Skip to content

F
ffmpeg-mt
Commit 09fcf89b authored by michael's avatar michael
Browse files
Options

adapting existing mmx/mmx2/sse/3dnow optimizations so they work on x86_64... 

adapting existing mmx/mmx2/sse/3dnow optimizations so they work on x86_64 patch by (Aurelien Jacobs <aurel at gnuage dot org>)


git-svn-id: file:///var/local/repositories/ffmpeg/trunk@3578 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b
parent fbfb4b78
Hide whitespace changes
Inline Side-by-side
Showing with 1129 additions and 1060 deletions
configure
View file @ 09fcf89b
......@@ -106,6 +106,14 @@ case "$cpu" in
i386|i486|i586|i686|i86pc|BePC)
cpu="x86"
;;
x86_64)
if [ "`$cc -dumpmachine | grep x86_64 | cut -d- -f1`" = "x86_64" -a \
-z "`echo $CFLAGS | grep -- -m32`" ]; then
cpu="x86_64"
else
cpu="x86"
fi
;;
# armv4l is a subset of armv5tel
armv4l|armv5tel)
cpu="armv4l"
......@@ -500,7 +508,7 @@ fi
# compute mmx state
if test $mmx = "default"; then
if test $cpu = "x86"; then
if test $cpu = "x86" -o $cpu = "x86_64"; then
mmx="yes"
else
mmx="no"
......@@ -827,6 +835,7 @@ done
# test gcc version to see if vector builtins can be used
# currently only used on i386 for MMX builtins
cat > $TMPC << EOF
#include <xmmintrin.h>
int main(void) {
#if __GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ >= 2)
return 0;
......@@ -985,7 +994,7 @@ echo "CPU $cpu ($tune)"
echo "Big Endian $bigendian"
echo "inttypes.h $inttypes"
echo "broken inttypes.h $emu_fast_int"
if test $cpu = "x86"; then
if test $cpu = "x86" -o $cpu = "x86_64"; then
echo "MMX enabled $mmx"
echo "Vector Builtins $builtin_vector"
fi
......@@ -1074,6 +1083,9 @@ echo "TARGET_OS=$TARGET_OS" >> config.mak
if test "$cpu" = "x86" ; then
echo "TARGET_ARCH_X86=yes" >> config.mak
echo "#define ARCH_X86 1" >> $TMPH
elif test "$cpu" = "x86_64" ; then
echo "TARGET_ARCH_X86_64=yes" >> config.mak
echo "#define ARCH_X86_64 1" >> $TMPH
elif test "$cpu" = "armv4l" ; then
echo "TARGET_ARCH_ARMV4L=yes" >> config.mak
echo "#define ARCH_ARMV4L 1" >> $TMPH
......
libavcodec/bswap.h
View file @ 09fcf89b
......@@ -10,17 +10,23 @@
#include <byteswap.h>
#else
#ifdef ARCH_X86
static inline unsigned short ByteSwap16(unsigned short x)
#ifdef ARCH_X86_64
# define LEGACY_REGS "=Q"
#else
# define LEGACY_REGS "=q"
#endif
#if defined(ARCH_X86) || defined(ARCH_X86_64)
static inline uint16_t ByteSwap16(uint16_t x)
{
__asm("xchgb %b0,%h0" :
"=q" (x) :
LEGACY_REGS (x) :
"0" (x));
return x;
}
#define bswap_16(x) ByteSwap16(x)
static inline unsigned int ByteSwap32(unsigned int x)
static inline uint32_t ByteSwap32(uint32_t x)
{
#if __CPU__ > 386
__asm("bswap %0":
......@@ -29,21 +35,28 @@ static inline unsigned int ByteSwap32(unsigned int x)
__asm("xchgb %b0,%h0\n"
" rorl $16,%0\n"
" xchgb %b0,%h0":
"=q" (x) :
LEGACY_REGS (x) :
#endif
"0" (x));
return x;
}
#define bswap_32(x) ByteSwap32(x)
static inline unsigned long long int ByteSwap64(unsigned long long int x)
static inline uint64_t ByteSwap64(uint64_t x)
{
#ifdef ARCH_X86_64
__asm("bswap %0":
"=r" (x) :
"0" (x));
return x;
#else
register union { __extension__ uint64_t __ll;
uint32_t __l[2]; } __x;
asm("xchgl %0,%1":
"=r"(__x.__l[0]),"=r"(__x.__l[1]):
"0"(bswap_32((unsigned long)x)),"1"(bswap_32((unsigned long)(x>>32))));
"0"(bswap_32((uint32_t)x)),"1"(bswap_32((uint32_t)(x>>32))));
return __x.__ll;
#endif
}
#define bswap_64(x) ByteSwap64(x)
......
libavcodec/common.h
View file @ 09fcf89b
......@@ -254,7 +254,7 @@ inline void dprintf(const char* fmt,...) {}
extern const uint32_t inverse[256];
#ifdef ARCH_X86
#if defined(ARCH_X86) || defined(ARCH_X86_64)
# define FASTDIV(a,b) \
({\
int ret,dmy;\
......@@ -271,7 +271,7 @@ extern const uint32_t inverse[256];
# define FASTDIV(a,b) ((a)/(b))
#endif
#ifdef ARCH_X86
#if defined(ARCH_X86) || defined(ARCH_X86_64)
// avoid +32 for shift optimization (gcc should do that ...)
static inline int32_t NEG_SSR32( int32_t a, int8_t s){
asm ("sarl %1, %0\n\t"
......@@ -390,7 +390,7 @@ typedef struct RL_VLC_ELEM {
#endif
/* used to avoid missaligned exceptions on some archs (alpha, ...) */
#ifdef ARCH_X86
#if defined(ARCH_X86) || defined(ARCH_X86_64)
# define unaligned32(a) (*(const uint32_t*)(a))
#else
# ifdef __GNUC__
......@@ -460,7 +460,7 @@ static inline void put_bits(PutBitContext *s, int n, unsigned int value)
static inline void put_bits(PutBitContext *s, int n, unsigned int value)
{
# ifdef ALIGNED_BITSTREAM_WRITER
# ifdef ARCH_X86
# if defined(ARCH_X86) || defined(ARCH_X86_64)
asm volatile(
"movl %0, %%ecx \n\t"
"xorl %%eax, %%eax \n\t"
......@@ -491,7 +491,7 @@ static inline void put_bits(PutBitContext *s, int n, unsigned int value)
s->index= index;
# endif
# else //ALIGNED_BITSTREAM_WRITER
# ifdef ARCH_X86
# if defined(ARCH_X86) || defined(ARCH_X86_64)
asm volatile(
"movl $7, %%ecx \n\t"
"andl %0, %%ecx \n\t"
......@@ -738,7 +738,7 @@ static inline int get_bits_count(GetBitContext *s){
name##_bit_count-= 32;\
}\
#ifdef ARCH_X86
#if defined(ARCH_X86) || defined(ARCH_X86_64)
# define SKIP_CACHE(name, gb, num)\
asm(\
"shldl %2, %1, %0 \n\t"\
......@@ -1218,7 +1218,7 @@ static inline int ff_get_fourcc(const char *s){
#define MKBETAG(a,b,c,d) (d | (c << 8) | (b << 16) | (a << 24))
#ifdef ARCH_X86
#if defined(ARCH_X86) || defined(ARCH_X86_64)
#define MASK_ABS(mask, level)\
asm volatile(\
"cdq \n\t"\
......@@ -1252,7 +1252,7 @@ if((y)<(x)){\
}
#endif
#ifdef ARCH_X86
#if defined(ARCH_X86) || defined(ARCH_X86_64)
static inline long long rdtsc(void)
{
long long l;
......
libavcodec/i386/cputest.c
View file @ 09fcf89b
......@@ -4,12 +4,20 @@
#include <stdlib.h>
#include "../dsputil.h"
#ifdef ARCH_X86_64
# define REG_b "rbx"
# define REG_S "rsi"
#else
# define REG_b "ebx"
# define REG_S "esi"
#endif
/* ebx saving is necessary for PIC. gcc seems unable to see it alone */
#define cpuid(index,eax,ebx,ecx,edx)\
__asm __volatile\
("movl %%ebx, %%esi\n\t"\
("mov %%"REG_b", %%"REG_S"\n\t"\
"cpuid\n\t"\
"xchgl %%ebx, %%esi"\
"xchg %%"REG_b", %%"REG_S\
: "=a" (eax), "=S" (ebx),\
"=c" (ecx), "=d" (edx)\
: "0" (index));
......@@ -24,7 +32,7 @@ int mm_support(void)
/* See if CPUID instruction is supported ... */
/* ... Get copies of EFLAGS into eax and ecx */
"pushf\n\t"
"popl %0\n\t"
"pop %0\n\t"
"movl %0, %1\n\t"
/* ... Toggle the ID bit in one copy and store */
......@@ -35,7 +43,7 @@ int mm_support(void)
/* ... Get the (hopefully modified) EFLAGS */
"pushf\n\t"
"popl %0\n\t"
"pop %0\n\t"
: "=a" (eax), "=c" (ecx)
:
: "cc"
......
libavcodec/i386/dsputil_mmx.c
View file @ 09fcf89b
......@@ -187,7 +187,7 @@ static const uint64_t ff_pb_FC attribute_used __attribute__ ((aligned(8))) = 0xF
static void get_pixels_mmx(DCTELEM *block, const uint8_t *pixels, int line_size)
{
asm volatile(
"movl $-128, %%eax \n\t"
"mov $-128, %%"REG_a" \n\t"
"pxor %%mm7, %%mm7 \n\t"
".balign 16 \n\t"
"1: \n\t"
......@@ -199,16 +199,16 @@ static void get_pixels_mmx(DCTELEM *block, const uint8_t *pixels, int line_size)
"punpckhbw %%mm7, %%mm1 \n\t"
"punpcklbw %%mm7, %%mm2 \n\t"
"punpckhbw %%mm7, %%mm3 \n\t"
"movq %%mm0, (%1, %%eax)\n\t"
"movq %%mm1, 8(%1, %%eax)\n\t"
"movq %%mm2, 16(%1, %%eax)\n\t"
"movq %%mm3, 24(%1, %%eax)\n\t"
"addl %3, %0 \n\t"
"addl $32, %%eax \n\t"
"movq %%mm0, (%1, %%"REG_a")\n\t"
"movq %%mm1, 8(%1, %%"REG_a")\n\t"
"movq %%mm2, 16(%1, %%"REG_a")\n\t"
"movq %%mm3, 24(%1, %%"REG_a")\n\t"
"add %3, %0 \n\t"
"add $32, %%"REG_a" \n\t"
"js 1b \n\t"
: "+r" (pixels)
: "r" (block+64), "r" (line_size), "r" (line_size*2)
: "%eax"
: "r" (block+64), "r" ((long)line_size), "r" ((long)line_size*2)
: "%"REG_a
);
}
......@@ -216,7 +216,7 @@ static inline void diff_pixels_mmx(DCTELEM *block, const uint8_t *s1, const uint
{
asm volatile(
"pxor %%mm7, %%mm7 \n\t"
"movl $-128, %%eax \n\t"
"mov $-128, %%"REG_a" \n\t"
".balign 16 \n\t"
"1: \n\t"
"movq (%0), %%mm0 \n\t"
......@@ -229,15 +229,15 @@ static inline void diff_pixels_mmx(DCTELEM *block, const uint8_t *s1, const uint
"punpckhbw %%mm7, %%mm3 \n\t"
"psubw %%mm2, %%mm0 \n\t"
"psubw %%mm3, %%mm1 \n\t"
"movq %%mm0, (%2, %%eax)\n\t"
"movq %%mm1, 8(%2, %%eax)\n\t"
"addl %3, %0 \n\t"
"addl %3, %1 \n\t"
"addl $16, %%eax \n\t"
"movq %%mm0, (%2, %%"REG_a")\n\t"
"movq %%mm1, 8(%2, %%"REG_a")\n\t"
"add %3, %0 \n\t"
"add %3, %1 \n\t"
"add $16, %%"REG_a" \n\t"
"jnz 1b \n\t"
: "+r" (s1), "+r" (s2)
: "r" (block+64), "r" (stride)
: "%eax"
: "r" (block+64), "r" ((long)stride)
: "%"REG_a
);
}
#endif //CONFIG_ENCODERS
......@@ -268,7 +268,7 @@ void put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size
"movq %%mm2, (%0, %1)\n\t"
"movq %%mm4, (%0, %1, 2)\n\t"
"movq %%mm6, (%0, %2)\n\t"
::"r" (pix), "r" (line_size), "r" (line_size*3), "m"(*p)
::"r" (pix), "r" ((long)line_size), "r" ((long)line_size*3), "m"(*p)
:"memory");
pix += line_size*4;
p += 32;
......@@ -293,7 +293,7 @@ void put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size
"movq %%mm2, (%0, %1)\n\t"
"movq %%mm4, (%0, %1, 2)\n\t"
"movq %%mm6, (%0, %2)\n\t"
::"r" (pix), "r" (line_size), "r" (line_size*3), "r"(p)
::"r" (pix), "r" ((long)line_size), "r" ((long)line_size*3), "r"(p)
:"memory");
}
......@@ -359,59 +359,59 @@ void add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size
static void put_pixels4_mmx(uint8_t *block, const uint8_t *pixels, int line_size, int h)
{
__asm __volatile(
"lea (%3, %3), %%eax \n\t"
"lea (%3, %3), %%"REG_a" \n\t"
".balign 8 \n\t"
"1: \n\t"
"movd (%1), %%mm0 \n\t"
"movd (%1, %3), %%mm1 \n\t"
"movd %%mm0, (%2) \n\t"
"movd %%mm1, (%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %1 \n\t"
"add %%"REG_a", %2 \n\t"
"movd (%1), %%mm0 \n\t"
"movd (%1, %3), %%mm1 \n\t"
"movd %%mm0, (%2) \n\t"
"movd %%mm1, (%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %1 \n\t"
"add %%"REG_a", %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
: "+g"(h), "+r" (pixels), "+r" (block)
: "r"(line_size)
: "%eax", "memory"
: "r"((long)line_size)
: "%"REG_a, "memory"
);
}
static void put_pixels8_mmx(uint8_t *block, const uint8_t *pixels, int line_size, int h)
{
__asm __volatile(
"lea (%3, %3), %%eax \n\t"
"lea (%3, %3), %%"REG_a" \n\t"
".balign 8 \n\t"
"1: \n\t"
"movq (%1), %%mm0 \n\t"
"movq (%1, %3), %%mm1 \n\t"
"movq %%mm0, (%2) \n\t"
"movq %%mm1, (%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %1 \n\t"
"add %%"REG_a", %2 \n\t"
"movq (%1), %%mm0 \n\t"
"movq (%1, %3), %%mm1 \n\t"
"movq %%mm0, (%2) \n\t"
"movq %%mm1, (%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %1 \n\t"
"add %%"REG_a", %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
: "+g"(h), "+r" (pixels), "+r" (block)
: "r"(line_size)
: "%eax", "memory"
: "r"((long)line_size)
: "%"REG_a, "memory"
);
}
static void put_pixels16_mmx(uint8_t *block, const uint8_t *pixels, int line_size, int h)
{
__asm __volatile(
"lea (%3, %3), %%eax \n\t"
"lea (%3, %3), %%"REG_a" \n\t"
".balign 8 \n\t"
"1: \n\t"
"movq (%1), %%mm0 \n\t"
......@@ -422,8 +422,8 @@ static void put_pixels16_mmx(uint8_t *block, const uint8_t *pixels, int line_siz
"movq %%mm4, 8(%2) \n\t"
"movq %%mm1, (%2, %3) \n\t"
"movq %%mm5, 8(%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %1 \n\t"
"add %%"REG_a", %2 \n\t"
"movq (%1), %%mm0 \n\t"
"movq 8(%1), %%mm4 \n\t"
"movq (%1, %3), %%mm1 \n\t"
......@@ -432,13 +432,13 @@ static void put_pixels16_mmx(uint8_t *block, const uint8_t *pixels, int line_siz
"movq %%mm4, 8(%2) \n\t"
"movq %%mm1, (%2, %3) \n\t"
"movq %%mm5, 8(%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %1 \n\t"
"add %%"REG_a", %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
: "+g"(h), "+r" (pixels), "+r" (block)
: "r"(line_size)
: "%eax", "memory"
: "r"((long)line_size)
: "%"REG_a, "memory"
);
}
......@@ -446,16 +446,16 @@ static void clear_blocks_mmx(DCTELEM *blocks)
{
__asm __volatile(
"pxor %%mm7, %%mm7 \n\t"
"movl $-128*6, %%eax \n\t"
"mov $-128*6, %%"REG_a" \n\t"
"1: \n\t"
"movq %%mm7, (%0, %%eax) \n\t"
"movq %%mm7, 8(%0, %%eax) \n\t"
"movq %%mm7, 16(%0, %%eax) \n\t"
"movq %%mm7, 24(%0, %%eax) \n\t"
"addl $32, %%eax \n\t"
"movq %%mm7, (%0, %%"REG_a") \n\t"
"movq %%mm7, 8(%0, %%"REG_a") \n\t"
"movq %%mm7, 16(%0, %%"REG_a") \n\t"
"movq %%mm7, 24(%0, %%"REG_a") \n\t"
"add $32, %%"REG_a" \n\t"
" js 1b \n\t"
: : "r" (((int)blocks)+128*6)
: "%eax"
: : "r" (((uint8_t *)blocks)+128*6)
: "%"REG_a
);
}
......@@ -463,7 +463,7 @@ static void clear_blocks_mmx(DCTELEM *blocks)
static int pix_sum16_mmx(uint8_t * pix, int line_size){
const int h=16;
int sum;
int index= -line_size*h;
long index= -line_size*h;
__asm __volatile(
"pxor %%mm7, %%mm7 \n\t"
......@@ -481,7 +481,7 @@ static int pix_sum16_mmx(uint8_t * pix, int line_size){
"paddw %%mm2, %%mm3 \n\t"
"paddw %%mm1, %%mm3 \n\t"
"paddw %%mm3, %%mm6 \n\t"
"addl %3, %1 \n\t"
"add %3, %1 \n\t"
" js 1b \n\t"
"movq %%mm6, %%mm5 \n\t"
"psrlq $32, %%mm6 \n\t"
......@@ -492,7 +492,7 @@ static int pix_sum16_mmx(uint8_t * pix, int line_size){
"movd %%mm6, %0 \n\t"
"andl $0xFFFF, %0 \n\t"
: "=&r" (sum), "+r" (index)
: "r" (pix - index), "r" (line_size)
: "r" (pix - index), "r" ((long)line_size)
);
return sum;
......@@ -500,7 +500,7 @@ static int pix_sum16_mmx(uint8_t * pix, int line_size){
#endif //CONFIG_ENCODERS
static void add_bytes_mmx(uint8_t *dst, uint8_t *src, int w){
int i=0;
long i=0;
asm volatile(
"1: \n\t"
"movq (%1, %0), %%mm0 \n\t"
......@@ -511,11 +511,11 @@ static void add_bytes_mmx(uint8_t *dst, uint8_t *src, int w){
"movq 8(%2, %0), %%mm1 \n\t"
"paddb %%mm0, %%mm1 \n\t"
"movq %%mm1, 8(%2, %0) \n\t"
"addl $16, %0 \n\t"
"cmpl %3, %0 \n\t"
"add $16, %0 \n\t"
"cmp %3, %0 \n\t"
" jb 1b \n\t"
: "+r" (i)
: "r"(src), "r"(dst), "r"(w-15)
: "r"(src), "r"(dst), "r"((long)w-15)
);
for(; i<w; i++)
dst[i+0] += src[i+0];
......@@ -726,7 +726,7 @@ static int pix_norm1_mmx(uint8_t *pix, int line_size) {
"paddd %%mm3,%%mm4\n"
"paddd %%mm2,%%mm7\n"
"addl %2, %0\n"
"add %2, %0\n"
"paddd %%mm4,%%mm7\n"
"dec %%ecx\n"
"jnz 1b\n"
......@@ -735,7 +735,7 @@ static int pix_norm1_mmx(uint8_t *pix, int line_size) {
"psrlq $32, %%mm7\n" /* shift hi dword to lo */
"paddd %%mm7,%%mm1\n"
"movd %%mm1,%1\n"
: "+r" (pix), "=r"(tmp) : "r" (line_size) : "%ecx" );
: "+r" (pix), "=r"(tmp) : "r" ((long)line_size) : "%ecx" );
return tmp;
}
......@@ -763,8 +763,8 @@ static int sse8_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int
"pmaddwd %%mm2,%%mm2\n"
"pmaddwd %%mm1,%%mm1\n"
"addl %3,%0\n"
"addl %3,%1\n"
"add %3,%0\n"
"add %3,%1\n"
"paddd %%mm2,%%mm1\n"
"paddd %%mm1,%%mm7\n"
......@@ -777,7 +777,7 @@ static int sse8_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int
"paddd %%mm7,%%mm1\n"
"movd %%mm1,%2\n"
: "+r" (pix1), "+r" (pix2), "=r"(tmp)
: "r" (line_size) , "m" (h)
: "r" ((long)line_size) , "m" (h)
: "%ecx");
return tmp;
}
......@@ -821,8 +821,8 @@ static int sse16_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int
"pmaddwd %%mm1,%%mm1\n"
"pmaddwd %%mm3,%%mm3\n"
"addl %3,%0\n"
"addl %3,%1\n"
"add %3,%0\n"
"add %3,%1\n"
"paddd %%mm2,%%mm1\n"
"paddd %%mm4,%%mm3\n"
......@@ -837,7 +837,7 @@ static int sse16_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int
"paddd %%mm7,%%mm1\n"
"movd %%mm1,%2\n"
: "+r" (pix1), "+r" (pix2), "=r"(tmp)
: "r" (line_size) , "m" (h)
: "r" ((long)line_size) , "m" (h)
: "%ecx");
return tmp;
}
......@@ -863,7 +863,7 @@ static int hf_noise8_mmx(uint8_t * pix1, int line_size, int h) {
"psubw %%mm1, %%mm0\n"
"psubw %%mm3, %%mm2\n"
"addl %2,%0\n"
"add %2,%0\n"
"movq (%0),%%mm4\n"
"movq %%mm4, %%mm1\n"
......@@ -891,7 +891,7 @@ static int hf_noise8_mmx(uint8_t * pix1, int line_size, int h) {
"paddw %%mm0, %%mm2\n"
"paddw %%mm2, %%mm6\n"
"addl %2,%0\n"
"add %2,%0\n"
"1:\n"
"movq (%0),%%mm0\n"
......@@ -920,7 +920,7 @@ static int hf_noise8_mmx(uint8_t * pix1, int line_size, int h) {
"paddw %%mm4, %%mm5\n"
"paddw %%mm5, %%mm6\n"
"addl %2,%0\n"
"add %2,%0\n"
"movq (%0),%%mm4\n"
"movq %%mm4, %%mm1\n"
......@@ -948,7 +948,7 @@ static int hf_noise8_mmx(uint8_t * pix1, int line_size, int h) {
"paddw %%mm0, %%mm2\n"
"paddw %%mm2, %%mm6\n"
"addl %2,%0\n"
"add %2,%0\n"
"subl $2, %%ecx\n"
" jnz 1b\n"
......@@ -962,7 +962,7 @@ static int hf_noise8_mmx(uint8_t * pix1, int line_size, int h) {
"paddd %%mm6,%%mm0\n"
"movd %%mm0,%1\n"
: "+r" (pix1), "=r"(tmp)
: "r" (line_size) , "g" (h-2)
: "r" ((long)line_size) , "g" (h-2)
: "%ecx");
return tmp;
}
......@@ -986,7 +986,7 @@ static int hf_noise16_mmx(uint8_t * pix1, int line_size, int h) {
"psubw %%mm1, %%mm0\n"
"psubw %%mm3, %%mm2\n"
"addl %2,%0\n"
"add %2,%0\n"
"movq (%0),%%mm4\n"
"movq 1(%0),%%mm1\n"
......@@ -1011,7 +1011,7 @@ static int hf_noise16_mmx(uint8_t * pix1, int line_size, int h) {
"paddw %%mm0, %%mm2\n"
"paddw %%mm2, %%mm6\n"
"addl %2,%0\n"
"add %2,%0\n"
"1:\n"
"movq (%0),%%mm0\n"
......@@ -1037,7 +1037,7 @@ static int hf_noise16_mmx(uint8_t * pix1, int line_size, int h) {
"paddw %%mm4, %%mm5\n"
"paddw %%mm5, %%mm6\n"
"addl %2,%0\n"
"add %2,%0\n"
"movq (%0),%%mm4\n"
"movq 1(%0),%%mm1\n"
......@@ -1062,7 +1062,7 @@ static int hf_noise16_mmx(uint8_t * pix1, int line_size, int h) {
"paddw %%mm0, %%mm2\n"
"paddw %%mm2, %%mm6\n"
"addl %2,%0\n"
"add %2,%0\n"
"subl $2, %%ecx\n"
" jnz 1b\n"
......@@ -1076,7 +1076,7 @@ static int hf_noise16_mmx(uint8_t * pix1, int line_size, int h) {
"paddd %%mm6,%%mm0\n"
"movd %%mm0,%1\n"
: "+r" (pix1), "=r"(tmp)
: "r" (line_size) , "g" (h-2)
: "r" ((long)line_size) , "g" (h-2)
: "%ecx");
return tmp + hf_noise8_mmx(pix+8, line_size, h);
}
......@@ -1106,7 +1106,7 @@ static int vsad_intra16_mmx(void *v, uint8_t * pix, uint8_t * dummy, int line_si
#define SUM(in0, in1, out0, out1) \
"movq (%0), %%mm2\n"\
"movq 8(%0), %%mm3\n"\
"addl %2,%0\n"\
"add %2,%0\n"\
"movq %%mm2, " #out0 "\n"\
"movq %%mm3, " #out1 "\n"\
"psubusb " #in0 ", %%mm2\n"\
......@@ -1133,7 +1133,7 @@ static int vsad_intra16_mmx(void *v, uint8_t * pix, uint8_t * dummy, int line_si
"pxor %%mm7,%%mm7\n"
"movq (%0),%%mm0\n"
"movq 8(%0),%%mm1\n"
"addl %2,%0\n"
"add %2,%0\n"
"subl $2, %%ecx\n"
SUM(%%mm0, %%mm1, %%mm4, %%mm5)
"1:\n"
......@@ -1153,7 +1153,7 @@ static int vsad_intra16_mmx(void *v, uint8_t * pix, uint8_t * dummy, int line_si
"paddw %%mm6,%%mm0\n"
"movd %%mm0,%1\n"
: "+r" (pix), "=r"(tmp)
: "r" (line_size) , "m" (h)
: "r" ((long)line_size) , "m" (h)
: "%ecx");
return tmp & 0xFFFF;
}
......@@ -1168,7 +1168,7 @@ static int vsad_intra16_mmx2(void *v, uint8_t * pix, uint8_t * dummy, int line_s
#define SUM(in0, in1, out0, out1) \
"movq (%0), " #out0 "\n"\
"movq 8(%0), " #out1 "\n"\
"addl %2,%0\n"\
"add %2,%0\n"\
"psadbw " #out0 ", " #in0 "\n"\
"psadbw " #out1 ", " #in1 "\n"\
"paddw " #in1 ", " #in0 "\n"\
......@@ -1180,7 +1180,7 @@ static int vsad_intra16_mmx2(void *v, uint8_t * pix, uint8_t * dummy, int line_s
"pxor %%mm7,%%mm7\n"
"movq (%0),%%mm0\n"
"movq 8(%0),%%mm1\n"
"addl %2,%0\n"
"add %2,%0\n"
"subl $2, %%ecx\n"
SUM(%%mm0, %%mm1, %%mm4, %%mm5)
"1:\n"
......@@ -1194,7 +1194,7 @@ static int vsad_intra16_mmx2(void *v, uint8_t * pix, uint8_t * dummy, int line_s
"movd %%mm6,%1\n"
: "+r" (pix), "=r"(tmp)
: "r" (line_size) , "m" (h)
: "r" ((long)line_size) , "m" (h)
: "%ecx");
return tmp;
}
......@@ -1212,8 +1212,8 @@ static int vsad16_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, in
"movq (%1)," #out0 "\n"\
"movq 8(%0),%%mm3\n"\
"movq 8(%1)," #out1 "\n"\
"addl %3,%0\n"\
"addl %3,%1\n"\
"add %3,%0\n"\
"add %3,%1\n"\
"psubb " #out0 ", %%mm2\n"\
"psubb " #out1 ", %%mm3\n"\
"pxor %%mm7, %%mm2\n"\
......@@ -1248,8 +1248,8 @@ static int vsad16_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, in
"movq (%1),%%mm2\n"
"movq 8(%0),%%mm1\n"
"movq 8(%1),%%mm3\n"
"addl %3,%0\n"
"addl %3,%1\n"
"add %3,%0\n"
"add %3,%1\n"
"subl $2, %%ecx\n"
"psubb %%mm2, %%mm0\n"
"psubb %%mm3, %%mm1\n"
......@@ -1273,7 +1273,7 @@ static int vsad16_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, in
"paddw %%mm6,%%mm0\n"
"movd %%mm0,%2\n"
: "+r" (pix1), "+r" (pix2), "=r"(tmp)
: "r" (line_size) , "m" (h)
: "r" ((long)line_size) , "m" (h)
: "%ecx");
return tmp & 0x7FFF;
}
......@@ -1291,8 +1291,8 @@ static int vsad16_mmx2(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, i
"movq (%1),%%mm2\n"\
"movq 8(%0)," #out1 "\n"\
"movq 8(%1),%%mm3\n"\
"addl %3,%0\n"\
"addl %3,%1\n"\
"add %3,%0\n"\
"add %3,%1\n"\
"psubb %%mm2, " #out0 "\n"\
"psubb %%mm3, " #out1 "\n"\
"pxor %%mm7, " #out0 "\n"\
......@@ -1312,8 +1312,8 @@ static int vsad16_mmx2(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, i
"movq (%1),%%mm2\n"
"movq 8(%0),%%mm1\n"
"movq 8(%1),%%mm3\n"
"addl %3,%0\n"
"addl %3,%1\n"
"add %3,%0\n"
"add %3,%1\n"
"subl $2, %%ecx\n"
"psubb %%mm2, %%mm0\n"
"psubb %%mm3, %%mm1\n"
......@@ -1331,14 +1331,14 @@ static int vsad16_mmx2(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, i
"movd %%mm6,%2\n"
: "+r" (pix1), "+r" (pix2), "=r"(tmp)
: "r" (line_size) , "m" (h)
: "r" ((long)line_size) , "m" (h)
: "%ecx");
return tmp;
}
#undef SUM
static void diff_bytes_mmx(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w){
int i=0;
long i=0;
asm volatile(
"1: \n\t"
"movq (%2, %0), %%mm0 \n\t"
......@@ -1349,18 +1349,18 @@ static void diff_bytes_mmx(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w){
"movq 8(%1, %0), %%mm1 \n\t"
"psubb %%mm0, %%mm1 \n\t"
"movq %%mm1, 8(%3, %0) \n\t"
"addl $16, %0 \n\t"
"cmpl %4, %0 \n\t"
"add $16, %0 \n\t"
"cmp %4, %0 \n\t"
" jb 1b \n\t"
: "+r" (i)
: "r"(src1), "r"(src2), "r"(dst), "r"(w-15)
: "r"(src1), "r"(src2), "r"(dst), "r"((long)w-15)
);
for(; i<w; i++)
dst[i+0] = src1[i+0]-src2[i+0];
}
static void sub_hfyu_median_prediction_mmx2(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w, int *left, int *left_top){
int i=0;
long i=0;
uint8_t l, lt;
asm volatile(
......@@ -1379,11 +1379,11 @@ static void sub_hfyu_median_prediction_mmx2(uint8_t *dst, uint8_t *src1, uint8_t
"pmaxub %%mm1, %%mm4 \n\t"
"psubb %%mm4, %%mm3 \n\t" // dst - pred
"movq %%mm3, (%3, %0) \n\t"
"addl $8, %0 \n\t"
"cmpl %4, %0 \n\t"
"add $8, %0 \n\t"
"cmp %4, %0 \n\t"
" jb 1b \n\t"
: "+r" (i)
: "r"(src1), "r"(src2), "r"(dst), "r"(w)
: "r"(src1), "r"(src2), "r"(dst), "r"((long)w)
);
l= *left;
......@@ -1772,12 +1772,12 @@ static void OPNAME ## mpeg4_qpel16_h_lowpass_mmx2(uint8_t *dst, uint8_t *src, in
"packuswb %%mm4, %%mm0 \n\t"\
OP_MMX2(%%mm0, 8(%1), %%mm4, q)\
\
"addl %3, %0 \n\t"\
"addl %4, %1 \n\t"\
"add %3, %0 \n\t"\
"add %4, %1 \n\t"\
"decl %2 \n\t"\
" jnz 1b \n\t"\
: "+a"(src), "+c"(dst), "+m"(h)\
: "d"(srcStride), "S"(dstStride), /*"m"(ff_pw_20), "m"(ff_pw_3),*/ "m"(temp), "m"(ROUNDER)\
: "d"((long)srcStride), "S"((long)dstStride), /*"m"(ff_pw_20), "m"(ff_pw_3),*/ "m"(temp), "m"(ROUNDER)\
: "memory"\
);\
}\
......@@ -1885,12 +1885,12 @@ static void OPNAME ## mpeg4_qpel8_h_lowpass_mmx2(uint8_t *dst, uint8_t *src, int
"packuswb %%mm3, %%mm0 \n\t"\
OP_MMX2(%%mm0, (%1), %%mm4, q)\
\
"addl %3, %0 \n\t"\
"addl %4, %1 \n\t"\
"add %3, %0 \n\t"\
"add %4, %1 \n\t"\
"decl %2 \n\t"\
" jnz 1b \n\t"\
: "+a"(src), "+c"(dst), "+m"(h)\
: "S"(srcStride), "D"(dstStride), /*"m"(ff_pw_20), "m"(ff_pw_3),*/ "m"(temp), "m"(ROUNDER)\
: "S"((long)srcStride), "D"((long)dstStride), /*"m"(ff_pw_20), "m"(ff_pw_3),*/ "m"(temp), "m"(ROUNDER)\
: "memory"\
);\
}\
......@@ -1949,12 +1949,12 @@ static void OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src,
"movq %%mm1, 17*8(%1) \n\t"\
"movq %%mm2, 2*17*8(%1) \n\t"\
"movq %%mm3, 3*17*8(%1) \n\t"\
"addl $8, %1 \n\t"\
"addl %3, %0 \n\t"\
"add $8, %1 \n\t"\
"add %3, %0 \n\t"\
"decl %2 \n\t"\
" jnz 1b \n\t"\
: "+r" (src), "+r" (temp_ptr), "+r"(count)\
: "r" (srcStride)\
: "r" ((long)srcStride)\
: "memory"\
);\
\
......@@ -1971,37 +1971,37 @@ static void OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src,
"movq 24(%0), %%mm3 \n\t"\
QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 16(%0), 8(%0), (%0), 32(%0), (%1), OP)\
QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 8(%0), (%0), (%0), 40(%0), (%1, %3), OP)\
"addl %4, %1 \n\t"\
"add %4, %1 \n\t"\
QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, (%0), (%0), 8(%0), 48(%0), (%1), OP)\
\
QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, (%0), 8(%0), 16(%0), 56(%0), (%1, %3), OP)\
"addl %4, %1 \n\t"\
"add %4, %1 \n\t"\
QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 8(%0), 16(%0), 24(%0), 64(%0), (%1), OP)\
QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 16(%0), 24(%0), 32(%0), 72(%0), (%1, %3), OP)\
"addl %4, %1 \n\t"\
"add %4, %1 \n\t"\
QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 24(%0), 32(%0), 40(%0), 80(%0), (%1), OP)\
QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 32(%0), 40(%0), 48(%0), 88(%0), (%1, %3), OP)\
"addl %4, %1 \n\t"\
"add %4, %1 \n\t"\
QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 40(%0), 48(%0), 56(%0), 96(%0), (%1), OP)\
QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 48(%0), 56(%0), 64(%0),104(%0), (%1, %3), OP)\
"addl %4, %1 \n\t"\
"add %4, %1 \n\t"\
QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 56(%0), 64(%0), 72(%0),112(%0), (%1), OP)\
QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 64(%0), 72(%0), 80(%0),120(%0), (%1, %3), OP)\
"addl %4, %1 \n\t"\
"add %4, %1 \n\t"\
QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 72(%0), 80(%0), 88(%0),128(%0), (%1), OP)\
\
QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 80(%0), 88(%0), 96(%0),128(%0), (%1, %3), OP)\
"addl %4, %1 \n\t" \
"add %4, %1 \n\t" \
QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 88(%0), 96(%0),104(%0),120(%0), (%1), OP)\
QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 96(%0),104(%0),112(%0),112(%0), (%1, %3), OP)\
\
"addl $136, %0 \n\t"\
"addl %6, %1 \n\t"\
"add $136, %0 \n\t"\
"add %6, %1 \n\t"\
"decl %2 \n\t"\
" jnz 1b \n\t"\
\
: "+r"(temp_ptr), "+r"(dst), "+g"(count)\
: "r"(dstStride), "r"(2*dstStride), /*"m"(ff_pw_20), "m"(ff_pw_3),*/ "m"(ROUNDER), "g"(4-14*dstStride)\
: "r"((long)dstStride), "r"(2*(long)dstStride), /*"m"(ff_pw_20), "m"(ff_pw_3),*/ "m"(ROUNDER), "g"(4-14*(long)dstStride)\
:"memory"\
);\
}\
......@@ -2021,12 +2021,12 @@ static void OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src,
"punpckhbw %%mm7, %%mm1 \n\t"\
"movq %%mm0, (%1) \n\t"\
"movq %%mm1, 9*8(%1) \n\t"\
"addl $8, %1 \n\t"\
"addl %3, %0 \n\t"\
"add $8, %1 \n\t"\
"add %3, %0 \n\t"\
"decl %2 \n\t"\
" jnz 1b \n\t"\
: "+r" (src), "+r" (temp_ptr), "+r"(count)\
: "r" (srcStride)\
: "r" ((long)srcStride)\
: "memory"\
);\
\
......@@ -2043,25 +2043,25 @@ static void OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src,
"movq 24(%0), %%mm3 \n\t"\
QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 16(%0), 8(%0), (%0), 32(%0), (%1), OP)\
QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 8(%0), (%0), (%0), 40(%0), (%1, %3), OP)\
"addl %4, %1 \n\t"\
"add %4, %1 \n\t"\
QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, (%0), (%0), 8(%0), 48(%0), (%1), OP)\
\
QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, (%0), 8(%0), 16(%0), 56(%0), (%1, %3), OP)\
"addl %4, %1 \n\t"\
"add %4, %1 \n\t"\
QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 8(%0), 16(%0), 24(%0), 64(%0), (%1), OP)\
\
QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 16(%0), 24(%0), 32(%0), 64(%0), (%1, %3), OP)\
"addl %4, %1 \n\t"\
"add %4, %1 \n\t"\
QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 24(%0), 32(%0), 40(%0), 56(%0), (%1), OP)\
QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 32(%0), 40(%0), 48(%0), 48(%0), (%1, %3), OP)\
\
"addl $72, %0 \n\t"\
"addl %6, %1 \n\t"\
"add $72, %0 \n\t"\
"add %6, %1 \n\t"\
"decl %2 \n\t"\
" jnz 1b \n\t"\
\
: "+r"(temp_ptr), "+r"(dst), "+g"(count)\
: "r"(dstStride), "r"(2*dstStride), /*"m"(ff_pw_20), "m"(ff_pw_3),*/ "m"(ROUNDER), "g"(4-6*dstStride)\
: "r"((long)dstStride), "r"(2*(long)dstStride), /*"m"(ff_pw_20), "m"(ff_pw_3),*/ "m"(ROUNDER), "g"(4-6*(long)dstStride)\
: "memory"\
);\
}\
......@@ -2297,7 +2297,7 @@ static void OPNAME ## qpel16_mc22_ ## MMX(uint8_t *dst, uint8_t *src, int stride
"psubw "#B", %%mm6 \n\t"\
"psubw "#E", %%mm6 \n\t"\
"pmullw %4, %%mm6 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"punpcklbw %%mm7, "#F" \n\t"\
"paddw %5, "#A" \n\t"\
"paddw "#F", "#A" \n\t"\
......@@ -2305,7 +2305,7 @@ static void OPNAME ## qpel16_mc22_ ## MMX(uint8_t *dst, uint8_t *src, int stride
"psraw $5, %%mm6 \n\t"\
"packuswb %%mm6, %%mm6 \n\t"\
OP(%%mm6, (%1), A, d)\
"addl %3, %1 \n\t"
"add %3, %1 \n\t"
#define QPEL_H264HV(A,B,C,D,E,F,OF)\
"movd (%0), "#F" \n\t"\
......@@ -2315,7 +2315,7 @@ static void OPNAME ## qpel16_mc22_ ## MMX(uint8_t *dst, uint8_t *src, int stride
"psubw "#B", %%mm6 \n\t"\
"psubw "#E", %%mm6 \n\t"\
"pmullw %3, %%mm6 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"punpcklbw %%mm7, "#F" \n\t"\
"paddw "#F", "#A" \n\t"\
"paddw "#A", %%mm6 \n\t"\
......@@ -2353,12 +2353,12 @@ static void OPNAME ## h264_qpel4_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, i
"psraw $5, %%mm0 \n\t"\
"packuswb %%mm0, %%mm0 \n\t"\
OP(%%mm0, (%1),%%mm6, d)\
"addl %3, %0 \n\t"\
"addl %4, %1 \n\t"\
"add %3, %0 \n\t"\
"add %4, %1 \n\t"\
"decl %2 \n\t"\
" jnz 1b \n\t"\
: "+a"(src), "+c"(dst), "+m"(h)\
: "d"(srcStride), "S"(dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
: "d"((long)srcStride), "S"((long)dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
: "memory"\
);\
}\
......@@ -2367,15 +2367,15 @@ static void OPNAME ## h264_qpel4_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, i
asm volatile(\
"pxor %%mm7, %%mm7 \n\t"\
"movd (%0), %%mm0 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"movd (%0), %%mm1 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"movd (%0), %%mm2 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"movd (%0), %%mm3 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"movd (%0), %%mm4 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"punpcklbw %%mm7, %%mm0 \n\t"\
"punpcklbw %%mm7, %%mm1 \n\t"\
"punpcklbw %%mm7, %%mm2 \n\t"\
......@@ -2387,7 +2387,7 @@ static void OPNAME ## h264_qpel4_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, i
QPEL_H264V(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP)\
\
: "+a"(src), "+c"(dst)\
: "S"(srcStride), "D"(dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
: "S"((long)srcStride), "D"((long)dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
: "memory"\
);\
}\
......@@ -2399,15 +2399,15 @@ static void OPNAME ## h264_qpel4_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp,
asm volatile(\
"pxor %%mm7, %%mm7 \n\t"\
"movd (%0), %%mm0 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"movd (%0), %%mm1 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"movd (%0), %%mm2 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"movd (%0), %%mm3 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"movd (%0), %%mm4 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"punpcklbw %%mm7, %%mm0 \n\t"\
"punpcklbw %%mm7, %%mm1 \n\t"\
"punpcklbw %%mm7, %%mm2 \n\t"\
......@@ -2419,7 +2419,7 @@ static void OPNAME ## h264_qpel4_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp,
QPEL_H264HV(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, 3*8*3)\
\
: "+a"(src)\
: "c"(tmp), "S"(srcStride), "m"(ff_pw_5)\
: "c"(tmp), "S"((long)srcStride), "m"(ff_pw_5)\
: "memory"\
);\
tmp += 4;\
......@@ -2445,12 +2445,12 @@ static void OPNAME ## h264_qpel4_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp,
"psraw $6, %%mm0 \n\t"\
"packuswb %%mm0, %%mm0 \n\t"\
OP(%%mm0, (%1),%%mm7, d)\
"addl $24, %0 \n\t"\
"addl %3, %1 \n\t"\
"add $24, %0 \n\t"\
"add %3, %1 \n\t"\
"decl %2 \n\t"\
" jnz 1b \n\t"\
: "+a"(tmp), "+c"(dst), "+m"(h)\
: "S"(dstStride), "m"(ff_pw_32)\
: "S"((long)dstStride), "m"(ff_pw_32)\
: "memory"\
);\
}\
......@@ -2502,12 +2502,12 @@ static void OPNAME ## h264_qpel8_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, i
"psraw $5, %%mm1 \n\t"\
"packuswb %%mm1, %%mm0 \n\t"\
OP(%%mm0, (%1),%%mm5, q)\
"addl %3, %0 \n\t"\
"addl %4, %1 \n\t"\
"add %3, %0 \n\t"\
"add %4, %1 \n\t"\
"decl %2 \n\t"\
" jnz 1b \n\t"\
: "+a"(src), "+c"(dst), "+m"(h)\
: "d"(srcStride), "S"(dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
: "d"((long)srcStride), "S"((long)dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
: "memory"\
);\
}\
......@@ -2520,15 +2520,15 @@ static void OPNAME ## h264_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, i
asm volatile(\
"pxor %%mm7, %%mm7 \n\t"\
"movd (%0), %%mm0 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"movd (%0), %%mm1 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"movd (%0), %%mm2 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"movd (%0), %%mm3 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"movd (%0), %%mm4 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"punpcklbw %%mm7, %%mm0 \n\t"\
"punpcklbw %%mm7, %%mm1 \n\t"\
"punpcklbw %%mm7, %%mm2 \n\t"\
......@@ -2544,7 +2544,7 @@ static void OPNAME ## h264_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, i
QPEL_H264V(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP)\
\
: "+a"(src), "+c"(dst)\
: "S"(srcStride), "D"(dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
: "S"((long)srcStride), "D"((long)dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
: "memory"\
);\
src += 4-13*srcStride;\
......@@ -2559,15 +2559,15 @@ static void OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp,
asm volatile(\
"pxor %%mm7, %%mm7 \n\t"\
"movd (%0), %%mm0 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"movd (%0), %%mm1 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"movd (%0), %%mm2 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"movd (%0), %%mm3 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"movd (%0), %%mm4 \n\t"\
"addl %2, %0 \n\t"\
"add %2, %0 \n\t"\
"punpcklbw %%mm7, %%mm0 \n\t"\
"punpcklbw %%mm7, %%mm1 \n\t"\
"punpcklbw %%mm7, %%mm2 \n\t"\
......@@ -2583,7 +2583,7 @@ static void OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp,
QPEL_H264HV(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, 7*8*4)\
\
: "+a"(src)\
: "c"(tmp), "S"(srcStride), "m"(ff_pw_5)\
: "c"(tmp), "S"((long)srcStride), "m"(ff_pw_5)\
: "memory"\
);\
tmp += 4;\
......@@ -2623,12 +2623,12 @@ static void OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp,
"psraw $6, %%mm3 \n\t"\
"packuswb %%mm3, %%mm0 \n\t"\
OP(%%mm0, (%1),%%mm7, q)\
"addl $32, %0 \n\t"\
"addl %3, %1 \n\t"\
"add $32, %0 \n\t"\
"add %3, %1 \n\t"\
"decl %2 \n\t"\
" jnz 1b \n\t"\
: "+a"(tmp), "+c"(dst), "+m"(h)\
: "S"(dstStride), "m"(ff_pw_32)\
: "S"((long)dstStride), "m"(ff_pw_32)\
: "memory"\
);\
}\
......@@ -2831,7 +2831,7 @@ static void just_return() { return; }
c->avg_ ## postfix1 = avg_ ## postfix2;
static int try_8x8basis_mmx(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale){
int i=0;
long i=0;
assert(ABS(scale) < 256);
scale<<= 16 + 1 - BASIS_SHIFT + RECON_SHIFT;
......@@ -2863,8 +2863,8 @@ static int try_8x8basis_mmx(int16_t rem[64], int16_t weight[64], int16_t basis[6
"paddd %%mm1, %%mm0 \n\t"
"psrld $4, %%mm0 \n\t"
"paddd %%mm0, %%mm7 \n\t"
"addl $16, %0 \n\t"
"cmpl $128, %0 \n\t" //FIXME optimize & bench
"add $16, %0 \n\t"
"cmp $128, %0 \n\t" //FIXME optimize & bench
" jb 1b \n\t"
"movq %%mm7, %%mm6 \n\t"
"psrlq $32, %%mm7 \n\t"
......@@ -2879,7 +2879,7 @@ static int try_8x8basis_mmx(int16_t rem[64], int16_t weight[64], int16_t basis[6
}
static void add_8x8basis_mmx(int16_t rem[64], int16_t basis[64], int scale){
int i=0;
long i=0;
if(ABS(scale) < 256){
scale<<= 16 + 1 - BASIS_SHIFT + RECON_SHIFT;
......@@ -2902,8 +2902,8 @@ static void add_8x8basis_mmx(int16_t rem[64], int16_t basis[64], int scale){
"paddw 8(%2, %0), %%mm1 \n\t"
"movq %%mm0, (%2, %0) \n\t"
"movq %%mm1, 8(%2, %0) \n\t"
"addl $16, %0 \n\t"
"cmpl $128, %0 \n\t" //FIXME optimize & bench
"add $16, %0 \n\t"
"cmp $128, %0 \n\t" //FIXME optimize & bench
" jb 1b \n\t"
: "+r" (i)
......
libavcodec/i386/dsputil_mmx_avg.h
View file @ 09fcf89b
......@@ -28,7 +28,7 @@
static void DEF(put_pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
{
__asm __volatile(
"lea (%3, %3), %%eax \n\t"
"lea (%3, %3), %%"REG_a" \n\t"
"1: \n\t"
"movq (%1), %%mm0 \n\t"
"movq (%1, %3), %%mm1 \n\t"
......@@ -36,21 +36,21 @@ static void DEF(put_pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line_
PAVGB" 1(%1, %3), %%mm1 \n\t"
"movq %%mm0, (%2) \n\t"
"movq %%mm1, (%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %1 \n\t"
"add %%"REG_a", %2 \n\t"
"movq (%1), %%mm0 \n\t"
"movq (%1, %3), %%mm1 \n\t"
PAVGB" 1(%1), %%mm0 \n\t"
PAVGB" 1(%1, %3), %%mm1 \n\t"
"addl %%eax, %1 \n\t"
"add %%"REG_a", %1 \n\t"
"movq %%mm0, (%2) \n\t"
"movq %%mm1, (%2, %3) \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
:"+g"(h), "+S"(pixels), "+D"(block)
:"r" (line_size)
:"%eax", "memory");
:"r" ((long)line_size)
:"%"REG_a, "memory");
}
static void DEF(put_pixels4_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h)
......@@ -60,34 +60,34 @@ static void DEF(put_pixels4_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int
" jz 1f \n\t"
"movd (%1), %%mm0 \n\t"
"movd (%2), %%mm1 \n\t"
"addl %4, %1 \n\t"
"addl $4, %2 \n\t"
"add %4, %1 \n\t"
"add $4, %2 \n\t"
PAVGB" %%mm1, %%mm0 \n\t"
"movd %%mm0, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"decl %0 \n\t"
"1: \n\t"
"movd (%1), %%mm0 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
"movd (%1), %%mm1 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
PAVGB" (%2), %%mm0 \n\t"
PAVGB" 4(%2), %%mm1 \n\t"
"movd %%mm0, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"movd %%mm1, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"movd (%1), %%mm0 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
"movd (%1), %%mm1 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
PAVGB" 8(%2), %%mm0 \n\t"
PAVGB" 12(%2), %%mm1 \n\t"
"movd %%mm0, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"movd %%mm1, (%3) \n\t"
"addl %5, %3 \n\t"
"addl $16, %2 \n\t"
"add %5, %3 \n\t"
"add $16, %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
#ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used
......@@ -95,7 +95,7 @@ static void DEF(put_pixels4_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int
#else
:"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)
#endif
:"S"(src1Stride), "D"(dstStride)
:"S"((long)src1Stride), "D"((long)dstStride)
:"memory");
}
......@@ -107,34 +107,34 @@ static void DEF(put_pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int
" jz 1f \n\t"
"movq (%1), %%mm0 \n\t"
"movq (%2), %%mm1 \n\t"
"addl %4, %1 \n\t"
"addl $8, %2 \n\t"
"add %4, %1 \n\t"
"add $8, %2 \n\t"
PAVGB" %%mm1, %%mm0 \n\t"
"movq %%mm0, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"decl %0 \n\t"
"1: \n\t"
"movq (%1), %%mm0 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
"movq (%1), %%mm1 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
PAVGB" (%2), %%mm0 \n\t"
PAVGB" 8(%2), %%mm1 \n\t"
"movq %%mm0, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"movq %%mm1, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"movq (%1), %%mm0 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
"movq (%1), %%mm1 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
PAVGB" 16(%2), %%mm0 \n\t"
PAVGB" 24(%2), %%mm1 \n\t"
"movq %%mm0, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"movq %%mm1, (%3) \n\t"
"addl %5, %3 \n\t"
"addl $32, %2 \n\t"
"add %5, %3 \n\t"
"add $32, %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
#ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used
......@@ -142,7 +142,7 @@ static void DEF(put_pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int
#else
:"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)
#endif
:"S"(src1Stride), "D"(dstStride)
:"S"((long)src1Stride), "D"((long)dstStride)
:"memory");
//the following should be used, though better not with gcc ...
/* :"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst)
......@@ -158,20 +158,20 @@ static void DEF(put_no_rnd_pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src
" jz 1f \n\t"
"movq (%1), %%mm0 \n\t"
"movq (%2), %%mm1 \n\t"
"addl %4, %1 \n\t"
"addl $8, %2 \n\t"
"add %4, %1 \n\t"
"add $8, %2 \n\t"
"pxor %%mm6, %%mm0 \n\t"
"pxor %%mm6, %%mm1 \n\t"
PAVGB" %%mm1, %%mm0 \n\t"
"pxor %%mm6, %%mm0 \n\t"
"movq %%mm0, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"decl %0 \n\t"
"1: \n\t"
"movq (%1), %%mm0 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
"movq (%1), %%mm1 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
"movq (%2), %%mm2 \n\t"
"movq 8(%2), %%mm3 \n\t"
"pxor %%mm6, %%mm0 \n\t"
......@@ -183,13 +183,13 @@ static void DEF(put_no_rnd_pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src
"pxor %%mm6, %%mm0 \n\t"
"pxor %%mm6, %%mm1 \n\t"
"movq %%mm0, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"movq %%mm1, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"movq (%1), %%mm0 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
"movq (%1), %%mm1 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
"movq 16(%2), %%mm2 \n\t"
"movq 24(%2), %%mm3 \n\t"
"pxor %%mm6, %%mm0 \n\t"
......@@ -201,10 +201,10 @@ static void DEF(put_no_rnd_pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src
"pxor %%mm6, %%mm0 \n\t"
"pxor %%mm6, %%mm1 \n\t"
"movq %%mm0, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"movq %%mm1, (%3) \n\t"
"addl %5, %3 \n\t"
"addl $32, %2 \n\t"
"add %5, %3 \n\t"
"add $32, %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
#ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used
......@@ -212,7 +212,7 @@ static void DEF(put_no_rnd_pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src
#else
:"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)
#endif
:"S"(src1Stride), "D"(dstStride)
:"S"((long)src1Stride), "D"((long)dstStride)
:"memory");
//the following should be used, though better not with gcc ...
/* :"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst)
......@@ -227,39 +227,39 @@ static void DEF(avg_pixels4_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int
" jz 1f \n\t"
"movd (%1), %%mm0 \n\t"
"movd (%2), %%mm1 \n\t"
"addl %4, %1 \n\t"
"addl $4, %2 \n\t"
"add %4, %1 \n\t"
"add $4, %2 \n\t"
PAVGB" %%mm1, %%mm0 \n\t"
PAVGB" (%3), %%mm0 \n\t"
"movd %%mm0, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"decl %0 \n\t"
"1: \n\t"
"movd (%1), %%mm0 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
"movd (%1), %%mm1 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
PAVGB" (%2), %%mm0 \n\t"
PAVGB" 4(%2), %%mm1 \n\t"
PAVGB" (%3), %%mm0 \n\t"
"movd %%mm0, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
PAVGB" (%3), %%mm1 \n\t"
"movd %%mm1, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"movd (%1), %%mm0 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
"movd (%1), %%mm1 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
PAVGB" 8(%2), %%mm0 \n\t"
PAVGB" 12(%2), %%mm1 \n\t"
PAVGB" (%3), %%mm0 \n\t"
"movd %%mm0, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
PAVGB" (%3), %%mm1 \n\t"
"movd %%mm1, (%3) \n\t"
"addl %5, %3 \n\t"
"addl $16, %2 \n\t"
"add %5, %3 \n\t"
"add $16, %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
#ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used
......@@ -267,7 +267,7 @@ static void DEF(avg_pixels4_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int
#else
:"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)
#endif
:"S"(src1Stride), "D"(dstStride)
:"S"((long)src1Stride), "D"((long)dstStride)
:"memory");
}
......@@ -279,39 +279,39 @@ static void DEF(avg_pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int
" jz 1f \n\t"
"movq (%1), %%mm0 \n\t"
"movq (%2), %%mm1 \n\t"
"addl %4, %1 \n\t"
"addl $8, %2 \n\t"
"add %4, %1 \n\t"
"add $8, %2 \n\t"
PAVGB" %%mm1, %%mm0 \n\t"
PAVGB" (%3), %%mm0 \n\t"
"movq %%mm0, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"decl %0 \n\t"
"1: \n\t"
"movq (%1), %%mm0 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
"movq (%1), %%mm1 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
PAVGB" (%2), %%mm0 \n\t"
PAVGB" 8(%2), %%mm1 \n\t"
PAVGB" (%3), %%mm0 \n\t"
"movq %%mm0, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
PAVGB" (%3), %%mm1 \n\t"
"movq %%mm1, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"movq (%1), %%mm0 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
"movq (%1), %%mm1 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
PAVGB" 16(%2), %%mm0 \n\t"
PAVGB" 24(%2), %%mm1 \n\t"
PAVGB" (%3), %%mm0 \n\t"
"movq %%mm0, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
PAVGB" (%3), %%mm1 \n\t"
"movq %%mm1, (%3) \n\t"
"addl %5, %3 \n\t"
"addl $32, %2 \n\t"
"add %5, %3 \n\t"
"add $32, %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
#ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used
......@@ -319,7 +319,7 @@ static void DEF(avg_pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int
#else
:"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)
#endif
:"S"(src1Stride), "D"(dstStride)
:"S"((long)src1Stride), "D"((long)dstStride)
:"memory");
//the following should be used, though better not with gcc ...
/* :"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst)
......@@ -330,7 +330,7 @@ static void DEF(avg_pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int
static void DEF(put_pixels16_x2)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
{
__asm __volatile(
"lea (%3, %3), %%eax \n\t"
"lea (%3, %3), %%"REG_a" \n\t"
"1: \n\t"
"movq (%1), %%mm0 \n\t"
"movq (%1, %3), %%mm1 \n\t"
......@@ -344,8 +344,8 @@ static void DEF(put_pixels16_x2)(uint8_t *block, const uint8_t *pixels, int line
"movq %%mm1, (%2, %3) \n\t"
"movq %%mm2, 8(%2) \n\t"
"movq %%mm3, 8(%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %1 \n\t"
"add %%"REG_a", %2 \n\t"
"movq (%1), %%mm0 \n\t"
"movq (%1, %3), %%mm1 \n\t"
"movq 8(%1), %%mm2 \n\t"
......@@ -354,17 +354,17 @@ static void DEF(put_pixels16_x2)(uint8_t *block, const uint8_t *pixels, int line
PAVGB" 1(%1, %3), %%mm1 \n\t"
PAVGB" 9(%1), %%mm2 \n\t"
PAVGB" 9(%1, %3), %%mm3 \n\t"
"addl %%eax, %1 \n\t"
"add %%"REG_a", %1 \n\t"
"movq %%mm0, (%2) \n\t"
"movq %%mm1, (%2, %3) \n\t"
"movq %%mm2, 8(%2) \n\t"
"movq %%mm3, 8(%2, %3) \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
:"+g"(h), "+S"(pixels), "+D"(block)
:"r" (line_size)
:"%eax", "memory");
:"r" ((long)line_size)
:"%"REG_a, "memory");
}
static void DEF(put_pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h)
......@@ -376,30 +376,30 @@ static void DEF(put_pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int
"movq 8(%1), %%mm1 \n\t"
PAVGB" (%2), %%mm0 \n\t"
PAVGB" 8(%2), %%mm1 \n\t"
"addl %4, %1 \n\t"
"addl $16, %2 \n\t"
"add %4, %1 \n\t"
"add $16, %2 \n\t"
"movq %%mm0, (%3) \n\t"
"movq %%mm1, 8(%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"decl %0 \n\t"
"1: \n\t"
"movq (%1), %%mm0 \n\t"
"movq 8(%1), %%mm1 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
PAVGB" (%2), %%mm0 \n\t"
PAVGB" 8(%2), %%mm1 \n\t"
"movq %%mm0, (%3) \n\t"
"movq %%mm1, 8(%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"movq (%1), %%mm0 \n\t"
"movq 8(%1), %%mm1 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
PAVGB" 16(%2), %%mm0 \n\t"
PAVGB" 24(%2), %%mm1 \n\t"
"movq %%mm0, (%3) \n\t"
"movq %%mm1, 8(%3) \n\t"
"addl %5, %3 \n\t"
"addl $32, %2 \n\t"
"add %5, %3 \n\t"
"add $32, %2 \n\t"
"subl $2, %0 \n\t"
"jnz 1b \n\t"
#ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used
......@@ -407,7 +407,7 @@ static void DEF(put_pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int
#else
:"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)
#endif
:"S"(src1Stride), "D"(dstStride)
:"S"((long)src1Stride), "D"((long)dstStride)
:"memory");
//the following should be used, though better not with gcc ...
/* :"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst)
......@@ -424,36 +424,36 @@ static void DEF(avg_pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int
"movq 8(%1), %%mm1 \n\t"
PAVGB" (%2), %%mm0 \n\t"
PAVGB" 8(%2), %%mm1 \n\t"
"addl %4, %1 \n\t"
"addl $16, %2 \n\t"
"add %4, %1 \n\t"
"add $16, %2 \n\t"
PAVGB" (%3), %%mm0 \n\t"
PAVGB" 8(%3), %%mm1 \n\t"
"movq %%mm0, (%3) \n\t"
"movq %%mm1, 8(%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"decl %0 \n\t"
"1: \n\t"
"movq (%1), %%mm0 \n\t"
"movq 8(%1), %%mm1 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
PAVGB" (%2), %%mm0 \n\t"
PAVGB" 8(%2), %%mm1 \n\t"
PAVGB" (%3), %%mm0 \n\t"
PAVGB" 8(%3), %%mm1 \n\t"
"movq %%mm0, (%3) \n\t"
"movq %%mm1, 8(%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"movq (%1), %%mm0 \n\t"
"movq 8(%1), %%mm1 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
PAVGB" 16(%2), %%mm0 \n\t"
PAVGB" 24(%2), %%mm1 \n\t"
PAVGB" (%3), %%mm0 \n\t"
PAVGB" 8(%3), %%mm1 \n\t"
"movq %%mm0, (%3) \n\t"
"movq %%mm1, 8(%3) \n\t"
"addl %5, %3 \n\t"
"addl $32, %2 \n\t"
"add %5, %3 \n\t"
"add $32, %2 \n\t"
"subl $2, %0 \n\t"
"jnz 1b \n\t"
#ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used
......@@ -461,7 +461,7 @@ static void DEF(avg_pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int
#else
:"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)
#endif
:"S"(src1Stride), "D"(dstStride)
:"S"((long)src1Stride), "D"((long)dstStride)
:"memory");
//the following should be used, though better not with gcc ...
/* :"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst)
......@@ -487,16 +487,16 @@ static void DEF(put_no_rnd_pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *sr
PAVGB" %%mm3, %%mm1 \n\t"
"pxor %%mm6, %%mm0 \n\t"
"pxor %%mm6, %%mm1 \n\t"
"addl %4, %1 \n\t"
"addl $16, %2 \n\t"
"add %4, %1 \n\t"
"add $16, %2 \n\t"
"movq %%mm0, (%3) \n\t"
"movq %%mm1, 8(%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"decl %0 \n\t"
"1: \n\t"
"movq (%1), %%mm0 \n\t"
"movq 8(%1), %%mm1 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
"movq (%2), %%mm2 \n\t"
"movq 8(%2), %%mm3 \n\t"
"pxor %%mm6, %%mm0 \n\t"
......@@ -509,10 +509,10 @@ static void DEF(put_no_rnd_pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *sr
"pxor %%mm6, %%mm1 \n\t"
"movq %%mm0, (%3) \n\t"
"movq %%mm1, 8(%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"movq (%1), %%mm0 \n\t"
"movq 8(%1), %%mm1 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
"movq 16(%2), %%mm2 \n\t"
"movq 24(%2), %%mm3 \n\t"
"pxor %%mm6, %%mm0 \n\t"
......@@ -525,8 +525,8 @@ static void DEF(put_no_rnd_pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *sr
"pxor %%mm6, %%mm1 \n\t"
"movq %%mm0, (%3) \n\t"
"movq %%mm1, 8(%3) \n\t"
"addl %5, %3 \n\t"
"addl $32, %2 \n\t"
"add %5, %3 \n\t"
"add $32, %2 \n\t"
"subl $2, %0 \n\t"
"jnz 1b \n\t"
#ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used
......@@ -534,7 +534,7 @@ static void DEF(put_no_rnd_pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *sr
#else
:"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)
#endif
:"S"(src1Stride), "D"(dstStride)
:"S"((long)src1Stride), "D"((long)dstStride)
:"memory");
//the following should be used, though better not with gcc ...
/* :"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst)
......@@ -547,13 +547,13 @@ static void DEF(put_no_rnd_pixels8_x2)(uint8_t *block, const uint8_t *pixels, in
{
MOVQ_BONE(mm6);
__asm __volatile(
"lea (%3, %3), %%eax \n\t"
"lea (%3, %3), %%"REG_a" \n\t"
"1: \n\t"
"movq (%1), %%mm0 \n\t"
"movq (%1, %3), %%mm2 \n\t"
"movq 1(%1), %%mm1 \n\t"
"movq 1(%1, %3), %%mm3 \n\t"
"addl %%eax, %1 \n\t"
"add %%"REG_a", %1 \n\t"
"psubusb %%mm6, %%mm0 \n\t"
"psubusb %%mm6, %%mm2 \n\t"
PAVGB" %%mm1, %%mm0 \n\t"
......@@ -564,50 +564,50 @@ static void DEF(put_no_rnd_pixels8_x2)(uint8_t *block, const uint8_t *pixels, in
"movq 1(%1), %%mm1 \n\t"
"movq (%1, %3), %%mm2 \n\t"
"movq 1(%1, %3), %%mm3 \n\t"
"addl %%eax, %2 \n\t"
"addl %%eax, %1 \n\t"
"add %%"REG_a", %2 \n\t"
"add %%"REG_a", %1 \n\t"
"psubusb %%mm6, %%mm0 \n\t"
"psubusb %%mm6, %%mm2 \n\t"
PAVGB" %%mm1, %%mm0 \n\t"
PAVGB" %%mm3, %%mm2 \n\t"
"movq %%mm0, (%2) \n\t"
"movq %%mm2, (%2, %3) \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
:"+g"(h), "+S"(pixels), "+D"(block)
:"r" (line_size)
:"%eax", "memory");
:"r" ((long)line_size)
:"%"REG_a, "memory");
}
static void DEF(put_pixels8_y2)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
{
__asm __volatile(
"lea (%3, %3), %%eax \n\t"
"lea (%3, %3), %%"REG_a" \n\t"
"movq (%1), %%mm0 \n\t"
"subl %3, %2 \n\t"
"sub %3, %2 \n\t"
"1: \n\t"
"movq (%1, %3), %%mm1 \n\t"
"movq (%1, %%eax), %%mm2 \n\t"
"addl %%eax, %1 \n\t"
"movq (%1, %%"REG_a"), %%mm2 \n\t"
"add %%"REG_a", %1 \n\t"
PAVGB" %%mm1, %%mm0 \n\t"
PAVGB" %%mm2, %%mm1 \n\t"
"movq %%mm0, (%2, %3) \n\t"
"movq %%mm1, (%2, %%eax) \n\t"
"movq %%mm1, (%2, %%"REG_a") \n\t"
"movq (%1, %3), %%mm1 \n\t"
"movq (%1, %%eax), %%mm0 \n\t"
"addl %%eax, %2 \n\t"
"addl %%eax, %1 \n\t"
"movq (%1, %%"REG_a"), %%mm0 \n\t"
"add %%"REG_a", %2 \n\t"
"add %%"REG_a", %1 \n\t"
PAVGB" %%mm1, %%mm2 \n\t"
PAVGB" %%mm0, %%mm1 \n\t"
"movq %%mm2, (%2, %3) \n\t"
"movq %%mm1, (%2, %%eax) \n\t"
"addl %%eax, %2 \n\t"
"movq %%mm1, (%2, %%"REG_a") \n\t"
"add %%"REG_a", %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
:"+g"(h), "+S"(pixels), "+D" (block)
:"r" (line_size)
:"%eax", "memory");
:"r" ((long)line_size)
:"%"REG_a, "memory");
}
/* GL: this function does incorrect rounding if overflow */
......@@ -615,39 +615,39 @@ static void DEF(put_no_rnd_pixels8_y2)(uint8_t *block, const uint8_t *pixels, in
{
MOVQ_BONE(mm6);
__asm __volatile(
"lea (%3, %3), %%eax \n\t"
"lea (%3, %3), %%"REG_a" \n\t"
"movq (%1), %%mm0 \n\t"
"subl %3, %2 \n\t"
"sub %3, %2 \n\t"
"1: \n\t"
"movq (%1, %3), %%mm1 \n\t"
"movq (%1, %%eax), %%mm2 \n\t"
"addl %%eax, %1 \n\t"
"movq (%1, %%"REG_a"), %%mm2 \n\t"
"add %%"REG_a", %1 \n\t"
"psubusb %%mm6, %%mm1 \n\t"
PAVGB" %%mm1, %%mm0 \n\t"
PAVGB" %%mm2, %%mm1 \n\t"
"movq %%mm0, (%2, %3) \n\t"
"movq %%mm1, (%2, %%eax) \n\t"
"movq %%mm1, (%2, %%"REG_a") \n\t"
"movq (%1, %3), %%mm1 \n\t"
"movq (%1, %%eax), %%mm0 \n\t"
"addl %%eax, %2 \n\t"
"addl %%eax, %1 \n\t"
"movq (%1, %%"REG_a"), %%mm0 \n\t"
"add %%"REG_a", %2 \n\t"
"add %%"REG_a", %1 \n\t"
"psubusb %%mm6, %%mm1 \n\t"
PAVGB" %%mm1, %%mm2 \n\t"
PAVGB" %%mm0, %%mm1 \n\t"
"movq %%mm2, (%2, %3) \n\t"
"movq %%mm1, (%2, %%eax) \n\t"
"addl %%eax, %2 \n\t"
"movq %%mm1, (%2, %%"REG_a") \n\t"
"add %%"REG_a", %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
:"+g"(h), "+S"(pixels), "+D" (block)
:"r" (line_size)
:"%eax", "memory");
:"r" ((long)line_size)
:"%"REG_a, "memory");
}
static void DEF(avg_pixels8)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
{
__asm __volatile(
"lea (%3, %3), %%eax \n\t"
"lea (%3, %3), %%"REG_a" \n\t"
"1: \n\t"
"movq (%2), %%mm0 \n\t"
"movq (%2, %3), %%mm1 \n\t"
......@@ -655,27 +655,27 @@ static void DEF(avg_pixels8)(uint8_t *block, const uint8_t *pixels, int line_siz
PAVGB" (%1, %3), %%mm1 \n\t"
"movq %%mm0, (%2) \n\t"
"movq %%mm1, (%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %1 \n\t"
"add %%"REG_a", %2 \n\t"
"movq (%2), %%mm0 \n\t"
"movq (%2, %3), %%mm1 \n\t"
PAVGB" (%1), %%mm0 \n\t"
PAVGB" (%1, %3), %%mm1 \n\t"
"addl %%eax, %1 \n\t"
"add %%"REG_a", %1 \n\t"
"movq %%mm0, (%2) \n\t"
"movq %%mm1, (%2, %3) \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
:"+g"(h), "+S"(pixels), "+D"(block)
:"r" (line_size)
:"%eax", "memory");
:"r" ((long)line_size)
:"%"REG_a, "memory");
}
static void DEF(avg_pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
{
__asm __volatile(
"lea (%3, %3), %%eax \n\t"
"lea (%3, %3), %%"REG_a" \n\t"
"1: \n\t"
"movq (%1), %%mm0 \n\t"
"movq (%1, %3), %%mm2 \n\t"
......@@ -683,63 +683,63 @@ static void DEF(avg_pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line_
PAVGB" 1(%1, %3), %%mm2 \n\t"
PAVGB" (%2), %%mm0 \n\t"
PAVGB" (%2, %3), %%mm2 \n\t"
"addl %%eax, %1 \n\t"
"add %%"REG_a", %1 \n\t"
"movq %%mm0, (%2) \n\t"
"movq %%mm2, (%2, %3) \n\t"
"movq (%1), %%mm0 \n\t"
"movq (%1, %3), %%mm2 \n\t"
PAVGB" 1(%1), %%mm0 \n\t"
PAVGB" 1(%1, %3), %%mm2 \n\t"
"addl %%eax, %2 \n\t"
"addl %%eax, %1 \n\t"
"add %%"REG_a", %2 \n\t"
"add %%"REG_a", %1 \n\t"
PAVGB" (%2), %%mm0 \n\t"
PAVGB" (%2, %3), %%mm2 \n\t"
"movq %%mm0, (%2) \n\t"
"movq %%mm2, (%2, %3) \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
:"+g"(h), "+S"(pixels), "+D"(block)
:"r" (line_size)
:"%eax", "memory");
:"r" ((long)line_size)
:"%"REG_a, "memory");
}
static void DEF(avg_pixels8_y2)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
{
__asm __volatile(
"lea (%3, %3), %%eax \n\t"
"lea (%3, %3), %%"REG_a" \n\t"
"movq (%1), %%mm0 \n\t"
"subl %3, %2 \n\t"
"sub %3, %2 \n\t"
"1: \n\t"
"movq (%1, %3), %%mm1 \n\t"
"movq (%1, %%eax), %%mm2 \n\t"
"addl %%eax, %1 \n\t"
"movq (%1, %%"REG_a"), %%mm2 \n\t"
"add %%"REG_a", %1 \n\t"
PAVGB" %%mm1, %%mm0 \n\t"
PAVGB" %%mm2, %%mm1 \n\t"
"movq (%2, %3), %%mm3 \n\t"
"movq (%2, %%eax), %%mm4 \n\t"
"movq (%2, %%"REG_a"), %%mm4 \n\t"
PAVGB" %%mm3, %%mm0 \n\t"
PAVGB" %%mm4, %%mm1 \n\t"
"movq %%mm0, (%2, %3) \n\t"
"movq %%mm1, (%2, %%eax) \n\t"
"movq %%mm1, (%2, %%"REG_a") \n\t"
"movq (%1, %3), %%mm1 \n\t"
"movq (%1, %%eax), %%mm0 \n\t"
"movq (%1, %%"REG_a"), %%mm0 \n\t"
PAVGB" %%mm1, %%mm2 \n\t"
PAVGB" %%mm0, %%mm1 \n\t"
"addl %%eax, %2 \n\t"
"addl %%eax, %1 \n\t"
"add %%"REG_a", %2 \n\t"
"add %%"REG_a", %1 \n\t"
"movq (%2, %3), %%mm3 \n\t"
"movq (%2, %%eax), %%mm4 \n\t"
"movq (%2, %%"REG_a"), %%mm4 \n\t"
PAVGB" %%mm3, %%mm2 \n\t"
PAVGB" %%mm4, %%mm1 \n\t"
"movq %%mm2, (%2, %3) \n\t"
"movq %%mm1, (%2, %%eax) \n\t"
"addl %%eax, %2 \n\t"
"movq %%mm1, (%2, %%"REG_a") \n\t"
"add %%"REG_a", %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
:"+g"(h), "+S"(pixels), "+D"(block)
:"r" (line_size)
:"%eax", "memory");
:"r" ((long)line_size)
:"%"REG_a, "memory");
}
// Note this is not correctly rounded, but this function is only used for b frames so it doesnt matter
......@@ -747,17 +747,17 @@ static void DEF(avg_pixels8_xy2)(uint8_t *block, const uint8_t *pixels, int line
{
MOVQ_BONE(mm6);
__asm __volatile(
"lea (%3, %3), %%eax \n\t"
"lea (%3, %3), %%"REG_a" \n\t"
"movq (%1), %%mm0 \n\t"
PAVGB" 1(%1), %%mm0 \n\t"
".balign 8 \n\t"
"1: \n\t"
"movq (%1, %%eax), %%mm2 \n\t"
"movq (%1, %%"REG_a"), %%mm2 \n\t"
"movq (%1, %3), %%mm1 \n\t"
"psubusb %%mm6, %%mm2 \n\t"
PAVGB" 1(%1, %3), %%mm1 \n\t"
PAVGB" 1(%1, %%eax), %%mm2 \n\t"
"addl %%eax, %1 \n\t"
PAVGB" 1(%1, %%"REG_a"), %%mm2 \n\t"
"add %%"REG_a", %1 \n\t"
PAVGB" %%mm1, %%mm0 \n\t"
PAVGB" %%mm2, %%mm1 \n\t"
PAVGB" (%2), %%mm0 \n\t"
......@@ -765,23 +765,23 @@ static void DEF(avg_pixels8_xy2)(uint8_t *block, const uint8_t *pixels, int line
"movq %%mm0, (%2) \n\t"
"movq %%mm1, (%2, %3) \n\t"
"movq (%1, %3), %%mm1 \n\t"
"movq (%1, %%eax), %%mm0 \n\t"
"movq (%1, %%"REG_a"), %%mm0 \n\t"
PAVGB" 1(%1, %3), %%mm1 \n\t"
PAVGB" 1(%1, %%eax), %%mm0 \n\t"
"addl %%eax, %2 \n\t"
"addl %%eax, %1 \n\t"
PAVGB" 1(%1, %%"REG_a"), %%mm0 \n\t"
"add %%"REG_a", %2 \n\t"
"add %%"REG_a", %1 \n\t"
PAVGB" %%mm1, %%mm2 \n\t"
PAVGB" %%mm0, %%mm1 \n\t"
PAVGB" (%2), %%mm2 \n\t"
PAVGB" (%2, %3), %%mm1 \n\t"
"movq %%mm2, (%2) \n\t"
"movq %%mm1, (%2, %3) \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
:"+g"(h), "+S"(pixels), "+D"(block)
:"r" (line_size)
:"%eax", "memory");
:"r" ((long)line_size)
:"%"REG_a, "memory");
}
//FIXME the following could be optimized too ...
......
libavcodec/i386/dsputil_mmx_rnd.h
View file @ 09fcf89b
......@@ -27,7 +27,7 @@ static void DEF(put, pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line
{
MOVQ_BFE(mm6);
__asm __volatile(
"lea (%3, %3), %%eax \n\t"
"lea (%3, %3), %%"REG_a" \n\t"
".balign 8 \n\t"
"1: \n\t"
"movq (%1), %%mm0 \n\t"
......@@ -37,8 +37,8 @@ static void DEF(put, pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line
PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5)
"movq %%mm4, (%2) \n\t"
"movq %%mm5, (%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %1 \n\t"
"add %%"REG_a", %2 \n\t"
"movq (%1), %%mm0 \n\t"
"movq 1(%1), %%mm1 \n\t"
"movq (%1, %3), %%mm2 \n\t"
......@@ -46,13 +46,13 @@ static void DEF(put, pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line
PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5)
"movq %%mm4, (%2) \n\t"
"movq %%mm5, (%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %1 \n\t"
"add %%"REG_a", %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
:"+g"(h), "+S"(pixels), "+D"(block)
:"r"(line_size)
:"eax", "memory");
:"r"((long)line_size)
:REG_a, "memory");
}
static void DEF(put, pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h)
......@@ -63,37 +63,37 @@ static void DEF(put, pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int
" jz 1f \n\t"
"movq (%1), %%mm0 \n\t"
"movq (%2), %%mm1 \n\t"
"addl %4, %1 \n\t"
"addl $8, %2 \n\t"
"add %4, %1 \n\t"
"add $8, %2 \n\t"
PAVGB(%%mm0, %%mm1, %%mm4, %%mm6)
"movq %%mm4, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"decl %0 \n\t"
".balign 8 \n\t"
"1: \n\t"
"movq (%1), %%mm0 \n\t"
"movq (%2), %%mm1 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
"movq (%1), %%mm2 \n\t"
"movq 8(%2), %%mm3 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5)
"movq %%mm4, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"movq %%mm5, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"movq (%1), %%mm0 \n\t"
"movq 16(%2), %%mm1 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
"movq (%1), %%mm2 \n\t"
"movq 24(%2), %%mm3 \n\t"
"addl %4, %1 \n\t"
"addl $32, %2 \n\t"
"add %4, %1 \n\t"
"add $32, %2 \n\t"
PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5)
"movq %%mm4, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"movq %%mm5, (%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
#ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used
......@@ -101,7 +101,7 @@ static void DEF(put, pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int
#else
:"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)
#endif
:"S"(src1Stride), "D"(dstStride)
:"S"((long)src1Stride), "D"((long)dstStride)
:"memory");
}
......@@ -109,7 +109,7 @@ static void DEF(put, pixels16_x2)(uint8_t *block, const uint8_t *pixels, int lin
{
MOVQ_BFE(mm6);
__asm __volatile(
"lea (%3, %3), %%eax \n\t"
"lea (%3, %3), %%"REG_a" \n\t"
".balign 8 \n\t"
"1: \n\t"
"movq (%1), %%mm0 \n\t"
......@@ -126,8 +126,8 @@ static void DEF(put, pixels16_x2)(uint8_t *block, const uint8_t *pixels, int lin
PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5)
"movq %%mm4, 8(%2) \n\t"
"movq %%mm5, 8(%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %1 \n\t"
"add %%"REG_a", %2 \n\t"
"movq (%1), %%mm0 \n\t"
"movq 1(%1), %%mm1 \n\t"
"movq (%1, %3), %%mm2 \n\t"
......@@ -142,13 +142,13 @@ static void DEF(put, pixels16_x2)(uint8_t *block, const uint8_t *pixels, int lin
PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5)
"movq %%mm4, 8(%2) \n\t"
"movq %%mm5, 8(%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %1 \n\t"
"add %%"REG_a", %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
:"+g"(h), "+S"(pixels), "+D"(block)
:"r"(line_size)
:"eax", "memory");
:"r"((long)line_size)
:REG_a, "memory");
}
static void DEF(put, pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h)
......@@ -161,12 +161,12 @@ static void DEF(put, pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, in
"movq (%2), %%mm1 \n\t"
"movq 8(%1), %%mm2 \n\t"
"movq 8(%2), %%mm3 \n\t"
"addl %4, %1 \n\t"
"addl $16, %2 \n\t"
"add %4, %1 \n\t"
"add $16, %2 \n\t"
PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5)
"movq %%mm4, (%3) \n\t"
"movq %%mm5, 8(%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"decl %0 \n\t"
".balign 8 \n\t"
"1: \n\t"
......@@ -174,21 +174,21 @@ static void DEF(put, pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, in
"movq (%2), %%mm1 \n\t"
"movq 8(%1), %%mm2 \n\t"
"movq 8(%2), %%mm3 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5)
"movq %%mm4, (%3) \n\t"
"movq %%mm5, 8(%3) \n\t"
"addl %5, %3 \n\t"
"add %5, %3 \n\t"
"movq (%1), %%mm0 \n\t"
"movq 16(%2), %%mm1 \n\t"
"movq 8(%1), %%mm2 \n\t"
"movq 24(%2), %%mm3 \n\t"
"addl %4, %1 \n\t"
"add %4, %1 \n\t"
PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5)
"movq %%mm4, (%3) \n\t"
"movq %%mm5, 8(%3) \n\t"
"addl %5, %3 \n\t"
"addl $32, %2 \n\t"
"add %5, %3 \n\t"
"add $32, %2 \n\t"
"subl $2, %0 \n\t"
"jnz 1b \n\t"
#ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used
......@@ -196,7 +196,7 @@ static void DEF(put, pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, in
#else
:"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)
#endif
:"S"(src1Stride), "D"(dstStride)
:"S"((long)src1Stride), "D"((long)dstStride)
:"memory");
}
......@@ -204,29 +204,29 @@ static void DEF(put, pixels8_y2)(uint8_t *block, const uint8_t *pixels, int line
{
MOVQ_BFE(mm6);
__asm __volatile(
"lea (%3, %3), %%eax \n\t"
"lea (%3, %3), %%"REG_a" \n\t"
"movq (%1), %%mm0 \n\t"
".balign 8 \n\t"
"1: \n\t"
"movq (%1, %3), %%mm1 \n\t"
"movq (%1, %%eax),%%mm2 \n\t"
"movq (%1, %%"REG_a"),%%mm2 \n\t"
PAVGBP(%%mm1, %%mm0, %%mm4, %%mm2, %%mm1, %%mm5)
"movq %%mm4, (%2) \n\t"
"movq %%mm5, (%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %1 \n\t"
"add %%"REG_a", %2 \n\t"
"movq (%1, %3), %%mm1 \n\t"
"movq (%1, %%eax),%%mm0 \n\t"
"movq (%1, %%"REG_a"),%%mm0 \n\t"
PAVGBP(%%mm1, %%mm2, %%mm4, %%mm0, %%mm1, %%mm5)
"movq %%mm4, (%2) \n\t"
"movq %%mm5, (%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %1 \n\t"
"add %%"REG_a", %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
:"+g"(h), "+S"(pixels), "+D"(block)
:"r"(line_size)
:"eax", "memory");
:"r"((long)line_size)
:REG_a, "memory");
}
static void DEF(put, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
......@@ -244,12 +244,12 @@ static void DEF(put, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, int lin
"punpckhbw %%mm7, %%mm5 \n\t"
"paddusw %%mm0, %%mm4 \n\t"
"paddusw %%mm1, %%mm5 \n\t"
"xorl %%eax, %%eax \n\t"
"addl %3, %1 \n\t"
"xor %%"REG_a", %%"REG_a" \n\t"
"add %3, %1 \n\t"
".balign 8 \n\t"
"1: \n\t"
"movq (%1, %%eax), %%mm0 \n\t"
"movq 1(%1, %%eax), %%mm2 \n\t"
"movq (%1, %%"REG_a"), %%mm0 \n\t"
"movq 1(%1, %%"REG_a"), %%mm2 \n\t"
"movq %%mm0, %%mm1 \n\t"
"movq %%mm2, %%mm3 \n\t"
"punpcklbw %%mm7, %%mm0 \n\t"
......@@ -265,11 +265,11 @@ static void DEF(put, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, int lin
"psrlw $2, %%mm4 \n\t"
"psrlw $2, %%mm5 \n\t"
"packuswb %%mm5, %%mm4 \n\t"
"movq %%mm4, (%2, %%eax) \n\t"
"addl %3, %%eax \n\t"
"movq %%mm4, (%2, %%"REG_a") \n\t"
"add %3, %%"REG_a" \n\t"
"movq (%1, %%eax), %%mm2 \n\t" // 0 <-> 2 1 <-> 3
"movq 1(%1, %%eax), %%mm4 \n\t"
"movq (%1, %%"REG_a"), %%mm2 \n\t" // 0 <-> 2 1 <-> 3
"movq 1(%1, %%"REG_a"), %%mm4 \n\t"
"movq %%mm2, %%mm3 \n\t"
"movq %%mm4, %%mm5 \n\t"
"punpcklbw %%mm7, %%mm2 \n\t"
......@@ -285,14 +285,14 @@ static void DEF(put, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, int lin
"psrlw $2, %%mm0 \n\t"
"psrlw $2, %%mm1 \n\t"
"packuswb %%mm1, %%mm0 \n\t"
"movq %%mm0, (%2, %%eax) \n\t"
"addl %3, %%eax \n\t"
"movq %%mm0, (%2, %%"REG_a") \n\t"
"add %3, %%"REG_a" \n\t"
"subl $2, %0 \n\t"
"jnz 1b \n\t"
:"+g"(h), "+S"(pixels)
:"D"(block), "r"(line_size)
:"eax", "memory");
:"D"(block), "r"((long)line_size)
:REG_a, "memory");
}
// avg_pixels
......@@ -456,12 +456,12 @@ static void DEF(avg, pixels8_y2)(uint8_t *block, const uint8_t *pixels, int line
{
MOVQ_BFE(mm6);
__asm __volatile(
"lea (%3, %3), %%eax \n\t"
"lea (%3, %3), %%"REG_a" \n\t"
"movq (%1), %%mm0 \n\t"
".balign 8 \n\t"
"1: \n\t"
"movq (%1, %3), %%mm1 \n\t"
"movq (%1, %%eax), %%mm2 \n\t"
"movq (%1, %%"REG_a"), %%mm2 \n\t"
PAVGBP(%%mm1, %%mm0, %%mm4, %%mm2, %%mm1, %%mm5)
"movq (%2), %%mm3 \n\t"
PAVGB(%%mm3, %%mm4, %%mm0, %%mm6)
......@@ -469,11 +469,11 @@ static void DEF(avg, pixels8_y2)(uint8_t *block, const uint8_t *pixels, int line
PAVGB(%%mm3, %%mm5, %%mm1, %%mm6)
"movq %%mm0, (%2) \n\t"
"movq %%mm1, (%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %1 \n\t"
"add %%"REG_a", %2 \n\t"
"movq (%1, %3), %%mm1 \n\t"
"movq (%1, %%eax), %%mm0 \n\t"
"movq (%1, %%"REG_a"), %%mm0 \n\t"
PAVGBP(%%mm1, %%mm2, %%mm4, %%mm0, %%mm1, %%mm5)
"movq (%2), %%mm3 \n\t"
PAVGB(%%mm3, %%mm4, %%mm2, %%mm6)
......@@ -481,14 +481,14 @@ static void DEF(avg, pixels8_y2)(uint8_t *block, const uint8_t *pixels, int line
PAVGB(%%mm3, %%mm5, %%mm1, %%mm6)
"movq %%mm2, (%2) \n\t"
"movq %%mm1, (%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"add %%"REG_a", %1 \n\t"
"add %%"REG_a", %2 \n\t"
"subl $4, %0 \n\t"
"jnz 1b \n\t"
:"+g"(h), "+S"(pixels), "+D"(block)
:"r"(line_size)
:"eax", "memory");
:"r"((long)line_size)
:REG_a, "memory");
}
// this routine is 'slightly' suboptimal but mostly unused
......@@ -507,12 +507,12 @@ static void DEF(avg, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, int lin
"punpckhbw %%mm7, %%mm5 \n\t"
"paddusw %%mm0, %%mm4 \n\t"
"paddusw %%mm1, %%mm5 \n\t"
"xorl %%eax, %%eax \n\t"
"addl %3, %1 \n\t"
"xor %%"REG_a", %%"REG_a" \n\t"
"add %3, %1 \n\t"
".balign 8 \n\t"
"1: \n\t"
"movq (%1, %%eax), %%mm0 \n\t"
"movq 1(%1, %%eax), %%mm2 \n\t"
"movq (%1, %%"REG_a"), %%mm0 \n\t"
"movq 1(%1, %%"REG_a"), %%mm2 \n\t"
"movq %%mm0, %%mm1 \n\t"
"movq %%mm2, %%mm3 \n\t"
"punpcklbw %%mm7, %%mm0 \n\t"
......@@ -527,16 +527,16 @@ static void DEF(avg, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, int lin
"paddusw %%mm1, %%mm5 \n\t"
"psrlw $2, %%mm4 \n\t"
"psrlw $2, %%mm5 \n\t"
"movq (%2, %%eax), %%mm3 \n\t"
"movq (%2, %%"REG_a"), %%mm3 \n\t"
"packuswb %%mm5, %%mm4 \n\t"
"pcmpeqd %%mm2, %%mm2 \n\t"
"paddb %%mm2, %%mm2 \n\t"
PAVGB(%%mm3, %%mm4, %%mm5, %%mm2)
"movq %%mm5, (%2, %%eax) \n\t"
"addl %3, %%eax \n\t"
"movq %%mm5, (%2, %%"REG_a") \n\t"
"add %3, %%"REG_a" \n\t"
"movq (%1, %%eax), %%mm2 \n\t" // 0 <-> 2 1 <-> 3
"movq 1(%1, %%eax), %%mm4 \n\t"
"movq (%1, %%"REG_a"), %%mm2 \n\t" // 0 <-> 2 1 <-> 3
"movq 1(%1, %%"REG_a"), %%mm4 \n\t"
"movq %%mm2, %%mm3 \n\t"
"movq %%mm4, %%mm5 \n\t"
"punpcklbw %%mm7, %%mm2 \n\t"
......@@ -551,19 +551,19 @@ static void DEF(avg, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, int lin
"paddusw %%mm5, %%mm1 \n\t"
"psrlw $2, %%mm0 \n\t"
"psrlw $2, %%mm1 \n\t"
"movq (%2, %%eax), %%mm3 \n\t"
"movq (%2, %%"REG_a"), %%mm3 \n\t"
"packuswb %%mm1, %%mm0 \n\t"
"pcmpeqd %%mm2, %%mm2 \n\t"
"paddb %%mm2, %%mm2 \n\t"
PAVGB(%%mm3, %%mm0, %%mm1, %%mm2)
"movq %%mm1, (%2, %%eax) \n\t"
"addl %3, %%eax \n\t"
"movq %%mm1, (%2, %%"REG_a") \n\t"
"add %3, %%"REG_a" \n\t"
"subl $2, %0 \n\t"
"jnz 1b \n\t"
:"+g"(h), "+S"(pixels)
:"D"(block), "r"(line_size)
:"eax", "memory");
:"D"(block), "r"((long)line_size)
:REG_a, "memory");
}
//FIXME optimize
......
libavcodec/i386/fdct_mmx.c
View file @ 09fcf89b
......@@ -47,13 +47,13 @@ static const int16_t ocos_4_16[4] ATTR_ALIGN(8) = {
23170, 23170, 23170, 23170, //cos * (2<<15) + 0.5
};
static const long long fdct_one_corr ATTR_ALIGN(8) = 0x0001000100010001LL;
static const int64_t fdct_one_corr ATTR_ALIGN(8) = 0x0001000100010001LL;
static const long fdct_r_row[2] ATTR_ALIGN(8) = {RND_FRW_ROW, RND_FRW_ROW };
static const int32_t fdct_r_row[2] ATTR_ALIGN(8) = {RND_FRW_ROW, RND_FRW_ROW };
struct
{
const long fdct_r_row_sse2[4] ATTR_ALIGN(16);
const int32_t fdct_r_row_sse2[4] ATTR_ALIGN(16);
} fdct_r_row_sse2 ATTR_ALIGN(16)=
{{
RND_FRW_ROW, RND_FRW_ROW, RND_FRW_ROW, RND_FRW_ROW
......
libavcodec/i386/mmx.h
View file @ 09fcf89b
......@@ -5,6 +5,12 @@
#ifndef AVCODEC_I386MMX_H
#define AVCODEC_I386MMX_H
#ifdef ARCH_X86_64
# define REG_a "rax"
#else
# define REG_a "eax"
#endif
/*
* The type of an value that fits in an MMX register (note that long
* long constant values MUST be suffixed by LL and unsigned long long
......
libavcodec/i386/motion_est_mmx.c
View file @ 09fcf89b
......@@ -20,6 +20,7 @@
* mostly by Michael Niedermayer <michaelni@gmx.at>
*/
#include "../dsputil.h"
#include "mmx.h"
static const __attribute__ ((aligned(8))) uint64_t round_tab[3]={
0x0000000000000000ULL,
......@@ -31,19 +32,19 @@ static attribute_used __attribute__ ((aligned(8))) uint64_t bone= 0x010101010101
static inline void sad8_1_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
{
int len= -(stride*h);
long len= -(stride*h);
asm volatile(
".balign 16 \n\t"
"1: \n\t"
"movq (%1, %%eax), %%mm0 \n\t"
"movq (%2, %%eax), %%mm2 \n\t"
"movq (%2, %%eax), %%mm4 \n\t"
"addl %3, %%eax \n\t"
"movq (%1, %%"REG_a"), %%mm0 \n\t"
"movq (%2, %%"REG_a"), %%mm2 \n\t"
"movq (%2, %%"REG_a"), %%mm4 \n\t"
"add %3, %%"REG_a" \n\t"
"psubusb %%mm0, %%mm2 \n\t"
"psubusb %%mm4, %%mm0 \n\t"
"movq (%1, %%eax), %%mm1 \n\t"
"movq (%2, %%eax), %%mm3 \n\t"
"movq (%2, %%eax), %%mm5 \n\t"
"movq (%1, %%"REG_a"), %%mm1 \n\t"
"movq (%2, %%"REG_a"), %%mm3 \n\t"
"movq (%2, %%"REG_a"), %%mm5 \n\t"
"psubusb %%mm1, %%mm3 \n\t"
"psubusb %%mm5, %%mm1 \n\t"
"por %%mm2, %%mm0 \n\t"
......@@ -58,116 +59,116 @@ static inline void sad8_1_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
"paddw %%mm3, %%mm2 \n\t"
"paddw %%mm2, %%mm0 \n\t"
"paddw %%mm0, %%mm6 \n\t"
"addl %3, %%eax \n\t"
"add %3, %%"REG_a" \n\t"
" js 1b \n\t"
: "+a" (len)
: "r" (blk1 - len), "r" (blk2 - len), "r" (stride)
: "r" (blk1 - len), "r" (blk2 - len), "r" ((long)stride)
);
}
static inline void sad8_1_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h)
{
int len= -(stride*h);
long len= -(stride*h);
asm volatile(
".balign 16 \n\t"
"1: \n\t"
"movq (%1, %%eax), %%mm0 \n\t"
"movq (%2, %%eax), %%mm2 \n\t"
"movq (%1, %%"REG_a"), %%mm0 \n\t"
"movq (%2, %%"REG_a"), %%mm2 \n\t"
"psadbw %%mm2, %%mm0 \n\t"
"addl %3, %%eax \n\t"
"movq (%1, %%eax), %%mm1 \n\t"
"movq (%2, %%eax), %%mm3 \n\t"
"add %3, %%"REG_a" \n\t"
"movq (%1, %%"REG_a"), %%mm1 \n\t"
"movq (%2, %%"REG_a"), %%mm3 \n\t"
"psadbw %%mm1, %%mm3 \n\t"
"paddw %%mm3, %%mm0 \n\t"
"paddw %%mm0, %%mm6 \n\t"
"addl %3, %%eax \n\t"
"add %3, %%"REG_a" \n\t"
" js 1b \n\t"
: "+a" (len)
: "r" (blk1 - len), "r" (blk2 - len), "r" (stride)
: "r" (blk1 - len), "r" (blk2 - len), "r" ((long)stride)
);
}
static inline void sad8_2_mmx2(uint8_t *blk1a, uint8_t *blk1b, uint8_t *blk2, int stride, int h)
{
int len= -(stride*h);
long len= -(stride*h);
asm volatile(
".balign 16 \n\t"
"1: \n\t"
"movq (%1, %%eax), %%mm0 \n\t"
"movq (%2, %%eax), %%mm2 \n\t"
"movq (%1, %%"REG_a"), %%mm0 \n\t"
"movq (%2, %%"REG_a"), %%mm2 \n\t"
"pavgb %%mm2, %%mm0 \n\t"
"movq (%3, %%eax), %%mm2 \n\t"
"movq (%3, %%"REG_a"), %%mm2 \n\t"
"psadbw %%mm2, %%mm0 \n\t"
"addl %4, %%eax \n\t"
"movq (%1, %%eax), %%mm1 \n\t"
"movq (%2, %%eax), %%mm3 \n\t"
"add %4, %%"REG_a" \n\t"
"movq (%1, %%"REG_a"), %%mm1 \n\t"
"movq (%2, %%"REG_a"), %%mm3 \n\t"
"pavgb %%mm1, %%mm3 \n\t"
"movq (%3, %%eax), %%mm1 \n\t"
"movq (%3, %%"REG_a"), %%mm1 \n\t"
"psadbw %%mm1, %%mm3 \n\t"
"paddw %%mm3, %%mm0 \n\t"
"paddw %%mm0, %%mm6 \n\t"
"addl %4, %%eax \n\t"
"add %4, %%"REG_a" \n\t"
" js 1b \n\t"
: "+a" (len)
: "r" (blk1a - len), "r" (blk1b -len), "r" (blk2 - len), "r" (stride)
: "r" (blk1a - len), "r" (blk1b -len), "r" (blk2 - len), "r" ((long)stride)
);
}
static inline void sad8_4_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h)
{ //FIXME reuse src
int len= -(stride*h);
long len= -(stride*h);
asm volatile(
".balign 16 \n\t"
"movq "MANGLE(bone)", %%mm5 \n\t"
"1: \n\t"
"movq (%1, %%eax), %%mm0 \n\t"
"movq (%2, %%eax), %%mm2 \n\t"
"movq 1(%1, %%eax), %%mm1 \n\t"
"movq 1(%2, %%eax), %%mm3 \n\t"
"movq (%1, %%"REG_a"), %%mm0 \n\t"
"movq (%2, %%"REG_a"), %%mm2 \n\t"
"movq 1(%1, %%"REG_a"), %%mm1 \n\t"
"movq 1(%2, %%"REG_a"), %%mm3 \n\t"
"pavgb %%mm2, %%mm0 \n\t"
"pavgb %%mm1, %%mm3 \n\t"
"psubusb %%mm5, %%mm3 \n\t"
"pavgb %%mm3, %%mm0 \n\t"
"movq (%3, %%eax), %%mm2 \n\t"
"movq (%3, %%"REG_a"), %%mm2 \n\t"
"psadbw %%mm2, %%mm0 \n\t"
"addl %4, %%eax \n\t"
"movq (%1, %%eax), %%mm1 \n\t"
"movq (%2, %%eax), %%mm3 \n\t"
"movq 1(%1, %%eax), %%mm2 \n\t"
"movq 1(%2, %%eax), %%mm4 \n\t"
"add %4, %%"REG_a" \n\t"
"movq (%1, %%"REG_a"), %%mm1 \n\t"
"movq (%2, %%"REG_a"), %%mm3 \n\t"
"movq 1(%1, %%"REG_a"), %%mm2 \n\t"
"movq 1(%2, %%"REG_a"), %%mm4 \n\t"
"pavgb %%mm3, %%mm1 \n\t"
"pavgb %%mm4, %%mm2 \n\t"
"psubusb %%mm5, %%mm2 \n\t"
"pavgb %%mm1, %%mm2 \n\t"
"movq (%3, %%eax), %%mm1 \n\t"
"movq (%3, %%"REG_a"), %%mm1 \n\t"
"psadbw %%mm1, %%mm2 \n\t"
"paddw %%mm2, %%mm0 \n\t"
"paddw %%mm0, %%mm6 \n\t"
"addl %4, %%eax \n\t"
"add %4, %%"REG_a" \n\t"
" js 1b \n\t"
: "+a" (len)
: "r" (blk1 - len), "r" (blk1 - len + stride), "r" (blk2 - len), "r" (stride)
: "r" (blk1 - len), "r" (blk1 - len + stride), "r" (blk2 - len), "r" ((long)stride)
);
}
static inline void sad8_2_mmx(uint8_t *blk1a, uint8_t *blk1b, uint8_t *blk2, int stride, int h)
{
int len= -(stride*h);
long len= -(stride*h);
asm volatile(
".balign 16 \n\t"
"1: \n\t"
"movq (%1, %%eax), %%mm0 \n\t"
"movq (%2, %%eax), %%mm1 \n\t"
"movq (%1, %%eax), %%mm2 \n\t"
"movq (%2, %%eax), %%mm3 \n\t"
"movq (%1, %%"REG_a"), %%mm0 \n\t"
"movq (%2, %%"REG_a"), %%mm1 \n\t"
"movq (%1, %%"REG_a"), %%mm2 \n\t"
"movq (%2, %%"REG_a"), %%mm3 \n\t"
"punpcklbw %%mm7, %%mm0 \n\t"
"punpcklbw %%mm7, %%mm1 \n\t"
"punpckhbw %%mm7, %%mm2 \n\t"
"punpckhbw %%mm7, %%mm3 \n\t"
"paddw %%mm0, %%mm1 \n\t"
"paddw %%mm2, %%mm3 \n\t"
"movq (%3, %%eax), %%mm4 \n\t"
"movq (%3, %%eax), %%mm2 \n\t"
"movq (%3, %%"REG_a"), %%mm4 \n\t"
"movq (%3, %%"REG_a"), %%mm2 \n\t"
"paddw %%mm5, %%mm1 \n\t"
"paddw %%mm5, %%mm3 \n\t"
"psrlw $1, %%mm1 \n\t"
......@@ -181,21 +182,21 @@ static inline void sad8_2_mmx(uint8_t *blk1a, uint8_t *blk1b, uint8_t *blk2, int
"punpckhbw %%mm7, %%mm1 \n\t"
"paddw %%mm1, %%mm0 \n\t"
"paddw %%mm0, %%mm6 \n\t"
"addl %4, %%eax \n\t"
"add %4, %%"REG_a" \n\t"
" js 1b \n\t"
: "+a" (len)
: "r" (blk1a - len), "r" (blk1b -len), "r" (blk2 - len), "r" (stride)
: "r" (blk1a - len), "r" (blk1b -len), "r" (blk2 - len), "r" ((long)stride)
);
}
static inline void sad8_4_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
{
int len= -(stride*h);
long len= -(stride*h);
asm volatile(
".balign 16 \n\t"
"1: \n\t"
"movq (%1, %%eax), %%mm0 \n\t"
"movq (%2, %%eax), %%mm1 \n\t"
"movq (%1, %%"REG_a"), %%mm0 \n\t"
"movq (%2, %%"REG_a"), %%mm1 \n\t"
"movq %%mm0, %%mm4 \n\t"
"movq %%mm1, %%mm2 \n\t"
"punpcklbw %%mm7, %%mm0 \n\t"
......@@ -204,8 +205,8 @@ static inline void sad8_4_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
"punpckhbw %%mm7, %%mm2 \n\t"
"paddw %%mm1, %%mm0 \n\t"
"paddw %%mm2, %%mm4 \n\t"
"movq 1(%1, %%eax), %%mm2 \n\t"
"movq 1(%2, %%eax), %%mm3 \n\t"
"movq 1(%1, %%"REG_a"), %%mm2 \n\t"
"movq 1(%2, %%"REG_a"), %%mm3 \n\t"
"movq %%mm2, %%mm1 \n\t"
"punpcklbw %%mm7, %%mm2 \n\t"
"punpckhbw %%mm7, %%mm1 \n\t"
......@@ -216,8 +217,8 @@ static inline void sad8_4_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
"punpckhbw %%mm7, %%mm4 \n\t"
"paddw %%mm3, %%mm2 \n\t"
"paddw %%mm4, %%mm1 \n\t"
"movq (%3, %%eax), %%mm3 \n\t"
"movq (%3, %%eax), %%mm4 \n\t"
"movq (%3, %%"REG_a"), %%mm3 \n\t"
"movq (%3, %%"REG_a"), %%mm4 \n\t"
"paddw %%mm5, %%mm2 \n\t"
"paddw %%mm5, %%mm1 \n\t"
"psrlw $2, %%mm2 \n\t"
......@@ -231,10 +232,10 @@ static inline void sad8_4_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
"punpckhbw %%mm7, %%mm2 \n\t"
"paddw %%mm2, %%mm0 \n\t"
"paddw %%mm0, %%mm6 \n\t"
"addl %4, %%eax \n\t"
"add %4, %%"REG_a" \n\t"
" js 1b \n\t"
: "+a" (len)
: "r" (blk1 - len), "r" (blk1 -len + stride), "r" (blk2 - len), "r" (stride)
: "r" (blk1 - len), "r" (blk1 -len + stride), "r" (blk2 - len), "r" ((long)stride)
);
}
......
libavcodec/i386/mpegvideo_mmx.c
View file @ 09fcf89b
......@@ -23,6 +23,7 @@
#include "../dsputil.h"
#include "../mpegvideo.h"
#include "../avcodec.h"
#include "mmx.h"
extern uint8_t zigzag_direct_noperm[64];
extern uint16_t inv_zigzag_direct16[64];
......@@ -34,7 +35,7 @@ static const unsigned long long int mm_wone __attribute__ ((aligned(8))) = 0x000
static void dct_unquantize_h263_intra_mmx(MpegEncContext *s,
DCTELEM *block, int n, int qscale)
{
int level, qmul, qadd, nCoeffs;
long level, qmul, qadd, nCoeffs;
qmul = qscale << 1;
......@@ -97,7 +98,7 @@ asm volatile(
"movq %%mm0, (%0, %3) \n\t"
"movq %%mm1, 8(%0, %3) \n\t"
"addl $16, %3 \n\t"
"add $16, %3 \n\t"
"jng 1b \n\t"
::"r" (block+nCoeffs), "g"(qmul), "g" (qadd), "r" (2*(-nCoeffs))
: "memory"
......@@ -109,7 +110,7 @@ asm volatile(
static void dct_unquantize_h263_inter_mmx(MpegEncContext *s,
DCTELEM *block, int n, int qscale)
{
int qmul, qadd, nCoeffs;
long qmul, qadd, nCoeffs;
qmul = qscale << 1;
qadd = (qscale - 1) | 1;
......@@ -160,7 +161,7 @@ asm volatile(
"movq %%mm0, (%0, %3) \n\t"
"movq %%mm1, 8(%0, %3) \n\t"
"addl $16, %3 \n\t"
"add $16, %3 \n\t"
"jng 1b \n\t"
::"r" (block+nCoeffs), "g"(qmul), "g" (qadd), "r" (2*(-nCoeffs))
: "memory"
......@@ -200,7 +201,7 @@ asm volatile(
static void dct_unquantize_mpeg1_intra_mmx(MpegEncContext *s,
DCTELEM *block, int n, int qscale)
{
int nCoeffs;
long nCoeffs;
const uint16_t *quant_matrix;
int block0;
......@@ -220,13 +221,13 @@ asm volatile(
"movd %2, %%mm6 \n\t"
"packssdw %%mm6, %%mm6 \n\t"
"packssdw %%mm6, %%mm6 \n\t"
"movl %3, %%eax \n\t"
"mov %3, %%"REG_a" \n\t"
".balign 16\n\t"
"1: \n\t"
"movq (%0, %%eax), %%mm0 \n\t"
"movq 8(%0, %%eax), %%mm1 \n\t"
"movq (%1, %%eax), %%mm4 \n\t"
"movq 8(%1, %%eax), %%mm5 \n\t"
"movq (%0, %%"REG_a"), %%mm0 \n\t"
"movq 8(%0, %%"REG_a"), %%mm1 \n\t"
"movq (%1, %%"REG_a"), %%mm4 \n\t"
"movq 8(%1, %%"REG_a"), %%mm5 \n\t"
"pmullw %%mm6, %%mm4 \n\t" // q=qscale*quant_matrix[i]
"pmullw %%mm6, %%mm5 \n\t" // q=qscale*quant_matrix[i]
"pxor %%mm2, %%mm2 \n\t"
......@@ -241,8 +242,8 @@ asm volatile(
"pmullw %%mm5, %%mm1 \n\t" // abs(block[i])*q
"pxor %%mm4, %%mm4 \n\t"
"pxor %%mm5, %%mm5 \n\t" // FIXME slow
"pcmpeqw (%0, %%eax), %%mm4 \n\t" // block[i] == 0 ? -1 : 0
"pcmpeqw 8(%0, %%eax), %%mm5 \n\t" // block[i] == 0 ? -1 : 0
"pcmpeqw (%0, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0
"pcmpeqw 8(%0, %%"REG_a"), %%mm5\n\t" // block[i] == 0 ? -1 : 0
"psraw $3, %%mm0 \n\t"
"psraw $3, %%mm1 \n\t"
"psubw %%mm7, %%mm0 \n\t"
......@@ -255,13 +256,13 @@ asm volatile(
"psubw %%mm3, %%mm1 \n\t"
"pandn %%mm0, %%mm4 \n\t"
"pandn %%mm1, %%mm5 \n\t"
"movq %%mm4, (%0, %%eax) \n\t"
"movq %%mm5, 8(%0, %%eax) \n\t"
"movq %%mm4, (%0, %%"REG_a") \n\t"
"movq %%mm5, 8(%0, %%"REG_a") \n\t"
"addl $16, %%eax \n\t"
"add $16, %%"REG_a" \n\t"
"js 1b \n\t"
::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "g" (-2*nCoeffs)
: "%eax", "memory"
: "%"REG_a, "memory"
);
block[0]= block0;
}
......@@ -269,7 +270,7 @@ asm volatile(
static void dct_unquantize_mpeg1_inter_mmx(MpegEncContext *s,
DCTELEM *block, int n, int qscale)
{
int nCoeffs;
long nCoeffs;
const uint16_t *quant_matrix;
assert(s->block_last_index[n]>=0);
......@@ -283,13 +284,13 @@ asm volatile(
"movd %2, %%mm6 \n\t"
"packssdw %%mm6, %%mm6 \n\t"
"packssdw %%mm6, %%mm6 \n\t"
"movl %3, %%eax \n\t"
"mov %3, %%"REG_a" \n\t"
".balign 16\n\t"
"1: \n\t"
"movq (%0, %%eax), %%mm0 \n\t"
"movq 8(%0, %%eax), %%mm1 \n\t"
"movq (%1, %%eax), %%mm4 \n\t"
"movq 8(%1, %%eax), %%mm5 \n\t"
"movq (%0, %%"REG_a"), %%mm0 \n\t"
"movq 8(%0, %%"REG_a"), %%mm1 \n\t"
"movq (%1, %%"REG_a"), %%mm4 \n\t"
"movq 8(%1, %%"REG_a"), %%mm5 \n\t"
"pmullw %%mm6, %%mm4 \n\t" // q=qscale*quant_matrix[i]
"pmullw %%mm6, %%mm5 \n\t" // q=qscale*quant_matrix[i]
"pxor %%mm2, %%mm2 \n\t"
......@@ -308,8 +309,8 @@ asm volatile(
"pmullw %%mm5, %%mm1 \n\t" // (abs(block[i])*2 + 1)*q
"pxor %%mm4, %%mm4 \n\t"
"pxor %%mm5, %%mm5 \n\t" // FIXME slow
"pcmpeqw (%0, %%eax), %%mm4 \n\t" // block[i] == 0 ? -1 : 0
"pcmpeqw 8(%0, %%eax), %%mm5 \n\t" // block[i] == 0 ? -1 : 0
"pcmpeqw (%0, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0
"pcmpeqw 8(%0, %%"REG_a"), %%mm5\n\t" // block[i] == 0 ? -1 : 0
"psraw $4, %%mm0 \n\t"
"psraw $4, %%mm1 \n\t"
"psubw %%mm7, %%mm0 \n\t"
......@@ -322,20 +323,20 @@ asm volatile(
"psubw %%mm3, %%mm1 \n\t"
"pandn %%mm0, %%mm4 \n\t"
"pandn %%mm1, %%mm5 \n\t"
"movq %%mm4, (%0, %%eax) \n\t"
"movq %%mm5, 8(%0, %%eax) \n\t"
"movq %%mm4, (%0, %%"REG_a") \n\t"
"movq %%mm5, 8(%0, %%"REG_a") \n\t"
"addl $16, %%eax \n\t"
"add $16, %%"REG_a" \n\t"
"js 1b \n\t"
::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "g" (-2*nCoeffs)
: "%eax", "memory"
: "%"REG_a, "memory"
);
}
static void dct_unquantize_mpeg2_intra_mmx(MpegEncContext *s,
DCTELEM *block, int n, int qscale)
{
int nCoeffs;
long nCoeffs;
const uint16_t *quant_matrix;
int block0;
......@@ -355,13 +356,13 @@ asm volatile(
"movd %2, %%mm6 \n\t"
"packssdw %%mm6, %%mm6 \n\t"
"packssdw %%mm6, %%mm6 \n\t"
"movl %3, %%eax \n\t"
"mov %3, %%"REG_a" \n\t"
".balign 16\n\t"
"1: \n\t"
"movq (%0, %%eax), %%mm0 \n\t"
"movq 8(%0, %%eax), %%mm1 \n\t"
"movq (%1, %%eax), %%mm4 \n\t"
"movq 8(%1, %%eax), %%mm5 \n\t"
"movq (%0, %%"REG_a"), %%mm0 \n\t"
"movq 8(%0, %%"REG_a"), %%mm1 \n\t"
"movq (%1, %%"REG_a"), %%mm4 \n\t"
"movq 8(%1, %%"REG_a"), %%mm5 \n\t"
"pmullw %%mm6, %%mm4 \n\t" // q=qscale*quant_matrix[i]
"pmullw %%mm6, %%mm5 \n\t" // q=qscale*quant_matrix[i]
"pxor %%mm2, %%mm2 \n\t"
......@@ -376,8 +377,8 @@ asm volatile(
"pmullw %%mm5, %%mm1 \n\t" // abs(block[i])*q
"pxor %%mm4, %%mm4 \n\t"
"pxor %%mm5, %%mm5 \n\t" // FIXME slow
"pcmpeqw (%0, %%eax), %%mm4 \n\t" // block[i] == 0 ? -1 : 0
"pcmpeqw 8(%0, %%eax), %%mm5 \n\t" // block[i] == 0 ? -1 : 0
"pcmpeqw (%0, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0
"pcmpeqw 8(%0, %%"REG_a"), %%mm5\n\t" // block[i] == 0 ? -1 : 0
"psraw $3, %%mm0 \n\t"
"psraw $3, %%mm1 \n\t"
"pxor %%mm2, %%mm0 \n\t"
......@@ -386,13 +387,13 @@ asm volatile(
"psubw %%mm3, %%mm1 \n\t"
"pandn %%mm0, %%mm4 \n\t"
"pandn %%mm1, %%mm5 \n\t"
"movq %%mm4, (%0, %%eax) \n\t"
"movq %%mm5, 8(%0, %%eax) \n\t"
"movq %%mm4, (%0, %%"REG_a") \n\t"
"movq %%mm5, 8(%0, %%"REG_a") \n\t"
"addl $16, %%eax \n\t"
"add $16, %%"REG_a" \n\t"
"jng 1b \n\t"
::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "g" (-2*nCoeffs)
: "%eax", "memory"
: "%"REG_a, "memory"
);
block[0]= block0;
//Note, we dont do mismatch control for intra as errors cannot accumulate
......@@ -401,7 +402,7 @@ asm volatile(
static void dct_unquantize_mpeg2_inter_mmx(MpegEncContext *s,
DCTELEM *block, int n, int qscale)
{
int nCoeffs;
long nCoeffs;
const uint16_t *quant_matrix;
assert(s->block_last_index[n]>=0);
......@@ -416,13 +417,13 @@ asm volatile(
"movd %2, %%mm6 \n\t"
"packssdw %%mm6, %%mm6 \n\t"
"packssdw %%mm6, %%mm6 \n\t"
"movl %3, %%eax \n\t"
"mov %3, %%"REG_a" \n\t"
".balign 16\n\t"
"1: \n\t"
"movq (%0, %%eax), %%mm0 \n\t"
"movq 8(%0, %%eax), %%mm1 \n\t"
"movq (%1, %%eax), %%mm4 \n\t"
"movq 8(%1, %%eax), %%mm5 \n\t"
"movq (%0, %%"REG_a"), %%mm0 \n\t"
"movq 8(%0, %%"REG_a"), %%mm1 \n\t"
"movq (%1, %%"REG_a"), %%mm4 \n\t"
"movq 8(%1, %%"REG_a"), %%mm5 \n\t"
"pmullw %%mm6, %%mm4 \n\t" // q=qscale*quant_matrix[i]
"pmullw %%mm6, %%mm5 \n\t" // q=qscale*quant_matrix[i]
"pxor %%mm2, %%mm2 \n\t"
......@@ -441,8 +442,8 @@ asm volatile(
"paddw %%mm5, %%mm1 \n\t" // (abs(block[i])*2 + 1)*q
"pxor %%mm4, %%mm4 \n\t"
"pxor %%mm5, %%mm5 \n\t" // FIXME slow
"pcmpeqw (%0, %%eax), %%mm4 \n\t" // block[i] == 0 ? -1 : 0
"pcmpeqw 8(%0, %%eax), %%mm5 \n\t" // block[i] == 0 ? -1 : 0
"pcmpeqw (%0, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0
"pcmpeqw 8(%0, %%"REG_a"), %%mm5\n\t" // block[i] == 0 ? -1 : 0
"psrlw $4, %%mm0 \n\t"
"psrlw $4, %%mm1 \n\t"
"pxor %%mm2, %%mm0 \n\t"
......@@ -453,10 +454,10 @@ asm volatile(
"pandn %%mm1, %%mm5 \n\t"
"pxor %%mm4, %%mm7 \n\t"
"pxor %%mm5, %%mm7 \n\t"
"movq %%mm4, (%0, %%eax) \n\t"
"movq %%mm5, 8(%0, %%eax) \n\t"
"movq %%mm4, (%0, %%"REG_a") \n\t"
"movq %%mm5, 8(%0, %%"REG_a") \n\t"
"addl $16, %%eax \n\t"
"add $16, %%"REG_a" \n\t"
"jng 1b \n\t"
"movd 124(%0, %3), %%mm0 \n\t"
"movq %%mm7, %%mm6 \n\t"
......@@ -471,7 +472,7 @@ asm volatile(
"movd %%mm0, 124(%0, %3) \n\t"
::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "r" (-2*nCoeffs)
: "%eax", "memory"
: "%"REG_a, "memory"
);
}
......@@ -499,11 +500,11 @@ static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w)
"punpckhwd %%mm1, %%mm1 \n\t"
"punpckhdq %%mm1, %%mm1 \n\t"
"movq %%mm1, (%0, %2) \n\t"
"addl %1, %0 \n\t"
"cmpl %3, %0 \n\t"
"add %1, %0 \n\t"
"cmp %3, %0 \n\t"
" jb 1b \n\t"
: "+r" (ptr)
: "r" (wrap), "r" (width), "r" (ptr + wrap*height)
: "r" ((long)wrap), "r" ((long)width), "r" (ptr + wrap*height)
);
}
else
......@@ -522,11 +523,11 @@ static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w)
"punpckhdq %%mm1, %%mm1 \n\t"
"movq %%mm1, (%0, %2) \n\t"
"movq %%mm1, 8(%0, %2) \n\t"
"addl %1, %0 \n\t"
"cmpl %3, %0 \n\t"
"add %1, %0 \n\t"
"cmp %3, %0 \n\t"
" jb 1b \n\t"
: "+r" (ptr)
: "r" (wrap), "r" (width), "r" (ptr + wrap*height)
: "r" ((long)wrap), "r" ((long)width), "r" (ptr + wrap*height)
);
}
......@@ -540,11 +541,11 @@ static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w)
"movq %%mm0, (%0, %2) \n\t"
"movq %%mm0, (%0, %2, 2) \n\t"
"movq %%mm0, (%0, %3) \n\t"
"addl $8, %0 \n\t"
"cmpl %4, %0 \n\t"
"add $8, %0 \n\t"
"cmp %4, %0 \n\t"
" jb 1b \n\t"
: "+r" (ptr)
: "r" ((int)buf - (int)ptr - w), "r" (-wrap), "r" (-wrap*3), "r" (ptr+width+2*w)
: "r" ((long)buf - (long)ptr - w), "r" ((long)-wrap), "r" ((long)-wrap*3), "r" (ptr+width+2*w)
);
ptr= last_line + (i + 1) * wrap - w;
asm volatile(
......@@ -554,11 +555,11 @@ static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w)
"movq %%mm0, (%0, %2) \n\t"
"movq %%mm0, (%0, %2, 2) \n\t"
"movq %%mm0, (%0, %3) \n\t"
"addl $8, %0 \n\t"
"cmpl %4, %0 \n\t"
"add $8, %0 \n\t"
"cmp %4, %0 \n\t"
" jb 1b \n\t"
: "+r" (ptr)
: "r" ((int)last_line - (int)ptr - w), "r" (wrap), "r" (wrap*3), "r" (ptr+width+2*w)
: "r" ((long)last_line - (long)ptr - w), "r" ((long)wrap), "r" ((long)wrap*3), "r" (ptr+width+2*w)
);
}
}
......@@ -607,10 +608,10 @@ static void denoise_dct_mmx(MpegEncContext *s, DCTELEM *block){
"movq %%mm2, 8(%1) \n\t"
"movq %%mm5, 16(%1) \n\t"
"movq %%mm3, 24(%1) \n\t"
"addl $16, %0 \n\t"
"addl $32, %1 \n\t"
"addl $16, %2 \n\t"
"cmpl %3, %0 \n\t"
"add $16, %0 \n\t"
"add $32, %1 \n\t"
"add $16, %2 \n\t"
"cmp %3, %0 \n\t"
" jb 1b \n\t"
: "+r" (block), "+r" (sum), "+r" (offset)
: "r"(block+64)
......@@ -661,10 +662,10 @@ static void denoise_dct_sse2(MpegEncContext *s, DCTELEM *block){
"movdqa %%xmm6, 16(%1) \n\t"
"movdqa %%xmm5, 32(%1) \n\t"
"movdqa %%xmm0, 48(%1) \n\t"
"addl $32, %0 \n\t"
"addl $64, %1 \n\t"
"addl $32, %2 \n\t"
"cmpl %3, %0 \n\t"
"add $32, %0 \n\t"
"add $64, %1 \n\t"
"add $32, %2 \n\t"
"cmp %3, %0 \n\t"
" jb 1b \n\t"
: "+r" (block), "+r" (sum), "+r" (offset)
: "r"(block+64)
......
libavcodec/i386/mpegvideo_mmx_template.c
View file @ 09fcf89b
......@@ -36,7 +36,8 @@ static int RENAME(dct_quantize)(MpegEncContext *s,
DCTELEM *block, int n,
int qscale, int *overflow)
{
int level=0, last_non_zero_p1, q; //=0 is cuz gcc says uninitalized ...
long last_non_zero_p1;
int level=0, q; //=0 is cuz gcc says uninitalized ...
const uint16_t *qmat, *bias;
__align8 int16_t temp_block[64];
......@@ -90,18 +91,18 @@ static int RENAME(dct_quantize)(MpegEncContext *s,
if(s->out_format == FMT_H263 && s->mpeg_quant==0){
asm volatile(
"movd %%eax, %%mm3 \n\t" // last_non_zero_p1
"movd %%"REG_a", %%mm3 \n\t" // last_non_zero_p1
SPREADW(%%mm3)
"pxor %%mm7, %%mm7 \n\t" // 0
"pxor %%mm4, %%mm4 \n\t" // 0
"movq (%2), %%mm5 \n\t" // qmat[0]
"pxor %%mm6, %%mm6 \n\t"
"psubw (%3), %%mm6 \n\t" // -bias[0]
"movl $-128, %%eax \n\t"
"mov $-128, %%"REG_a" \n\t"
".balign 16 \n\t"
"1: \n\t"
"pxor %%mm1, %%mm1 \n\t" // 0
"movq (%1, %%eax), %%mm0 \n\t" // block[i]
"movq (%1, %%"REG_a"), %%mm0 \n\t" // block[i]
"pcmpgtw %%mm0, %%mm1 \n\t" // block[i] <= 0 ? 0xFF : 0x00
"pxor %%mm1, %%mm0 \n\t"
"psubw %%mm1, %%mm0 \n\t" // ABS(block[i])
......@@ -110,13 +111,13 @@ static int RENAME(dct_quantize)(MpegEncContext *s,
"por %%mm0, %%mm4 \n\t"
"pxor %%mm1, %%mm0 \n\t"
"psubw %%mm1, %%mm0 \n\t" // out=((ABS(block[i])*qmat[0] - bias[0]*qmat[0])>>16)*sign(block[i])
"movq %%mm0, (%5, %%eax) \n\t"
"movq %%mm0, (%5, %%"REG_a") \n\t"
"pcmpeqw %%mm7, %%mm0 \n\t" // out==0 ? 0xFF : 0x00
"movq (%4, %%eax), %%mm1 \n\t"
"movq %%mm7, (%1, %%eax) \n\t" // 0
"movq (%4, %%"REG_a"), %%mm1 \n\t"
"movq %%mm7, (%1, %%"REG_a") \n\t" // 0
"pandn %%mm1, %%mm0 \n\t"
PMAXW(%%mm0, %%mm3)
"addl $8, %%eax \n\t"
"add $8, %%"REG_a" \n\t"
" js 1b \n\t"
"movq %%mm3, %%mm0 \n\t"
"psrlq $32, %%mm3 \n\t"
......@@ -124,8 +125,8 @@ static int RENAME(dct_quantize)(MpegEncContext *s,
"movq %%mm3, %%mm0 \n\t"
"psrlq $16, %%mm3 \n\t"
PMAXW(%%mm0, %%mm3)
"movd %%mm3, %%eax \n\t"
"movzbl %%al, %%eax \n\t" // last_non_zero_p1
"movd %%mm3, %%"REG_a" \n\t"
"movzb %%al, %%"REG_a" \n\t" // last_non_zero_p1
: "+a" (last_non_zero_p1)
: "r" (block+64), "r" (qmat), "r" (bias),
"r" (inv_zigzag_direct16+64), "r" (temp_block+64)
......@@ -142,32 +143,32 @@ static int RENAME(dct_quantize)(MpegEncContext *s,
);
}else{ // FMT_H263
asm volatile(
"movd %%eax, %%mm3 \n\t" // last_non_zero_p1
"movd %%"REG_a", %%mm3 \n\t" // last_non_zero_p1
SPREADW(%%mm3)
"pxor %%mm7, %%mm7 \n\t" // 0
"pxor %%mm4, %%mm4 \n\t" // 0
"movl $-128, %%eax \n\t"
"mov $-128, %%"REG_a" \n\t"
".balign 16 \n\t"
"1: \n\t"
"pxor %%mm1, %%mm1 \n\t" // 0
"movq (%1, %%eax), %%mm0 \n\t" // block[i]
"movq (%1, %%"REG_a"), %%mm0 \n\t" // block[i]
"pcmpgtw %%mm0, %%mm1 \n\t" // block[i] <= 0 ? 0xFF : 0x00
"pxor %%mm1, %%mm0 \n\t"
"psubw %%mm1, %%mm0 \n\t" // ABS(block[i])
"movq (%3, %%eax), %%mm6 \n\t" // bias[0]
"movq (%3, %%"REG_a"), %%mm6 \n\t" // bias[0]
"paddusw %%mm6, %%mm0 \n\t" // ABS(block[i]) + bias[0]
"movq (%2, %%eax), %%mm5 \n\t" // qmat[i]
"movq (%2, %%"REG_a"), %%mm5 \n\t" // qmat[i]
"pmulhw %%mm5, %%mm0 \n\t" // (ABS(block[i])*qmat[0] + bias[0]*qmat[0])>>16
"por %%mm0, %%mm4 \n\t"
"pxor %%mm1, %%mm0 \n\t"
"psubw %%mm1, %%mm0 \n\t" // out=((ABS(block[i])*qmat[0] - bias[0]*qmat[0])>>16)*sign(block[i])
"movq %%mm0, (%5, %%eax) \n\t"
"movq %%mm0, (%5, %%"REG_a") \n\t"
"pcmpeqw %%mm7, %%mm0 \n\t" // out==0 ? 0xFF : 0x00
"movq (%4, %%eax), %%mm1 \n\t"
"movq %%mm7, (%1, %%eax) \n\t" // 0
"movq (%4, %%"REG_a"), %%mm1 \n\t"
"movq %%mm7, (%1, %%"REG_a") \n\t" // 0
"pandn %%mm1, %%mm0 \n\t"
PMAXW(%%mm0, %%mm3)
"addl $8, %%eax \n\t"
"add $8, %%"REG_a" \n\t"
" js 1b \n\t"
"movq %%mm3, %%mm0 \n\t"
"psrlq $32, %%mm3 \n\t"
......@@ -175,8 +176,8 @@ static int RENAME(dct_quantize)(MpegEncContext *s,
"movq %%mm3, %%mm0 \n\t"
"psrlq $16, %%mm3 \n\t"
PMAXW(%%mm0, %%mm3)
"movd %%mm3, %%eax \n\t"
"movzbl %%al, %%eax \n\t" // last_non_zero_p1
"movd %%mm3, %%"REG_a" \n\t"
"movzb %%al, %%"REG_a" \n\t" // last_non_zero_p1
: "+a" (last_non_zero_p1)
: "r" (block+64), "r" (qmat+64), "r" (bias+64),
"r" (inv_zigzag_direct16+64), "r" (temp_block+64)
......
libavcodec/libpostproc/postprocess.c
View file @ 09fcf89b
......@@ -119,7 +119,7 @@ try to unroll inner for(x=0 ... loop to avoid these damn if(x ... checks
# define always_inline inline
#endif
#ifdef ARCH_X86
#if defined(ARCH_X86) || defined(ARCH_X86_64)
static uint64_t __attribute__((aligned(8))) attribute_used w05= 0x0005000500050005LL;
static uint64_t __attribute__((aligned(8))) attribute_used w04= 0x0004000400040004LL;
static uint64_t __attribute__((aligned(8))) attribute_used w20= 0x0020002000200020LL;
......@@ -172,7 +172,7 @@ static char *replaceTable[]=
};
#ifdef ARCH_X86
#if defined(ARCH_X86) || defined(ARCH_X86_64)
static inline void prefetchnta(void *p)
{
asm volatile( "prefetchnta (%0)\n\t"
......@@ -597,7 +597,7 @@ static always_inline void do_a_deblock_C(uint8_t *src, int step, int stride, PPC
#endif //HAVE_ALTIVEC
#endif //ARCH_POWERPC
#ifdef ARCH_X86
#if defined(ARCH_X86) || defined(ARCH_X86_64)
#if (defined (HAVE_MMX) && !defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)
#define COMPILE_MMX
......@@ -616,13 +616,11 @@ static always_inline void do_a_deblock_C(uint8_t *src, int step, int stride, PPC
#undef HAVE_MMX2
#undef HAVE_3DNOW
#undef HAVE_ALTIVEC
#undef ARCH_X86
#ifdef COMPILE_C
#undef HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
#undef ARCH_X86
#define RENAME(a) a ## _C
#include "postprocess_template.c"
#endif
......@@ -643,7 +641,6 @@ static always_inline void do_a_deblock_C(uint8_t *src, int step, int stride, PPC
#define HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
#define ARCH_X86
#define RENAME(a) a ## _MMX
#include "postprocess_template.c"
#endif
......@@ -654,7 +651,6 @@ static always_inline void do_a_deblock_C(uint8_t *src, int step, int stride, PPC
#define HAVE_MMX
#define HAVE_MMX2
#undef HAVE_3DNOW
#define ARCH_X86
#define RENAME(a) a ## _MMX2
#include "postprocess_template.c"
#endif
......@@ -665,7 +661,6 @@ static always_inline void do_a_deblock_C(uint8_t *src, int step, int stride, PPC
#define HAVE_MMX
#undef HAVE_MMX2
#define HAVE_3DNOW
#define ARCH_X86
#define RENAME(a) a ## _3DNow
#include "postprocess_template.c"
#endif
......@@ -683,7 +678,7 @@ static inline void postProcess(uint8_t src[], int srcStride, uint8_t dst[], int
// difference wouldnt be messureable here but its much better because
// someone might exchange the cpu whithout restarting mplayer ;)
#ifdef RUNTIME_CPUDETECT
#ifdef ARCH_X86
#if defined(ARCH_X86) || defined(ARCH_X86_64)
// ordered per speed fasterst first
if(c->cpuCaps & PP_CPU_CAPS_MMX2)
postProcess_MMX2(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c);
......
libavcodec/libpostproc/postprocess_template.c
View file @ 09fcf89b
......@@ -22,15 +22,37 @@
*/
#ifdef ARCH_X86_64
# define REGa rax
# define REGc rcx
# define REGd rdx
# define REG_a "rax"
# define REG_c "rcx"
# define REG_d "rdx"
# define REG_SP "rsp"
# define ALIGN_MASK "$0xFFFFFFFFFFFFFFF8"
#else
# define REGa eax
# define REGc ecx
# define REGd edx
# define REG_a "eax"
# define REG_c "ecx"
# define REG_d "edx"
# define REG_SP "esp"
# define ALIGN_MASK "$0xFFFFFFF8"
#endif
#undef PAVGB
#undef PMINUB
#undef PMAXUB
#ifdef HAVE_MMX2
#define PAVGB(a,b) "pavgb " #a ", " #b " \n\t"
#define REAL_PAVGB(a,b) "pavgb " #a ", " #b " \n\t"
#elif defined (HAVE_3DNOW)
#define PAVGB(a,b) "pavgusb " #a ", " #b " \n\t"
#define REAL_RPAVGB(a,b) "pavgusb " #a ", " #b " \n\t"
#endif
#define PAVGB(a,b) REAL_PAVGB(a,b)
#ifdef HAVE_MMX2
#define PMINUB(a,b,t) "pminub " #a ", " #b " \n\t"
......@@ -64,12 +86,12 @@ asm volatile(
);
asm volatile(
"leal (%2, %3), %%eax \n\t"
"lea (%2, %3), %%"REG_a" \n\t"
// 0 1 2 3 4 5 6 7 8 9
// %1 eax eax+%2 eax+2%2 %1+4%2 ecx ecx+%2 ecx+2%2 %1+8%2 ecx+4%2
"movq (%2), %%mm0 \n\t"
"movq (%%eax), %%mm1 \n\t"
"movq (%%"REG_a"), %%mm1 \n\t"
"movq %%mm0, %%mm3 \n\t"
"movq %%mm0, %%mm4 \n\t"
PMAXUB(%%mm1, %%mm4)
......@@ -78,7 +100,7 @@ asm volatile(
"paddb %%mm7, %%mm0 \n\t"
"pcmpgtb %%mm6, %%mm0 \n\t"
"movq (%%eax,%3), %%mm2 \n\t"
"movq (%%"REG_a",%3), %%mm2 \n\t"
PMAXUB(%%mm2, %%mm4)
PMINUB(%%mm2, %%mm3, %%mm5)
"psubb %%mm2, %%mm1 \n\t"
......@@ -86,7 +108,7 @@ asm volatile(
"pcmpgtb %%mm6, %%mm1 \n\t"
"paddb %%mm1, %%mm0 \n\t"
"movq (%%eax, %3, 2), %%mm1 \n\t"
"movq (%%"REG_a", %3, 2), %%mm1 \n\t"
PMAXUB(%%mm1, %%mm4)
PMINUB(%%mm1, %%mm3, %%mm5)
"psubb %%mm1, %%mm2 \n\t"
......@@ -94,7 +116,7 @@ asm volatile(
"pcmpgtb %%mm6, %%mm2 \n\t"
"paddb %%mm2, %%mm0 \n\t"
"leal (%%eax, %3, 4), %%eax \n\t"
"lea (%%"REG_a", %3, 4), %%"REG_a" \n\t"
"movq (%2, %3, 4), %%mm2 \n\t"
PMAXUB(%%mm2, %%mm4)
......@@ -104,7 +126,7 @@ asm volatile(
"pcmpgtb %%mm6, %%mm1 \n\t"
"paddb %%mm1, %%mm0 \n\t"
"movq (%%eax), %%mm1 \n\t"
"movq (%%"REG_a"), %%mm1 \n\t"
PMAXUB(%%mm1, %%mm4)
PMINUB(%%mm1, %%mm3, %%mm5)
"psubb %%mm1, %%mm2 \n\t"
......@@ -112,7 +134,7 @@ asm volatile(
"pcmpgtb %%mm6, %%mm2 \n\t"
"paddb %%mm2, %%mm0 \n\t"
"movq (%%eax, %3), %%mm2 \n\t"
"movq (%%"REG_a", %3), %%mm2 \n\t"
PMAXUB(%%mm2, %%mm4)
PMINUB(%%mm2, %%mm3, %%mm5)
"psubb %%mm2, %%mm1 \n\t"
......@@ -120,7 +142,7 @@ asm volatile(
"pcmpgtb %%mm6, %%mm1 \n\t"
"paddb %%mm1, %%mm0 \n\t"
"movq (%%eax, %3, 2), %%mm1 \n\t"
"movq (%%"REG_a", %3, 2), %%mm1 \n\t"
PMAXUB(%%mm1, %%mm4)
PMINUB(%%mm1, %%mm3, %%mm5)
"psubb %%mm1, %%mm2 \n\t"
......@@ -152,8 +174,8 @@ asm volatile(
"movd %%mm4, %1 \n\t"
: "=r" (numEq), "=r" (dcOk)
: "r" (src), "r" (stride), "m" (c->pQPb)
: "%eax"
: "r" (src), "r" ((long)stride), "m" (c->pQPb)
: "%"REG_a
);
numEq= (-numEq) &0xFF;
......@@ -194,10 +216,10 @@ static inline void RENAME(doVertLowPass)(uint8_t *src, int stride, PPContext *c)
"por %%mm2, %%mm6 \n\t"// First Line to Filter
"movq (%0, %1, 8), %%mm5 \n\t"
"leal (%0, %1, 4), %%eax \n\t"
"leal (%0, %1, 8), %%ecx \n\t"
"subl %1, %%ecx \n\t"
"addl %1, %0 \n\t" // %0 points to line 1 not 0
"lea (%0, %1, 4), %%"REG_a" \n\t"
"lea (%0, %1, 8), %%"REG_c" \n\t"
"sub %1, %%"REG_c" \n\t"
"add %1, %0 \n\t" // %0 points to line 1 not 0
"movq (%0, %1, 8), %%mm7 \n\t"
"movq %%mm5, %%mm1 \n\t"
"movq %%mm7, %%mm2 \n\t"
......@@ -225,7 +247,7 @@ static inline void RENAME(doVertLowPass)(uint8_t *src, int stride, PPContext *c)
"movq (%0, %1, 4), %%mm2 \n\t" // 1
"movq %%mm2, %%mm5 \n\t" // 1
PAVGB((%%eax), %%mm2) // 11 /2
PAVGB((%%REGa), %%mm2) // 11 /2
PAVGB((%0, %1, 2), %%mm2) // 211 /4
"movq %%mm2, %%mm3 \n\t" // 211 /4
"movq (%0), %%mm4 \n\t" // 1
......@@ -237,15 +259,15 @@ static inline void RENAME(doVertLowPass)(uint8_t *src, int stride, PPContext *c)
PAVGB(%%mm6, %%mm0) //1 1 /2
"movq %%mm4, %%mm3 \n\t" // 1
PAVGB((%0,%1,2), %%mm3) // 1 1 /2
PAVGB((%%eax,%1,2), %%mm5) // 11 /2
PAVGB((%%eax), %%mm5) // 211 /4
PAVGB((%%REGa,%1,2), %%mm5) // 11 /2
PAVGB((%%REGa), %%mm5) // 211 /4
PAVGB(%%mm5, %%mm3) // 2 2211 /8
PAVGB(%%mm0, %%mm3) //4242211 /16
"movq %%mm3, (%0,%1) \n\t" // X
// mm1=2 mm2=3(211) mm4=1 mm5=4(211) mm6=0 mm7=9
PAVGB(%%mm4, %%mm6) //11 /2
"movq (%%ecx), %%mm0 \n\t" // 1
PAVGB((%%eax, %1, 2), %%mm0) // 11/2
"movq (%%"REG_c"), %%mm0 \n\t" // 1
PAVGB((%%REGa, %1, 2), %%mm0) // 11/2
"movq %%mm0, %%mm3 \n\t" // 11/2
PAVGB(%%mm1, %%mm0) // 2 11/4
PAVGB(%%mm6, %%mm0) //222 11/8
......@@ -253,17 +275,17 @@ static inline void RENAME(doVertLowPass)(uint8_t *src, int stride, PPContext *c)
"movq (%0, %1, 2), %%mm2 \n\t" // 1
"movq %%mm0, (%0, %1, 2) \n\t" // X
// mm1=2 mm2=3 mm3=6(11) mm4=1 mm5=4(211) mm6=0(11) mm7=9
"movq (%%eax, %1, 4), %%mm0 \n\t" // 1
PAVGB((%%ecx), %%mm0) // 11 /2
"movq (%%"REG_a", %1, 4), %%mm0 \n\t" // 1
PAVGB((%%REGc), %%mm0) // 11 /2
PAVGB(%%mm0, %%mm6) //11 11 /4
PAVGB(%%mm1, %%mm4) // 11 /2
PAVGB(%%mm2, %%mm1) // 11 /2
PAVGB(%%mm1, %%mm6) //1122 11 /8
PAVGB(%%mm5, %%mm6) //112242211 /16
"movq (%%eax), %%mm5 \n\t" // 1
"movq %%mm6, (%%eax) \n\t" // X
"movq (%%"REG_a"), %%mm5 \n\t" // 1
"movq %%mm6, (%%"REG_a") \n\t" // X
// mm0=7(11) mm1=2(11) mm2=3 mm3=6(11) mm4=1(11) mm5=4 mm7=9
"movq (%%eax, %1, 4), %%mm6 \n\t" // 1
"movq (%%"REG_a", %1, 4), %%mm6 \n\t" // 1
PAVGB(%%mm7, %%mm6) // 11 /2
PAVGB(%%mm4, %%mm6) // 11 11 /4
PAVGB(%%mm3, %%mm6) // 11 2211 /8
......@@ -276,29 +298,29 @@ static inline void RENAME(doVertLowPass)(uint8_t *src, int stride, PPContext *c)
PAVGB(%%mm7, %%mm1) // 11 2 /4
PAVGB(%%mm4, %%mm5) // 11 /2
PAVGB(%%mm5, %%mm0) // 11 11 /4
"movq (%%eax, %1, 2), %%mm6 \n\t" // 1
"movq (%%"REG_a", %1, 2), %%mm6 \n\t" // 1
PAVGB(%%mm6, %%mm1) // 11 4 2 /8
PAVGB(%%mm0, %%mm1) // 11224222 /16
"movq %%mm1, (%%eax, %1, 2) \n\t" // X
"movq %%mm1, (%%"REG_a", %1, 2) \n\t" // X
// mm2=3(112) mm3=6(11) mm4=5 mm5=4(11) mm6=6 mm7=9
PAVGB((%%ecx), %%mm2) // 112 4 /8
"movq (%%eax, %1, 4), %%mm0 \n\t" // 1
PAVGB((%%REGc), %%mm2) // 112 4 /8
"movq (%%"REG_a", %1, 4), %%mm0 \n\t" // 1
PAVGB(%%mm0, %%mm6) // 1 1 /2
PAVGB(%%mm7, %%mm6) // 1 12 /4
PAVGB(%%mm2, %%mm6) // 1122424 /4
"movq %%mm6, (%%ecx) \n\t" // X
"movq %%mm6, (%%"REG_c") \n\t" // X
// mm0=8 mm3=6(11) mm4=5 mm5=4(11) mm7=9
PAVGB(%%mm7, %%mm5) // 11 2 /4
PAVGB(%%mm7, %%mm5) // 11 6 /8
PAVGB(%%mm3, %%mm0) // 112 /4
PAVGB(%%mm0, %%mm5) // 112246 /16
"movq %%mm5, (%%eax, %1, 4) \n\t" // X
"subl %1, %0 \n\t"
"movq %%mm5, (%%"REG_a", %1, 4) \n\t" // X
"sub %1, %0 \n\t"
:
: "r" (src), "r" (stride), "m" (c->pQPb)
: "%eax", "%ecx"
: "r" (src), "r" ((long)stride), "m" (c->pQPb)
: "%"REG_a, "%"REG_c
);
#else
const int l1= stride;
......@@ -364,8 +386,8 @@ static inline void RENAME(vertRK1Filter)(uint8_t *src, int stride, int QP)
asm volatile(
"pxor %%mm7, %%mm7 \n\t" // 0
"movq "MANGLE(b80)", %%mm6 \n\t" // MIN_SIGNED_BYTE
"leal (%0, %1), %%eax \n\t"
"leal (%%eax, %1, 4), %%ecx \n\t"
"leal (%0, %1), %%"REG_a" \n\t"
"leal (%%"REG_a", %1, 4), %%"REG_c" \n\t"
// 0 1 2 3 4 5 6 7 8 9
// %0 eax eax+%1 eax+2%1 %0+4%1 ecx ecx+%1 ecx+2%1 %0+8%1 ecx+4%1
"movq "MANGLE(pQPb)", %%mm0 \n\t" // QP,..., QP
......@@ -375,7 +397,7 @@ static inline void RENAME(vertRK1Filter)(uint8_t *src, int stride, int QP)
"pand "MANGLE(b3F)", %%mm0 \n\t" // QP/4,..., QP/4
"paddusb %%mm1, %%mm0 \n\t" // QP*1.25 ...
"movq (%0, %1, 4), %%mm2 \n\t" // line 4
"movq (%%ecx), %%mm3 \n\t" // line 5
"movq (%%"REG_c"), %%mm3 \n\t" // line 5
"movq %%mm2, %%mm4 \n\t" // line 4
"pcmpeqb %%mm5, %%mm5 \n\t" // -1
"pxor %%mm2, %%mm5 \n\t" // -line 4 - 1
......@@ -393,32 +415,32 @@ static inline void RENAME(vertRK1Filter)(uint8_t *src, int stride, int QP)
// "psubb %%mm6, %%mm2 \n\t"
"movq %%mm2, (%0,%1, 4) \n\t"
"movq (%%ecx), %%mm2 \n\t"
"movq (%%"REG_c"), %%mm2 \n\t"
// "paddb %%mm6, %%mm2 \n\t" // line 5 + 0x80
"psubb %%mm5, %%mm2 \n\t"
// "psubb %%mm6, %%mm2 \n\t"
"movq %%mm2, (%%ecx) \n\t"
"movq %%mm2, (%%"REG_c") \n\t"
"paddb %%mm6, %%mm5 \n\t"
"psrlw $2, %%mm5 \n\t"
"pand "MANGLE(b3F)", %%mm5 \n\t"
"psubb "MANGLE(b20)", %%mm5 \n\t" // (l5-l4)/8
"movq (%%eax, %1, 2), %%mm2 \n\t"
"movq (%%"REG_a", %1, 2), %%mm2 \n\t"
"paddb %%mm6, %%mm2 \n\t" // line 3 + 0x80
"paddsb %%mm5, %%mm2 \n\t"
"psubb %%mm6, %%mm2 \n\t"
"movq %%mm2, (%%eax, %1, 2) \n\t"
"movq %%mm2, (%%"REG_a", %1, 2) \n\t"
"movq (%%ecx, %1), %%mm2 \n\t"
"movq (%%"REG_c", %1), %%mm2 \n\t"
"paddb %%mm6, %%mm2 \n\t" // line 6 + 0x80
"psubsb %%mm5, %%mm2 \n\t"
"psubb %%mm6, %%mm2 \n\t"
"movq %%mm2, (%%ecx, %1) \n\t"
"movq %%mm2, (%%"REG_c", %1) \n\t"
:
: "r" (src), "r" (stride)
: "%eax", "%ecx"
: "r" (src), "r" ((long)stride)
: "%"REG_a, "%"REG_c
);
#else
const int l1= stride;
......@@ -464,18 +486,18 @@ static inline void RENAME(vertX1Filter)(uint8_t *src, int stride, PPContext *co)
asm volatile(
"pxor %%mm7, %%mm7 \n\t" // 0
"leal (%0, %1), %%eax \n\t"
"leal (%%eax, %1, 4), %%ecx \n\t"
"lea (%0, %1), %%"REG_a" \n\t"
"lea (%%"REG_a", %1, 4), %%"REG_c" \n\t"
// 0 1 2 3 4 5 6 7 8 9
// %0 eax eax+%1 eax+2%1 %0+4%1 ecx ecx+%1 ecx+2%1 %0+8%1 ecx+4%1
"movq (%%eax, %1, 2), %%mm0 \n\t" // line 3
"movq (%%"REG_a", %1, 2), %%mm0 \n\t" // line 3
"movq (%0, %1, 4), %%mm1 \n\t" // line 4
"movq %%mm1, %%mm2 \n\t" // line 4
"psubusb %%mm0, %%mm1 \n\t"
"psubusb %%mm2, %%mm0 \n\t"
"por %%mm1, %%mm0 \n\t" // |l2 - l3|
"movq (%%ecx), %%mm3 \n\t" // line 5
"movq (%%ecx, %1), %%mm4 \n\t" // line 6
"movq (%%"REG_c"), %%mm3 \n\t" // line 5
"movq (%%"REG_c", %1), %%mm4 \n\t" // line 6
"movq %%mm3, %%mm5 \n\t" // line 5
"psubusb %%mm4, %%mm3 \n\t"
"psubusb %%mm5, %%mm4 \n\t"
......@@ -507,43 +529,43 @@ static inline void RENAME(vertX1Filter)(uint8_t *src, int stride, PPContext *co)
"pxor %%mm2, %%mm0 \n\t"
"movq %%mm0, (%0, %1, 4) \n\t" // line 4
"movq (%%ecx), %%mm0 \n\t" // line 5
"movq (%%"REG_c"), %%mm0 \n\t" // line 5
"pxor %%mm2, %%mm0 \n\t" //(l4 - l5) <= 0 ? -l5-1 : l5
"paddusb %%mm3, %%mm0 \n\t"
"pxor %%mm2, %%mm0 \n\t"
"movq %%mm0, (%%ecx) \n\t" // line 5
"movq %%mm0, (%%"REG_c") \n\t" // line 5
PAVGB(%%mm7, %%mm1) // d/4
"movq (%%eax, %1, 2), %%mm0 \n\t" // line 3
"movq (%%"REG_a", %1, 2), %%mm0 \n\t" // line 3
"pxor %%mm2, %%mm0 \n\t" //(l4 - l5) <= 0 ? -l4-1 : l4
"psubusb %%mm1, %%mm0 \n\t"
"pxor %%mm2, %%mm0 \n\t"
"movq %%mm0, (%%eax, %1, 2) \n\t" // line 3
"movq %%mm0, (%%"REG_a", %1, 2) \n\t" // line 3
"movq (%%ecx, %1), %%mm0 \n\t" // line 6
"movq (%%"REG_c", %1), %%mm0 \n\t" // line 6
"pxor %%mm2, %%mm0 \n\t" //(l4 - l5) <= 0 ? -l5-1 : l5
"paddusb %%mm1, %%mm0 \n\t"
"pxor %%mm2, %%mm0 \n\t"
"movq %%mm0, (%%ecx, %1) \n\t" // line 6
"movq %%mm0, (%%"REG_c", %1) \n\t" // line 6
PAVGB(%%mm7, %%mm1) // d/8
"movq (%%eax, %1), %%mm0 \n\t" // line 2
"movq (%%"REG_a", %1), %%mm0 \n\t" // line 2
"pxor %%mm2, %%mm0 \n\t" //(l4 - l5) <= 0 ? -l2-1 : l2
"psubusb %%mm1, %%mm0 \n\t"
"pxor %%mm2, %%mm0 \n\t"
"movq %%mm0, (%%eax, %1) \n\t" // line 2
"movq %%mm0, (%%"REG_a", %1) \n\t" // line 2
"movq (%%ecx, %1, 2), %%mm0 \n\t" // line 7
"movq (%%"REG_c", %1, 2), %%mm0 \n\t" // line 7
"pxor %%mm2, %%mm0 \n\t" //(l4 - l5) <= 0 ? -l7-1 : l7
"paddusb %%mm1, %%mm0 \n\t"
"pxor %%mm2, %%mm0 \n\t"
"movq %%mm0, (%%ecx, %1, 2) \n\t" // line 7
"movq %%mm0, (%%"REG_c", %1, 2) \n\t" // line 7
:
: "r" (src), "r" (stride), "m" (co->pQPb)
: "%eax", "%ecx"
: "r" (src), "r" ((long)stride), "m" (co->pQPb)
: "%"REG_a, "%"REG_c
);
#else
......@@ -608,8 +630,8 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
#if 0 //sligtly more accurate and slightly slower
"pxor %%mm7, %%mm7 \n\t" // 0
"leal (%0, %1), %%eax \n\t"
"leal (%%eax, %1, 4), %%ecx \n\t"
"lea (%0, %1), %%"REG_a" \n\t"
"lea (%%"REG_a", %1, 4), %%"REG_c" \n\t"
// 0 1 2 3 4 5 6 7
// %0 %0+%1 %0+2%1 eax+2%1 %0+4%1 eax+4%1 ecx+%1 ecx+2%1
// %0 eax eax+%1 eax+2%1 %0+4%1 ecx ecx+%1 ecx+2%1
......@@ -622,8 +644,8 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
PAVGB(%%mm1, %%mm0) // ~(l2 + 2l0)/4
PAVGB(%%mm2, %%mm0) // ~(5l2 + 2l0)/8
"movq (%%eax), %%mm1 \n\t" // l1
"movq (%%eax, %1, 2), %%mm3 \n\t" // l3
"movq (%%"REG_a"), %%mm1 \n\t" // l1
"movq (%%"REG_a", %1, 2), %%mm3 \n\t" // l3
"movq %%mm1, %%mm4 \n\t" // l1
PAVGB(%%mm7, %%mm1) // ~l1/2
PAVGB(%%mm3, %%mm1) // ~(l1 + 2l3)/4
......@@ -641,7 +663,7 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
PAVGB(%%mm2, %%mm0) // ~(l4 + 2l2)/4
PAVGB(%%mm4, %%mm0) // ~(5l4 + 2l2)/8
"movq (%%ecx), %%mm2 \n\t" // l5
"movq (%%"REG_c"), %%mm2 \n\t" // l5
"movq %%mm3, %%mm5 \n\t" // l3
PAVGB(%%mm7, %%mm3) // ~l3/2
PAVGB(%%mm2, %%mm3) // ~(l3 + 2l5)/4
......@@ -654,13 +676,13 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
"pcmpeqb %%mm7, %%mm0 \n\t" // SIGN(2l2 - 5l3 + 5l4 - 2l5)
// mm0= SIGN(menergy), mm1= |lenergy|, mm2= l5, mm3= |menergy|, mm4=l4, mm5= l3, mm7=0
"movq (%%ecx, %1), %%mm6 \n\t" // l6
"movq (%%"REG_c", %1), %%mm6 \n\t" // l6
"movq %%mm6, %%mm5 \n\t" // l6
PAVGB(%%mm7, %%mm6) // ~l6/2
PAVGB(%%mm4, %%mm6) // ~(l6 + 2l4)/4
PAVGB(%%mm5, %%mm6) // ~(5l6 + 2l4)/8
"movq (%%ecx, %1, 2), %%mm5 \n\t" // l7
"movq (%%"REG_c", %1, 2), %%mm5 \n\t" // l7
"movq %%mm2, %%mm4 \n\t" // l5
PAVGB(%%mm7, %%mm2) // ~l5/2
PAVGB(%%mm5, %%mm2) // ~(l5 + 2l7)/4
......@@ -687,7 +709,7 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
"paddusb %%mm1, %%mm3 \n\t"
// "paddusb "MANGLE(b01)", %%mm3 \n\t"
"movq (%%eax, %1, 2), %%mm6 \n\t" //l3
"movq (%%"REG_a", %1, 2), %%mm6 \n\t" //l3
"movq (%0, %1, 4), %%mm5 \n\t" //l4
"movq (%0, %1, 4), %%mm4 \n\t" //l4
"psubusb %%mm6, %%mm5 \n\t"
......@@ -701,7 +723,7 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
"psubusb "MANGLE(b01)", %%mm3 \n\t"
PAVGB(%%mm7, %%mm3)
"movq (%%eax, %1, 2), %%mm0 \n\t"
"movq (%%"REG_a", %1, 2), %%mm0 \n\t"
"movq (%0, %1, 4), %%mm2 \n\t"
"pxor %%mm6, %%mm0 \n\t"
"pxor %%mm6, %%mm2 \n\t"
......@@ -709,36 +731,36 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
"paddb %%mm3, %%mm2 \n\t"
"pxor %%mm6, %%mm0 \n\t"
"pxor %%mm6, %%mm2 \n\t"
"movq %%mm0, (%%eax, %1, 2) \n\t"
"movq %%mm0, (%%"REG_a", %1, 2) \n\t"
"movq %%mm2, (%0, %1, 4) \n\t"
#endif
"leal (%0, %1), %%eax \n\t"
"lea (%0, %1), %%"REG_a" \n\t"
"pcmpeqb %%mm6, %%mm6 \n\t" // -1
// 0 1 2 3 4 5 6 7
// %0 %0+%1 %0+2%1 eax+2%1 %0+4%1 eax+4%1 ecx+%1 ecx+2%1
// %0 eax eax+%1 eax+2%1 %0+4%1 ecx ecx+%1 ecx+2%1
"movq (%%eax, %1, 2), %%mm1 \n\t" // l3
"movq (%%"REG_a", %1, 2), %%mm1 \n\t" // l3
"movq (%0, %1, 4), %%mm0 \n\t" // l4
"pxor %%mm6, %%mm1 \n\t" // -l3-1
PAVGB(%%mm1, %%mm0) // -q+128 = (l4-l3+256)/2
// mm1=-l3-1, mm0=128-q
"movq (%%eax, %1, 4), %%mm2 \n\t" // l5
"movq (%%eax, %1), %%mm3 \n\t" // l2
"movq (%%"REG_a", %1, 4), %%mm2 \n\t" // l5
"movq (%%"REG_a", %1), %%mm3 \n\t" // l2
"pxor %%mm6, %%mm2 \n\t" // -l5-1
"movq %%mm2, %%mm5 \n\t" // -l5-1
"movq "MANGLE(b80)", %%mm4 \n\t" // 128
"leal (%%eax, %1, 4), %%ecx \n\t"
"lea (%%"REG_a", %1, 4), %%"REG_c" \n\t"
PAVGB(%%mm3, %%mm2) // (l2-l5+256)/2
PAVGB(%%mm0, %%mm4) // ~(l4-l3)/4 + 128
PAVGB(%%mm2, %%mm4) // ~(l2-l5)/4 +(l4-l3)/8 + 128
PAVGB(%%mm0, %%mm4) // ~(l2-l5)/8 +5(l4-l3)/16 + 128
// mm1=-l3-1, mm0=128-q, mm3=l2, mm4=menergy/16 + 128, mm5= -l5-1
"movq (%%eax), %%mm2 \n\t" // l1
"movq (%%"REG_a"), %%mm2 \n\t" // l1
"pxor %%mm6, %%mm2 \n\t" // -l1-1
PAVGB(%%mm3, %%mm2) // (l2-l1+256)/2
PAVGB((%0), %%mm1) // (l0-l3+256)/2
......@@ -748,8 +770,8 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
PAVGB(%%mm2, %%mm3) // ~(l0-l3)/8 +5(l2-l1)/16 + 128
// mm0=128-q, mm3=lenergy/16 + 128, mm4= menergy/16 + 128, mm5= -l5-1
PAVGB((%%ecx, %1), %%mm5) // (l6-l5+256)/2
"movq (%%ecx, %1, 2), %%mm1 \n\t" // l7
PAVGB((%%REGc, %1), %%mm5) // (l6-l5+256)/2
"movq (%%"REG_c", %1, 2), %%mm1 \n\t" // l7
"pxor %%mm6, %%mm1 \n\t" // -l7-1
PAVGB((%0, %1, 4), %%mm1) // (l4-l7+256)/2
"movq "MANGLE(b80)", %%mm2 \n\t" // 128
......@@ -798,7 +820,7 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
"pxor %%mm1, %%mm7 \n\t" // SIGN(d*q)
"pand %%mm7, %%mm4 \n\t"
"movq (%%eax, %1, 2), %%mm0 \n\t"
"movq (%%"REG_a", %1, 2), %%mm0 \n\t"
"movq (%0, %1, 4), %%mm2 \n\t"
"pxor %%mm1, %%mm0 \n\t"
"pxor %%mm1, %%mm2 \n\t"
......@@ -806,12 +828,12 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
"psubb %%mm4, %%mm2 \n\t"
"pxor %%mm1, %%mm0 \n\t"
"pxor %%mm1, %%mm2 \n\t"
"movq %%mm0, (%%eax, %1, 2) \n\t"
"movq %%mm0, (%%"REG_a", %1, 2) \n\t"
"movq %%mm2, (%0, %1, 4) \n\t"
:
: "r" (src), "r" (stride), "m" (c->pQPb)
: "%eax", "%ecx"
: "r" (src), "r" ((long)stride), "m" (c->pQPb)
: "%"REG_a, "%"REG_c
);
/*
......@@ -882,8 +904,8 @@ src-=8;
src+= stride*4;
asm volatile(
"pxor %%mm7, %%mm7 \n\t"
"leal -40(%%esp), %%ecx \n\t" // make space for 4 8-byte vars
"andl $0xFFFFFFF8, %%ecx \n\t" // align
"lea -40(%%"REG_SP"), %%"REG_c" \n\t" // make space for 4 8-byte vars
"and "ALIGN_MASK", %%"REG_c" \n\t" // align
// 0 1 2 3 4 5 6 7
// %0 %0+%1 %0+2%1 eax+2%1 %0+4%1 eax+4%1 edx+%1 edx+2%1
// %0 eax eax+%1 eax+2%1 %0+4%1 edx edx+%1 edx+2%1
......@@ -894,12 +916,12 @@ src-=8;
"punpckhbw %%mm7, %%mm1 \n\t" // high part of line 0
"movq (%0, %1), %%mm2 \n\t"
"leal (%0, %1, 2), %%eax \n\t"
"lea (%0, %1, 2), %%"REG_a" \n\t"
"movq %%mm2, %%mm3 \n\t"
"punpcklbw %%mm7, %%mm2 \n\t" // low part of line 1
"punpckhbw %%mm7, %%mm3 \n\t" // high part of line 1
"movq (%%eax), %%mm4 \n\t"
"movq (%%"REG_a"), %%mm4 \n\t"
"movq %%mm4, %%mm5 \n\t"
"punpcklbw %%mm7, %%mm4 \n\t" // low part of line 2
"punpckhbw %%mm7, %%mm5 \n\t" // high part of line 2
......@@ -916,7 +938,7 @@ src-=8;
"psubw %%mm2, %%mm0 \n\t" // 2L0 - 5L1 + 5L2
"psubw %%mm3, %%mm1 \n\t" // 2H0 - 5H1 + 5H2
"movq (%%eax, %1), %%mm2 \n\t"
"movq (%%"REG_a", %1), %%mm2 \n\t"
"movq %%mm2, %%mm3 \n\t"
"punpcklbw %%mm7, %%mm2 \n\t" // L3
"punpckhbw %%mm7, %%mm3 \n\t" // H3
......@@ -925,24 +947,24 @@ src-=8;
"psubw %%mm3, %%mm1 \n\t" // 2H0 - 5H1 + 5H2 - H3
"psubw %%mm2, %%mm0 \n\t" // 2L0 - 5L1 + 5L2 - 2L3
"psubw %%mm3, %%mm1 \n\t" // 2H0 - 5H1 + 5H2 - 2H3
"movq %%mm0, (%%ecx) \n\t" // 2L0 - 5L1 + 5L2 - 2L3
"movq %%mm1, 8(%%ecx) \n\t" // 2H0 - 5H1 + 5H2 - 2H3
"movq %%mm0, (%%"REG_c") \n\t" // 2L0 - 5L1 + 5L2 - 2L3
"movq %%mm1, 8(%%"REG_c") \n\t" // 2H0 - 5H1 + 5H2 - 2H3
"movq (%%eax, %1, 2), %%mm0 \n\t"
"movq (%%"REG_a", %1, 2), %%mm0 \n\t"
"movq %%mm0, %%mm1 \n\t"
"punpcklbw %%mm7, %%mm0 \n\t" // L4
"punpckhbw %%mm7, %%mm1 \n\t" // H4
"psubw %%mm0, %%mm2 \n\t" // L3 - L4
"psubw %%mm1, %%mm3 \n\t" // H3 - H4
"movq %%mm2, 16(%%ecx) \n\t" // L3 - L4
"movq %%mm3, 24(%%ecx) \n\t" // H3 - H4
"movq %%mm2, 16(%%"REG_c") \n\t" // L3 - L4
"movq %%mm3, 24(%%"REG_c") \n\t" // H3 - H4
"paddw %%mm4, %%mm4 \n\t" // 2L2
"paddw %%mm5, %%mm5 \n\t" // 2H2
"psubw %%mm2, %%mm4 \n\t" // 2L2 - L3 + L4
"psubw %%mm3, %%mm5 \n\t" // 2H2 - H3 + H4
"leal (%%eax, %1), %0 \n\t"
"lea (%%"REG_a", %1), %0 \n\t"
"psllw $2, %%mm2 \n\t" // 4L3 - 4L4
"psllw $2, %%mm3 \n\t" // 4H3 - 4H4
"psubw %%mm2, %%mm4 \n\t" // 2L2 - 5L3 + 5L4
......@@ -957,10 +979,10 @@ src-=8;
"psubw %%mm2, %%mm4 \n\t" // 2L2 - 5L3 + 5L4 - 2L5
"psubw %%mm3, %%mm5 \n\t" // 2H2 - 5H3 + 5H4 - 2H5
"movq (%%eax, %1, 4), %%mm6 \n\t"
"movq (%%"REG_a", %1, 4), %%mm6 \n\t"
"punpcklbw %%mm7, %%mm6 \n\t" // L6
"psubw %%mm6, %%mm2 \n\t" // L5 - L6
"movq (%%eax, %1, 4), %%mm6 \n\t"
"movq (%%"REG_a", %1, 4), %%mm6 \n\t"
"punpckhbw %%mm7, %%mm6 \n\t" // H6
"psubw %%mm6, %%mm3 \n\t" // H5 - H6
......@@ -984,8 +1006,8 @@ src-=8;
"psubw %%mm2, %%mm0 \n\t" // 2L4 - 5L5 + 5L6 - 2L7
"psubw %%mm3, %%mm1 \n\t" // 2H4 - 5H5 + 5H6 - 2H7
"movq (%%ecx), %%mm2 \n\t" // 2L0 - 5L1 + 5L2 - 2L3
"movq 8(%%ecx), %%mm3 \n\t" // 2H0 - 5H1 + 5H2 - 2H3
"movq (%%"REG_c"), %%mm2 \n\t" // 2L0 - 5L1 + 5L2 - 2L3
"movq 8(%%"REG_c"), %%mm3 \n\t" // 2H0 - 5H1 + 5H2 - 2H3
#ifdef HAVE_MMX2
"movq %%mm7, %%mm6 \n\t" // 0
......@@ -1063,8 +1085,8 @@ src-=8;
"psrlw $6, %%mm4 \n\t"
"psrlw $6, %%mm5 \n\t"
"movq 16(%%ecx), %%mm0 \n\t" // L3 - L4
"movq 24(%%ecx), %%mm1 \n\t" // H3 - H4
"movq 16(%%"REG_c"), %%mm0 \n\t" // L3 - L4
"movq 24(%%"REG_c"), %%mm1 \n\t" // H3 - H4
"pxor %%mm2, %%mm2 \n\t"
"pxor %%mm3, %%mm3 \n\t"
......@@ -1107,8 +1129,8 @@ src-=8;
"movq %%mm0, (%0, %1) \n\t"
: "+r" (src)
: "r" (stride), "m" (c->pQPb)
: "%eax", "%ecx"
: "r" ((long)stride), "m" (c->pQPb)
: "%"REG_a, "%"REG_c
);
#else
const int l1= stride;
......@@ -1171,20 +1193,20 @@ static inline void RENAME(dering)(uint8_t src[], int stride, PPContext *c)
"packuswb %%mm0, %%mm0 \n\t"
"movq %%mm0, %3 \n\t"
"leal (%0, %1), %%eax \n\t"
"leal (%%eax, %1, 4), %%edx \n\t"
"lea (%0, %1), %%"REG_a" \n\t"
"lea (%%"REG_a", %1, 4), %%"REG_d" \n\t"
// 0 1 2 3 4 5 6 7 8 9
// %0 eax eax+%1 eax+2%1 %0+4%1 edx edx+%1 edx+2%1 %0+8%1 edx+4%1
#undef FIND_MIN_MAX
#ifdef HAVE_MMX2
#define FIND_MIN_MAX(addr)\
#define REAL_FIND_MIN_MAX(addr)\
"movq " #addr ", %%mm0 \n\t"\
"pminub %%mm0, %%mm7 \n\t"\
"pmaxub %%mm0, %%mm6 \n\t"
#else
#define FIND_MIN_MAX(addr)\
#define REAL_FIND_MIN_MAX(addr)\
"movq " #addr ", %%mm0 \n\t"\
"movq %%mm7, %%mm1 \n\t"\
"psubusb %%mm0, %%mm6 \n\t"\
......@@ -1192,14 +1214,15 @@ static inline void RENAME(dering)(uint8_t src[], int stride, PPContext *c)
"psubusb %%mm0, %%mm1 \n\t"\
"psubb %%mm1, %%mm7 \n\t"
#endif
#define FIND_MIN_MAX(addr) REAL_FIND_MIN_MAX(addr)
FIND_MIN_MAX((%%eax))
FIND_MIN_MAX((%%eax, %1))
FIND_MIN_MAX((%%eax, %1, 2))
FIND_MIN_MAX((%%REGa))
FIND_MIN_MAX((%%REGa, %1))
FIND_MIN_MAX((%%REGa, %1, 2))
FIND_MIN_MAX((%0, %1, 4))
FIND_MIN_MAX((%%edx))
FIND_MIN_MAX((%%edx, %1))
FIND_MIN_MAX((%%edx, %1, 2))
FIND_MIN_MAX((%%REGd))
FIND_MIN_MAX((%%REGd, %1))
FIND_MIN_MAX((%%REGd, %1, 2))
FIND_MIN_MAX((%0, %1, 8))
"movq %%mm7, %%mm4 \n\t"
......@@ -1252,13 +1275,13 @@ FIND_MIN_MAX((%0, %1, 8))
"movd %%mm6, %%ecx \n\t"
"cmpb "MANGLE(deringThreshold)", %%cl \n\t"
" jb 1f \n\t"
"leal -24(%%esp), %%ecx \n\t"
"andl $0xFFFFFFF8, %%ecx \n\t"
"lea -24(%%"REG_SP"), %%"REG_c" \n\t"
"and "ALIGN_MASK", %%"REG_c" \n\t"
PAVGB(%%mm0, %%mm7) // a=(max + min)/2
"punpcklbw %%mm7, %%mm7 \n\t"
"punpcklbw %%mm7, %%mm7 \n\t"
"punpcklbw %%mm7, %%mm7 \n\t"
"movq %%mm7, (%%ecx) \n\t"
"movq %%mm7, (%%"REG_c") \n\t"
"movq (%0), %%mm0 \n\t" // L10
"movq %%mm0, %%mm1 \n\t" // L10
......@@ -1283,13 +1306,13 @@ FIND_MIN_MAX((%0, %1, 8))
"paddb %%mm2, %%mm0 \n\t"
"paddb %%mm3, %%mm0 \n\t"
"movq (%%eax), %%mm2 \n\t" // L11
"movq (%%"REG_a"), %%mm2 \n\t" // L11
"movq %%mm2, %%mm3 \n\t" // L11
"movq %%mm2, %%mm4 \n\t" // L11
"psllq $8, %%mm3 \n\t"
"psrlq $8, %%mm4 \n\t"
"movd -4(%%eax), %%mm5 \n\t"
"movd 8(%%eax), %%mm6 \n\t"
"movd -4(%%"REG_a"), %%mm5 \n\t"
"movd 8(%%"REG_a"), %%mm6 \n\t"
"psrlq $24, %%mm5 \n\t"
"psllq $56, %%mm6 \n\t"
"por %%mm5, %%mm3 \n\t" // L01
......@@ -1306,7 +1329,7 @@ FIND_MIN_MAX((%0, %1, 8))
"paddb %%mm4, %%mm2 \n\t"
"paddb %%mm5, %%mm2 \n\t"
// 0, 2, 3, 1
#define DERING_CORE(dst,src,ppsx,psx,sx,pplx,plx,lx,t0,t1) \
#define REAL_DERING_CORE(dst,src,ppsx,psx,sx,pplx,plx,lx,t0,t1) \
"movq " #src ", " #sx " \n\t" /* src[0] */\
"movq " #sx ", " #lx " \n\t" /* src[0] */\
"movq " #sx ", " #t0 " \n\t" /* src[0] */\
......@@ -1322,8 +1345,8 @@ FIND_MIN_MAX((%0, %1, 8))
PAVGB(t0, lx) /* (src[-1] + src[+1])/2 */\
PAVGB(sx, lx) /* (src[-1] + 2src[0] + src[+1])/4 */\
PAVGB(lx, pplx) \
"movq " #lx ", 8(%%ecx) \n\t"\
"movq (%%ecx), " #lx " \n\t"\
"movq " #lx ", 8(%%"REG_c") \n\t"\
"movq (%%"REG_c"), " #lx " \n\t"\
"psubusb " #lx ", " #t1 " \n\t"\
"psubusb " #lx ", " #t0 " \n\t"\
"psubusb " #lx ", " #sx " \n\t"\
......@@ -1350,8 +1373,10 @@ FIND_MIN_MAX((%0, %1, 8))
"pandn " #dst ", " #ppsx " \n\t"\
"por " #pplx ", " #ppsx " \n\t"\
"movq " #ppsx ", " #dst " \n\t"\
"movq 8(%%ecx), " #lx " \n\t"
"movq 8(%%"REG_c"), " #lx " \n\t"
#define DERING_CORE(dst,src,ppsx,psx,sx,pplx,plx,lx,t0,t1) \
REAL_DERING_CORE(dst,src,ppsx,psx,sx,pplx,plx,lx,t0,t1)
/*
0000000
1111111
......@@ -1368,18 +1393,18 @@ FIND_MIN_MAX((%0, %1, 8))
*/
//DERING_CORE(dst,src ,ppsx ,psx ,sx ,pplx ,plx ,lx ,t0 ,t1)
DERING_CORE((%%eax),(%%eax, %1) ,%%mm0,%%mm2,%%mm4,%%mm1,%%mm3,%%mm5,%%mm6,%%mm7)
DERING_CORE((%%eax, %1),(%%eax, %1, 2) ,%%mm2,%%mm4,%%mm0,%%mm3,%%mm5,%%mm1,%%mm6,%%mm7)
DERING_CORE((%%eax, %1, 2),(%0, %1, 4) ,%%mm4,%%mm0,%%mm2,%%mm5,%%mm1,%%mm3,%%mm6,%%mm7)
DERING_CORE((%0, %1, 4),(%%edx) ,%%mm0,%%mm2,%%mm4,%%mm1,%%mm3,%%mm5,%%mm6,%%mm7)
DERING_CORE((%%edx),(%%edx, %1) ,%%mm2,%%mm4,%%mm0,%%mm3,%%mm5,%%mm1,%%mm6,%%mm7)
DERING_CORE((%%edx, %1), (%%edx, %1, 2),%%mm4,%%mm0,%%mm2,%%mm5,%%mm1,%%mm3,%%mm6,%%mm7)
DERING_CORE((%%edx, %1, 2),(%0, %1, 8) ,%%mm0,%%mm2,%%mm4,%%mm1,%%mm3,%%mm5,%%mm6,%%mm7)
DERING_CORE((%0, %1, 8),(%%edx, %1, 4) ,%%mm2,%%mm4,%%mm0,%%mm3,%%mm5,%%mm1,%%mm6,%%mm7)
DERING_CORE((%%REGa),(%%REGa, %1) ,%%mm0,%%mm2,%%mm4,%%mm1,%%mm3,%%mm5,%%mm6,%%mm7)
DERING_CORE((%%REGa, %1),(%%REGa, %1, 2) ,%%mm2,%%mm4,%%mm0,%%mm3,%%mm5,%%mm1,%%mm6,%%mm7)
DERING_CORE((%%REGa, %1, 2),(%0, %1, 4) ,%%mm4,%%mm0,%%mm2,%%mm5,%%mm1,%%mm3,%%mm6,%%mm7)
DERING_CORE((%0, %1, 4),(%%REGd) ,%%mm0,%%mm2,%%mm4,%%mm1,%%mm3,%%mm5,%%mm6,%%mm7)
DERING_CORE((%%REGd),(%%REGd, %1) ,%%mm2,%%mm4,%%mm0,%%mm3,%%mm5,%%mm1,%%mm6,%%mm7)
DERING_CORE((%%REGd, %1), (%%REGd, %1, 2),%%mm4,%%mm0,%%mm2,%%mm5,%%mm1,%%mm3,%%mm6,%%mm7)
DERING_CORE((%%REGd, %1, 2),(%0, %1, 8) ,%%mm0,%%mm2,%%mm4,%%mm1,%%mm3,%%mm5,%%mm6,%%mm7)
DERING_CORE((%0, %1, 8),(%%REGd, %1, 4) ,%%mm2,%%mm4,%%mm0,%%mm3,%%mm5,%%mm1,%%mm6,%%mm7)
"1: \n\t"
: : "r" (src), "r" (stride), "m" (c->pQPb), "m"(c->pQPb2)
: "%eax", "%edx", "%ecx"
: : "r" (src), "r" ((long)stride), "m" (c->pQPb), "m"(c->pQPb2)
: "%"REG_a, "%"REG_d, "%"REG_c
);
#else
int y;
......@@ -1526,27 +1551,27 @@ static inline void RENAME(deInterlaceInterpolateLinear)(uint8_t src[], int strid
#if defined (HAVE_MMX2) || defined (HAVE_3DNOW)
src+= 4*stride;
asm volatile(
"leal (%0, %1), %%eax \n\t"
"leal (%%eax, %1, 4), %%ecx \n\t"
"lea (%0, %1), %%"REG_a" \n\t"
"lea (%%"REG_a", %1, 4), %%"REG_c" \n\t"
// 0 1 2 3 4 5 6 7 8 9
// %0 eax eax+%1 eax+2%1 %0+4%1 ecx ecx+%1 ecx+2%1 %0+8%1 ecx+4%1
"movq (%0), %%mm0 \n\t"
"movq (%%eax, %1), %%mm1 \n\t"
"movq (%%"REG_a", %1), %%mm1 \n\t"
PAVGB(%%mm1, %%mm0)
"movq %%mm0, (%%eax) \n\t"
"movq %%mm0, (%%"REG_a") \n\t"
"movq (%0, %1, 4), %%mm0 \n\t"
PAVGB(%%mm0, %%mm1)
"movq %%mm1, (%%eax, %1, 2) \n\t"
"movq (%%ecx, %1), %%mm1 \n\t"
"movq %%mm1, (%%"REG_a", %1, 2) \n\t"
"movq (%%"REG_c", %1), %%mm1 \n\t"
PAVGB(%%mm1, %%mm0)
"movq %%mm0, (%%ecx) \n\t"
"movq %%mm0, (%%"REG_c") \n\t"
"movq (%0, %1, 8), %%mm0 \n\t"
PAVGB(%%mm0, %%mm1)
"movq %%mm1, (%%ecx, %1, 2) \n\t"
"movq %%mm1, (%%"REG_c", %1, 2) \n\t"
: : "r" (src), "r" (stride)
: "%eax", "%ecx"
: : "r" (src), "r" ((long)stride)
: "%"REG_a, "%"REG_c
);
#else
int a, b, x;
......@@ -1579,15 +1604,15 @@ static inline void RENAME(deInterlaceInterpolateCubic)(uint8_t src[], int stride
#if defined (HAVE_MMX2) || defined (HAVE_3DNOW)
src+= stride*3;
asm volatile(
"leal (%0, %1), %%eax \n\t"
"leal (%%eax, %1, 4), %%edx \n\t"
"leal (%%edx, %1, 4), %%ecx \n\t"
"addl %1, %%ecx \n\t"
"lea (%0, %1), %%"REG_a" \n\t"
"lea (%%"REG_a", %1, 4), %%"REG_d" \n\t"
"lea (%%"REG_d", %1, 4), %%"REG_c" \n\t"
"add %1, %%"REG_c" \n\t"
"pxor %%mm7, %%mm7 \n\t"
// 0 1 2 3 4 5 6 7 8 9 10
// %0 eax eax+%1 eax+2%1 %0+4%1 edx edx+%1 edx+2%1 %0+8%1 edx+4%1 ecx
#define DEINT_CUBIC(a,b,c,d,e)\
#define REAL_DEINT_CUBIC(a,b,c,d,e)\
"movq " #a ", %%mm0 \n\t"\
"movq " #b ", %%mm1 \n\t"\
"movq " #d ", %%mm2 \n\t"\
......@@ -1608,14 +1633,15 @@ static inline void RENAME(deInterlaceInterpolateCubic)(uint8_t src[], int stride
"psubw %%mm2, %%mm3 \n\t" /* H(9b + 9d - a - e)/16 */\
"packuswb %%mm3, %%mm1 \n\t"\
"movq %%mm1, " #c " \n\t"
#define DEINT_CUBIC(a,b,c,d,e) REAL_DEINT_CUBIC(a,b,c,d,e)
DEINT_CUBIC((%0), (%%eax, %1), (%%eax, %1, 2), (%0, %1, 4), (%%edx, %1))
DEINT_CUBIC((%%eax, %1), (%0, %1, 4), (%%edx), (%%edx, %1), (%0, %1, 8))
DEINT_CUBIC((%0, %1, 4), (%%edx, %1), (%%edx, %1, 2), (%0, %1, 8), (%%ecx))
DEINT_CUBIC((%%edx, %1), (%0, %1, 8), (%%edx, %1, 4), (%%ecx), (%%ecx, %1, 2))
DEINT_CUBIC((%0), (%%REGa, %1), (%%REGa, %1, 2), (%0, %1, 4), (%%REGd, %1))
DEINT_CUBIC((%%REGa, %1), (%0, %1, 4), (%%REGd), (%%REGd, %1), (%0, %1, 8))
DEINT_CUBIC((%0, %1, 4), (%%REGd, %1), (%%REGd, %1, 2), (%0, %1, 8), (%%REGc))
DEINT_CUBIC((%%REGd, %1), (%0, %1, 8), (%%REGd, %1, 4), (%%REGc), (%%REGc, %1, 2))
: : "r" (src), "r" (stride)
: "%eax", "%edx", "ecx"
: : "r" (src), "r" ((long)stride)
: "%"REG_a, "%"REG_d, "%"REG_c
);
#else
int x;
......@@ -1643,14 +1669,14 @@ static inline void RENAME(deInterlaceFF)(uint8_t src[], int stride, uint8_t *tmp
#if defined (HAVE_MMX2) || defined (HAVE_3DNOW)
src+= stride*4;
asm volatile(
"leal (%0, %1), %%eax \n\t"
"leal (%%eax, %1, 4), %%edx \n\t"
"lea (%0, %1), %%"REG_a" \n\t"
"lea (%%"REG_a", %1, 4), %%"REG_d" \n\t"
"pxor %%mm7, %%mm7 \n\t"
"movq (%2), %%mm0 \n\t"
// 0 1 2 3 4 5 6 7 8 9 10
// %0 eax eax+%1 eax+2%1 %0+4%1 edx edx+%1 edx+2%1 %0+8%1 edx+4%1 ecx
#define DEINT_FF(a,b,c,d)\
#define REAL_DEINT_FF(a,b,c,d)\
"movq " #a ", %%mm1 \n\t"\
"movq " #b ", %%mm2 \n\t"\
"movq " #c ", %%mm3 \n\t"\
......@@ -1678,14 +1704,16 @@ static inline void RENAME(deInterlaceFF)(uint8_t src[], int stride, uint8_t *tmp
"packuswb %%mm4, %%mm1 \n\t"\
"movq %%mm1, " #b " \n\t"\
DEINT_FF((%0) , (%%eax) , (%%eax, %1), (%%eax, %1, 2))
DEINT_FF((%%eax, %1), (%%eax, %1, 2), (%0, %1, 4), (%%edx) )
DEINT_FF((%0, %1, 4), (%%edx) , (%%edx, %1), (%%edx, %1, 2))
DEINT_FF((%%edx, %1), (%%edx, %1, 2), (%0, %1, 8), (%%edx, %1, 4))
#define DEINT_FF(a,b,c,d) REAL_DEINT_FF(a,b,c,d)
DEINT_FF((%0) , (%%REGa) , (%%REGa, %1), (%%REGa, %1, 2))
DEINT_FF((%%REGa, %1), (%%REGa, %1, 2), (%0, %1, 4), (%%REGd) )
DEINT_FF((%0, %1, 4), (%%REGd) , (%%REGd, %1), (%%REGd, %1, 2))
DEINT_FF((%%REGd, %1), (%%REGd, %1, 2), (%0, %1, 8), (%%REGd, %1, 4))
"movq %%mm0, (%2) \n\t"
: : "r" (src), "r" (stride), "r"(tmp)
: "%eax", "%edx"
: : "r" (src), "r" ((long)stride), "r"(tmp)
: "%"REG_a, "%"REG_d
);
#else
int x;
......@@ -1721,15 +1749,15 @@ static inline void RENAME(deInterlaceL5)(uint8_t src[], int stride, uint8_t *tmp
#if defined (HAVE_MMX2) || defined (HAVE_3DNOW)
src+= stride*4;
asm volatile(
"leal (%0, %1), %%eax \n\t"
"leal (%%eax, %1, 4), %%edx \n\t"
"lea (%0, %1), %%"REG_a" \n\t"
"lea (%%"REG_a", %1, 4), %%"REG_d" \n\t"
"pxor %%mm7, %%mm7 \n\t"
"movq (%2), %%mm0 \n\t"
"movq (%3), %%mm1 \n\t"
// 0 1 2 3 4 5 6 7 8 9 10
// %0 eax eax+%1 eax+2%1 %0+4%1 edx edx+%1 edx+2%1 %0+8%1 edx+4%1 ecx
#define DEINT_L5(t1,t2,a,b,c)\
#define REAL_DEINT_L5(t1,t2,a,b,c)\
"movq " #a ", %%mm2 \n\t"\
"movq " #b ", %%mm3 \n\t"\
"movq " #c ", %%mm4 \n\t"\
......@@ -1762,19 +1790,21 @@ static inline void RENAME(deInterlaceL5)(uint8_t src[], int stride, uint8_t *tmp
"packuswb %%mm5, %%mm2 \n\t"\
"movq %%mm2, " #a " \n\t"\
DEINT_L5(%%mm0, %%mm1, (%0) , (%%eax) , (%%eax, %1) )
DEINT_L5(%%mm1, %%mm0, (%%eax) , (%%eax, %1) , (%%eax, %1, 2))
DEINT_L5(%%mm0, %%mm1, (%%eax, %1) , (%%eax, %1, 2), (%0, %1, 4) )
DEINT_L5(%%mm1, %%mm0, (%%eax, %1, 2), (%0, %1, 4) , (%%edx) )
DEINT_L5(%%mm0, %%mm1, (%0, %1, 4) , (%%edx) , (%%edx, %1) )
DEINT_L5(%%mm1, %%mm0, (%%edx) , (%%edx, %1) , (%%edx, %1, 2))
DEINT_L5(%%mm0, %%mm1, (%%edx, %1) , (%%edx, %1, 2), (%0, %1, 8) )
DEINT_L5(%%mm1, %%mm0, (%%edx, %1, 2), (%0, %1, 8) , (%%edx, %1, 4))
#define DEINT_L5(t1,t2,a,b,c) REAL_DEINT_L5(t1,t2,a,b,c)
DEINT_L5(%%mm0, %%mm1, (%0) , (%%REGa) , (%%REGa, %1) )
DEINT_L5(%%mm1, %%mm0, (%%REGa) , (%%REGa, %1) , (%%REGa, %1, 2))
DEINT_L5(%%mm0, %%mm1, (%%REGa, %1) , (%%REGa, %1, 2), (%0, %1, 4) )
DEINT_L5(%%mm1, %%mm0, (%%REGa, %1, 2), (%0, %1, 4) , (%%REGd) )
DEINT_L5(%%mm0, %%mm1, (%0, %1, 4) , (%%REGd) , (%%REGd, %1) )
DEINT_L5(%%mm1, %%mm0, (%%REGd) , (%%REGd, %1) , (%%REGd, %1, 2))
DEINT_L5(%%mm0, %%mm1, (%%REGd, %1) , (%%REGd, %1, 2), (%0, %1, 8) )
DEINT_L5(%%mm1, %%mm0, (%%REGd, %1, 2), (%0, %1, 8) , (%%REGd, %1, 4))
"movq %%mm0, (%2) \n\t"
"movq %%mm1, (%3) \n\t"
: : "r" (src), "r" (stride), "r"(tmp), "r"(tmp2)
: "%eax", "%edx"
: : "r" (src), "r" ((long)stride), "r"(tmp), "r"(tmp2)
: "%"REG_a, "%"REG_d
);
#else
int x;
......@@ -1821,49 +1851,49 @@ static inline void RENAME(deInterlaceBlendLinear)(uint8_t src[], int stride, uin
#if defined (HAVE_MMX2) || defined (HAVE_3DNOW)
src+= 4*stride;
asm volatile(
"leal (%0, %1), %%eax \n\t"
"leal (%%eax, %1, 4), %%edx \n\t"
"lea (%0, %1), %%"REG_a" \n\t"
"lea (%%"REG_a", %1, 4), %%"REG_d" \n\t"
// 0 1 2 3 4 5 6 7 8 9
// %0 eax eax+%1 eax+2%1 %0+4%1 edx edx+%1 edx+2%1 %0+8%1 edx+4%1
"movq (%2), %%mm0 \n\t" // L0
"movq (%%eax), %%mm1 \n\t" // L2
"movq (%%"REG_a"), %%mm1 \n\t" // L2
PAVGB(%%mm1, %%mm0) // L0+L2
"movq (%0), %%mm2 \n\t" // L1
PAVGB(%%mm2, %%mm0)
"movq %%mm0, (%0) \n\t"
"movq (%%eax, %1), %%mm0 \n\t" // L3
"movq (%%"REG_a", %1), %%mm0 \n\t" // L3
PAVGB(%%mm0, %%mm2) // L1+L3
PAVGB(%%mm1, %%mm2) // 2L2 + L1 + L3
"movq %%mm2, (%%eax) \n\t"
"movq (%%eax, %1, 2), %%mm2 \n\t" // L4
"movq %%mm2, (%%"REG_a") \n\t"
"movq (%%"REG_a", %1, 2), %%mm2 \n\t" // L4
PAVGB(%%mm2, %%mm1) // L2+L4
PAVGB(%%mm0, %%mm1) // 2L3 + L2 + L4
"movq %%mm1, (%%eax, %1) \n\t"
"movq %%mm1, (%%"REG_a", %1) \n\t"
"movq (%0, %1, 4), %%mm1 \n\t" // L5
PAVGB(%%mm1, %%mm0) // L3+L5
PAVGB(%%mm2, %%mm0) // 2L4 + L3 + L5
"movq %%mm0, (%%eax, %1, 2) \n\t"
"movq (%%edx), %%mm0 \n\t" // L6
"movq %%mm0, (%%"REG_a", %1, 2) \n\t"
"movq (%%"REG_d"), %%mm0 \n\t" // L6
PAVGB(%%mm0, %%mm2) // L4+L6
PAVGB(%%mm1, %%mm2) // 2L5 + L4 + L6
"movq %%mm2, (%0, %1, 4) \n\t"
"movq (%%edx, %1), %%mm2 \n\t" // L7
"movq (%%"REG_d", %1), %%mm2 \n\t" // L7
PAVGB(%%mm2, %%mm1) // L5+L7
PAVGB(%%mm0, %%mm1) // 2L6 + L5 + L7
"movq %%mm1, (%%edx) \n\t"
"movq (%%edx, %1, 2), %%mm1 \n\t" // L8
"movq %%mm1, (%%"REG_d") \n\t"
"movq (%%"REG_d", %1, 2), %%mm1 \n\t" // L8
PAVGB(%%mm1, %%mm0) // L6+L8
PAVGB(%%mm2, %%mm0) // 2L7 + L6 + L8
"movq %%mm0, (%%edx, %1) \n\t"
"movq %%mm0, (%%"REG_d", %1) \n\t"
"movq (%0, %1, 8), %%mm0 \n\t" // L9
PAVGB(%%mm0, %%mm2) // L7+L9
PAVGB(%%mm1, %%mm2) // 2L8 + L7 + L9
"movq %%mm2, (%%edx, %1, 2) \n\t"
"movq %%mm2, (%%"REG_d", %1, 2) \n\t"
"movq %%mm1, (%2) \n\t"
: : "r" (src), "r" (stride), "r" (tmp)
: "%eax", "%edx"
: : "r" (src), "r" ((long)stride), "r" (tmp)
: "%"REG_a, "%"REG_d
);
#else
int a, b, c, x;
......@@ -1923,62 +1953,62 @@ static inline void RENAME(deInterlaceMedian)(uint8_t src[], int stride)
src+= 4*stride;
#ifdef HAVE_MMX2
asm volatile(
"leal (%0, %1), %%eax \n\t"
"leal (%%eax, %1, 4), %%edx \n\t"
"lea (%0, %1), %%"REG_a" \n\t"
"lea (%%"REG_a", %1, 4), %%"REG_d" \n\t"
// 0 1 2 3 4 5 6 7 8 9
// %0 eax eax+%1 eax+2%1 %0+4%1 edx edx+%1 edx+2%1 %0+8%1 edx+4%1
"movq (%0), %%mm0 \n\t" //
"movq (%%eax, %1), %%mm2 \n\t" //
"movq (%%eax), %%mm1 \n\t" //
"movq (%%"REG_a", %1), %%mm2 \n\t" //
"movq (%%"REG_a"), %%mm1 \n\t" //
"movq %%mm0, %%mm3 \n\t"
"pmaxub %%mm1, %%mm0 \n\t" //
"pminub %%mm3, %%mm1 \n\t" //
"pmaxub %%mm2, %%mm1 \n\t" //
"pminub %%mm1, %%mm0 \n\t"
"movq %%mm0, (%%eax) \n\t"
"movq %%mm0, (%%"REG_a") \n\t"
"movq (%0, %1, 4), %%mm0 \n\t" //
"movq (%%eax, %1, 2), %%mm1 \n\t" //
"movq (%%"REG_a", %1, 2), %%mm1 \n\t" //
"movq %%mm2, %%mm3 \n\t"
"pmaxub %%mm1, %%mm2 \n\t" //
"pminub %%mm3, %%mm1 \n\t" //
"pmaxub %%mm0, %%mm1 \n\t" //
"pminub %%mm1, %%mm2 \n\t"
"movq %%mm2, (%%eax, %1, 2) \n\t"
"movq %%mm2, (%%"REG_a", %1, 2) \n\t"
"movq (%%edx), %%mm2 \n\t" //
"movq (%%edx, %1), %%mm1 \n\t" //
"movq (%%"REG_d"), %%mm2 \n\t" //
"movq (%%"REG_d", %1), %%mm1 \n\t" //
"movq %%mm2, %%mm3 \n\t"
"pmaxub %%mm0, %%mm2 \n\t" //
"pminub %%mm3, %%mm0 \n\t" //
"pmaxub %%mm1, %%mm0 \n\t" //
"pminub %%mm0, %%mm2 \n\t"
"movq %%mm2, (%%edx) \n\t"
"movq %%mm2, (%%"REG_d") \n\t"
"movq (%%edx, %1, 2), %%mm2 \n\t" //
"movq (%%"REG_d", %1, 2), %%mm2 \n\t" //
"movq (%0, %1, 8), %%mm0 \n\t" //
"movq %%mm2, %%mm3 \n\t"
"pmaxub %%mm0, %%mm2 \n\t" //
"pminub %%mm3, %%mm0 \n\t" //
"pmaxub %%mm1, %%mm0 \n\t" //
"pminub %%mm0, %%mm2 \n\t"
"movq %%mm2, (%%edx, %1, 2) \n\t"
"movq %%mm2, (%%"REG_d", %1, 2) \n\t"
: : "r" (src), "r" (stride)
: "%eax", "%edx"
: : "r" (src), "r" ((long)stride)
: "%"REG_a, "%"REG_d
);
#else // MMX without MMX2
asm volatile(
"leal (%0, %1), %%eax \n\t"
"leal (%%eax, %1, 4), %%edx \n\t"
"lea (%0, %1), %%"REG_a" \n\t"
"lea (%%"REG_a", %1, 4), %%"REG_d" \n\t"
// 0 1 2 3 4 5 6 7 8 9
// %0 eax eax+%1 eax+2%1 %0+4%1 edx edx+%1 edx+2%1 %0+8%1 edx+4%1
"pxor %%mm7, %%mm7 \n\t"
#define MEDIAN(a,b,c)\
#define REAL_MEDIAN(a,b,c)\
"movq " #a ", %%mm0 \n\t"\
"movq " #b ", %%mm2 \n\t"\
"movq " #c ", %%mm1 \n\t"\
......@@ -2001,14 +2031,15 @@ static inline void RENAME(deInterlaceMedian)(uint8_t src[], int stride)
"pand %%mm2, %%mm0 \n\t"\
"pand %%mm1, %%mm0 \n\t"\
"movq %%mm0, " #b " \n\t"
#define MEDIAN(a,b,c) REAL_MEDIAN(a,b,c)
MEDIAN((%0), (%%eax), (%%eax, %1))
MEDIAN((%%eax, %1), (%%eax, %1, 2), (%0, %1, 4))
MEDIAN((%0, %1, 4), (%%edx), (%%edx, %1))
MEDIAN((%%edx, %1), (%%edx, %1, 2), (%0, %1, 8))
MEDIAN((%0), (%%REGa), (%%REGa, %1))
MEDIAN((%%REGa, %1), (%%REGa, %1, 2), (%0, %1, 4))
MEDIAN((%0, %1, 4), (%%REGd), (%%REGd, %1))
MEDIAN((%%REGd, %1), (%%REGd, %1, 2), (%0, %1, 8))
: : "r" (src), "r" (stride)
: "%eax", "%edx"
: : "r" (src), "r" ((long)stride)
: "%"REG_a, "%"REG_d
);
#endif // MMX
#else
......@@ -2042,17 +2073,17 @@ MEDIAN((%%edx, %1), (%%edx, %1, 2), (%0, %1, 8))
static inline void RENAME(transpose1)(uint8_t *dst1, uint8_t *dst2, uint8_t *src, int srcStride)
{
asm(
"leal (%0, %1), %%eax \n\t"
"lea (%0, %1), %%"REG_a" \n\t"
// 0 1 2 3 4 5 6 7 8 9
// %0 eax eax+%1 eax+2%1 %0+4%1 edx edx+%1 edx+2%1 %0+8%1 edx+4%1
"movq (%0), %%mm0 \n\t" // 12345678
"movq (%%eax), %%mm1 \n\t" // abcdefgh
"movq (%%"REG_a"), %%mm1 \n\t" // abcdefgh
"movq %%mm0, %%mm2 \n\t" // 12345678
"punpcklbw %%mm1, %%mm0 \n\t" // 1a2b3c4d
"punpckhbw %%mm1, %%mm2 \n\t" // 5e6f7g8h
"movq (%%eax, %1), %%mm1 \n\t"
"movq (%%eax, %1, 2), %%mm3 \n\t"
"movq (%%"REG_a", %1), %%mm1 \n\t"
"movq (%%"REG_a", %1, 2), %%mm3 \n\t"
"movq %%mm1, %%mm4 \n\t"
"punpcklbw %%mm3, %%mm1 \n\t"
"punpckhbw %%mm3, %%mm4 \n\t"
......@@ -2079,16 +2110,16 @@ static inline void RENAME(transpose1)(uint8_t *dst1, uint8_t *dst2, uint8_t *src
"psrlq $32, %%mm1 \n\t"
"movd %%mm1, 112(%3) \n\t"
"leal (%%eax, %1, 4), %%eax \n\t"
"lea (%%"REG_a", %1, 4), %%"REG_a" \n\t"
"movq (%0, %1, 4), %%mm0 \n\t" // 12345678
"movq (%%eax), %%mm1 \n\t" // abcdefgh
"movq (%%"REG_a"), %%mm1 \n\t" // abcdefgh
"movq %%mm0, %%mm2 \n\t" // 12345678
"punpcklbw %%mm1, %%mm0 \n\t" // 1a2b3c4d
"punpckhbw %%mm1, %%mm2 \n\t" // 5e6f7g8h
"movq (%%eax, %1), %%mm1 \n\t"
"movq (%%eax, %1, 2), %%mm3 \n\t"
"movq (%%"REG_a", %1), %%mm1 \n\t"
"movq (%%"REG_a", %1, 2), %%mm3 \n\t"
"movq %%mm1, %%mm4 \n\t"
"punpcklbw %%mm3, %%mm1 \n\t"
"punpckhbw %%mm3, %%mm4 \n\t"
......@@ -2116,8 +2147,8 @@ static inline void RENAME(transpose1)(uint8_t *dst1, uint8_t *dst2, uint8_t *src
"movd %%mm1, 116(%3) \n\t"
:: "r" (src), "r" (srcStride), "r" (dst1), "r" (dst2)
: "%eax"
:: "r" (src), "r" ((long)srcStride), "r" (dst1), "r" (dst2)
: "%"REG_a
);
}
......@@ -2127,8 +2158,8 @@ static inline void RENAME(transpose1)(uint8_t *dst1, uint8_t *dst2, uint8_t *src
static inline void RENAME(transpose2)(uint8_t *dst, int dstStride, uint8_t *src)
{
asm(
"leal (%0, %1), %%eax \n\t"
"leal (%%eax, %1, 4), %%edx \n\t"
"lea (%0, %1), %%"REG_a" \n\t"
"lea (%%"REG_a",%1,4), %%"REG_d"\n\t"
// 0 1 2 3 4 5 6 7 8 9
// %0 eax eax+%1 eax+2%1 %0+4%1 edx edx+%1 edx+2%1 %0+8%1 edx+4%1
"movq (%2), %%mm0 \n\t" // 12345678
......@@ -2152,16 +2183,16 @@ static inline void RENAME(transpose2)(uint8_t *dst, int dstStride, uint8_t *src)
"movd %%mm0, (%0) \n\t"
"psrlq $32, %%mm0 \n\t"
"movd %%mm0, (%%eax) \n\t"
"movd %%mm3, (%%eax, %1) \n\t"
"movd %%mm0, (%%"REG_a") \n\t"
"movd %%mm3, (%%"REG_a", %1) \n\t"
"psrlq $32, %%mm3 \n\t"
"movd %%mm3, (%%eax, %1, 2) \n\t"
"movd %%mm3, (%%"REG_a", %1, 2) \n\t"
"movd %%mm2, (%0, %1, 4) \n\t"
"psrlq $32, %%mm2 \n\t"
"movd %%mm2, (%%edx) \n\t"
"movd %%mm1, (%%edx, %1) \n\t"
"movd %%mm2, (%%"REG_d") \n\t"
"movd %%mm1, (%%"REG_d", %1) \n\t"
"psrlq $32, %%mm1 \n\t"
"movd %%mm1, (%%edx, %1, 2) \n\t"
"movd %%mm1, (%%"REG_d", %1, 2) \n\t"
"movq 64(%2), %%mm0 \n\t" // 12345678
......@@ -2185,23 +2216,23 @@ static inline void RENAME(transpose2)(uint8_t *dst, int dstStride, uint8_t *src)
"movd %%mm0, 4(%0) \n\t"
"psrlq $32, %%mm0 \n\t"
"movd %%mm0, 4(%%eax) \n\t"
"movd %%mm3, 4(%%eax, %1) \n\t"
"movd %%mm0, 4(%%"REG_a") \n\t"
"movd %%mm3, 4(%%"REG_a", %1) \n\t"
"psrlq $32, %%mm3 \n\t"
"movd %%mm3, 4(%%eax, %1, 2) \n\t"
"movd %%mm3, 4(%%"REG_a", %1, 2) \n\t"
"movd %%mm2, 4(%0, %1, 4) \n\t"
"psrlq $32, %%mm2 \n\t"
"movd %%mm2, 4(%%edx) \n\t"
"movd %%mm1, 4(%%edx, %1) \n\t"
"movd %%mm2, 4(%%"REG_d") \n\t"
"movd %%mm1, 4(%%"REG_d", %1) \n\t"
"psrlq $32, %%mm1 \n\t"
"movd %%mm1, 4(%%edx, %1, 2) \n\t"
"movd %%mm1, 4(%%"REG_d", %1, 2) \n\t"
:: "r" (dst), "r" (dstStride), "r" (src)
: "%eax", "%edx"
:: "r" (dst), "r" ((long)dstStride), "r" (src)
: "%"REG_a, "%"REG_d
);
}
#endif
//static int test=0;
//static long test=0;
#ifndef HAVE_ALTIVEC
static inline void RENAME(tempNoiseReducer)(uint8_t *src, int stride,
......@@ -2216,9 +2247,9 @@ static inline void RENAME(tempNoiseReducer)(uint8_t *src, int stride,
//#define L1_DIFF //u should change the thresholds too if u try that one
#if defined (HAVE_MMX2) || defined (HAVE_3DNOW)
asm volatile(
"leal (%2, %2, 2), %%eax \n\t" // 3*stride
"leal (%2, %2, 4), %%edx \n\t" // 5*stride
"leal (%%edx, %2, 2), %%ecx \n\t" // 7*stride
"lea (%2, %2, 2), %%"REG_a" \n\t" // 3*stride
"lea (%2, %2, 4), %%"REG_d" \n\t" // 5*stride
"lea (%%"REG_d", %2, 2), %%"REG_c" \n\t" // 7*stride
// 0 1 2 3 4 5 6 7 8 9
// %x %x+%2 %x+2%2 %x+eax %x+4%2 %x+edx %x+2eax %x+ecx %x+8%2
//FIXME reorder?
......@@ -2229,29 +2260,30 @@ static inline void RENAME(tempNoiseReducer)(uint8_t *src, int stride,
"psadbw (%1, %2), %%mm1 \n\t" // |L1-R1|
"movq (%0, %2, 2), %%mm2 \n\t" // L2
"psadbw (%1, %2, 2), %%mm2 \n\t" // |L2-R2|
"movq (%0, %%eax), %%mm3 \n\t" // L3
"psadbw (%1, %%eax), %%mm3 \n\t" // |L3-R3|
"movq (%0, %%"REG_a"), %%mm3 \n\t" // L3
"psadbw (%1, %%"REG_a"), %%mm3 \n\t" // |L3-R3|
"movq (%0, %2, 4), %%mm4 \n\t" // L4
"paddw %%mm1, %%mm0 \n\t"
"psadbw (%1, %2, 4), %%mm4 \n\t" // |L4-R4|
"movq (%0, %%edx), %%mm5 \n\t" // L5
"movq (%0, %%"REG_d"), %%mm5 \n\t" // L5
"paddw %%mm2, %%mm0 \n\t"
"psadbw (%1, %%edx), %%mm5 \n\t" // |L5-R5|
"movq (%0, %%eax, 2), %%mm6 \n\t" // L6
"psadbw (%1, %%"REG_d"), %%mm5 \n\t" // |L5-R5|
"movq (%0, %%"REG_a", 2), %%mm6 \n\t" // L6
"paddw %%mm3, %%mm0 \n\t"
"psadbw (%1, %%eax, 2), %%mm6 \n\t" // |L6-R6|
"movq (%0, %%ecx), %%mm7 \n\t" // L7
"psadbw (%1, %%"REG_a", 2), %%mm6 \n\t" // |L6-R6|
"movq (%0, %%"REG_c"), %%mm7 \n\t" // L7
"paddw %%mm4, %%mm0 \n\t"
"psadbw (%1, %%ecx), %%mm7 \n\t" // |L7-R7|
"psadbw (%1, %%"REG_c"), %%mm7 \n\t" // |L7-R7|
"paddw %%mm5, %%mm6 \n\t"
"paddw %%mm7, %%mm6 \n\t"
"paddw %%mm6, %%mm0 \n\t"
#elif defined (FAST_L2_DIFF)
#else
#if defined (FAST_L2_DIFF)
"pcmpeqb %%mm7, %%mm7 \n\t"
"movq "MANGLE(b80)", %%mm6 \n\t"
"pxor %%mm0, %%mm0 \n\t"
#define L2_DIFF_CORE(a, b)\
#define REAL_L2_DIFF_CORE(a, b)\
"movq " #a ", %%mm5 \n\t"\
"movq " #b ", %%mm2 \n\t"\
"pxor %%mm7, %%mm2 \n\t"\
......@@ -2265,19 +2297,10 @@ static inline void RENAME(tempNoiseReducer)(uint8_t *src, int stride,
"psrld $14, %%mm5 \n\t"\
"paddd %%mm5, %%mm0 \n\t"
L2_DIFF_CORE((%0), (%1))
L2_DIFF_CORE((%0, %2), (%1, %2))
L2_DIFF_CORE((%0, %2, 2), (%1, %2, 2))
L2_DIFF_CORE((%0, %%eax), (%1, %%eax))
L2_DIFF_CORE((%0, %2, 4), (%1, %2, 4))
L2_DIFF_CORE((%0, %%edx), (%1, %%edx))
L2_DIFF_CORE((%0, %%eax,2), (%1, %%eax,2))
L2_DIFF_CORE((%0, %%ecx), (%1, %%ecx))
#else
"pxor %%mm7, %%mm7 \n\t"
"pxor %%mm0, %%mm0 \n\t"
#define L2_DIFF_CORE(a, b)\
#define REAL_L2_DIFF_CORE(a, b)\
"movq " #a ", %%mm5 \n\t"\
"movq " #b ", %%mm2 \n\t"\
"movq %%mm5, %%mm1 \n\t"\
......@@ -2293,14 +2316,18 @@ L2_DIFF_CORE((%0, %%ecx), (%1, %%ecx))
"paddd %%mm1, %%mm5 \n\t"\
"paddd %%mm5, %%mm0 \n\t"
#endif
#define L2_DIFF_CORE(a, b) REAL_L2_DIFF_CORE(a, b)
L2_DIFF_CORE((%0), (%1))
L2_DIFF_CORE((%0, %2), (%1, %2))
L2_DIFF_CORE((%0, %2, 2), (%1, %2, 2))
L2_DIFF_CORE((%0, %%eax), (%1, %%eax))
L2_DIFF_CORE((%0, %%REGa), (%1, %%REGa))
L2_DIFF_CORE((%0, %2, 4), (%1, %2, 4))
L2_DIFF_CORE((%0, %%edx), (%1, %%edx))
L2_DIFF_CORE((%0, %%eax,2), (%1, %%eax,2))
L2_DIFF_CORE((%0, %%ecx), (%1, %%ecx))
L2_DIFF_CORE((%0, %%REGd), (%1, %%REGd))
L2_DIFF_CORE((%0, %%REGa,2), (%1, %%REGa,2))
L2_DIFF_CORE((%0, %%REGc), (%1, %%REGc))
#endif
......@@ -2309,94 +2336,94 @@ L2_DIFF_CORE((%0, %%ecx), (%1, %%ecx))
"paddd %%mm0, %%mm4 \n\t"
"movd %%mm4, %%ecx \n\t"
"shll $2, %%ecx \n\t"
"movl %3, %%edx \n\t"
"addl -4(%%edx), %%ecx \n\t"
"addl 4(%%edx), %%ecx \n\t"
"addl -1024(%%edx), %%ecx \n\t"
"mov %3, %%"REG_d" \n\t"
"addl -4(%%"REG_d"), %%ecx \n\t"
"addl 4(%%"REG_d"), %%ecx \n\t"
"addl -1024(%%"REG_d"), %%ecx \n\t"
"addl $4, %%ecx \n\t"
"addl 1024(%%edx), %%ecx \n\t"
"addl 1024(%%"REG_d"), %%ecx \n\t"
"shrl $3, %%ecx \n\t"
"movl %%ecx, (%%edx) \n\t"
"movl %%ecx, (%%"REG_d") \n\t"
// "movl %3, %%ecx \n\t"
// "movl %%ecx, test \n\t"
// "mov %3, %%"REG_c" \n\t"
// "mov %%"REG_c", test \n\t"
// "jmp 4f \n\t"
"cmpl 512(%%edx), %%ecx \n\t"
"cmpl 512(%%"REG_d"), %%ecx \n\t"
" jb 2f \n\t"
"cmpl 516(%%edx), %%ecx \n\t"
"cmpl 516(%%"REG_d"), %%ecx \n\t"
" jb 1f \n\t"
"leal (%%eax, %2, 2), %%edx \n\t" // 5*stride
"leal (%%edx, %2, 2), %%ecx \n\t" // 7*stride
"lea (%%"REG_a", %2, 2), %%"REG_d" \n\t" // 5*stride
"lea (%%"REG_d", %2, 2), %%"REG_c" \n\t" // 7*stride
"movq (%0), %%mm0 \n\t" // L0
"movq (%0, %2), %%mm1 \n\t" // L1
"movq (%0, %2, 2), %%mm2 \n\t" // L2
"movq (%0, %%eax), %%mm3 \n\t" // L3
"movq (%0, %%"REG_a"), %%mm3 \n\t" // L3
"movq (%0, %2, 4), %%mm4 \n\t" // L4
"movq (%0, %%edx), %%mm5 \n\t" // L5
"movq (%0, %%eax, 2), %%mm6 \n\t" // L6
"movq (%0, %%ecx), %%mm7 \n\t" // L7
"movq (%0, %%"REG_d"), %%mm5 \n\t" // L5
"movq (%0, %%"REG_a", 2), %%mm6 \n\t" // L6
"movq (%0, %%"REG_c"), %%mm7 \n\t" // L7
"movq %%mm0, (%1) \n\t" // L0
"movq %%mm1, (%1, %2) \n\t" // L1
"movq %%mm2, (%1, %2, 2) \n\t" // L2
"movq %%mm3, (%1, %%eax) \n\t" // L3
"movq %%mm3, (%1, %%"REG_a") \n\t" // L3
"movq %%mm4, (%1, %2, 4) \n\t" // L4
"movq %%mm5, (%1, %%edx) \n\t" // L5
"movq %%mm6, (%1, %%eax, 2) \n\t" // L6
"movq %%mm7, (%1, %%ecx) \n\t" // L7
"movq %%mm5, (%1, %%"REG_d") \n\t" // L5
"movq %%mm6, (%1, %%"REG_a", 2) \n\t" // L6
"movq %%mm7, (%1, %%"REG_c") \n\t" // L7
"jmp 4f \n\t"
"1: \n\t"
"leal (%%eax, %2, 2), %%edx \n\t" // 5*stride
"leal (%%edx, %2, 2), %%ecx \n\t" // 7*stride
"lea (%%"REG_a", %2, 2), %%"REG_d" \n\t" // 5*stride
"lea (%%"REG_d", %2, 2), %%"REG_c" \n\t" // 7*stride
"movq (%0), %%mm0 \n\t" // L0
PAVGB((%1), %%mm0) // L0
"movq (%0, %2), %%mm1 \n\t" // L1
PAVGB((%1, %2), %%mm1) // L1
"movq (%0, %2, 2), %%mm2 \n\t" // L2
PAVGB((%1, %2, 2), %%mm2) // L2
"movq (%0, %%eax), %%mm3 \n\t" // L3
PAVGB((%1, %%eax), %%mm3) // L3
"movq (%0, %%"REG_a"), %%mm3 \n\t" // L3
PAVGB((%1, %%REGa), %%mm3) // L3
"movq (%0, %2, 4), %%mm4 \n\t" // L4
PAVGB((%1, %2, 4), %%mm4) // L4
"movq (%0, %%edx), %%mm5 \n\t" // L5
PAVGB((%1, %%edx), %%mm5) // L5
"movq (%0, %%eax, 2), %%mm6 \n\t" // L6
PAVGB((%1, %%eax, 2), %%mm6) // L6
"movq (%0, %%ecx), %%mm7 \n\t" // L7
PAVGB((%1, %%ecx), %%mm7) // L7
"movq (%0, %%"REG_d"), %%mm5 \n\t" // L5
PAVGB((%1, %%REGd), %%mm5) // L5
"movq (%0, %%"REG_a", 2), %%mm6 \n\t" // L6
PAVGB((%1, %%REGa, 2), %%mm6) // L6
"movq (%0, %%"REG_c"), %%mm7 \n\t" // L7
PAVGB((%1, %%REGc), %%mm7) // L7
"movq %%mm0, (%1) \n\t" // R0
"movq %%mm1, (%1, %2) \n\t" // R1
"movq %%mm2, (%1, %2, 2) \n\t" // R2
"movq %%mm3, (%1, %%eax) \n\t" // R3
"movq %%mm3, (%1, %%"REG_a") \n\t" // R3
"movq %%mm4, (%1, %2, 4) \n\t" // R4
"movq %%mm5, (%1, %%edx) \n\t" // R5
"movq %%mm6, (%1, %%eax, 2) \n\t" // R6
"movq %%mm7, (%1, %%ecx) \n\t" // R7
"movq %%mm5, (%1, %%"REG_d") \n\t" // R5
"movq %%mm6, (%1, %%"REG_a", 2) \n\t" // R6
"movq %%mm7, (%1, %%"REG_c") \n\t" // R7
"movq %%mm0, (%0) \n\t" // L0
"movq %%mm1, (%0, %2) \n\t" // L1
"movq %%mm2, (%0, %2, 2) \n\t" // L2
"movq %%mm3, (%0, %%eax) \n\t" // L3
"movq %%mm3, (%0, %%"REG_a") \n\t" // L3
"movq %%mm4, (%0, %2, 4) \n\t" // L4
"movq %%mm5, (%0, %%edx) \n\t" // L5
"movq %%mm6, (%0, %%eax, 2) \n\t" // L6
"movq %%mm7, (%0, %%ecx) \n\t" // L7
"movq %%mm5, (%0, %%"REG_d") \n\t" // L5
"movq %%mm6, (%0, %%"REG_a", 2) \n\t" // L6
"movq %%mm7, (%0, %%"REG_c") \n\t" // L7
"jmp 4f \n\t"
"2: \n\t"
"cmpl 508(%%edx), %%ecx \n\t"
"cmpl 508(%%"REG_d"), %%ecx \n\t"
" jb 3f \n\t"
"leal (%%eax, %2, 2), %%edx \n\t" // 5*stride
"leal (%%edx, %2, 2), %%ecx \n\t" // 7*stride
"lea (%%"REG_a", %2, 2), %%"REG_d" \n\t" // 5*stride
"lea (%%"REG_d", %2, 2), %%"REG_c" \n\t" // 7*stride
"movq (%0), %%mm0 \n\t" // L0
"movq (%0, %2), %%mm1 \n\t" // L1
"movq (%0, %2, 2), %%mm2 \n\t" // L2
"movq (%0, %%eax), %%mm3 \n\t" // L3
"movq (%0, %%"REG_a"), %%mm3 \n\t" // L3
"movq (%1), %%mm4 \n\t" // R0
"movq (%1, %2), %%mm5 \n\t" // R1
"movq (%1, %2, 2), %%mm6 \n\t" // R2
"movq (%1, %%eax), %%mm7 \n\t" // R3
"movq (%1, %%"REG_a"), %%mm7 \n\t" // R3
PAVGB(%%mm4, %%mm0)
PAVGB(%%mm5, %%mm1)
PAVGB(%%mm6, %%mm2)
......@@ -2408,20 +2435,20 @@ L2_DIFF_CORE((%0, %%ecx), (%1, %%ecx))
"movq %%mm0, (%1) \n\t" // R0
"movq %%mm1, (%1, %2) \n\t" // R1
"movq %%mm2, (%1, %2, 2) \n\t" // R2
"movq %%mm3, (%1, %%eax) \n\t" // R3
"movq %%mm3, (%1, %%"REG_a") \n\t" // R3
"movq %%mm0, (%0) \n\t" // L0
"movq %%mm1, (%0, %2) \n\t" // L1
"movq %%mm2, (%0, %2, 2) \n\t" // L2
"movq %%mm3, (%0, %%eax) \n\t" // L3
"movq %%mm3, (%0, %%"REG_a") \n\t" // L3
"movq (%0, %2, 4), %%mm0 \n\t" // L4
"movq (%0, %%edx), %%mm1 \n\t" // L5
"movq (%0, %%eax, 2), %%mm2 \n\t" // L6
"movq (%0, %%ecx), %%mm3 \n\t" // L7
"movq (%0, %%"REG_d"), %%mm1 \n\t" // L5
"movq (%0, %%"REG_a", 2), %%mm2 \n\t" // L6
"movq (%0, %%"REG_c"), %%mm3 \n\t" // L7
"movq (%1, %2, 4), %%mm4 \n\t" // R4
"movq (%1, %%edx), %%mm5 \n\t" // R5
"movq (%1, %%eax, 2), %%mm6 \n\t" // R6
"movq (%1, %%ecx), %%mm7 \n\t" // R7
"movq (%1, %%"REG_d"), %%mm5 \n\t" // R5
"movq (%1, %%"REG_a", 2), %%mm6 \n\t" // R6
"movq (%1, %%"REG_c"), %%mm7 \n\t" // R7
PAVGB(%%mm4, %%mm0)
PAVGB(%%mm5, %%mm1)
PAVGB(%%mm6, %%mm2)
......@@ -2431,26 +2458,26 @@ L2_DIFF_CORE((%0, %%ecx), (%1, %%ecx))
PAVGB(%%mm6, %%mm2)
PAVGB(%%mm7, %%mm3)
"movq %%mm0, (%1, %2, 4) \n\t" // R4
"movq %%mm1, (%1, %%edx) \n\t" // R5
"movq %%mm2, (%1, %%eax, 2) \n\t" // R6
"movq %%mm3, (%1, %%ecx) \n\t" // R7
"movq %%mm1, (%1, %%"REG_d") \n\t" // R5
"movq %%mm2, (%1, %%"REG_a", 2) \n\t" // R6
"movq %%mm3, (%1, %%"REG_c") \n\t" // R7
"movq %%mm0, (%0, %2, 4) \n\t" // L4
"movq %%mm1, (%0, %%edx) \n\t" // L5
"movq %%mm2, (%0, %%eax, 2) \n\t" // L6
"movq %%mm3, (%0, %%ecx) \n\t" // L7
"movq %%mm1, (%0, %%"REG_d") \n\t" // L5
"movq %%mm2, (%0, %%"REG_a", 2) \n\t" // L6
"movq %%mm3, (%0, %%"REG_c") \n\t" // L7
"jmp 4f \n\t"
"3: \n\t"
"leal (%%eax, %2, 2), %%edx \n\t" // 5*stride
"leal (%%edx, %2, 2), %%ecx \n\t" // 7*stride
"lea (%%"REG_a", %2, 2), %%"REG_d" \n\t" // 5*stride
"lea (%%"REG_d", %2, 2), %%"REG_c" \n\t" // 7*stride
"movq (%0), %%mm0 \n\t" // L0
"movq (%0, %2), %%mm1 \n\t" // L1
"movq (%0, %2, 2), %%mm2 \n\t" // L2
"movq (%0, %%eax), %%mm3 \n\t" // L3
"movq (%0, %%"REG_a"), %%mm3 \n\t" // L3
"movq (%1), %%mm4 \n\t" // R0
"movq (%1, %2), %%mm5 \n\t" // R1
"movq (%1, %2, 2), %%mm6 \n\t" // R2
"movq (%1, %%eax), %%mm7 \n\t" // R3
"movq (%1, %%"REG_a"), %%mm7 \n\t" // R3
PAVGB(%%mm4, %%mm0)
PAVGB(%%mm5, %%mm1)
PAVGB(%%mm6, %%mm2)
......@@ -2466,20 +2493,20 @@ L2_DIFF_CORE((%0, %%ecx), (%1, %%ecx))
"movq %%mm0, (%1) \n\t" // R0
"movq %%mm1, (%1, %2) \n\t" // R1
"movq %%mm2, (%1, %2, 2) \n\t" // R2
"movq %%mm3, (%1, %%eax) \n\t" // R3
"movq %%mm3, (%1, %%"REG_a") \n\t" // R3
"movq %%mm0, (%0) \n\t" // L0
"movq %%mm1, (%0, %2) \n\t" // L1
"movq %%mm2, (%0, %2, 2) \n\t" // L2
"movq %%mm3, (%0, %%eax) \n\t" // L3
"movq %%mm3, (%0, %%"REG_a") \n\t" // L3
"movq (%0, %2, 4), %%mm0 \n\t" // L4
"movq (%0, %%edx), %%mm1 \n\t" // L5
"movq (%0, %%eax, 2), %%mm2 \n\t" // L6
"movq (%0, %%ecx), %%mm3 \n\t" // L7
"movq (%0, %%"REG_d"), %%mm1 \n\t" // L5
"movq (%0, %%"REG_a", 2), %%mm2 \n\t" // L6
"movq (%0, %%"REG_c"), %%mm3 \n\t" // L7
"movq (%1, %2, 4), %%mm4 \n\t" // R4
"movq (%1, %%edx), %%mm5 \n\t" // R5
"movq (%1, %%eax, 2), %%mm6 \n\t" // R6
"movq (%1, %%ecx), %%mm7 \n\t" // R7
"movq (%1, %%"REG_d"), %%mm5 \n\t" // R5
"movq (%1, %%"REG_a", 2), %%mm6 \n\t" // R6
"movq (%1, %%"REG_c"), %%mm7 \n\t" // R7
PAVGB(%%mm4, %%mm0)
PAVGB(%%mm5, %%mm1)
PAVGB(%%mm6, %%mm2)
......@@ -2493,18 +2520,18 @@ L2_DIFF_CORE((%0, %%ecx), (%1, %%ecx))
PAVGB(%%mm6, %%mm2)
PAVGB(%%mm7, %%mm3)
"movq %%mm0, (%1, %2, 4) \n\t" // R4
"movq %%mm1, (%1, %%edx) \n\t" // R5
"movq %%mm2, (%1, %%eax, 2) \n\t" // R6
"movq %%mm3, (%1, %%ecx) \n\t" // R7
"movq %%mm1, (%1, %%"REG_d") \n\t" // R5
"movq %%mm2, (%1, %%"REG_a", 2) \n\t" // R6
"movq %%mm3, (%1, %%"REG_c") \n\t" // R7
"movq %%mm0, (%0, %2, 4) \n\t" // L4
"movq %%mm1, (%0, %%edx) \n\t" // L5
"movq %%mm2, (%0, %%eax, 2) \n\t" // L6
"movq %%mm3, (%0, %%ecx) \n\t" // L7
"movq %%mm1, (%0, %%"REG_d") \n\t" // L5
"movq %%mm2, (%0, %%"REG_a", 2) \n\t" // L6
"movq %%mm3, (%0, %%"REG_c") \n\t" // L7
"4: \n\t"
:: "r" (src), "r" (tempBlured), "r"(stride), "m" (tempBluredPast)
: "%eax", "%edx", "%ecx", "memory"
:: "r" (src), "r" (tempBlured), "r"((long)stride), "m" (tempBluredPast)
: "%"REG_a, "%"REG_d, "%"REG_c, "memory"
);
//printf("%d\n", test);
#else
......@@ -2630,19 +2657,19 @@ asm volatile(
);
asm volatile(
"leal (%2, %3), %%eax \n\t"
"lea (%2, %3), %%"REG_a" \n\t"
// 0 1 2 3 4 5 6 7 8 9
// %1 eax eax+%2 eax+2%2 %1+4%2 ecx ecx+%2 ecx+2%2 %1+8%2 ecx+4%2
"movq (%2), %%mm0 \n\t"
"movq (%%eax), %%mm1 \n\t"
"movq (%%"REG_a"), %%mm1 \n\t"
"movq %%mm1, %%mm3 \n\t"
"movq %%mm1, %%mm4 \n\t"
"psubb %%mm1, %%mm0 \n\t" // mm0 = differnece
"paddb %%mm7, %%mm0 \n\t"
"pcmpgtb %%mm6, %%mm0 \n\t"
"movq (%%eax,%3), %%mm2 \n\t"
"movq (%%"REG_a",%3), %%mm2 \n\t"
PMAXUB(%%mm2, %%mm4)
PMINUB(%%mm2, %%mm3, %%mm5)
"psubb %%mm2, %%mm1 \n\t"
......@@ -2650,7 +2677,7 @@ asm volatile(
"pcmpgtb %%mm6, %%mm1 \n\t"
"paddb %%mm1, %%mm0 \n\t"
"movq (%%eax, %3, 2), %%mm1 \n\t"
"movq (%%"REG_a", %3, 2), %%mm1 \n\t"
PMAXUB(%%mm1, %%mm4)
PMINUB(%%mm1, %%mm3, %%mm5)
"psubb %%mm1, %%mm2 \n\t"
......@@ -2658,7 +2685,7 @@ asm volatile(
"pcmpgtb %%mm6, %%mm2 \n\t"
"paddb %%mm2, %%mm0 \n\t"
"leal (%%eax, %3, 4), %%eax \n\t"
"lea (%%"REG_a", %3, 4), %%"REG_a" \n\t"
"movq (%2, %3, 4), %%mm2 \n\t"
PMAXUB(%%mm2, %%mm4)
......@@ -2668,7 +2695,7 @@ asm volatile(
"pcmpgtb %%mm6, %%mm1 \n\t"
"paddb %%mm1, %%mm0 \n\t"
"movq (%%eax), %%mm1 \n\t"
"movq (%%"REG_a"), %%mm1 \n\t"
PMAXUB(%%mm1, %%mm4)
PMINUB(%%mm1, %%mm3, %%mm5)
"psubb %%mm1, %%mm2 \n\t"
......@@ -2676,7 +2703,7 @@ asm volatile(
"pcmpgtb %%mm6, %%mm2 \n\t"
"paddb %%mm2, %%mm0 \n\t"
"movq (%%eax, %3), %%mm2 \n\t"
"movq (%%"REG_a", %3), %%mm2 \n\t"
PMAXUB(%%mm2, %%mm4)
PMINUB(%%mm2, %%mm3, %%mm5)
"psubb %%mm2, %%mm1 \n\t"
......@@ -2684,7 +2711,7 @@ asm volatile(
"pcmpgtb %%mm6, %%mm1 \n\t"
"paddb %%mm1, %%mm0 \n\t"
"movq (%%eax, %3, 2), %%mm1 \n\t"
"movq (%%"REG_a", %3, 2), %%mm1 \n\t"
PMAXUB(%%mm1, %%mm4)
PMINUB(%%mm1, %%mm3, %%mm5)
"psubb %%mm1, %%mm2 \n\t"
......@@ -2700,7 +2727,7 @@ asm volatile(
"pcmpgtb %%mm6, %%mm1 \n\t"
"paddb %%mm1, %%mm0 \n\t"
"movq (%%eax, %3, 4), %%mm1 \n\t"
"movq (%%"REG_a", %3, 4), %%mm1 \n\t"
"psubb %%mm1, %%mm2 \n\t"
"paddb %%mm7, %%mm2 \n\t"
"pcmpgtb %%mm6, %%mm2 \n\t"
......@@ -2724,12 +2751,12 @@ asm volatile(
"movq %%mm6, %0 \n\t"
: "=m" (eq_mask), "=m" (dc_mask)
: "r" (src), "r" (step), "m" (c->pQPb), "m"(c->ppMode.flatnessThreshold)
: "%eax"
: "r" (src), "r" ((long)step), "m" (c->pQPb), "m"(c->ppMode.flatnessThreshold)
: "%"REG_a
);
if(dc_mask & eq_mask){
int offset= -8*step;
long offset= -8*step;
int64_t *temp_sums= sums;
asm volatile(
......@@ -2752,7 +2779,7 @@ asm volatile(
// 0:QP 6:First
"movq (%0, %1, 8), %%mm5 \n\t"
"addl %1, %0 \n\t" // %0 points to line 1 not 0
"add %1, %0 \n\t" // %0 points to line 1 not 0
"movq (%0, %1, 8), %%mm7 \n\t"
"movq %%mm5, %%mm1 \n\t"
"movq %%mm7, %%mm2 \n\t"
......@@ -2782,7 +2809,7 @@ asm volatile(
#define NEXT\
"movq (%0), %%mm2 \n\t"\
"movq (%0), %%mm3 \n\t"\
"addl %1, %0 \n\t"\
"add %1, %0 \n\t"\
"punpcklbw %%mm4, %%mm2 \n\t"\
"punpckhbw %%mm4, %%mm3 \n\t"\
"paddw %%mm2, %%mm0 \n\t"\
......@@ -2791,7 +2818,7 @@ asm volatile(
#define PREV\
"movq (%0), %%mm2 \n\t"\
"movq (%0), %%mm3 \n\t"\
"addl %1, %0 \n\t"\
"add %1, %0 \n\t"\
"punpcklbw %%mm4, %%mm2 \n\t"\
"punpckhbw %%mm4, %%mm3 \n\t"\
"psubw %%mm2, %%mm0 \n\t"\
......@@ -2833,8 +2860,8 @@ asm volatile(
"punpcklbw %%mm4, %%mm6 \n\t"
NEXT //7
"movl %4, %0 \n\t"
"addl %1, %0 \n\t"
"mov %4, %0 \n\t"
"add %1, %0 \n\t"
PREV //0
"movq %%mm0, 80(%3) \n\t"
"movq %%mm1, 88(%3) \n\t"
......@@ -2863,10 +2890,10 @@ asm volatile(
"movq %%mm0, 144(%3) \n\t"
"movq %%mm1, 152(%3) \n\t"
"movl %4, %0 \n\t" //FIXME
"mov %4, %0 \n\t" //FIXME
: "+&r"(src)
: "r" (step), "m" (c->pQPb), "r"(sums), "g"(src)
: "r" ((long)step), "m" (c->pQPb), "r"(sums), "g"(src)
);
src+= step; // src points to begin of the 8x8 Block
......@@ -2898,12 +2925,12 @@ asm volatile(
"pand %%mm5, %%mm4 \n\t"
"por %%mm4, %%mm0 \n\t"
"movq %%mm0, (%0, %3) \n\t"
"addl $16, %1 \n\t"
"addl %2, %0 \n\t"
"add $16, %1 \n\t"
"add %2, %0 \n\t"
" js 1b \n\t"
: "+r"(offset), "+r"(temp_sums)
: "r" (step), "r"(src - offset), "m"(dc_mask & eq_mask)
: "r" ((long)step), "r"(src - offset), "m"(dc_mask & eq_mask)
);
}else
src+= step; // src points to begin of the 8x8 Block
......@@ -2912,8 +2939,8 @@ asm volatile(
uint8_t *temp_src= src;
asm volatile(
"pxor %%mm7, %%mm7 \n\t"
"leal -40(%%esp), %%ecx \n\t" // make space for 4 8-byte vars
"andl $0xFFFFFFF8, %%ecx \n\t" // align
"lea -40(%%"REG_SP"), %%"REG_c" \n\t" // make space for 4 8-byte vars
"and "ALIGN_MASK", %%"REG_c" \n\t" // align
// 0 1 2 3 4 5 6 7 8 9
// %0 eax eax+%1 eax+2%1 %0+4%1 ecx ecx+%1 ecx+2%1 %1+8%1 ecx+4%1
......@@ -2923,12 +2950,12 @@ asm volatile(
"punpckhbw %%mm7, %%mm1 \n\t" // high part of line 0
"movq (%0, %1), %%mm2 \n\t"
"leal (%0, %1, 2), %%eax \n\t"
"lea (%0, %1, 2), %%"REG_a" \n\t"
"movq %%mm2, %%mm3 \n\t"
"punpcklbw %%mm7, %%mm2 \n\t" // low part of line 1
"punpckhbw %%mm7, %%mm3 \n\t" // high part of line 1
"movq (%%eax), %%mm4 \n\t"
"movq (%%"REG_a"), %%mm4 \n\t"
"movq %%mm4, %%mm5 \n\t"
"punpcklbw %%mm7, %%mm4 \n\t" // low part of line 2
"punpckhbw %%mm7, %%mm5 \n\t" // high part of line 2
......@@ -2945,7 +2972,7 @@ asm volatile(
"psubw %%mm2, %%mm0 \n\t" // 2L0 - 5L1 + 5L2
"psubw %%mm3, %%mm1 \n\t" // 2H0 - 5H1 + 5H2
"movq (%%eax, %1), %%mm2 \n\t"
"movq (%%"REG_a", %1), %%mm2 \n\t"
"movq %%mm2, %%mm3 \n\t"
"punpcklbw %%mm7, %%mm2 \n\t" // L3
"punpckhbw %%mm7, %%mm3 \n\t" // H3
......@@ -2954,24 +2981,24 @@ asm volatile(
"psubw %%mm3, %%mm1 \n\t" // 2H0 - 5H1 + 5H2 - H3
"psubw %%mm2, %%mm0 \n\t" // 2L0 - 5L1 + 5L2 - 2L3
"psubw %%mm3, %%mm1 \n\t" // 2H0 - 5H1 + 5H2 - 2H3
"movq %%mm0, (%%ecx) \n\t" // 2L0 - 5L1 + 5L2 - 2L3
"movq %%mm1, 8(%%ecx) \n\t" // 2H0 - 5H1 + 5H2 - 2H3
"movq %%mm0, (%%"REG_c") \n\t" // 2L0 - 5L1 + 5L2 - 2L3
"movq %%mm1, 8(%%"REG_c") \n\t" // 2H0 - 5H1 + 5H2 - 2H3
"movq (%%eax, %1, 2), %%mm0 \n\t"
"movq (%%"REG_a", %1, 2), %%mm0 \n\t"
"movq %%mm0, %%mm1 \n\t"
"punpcklbw %%mm7, %%mm0 \n\t" // L4
"punpckhbw %%mm7, %%mm1 \n\t" // H4
"psubw %%mm0, %%mm2 \n\t" // L3 - L4
"psubw %%mm1, %%mm3 \n\t" // H3 - H4
"movq %%mm2, 16(%%ecx) \n\t" // L3 - L4
"movq %%mm3, 24(%%ecx) \n\t" // H3 - H4
"movq %%mm2, 16(%%"REG_c") \n\t" // L3 - L4
"movq %%mm3, 24(%%"REG_c") \n\t" // H3 - H4
"paddw %%mm4, %%mm4 \n\t" // 2L2
"paddw %%mm5, %%mm5 \n\t" // 2H2
"psubw %%mm2, %%mm4 \n\t" // 2L2 - L3 + L4
"psubw %%mm3, %%mm5 \n\t" // 2H2 - H3 + H4
"leal (%%eax, %1), %0 \n\t"
"lea (%%"REG_a", %1), %0 \n\t"
"psllw $2, %%mm2 \n\t" // 4L3 - 4L4
"psllw $2, %%mm3 \n\t" // 4H3 - 4H4
"psubw %%mm2, %%mm4 \n\t" // 2L2 - 5L3 + 5L4
......@@ -2986,10 +3013,10 @@ asm volatile(
"psubw %%mm2, %%mm4 \n\t" // 2L2 - 5L3 + 5L4 - 2L5
"psubw %%mm3, %%mm5 \n\t" // 2H2 - 5H3 + 5H4 - 2H5
"movq (%%eax, %1, 4), %%mm6 \n\t"
"movq (%%"REG_a", %1, 4), %%mm6 \n\t"
"punpcklbw %%mm7, %%mm6 \n\t" // L6
"psubw %%mm6, %%mm2 \n\t" // L5 - L6
"movq (%%eax, %1, 4), %%mm6 \n\t"
"movq (%%"REG_a", %1, 4), %%mm6 \n\t"
"punpckhbw %%mm7, %%mm6 \n\t" // H6
"psubw %%mm6, %%mm3 \n\t" // H5 - H6
......@@ -3013,8 +3040,8 @@ asm volatile(
"psubw %%mm2, %%mm0 \n\t" // 2L4 - 5L5 + 5L6 - 2L7
"psubw %%mm3, %%mm1 \n\t" // 2H4 - 5H5 + 5H6 - 2H7
"movq (%%ecx), %%mm2 \n\t" // 2L0 - 5L1 + 5L2 - 2L3
"movq 8(%%ecx), %%mm3 \n\t" // 2H0 - 5H1 + 5H2 - 2H3
"movq (%%"REG_c"), %%mm2 \n\t" // 2L0 - 5L1 + 5L2 - 2L3
"movq 8(%%"REG_c"), %%mm3 \n\t" // 2H0 - 5H1 + 5H2 - 2H3
#ifdef HAVE_MMX2
"movq %%mm7, %%mm6 \n\t" // 0
......@@ -3092,8 +3119,8 @@ asm volatile(
"psrlw $6, %%mm4 \n\t"
"psrlw $6, %%mm5 \n\t"
"movq 16(%%ecx), %%mm0 \n\t" // L3 - L4
"movq 24(%%ecx), %%mm1 \n\t" // H3 - H4
"movq 16(%%"REG_c"), %%mm0 \n\t" // L3 - L4
"movq 24(%%"REG_c"), %%mm1 \n\t" // H3 - H4
"pxor %%mm2, %%mm2 \n\t"
"pxor %%mm3, %%mm3 \n\t"
......@@ -3138,8 +3165,8 @@ asm volatile(
"movq %%mm0, (%0, %1) \n\t"
: "+r" (temp_src)
: "r" (step), "m" (c->pQPb), "m"(eq_mask)
: "%eax", "%ecx"
: "r" ((long)step), "m" (c->pQPb), "m"(eq_mask)
: "%"REG_a, "%"REG_c
);
}
/*if(step==16){
......@@ -3169,13 +3196,13 @@ static inline void RENAME(blockCopy)(uint8_t dst[], int dstStride, uint8_t src[]
{
#ifdef HAVE_MMX
asm volatile(
"movq (%%eax), %%mm2 \n\t" // packedYOffset
"movq 8(%%eax), %%mm3 \n\t" // packedYScale
"leal (%2,%4), %%eax \n\t"
"leal (%3,%5), %%edx \n\t"
"movq (%%"REG_a"), %%mm2 \n\t" // packedYOffset
"movq 8(%%"REG_a"), %%mm3 \n\t" // packedYScale
"lea (%2,%4), %%"REG_a" \n\t"
"lea (%3,%5), %%"REG_d" \n\t"
"pxor %%mm4, %%mm4 \n\t"
#ifdef HAVE_MMX2
#define SCALED_CPY(src1, src2, dst1, dst2) \
#define REAL_SCALED_CPY(src1, src2, dst1, dst2) \
"movq " #src1 ", %%mm0 \n\t"\
"movq " #src1 ", %%mm5 \n\t"\
"movq " #src2 ", %%mm1 \n\t"\
......@@ -3198,7 +3225,7 @@ static inline void RENAME(blockCopy)(uint8_t dst[], int dstStride, uint8_t src[]
"movq %%mm1, " #dst2 " \n\t"\
#else //HAVE_MMX2
#define SCALED_CPY(src1, src2, dst1, dst2) \
#define REAL_SCALED_CPY(src1, src2, dst1, dst2) \
"movq " #src1 ", %%mm0 \n\t"\
"movq " #src1 ", %%mm5 \n\t"\
"punpcklbw %%mm4, %%mm0 \n\t"\
......@@ -3225,22 +3252,24 @@ static inline void RENAME(blockCopy)(uint8_t dst[], int dstStride, uint8_t src[]
"movq %%mm1, " #dst2 " \n\t"\
#endif //!HAVE_MMX2
#define SCALED_CPY(src1, src2, dst1, dst2)\
REAL_SCALED_CPY(src1, src2, dst1, dst2)
SCALED_CPY((%2) , (%2, %4) , (%3) , (%3, %5))
SCALED_CPY((%2, %4, 2), (%%eax, %4, 2), (%3, %5, 2), (%%edx, %5, 2))
SCALED_CPY((%2, %4, 4), (%%eax, %4, 4), (%3, %5, 4), (%%edx, %5, 4))
"leal (%%eax,%4,4), %%eax \n\t"
"leal (%%edx,%5,4), %%edx \n\t"
SCALED_CPY((%%eax, %4), (%%eax, %4, 2), (%%edx, %5), (%%edx, %5, 2))
SCALED_CPY((%2, %4, 2), (%%REGa, %4, 2), (%3, %5, 2), (%%REGd, %5, 2))
SCALED_CPY((%2, %4, 4), (%%REGa, %4, 4), (%3, %5, 4), (%%REGd, %5, 4))
"lea (%%"REG_a",%4,4), %%"REG_a" \n\t"
"lea (%%"REG_d",%5,4), %%"REG_d" \n\t"
SCALED_CPY((%%REGa, %4), (%%REGa, %4, 2), (%%REGd, %5), (%%REGd, %5, 2))
: "=&a" (packedOffsetAndScale)
: "0" (packedOffsetAndScale),
"r"(src),
"r"(dst),
"r" (srcStride),
"r" (dstStride)
: "%edx"
"r" ((long)srcStride),
"r" ((long)dstStride)
: "%"REG_d
);
#else
for(i=0; i<8; i++)
......@@ -3252,27 +3281,30 @@ SCALED_CPY((%%eax, %4), (%%eax, %4, 2), (%%edx, %5), (%%edx, %5, 2))
{
#ifdef HAVE_MMX
asm volatile(
"leal (%0,%2), %%eax \n\t"
"leal (%1,%3), %%edx \n\t"
"lea (%0,%2), %%"REG_a" \n\t"
"lea (%1,%3), %%"REG_d" \n\t"
#define SIMPLE_CPY(src1, src2, dst1, dst2) \
#define REAL_SIMPLE_CPY(src1, src2, dst1, dst2) \
"movq " #src1 ", %%mm0 \n\t"\
"movq " #src2 ", %%mm1 \n\t"\
"movq %%mm0, " #dst1 " \n\t"\
"movq %%mm1, " #dst2 " \n\t"\
#define SIMPLE_CPY(src1, src2, dst1, dst2)\
REAL_SIMPLE_CPY(src1, src2, dst1, dst2)
SIMPLE_CPY((%0) , (%0, %2) , (%1) , (%1, %3))
SIMPLE_CPY((%0, %2, 2), (%%eax, %2, 2), (%1, %3, 2), (%%edx, %3, 2))
SIMPLE_CPY((%0, %2, 4), (%%eax, %2, 4), (%1, %3, 4), (%%edx, %3, 4))
"leal (%%eax,%2,4), %%eax \n\t"
"leal (%%edx,%3,4), %%edx \n\t"
SIMPLE_CPY((%%eax, %2), (%%eax, %2, 2), (%%edx, %3), (%%edx, %3, 2))
SIMPLE_CPY((%0, %2, 2), (%%REGa, %2, 2), (%1, %3, 2), (%%REGd, %3, 2))
SIMPLE_CPY((%0, %2, 4), (%%REGa, %2, 4), (%1, %3, 4), (%%REGd, %3, 4))
"lea (%%"REG_a",%2,4), %%"REG_a" \n\t"
"lea (%%"REG_d",%3,4), %%"REG_d" \n\t"
SIMPLE_CPY((%%REGa, %2), (%%REGa, %2, 2), (%%REGd, %3), (%%REGd, %3, 2))
: : "r" (src),
"r" (dst),
"r" (srcStride),
"r" (dstStride)
: "%eax", "%edx"
"r" ((long)srcStride),
"r" ((long)dstStride)
: "%"REG_a, "%"REG_d
);
#else
for(i=0; i<8; i++)
......@@ -3290,12 +3322,12 @@ static inline void RENAME(duplicate)(uint8_t src[], int stride)
#ifdef HAVE_MMX
asm volatile(
"movq (%0), %%mm0 \n\t"
"addl %1, %0 \n\t"
"add %1, %0 \n\t"
"movq %%mm0, (%0) \n\t"
"movq %%mm0, (%0, %1) \n\t"
"movq %%mm0, (%0, %1, 2) \n\t"
: "+r" (src)
: "r" (-stride)
: "r" ((long)-stride)
);
#else
int i;
......@@ -3447,22 +3479,22 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int
*/
asm(
"movl %4, %%eax \n\t"
"shrl $2, %%eax \n\t"
"andl $6, %%eax \n\t"
"addl %5, %%eax \n\t"
"movl %%eax, %%edx \n\t"
"imul %1, %%eax \n\t"
"imul %3, %%edx \n\t"
"prefetchnta 32(%%eax, %0) \n\t"
"prefetcht0 32(%%edx, %2) \n\t"
"addl %1, %%eax \n\t"
"addl %3, %%edx \n\t"
"prefetchnta 32(%%eax, %0) \n\t"
"prefetcht0 32(%%edx, %2) \n\t"
:: "r" (srcBlock), "r" (srcStride), "r" (dstBlock), "r" (dstStride),
"m" (x), "m" (copyAhead)
: "%eax", "%edx"
"mov %4, %%"REG_a" \n\t"
"shr $2, %%"REG_a" \n\t"
"and $6, %%"REG_a" \n\t"
"add %5, %%"REG_a" \n\t"
"mov %%"REG_a", %%"REG_d" \n\t"
"imul %1, %%"REG_a" \n\t"
"imul %3, %%"REG_d" \n\t"
"prefetchnta 32(%%"REG_a", %0) \n\t"
"prefetcht0 32(%%"REG_d", %2) \n\t"
"add %1, %%"REG_a" \n\t"
"add %3, %%"REG_d" \n\t"
"prefetchnta 32(%%"REG_a", %0) \n\t"
"prefetcht0 32(%%"REG_d", %2) \n\t"
:: "r" (srcBlock), "r" ((long)srcStride), "r" (dstBlock), "r" ((long)dstStride),
"m" ((long)x), "m" ((long)copyAhead)
: "%"REG_a, "%"REG_d
);
#elif defined(HAVE_3DNOW)
......@@ -3593,22 +3625,22 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int
*/
asm(
"movl %4, %%eax \n\t"
"shrl $2, %%eax \n\t"
"andl $6, %%eax \n\t"
"addl %5, %%eax \n\t"
"movl %%eax, %%edx \n\t"
"imul %1, %%eax \n\t"
"imul %3, %%edx \n\t"
"prefetchnta 32(%%eax, %0) \n\t"
"prefetcht0 32(%%edx, %2) \n\t"
"addl %1, %%eax \n\t"
"addl %3, %%edx \n\t"
"prefetchnta 32(%%eax, %0) \n\t"
"prefetcht0 32(%%edx, %2) \n\t"
:: "r" (srcBlock), "r" (srcStride), "r" (dstBlock), "r" (dstStride),
"m" (x), "m" (copyAhead)
: "%eax", "%edx"
"mov %4, %%"REG_a" \n\t"
"shr $2, %%"REG_a" \n\t"
"and $6, %%"REG_a" \n\t"
"add %5, %%"REG_a" \n\t"
"mov %%"REG_a", %%"REG_d" \n\t"
"imul %1, %%"REG_a" \n\t"
"imul %3, %%"REG_d" \n\t"
"prefetchnta 32(%%"REG_a", %0) \n\t"
"prefetcht0 32(%%"REG_d", %2) \n\t"
"add %1, %%"REG_a" \n\t"
"add %3, %%"REG_d" \n\t"
"prefetchnta 32(%%"REG_a", %0) \n\t"
"prefetcht0 32(%%"REG_d", %2) \n\t"
:: "r" (srcBlock), "r" ((long)srcStride), "r" (dstBlock), "r" ((long)dstStride),
"m" ((long)x), "m" ((long)copyAhead)
: "%"REG_a, "%"REG_d
);
#elif defined(HAVE_3DNOW)
......
libavcodec/msmpeg4.c
View file @ 09fcf89b
......@@ -716,7 +716,7 @@ static inline int msmpeg4_pred_dc(MpegEncContext * s, int n,
necessitate to modify mpegvideo.c. The problem comes from the
fact they decided to store the quantized DC (which would lead
to problems if Q could vary !) */
#if defined ARCH_X86 && !defined PIC
#if (defined(ARCH_X86) || defined(ARCH_X86_64)) && !defined PIC
asm volatile(
"movl %3, %%eax \n\t"
"shrl $1, %%eax \n\t"
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment