[patch] updated pic-fix for theora 1.0-beta2

libtheora-1.0_beta2-pic-fix.patch (text/plain), 4.47 KB, created by PaX Team on 2007-12-17 09:46:44 UTC

diff -urp libtheora-1.0beta2-orig/lib/enc/x86_32/dct_decode_mmx.c libtheora-1.0beta2/lib/enc/x86_32/dct_decode_mmx.c
--- libtheora-1.0beta2-orig/lib/enc/x86_32/dct_decode_mmx.c	2007-10-04 20:37:01.000000000 +0200
+++ libtheora-1.0beta2/lib/enc/x86_32/dct_decode_mmx.c	2007-12-17 10:32:44.000000000 +0100
@@ -57,9 +57,9 @@ static void FilterHoriz__mmx(unsigned ch
     "psubw %%mm3,%%mm1\n"       /* mm1 = pix[0]-pix[3] mm1 - mm3 */     \
     "movq %%mm0,%%mm7\n"        /* mm7 = pix[2]*/                       \
     "psubw %%mm5,%%mm0\n"       /* mm0 = pix[2]-pix[1] mm0 - mm5*/      \
-    "PMULLW "MANGLE(V3)",%%mm0\n" /* *3 */                              \
+    "PMULLW %3,%%mm0\n" /* *3 */                              \
     "paddw %%mm0,%%mm1\n"         /* mm1 has f[0] ... f[4]*/            \
-    "paddw "MANGLE(V804)",%%mm1\n"/* add 4 */ /* add 256 after shift */ \
+    "paddw %4,%%mm1\n"/* add 4 */ /* add 256 after shift */ \
     "psraw $3,%%mm1\n"          /* >>3 */                               \
     " pextrw $0,%%mm1,%%esi\n"  /* In MM1 we have 4 f coefs (16bits) */ \
     " pextrw $1,%%mm1,%%edi\n"  /* now perform MM4 = *(_bv+ f) */       \
@@ -75,20 +75,19 @@ static void FilterHoriz__mmx(unsigned ch
     " packuswb %%mm0,%%mm5\n"   /* mm5 = x x x x newpix1 */             \
     " packuswb %%mm0,%%mm7\n"   /* mm7 = x x x x newpix2 */             \
     " punpcklbw %%mm7,%%mm5\n"  /* 2 1 2 1 2 1 2 1 */                   \
-    " movd %%mm5,%%eax\n"       /* eax = newpix21 */                    \
-    " movw %%ax,1(%0)\n"                                                \
+    " movd %%mm5,%%edi\n"       /* eax = newpix21 */                    \
+    " movw %%di,1(%0)\n"                                                \
     " psrlq $32,%%mm5\n"        /* why is so big stall here ? */        \
-    " shrl $16,%%eax\n"                                                 \
-    " lea 1(%0,%1,2),%%edi\n"                                           \
-    " movw %%ax,1(%0,%1,1)\n"                                           \
-    " movd %%mm5,%%eax\n"       /* eax = newpix21 high part */          \
+    " shrl $16,%%edi\n"                                                 \
+    " movw %%di,1(%0,%1,1)\n"                                           \
+    " movd %%mm5,%%edi\n"       /* eax = newpix21 high part */          \
     " lea (%1,%1,2),%%esi\n"                                            \
-    " movw %%ax,(%%edi)\n"                                              \
-    " shrl $16,%%eax\n"                                                 \
-    " movw %%ax,1(%0,%%esi)\n"                                          \
+    " movw %%di,1(%0,%1,2)\n"                                              \
+    " shrl $16,%%edi\n"                                                 \
+    " movw %%di,1(%0,%%esi)\n"                                          \
     :                                                                   \
-    : "r" (PixelPtr), "r" (LineLength), "r" (BoundingValuePtr-256)      \
-    : "esi", "edi" , "memory", "eax"                                    \
+    : "r" (PixelPtr), "r" (LineLength), "r" (BoundingValuePtr-256), "m" (V3), "m" (V804) \
+    : "esi", "edi" , "memory"                                           \
     );
 
     OC_LOOP_H_4x4
@@ -126,12 +125,12 @@ static void FilterVert__mmx(unsigned cha
     "psubw %%mm5,%%mm3\n"
     "psubw %%mm4,%%mm2\n"
                 /* mm3:mm2 = (pix[ystride*2]-pix[ystride]); */
-    "PMULLW "MANGLE(V3)",%%mm3\n"    /* *3 */
-    "PMULLW "MANGLE(V3)",%%mm2\n"    /* *3 */
+    "PMULLW %3,%%mm3\n"    /* *3 */
+    "PMULLW %3,%%mm2\n"    /* *3 */
     "paddw %%mm7,%%mm3\n"            /* highpart */
     "paddw %%mm6,%%mm2\n"            /* lowpart of pix[0]-pix[ystride*3]+3*(pix[ystride*2]-pix[ystride]);  */
-    "paddw "MANGLE(V804)",%%mm3\n"   /* add 4 */ /* add 256 after shift */
-    "paddw "MANGLE(V804)",%%mm2\n"   /* add 4 */ /* add 256 after shift */
+    "paddw %4,%%mm3\n"   /* add 4 */ /* add 256 after shift */
+    "paddw %4,%%mm2\n"   /* add 4 */ /* add 256 after shift */
     "psraw $3,%%mm3\n"               /* >>3 f coefs high */
     "psraw $3,%%mm2\n"               /* >>3 f coefs low */
 
@@ -168,7 +167,7 @@ static void FilterVert__mmx(unsigned cha
     "movq %%mm4,(%0,%1)\n"      /* pix[ystride]= */
     "emms\n"
     :
-    : "r" (PixelPtr-2*LineLength), "r" (LineLength), "r" (BoundingValuePtr-256)
+    : "r" (PixelPtr-2*LineLength), "r" (LineLength), "r" (BoundingValuePtr-256), "m" (V3), "m" (V804)
     : "esi", "edi" , "memory"
     );
 }