 8) return e("8bit only","PNG not supported: 8-bit only"); + color = get8(s); if (color > 6) return e("bad ctype","Corrupt PNG"); + if (color == 3) pal_img_n = 3; else if (color & 1) return e("bad ctype","Corrupt PNG"); + comp = get8(s); if (comp) return e("bad comp method","Corrupt PNG"); + filter= get8(s); if (filter) return e("bad filter method","Corrupt PNG"); + interlace = get8(s); if (interlace) return e("interlaced","PNG not supported: interlaced mode"); + if (!s->img_x || !s->img_y) return e("0-pixel image","Corrupt PNG"); + if (!pal_img_n) { + s->img_n = (color & 2 ? 3 : 1) + (color & 4 ? 1 : 0); + if ((1 << 30) / s->img_x / s->img_n < s->img_y) return e("too large", "Image too large to decode"); + if (scan == SCAN_header) return 1; + } else { + // if paletted, then pal_n is our final components, and + // img_n is # components to decompress/filter. + s->img_n = 1; + if ((1 << 30) / s->img_x / 4 < s->img_y) return e("too large","Corrupt PNG"); + // if SCAN_header, have to scan to see if we have a tRNS + } + break; + } + + case PNG_TYPE('P','L','T','E'): { + if (c.length > 256*3) return e("invalid PLTE","Corrupt PNG"); + pal_len = c.length / 3; + if (pal_len * 3 != c.length) return e("invalid PLTE","Corrupt PNG"); + for (i=0; i < pal_len; ++i) { + palette[i*4+0] = get8u(s); + palette[i*4+1] = get8u(s); + palette[i*4+2] = get8u(s); + palette[i*4+3] = 255; + } + break; + } + + case PNG_TYPE('t','R','N','S'): { + if (z->idata) return e("tRNS after IDAT","Corrupt PNG"); + if (pal_img_n) { + if (scan == SCAN_header) { s->img_n = 4; return 1; } + if (pal_len == 0) return e("tRNS before PLTE","Corrupt PNG"); + if (c.length > pal_len) return e("bad tRNS len","Corrupt PNG"); + pal_img_n = 4; + for (i=0; i < c.length; ++i) + palette[i*4+3] = get8u(s); + } else { + if (!(s->img_n & 1)) return e("tRNS with alpha","Corrupt PNG"); + if (c.length != (uint32) s->img_n*2) return e("bad tRNS len","Corrupt PNG"); + has_trans = 1; + for (k=0; k < s->img_n; ++k) + tc[k] = (uint8) get16(s); // non 8-bit images will be larger + } + break; + } + + case PNG_TYPE('I','D','A','T'): { + if (pal_img_n && !pal_len) return e("no PLTE","Corrupt PNG"); + if (scan == SCAN_header) { s->img_n = pal_img_n; return 1; } + if (ioff + c.length > idata_limit) { + uint8 *p; + if (idata_limit == 0) idata_limit = c.length > 4096 ? c.length : 4096; + while (ioff + c.length > idata_limit) + idata_limit *= 2; + p = (uint8 *) realloc(z->idata, idata_limit); if (p == NULL) return e("outofmem", "Out of memory"); + z->idata = p; + } + #ifndef STBI_NO_STDIO + if (s->img_file) + { + if (fread(z->idata+ioff,1,c.length,s->img_file) != c.length) return e("outofdata","Corrupt PNG"); + } + else + #endif + { + memcpy(z->idata+ioff, s->img_buffer, c.length); + s->img_buffer += c.length; + } + ioff += c.length; + break; + } + + case PNG_TYPE('I','E','N','D'): { + uint32 raw_len; + if (scan != SCAN_load) return 1; + if (z->idata == NULL) return e("no IDAT","Corrupt PNG"); + z->expanded = (uint8 *) stbi_zlib_decode_malloc((char *) z->idata, ioff, (int *) &raw_len); + if (z->expanded == NULL) return 0; // zlib should set error + free(z->idata); z->idata = NULL; + if ((req_comp == s->img_n+1 && req_comp != 3 && !pal_img_n) || has_trans) + s->img_out_n = s->img_n+1; + else + s->img_out_n = s->img_n; + if (!create_png_image(z, z->expanded, raw_len, s->img_out_n)) return 0; + if (has_trans) + if (!compute_transparency(z, tc, s->img_out_n)) return 0; + if (pal_img_n) { + // pal_img_n == 3 or 4 + s->img_n = pal_img_n; // record the actual colors we had + s->img_out_n = pal_img_n; + if (req_comp >= 3) s->img_out_n = req_comp; + if (!expand_palette(z, palette, pal_len, s->img_out_n)) + return 0; + } + free(z->expanded); z->expanded = NULL; + return 1; + } + + default: + // if critical, fail + if ((c.type & (1 << 29)) == 0) { + #ifndef STBI_NO_FAILURE_STRINGS + // not threadsafe + static char invalid_chunk[] = "XXXX chunk not known"; + invalid_chunk[0] = (uint8) (c.type >> 24); + invalid_chunk[1] = (uint8) (c.type >> 16); + invalid_chunk[2] = (uint8) (c.type >> 8); + invalid_chunk[3] = (uint8) (c.type >> 0); + #endif + return e(invalid_chunk, "PNG not supported: unknown chunk type"); + } + skip(s, c.length); + break; + } + // end of chunk, read and skip CRC + get32(s); + } +} + +static unsigned char *do_png(png *p, int *x, int *y, int *n, int req_comp) +{ + unsigned char *result=NULL; + p->expanded = NULL; + p->idata = NULL; + p->out = NULL; + if (req_comp < 0 || req_comp > 4) return epuc("bad req_comp", "Internal error"); + if (parse_png_file(p, SCAN_load, req_comp)) { + result = p->out; + p->out = NULL; + if (req_comp && req_comp != p->s.img_out_n) { + result = convert_format(result, p->s.img_out_n, req_comp, p->s.img_x, p->s.img_y); + p->s.img_out_n = req_comp; + if (result == NULL) return result; + } + *x = p->s.img_x; + *y = p->s.img_y; + if (n) *n = p->s.img_n; + } + free(p->out); p->out = NULL; + free(p->expanded); p->expanded = NULL; + free(p->idata); p->idata = NULL; + + return result; +} + +#ifndef STBI_NO_STDIO +unsigned char *stbi_png_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + png p; + start_file(&p.s, f); + return do_png(&p, x,y,comp,req_comp); +} + +unsigned char *stbi_png_load(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *data; + FILE *f = fopen(filename, "rb"); + if (!f) return NULL; + data = stbi_png_load_from_file(f,x,y,comp,req_comp); + fclose(f); + return data; +} +#endif + +unsigned char *stbi_png_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + png p; + start_mem(&p.s, buffer,len); + return do_png(&p, x,y,comp,req_comp); +} + +#ifndef STBI_NO_STDIO +int stbi_png_test_file(FILE *f) +{ + png p; + int n,r; + n = ftell(f); + start_file(&p.s, f); + r = parse_png_file(&p, SCAN_type,STBI_default); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +int stbi_png_test_memory(stbi_uc const *buffer, int len) +{ + png p; + start_mem(&p.s, buffer, len); + return parse_png_file(&p, SCAN_type,STBI_default); +} + +// TODO: load header from png +#ifndef STBI_NO_STDIO +extern int stbi_png_info (char const *filename, int *x, int *y, int *comp); +extern int stbi_png_info_from_file (FILE *f, int *x, int *y, int *comp); +#endif +extern int stbi_png_info_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp); + +// Microsoft/Windows BMP image + +static int bmp_test(stbi *s) +{ + int sz; + if (get8(s) != 'B') return 0; + if (get8(s) != 'M') return 0; + get32le(s); // discard filesize + get16le(s); // discard reserved + get16le(s); // discard reserved + get32le(s); // discard data offset + sz = get32le(s); + if (sz == 12 || sz == 40 || sz == 56 || sz == 108) return 1; + return 0; +} + +#ifndef STBI_NO_STDIO +int stbi_bmp_test_file (FILE *f) +{ + stbi s; + int r,n = ftell(f); + start_file(&s,f); + r = bmp_test(&s); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +int stbi_bmp_test_memory (stbi_uc const *buffer, int len) +{ + stbi s; + start_mem(&s, buffer, len); + return bmp_test(&s); +} + +// returns 0..31 for the highest set bit +static int high_bit(unsigned int z) +{ + int n=0; + if (z == 0) return -1; + if (z >= 0x10000) n += 16, z >>= 16; + if (z >= 0x00100) n += 8, z >>= 8; + if (z >= 0x00010) n += 4, z >>= 4; + if (z >= 0x00004) n += 2, z >>= 2; + if (z >= 0x00002) n += 1, z >>= 1; + return n; +} + +static int bitcount(unsigned int a) +{ + a = (a & 0x55555555) + ((a >> 1) & 0x55555555); // max 2 + a = (a & 0x33333333) + ((a >> 2) & 0x33333333); // max 4 + a = (a + (a >> 4)) & 0x0f0f0f0f; // max 8 per 4, now 8 bits + a = (a + (a >> 8)); // max 16 per 8 bits + a = (a + (a >> 16)); // max 32 per 8 bits + return a & 0xff; +} + +static int shiftsigned(int v, int shift, int bits) +{ + int result; + int z=0; + + if (shift < 0) v <<= -shift; + else v >>= shift; + result = v; + + z = bits; + while (z < 8) { + result += v >> z; + z += bits; + } + return result; +} + +static stbi_uc *bmp_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + uint8 *out; + unsigned int mr=0,mg=0,mb=0,ma=0; + stbi_uc pal[256][4]; + int psize=0,i,j,compress=0,width; + int bpp, flip_vertically, pad, target, offset, hsz; + if (get8(s) != 'B' || get8(s) != 'M') return epuc("not BMP", "Corrupt BMP"); + get32le(s); // discard filesize + get16le(s); // discard reserved + get16le(s); // discard reserved + offset = get32le(s); + hsz = get32le(s); + if (hsz != 12 && hsz != 40 && hsz != 56 && hsz != 108) return epuc("unknown BMP", "BMP type not supported: unknown"); + failure_reason = "bad BMP"; + if (hsz == 12) { + s->img_x = get16le(s); + s->img_y = get16le(s); + } else { + s->img_x = get32le(s); + s->img_y = get32le(s); + } + if (get16le(s) != 1) return 0; + bpp = get16le(s); + if (bpp == 1) return epuc("monochrome", "BMP type not supported: 1-bit"); + flip_vertically = ((int) s->img_y) > 0; + s->img_y = abs((int) s->img_y); + if (hsz == 12) { + if (bpp < 24) + psize = (offset - 14 - 24) / 3; + } else { + compress = get32le(s); + if (compress == 1 || compress == 2) return epuc("BMP RLE", "BMP type not supported: RLE"); + get32le(s); // discard sizeof + get32le(s); // discard hres + get32le(s); // discard vres + get32le(s); // discard colorsused + get32le(s); // discard max important + if (hsz == 40 || hsz == 56) { + if (hsz == 56) { + get32le(s); + get32le(s); + get32le(s); + get32le(s); + } + if (bpp == 16 || bpp == 32) { + mr = mg = mb = 0; + if (compress == 0) { + if (bpp == 32) { + mr = 0xff << 16; + mg = 0xff << 8; + mb = 0xff << 0; + } else { + mr = 31 << 10; + mg = 31 << 5; + mb = 31 << 0; + } + } else if (compress == 3) { + mr = get32le(s); + mg = get32le(s); + mb = get32le(s); + // not documented, but generated by photoshop and handled by mspaint + if (mr == mg && mg == mb) { + // ?!?!? + return NULL; + } + } else + return NULL; + } + } else { + assert(hsz == 108); + mr = get32le(s); + mg = get32le(s); + mb = get32le(s); + ma = get32le(s); + get32le(s); // discard color space + for (i=0; i < 12; ++i) + get32le(s); // discard color space parameters + } + if (bpp < 16) + psize = (offset - 14 - hsz) >> 2; + } + s->img_n = ma ? 4 : 3; + if (req_comp && req_comp >= 3) // we can directly decode 3 or 4 + target = req_comp; + else + target = s->img_n; // if they want monochrome, we'll post-convert + out = (stbi_uc *) malloc(target * s->img_x * s->img_y); + if (!out) return epuc("outofmem", "Out of memory"); + if (bpp < 16) { + int z=0; + if (psize == 0 || psize > 256) { free(out); return epuc("invalid", "Corrupt BMP"); } + for (i=0; i < psize; ++i) { + pal[i][2] = get8(s); + pal[i][1] = get8(s); + pal[i][0] = get8(s); + if (hsz != 12) get8(s); + pal[i][3] = 255; + } + skip(s, offset - 14 - hsz - psize * (hsz == 12 ? 3 : 4)); + if (bpp == 4) width = (s->img_x + 1) >> 1; + else if (bpp == 8) width = s->img_x; + else { free(out); return epuc("bad bpp", "Corrupt BMP"); } + pad = (-width)&3; + for (j=0; j < (int) s->img_y; ++j) { + for (i=0; i < (int) s->img_x; i += 2) { + int v=get8(s),v2=0; + if (bpp == 4) { + v2 = v & 15; + v >>= 4; + } + out[z++] = pal[v][0]; + out[z++] = pal[v][1]; + out[z++] = pal[v][2]; + if (target == 4) out[z++] = 255; + if (i+1 == (int) s->img_x) break; + v = (bpp == 8) ? get8(s) : v2; + out[z++] = pal[v][0]; + out[z++] = pal[v][1]; + out[z++] = pal[v][2]; + if (target == 4) out[z++] = 255; + } + skip(s, pad); + } + } else { + int rshift=0,gshift=0,bshift=0,ashift=0,rcount=0,gcount=0,bcount=0,acount=0; + int z = 0; + int easy=0; + skip(s, offset - 14 - hsz); + if (bpp == 24) width = 3 * s->img_x; + else if (bpp == 16) width = 2*s->img_x; + else /* bpp = 32 and pad = 0 */ width=0; + pad = (-width) & 3; + if (bpp == 24) { + easy = 1; + } else if (bpp == 32) { + if (mb == 0xff && mg == 0xff00 && mr == 0xff000000 && ma == 0xff000000) + easy = 2; + } + if (!easy) { + if (!mr || !mg || !mb) return epuc("bad masks", "Corrupt BMP"); + // right shift amt to put high bit in position #7 + rshift = high_bit(mr)-7; rcount = bitcount(mr); + gshift = high_bit(mg)-7; gcount = bitcount(mr); + bshift = high_bit(mb)-7; bcount = bitcount(mr); + ashift = high_bit(ma)-7; acount = bitcount(mr); + } + for (j=0; j < (int) s->img_y; ++j) { + if (easy) { + for (i=0; i < (int) s->img_x; ++i) { + int a; + out[z+2] = get8(s); + out[z+1] = get8(s); + out[z+0] = get8(s); + z += 3; + a = (easy == 2 ? get8(s) : 255); + if (target == 4) out[z++] = a; + } + } else { + for (i=0; i < (int) s->img_x; ++i) { + uint32 v = (bpp == 16 ? get16le(s) : get32le(s)); + int a; + out[z++] = shiftsigned(v & mr, rshift, rcount); + out[z++] = shiftsigned(v & mg, gshift, gcount); + out[z++] = shiftsigned(v & mb, bshift, bcount); + a = (ma ? shiftsigned(v & ma, ashift, acount) : 255); + if (target == 4) out[z++] = a; + } + } + skip(s, pad); + } + } + if (flip_vertically) { + stbi_uc t; + for (j=0; j < (int) s->img_y>>1; ++j) { + stbi_uc *p1 = out + j *s->img_x*target; + stbi_uc *p2 = out + (s->img_y-1-j)*s->img_x*target; + for (i=0; i < (int) s->img_x*target; ++i) { + t = p1[i], p1[i] = p2[i], p2[i] = t; + } + } + } + + if (req_comp && req_comp != target) { + out = convert_format(out, target, req_comp, s->img_x, s->img_y); + if (out == NULL) return out; // convert_format frees input on failure + } + + *x = s->img_x; + *y = s->img_y; + if (comp) *comp = target; + return out; +} + +#ifndef STBI_NO_STDIO +stbi_uc *stbi_bmp_load (char const *filename, int *x, int *y, int *comp, int req_comp) +{ + stbi_uc *data; + FILE *f = fopen(filename, "rb"); + if (!f) return NULL; + data = stbi_bmp_load_from_file(f, x,y,comp,req_comp); + fclose(f); + return data; +} + +stbi_uc *stbi_bmp_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_file(&s, f); + return bmp_load(&s, x,y,comp,req_comp); +} +#endif + +stbi_uc *stbi_bmp_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_mem(&s, buffer, len); + return bmp_load(&s, x,y,comp,req_comp); +} + +// Targa Truevision - TGA +// by Jonathan Dummer + +static int tga_test(stbi *s) +{ + int sz; + get8u(s); // discard Offset + sz = get8u(s); // color type + if( sz > 1 ) return 0; // only RGB or indexed allowed + sz = get8u(s); // image type + if( (sz != 1) && (sz != 2) && (sz != 3) && (sz != 9) && (sz != 10) && (sz != 11) ) return 0; // only RGB or grey allowed, +/- RLE + get16(s); // discard palette start + get16(s); // discard palette length + get8(s); // discard bits per palette color entry + get16(s); // discard x origin + get16(s); // discard y origin + if( get16(s) < 1 ) return 0; // test width + if( get16(s) < 1 ) return 0; // test height + sz = get8(s); // bits per pixel + if( (sz != 8) && (sz != 16) && (sz != 24) && (sz != 32) ) return 0; // only RGB or RGBA or grey allowed + return 1; // seems to have passed everything +} + +#ifndef STBI_NO_STDIO +int stbi_tga_test_file (FILE *f) +{ + stbi s; + int r,n = ftell(f); + start_file(&s, f); + r = tga_test(&s); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +int stbi_tga_test_memory (stbi_uc const *buffer, int len) +{ + stbi s; + start_mem(&s, buffer, len); + return tga_test(&s); +} + +static stbi_uc *tga_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + // read in the TGA header stuff + int tga_offset = get8u(s); + int tga_indexed = get8u(s); + int tga_image_type = get8u(s); + int tga_is_RLE = 0; + int tga_palette_start = get16le(s); + int tga_palette_len = get16le(s); + int tga_palette_bits = get8u(s); + int tga_x_origin = get16le(s); + int tga_y_origin = get16le(s); + int tga_width = get16le(s); + int tga_height = get16le(s); + int tga_bits_per_pixel = get8u(s); + int tga_inverted = get8u(s); + // image data + unsigned char *tga_data; + unsigned char *tga_palette = NULL; + int i, j; + unsigned char raw_data[4]; + unsigned char trans_data[] = { 0,0,0,0 }; + int RLE_count = 0; + int RLE_repeating = 0; + int read_next_pixel = 1; + // do a tiny bit of precessing + if( tga_image_type >= 8 ) + { + tga_image_type -= 8; + tga_is_RLE = 1; + } + /* int tga_alpha_bits = tga_inverted & 15; */ + tga_inverted = 1 - ((tga_inverted >> 5) & 1); + + // error check + if( //(tga_indexed) || + (tga_width < 1) || (tga_height < 1) || + (tga_image_type < 1) || (tga_image_type > 3) || + ((tga_bits_per_pixel != 8) && (tga_bits_per_pixel != 16) && + (tga_bits_per_pixel != 24) && (tga_bits_per_pixel != 32)) + ) + { + return NULL; + } + + // If I'm paletted, then I'll use the number of bits from the palette + if( tga_indexed ) + { + tga_bits_per_pixel = tga_palette_bits; + } + + // tga info + *x = tga_width; + *y = tga_height; + if( (req_comp < 1) || (req_comp > 4) ) + { + // just use whatever the file was + req_comp = tga_bits_per_pixel / 8; + *comp = req_comp; + } else + { + // force a new number of components + *comp = tga_bits_per_pixel/8; + } + tga_data = (unsigned char*)malloc( tga_width * tga_height * req_comp ); + + // skip to the data's starting position (offset usually = 0) + skip(s, tga_offset ); + // do I need to load a palette? + if( tga_indexed ) + { + // any data to skip? (offset usually = 0) + skip(s, tga_palette_start ); + // load the palette + tga_palette = (unsigned char*)malloc( tga_palette_len * tga_palette_bits / 8 ); + getn(s, tga_palette, tga_palette_len * tga_palette_bits / 8 ); + } + // load the data + for( i = 0; i < tga_width * tga_height; ++i ) + { + // if I'm in RLE mode, do I need to get a RLE chunk? + if( tga_is_RLE ) + { + if( RLE_count == 0 ) + { + // yep, get the next byte as a RLE command + int RLE_cmd = get8u(s); + RLE_count = 1 + (RLE_cmd & 127); + RLE_repeating = RLE_cmd >> 7; + read_next_pixel = 1; + } else if( !RLE_repeating ) + { + read_next_pixel = 1; + } + } else + { + read_next_pixel = 1; + } + // OK, if I need to read a pixel, do it now + if( read_next_pixel ) + { + // load however much data we did have + if( tga_indexed ) + { + // read in 1 byte, then perform the lookup + int pal_idx = get8u(s); + if( pal_idx >= tga_palette_len ) + { + // invalid index + pal_idx = 0; + } + pal_idx *= tga_bits_per_pixel / 8; + for( j = 0; j*8 < tga_bits_per_pixel; ++j ) + { + raw_data[j] = tga_palette[pal_idx+j]; + } + } else + { + // read in the data raw + for( j = 0; j*8 < tga_bits_per_pixel; ++j ) + { + raw_data[j] = get8u(s); + } + } + // convert raw to the intermediate format + switch( tga_bits_per_pixel ) + { + case 8: + // Luminous => RGBA + trans_data[0] = raw_data[0]; + trans_data[1] = raw_data[0]; + trans_data[2] = raw_data[0]; + trans_data[3] = 255; + break; + case 16: + // Luminous,Alpha => RGBA + trans_data[0] = raw_data[0]; + trans_data[1] = raw_data[0]; + trans_data[2] = raw_data[0]; + trans_data[3] = raw_data[1]; + break; + case 24: + // BGR => RGBA + trans_data[0] = raw_data[2]; + trans_data[1] = raw_data[1]; + trans_data[2] = raw_data[0]; + trans_data[3] = 255; + break; + case 32: + // BGRA => RGBA + trans_data[0] = raw_data[2]; + trans_data[1] = raw_data[1]; + trans_data[2] = raw_data[0]; + trans_data[3] = raw_data[3]; + break; + } + // clear the reading flag for the next pixel + read_next_pixel = 0; + } // end of reading a pixel + // convert to final format + switch( req_comp ) + { + case 1: + // RGBA => Luminance + tga_data[i*req_comp+0] = compute_y(trans_data[0],trans_data[1],trans_data[2]); + break; + case 2: + // RGBA => Luminance,Alpha + tga_data[i*req_comp+0] = compute_y(trans_data[0],trans_data[1],trans_data[2]); + tga_data[i*req_comp+1] = trans_data[3]; + break; + case 3: + // RGBA => RGB + tga_data[i*req_comp+0] = trans_data[0]; + tga_data[i*req_comp+1] = trans_data[1]; + tga_data[i*req_comp+2] = trans_data[2]; + break; + case 4: + // RGBA => RGBA + tga_data[i*req_comp+0] = trans_data[0]; + tga_data[i*req_comp+1] = trans_data[1]; + tga_data[i*req_comp+2] = trans_data[2]; + tga_data[i*req_comp+3] = trans_data[3]; + break; + } + // in case we're in RLE mode, keep counting down + --RLE_count; + } + // do I need to invert the image? + if( tga_inverted ) + { + for( j = 0; j*2 < tga_height; ++j ) + { + int index1 = j * tga_width * req_comp; + int index2 = (tga_height - 1 - j) * tga_width * req_comp; + for( i = tga_width * req_comp; i > 0; --i ) + { + unsigned char temp = tga_data[index1]; + tga_data[index1] = tga_data[index2]; + tga_data[index2] = temp; + ++index1; + ++index2; + } + } + } + // clear my palette, if I had one + if( tga_palette != NULL ) + { + free( tga_palette ); + } + // the things I do to get rid of an error message, and yet keep + // Microsoft's C compilers happy... [8^( + tga_palette_start = tga_palette_len = tga_palette_bits = + tga_x_origin = tga_y_origin = 0; + // OK, done + return tga_data; +} + +#ifndef STBI_NO_STDIO +stbi_uc *stbi_tga_load (char const *filename, int *x, int *y, int *comp, int req_comp) +{ + stbi_uc *data; + FILE *f = fopen(filename, "rb"); + if (!f) return NULL; + data = stbi_tga_load_from_file(f, x,y,comp,req_comp); + fclose(f); + return data; +} + +stbi_uc *stbi_tga_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_file(&s, f); + return tga_load(&s, x,y,comp,req_comp); +} +#endif + +stbi_uc *stbi_tga_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_mem(&s, buffer, len); + return tga_load(&s, x,y,comp,req_comp); +} + + +// ************************************************************************************************* +// Photoshop PSD loader -- PD by Thatcher Ulrich, integration by Nicholas Schulz, tweaked by STB + +static int psd_test(stbi *s) +{ + if (get32(s) != 0x38425053) return 0; // "8BPS" + else return 1; +} + +#ifndef STBI_NO_STDIO +int stbi_psd_test_file(FILE *f) +{ + stbi s; + int r,n = ftell(f); + start_file(&s, f); + r = psd_test(&s); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +int stbi_psd_test_memory(stbi_uc const *buffer, int len) +{ + stbi s; + start_mem(&s, buffer, len); + return psd_test(&s); +} + +static stbi_uc *psd_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + int pixelCount; + int channelCount, compression; + int channel, i, count, len; + int w,h; + uint8 *out; + + // Check identifier + if (get32(s) != 0x38425053) // "8BPS" + return epuc("not PSD", "Corrupt PSD image"); + + // Check file type version. + if (get16(s) != 1) + return epuc("wrong version", "Unsupported version of PSD image"); + + // Skip 6 reserved bytes. + skip(s, 6 ); + + // Read the number of channels (R, G, B, A, etc). + channelCount = get16(s); + if (channelCount < 0 || channelCount > 16) + return epuc("wrong channel count", "Unsupported number of channels in PSD image"); + + // Read the rows and columns of the image. + h = get32(s); + w = get32(s); + + // Make sure the depth is 8 bits. + if (get16(s) != 8) + return epuc("unsupported bit depth", "PSD bit depth is not 8 bit"); + + // Make sure the color mode is RGB. + // Valid options are: + // 0: Bitmap + // 1: Grayscale + // 2: Indexed color + // 3: RGB color + // 4: CMYK color + // 7: Multichannel + // 8: Duotone + // 9: Lab color + if (get16(s) != 3) + return epuc("wrong color format", "PSD is not in RGB color format"); + + // Skip the Mode Data. (It's the palette for indexed color; other info for other modes.) + skip(s,get32(s) ); + + // Skip the image resources. (resolution, pen tool paths, etc) + skip(s, get32(s) ); + + // Skip the reserved data. + skip(s, get32(s) ); + + // Find out if the data is compressed. + // Known values: + // 0: no compression + // 1: RLE compressed + compression = get16(s); + if (compression > 1) + return epuc("bad compression", "PSD has an unknown compression format"); + + // Create the destination image. + out = (stbi_uc *) malloc(4 * w*h); + if (!out) return epuc("outofmem", "Out of memory"); + pixelCount = w*h; + + // Initialize the data to zero. + //memset( out, 0, pixelCount * 4 ); + + // Finally, the image data. + if (compression) { + // RLE as used by .PSD and .TIFF + // Loop until you get the number of unpacked bytes you are expecting: + // Read the next source byte into n. + // If n is between 0 and 127 inclusive, copy the next n+1 bytes literally. + // Else if n is between -127 and -1 inclusive, copy the next byte -n+1 times. + // Else if n is 128, noop. + // Endloop + + // The RLE-compressed data is preceeded by a 2-byte data count for each row in the data, + // which we're going to just skip. + skip(s, h * channelCount * 2 ); + + // Read the RLE data by channel. + for (channel = 0; channel < 4; channel++) { + uint8 *p; + + p = out+channel; + if (channel >= channelCount) { + // Fill this channel with default data. + for (i = 0; i < pixelCount; i++) *p = (channel == 3 ? 255 : 0), p += 4; + } else { + // Read the RLE data. + count = 0; + while (count < pixelCount) { + len = get8(s); + if (len == 128) { + // No-op. + } else if (len < 128) { + // Copy next len+1 bytes literally. + len++; + count += len; + while (len) { + *p = get8(s); + p += 4; + len--; + } + } else if (len > 128) { + uint32 val; + // Next -len+1 bytes in the dest are replicated from next source byte. + // (Interpret len as a negative 8-bit int.) + len ^= 0x0FF; + len += 2; + val = get8(s); + count += len; + while (len) { + *p = val; + p += 4; + len--; + } + } + } + } + } + + } else { + // We're at the raw image data. It's each channel in order (Red, Green, Blue, Alpha, ...) + // where each channel consists of an 8-bit value for each pixel in the image. + + // Read the data by channel. + for (channel = 0; channel < 4; channel++) { + uint8 *p; + + p = out + channel; + if (channel > channelCount) { + // Fill this channel with default data. + for (i = 0; i < pixelCount; i++) *p = channel == 3 ? 255 : 0, p += 4; + } else { + // Read the data. + count = 0; + for (i = 0; i < pixelCount; i++) + *p = get8(s), p += 4; + } + } + } + + if (req_comp && req_comp != 4) { + out = convert_format(out, 4, req_comp, w, h); + if (out == NULL) return out; // convert_format frees input on failure + } + + if (comp) *comp = channelCount; + *y = h; + *x = w; + + return out; +} + +#ifndef STBI_NO_STDIO +stbi_uc *stbi_psd_load(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + stbi_uc *data; + FILE *f = fopen(filename, "rb"); + if (!f) return NULL; + data = stbi_psd_load_from_file(f, x,y,comp,req_comp); + fclose(f); + return data; +} + +stbi_uc *stbi_psd_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_file(&s, f); + return psd_load(&s, x,y,comp,req_comp); +} +#endif + +stbi_uc *stbi_psd_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_mem(&s, buffer, len); + return psd_load(&s, x,y,comp,req_comp); +} + + +// ************************************************************************************************* +// Radiance RGBE HDR loader +// originally by Nicolas Schulz +#ifndef STBI_NO_HDR +static int hdr_test(stbi *s) +{ + char *signature = "#?RADIANCE\n"; + int i; + for (i=0; signature[i]; ++i) + if (get8(s) != signature[i]) + return 0; + return 1; +} + +int stbi_hdr_test_memory(stbi_uc const *buffer, int len) +{ + stbi s; + start_mem(&s, buffer, len); + return hdr_test(&s); +} + +#ifndef STBI_NO_STDIO +int stbi_hdr_test_file(FILE *f) +{ + stbi s; + int r,n = ftell(f); + start_file(&s, f); + r = hdr_test(&s); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +#define HDR_BUFLEN 1024 +static char *hdr_gettoken(stbi *z, char *buffer) +{ + int len=0; + //char *s = buffer, + char c = '\0'; + + c = get8(z); + + while (!at_eof(z) && c != '\n') { + buffer[len++] = c; + if (len == HDR_BUFLEN-1) { + // flush to end of line + while (!at_eof(z) && get8(z) != '\n') + ; + break; + } + c = get8(z); + } + + buffer[len] = 0; + return buffer; +} + +static void hdr_convert(float *output, stbi_uc *input, int req_comp) +{ + if( input[3] != 0 ) { + float f1; + // Exponent + f1 = (float) ldexp(1.0f, input[3] - (int)(128 + 8)); + if (req_comp <= 2) + output[0] = (input[0] + input[1] + input[2]) * f1 / 3; + else { + output[0] = input[0] * f1; + output[1] = input[1] * f1; + output[2] = input[2] * f1; + } + if (req_comp == 2) output[1] = 1; + if (req_comp == 4) output[3] = 1; + } else { + switch (req_comp) { + case 4: output[3] = 1; /* fallthrough */ + case 3: output[0] = output[1] = output[2] = 0; + break; + case 2: output[1] = 1; /* fallthrough */ + case 1: output[0] = 0; + break; + } + } +} + + +static float *hdr_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + char buffer[HDR_BUFLEN]; + char *token; + int valid = 0; + int width, height; + stbi_uc *scanline; + float *hdr_data; + int len; + unsigned char count, value; + int i, j, k, c1,c2, z; + + + // Check identifier + if (strcmp(hdr_gettoken(s,buffer), "#?RADIANCE") != 0) + return epf("not HDR", "Corrupt HDR image"); + + // Parse header + while(1) { + token = hdr_gettoken(s,buffer); + if (token[0] == 0) break; + if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1; + } + + if (!valid) return epf("unsupported format", "Unsupported HDR format"); + + // Parse width and height + // can't use sscanf() if we're not using stdio! + token = hdr_gettoken(s,buffer); + if (strncmp(token, "-Y ", 3)) return epf("unsupported data layout", "Unsupported HDR format"); + token += 3; + height = strtol(token, &token, 10); + while (*token == ' ') ++token; + if (strncmp(token, "+X ", 3)) return epf("unsupported data layout", "Unsupported HDR format"); + token += 3; + width = strtol(token, NULL, 10); + + *x = width; + *y = height; + + *comp = 3; + if (req_comp == 0) req_comp = 3; + + // Read data + hdr_data = (float *) malloc(height * width * req_comp * sizeof(float)); + + // Load image data + // image data is stored as some number of sca + if( width < 8 || width >= 32768) { + // Read flat data + for (j=0; j < height; ++j) { + for (i=0; i < width; ++i) { + stbi_uc rgbe[4]; + main_decode_loop: + getn(s, rgbe, 4); + hdr_convert(hdr_data + j * width * req_comp + i * req_comp, rgbe, req_comp); + } + } + } else { + // Read RLE-encoded data + scanline = NULL; + + for (j = 0; j < height; ++j) { + c1 = get8(s); + c2 = get8(s); + len = get8(s); + if (c1 != 2 || c2 != 2 || (len & 0x80)) { + // not run-length encoded, so we have to actually use THIS data as a decoded + // pixel (note this can't be a valid pixel--one of RGB must be >= 128) + stbi_uc rgbe[4] = { c1,c2,len, get8(s) }; + hdr_convert(hdr_data, rgbe, req_comp); + i = 1; + j = 0; + free(scanline); + goto main_decode_loop; // yes, this is fucking insane; blame the fucking insane format + } + len <<= 8; + len |= get8(s); + if (len != width) { free(hdr_data); free(scanline); return epf("invalid decoded scanline length", "corrupt HDR"); } + if (scanline == NULL) scanline = (stbi_uc *) malloc(width * 4); + + for (k = 0; k < 4; ++k) { + i = 0; + while (i < width) { + count = get8(s); + if (count > 128) { + // Run + value = get8(s); + count -= 128; + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = value; + } else { + // Dump + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = get8(s); + } + } + } + for (i=0; i < width; ++i) + hdr_convert(hdr_data+(j*width + i)*req_comp, scanline + i*4, req_comp); + } + free(scanline); + } + + return hdr_data; +} + +static stbi_uc *hdr_load_rgbe(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + char buffer[HDR_BUFLEN]; + char *token; + int valid = 0; + int width, height; + stbi_uc *scanline; + stbi_uc *rgbe_data; + int len; + unsigned char count, value; + int i, j, k, c1,c2, z; + + + // Check identifier + if (strcmp(hdr_gettoken(s,buffer), "#?RADIANCE") != 0) + return epuc("not HDR", "Corrupt HDR image"); + + // Parse header + while(1) { + token = hdr_gettoken(s,buffer); + if (token[0] == 0) break; + if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1; + } + + if (!valid) return epuc("unsupported format", "Unsupported HDR format"); + + // Parse width and height + // can't use sscanf() if we're not using stdio! + token = hdr_gettoken(s,buffer); + if (strncmp(token, "-Y ", 3)) return epuc("unsupported data layout", "Unsupported HDR format"); + token += 3; + height = strtol(token, &token, 10); + while (*token == ' ') ++token; + if (strncmp(token, "+X ", 3)) return epuc("unsupported data layout", "Unsupported HDR format"); + token += 3; + width = strtol(token, NULL, 10); + + *x = width; + *y = height; + + // RGBE _MUST_ come out as 4 components + *comp = 4; + req_comp = 4; + + // Read data + rgbe_data = (stbi_uc *) malloc(height * width * req_comp * sizeof(stbi_uc)); + // point to the beginning + scanline = rgbe_data; + + // Load image data + // image data is stored as some number of scan lines + if( width < 8 || width >= 32768) { + // Read flat data + for (j=0; j < height; ++j) { + for (i=0; i < width; ++i) { + main_decode_loop: + //getn(rgbe, 4); + getn(s,scanline, 4); + scanline += 4; + } + } + } else { + // Read RLE-encoded data + for (j = 0; j < height; ++j) { + c1 = get8(s); + c2 = get8(s); + len = get8(s); + if (c1 != 2 || c2 != 2 || (len & 0x80)) { + // not run-length encoded, so we have to actually use THIS data as a decoded + // pixel (note this can't be a valid pixel--one of RGB must be >= 128) + scanline[0] = c1; + scanline[1] = c2; + scanline[2] = len; + scanline[3] = get8(s); + scanline += 4; + i = 1; + j = 0; + goto main_decode_loop; // yes, this is insane; blame the insane format + } + len <<= 8; + len |= get8(s); + if (len != width) { free(rgbe_data); return epuc("invalid decoded scanline length", "corrupt HDR"); } + for (k = 0; k < 4; ++k) { + i = 0; + while (i < width) { + count = get8(s); + if (count > 128) { + // Run + value = get8(s); + count -= 128; + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = value; + } else { + // Dump + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = get8(s); + } + } + } + // move the scanline on + scanline += 4 * width; + } + } + + return rgbe_data; +} + +#ifndef STBI_NO_STDIO +float *stbi_hdr_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_file(&s,f); + return hdr_load(&s,x,y,comp,req_comp); +} + +stbi_uc *stbi_hdr_load_rgbe_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_file(&s,f); + return hdr_load_rgbe(&s,x,y,comp,req_comp); +} + +stbi_uc *stbi_hdr_load_rgbe (char const *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = fopen(filename, "rb"); + unsigned char *result; + if (!f) return epuc("can't fopen", "Unable to open file"); + result = stbi_hdr_load_rgbe_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} +#endif + +float *stbi_hdr_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_mem(&s,buffer, len); + return hdr_load(&s,x,y,comp,req_comp); +} + +stbi_uc *stbi_hdr_load_rgbe_memory(stbi_uc *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_mem(&s,buffer, len); + return hdr_load_rgbe(&s,x,y,comp,req_comp); +} + +#endif // STBI_NO_HDR + +/////////////////////// write image /////////////////////// + +#ifndef STBI_NO_WRITE + +static void write8(FILE *f, int x) { uint8 z = (uint8) x; fwrite(&z,1,1,f); } + +static void writefv(FILE *f, char *fmt, va_list v) +{ + while (*fmt) { + switch (*fmt++) { + case ' ': break; + case '1': { uint8 x = va_arg(v, int); write8(f,x); break; } + case '2': { int16 x = va_arg(v, int); write8(f,x); write8(f,x>>8); break; } + case '4': { int32 x = va_arg(v, int); write8(f,x); write8(f,x>>8); write8(f,x>>16); write8(f,x>>24); break; } + default: + assert(0); + va_end(v); + return; + } + } +} + +static void writef(FILE *f, char *fmt, ...) +{ + va_list v; + va_start(v, fmt); + writefv(f,fmt,v); + va_end(v); +} + +static void write_pixels(FILE *f, int rgb_dir, int vdir, int x, int y, int comp, void *data, int write_alpha, int scanline_pad) +{ + uint8 bg[3] = { 255, 0, 255}, px[3]; + uint32 zero = 0; + int i,j,k, j_end; + + if (vdir < 0) + j_end = -1, j = y-1; + else + j_end = y, j = 0; + + for (; j != j_end; j += vdir) { + for (i=0; i < x; ++i) { + uint8 *d = (uint8 *) data + (j*x+i)*comp; + if (write_alpha < 0) + fwrite(&d[comp-1], 1, 1, f); + switch (comp) { + case 1: + case 2: writef(f, "111", d[0],d[0],d[0]); + break; + case 4: + if (!write_alpha) { + for (k=0; k < 3; ++k) + px[k] = bg[k] + ((d[k] - bg[k]) * d[3])/255; + writef(f, "111", px[1-rgb_dir],px[1],px[1+rgb_dir]); + break; + } + /* FALLTHROUGH */ + case 3: + writef(f, "111", d[1-rgb_dir],d[1],d[1+rgb_dir]); + break; + } + if (write_alpha > 0) + fwrite(&d[comp-1], 1, 1, f); + } + fwrite(&zero,scanline_pad,1,f); + } +} + +static int outfile(char const *filename, int rgb_dir, int vdir, int x, int y, int comp, void *data, int alpha, int pad, char *fmt, ...) +{ + FILE *f = fopen(filename, "wb"); + if (f) { + va_list v; + va_start(v, fmt); + writefv(f, fmt, v); + va_end(v); + write_pixels(f,rgb_dir,vdir,x,y,comp,data,alpha,pad); + fclose(f); + } + return f != NULL; +} + +int stbi_write_bmp(char const *filename, int x, int y, int comp, void *data) +{ + int pad = (-x*3) & 3; + return outfile(filename,-1,-1,x,y,comp,data,0,pad, + "11 4 22 4" "4 44 22 444444", + 'B', 'M', 14+40+(x*3+pad)*y, 0,0, 14+40, // file header + 40, x,y, 1,24, 0,0,0,0,0,0); // bitmap header +} + +int stbi_write_tga(char const *filename, int x, int y, int comp, void *data) +{ + int has_alpha = !(comp & 1); + return outfile(filename, -1,-1, x, y, comp, data, has_alpha, 0, + "111 221 2222 11", 0,0,2, 0,0,0, 0,0,x,y, 24+8*has_alpha, 8*has_alpha); +} + +// any other image formats that do interleaved rgb data? +// PNG: requires adler32,crc32 -- significant amount of code +// PSD: no, channels output separately +// TIFF: no, stripwise-interleaved... i think + +#endif // STBI_NO_WRITE + +// add in my DDS loading support +#ifndef STBI_NO_DDS +#include "stbi_DDS_aug_c.h" +#endif diff --git a/lib/soil/soil/stb_image_aug.h b/lib/soil/soil/stb_image_aug.h new file mode 100644 index 0000000..52ea75c --- /dev/null +++ b/lib/soil/soil/stb_image_aug.h @@ -0,0 +1,354 @@ +/* stbi-1.16 - public domain JPEG/PNG reader - http://nothings.org/stb_image.c + when you control the images you're loading + + QUICK NOTES: + Primarily of interest to game developers and other people who can + avoid problematic images and only need the trivial interface + + JPEG baseline (no JPEG progressive, no oddball channel decimations) + PNG non-interlaced + BMP non-1bpp, non-RLE + TGA (not sure what subset, if a subset) + PSD (composited view only, no extra channels) + HDR (radiance rgbE format) + writes BMP,TGA (define STBI_NO_WRITE to remove code) + decoded from memory or through stdio FILE (define STBI_NO_STDIO to remove code) + supports installable dequantizing-IDCT, YCbCr-to-RGB conversion (define STBI_SIMD) + + TODO: + stbi_info_* + + history: + 1.16 major bugfix - convert_format converted one too many pixels + 1.15 initialize some fields for thread safety + 1.14 fix threadsafe conversion bug; header-file-only version (#define STBI_HEADER_FILE_ONLY before including) + 1.13 threadsafe + 1.12 const qualifiers in the API + 1.11 Support installable IDCT, colorspace conversion routines + 1.10 Fixes for 64-bit (don't use "unsigned long") + optimized upsampling by Fabian "ryg" Giesen + 1.09 Fix format-conversion for PSD code (bad global variables!) + 1.08 Thatcher Ulrich's PSD code integrated by Nicolas Schulz + 1.07 attempt to fix C++ warning/errors again + 1.06 attempt to fix C++ warning/errors again + 1.05 fix TGA loading to return correct *comp and use good luminance calc + 1.04 default float alpha is 1, not 255; use 'void *' for stbi_image_free + 1.03 bugfixes to STBI_NO_STDIO, STBI_NO_HDR + 1.02 support for (subset of) HDR files, float interface for preferred access to them + 1.01 fix bug: possible bug in handling right-side up bmps... not sure + fix bug: the stbi_bmp_load() and stbi_tga_load() functions didn't work at all + 1.00 interface to zlib that skips zlib header + 0.99 correct handling of alpha in palette + 0.98 TGA loader by lonesock; dynamically add loaders (untested) + 0.97 jpeg errors on too large a file; also catch another malloc failure + 0.96 fix detection of invalid v value - particleman@mollyrocket forum + 0.95 during header scan, seek to markers in case of padding + 0.94 STBI_NO_STDIO to disable stdio usage; rename all #defines the same + 0.93 handle jpegtran output; verbose errors + 0.92 read 4,8,16,24,32-bit BMP files of several formats + 0.91 output 24-bit Windows 3.0 BMP files + 0.90 fix a few more warnings; bump version number to approach 1.0 + 0.61 bugfixes due to Marc LeBlanc, Christopher Lloyd + 0.60 fix compiling as c++ + 0.59 fix warnings: merge Dave Moore's -Wall fixes + 0.58 fix bug: zlib uncompressed mode len/nlen was wrong endian + 0.57 fix bug: jpg last huffman symbol before marker was >9 bits but less + than 16 available + 0.56 fix bug: zlib uncompressed mode len vs. nlen + 0.55 fix bug: restart_interval not initialized to 0 + 0.54 allow NULL for 'int *comp' + 0.53 fix bug in png 3->4; speedup png decoding + 0.52 png handles req_comp=3,4 directly; minor cleanup; jpeg comments + 0.51 obey req_comp requests, 1-component jpegs return as 1-component, + on 'test' only check type, not whether we support this variant +*/ + +#ifndef HEADER_STB_IMAGE_AUGMENTED +#define HEADER_STB_IMAGE_AUGMENTED + +//// begin header file //////////////////////////////////////////////////// +// +// Limitations: +// - no progressive/interlaced support (jpeg, png) +// - 8-bit samples only (jpeg, png) +// - not threadsafe +// - channel subsampling of at most 2 in each dimension (jpeg) +// - no delayed line count (jpeg) -- IJG doesn't support either +// +// Basic usage (see HDR discussion below): +// int x,y,n; +// unsigned char *data = stbi_load(filename, &x, &y, &n, 0); +// // ... process data if not NULL ... +// // ... x = width, y = height, n = # 8-bit components per pixel ... +// // ... replace '0' with '1'..'4' to force that many components per pixel +// stbi_image_free(data) +// +// Standard parameters: +// int *x -- outputs image width in pixels +// int *y -- outputs image height in pixels +// int *comp -- outputs # of image components in image file +// int req_comp -- if non-zero, # of image components requested in result +// +// The return value from an image loader is an 'unsigned char *' which points +// to the pixel data. The pixel data consists of *y scanlines of *x pixels, +// with each pixel consisting of N interleaved 8-bit components; the first +// pixel pointed to is top-left-most in the image. There is no padding between +// image scanlines or between pixels, regardless of format. The number of +// components N is 'req_comp' if req_comp is non-zero, or *comp otherwise. +// If req_comp is non-zero, *comp has the number of components that _would_ +// have been output otherwise. E.g. if you set req_comp to 4, you will always +// get RGBA output, but you can check *comp to easily see if it's opaque. +// +// An output image with N components has the following components interleaved +// in this order in each pixel: +// +// N=#comp components +// 1 grey +// 2 grey, alpha +// 3 red, green, blue +// 4 red, green, blue, alpha +// +// If image loading fails for any reason, the return value will be NULL, +// and *x, *y, *comp will be unchanged. The function stbi_failure_reason() +// can be queried for an extremely brief, end-user unfriendly explanation +// of why the load failed. Define STBI_NO_FAILURE_STRINGS to avoid +// compiling these strings at all, and STBI_FAILURE_USERMSG to get slightly +// more user-friendly ones. +// +// Paletted PNG and BMP images are automatically depalettized. +// +// +// =========================================================================== +// +// HDR image support (disable by defining STBI_NO_HDR) +// +// stb_image now supports loading HDR images in general, and currently +// the Radiance .HDR file format, although the support is provided +// generically. You can still load any file through the existing interface; +// if you attempt to load an HDR file, it will be automatically remapped to +// LDR, assuming gamma 2.2 and an arbitrary scale factor defaulting to 1; +// both of these constants can be reconfigured through this interface: +// +// stbi_hdr_to_ldr_gamma(2.2f); +// stbi_hdr_to_ldr_scale(1.0f); +// +// (note, do not use _inverse_ constants; stbi_image will invert them +// appropriately). +// +// Additionally, there is a new, parallel interface for loading files as +// (linear) floats to preserve the full dynamic range: +// +// float *data = stbi_loadf(filename, &x, &y, &n, 0); +// +// If you load LDR images through this interface, those images will +// be promoted to floating point values, run through the inverse of +// constants corresponding to the above: +// +// stbi_ldr_to_hdr_scale(1.0f); +// stbi_ldr_to_hdr_gamma(2.2f); +// +// Finally, given a filename (or an open file or memory block--see header +// file for details) containing image data, you can query for the "most +// appropriate" interface to use (that is, whether the image is HDR or +// not), using: +// +// stbi_is_hdr(char *filename); + +#ifndef STBI_NO_STDIO +#include +#endif + +#define STBI_VERSION 1 + +enum +{ + STBI_default = 0, // only used for req_comp + + STBI_grey = 1, + STBI_grey_alpha = 2, + STBI_rgb = 3, + STBI_rgb_alpha = 4, +}; + +typedef unsigned char stbi_uc; + +#ifdef __cplusplus +extern "C" { +#endif + +// WRITING API + +#if !defined(STBI_NO_WRITE) && !defined(STBI_NO_STDIO) +// write a BMP/TGA file given tightly packed 'comp' channels (no padding, nor bmp-stride-padding) +// (you must include the appropriate extension in the filename). +// returns TRUE on success, FALSE if couldn't open file, error writing file +extern int stbi_write_bmp (char const *filename, int x, int y, int comp, void *data); +extern int stbi_write_tga (char const *filename, int x, int y, int comp, void *data); +#endif + +// PRIMARY API - works on images of any type + +// load image by filename, open file, or memory buffer +#ifndef STBI_NO_STDIO +extern stbi_uc *stbi_load (char const *filename, int *x, int *y, int *comp, int req_comp); +extern stbi_uc *stbi_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +extern int stbi_info_from_file (FILE *f, int *x, int *y, int *comp); +#endif +extern stbi_uc *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); +// for stbi_load_from_file, file pointer is left pointing immediately after image + +#ifndef STBI_NO_HDR +#ifndef STBI_NO_STDIO +extern float *stbi_loadf (char const *filename, int *x, int *y, int *comp, int req_comp); +extern float *stbi_loadf_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +#endif +extern float *stbi_loadf_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); + +extern void stbi_hdr_to_ldr_gamma(float gamma); +extern void stbi_hdr_to_ldr_scale(float scale); + +extern void stbi_ldr_to_hdr_gamma(float gamma); +extern void stbi_ldr_to_hdr_scale(float scale); + +#endif // STBI_NO_HDR + +// get a VERY brief reason for failure +// NOT THREADSAFE +extern char *stbi_failure_reason (void); + +// free the loaded image -- this is just free() +extern void stbi_image_free (void *retval_from_stbi_load); + +// get image dimensions & components without fully decoding +extern int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp); +extern int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len); +#ifndef STBI_NO_STDIO +extern int stbi_info (char const *filename, int *x, int *y, int *comp); +extern int stbi_is_hdr (char const *filename); +extern int stbi_is_hdr_from_file(FILE *f); +#endif + +// ZLIB client - used by PNG, available for other purposes + +extern char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen); +extern char *stbi_zlib_decode_malloc(const char *buffer, int len, int *outlen); +extern int stbi_zlib_decode_buffer(char *obuffer, int olen, const char *ibuffer, int ilen); + +extern char *stbi_zlib_decode_noheader_malloc(const char *buffer, int len, int *outlen); +extern int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen); + +// TYPE-SPECIFIC ACCESS + +// is it a jpeg? +extern int stbi_jpeg_test_memory (stbi_uc const *buffer, int len); +extern stbi_uc *stbi_jpeg_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); +extern int stbi_jpeg_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp); + +#ifndef STBI_NO_STDIO +extern stbi_uc *stbi_jpeg_load (char const *filename, int *x, int *y, int *comp, int req_comp); +extern int stbi_jpeg_test_file (FILE *f); +extern stbi_uc *stbi_jpeg_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); + +extern int stbi_jpeg_info (char const *filename, int *x, int *y, int *comp); +extern int stbi_jpeg_info_from_file (FILE *f, int *x, int *y, int *comp); +#endif + +// is it a png? +extern int stbi_png_test_memory (stbi_uc const *buffer, int len); +extern stbi_uc *stbi_png_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); +extern int stbi_png_info_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp); + +#ifndef STBI_NO_STDIO +extern stbi_uc *stbi_png_load (char const *filename, int *x, int *y, int *comp, int req_comp); +extern int stbi_png_info (char const *filename, int *x, int *y, int *comp); +extern int stbi_png_test_file (FILE *f); +extern stbi_uc *stbi_png_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +extern int stbi_png_info_from_file (FILE *f, int *x, int *y, int *comp); +#endif + +// is it a bmp? +extern int stbi_bmp_test_memory (stbi_uc const *buffer, int len); + +extern stbi_uc *stbi_bmp_load (char const *filename, int *x, int *y, int *comp, int req_comp); +extern stbi_uc *stbi_bmp_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); +#ifndef STBI_NO_STDIO +extern int stbi_bmp_test_file (FILE *f); +extern stbi_uc *stbi_bmp_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +#endif + +// is it a tga? +extern int stbi_tga_test_memory (stbi_uc const *buffer, int len); + +extern stbi_uc *stbi_tga_load (char const *filename, int *x, int *y, int *comp, int req_comp); +extern stbi_uc *stbi_tga_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); +#ifndef STBI_NO_STDIO +extern int stbi_tga_test_file (FILE *f); +extern stbi_uc *stbi_tga_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +#endif + +// is it a psd? +extern int stbi_psd_test_memory (stbi_uc const *buffer, int len); + +extern stbi_uc *stbi_psd_load (char const *filename, int *x, int *y, int *comp, int req_comp); +extern stbi_uc *stbi_psd_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); +#ifndef STBI_NO_STDIO +extern int stbi_psd_test_file (FILE *f); +extern stbi_uc *stbi_psd_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +#endif + +// is it an hdr? +extern int stbi_hdr_test_memory (stbi_uc const *buffer, int len); + +extern float * stbi_hdr_load (char const *filename, int *x, int *y, int *comp, int req_comp); +extern float * stbi_hdr_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); +extern stbi_uc *stbi_hdr_load_rgbe (char const *filename, int *x, int *y, int *comp, int req_comp); +extern float * stbi_hdr_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); +#ifndef STBI_NO_STDIO +extern int stbi_hdr_test_file (FILE *f); +extern float * stbi_hdr_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +extern stbi_uc *stbi_hdr_load_rgbe_file (FILE *f, int *x, int *y, int *comp, int req_comp); +#endif + +// define new loaders +typedef struct +{ + int (*test_memory)(stbi_uc const *buffer, int len); + stbi_uc * (*load_from_memory)(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); + #ifndef STBI_NO_STDIO + int (*test_file)(FILE *f); + stbi_uc * (*load_from_file)(FILE *f, int *x, int *y, int *comp, int req_comp); + #endif +} stbi_loader; + +// register a loader by filling out the above structure (you must defined ALL functions) +// returns 1 if added or already added, 0 if not added (too many loaders) +// NOT THREADSAFE +extern int stbi_register_loader(stbi_loader *loader); + +// define faster low-level operations (typically SIMD support) +#if STBI_SIMD +typedef void (*stbi_idct_8x8)(uint8 *out, int out_stride, short data[64], unsigned short *dequantize); +// compute an integer IDCT on "input" +// input[x] = data[x] * dequantize[x] +// write results to 'out': 64 samples, each run of 8 spaced by 'out_stride' +// CLAMP results to 0..255 +typedef void (*stbi_YCbCr_to_RGB_run)(uint8 *output, uint8 const *y, uint8 const *cb, uint8 const *cr, int count, int step); +// compute a conversion from YCbCr to RGB +// 'count' pixels +// write pixels to 'output'; each pixel is 'step' bytes (either 3 or 4; if 4, write '255' as 4th), order R,G,B +// y: Y input channel +// cb: Cb input channel; scale/biased to be 0..255 +// cr: Cr input channel; scale/biased to be 0..255 + +extern void stbi_install_idct(stbi_idct_8x8 func); +extern void stbi_install_YCbCr_to_RGB(stbi_YCbCr_to_RGB_run func); +#endif // STBI_SIMD + +#ifdef __cplusplus +} +#endif + +// +// +//// end header file ///////////////////////////////////////////////////// +#endif // STBI_INCLUDE_STB_IMAGE_H diff --git a/lib/soil/soil/stbi_DDS_aug.h b/lib/soil/soil/stbi_DDS_aug.h new file mode 100644 index 0000000..7317d63 --- /dev/null +++ b/lib/soil/soil/stbi_DDS_aug.h @@ -0,0 +1,21 @@ +/* + adding DDS loading support to stbi +*/ + +#ifndef HEADER_STB_IMAGE_DDS_AUGMENTATION +#define HEADER_STB_IMAGE_DDS_AUGMENTATION + +// is it a DDS file? +extern int stbi_dds_test_memory (stbi_uc const *buffer, int len); + +extern stbi_uc *stbi_dds_load (char *filename, int *x, int *y, int *comp, int req_comp); +extern stbi_uc *stbi_dds_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); +#ifndef STBI_NO_STDIO +extern int stbi_dds_test_file (FILE *f); +extern stbi_uc *stbi_dds_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +#endif + +// +// +//// end header file ///////////////////////////////////////////////////// +#endif // HEADER_STB_IMAGE_DDS_AUGMENTATION diff --git a/lib/soil/soil/stbi_DDS_aug_c.h b/lib/soil/soil/stbi_DDS_aug_c.h new file mode 100644 index 0000000..683d1cf --- /dev/null +++ b/lib/soil/soil/stbi_DDS_aug_c.h @@ -0,0 +1,511 @@ + +/// DDS file support, does decoding, _not_ direct uploading +/// (use SOIL for that ;-) + +/// A bunch of DirectDraw Surface structures and flags +typedef struct { + unsigned int dwMagic; + unsigned int dwSize; + unsigned int dwFlags; + unsigned int dwHeight; + unsigned int dwWidth; + unsigned int dwPitchOrLinearSize; + unsigned int dwDepth; + unsigned int dwMipMapCount; + unsigned int dwReserved1[ 11 ]; + + // DDPIXELFORMAT + struct { + unsigned int dwSize; + unsigned int dwFlags; + unsigned int dwFourCC; + unsigned int dwRGBBitCount; + unsigned int dwRBitMask; + unsigned int dwGBitMask; + unsigned int dwBBitMask; + unsigned int dwAlphaBitMask; + } sPixelFormat; + + // DDCAPS2 + struct { + unsigned int dwCaps1; + unsigned int dwCaps2; + unsigned int dwDDSX; + unsigned int dwReserved; + } sCaps; + unsigned int dwReserved2; +} DDS_header ; + +// the following constants were copied directly off the MSDN website + +// The dwFlags member of the original DDSURFACEDESC2 structure +// can be set to one or more of the following values. +#define DDSD_CAPS 0x00000001 +#define DDSD_HEIGHT 0x00000002 +#define DDSD_WIDTH 0x00000004 +#define DDSD_PITCH 0x00000008 +#define DDSD_PIXELFORMAT 0x00001000 +#define DDSD_MIPMAPCOUNT 0x00020000 +#define DDSD_LINEARSIZE 0x00080000 +#define DDSD_DEPTH 0x00800000 + +// DirectDraw Pixel Format +#define DDPF_ALPHAPIXELS 0x00000001 +#define DDPF_FOURCC 0x00000004 +#define DDPF_RGB 0x00000040 + +// The dwCaps1 member of the DDSCAPS2 structure can be +// set to one or more of the following values. +#define DDSCAPS_COMPLEX 0x00000008 +#define DDSCAPS_TEXTURE 0x00001000 +#define DDSCAPS_MIPMAP 0x00400000 + +// The dwCaps2 member of the DDSCAPS2 structure can be +// set to one or more of the following values. +#define DDSCAPS2_CUBEMAP 0x00000200 +#define DDSCAPS2_CUBEMAP_POSITIVEX 0x00000400 +#define DDSCAPS2_CUBEMAP_NEGATIVEX 0x00000800 +#define DDSCAPS2_CUBEMAP_POSITIVEY 0x00001000 +#define DDSCAPS2_CUBEMAP_NEGATIVEY 0x00002000 +#define DDSCAPS2_CUBEMAP_POSITIVEZ 0x00004000 +#define DDSCAPS2_CUBEMAP_NEGATIVEZ 0x00008000 +#define DDSCAPS2_VOLUME 0x00200000 + +static int dds_test(stbi *s) +{ + // check the magic number + if (get8(s) != 'D') return 0; + if (get8(s) != 'D') return 0; + if (get8(s) != 'S') return 0; + if (get8(s) != ' ') return 0; + // check header size + if (get32le(s) != 124) return 0; + return 1; +} +#ifndef STBI_NO_STDIO +int stbi_dds_test_file (FILE *f) +{ + stbi s; + int r,n = ftell(f); + start_file(&s,f); + r = dds_test(&s); + fseek(f,n,SEEK_SET); + return r; +} +#endif + +int stbi_dds_test_memory (stbi_uc const *buffer, int len) +{ + stbi s; + start_mem(&s,buffer, len); + return dds_test(&s); +} + +// helper functions +int stbi_convert_bit_range( int c, int from_bits, int to_bits ) +{ + int b = (1 << (from_bits - 1)) + c * ((1 << to_bits) - 1); + return (b + (b >> from_bits)) >> from_bits; +} +void stbi_rgb_888_from_565( unsigned int c, int *r, int *g, int *b ) +{ + *r = stbi_convert_bit_range( (c >> 11) & 31, 5, 8 ); + *g = stbi_convert_bit_range( (c >> 05) & 63, 6, 8 ); + *b = stbi_convert_bit_range( (c >> 00) & 31, 5, 8 ); +} +void stbi_decode_DXT1_block( + unsigned char uncompressed[16*4], + unsigned char compressed[8] ) +{ + int next_bit = 4*8; + int i, r, g, b; + int c0, c1; + unsigned char decode_colors[4*4]; + // find the 2 primary colors + c0 = compressed[0] + (compressed[1] << 8); + c1 = compressed[2] + (compressed[3] << 8); + stbi_rgb_888_from_565( c0, &r, &g, &b ); + decode_colors[0] = r; + decode_colors[1] = g; + decode_colors[2] = b; + decode_colors[3] = 255; + stbi_rgb_888_from_565( c1, &r, &g, &b ); + decode_colors[4] = r; + decode_colors[5] = g; + decode_colors[6] = b; + decode_colors[7] = 255; + if( c0 > c1 ) + { + // no alpha, 2 interpolated colors + decode_colors[8] = (2*decode_colors[0] + decode_colors[4]) / 3; + decode_colors[9] = (2*decode_colors[1] + decode_colors[5]) / 3; + decode_colors[10] = (2*decode_colors[2] + decode_colors[6]) / 3; + decode_colors[11] = 255; + decode_colors[12] = (decode_colors[0] + 2*decode_colors[4]) / 3; + decode_colors[13] = (decode_colors[1] + 2*decode_colors[5]) / 3; + decode_colors[14] = (decode_colors[2] + 2*decode_colors[6]) / 3; + decode_colors[15] = 255; + } else + { + // 1 interpolated color, alpha + decode_colors[8] = (decode_colors[0] + decode_colors[4]) / 2; + decode_colors[9] = (decode_colors[1] + decode_colors[5]) / 2; + decode_colors[10] = (decode_colors[2] + decode_colors[6]) / 2; + decode_colors[11] = 255; + decode_colors[12] = 0; + decode_colors[13] = 0; + decode_colors[14] = 0; + decode_colors[15] = 0; + } + // decode the block + for( i = 0; i < 16*4; i += 4 ) + { + int idx = ((compressed[next_bit>>3] >> (next_bit & 7)) & 3) * 4; + next_bit += 2; + uncompressed[i+0] = decode_colors[idx+0]; + uncompressed[i+1] = decode_colors[idx+1]; + uncompressed[i+2] = decode_colors[idx+2]; + uncompressed[i+3] = decode_colors[idx+3]; + } + // done +} +void stbi_decode_DXT23_alpha_block( + unsigned char uncompressed[16*4], + unsigned char compressed[8] ) +{ + int i, next_bit = 0; + // each alpha value gets 4 bits + for( i = 3; i < 16*4; i += 4 ) + { + uncompressed[i] = stbi_convert_bit_range( + (compressed[next_bit>>3] >> (next_bit&7)) & 15, + 4, 8 ); + next_bit += 4; + } +} +void stbi_decode_DXT45_alpha_block( + unsigned char uncompressed[16*4], + unsigned char compressed[8] ) +{ + int i, next_bit = 8*2; + unsigned char decode_alpha[8]; + // each alpha value gets 3 bits, and the 1st 2 bytes are the range + decode_alpha[0] = compressed[0]; + decode_alpha[1] = compressed[1]; + if( decode_alpha[0] > decode_alpha[1] ) + { + // 6 step intermediate + decode_alpha[2] = (6*decode_alpha[0] + 1*decode_alpha[1]) / 7; + decode_alpha[3] = (5*decode_alpha[0] + 2*decode_alpha[1]) / 7; + decode_alpha[4] = (4*decode_alpha[0] + 3*decode_alpha[1]) / 7; + decode_alpha[5] = (3*decode_alpha[0] + 4*decode_alpha[1]) / 7; + decode_alpha[6] = (2*decode_alpha[0] + 5*decode_alpha[1]) / 7; + decode_alpha[7] = (1*decode_alpha[0] + 6*decode_alpha[1]) / 7; + } else + { + // 4 step intermediate, pluss full and none + decode_alpha[2] = (4*decode_alpha[0] + 1*decode_alpha[1]) / 5; + decode_alpha[3] = (3*decode_alpha[0] + 2*decode_alpha[1]) / 5; + decode_alpha[4] = (2*decode_alpha[0] + 3*decode_alpha[1]) / 5; + decode_alpha[5] = (1*decode_alpha[0] + 4*decode_alpha[1]) / 5; + decode_alpha[6] = 0; + decode_alpha[7] = 255; + } + for( i = 3; i < 16*4; i += 4 ) + { + int idx = 0, bit; + bit = (compressed[next_bit>>3] >> (next_bit&7)) & 1; + idx += bit << 0; + ++next_bit; + bit = (compressed[next_bit>>3] >> (next_bit&7)) & 1; + idx += bit << 1; + ++next_bit; + bit = (compressed[next_bit>>3] >> (next_bit&7)) & 1; + idx += bit << 2; + ++next_bit; + uncompressed[i] = decode_alpha[idx & 7]; + } + // done +} +void stbi_decode_DXT_color_block( + unsigned char uncompressed[16*4], + unsigned char compressed[8] ) +{ + int next_bit = 4*8; + int i, r, g, b; + int c0, c1; + unsigned char decode_colors[4*3]; + // find the 2 primary colors + c0 = compressed[0] + (compressed[1] << 8); + c1 = compressed[2] + (compressed[3] << 8); + stbi_rgb_888_from_565( c0, &r, &g, &b ); + decode_colors[0] = r; + decode_colors[1] = g; + decode_colors[2] = b; + stbi_rgb_888_from_565( c1, &r, &g, &b ); + decode_colors[3] = r; + decode_colors[4] = g; + decode_colors[5] = b; + // Like DXT1, but no choicees: + // no alpha, 2 interpolated colors + decode_colors[6] = (2*decode_colors[0] + decode_colors[3]) / 3; + decode_colors[7] = (2*decode_colors[1] + decode_colors[4]) / 3; + decode_colors[8] = (2*decode_colors[2] + decode_colors[5]) / 3; + decode_colors[9] = (decode_colors[0] + 2*decode_colors[3]) / 3; + decode_colors[10] = (decode_colors[1] + 2*decode_colors[4]) / 3; + decode_colors[11] = (decode_colors[2] + 2*decode_colors[5]) / 3; + // decode the block + for( i = 0; i < 16*4; i += 4 ) + { + int idx = ((compressed[next_bit>>3] >> (next_bit & 7)) & 3) * 3; + next_bit += 2; + uncompressed[i+0] = decode_colors[idx+0]; + uncompressed[i+1] = decode_colors[idx+1]; + uncompressed[i+2] = decode_colors[idx+2]; + } + // done +} +static stbi_uc *dds_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + // all variables go up front + stbi_uc *dds_data = NULL; + stbi_uc block[16*4]; + stbi_uc compressed[8]; + int flags, DXT_family; + int has_alpha, has_mipmap; + int is_compressed, cubemap_faces; + int block_pitch, num_blocks; + DDS_header header; + int i, sz, cf; + // load the header + if( sizeof( DDS_header ) != 128 ) + { + return NULL; + } + getn( s, (stbi_uc*)(&header), 128 ); + // and do some checking + if( header.dwMagic != (('D' << 0) | ('D' << 8) | ('S' << 16) | (' ' << 24)) ) return NULL; + if( header.dwSize != 124 ) return NULL; + flags = DDSD_CAPS | DDSD_HEIGHT | DDSD_WIDTH | DDSD_PIXELFORMAT; + if( (header.dwFlags & flags) != flags ) return NULL; + /* According to the MSDN spec, the dwFlags should contain + DDSD_LINEARSIZE if it's compressed, or DDSD_PITCH if + uncompressed. Some DDS writers do not conform to the + spec, so I need to make my reader more tolerant */ + if( header.sPixelFormat.dwSize != 32 ) return NULL; + flags = DDPF_FOURCC | DDPF_RGB; + if( (header.sPixelFormat.dwFlags & flags) == 0 ) return NULL; + if( (header.sCaps.dwCaps1 & DDSCAPS_TEXTURE) == 0 ) return NULL; + // get the image data + s->img_x = header.dwWidth; + s->img_y = header.dwHeight; + s->img_n = 4; + is_compressed = (header.sPixelFormat.dwFlags & DDPF_FOURCC) / DDPF_FOURCC; + has_alpha = (header.sPixelFormat.dwFlags & DDPF_ALPHAPIXELS) / DDPF_ALPHAPIXELS; + has_mipmap = (header.sCaps.dwCaps1 & DDSCAPS_MIPMAP) && (header.dwMipMapCount > 1); + cubemap_faces = (header.sCaps.dwCaps2 & DDSCAPS2_CUBEMAP) / DDSCAPS2_CUBEMAP; + /* I need cubemaps to have square faces */ + cubemap_faces &= (s->img_x == s->img_y); + cubemap_faces *= 5; + cubemap_faces += 1; + block_pitch = (s->img_x+3) >> 2; + num_blocks = block_pitch * ((s->img_y+3) >> 2); + /* let the user know what's going on */ + *x = s->img_x; + *y = s->img_y; + *comp = s->img_n; + /* is this uncompressed? */ + if( is_compressed ) + { + /* compressed */ + // note: header.sPixelFormat.dwFourCC is something like (('D'<<0)|('X'<<8)|('T'<<16)|('1'<<24)) + DXT_family = 1 + (header.sPixelFormat.dwFourCC >> 24) - '1'; + if( (DXT_family < 1) || (DXT_family > 5) ) return NULL; + /* check the expected size...oops, nevermind... + those non-compliant writers leave + dwPitchOrLinearSize == 0 */ + // passed all the tests, get the RAM for decoding + sz = (s->img_x)*(s->img_y)*4*cubemap_faces; + dds_data = (unsigned char*)malloc( sz ); + /* do this once for each face */ + for( cf = 0; cf < cubemap_faces; ++ cf ) + { + // now read and decode all the blocks + for( i = 0; i < num_blocks; ++i ) + { + // where are we? + int bx, by, bw=4, bh=4; + int ref_x = 4 * (i % block_pitch); + int ref_y = 4 * (i / block_pitch); + // get the next block's worth of compressed data, and decompress it + if( DXT_family == 1 ) + { + // DXT1 + getn( s, compressed, 8 ); + stbi_decode_DXT1_block( block, compressed ); + } else if( DXT_family < 4 ) + { + // DXT2/3 + getn( s, compressed, 8 ); + stbi_decode_DXT23_alpha_block ( block, compressed ); + getn( s, compressed, 8 ); + stbi_decode_DXT_color_block ( block, compressed ); + } else + { + // DXT4/5 + getn( s, compressed, 8 ); + stbi_decode_DXT45_alpha_block ( block, compressed ); + getn( s, compressed, 8 ); + stbi_decode_DXT_color_block ( block, compressed ); + } + // is this a partial block? + if( ref_x + 4 > s->img_x ) + { + bw = s->img_x - ref_x; + } + if( ref_y + 4 > s->img_y ) + { + bh = s->img_y - ref_y; + } + // now drop our decompressed data into the buffer + for( by = 0; by < bh; ++by ) + { + int idx = 4*((ref_y+by+cf*s->img_x)*s->img_x + ref_x); + for( bx = 0; bx < bw*4; ++bx ) + { + + dds_data[idx+bx] = block[by*16+bx]; + } + } + } + /* done reading and decoding the main image... + skip MIPmaps if present */ + if( has_mipmap ) + { + int block_size = 16; + if( DXT_family == 1 ) + { + block_size = 8; + } + for( i = 1; i < header.dwMipMapCount; ++i ) + { + int mx = s->img_x >> (i + 2); + int my = s->img_y >> (i + 2); + if( mx < 1 ) + { + mx = 1; + } + if( my < 1 ) + { + my = 1; + } + skip( s, mx*my*block_size ); + } + } + }/* per cubemap face */ + } else + { + /* uncompressed */ + DXT_family = 0; + s->img_n = 3; + if( has_alpha ) + { + s->img_n = 4; + } + *comp = s->img_n; + sz = s->img_x*s->img_y*s->img_n*cubemap_faces; + dds_data = (unsigned char*)malloc( sz ); + /* do this once for each face */ + for( cf = 0; cf < cubemap_faces; ++ cf ) + { + /* read the main image for this face */ + getn( s, &dds_data[cf*s->img_x*s->img_y*s->img_n], s->img_x*s->img_y*s->img_n ); + /* done reading and decoding the main image... + skip MIPmaps if present */ + if( has_mipmap ) + { + for( i = 1; i < header.dwMipMapCount; ++i ) + { + int mx = s->img_x >> i; + int my = s->img_y >> i; + if( mx < 1 ) + { + mx = 1; + } + if( my < 1 ) + { + my = 1; + } + skip( s, mx*my*s->img_n ); + } + } + } + /* data was BGR, I need it RGB */ + for( i = 0; i < sz; i += s->img_n ) + { + unsigned char temp = dds_data[i]; + dds_data[i] = dds_data[i+2]; + dds_data[i+2] = temp; + } + } + /* finished decompressing into RGBA, + adjust the y size if we have a cubemap + note: sz is already up to date */ + s->img_y *= cubemap_faces; + *y = s->img_y; + // did the user want something else, or + // see if all the alpha values are 255 (i.e. no transparency) + has_alpha = 0; + if( s->img_n == 4) + { + for( i = 3; (i < sz) && (has_alpha == 0); i += 4 ) + { + has_alpha |= (dds_data[i] < 255); + } + } + if( (req_comp <= 4) && (req_comp >= 1) ) + { + // user has some requirements, meet them + if( req_comp != s->img_n ) + { + dds_data = convert_format( dds_data, s->img_n, req_comp, s->img_x, s->img_y ); + *comp = s->img_n; + } + } else + { + // user had no requirements, only drop to RGB is no alpha + if( (has_alpha == 0) && (s->img_n == 4) ) + { + dds_data = convert_format( dds_data, 4, 3, s->img_x, s->img_y ); + *comp = 3; + } + } + // OK, done + return dds_data; +} + +#ifndef STBI_NO_STDIO +stbi_uc *stbi_dds_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_file(&s,f); + return dds_load(&s,x,y,comp,req_comp); +} + +stbi_uc *stbi_dds_load (char *filename, int *x, int *y, int *comp, int req_comp) +{ + stbi_uc *data; + FILE *f = fopen(filename, "rb"); + if (!f) return NULL; + data = stbi_dds_load_from_file(f,x,y,comp,req_comp); + fclose(f); + return data; +} +#endif + +stbi_uc *stbi_dds_load_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_mem(&s,buffer, len); + return dds_load(&s,x,y,comp,req_comp); +} diff --git a/lib/soil/soil/test_SOIL.cpp b/lib/soil/soil/test_SOIL.cpp new file mode 100644 index 0000000..9c4c091 --- /dev/null +++ b/lib/soil/soil/test_SOIL.cpp @@ -0,0 +1,379 @@ +#include +#include + +#include +#include +#include +#include + +#include "SOIL.h" + +LRESULT CALLBACK WindowProc(HWND, UINT, WPARAM, LPARAM); +void EnableOpenGL(HWND hwnd, HDC*, HGLRC*); +void DisableOpenGL(HWND, HDC, HGLRC); + +int WINAPI WinMain(HINSTANCE hInstance, + HINSTANCE hPrevInstance, + LPSTR lpCmdLine, + int nCmdShow) +{ + WNDCLASSEX wcex; + HWND hwnd; + HDC hDC; + HGLRC hRC; + MSG msg; + BOOL bQuit = FALSE; + float theta = 0.0f; + + // register window class + wcex.cbSize = sizeof(WNDCLASSEX); + wcex.style = CS_OWNDC; + wcex.lpfnWndProc = WindowProc; + wcex.cbClsExtra = 0; + wcex.cbWndExtra = 0; + wcex.hInstance = hInstance; + wcex.hIcon = LoadIcon(NULL, IDI_APPLICATION); + wcex.hCursor = LoadCursor(NULL, IDC_ARROW); + wcex.hbrBackground = (HBRUSH)GetStockObject(BLACK_BRUSH); + wcex.lpszMenuName = NULL; + wcex.lpszClassName = "GLSample"; + wcex.hIconSm = LoadIcon(NULL, IDI_APPLICATION); + + + if (!RegisterClassEx(&wcex)) + return 0; + + // create main window + hwnd = CreateWindowEx(0, + "GLSample", + "SOIL Sample", + WS_OVERLAPPEDWINDOW, + CW_USEDEFAULT, + CW_USEDEFAULT, + 512, + 512, + NULL, + NULL, + hInstance, + NULL); + + ShowWindow(hwnd, nCmdShow); + + // check my error handling + /* + SOIL_load_OGL_texture( "img_test.png", SOIL_LOAD_AUTO, SOIL_CREATE_NEW_ID, 0 ); + std::cout << "'" << SOIL_last_result() << "'" << std::endl; + */ + + + // enable OpenGL for the window + EnableOpenGL(hwnd, &hDC, &hRC); + + glEnable( GL_BLEND ); + //glDisable( GL_BLEND ); + // straight alpha + glBlendFunc( GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA ); + // premultiplied alpha (remember to do the same in glColor!!) + //glBlendFunc( GL_ONE, GL_ONE_MINUS_SRC_ALPHA ); + + // do I want alpha thresholding? + glEnable( GL_ALPHA_TEST ); + glAlphaFunc( GL_GREATER, 0.5f ); + + // log what the use is asking us to load + std::string load_me = lpCmdLine; + if( load_me.length() > 2 ) + { + //load_me = load_me.substr( 1, load_me.length() - 2 ); + load_me = load_me.substr( 0, load_me.length() - 0 ); + } else + { + //load_me = "img_test_uncompressed.dds"; + //load_me = "img_test_indexed.tga"; + //load_me = "img_test.dds"; + load_me = "img_test.png"; + //load_me = "odd_size.jpg"; + //load_me = "img_cheryl.jpg"; + //load_me = "oak_odd.png"; + //load_me = "field_128_cube.dds"; + //load_me = "field_128_cube_nomip.dds"; + //load_me = "field_128_cube_uc.dds"; + //load_me = "field_128_cube_uc_nomip.dds"; + //load_me = "Goblin.dds"; + //load_me = "parquet.dds"; + //load_me = "stpeters_probe.hdr"; + //load_me = "VeraMoBI_sdf.png"; + + // for testing the texture rectangle code + //load_me = "test_rect.png"; + } + std::cout << "'" << load_me << "'" << std::endl; + + // 1st try to load it as a single-image-cubemap + // (note, need DDS ordered faces: "EWUDNS") + GLuint tex_ID; + int time_me; + + std::cout << "Attempting to load as a cubemap" << std::endl; + time_me = clock(); + tex_ID = SOIL_load_OGL_single_cubemap( + load_me.c_str(), + SOIL_DDS_CUBEMAP_FACE_ORDER, + SOIL_LOAD_AUTO, + SOIL_CREATE_NEW_ID, + SOIL_FLAG_POWER_OF_TWO + | SOIL_FLAG_MIPMAPS + //| SOIL_FLAG_COMPRESS_TO_DXT + //| SOIL_FLAG_TEXTURE_REPEATS + //| SOIL_FLAG_INVERT_Y + | SOIL_FLAG_DDS_LOAD_DIRECT + ); + time_me = clock() - time_me; + std::cout << "the load time was " << 0.001f * time_me << " seconds (warning: low resolution timer)" << std::endl; + if( tex_ID > 0 ) + { + glEnable( GL_TEXTURE_CUBE_MAP ); + glEnable( GL_TEXTURE_GEN_S ); + glEnable( GL_TEXTURE_GEN_T ); + glEnable( GL_TEXTURE_GEN_R ); + glTexGeni( GL_S, GL_TEXTURE_GEN_MODE, GL_REFLECTION_MAP ); + glTexGeni( GL_T, GL_TEXTURE_GEN_MODE, GL_REFLECTION_MAP ); + glTexGeni( GL_R, GL_TEXTURE_GEN_MODE, GL_REFLECTION_MAP ); + glBindTexture( GL_TEXTURE_CUBE_MAP, tex_ID ); + // report + std::cout << "the loaded single cube map ID was " << tex_ID << std::endl; + //std::cout << "the load time was " << 0.001f * time_me << " seconds (warning: low resolution timer)" << std::endl; + } else + { + std::cout << "Attempting to load as a HDR texture" << std::endl; + time_me = clock(); + tex_ID = SOIL_load_OGL_HDR_texture( + load_me.c_str(), + //SOIL_HDR_RGBE, + //SOIL_HDR_RGBdivA, + SOIL_HDR_RGBdivA2, + 0, + SOIL_CREATE_NEW_ID, + SOIL_FLAG_POWER_OF_TWO + | SOIL_FLAG_MIPMAPS + //| SOIL_FLAG_COMPRESS_TO_DXT + ); + time_me = clock() - time_me; + std::cout << "the load time was " << 0.001f * time_me << " seconds (warning: low resolution timer)" << std::endl; + + // did I fail? + if( tex_ID < 1 ) + { + // loading of the single-image-cubemap failed, try it as a simple texture + std::cout << "Attempting to load as a simple 2D texture" << std::endl; + // load the texture, if specified + time_me = clock(); + tex_ID = SOIL_load_OGL_texture( + load_me.c_str(), + SOIL_LOAD_AUTO, + SOIL_CREATE_NEW_ID, + SOIL_FLAG_POWER_OF_TWO + | SOIL_FLAG_MIPMAPS + //| SOIL_FLAG_MULTIPLY_ALPHA + //| SOIL_FLAG_COMPRESS_TO_DXT + | SOIL_FLAG_DDS_LOAD_DIRECT + //| SOIL_FLAG_NTSC_SAFE_RGB + //| SOIL_FLAG_CoCg_Y + //| SOIL_FLAG_TEXTURE_RECTANGLE + ); + time_me = clock() - time_me; + std::cout << "the load time was " << 0.001f * time_me << " seconds (warning: low resolution timer)" << std::endl; + } + + if( tex_ID > 0 ) + { + // enable texturing + glEnable( GL_TEXTURE_2D ); + //glEnable( 0x84F5 );// enables texture rectangle + // bind an OpenGL texture ID + glBindTexture( GL_TEXTURE_2D, tex_ID ); + // report + std::cout << "the loaded texture ID was " << tex_ID << std::endl; + //std::cout << "the load time was " << 0.001f * time_me << " seconds (warning: low resolution timer)" << std::endl; + } else + { + // loading of the texture failed...why? + glDisable( GL_TEXTURE_2D ); + std::cout << "Texture loading failed: '" << SOIL_last_result() << "'" << std::endl; + } + } + + // program main loop + const float ref_mag = 0.1f; + while (!bQuit) + { + // check for messages + if (PeekMessage(&msg, NULL, 0, 0, PM_REMOVE)) + { + // handle or dispatch messages + if (msg.message == WM_QUIT) + { + bQuit = TRUE; + } + else + { + TranslateMessage(&msg); + DispatchMessage(&msg); + } + } + else + { + // OpenGL animation code goes here + theta = clock() * 0.1; + + float tex_u_max = 1.0f;//0.2f; + float tex_v_max = 1.0f;//0.2f; + + glClearColor(0.0f, 0.0f, 0.0f, 0.0f); + glClear(GL_COLOR_BUFFER_BIT); + + glPushMatrix(); + glScalef( 0.8f, 0.8f, 0.8f ); + //glRotatef(-0.314159f*theta, 0.0f, 0.0f, 1.0f); + glColor4f( 1.0f, 1.0f, 1.0f, 1.0f ); + glNormal3f( 0.0f, 0.0f, 1.0f ); + glBegin(GL_QUADS); + glNormal3f( -ref_mag, -ref_mag, 1.0f ); + glTexCoord2f( 0.0f, tex_v_max ); + glVertex3f( -1.0f, -1.0f, -0.1f ); + + glNormal3f( ref_mag, -ref_mag, 1.0f ); + glTexCoord2f( tex_u_max, tex_v_max ); + glVertex3f( 1.0f, -1.0f, -0.1f ); + + glNormal3f( ref_mag, ref_mag, 1.0f ); + glTexCoord2f( tex_u_max, 0.0f ); + glVertex3f( 1.0f, 1.0f, -0.1f ); + + glNormal3f( -ref_mag, ref_mag, 1.0f ); + glTexCoord2f( 0.0f, 0.0f ); + glVertex3f( -1.0f, 1.0f, -0.1f ); + glEnd(); + glPopMatrix(); + + tex_u_max = 1.0f; + tex_v_max = 1.0f; + glPushMatrix(); + glScalef( 0.8f, 0.8f, 0.8f ); + glRotatef(theta, 0.0f, 0.0f, 1.0f); + glColor4f( 1.0f, 1.0f, 1.0f, 1.0f ); + glNormal3f( 0.0f, 0.0f, 1.0f ); + glBegin(GL_QUADS); + glTexCoord2f( 0.0f, tex_v_max ); glVertex3f( 0.0f, 0.0f, 0.1f ); + glTexCoord2f( tex_u_max, tex_v_max ); glVertex3f( 1.0f, 0.0f, 0.1f ); + glTexCoord2f( tex_u_max, 0.0f ); glVertex3f( 1.0f, 1.0f, 0.1f ); + glTexCoord2f( 0.0f, 0.0f ); glVertex3f( 0.0f, 1.0f, 0.1f ); + glEnd(); + glPopMatrix(); + + { + /* check for errors */ + GLenum err_code = glGetError(); + while( GL_NO_ERROR != err_code ) + { + printf( "OpenGL Error @ %s: %i", "drawing loop", err_code ); + err_code = glGetError(); + } + } + + SwapBuffers(hDC); + + Sleep (1); + } + } + + // and show off the screenshot capability + /* + load_me += "-screenshot.tga"; + SOIL_save_screenshot( load_me.c_str(), SOIL_SAVE_TYPE_TGA, 0, 0, 512, 512 ); + //*/ + //* + load_me += "-screenshot.bmp"; + SOIL_save_screenshot( load_me.c_str(), SOIL_SAVE_TYPE_BMP, 0, 0, 512, 512 ); + //*/ + /* + load_me += "-screenshot.dds"; + SOIL_save_screenshot( load_me.c_str(), SOIL_SAVE_TYPE_DDS, 0, 0, 512, 512 ); + //*/ + + // shutdown OpenGL + DisableOpenGL(hwnd, hDC, hRC); + + // destroy the window explicitly + DestroyWindow(hwnd); + + return msg.wParam; +} + +LRESULT CALLBACK WindowProc(HWND hwnd, UINT uMsg, WPARAM wParam, LPARAM lParam) +{ + switch (uMsg) + { + case WM_CLOSE: + PostQuitMessage(0); + break; + + case WM_DESTROY: + return 0; + + case WM_KEYDOWN: + { + switch (wParam) + { + case VK_ESCAPE: + PostQuitMessage(0); + break; + } + } + break; + + default: + return DefWindowProc(hwnd, uMsg, wParam, lParam); + } + + return 0; +} + +void EnableOpenGL(HWND hwnd, HDC* hDC, HGLRC* hRC) +{ + PIXELFORMATDESCRIPTOR pfd; + + int iFormat; + + /* get the device context (DC) */ + *hDC = GetDC(hwnd); + + /* set the pixel format for the DC */ + ZeroMemory(&pfd, sizeof(pfd)); + + pfd.nSize = sizeof(pfd); + pfd.nVersion = 1; + pfd.dwFlags = PFD_DRAW_TO_WINDOW | + PFD_SUPPORT_OPENGL | PFD_DOUBLEBUFFER; + pfd.iPixelType = PFD_TYPE_RGBA; + pfd.cColorBits = 24; + pfd.cDepthBits = 16; + pfd.iLayerType = PFD_MAIN_PLANE; + + iFormat = ChoosePixelFormat(*hDC, &pfd); + + SetPixelFormat(*hDC, iFormat, &pfd); + + /* create and enable the render context (RC) */ + *hRC = wglCreateContext(*hDC); + + wglMakeCurrent(*hDC, *hRC); +} + +void DisableOpenGL (HWND hwnd, HDC hDC, HGLRC hRC) +{ + wglMakeCurrent(NULL, NULL); + wglDeleteContext(hRC); + ReleaseDC(hwnd, hDC); +} + -- 2.39.5