From ef9e75d4ad2c708c1f5baff6518c811b9cb5e58d Mon Sep 17 00:00:00 2001 From: Danny Rawlins Date: Sun, 3 Aug 2008 01:10:16 +1000 Subject: [PATCH] libfame: various fixes --- libfame/.md5sum | 6 + libfame/Pkgfile | 16 +- .../libfame-0.9.1-fame_motion_pmvfast.c.patch | 11 + libfame/libfame-0.9.1-gcc-version-check.patch | 19 + libfame/libfame-0.9.1-missing-includes.patch | 45 +++ libfame/libfame-0.9.1-mmx-pic.patch | 366 ++++++++++++++++++ libfame/libfame-0.9.1-mmx.patch | 133 +++++++ libfame/libfame-0.9.1-ptr-null-check.patch | 19 + 8 files changed, 613 insertions(+), 2 deletions(-) create mode 100644 libfame/libfame-0.9.1-fame_motion_pmvfast.c.patch create mode 100644 libfame/libfame-0.9.1-gcc-version-check.patch create mode 100644 libfame/libfame-0.9.1-missing-includes.patch create mode 100644 libfame/libfame-0.9.1-mmx-pic.patch create mode 100644 libfame/libfame-0.9.1-mmx.patch create mode 100644 libfame/libfame-0.9.1-ptr-null-check.patch diff --git a/libfame/.md5sum b/libfame/.md5sum index 37248d29e..c2309c493 100644 --- a/libfame/.md5sum +++ b/libfame/.md5sum @@ -1,2 +1,8 @@ +fb98aaf4ca928bc10f6d1fc92983b6ec libfame-0.9.1-fame_motion_pmvfast.c.patch +9022ea48cc238fb071b9b9c29a85fc7a libfame-0.9.1-gcc-version-check.patch af9a3ce19e852874fd602c9c451d3160 libfame-0.9.1-m4.patch +f01512be0252259fc94ec638f9837580 libfame-0.9.1-missing-includes.patch +84bf7e9820d228ef3eaa2873992200aa libfame-0.9.1-mmx-pic.patch +d7f06d0ab862883eff73dbb0ebf6a14c libfame-0.9.1-mmx.patch +ee57a8744b4c559a7b9f8d3d50b28baa libfame-0.9.1-ptr-null-check.patch 880085761e17a3b4fc41f4f6f198fd3b libfame-0.9.1.tar.gz diff --git a/libfame/Pkgfile b/libfame/Pkgfile index 22ff39b17..5dd98d945 100644 --- a/libfame/Pkgfile +++ b/libfame/Pkgfile @@ -6,13 +6,25 @@ name=libfame version=0.9.1 -release=1 +release=2 source=(http://dl.sourceforge.net/sourceforge/fame/libfame-$version.tar.gz - $name-$version-m4.patch) + $name-$version-m4.patch + $name-$version-gcc-version-check.patch + $name-$version-mmx.patch + $name-$version-mmx-pic.patch + $name-$version-fame_motion_pmvfast.c.patch + $name-$version-missing-includes.patch + $name-$version-ptr-null-check.patch) build() { cd libfame-$version patch -p 1 -i $SRC/$name-$version-m4.patch + patch -p 1 -i $SRC/$name-$version-gcc-version-check.patch + patch -p 1 -i $SRC/$name-$version-mmx.patch + patch -p 1 -i $SRC/$name-$version-mmx-pic.patch + patch -p 1 -i $SRC/$name-$version-fame_motion_pmvfast.c.patch + patch -p 1 -i $SRC/$name-$version-missing-includes.patch + patch -p 1 -i $SRC/$name-$version-ptr-null-check.patch ./configure \ --prefix=/usr \ diff --git a/libfame/libfame-0.9.1-fame_motion_pmvfast.c.patch b/libfame/libfame-0.9.1-fame_motion_pmvfast.c.patch new file mode 100644 index 000000000..90a4e441f --- /dev/null +++ b/libfame/libfame-0.9.1-fame_motion_pmvfast.c.patch @@ -0,0 +1,11 @@ +diff -pruN libfame-0.9.1.orig/src/fame_motion_pmvfast.c libfame-0.9.1/src/fame_motion_pmvfast.c +--- libfame-0.9.1.orig/src/fame_motion_pmvfast.c 2008-08-02 22:44:06.857141618 +1000 ++++ libfame-0.9.1/src/fame_motion_pmvfast.c 2008-08-02 23:00:26.842141204 +1000 +@@ -921,7 +921,6 @@ static fame_motion_coding_t pmvfast_esti + fame_motion_vector_t *plast; + fame_motion_vector_t *pvector_left, *pvector_topleft; + fame_motion_vector_t *pvector_top, *pvector_topright; +- unsigned int weight_left, weight_top, weight_topright; + + unsigned char *shape; + unsigned char *current; diff --git a/libfame/libfame-0.9.1-gcc-version-check.patch b/libfame/libfame-0.9.1-gcc-version-check.patch new file mode 100644 index 000000000..19b919632 --- /dev/null +++ b/libfame/libfame-0.9.1-gcc-version-check.patch @@ -0,0 +1,19 @@ +diff -pruN libfame-0.9.1.orig/configure libfame-0.9.1/configure +--- libfame-0.9.1.orig/configure 2008-08-02 22:04:18.375141589 +1000 ++++ libfame-0.9.1/configure 2008-08-02 22:05:49.940140531 +1000 +@@ -4607,12 +4607,9 @@ fi + if test x$ac_cv_c_compiler_gnu = xyes; then + CFLAGS="$CFLAGS -Wall -fexpensive-optimizations -funroll-loops -ffast-math" + +- gcc_major_version=`$CC --version | \ +- sed -e 's,[^0-9.],,g' -e 's/\([0-9]*\).\([0-9]*\).\([0-9]*\)/\1/'` +- gcc_minor_version=`$CC --version | \ +- sed -e 's,[^0-9.],,g' -e 's/\([0-9]*\).\([0-9]*\).\([0-9]*\)/\2/'` +- gcc_micro_version=`$CC --version | \ +- sed -e 's,[^0-9.],,g' -e 's/\([0-9]*\).\([0-9]*\).\([0-9]*\)/\3/'` ++ gcc_major_version=`${CC} -dumpversion | cut -f1 -d.` ++ gcc_minor_version=`${CC} -dumpversion | cut -f2 -d.` ++ gcc_micro_version=`${CC} -dumpversion | cut -f3 -d.` + + if test $gcc_major_version -eq 2; then + if test $gcc_minor_version -gt 94; then diff --git a/libfame/libfame-0.9.1-missing-includes.patch b/libfame/libfame-0.9.1-missing-includes.patch new file mode 100644 index 000000000..8011da1ba --- /dev/null +++ b/libfame/libfame-0.9.1-missing-includes.patch @@ -0,0 +1,45 @@ +diff -pruN libfame-0.9.1.orig/src/fame_malloc.c libfame-0.9.1/src/fame_malloc.c +--- libfame-0.9.1.orig/src/fame_malloc.c 2008-08-02 22:44:06.858140887 +1000 ++++ libfame-0.9.1/src/fame_malloc.c 2008-08-02 22:44:35.795140914 +1000 +@@ -21,6 +21,8 @@ + #define ALIGN 32 + + #include ++#include ++#include + + void* fame_malloc(size_t size) + { +diff -pruN libfame-0.9.1.orig/src/fame_motion_pmvfast.c libfame-0.9.1/src/fame_motion_pmvfast.c +--- libfame-0.9.1.orig/src/fame_motion_pmvfast.c 2008-08-02 22:44:06.857141618 +1000 ++++ libfame-0.9.1/src/fame_motion_pmvfast.c 2008-08-02 22:44:35.795140914 +1000 +@@ -19,6 +19,7 @@ + + #include + #include ++#include /* memcpy, memset */ + #include "fame.h" + #include "fame_malloc.h" + #include "fame_motion.h" +diff -pruN libfame-0.9.1.orig/src/fame_rate_1param.c libfame-0.9.1/src/fame_rate_1param.c +--- libfame-0.9.1.orig/src/fame_rate_1param.c 2008-08-02 22:44:06.860141435 +1000 ++++ libfame-0.9.1/src/fame_rate_1param.c 2008-08-02 22:44:35.794142186 +1000 +@@ -19,6 +19,7 @@ + + #include + #include ++#include /* strlen() */ + #include + #include "fame.h" + #include "fame_rate_1param.h" +diff -pruN libfame-0.9.1.orig/src/fame_rate_simple.c libfame-0.9.1/src/fame_rate_simple.c +--- libfame-0.9.1.orig/src/fame_rate_simple.c 2008-08-02 22:44:06.860141435 +1000 ++++ libfame-0.9.1/src/fame_rate_simple.c 2008-08-02 22:44:35.793143865 +1000 +@@ -19,6 +19,7 @@ + + #include + #include ++#include /* strlen() */ + #include + #include "fame.h" + #include "fame_rate_simple.h" diff --git a/libfame/libfame-0.9.1-mmx-pic.patch b/libfame/libfame-0.9.1-mmx-pic.patch new file mode 100644 index 000000000..838ee19ae --- /dev/null +++ b/libfame/libfame-0.9.1-mmx-pic.patch @@ -0,0 +1,366 @@ +diff -pruN libfame-0.9.1.orig/src/dct_mmx.h libfame-0.9.1/src/dct_mmx.h +--- libfame-0.9.1.orig/src/dct_mmx.h 2008-08-02 23:18:19.164140803 +1000 ++++ libfame-0.9.1/src/dct_mmx.h 2008-08-02 23:19:00.554142886 +1000 +@@ -22,6 +22,9 @@ + + #define precision + ++extern FAME_ALIGNED short const _mmx_1[]; ++extern FAME_ALIGNED short const _mmx_cos[]; ++ + static void inline dct_aan_pass(dct_t *cache) + { + // register unsigned short const *mmx_cos = _mmx_cos; +@@ -66,42 +69,42 @@ static void inline dct_aan_pass(dct_t *c + #ifdef precision + "psllw $0x01, %%mm5\n" /* precision(va0) += 1 bit */ + #endif +- "paddw " ASMSYM "_mmx_1, %%mm4\n" /* + 1 */ ++ "paddw (%2), %%mm4\n" /* + 1 */ + // "pmulhw 16(%1), %%mm5\n" /* (v14+v16)*COS6 -> mm5 (va0) */ +- "pmulhw " ASMSYM "_mmx_cos+16, %%mm5\n" /* (v14+v16)*COS6 -> mm5 (va0) */ ++ "pmulhw 16(%3), %%mm5\n" /* (v14+v16)*COS6 -> mm5 (va0) */ + "" /* STEP 4 */ + #ifdef precision + "psllw $0x02, %%mm6\n" /* precision(v22) += 1 bit */ + #else + "psllw $0x01, %%mm6\n" /* */ + #endif +- "paddw " ASMSYM "_mmx_1, %%mm4\n" /* + 1 */ ++ "paddw (%2), %%mm4\n" /* + 1 */ + // "pmulhw 8(%1), %%mm6\n" /* 2*v22*COS4/2 -> mm6 (v32)*/ +- "pmulhw " ASMSYM "_mmx_cos+8, %%mm6\n" /* 2*v22*COS4/2 -> mm6 (v32)*/ ++ "pmulhw 8(%3), %%mm6\n" /* 2*v22*COS4/2 -> mm6 (v32)*/ + #ifdef precision + "psllw $0x02, %%mm2\n" /* precision(v15) += 1 bit */ + #else + "psllw $0x01, %%mm2\n" /* */ + #endif +- "paddw " ASMSYM "_mmx_1, %%mm4\n" /* + 1 */ ++ "paddw (%2), %%mm4\n" /* + 1 */ + // "pmulhw 8(%1), %%mm2\n" /* 2*v15*COS4/2 -> mm2 (v35) */ +- "pmulhw " ASMSYM "_mmx_cos+8, %%mm2\n" /* 2*v15*COS4/2 -> mm2 (v35) */ ++ "pmulhw 8(%3), %%mm2\n" /* 2*v15*COS4/2 -> mm2 (v35) */ + #ifdef precision + "psllw $0x02, %%mm4\n" /* precision(v14) += 1 bit */ + #else + "psllw $0x01, %%mm4\n" /* */ + #endif +- "paddw " ASMSYM "_mmx_1, %%mm4\n" /* + 1 */ ++ "paddw (%2), %%mm4\n" /* + 1 */ + // "pmulhw 0(%1), %%mm4\n" /* 2 * v14 * -COS2/2 -> mm4 */ +- "pmulhw " ASMSYM "_mmx_cos, %%mm4\n" /* 2 * v14 * -COS2/2 -> mm4 */ ++ "pmulhw (%3), %%mm4\n" /* 2 * v14 * -COS2/2 -> mm4 */ + "psubsw %%mm5, %%mm4\n" /* v14*-COS2 - va0 -> mm4 (v34) */ + #ifdef precision + "psllw $0x01, %%mm1\n" /* precision(v16) += 1 bit */ + #endif + "psubsw %%mm1, %%mm5\n" /* va0 - v16 -> mm5 */ +- "paddw " ASMSYM "_mmx_1, %%mm4\n" /* + 1 */ ++ "paddw (%2), %%mm4\n" /* + 1 */ + // "pmulhw 24(%1), %%mm1\n" /* v16 * (COS8 - 1) -> mm1 */ +- "pmulhw " ASMSYM "_mmx_cos+24, %%mm1\n" /* v16 * (COS8 - 1) -> mm1 */ ++ "pmulhw 24(%3), %%mm1\n" /* v16 * (COS8 - 1) -> mm1 */ + "psubsw %%mm5, %%mm1\n" /* v16 * COS8 - va0 -> mm1 (v36)*/ + "" /* STEP 5 */ + "movq 0x70(%0), %%mm0\n" /* retrieve v07 -> mm0 */ +@@ -138,8 +141,8 @@ static void inline dct_aan_pass(dct_t *c + "movq %%mm0, 0x30(%0)\n" /* store line 3 */ + "movq %%mm4, 0x50(%0)\n" /* store line 5 */ + "movq %%mm2, 0x70(%0)\n" /* store line 7 */ +- : "=r"(cache)/*, "=r"(mmx_cos)*/ +- : "0"(cache)/*, "1"(mmx_cos)*/ ++ : "=r"(cache) ++ : "0"(cache), "r"(_mmx_1), "r"(_mmx_cos) + : "memory"); + } + +diff -pruN libfame-0.9.1.orig/src/dequantize_mmx.h libfame-0.9.1/src/dequantize_mmx.h +--- libfame-0.9.1.orig/src/dequantize_mmx.h 2008-08-02 23:18:19.164140803 +1000 ++++ libfame-0.9.1/src/dequantize_mmx.h 2008-08-02 23:19:00.555141217 +1000 +@@ -27,8 +27,8 @@ + "pmullw 0x" #x "8(%3), %%mm5\n" /* premultiply for iDCT */ \ + "psrlw $0x0b, %%mm4\n" /* keep 5 bits */ \ + "psrlw $0x0b, %%mm5\n" /* keep 5 bits */ \ +- "paddw " ASMSYM "_mmx_1, %%mm4\n" /* + 1 */ \ +- "paddw " ASMSYM "_mmx_1, %%mm5\n" /* + 1 */ \ ++ "paddw (%8), %%mm4\n" /* + 1 */ \ ++ "paddw (%8), %%mm5\n" /* + 1 */ \ + "psrlw $0x01, %%mm4\n" /* keep 4 bits rounded */ \ + "psrlw $0x01, %%mm5\n" /* keep 4 bits rounded */ \ + "psllw $0x04, %%mm0\n" /* multiply by 16 for iDCT */ \ +@@ -107,7 +107,7 @@ static void inline dequantize_intra_glob + DEQUANTIZE_GLOBAL_MISMATCH_CONTROL() + DEQUANTIZE_PRESCALE_STEP(7) + : "=r"(block), "=r"(dqmatrix), "=r"(cache), "=r"(psmatrix) +- : "0"(block), "1"(dqmatrix), "2"(cache), "3"(psmatrix) ++ : "0"(block), "1"(dqmatrix), "2"(cache), "3"(psmatrix), "r"(_mmx_1) + : "memory"); + + asm volatile("movd %%mm6, %0\n" /* export mismatch */ +@@ -160,8 +160,8 @@ static void inline dequantize_intra_loca + "pcmpeqw %%mm7, %%mm3\n" /* invert sign */ \ + "paddw %%mm2, %%mm0\n" /* sub 1 if >0 */ \ + "paddw %%mm3, %%mm1\n" /* sub 1 if >0 */ \ +- "por " ASMSYM "_mmx_1, %%mm0\n" /* or 1 */ \ +- "por " ASMSYM "_mmx_1, %%mm1\n" /* or 1 */ \ ++ "por (%8), %%mm0\n" /* or 1 */ \ ++ "por (%8), %%mm1\n" /* or 1 */ \ + "pand %%mm4, %%mm0\n" /* [0-3]=0 if [0-3] was zero */ \ + "pand %%mm5, %%mm1\n" /* [4-7]=0 if [4-7] was zero */ + +@@ -184,7 +184,7 @@ static void inline dequantize_intra_loca + DEQUANTIZE_INTRA_LOCAL_STEP(7) + DEQUANTIZE_PRESCALE_STEP(7) + : "=r"(block), "=r"(dqmatrix), "=r"(cache), "=r"(psmatrix) +- : "0"(block), "1"(dqmatrix), "2"(cache), "3"(psmatrix) ++ : "0"(block), "1"(dqmatrix), "2"(cache), "3"(psmatrix), "r"(_mmx_1) + : "memory"); + } + +@@ -256,7 +256,7 @@ static void inline dequantize_inter_glob + /* resetting the accumulator when the block is coded intra */ + DEQUANTIZE_PRESCALE_STEP(7) + : "=r"(block), "=r"(dqmatrix), "=r"(cache), "=r"(psmatrix) +- : "0"(block), "1"(dqmatrix), "2"(cache), "3"(psmatrix) ++ : "0"(block), "1"(dqmatrix), "2"(cache), "3"(psmatrix), "r"(_mmx_1) + : "memory"); + + asm volatile("movd %%mm6, %0\n" /* export mismatch */ +@@ -324,8 +324,8 @@ static void inline dequantize_inter_loca + "pcmpeqw %%mm7, %%mm3\n" /* invert sign */ \ + "paddw %%mm2, %%mm0\n" /* sub 1 if >0 */ \ + "paddw %%mm3, %%mm1\n" /* sub 1 if >0 */ \ +- "por " ASMSYM "_mmx_1, %%mm0\n" /* or 1 */ \ +- "por " ASMSYM "_mmx_1, %%mm1\n" /* or 1 */ \ ++ "por (%8), %%mm0\n" /* or 1 */ \ ++ "por (%8), %%mm1\n" /* or 1 */ \ + "pand %%mm4, %%mm0\n" /* [0-3]=0 if [0-3] was zero */ \ + "pand %%mm5, %%mm1\n" /* [4-7]=0 if [4-7] was zero */ + +@@ -348,6 +348,6 @@ static void inline dequantize_inter_loca + DEQUANTIZE_INTER_LOCAL_STEP(7) + DEQUANTIZE_PRESCALE_STEP(7) + : "=r"(block), "=r"(dqmatrix), "=r"(cache), "=r"(psmatrix) +- : "0"(block), "1"(dqmatrix), "2"(cache), "3"(psmatrix) ++ : "0"(block), "1"(dqmatrix), "2"(cache), "3"(psmatrix), "r"(_mmx_1) + : "memory"); + } +diff -pruN libfame-0.9.1.orig/src/fame_syntax_mpeg1.c libfame-0.9.1/src/fame_syntax_mpeg1.c +--- libfame-0.9.1.orig/src/fame_syntax_mpeg1.c 2008-08-02 23:18:19.167141250 +1000 ++++ libfame-0.9.1/src/fame_syntax_mpeg1.c 2008-08-02 23:19:00.556140239 +1000 +@@ -469,89 +469,6 @@ static void mpeg1_block_intra(fame_synta + fast_bitbuffer_write(data, shift, table[v+255].code, table[v+255].length); + + /* encode AC coefficients */ +-#if defined(HAS_BSWAP) +- { +- unsigned long dummy1, dummy2; +- +- /* Note: +- movsx mpeg1_table_clip+4096(, %%eax ,2), %%eax +- has been replaced by +- movw mpeg1_table_clip+4096(, %%eax ,2), %%ax +- movsx %%ax, %%eax +- because the first instruction failed on a PIII!! (wrong sign extension) +- whereas it worked well on my P75 :) +- */ +- /* Ok, a bit of explanations for a couple of tricks: +- The DC value of block is already coded and stored in v so we can use it to store something. +- We add one index to the zigzag table so that after coding block[63] we go to index 0. There +- we need to escape the zero counting loop (1), what we ensure by putting a non-zero value in +- the DC coefficient. Then we can test for index == 0 to exit. +- Now this non-zero value is a bit special :) +- In order to have one more 'half' register, we store sp value (16 less significant bit of the +- 32 bit register esp) *plus one* in the DC coefficient. Since the stack is aligned at an +- address multiple of 4 bytes (at least), we are sure that sp != 0xffff and thus sp+1 will +- never be zero. We then retrieve sp at the end for it is needed by 'pop' instructions. +- */ +- /* TODO : echange the role of edx and esp */ +- __asm__ __volatile__ ("pushl %%ebx\n" /* save ebx */ +- "pushl %%ebp\n" /* save stack pointer */ +- "inc %%sp\n" /* make sure sp != 0 */ +- "movw %%sp, (%%edx)\n" /* store sp+1 in DC ;) */ +- "movl %%esi, %%ebp\n" /* ebp = vlc_table */ +- "xorl %%eax, %%eax\n" /* eax = 0 */ +- "movl $" ASMSYM "mpeg1_zigzag_table+1, %%esi\n" /*esi = zigzag*/ +- "lea 1(%%esi), %%ebx\n" /* ebx = zigzag_table+1*/ +- "neg %%ebx\n" /* ebx = -(esi+1) */ +- ".p2align 4,,7\n" /* align for jump */ +- "0: xorw %%sp, %%sp\n" /* sp = 0 */ +- "1: movb (%%esi), %%al\n" /* eax = index in block*/ +- "incl %%esi\n" /* (faster than lodsb) */ +- "addw (%%edx, %%eax, 2), %%sp\n" /* sp = unzig */ +- "jz 1b\n" /* coeff == 0 then loop*/ +- "orl %%eax, %%eax\n" /* index == 0 then quit*/ +- "jz 2f\n" /* (faster than jcxz) */ +- "movsx %%sp, %%eax\n" /* extend sign */ +- "movw " ASMSYM "mpeg1_table_clip_data+4096(, %%eax ,2), %%ax\n" /*clip*/ +- "movsx %%ax, %%eax\n" /* extend sign */ +- "addl %%esi, %%ebx\n" /* ebx = run */ +- "shll $7, %%eax\n" /* eax *= 128(indexing)*/ +- "lea (%%eax, %%ebx, 2), %%eax\n" /*eax = 2 * offset*/ +- "lea (%%ebp, %%eax, 4), %%ebx\n" /* ebx = &vlc */ +- "movl (%%ebx), %%eax\n" /* eax = code */ +- "addl 4(%%ebx), %%ecx\n" /* ecx = shift+=length */ +- "xorl %%ebx, %%ebx\n" /* ebx = 0 */ +- "shrd %%cl, %%eax, %%ebx\n" /* adjust code to fit */ +- "shr %%cl, %%eax\n" /* adjust code to fit */ +- "bswap %%eax\n" /* reverse byte order of code */ +- "bswap %%ebx\n" /* reverse byte order of code */ +- "or %%eax, (%%edi)\n" /* put first 32 bits */ +- "movl %%ecx, %%eax\n" /* eax = shift + length*/ +- "shrl $5, %%eax\n" /* get dword increment */ +- "andl $31, %%ecx\n" /* mask shift */ +- "lea (%%edi, %%eax, 4), %%edi\n"/* data+=(ecx>32)*/ +- "orl %%ebx, (%%edi)\n" /* put last 32 bits */ +- "xorl %%eax, %%eax\n" /* eax = 0 */ +- "lea 1(%%esi), %%ebx\n" /* ebx = esi + 1 (last)*/ +- "neg %%ebx\n" /* ebx = -(esi + 1) */ +- "jmp 0b\n" /* loop */ +- "2:\n" +- "movw (%%edx), %%sp\n" /* retrieve sp+1 */ +- "dec %%sp\n" /* restore esp */ +- "popl %%ebp\n" /* reload stack pointer*/ +- "popl %%ebx\n" /* reload ebx */ +- : "=c"(shift), +- "=a"(dummy1), +- "=d"(block), +- "=D"(data), +- "=S"(dummy2) +- : "d"(block), +- "c"(shift), +- "D"(data), +- "S"(syntax_mpeg1->vlc_table) +- : "memory"); +- block[0] = v; /* restore DC value */ +- } +-#else + { + short i; + unsigned long last; +@@ -573,7 +490,6 @@ static void mpeg1_block_intra(fame_synta + } + } + } +-#endif /* HAS_BSWAP */ + + /* mark end of block */ + fast_bitbuffer_write(data, shift, 2, 2); +diff -pruN libfame-0.9.1.orig/src/half_mmx.h libfame-0.9.1/src/half_mmx.h +--- libfame-0.9.1.orig/src/half_mmx.h 2008-08-02 23:18:19.161140026 +1000 ++++ libfame-0.9.1/src/half_mmx.h 2008-08-02 23:19:00.556140239 +1000 +@@ -68,8 +68,8 @@ static void inline mmx_interpolate(unsig + "paddw %%mm5, %%mm6\n" /* mm6 = ref00+ref10+ref11+1-r 4-7*/ + "psrlw $1, %%mm4\n" /* divide by 2 */ + "psrlw $1, %%mm5\n" /* divide by 2 */ +- "paddw " ASMSYM "_mmx_one, %%mm3\n" /* add 1 */ +- "paddw " ASMSYM "_mmx_one, %%mm6\n" /* add 1 */ ++ "paddw (%8), %%mm3\n" /* add 1 */ ++ "paddw (%8), %%mm6\n" /* add 1 */ + "packuswb %%mm5, %%mm4\n" /* pack to byte and saturate */ + "movq 1(%3), %%mm1\n" /* mm1 = [ref+1] */ + "movq %%mm1, %%mm2\n" /* mm2 = mm1 */ +@@ -87,7 +87,7 @@ static void inline mmx_interpolate(unsig + "movl 12(%0), %3\n" /* %3 = ref[3] */ + "movq %%mm3, (%3)\n" /* store in frame */ + : "=r"(ref), "=r"(pitch), "=r"(rc), "=r"(dummy) +- : "0"(ref), "1"(pitch), "2"(rc), "3"(dummy) ++ : "0"(ref), "1"(pitch), "2"(rc), "3"(dummy), "r"(_mmx_one) + : "memory"); + } + +diff -pruN libfame-0.9.1.orig/src/half_sse.h libfame-0.9.1/src/half_sse.h +--- libfame-0.9.1.orig/src/half_sse.h 2008-08-02 23:18:19.160143502 +1000 ++++ libfame-0.9.1/src/half_sse.h 2008-08-02 23:19:00.557139019 +1000 +@@ -71,8 +71,8 @@ static void inline mmx_interpolate_signe + "paddw %%mm5, %%mm6\n" /* mm6 = ref00+ref10+ref11+1-r 4-7*/ + "psrlw $1, %%mm4\n" /* divide by 2 */ + "psrlw $1, %%mm5\n" /* divide by 2 */ +- "paddw " ASMSYM "_mmx_one, %%mm3\n" /* add 1 */ +- "paddw " ASMSYM "_mmx_one, %%mm6\n" /* add 1 */ ++ "paddw (%8), %%mm3\n" /* add 1 */ ++ "paddw (%8), %%mm6\n" /* add 1 */ + "packuswb %%mm5, %%mm4\n" /* pack to byte and saturate */ + "movq 1(%3), %%mm1\n" /* mm1 = [ref+1] */ + "movq %%mm1, %%mm2\n" /* mm2 = mm1 */ +@@ -90,7 +90,7 @@ static void inline mmx_interpolate_signe + "movl 12(%0), %3\n" /* %3 = ref[3] */ + "movq %%mm3, (%3)\n" /* store in frame */ + : "=r"(ref), "=r"(pitch), "=r"(rc), "=r"(dummy) +- : "0"(ref), "1"(pitch), "2"(rc), "3"(dummy) ++ : "0"(ref), "1"(pitch), "2"(rc), "3"(dummy), "r"(_mmx_one) + : "memory"); + } + +diff -pruN libfame-0.9.1.orig/src/idct_mmx.h libfame-0.9.1/src/idct_mmx.h +--- libfame-0.9.1.orig/src/idct_mmx.h 2008-08-02 23:18:19.160143502 +1000 ++++ libfame-0.9.1/src/idct_mmx.h 2008-08-02 23:19:00.557139019 +1000 +@@ -18,6 +18,10 @@ + */ + /*************************** MMX accelerated iDCT ****************************/ + ++extern FAME_ALIGNED short const _mmx_1[]; ++extern FAME_ALIGNED short const _mmx_cos[]; ++extern FAME_ALIGNED short const _mmx_icos[]; ++ + static void inline idct_aan_pass(dct_t * block) + { + // register unsigned short const *mmx_icos = _mmx_icos; +@@ -65,9 +69,9 @@ static void inline idct_aan_pass(dct_t * + block[row*8+6] = v45; - v71, v11, v44, v65, v24 - + */ + "psllw $0x02, %%mm5\n" /* adjust v22 for multiply */ +- "paddw " ASMSYM "_mmx_1, %%mm5\n" /* + 1 for rounding */ ++ "paddw (%2), %%mm5\n" /* + 1 for rounding */ + // "pmulhw 8(%1), %%mm5\n" /* 4*v15*ICOS4/4 -> mm5 (v23) */ +- "pmulhw " ASMSYM "_mmx_icos+8, %%mm5\n" /* 4*v15*ICOS4/4 -> mm5 (v23)*/ ++ "pmulhw 8(%3), %%mm5\n" /* 4*v15*ICOS4/4 -> mm5 (v23)*/ + "psubsw %%mm4, %%mm5\n" /* v23 - v62 -> mm5 (v24) */ + "movq %%mm3, %%mm6\n" /* v44 -> mm6 */ + "paddsw %%mm5, %%mm6\n" /* v44 + v24 -> mm6 (v45) */ +@@ -125,25 +129,25 @@ static void inline idct_aan_pass(dct_t * + block[row*8+4] += v55; - - + */ + "psllw $0x02, %%mm0\n" /* adjust v12 for multiply */ +- "paddw " ASMSYM "_mmx_1, %%mm0\n" /* + 1 for rounding */ ++ "paddw (%2), %%mm0\n" /* + 1 for rounding */ + // "pmulhw 8(%1), %%mm0\n" /* 4*v12*ICOS4/4 -> mm0 (v13) */ +- "pmulhw " ASMSYM "_mmx_icos+8, %%mm0\n" /* 4*v12*ICOS4/4 -> mm0 (v13) */ ++ "pmulhw 8(%3), %%mm0\n" /* 4*v12*ICOS4/4 -> mm0 (v13) */ + "movq %%mm2, %%mm6\n" /* v51 -> mm6 */ + "psubsw %%mm1, %%mm6\n" /* v51 - v71 -> mm6 (va2) */ + "psllw $0x03, %%mm2\n" /* adjust v51 for multiply */ +- "paddw " ASMSYM "_mmx_1, %%mm2\n" /* + 1 for rounding */ ++ "paddw (%2), %%mm2\n" /* + 1 for rounding */ + /* should add another one here but it seems to look better without */ + // "pmulhw 16(%1), %%mm2\n" /* 8*v51*ICOS6/8 -> mm2 (v53) */ +- "pmulhw " ASMSYM "_mmx_icos+16, %%mm2\n" /* 8*v51*ICOS6/8 -> mm2 (v53) */ ++ "pmulhw 16(%3), %%mm2\n" /* 8*v51*ICOS6/8 -> mm2 (v53) */ + "psllw $0x02, %%mm1\n" /* adjust v71 for multiply */ +- "paddw " ASMSYM "_mmx_1, %%mm1\n" /* + 1 for rounding */ ++ "paddw (%2), %%mm1\n" /* + 1 for rounding */ + /* should add another one here but it seems to look better without */ + // "pmulhw 0(%1), %%mm1\n" /* 4*v71*ICOS2/4 -> mm1 (v73) */ +- "pmulhw " ASMSYM "_mmx_icos, %%mm1\n" /* 4*v71*ICOS2/4 -> mm1 (v73) */ ++ "pmulhw (%3), %%mm1\n" /* 4*v71*ICOS2/4 -> mm1 (v73) */ + "psllw $0x01, %%mm6\n" /* adjust va2 for multiply */ +- "paddw " ASMSYM "_mmx_1, %%mm6\n" /* + 1 for rounding */ ++ "paddw (%2), %%mm6\n" /* + 1 for rounding */ + // "pmulhw 24(%1), %%mm6\n" /* 2*v12*ICOS8/2 -> mm6 (va3) */ +- "pmulhw " ASMSYM "_mmx_icos+24, %%mm6\n" /* 2*v12*ICOS8/2 -> mm6 (va3) */ ++ "pmulhw 24(%3), %%mm6\n" /* 2*v12*ICOS8/2 -> mm6 (va3) */ + "psubsw %%mm6, %%mm2\n" /* v53 - va3 -> mm2 (v54) */ + "psubsw %%mm6, %%mm1\n" /* v73 - va3 -> mm1 (v74) */ + "psubsw %%mm3, %%mm1\n" /* v74 - v32 -> mm3 (v75) */ +@@ -167,8 +171,8 @@ static void inline idct_aan_pass(dct_t * + "paddsw %%mm0, %%mm7\n" /* v65 + v55 -> mm7 */ + "movq %%mm6, 0x30(%0)\n" /* mm6 -> line 3 */ + "movq %%mm7, 0x40(%0)\n" /* mm7 -> line 4 */ +- : "=r"(block)/*, "=r"(mmx_icos)*/ +- : "0"(block)/*, "1"(mmx_icos)*/ ++ : "=r"(block) ++ : "0"(block), "r"(_mmx_1), "r"(_mmx_icos) + : "memory"); + } + diff --git a/libfame/libfame-0.9.1-mmx.patch b/libfame/libfame-0.9.1-mmx.patch new file mode 100644 index 000000000..6d20bc9e8 --- /dev/null +++ b/libfame/libfame-0.9.1-mmx.patch @@ -0,0 +1,133 @@ +diff -pruN libfame-0.9.1.orig/src/fame_decoder_mpeg.c libfame-0.9.1/src/fame_decoder_mpeg.c +--- libfame-0.9.1.orig/src/fame_decoder_mpeg.c 2008-08-02 22:44:06.860141435 +1000 ++++ libfame-0.9.1/src/fame_decoder_mpeg.c 2008-08-02 22:55:42.516142586 +1000 +@@ -78,7 +78,9 @@ static void mpeg_pad(fame_decoder_t *dec + fame_box_t *box); + static void mpeg_interpolate(fame_decoder_t *decoder, int rounding); + static void mpeg_leave(fame_decoder_t *decoder); ++#ifdef HAS_MMX + static void mpeg_close(fame_decoder_t *decoder); ++#endif + + FAME_CONSTRUCTOR(fame_decoder_mpeg_t) + { +@@ -90,7 +92,11 @@ FAME_CONSTRUCTOR(fame_decoder_mpeg_t) + FAME_DECODER(this)->pad = mpeg_pad; + FAME_DECODER(this)->interpolate = mpeg_interpolate; + FAME_DECODER(this)->leave = mpeg_leave; ++#ifdef HAS_MMX + FAME_DECODER(this)->close = mpeg_close; ++#else ++ FAME_DECODER(this)->close = NULL; ++#endif + return(this); + } + +@@ -603,6 +609,7 @@ static void mpeg_leave(fame_decoder_t *d + arch_leave_state(); + } + ++#ifdef HAS_MMX + /* mpeg_close */ + /* */ + /* Description: */ +@@ -618,7 +625,6 @@ static void mpeg_close(fame_decoder_t *d + { + fame_decoder_mpeg_t *decoder_mpeg = FAME_DECODER_MPEG(decoder); + +-#ifdef HAS_MMX + /* free mismatch accumulator */ + { + int i; +@@ -626,5 +632,5 @@ static void mpeg_close(fame_decoder_t *d + for(i = 0; i < 6; i++) + fame_free(decoder_mpeg->mismatch_accumulator[i]); + } +-#endif + } ++#endif +diff -pruN libfame-0.9.1.orig/src/half_mmx.h libfame-0.9.1/src/half_mmx.h +--- libfame-0.9.1.orig/src/half_mmx.h 2008-08-02 22:44:06.856141130 +1000 ++++ libfame-0.9.1/src/half_mmx.h 2008-08-02 22:51:49.078142536 +1000 +@@ -28,10 +28,10 @@ static void inline mmx_interpolate(unsig + short _mmx_rc[4]; /* rounding control */ + register short *rc = _mmx_rc; + +- _mmx_rc[0] = 1 - rounding; +- _mmx_rc[1] = 1 - rounding; +- _mmx_rc[2] = 1 - rounding; +- _mmx_rc[3] = 1 - rounding; ++ _mmx_rc[0] = _mmx_one[0] - rounding; ++ _mmx_rc[1] = _mmx_one[1] - rounding; ++ _mmx_rc[2] = _mmx_one[2] - rounding; ++ _mmx_rc[3] = _mmx_one[3] - rounding; + + asm volatile ("movl (%0), %3\n" /* %3 = ref[0] */ + "pxor %%mm7, %%mm7\n" /* mm7 = 0 */ +diff -pruN libfame-0.9.1.orig/src/mae_int.h libfame-0.9.1/src/mae_int.h +--- libfame-0.9.1.orig/src/mae_int.h 2008-08-02 22:44:06.862141891 +1000 ++++ libfame-0.9.1/src/mae_int.h 2008-08-02 22:55:42.517140304 +1000 +@@ -18,10 +18,10 @@ + Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + */ + +-static unsigned int MAE8x8_withmask(unsigned char *ref, +- unsigned char *input, +- unsigned char *shape, +- int pitch) ++static inline unsigned int ++MAE8x8_withmask(unsigned char *ref, unsigned char *input, ++ unsigned char *shape, ++ int pitch) + { + int j,i; + +@@ -41,10 +41,11 @@ static unsigned int MAE8x8_withmask(unsi + return error; + } + +-static unsigned int MAE8x8_withoutmask(unsigned char *ref, +- unsigned char *input, +- unsigned char *shape, +- int pitch) ++static inline unsigned int ++MAE8x8_withoutmask(unsigned char *ref, ++ unsigned char *input, ++ unsigned char *shape, ++ int pitch) + { + int j,i; + +diff -pruN libfame-0.9.1.orig/src/mae_mmx.h libfame-0.9.1/src/mae_mmx.h +--- libfame-0.9.1.orig/src/mae_mmx.h 2008-08-02 22:44:06.855143632 +1000 ++++ libfame-0.9.1/src/mae_mmx.h 2008-08-02 22:55:42.517140304 +1000 +@@ -18,10 +18,10 @@ + Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + */ + +-static unsigned int MAE8x8_withmask(unsigned char *ref, +- unsigned char *input, +- unsigned char *shape, +- int pitch) ++static inline unsigned int ++MAE8x8_withmask(unsigned char *ref, unsigned char *input, ++ unsigned char *shape, ++ int pitch) + { + int dummy; + unsigned long retval; +@@ -70,10 +70,10 @@ static unsigned int MAE8x8_withmask(unsi + return (retval&65535); + } + +-static unsigned int MAE8x8_withoutmask(unsigned char *ref, +- unsigned char *input, +- unsigned char *shape, +- int pitch) ++static inline unsigned int ++MAE8x8_withoutmask(unsigned char *ref, unsigned char *input, ++ unsigned char *shape, ++ int pitch) + { + int dummy; + unsigned long retval; diff --git a/libfame/libfame-0.9.1-ptr-null-check.patch b/libfame/libfame-0.9.1-ptr-null-check.patch new file mode 100644 index 000000000..349b04558 --- /dev/null +++ b/libfame/libfame-0.9.1-ptr-null-check.patch @@ -0,0 +1,19 @@ +diff -pruN libfame-0.9.1.orig/src/fame_malloc.c libfame-0.9.1/src/fame_malloc.c +--- libfame-0.9.1.orig/src/fame_malloc.c 2008-08-02 22:44:06.858140887 +1000 ++++ libfame-0.9.1/src/fame_malloc.c 2008-08-02 23:05:08.693143429 +1000 +@@ -36,9 +36,13 @@ void* fame_malloc(size_t size) + */ + + ptr = (unsigned char*) malloc(size+ALIGN); +- aligned = (unsigned char*) (((unsigned int)ptr & (~(ALIGN-1))) + ALIGN ); ++ if (ptr == NULL) { ++ perror("fame_malloc"); ++ exit(1); ++ } ++ aligned = (unsigned char*) (((uintptr_t)ptr & (~(ALIGN-1))) + ALIGN ); + padding = aligned - 1; +- *padding = (ALIGN-1) - ((unsigned int)ptr & (ALIGN-1)); ++ *padding = (ALIGN-1) - ((uintptr_t)ptr & (ALIGN-1)); + + return ((void*)aligned); + }