RepoMirrors/mpv
1
0
Fork 0
mirror of https://github.com/mpv-player/mpv synced 2025-03-20 18:28:01 +00:00
Code Issues Releases Wiki Activity

Change all occurrences of asm and __asm to __asm__, same as was done for FFmpeg.

Browse Source 
Neither variant is valid C99 syntax, but __asm__ is the most portable variant.


git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@27788 b3059339-0415-0410-9bf9-f77b7e298cf2
This commit is contained in:
diego 2008-10-16 18:59:27 +00:00
parent 629db77d89
commit 6b52a2e974
43 changed files with 238 additions and 243 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
TOOLS/fastmemcpybench.c
View File

@ -86,7 +86,7 @@ static unsigned int GetTimer(){
static inline unsigned long long int read_tsc( void )
{
unsigned long long int retval;
__asm __volatile ("rdtsc":"=A"(retval)::"memory");
__asm__ __volatile ("rdtsc":"=A"(retval)::"memory");
return retval;
}

6
configure vendored
View File

@ -2124,8 +2124,8 @@ EOF
cat > $TMPC << EOF
int main(void) {
unsigned long ver, mask;
asm ("implver %0" : "=r" (ver));
asm ("amask %1, %0" : "=r" (mask) : "r" (-1));
__asm__ ("implver %0" : "=r" (ver));
__asm__ ("amask %1, %0" : "=r" (mask) : "r" (-1));
printf("%ld-%x\n", ver, ~mask);
return 0;
}
@ -2374,7 +2374,7 @@ echocheck ".align is a power of two"
if test "$_asmalign_pot" = auto ; then
_asmalign_pot=no
cat > $TMPC << EOF
int main(void) { asm (".align 3"); return 0; }
int main(void) { __asm__ (".align 3"); return 0; }
EOF
cc_check && _asmalign_pot=yes
fi

14
cpudetect.c
View File

@ -85,14 +85,14 @@ static void
do_cpuid(unsigned int ax, unsigned int *p)
{
#if 0
__asm __volatile(
__asm__ __volatile(
"cpuid;"
: "=a" (p[0]), "=b" (p[1]), "=c" (p[2]), "=d" (p[3])
: "0" (ax)
);
#else
// code from libavcodec:
__asm __volatile
__asm__ __volatile__
("mov %%"REG_b", %%"REG_S"\n\t"
"cpuid\n\t"
"xchg %%"REG_b", %%"REG_S
@ -400,7 +400,7 @@ static void check_os_katmai_support( void )
if ( gCpuCaps.hasSSE ) {
mp_msg(MSGT_CPUDETECT,MSGL_V, "Testing OS support for SSE... " );
exc_fil = SetUnhandledExceptionFilter(win32_sig_handler_sse);
__asm __volatile ("xorps %xmm0, %xmm0");
__asm__ __volatile ("xorps %xmm0, %xmm0");
SetUnhandledExceptionFilter(exc_fil);
mp_msg(MSGT_CPUDETECT,MSGL_V, gCpuCaps.hasSSE ? "yes.\n" : "no!\n" );
}
@ -409,7 +409,7 @@ static void check_os_katmai_support( void )
if ( gCpuCaps.hasSSE ) {
mp_msg(MSGT_CPUDETECT,MSGL_V, "Testing OS support for SSE... " );
DosSetExceptionHandler( &RegRec );
__asm __volatile ("xorps %xmm0, %xmm0");
__asm__ __volatile ("xorps %xmm0, %xmm0");
DosUnsetExceptionHandler( &RegRec );
mp_msg(MSGT_CPUDETECT,MSGL_V, gCpuCaps.hasSSE ? "yes.\n" : "no!\n" );
}
@ -432,8 +432,8 @@ static void check_os_katmai_support( void )
if ( gCpuCaps.hasSSE ) {
mp_msg(MSGT_CPUDETECT,MSGL_V, "Testing OS support for SSE... " );
// __asm __volatile ("xorps %%xmm0, %%xmm0");
__asm __volatile ("xorps %xmm0, %xmm0");
// __asm__ __volatile ("xorps %%xmm0, %%xmm0");
__asm__ __volatile ("xorps %xmm0, %xmm0");
mp_msg(MSGT_CPUDETECT,MSGL_V, gCpuCaps.hasSSE ? "yes.\n" : "no!\n" );
}
@ -532,7 +532,7 @@ void GetCpuCaps( CpuCaps *caps)
} else {
canjump = 1;
asm volatile ("mtspr 256, %0\n\t"
__asm__ volatile ("mtspr 256, %0\n\t"
"vand %%v0, %%v0, %%v0"
:
: "r" (-1));

6
cpuinfo.c
View File

@ -49,9 +49,9 @@ cpuid(int func) {
cpuid_regs_t regs;
#define CPUID ".byte 0x0f, 0xa2; "
#ifdef __x86_64__
asm("mov %%rbx, %%rsi\n\t"
__asm__("mov %%rbx, %%rsi\n\t"
#else
asm("mov %%ebx, %%esi\n\t"
__asm__("mov %%ebx, %%esi\n\t"
#endif
CPUID"\n\t"
#ifdef __x86_64__
@ -70,7 +70,7 @@ rdtsc(void)
{
uint64_t i;
#define RDTSC ".byte 0x0f, 0x31; "
asm volatile (RDTSC : "=A"(i) : );
__asm__ volatile (RDTSC : "=A"(i) : );
return i;
}

4
drivers/radeon_vid.c
View File

@ -129,7 +129,7 @@ static int swap_fourcc __initdata = 0;
static inline double FastSin(double x)
{
register double res;
__asm __volatile("fsin":"=t"(res):"0"(x));
__asm__ __volatile("fsin":"=t"(res):"0"(x));
return res;
}
#undef sin
@ -138,7 +138,7 @@ static inline double FastSin(double x)
static inline double FastCos(double x)
{
register double res;
__asm __volatile("fcos":"=t"(res):"0"(x));
__asm__ __volatile("fcos":"=t"(res):"0"(x));
return res;
}
#undef cos

56
liba52/downmix.c
View File

@ -689,7 +689,7 @@ void upmix_C (sample_t * samples, int acmod, int output)
#if defined(ARCH_X86) || defined(ARCH_X86_64)
static void mix2to1_SSE (sample_t * dest, sample_t * src, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movlps %2, %%xmm7 \n\t"
"shufps $0x00, %%xmm7, %%xmm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -712,7 +712,7 @@ static void mix2to1_SSE (sample_t * dest, sample_t * src, sample_t bias)
static void mix3to1_SSE (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movlps %1, %%xmm7 \n\t"
"shufps $0x00, %%xmm7, %%xmm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -733,7 +733,7 @@ static void mix3to1_SSE (sample_t * samples, sample_t bias)
static void mix4to1_SSE (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movlps %1, %%xmm7 \n\t"
"shufps $0x00, %%xmm7, %%xmm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -755,7 +755,7 @@ static void mix4to1_SSE (sample_t * samples, sample_t bias)
static void mix5to1_SSE (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movlps %1, %%xmm7 \n\t"
"shufps $0x00, %%xmm7, %%xmm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -778,7 +778,7 @@ static void mix5to1_SSE (sample_t * samples, sample_t bias)
static void mix3to2_SSE (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movlps %1, %%xmm7 \n\t"
"shufps $0x00, %%xmm7, %%xmm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -801,7 +801,7 @@ static void mix3to2_SSE (sample_t * samples, sample_t bias)
static void mix21to2_SSE (sample_t * left, sample_t * right, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movlps %2, %%xmm7 \n\t"
"shufps $0x00, %%xmm7, %%xmm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -824,7 +824,7 @@ static void mix21to2_SSE (sample_t * left, sample_t * right, sample_t bias)
static void mix21toS_SSE (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movlps %1, %%xmm7 \n\t"
"shufps $0x00, %%xmm7, %%xmm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -848,7 +848,7 @@ static void mix21toS_SSE (sample_t * samples, sample_t bias)
static void mix31to2_SSE (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movlps %1, %%xmm7 \n\t"
"shufps $0x00, %%xmm7, %%xmm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -872,7 +872,7 @@ static void mix31to2_SSE (sample_t * samples, sample_t bias)
static void mix31toS_SSE (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movlps %1, %%xmm7 \n\t"
"shufps $0x00, %%xmm7, %%xmm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -898,7 +898,7 @@ static void mix31toS_SSE (sample_t * samples, sample_t bias)
static void mix22toS_SSE (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movlps %1, %%xmm7 \n\t"
"shufps $0x00, %%xmm7, %%xmm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -923,7 +923,7 @@ static void mix22toS_SSE (sample_t * samples, sample_t bias)
static void mix32to2_SSE (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movlps %1, %%xmm7 \n\t"
"shufps $0x00, %%xmm7, %%xmm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -947,7 +947,7 @@ static void mix32to2_SSE (sample_t * samples, sample_t bias)
static void mix32toS_SSE (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movlps %1, %%xmm7 \n\t"
"shufps $0x00, %%xmm7, %%xmm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -974,7 +974,7 @@ static void mix32toS_SSE (sample_t * samples, sample_t bias)
static void move2to1_SSE (sample_t * src, sample_t * dest, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movlps %2, %%xmm7 \n\t"
"shufps $0x00, %%xmm7, %%xmm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -997,7 +997,7 @@ static void move2to1_SSE (sample_t * src, sample_t * dest, sample_t bias)
static void zero_MMX(sample_t * samples)
{
asm volatile(
__asm__ volatile(
"mov $-1024, %%"REG_S" \n\t"
"pxor %%mm0, %%mm0 \n\t"
ASMALIGN(4)
@ -1223,7 +1223,7 @@ static void upmix_MMX (sample_t * samples, int acmod, int output)
static void mix2to1_3dnow (sample_t * dest, sample_t * src, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movd %2, %%mm7 \n\t"
"punpckldq %2, %%mm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -1254,7 +1254,7 @@ static void mix2to1_3dnow (sample_t * dest, sample_t * src, sample_t bias)
static void mix3to1_3dnow (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movd %1, %%mm7 \n\t"
"punpckldq %1, %%mm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -1281,7 +1281,7 @@ static void mix3to1_3dnow (sample_t * samples, sample_t bias)
static void mix4to1_3dnow (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movd %1, %%mm7 \n\t"
"punpckldq %1, %%mm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -1310,7 +1310,7 @@ static void mix4to1_3dnow (sample_t * samples, sample_t bias)
static void mix5to1_3dnow (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movd %1, %%mm7 \n\t"
"punpckldq %1, %%mm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -1341,7 +1341,7 @@ static void mix5to1_3dnow (sample_t * samples, sample_t bias)
static void mix3to2_3dnow (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movd %1, %%mm7 \n\t"
"punpckldq %1, %%mm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -1372,7 +1372,7 @@ static void mix3to2_3dnow (sample_t * samples, sample_t bias)
static void mix21to2_3dnow (sample_t * left, sample_t * right, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movd %2, %%mm7 \n\t"
"punpckldq %2, %%mm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -1403,7 +1403,7 @@ static void mix21to2_3dnow (sample_t * left, sample_t * right, sample_t bias)
static void mix21toS_3dnow (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movd %1, %%mm7 \n\t"
"punpckldq %1, %%mm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -1436,7 +1436,7 @@ static void mix21toS_3dnow (sample_t * samples, sample_t bias)
static void mix31to2_3dnow (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movd %1, %%mm7 \n\t"
"punpckldq %1, %%mm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -1469,7 +1469,7 @@ static void mix31to2_3dnow (sample_t * samples, sample_t bias)
static void mix31toS_3dnow (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movd %1, %%mm7 \n\t"
"punpckldq %1, %%mm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -1506,7 +1506,7 @@ static void mix31toS_3dnow (sample_t * samples, sample_t bias)
static void mix22toS_3dnow (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movd %1, %%mm7 \n\t"
"punpckldq %1, %%mm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -1541,7 +1541,7 @@ static void mix22toS_3dnow (sample_t * samples, sample_t bias)
static void mix32to2_3dnow (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movd %1, %%mm7 \n\t"
"punpckldq %1, %%mm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -1575,7 +1575,7 @@ static void mix32to2_3dnow (sample_t * samples, sample_t bias)
/* todo: should be optimized better */
static void mix32toS_3dnow (sample_t * samples, sample_t bias)
{
asm volatile(
__asm__ volatile(
"mov $-1024, %%"REG_S" \n\t"
ASMALIGN(4)
"1: \n\t"
@ -1614,7 +1614,7 @@ static void mix32toS_3dnow (sample_t * samples, sample_t bias)
static void move2to1_3dnow (sample_t * src, sample_t * dest, sample_t bias)
{
asm volatile(
__asm__ volatile(
"movd %2, %%mm7 \n\t"
"punpckldq %2, %%mm7 \n\t"
"mov $-1024, %%"REG_S" \n\t"
@ -1782,7 +1782,7 @@ static void downmix_3dnow (sample_t * samples, int acmod, int output, sample_t b
memcpy (samples + 1024, samples + 768, 256 * sizeof (sample_t));
break;
}
__asm __volatile("femms":::"memory");
__asm__ volatile("femms":::"memory");
}
#endif // ARCH_X86 || ARCH_X86_64

20
liba52/imdct.c
View File

@ -750,7 +750,7 @@ imdct_do_512_sse(sample_t data[],sample_t delay[], sample_t bias)
/* Pre IFFT complex multiply plus IFFT cmplx conjugate */
/* Bit reversed shuffling */
asm volatile(
__asm__ volatile(
"xor %%"REG_S", %%"REG_S" \n\t"
"lea "MANGLE(bit_reverse_512)", %%"REG_a"\n\t"
"mov $1008, %%"REG_D" \n\t"
@ -810,7 +810,7 @@ imdct_do_512_sse(sample_t data[],sample_t delay[], sample_t bias)
/* 1. iteration */
// Note w[0][0]={1,0}
asm volatile(
__asm__ volatile(
"xorps %%xmm1, %%xmm1 \n\t"
"xorps %%xmm2, %%xmm2 \n\t"
"mov %0, %%"REG_S" \n\t"
@ -832,7 +832,7 @@ imdct_do_512_sse(sample_t data[],sample_t delay[], sample_t bias)
/* 2. iteration */
// Note w[1]={{1,0}, {0,-1}}
asm volatile(
__asm__ volatile(
"movaps "MANGLE(ps111_1)", %%xmm7\n\t" // 1,1,1,-1
"mov %0, %%"REG_S" \n\t"
ASMALIGN(4)
@ -860,7 +860,7 @@ imdct_do_512_sse(sample_t data[],sample_t delay[], sample_t bias)
Note sseW2+32={0,0,-sqrt(2),-sqrt(2))
Note sseW2+48={1,-1,sqrt(2),-sqrt(2))
*/
asm volatile(
__asm__ volatile(
"movaps 48+"MANGLE(sseW2)", %%xmm6\n\t"
"movaps 16+"MANGLE(sseW2)", %%xmm7\n\t"
"xorps %%xmm5, %%xmm5 \n\t"
@ -905,7 +905,7 @@ imdct_do_512_sse(sample_t data[],sample_t delay[], sample_t bias)
two_m_plus_one = two_m<<1;
two_m_plus_one_shl3 = (two_m_plus_one<<3);
buf_offset = buf+128;
asm volatile(
__asm__ volatile(
"mov %0, %%"REG_S" \n\t"
ASMALIGN(4)
"1: \n\t"
@ -937,7 +937,7 @@ imdct_do_512_sse(sample_t data[],sample_t delay[], sample_t bias)
}
/* Post IFFT complex multiply plus IFFT complex conjugate*/
asm volatile(
__asm__ volatile(
"mov $-1024, %%"REG_S" \n\t"
ASMALIGN(4)
"1: \n\t"
@ -960,7 +960,7 @@ imdct_do_512_sse(sample_t data[],sample_t delay[], sample_t bias)
window_ptr = a52_imdct_window;
/* Window and convert to real valued signal */
asm volatile(
__asm__ volatile(
"xor %%"REG_D", %%"REG_D" \n\t" // 0
"xor %%"REG_S", %%"REG_S" \n\t" // 0
"movss %3, %%xmm2 \n\t" // bias
@ -987,7 +987,7 @@ imdct_do_512_sse(sample_t data[],sample_t delay[], sample_t bias)
delay_ptr+=128;
// window_ptr+=128;
asm volatile(
__asm__ volatile(
"mov $1024, %%"REG_D" \n\t" // 512
"xor %%"REG_S", %%"REG_S" \n\t" // 0
"movss %3, %%xmm2 \n\t" // bias
@ -1016,7 +1016,7 @@ imdct_do_512_sse(sample_t data[],sample_t delay[], sample_t bias)
/* The trailing edge of the window goes into the delay line */
delay_ptr = delay;
asm volatile(
__asm__ volatile(
"xor %%"REG_D", %%"REG_D" \n\t" // 0
"xor %%"REG_S", %%"REG_S" \n\t" // 0
ASMALIGN(4)
@ -1038,7 +1038,7 @@ imdct_do_512_sse(sample_t data[],sample_t delay[], sample_t bias)
delay_ptr+=128;
// window_ptr-=128;
asm volatile(
__asm__ volatile(
"mov $1024, %%"REG_D" \n\t" // 1024
"xor %%"REG_S", %%"REG_S" \n\t" // 0
ASMALIGN(4)

4
liba52/imdct_3dnow.h
View File

@ -431,7 +431,7 @@ imdct_do_512_3dnow
*/
FFT_128P_3DNOW (&buf[0]);
// asm volatile ("femms \n\t":::"memory");
// __asm__ volatile ("femms \n\t":::"memory");
/* Post IFFT complex multiply plus IFFT complex conjugate*/
#if 1
@ -489,7 +489,7 @@ imdct_do_512_3dnow
/* Window and convert to real valued signal */
#if 1
asm volatile (
__asm__ volatile (
"movd (%0), %%mm3 \n\t"
"punpckldq %%mm3, %%mm3 \n\t"
:: "r" (&bias)

122
liba52/liba52_changes.diff
View File

@ -1,6 +1,6 @@
--- include/a52.h 2006-06-12 15:04:57.000000000 +0200
+++ liba52/a52.h 2006-06-05 02:23:02.000000000 +0200
@@ -59,4 +63,9 @@
@@ -63,4 +63,9 @@
int a52_block (a52_state_t * state);
void a52_free (a52_state_t * state);
@ -12,7 +12,7 @@
#endif /* A52_H */
--- liba52/a52_internal.h 2006-06-12 15:05:07.000000000 +0200
+++ liba52/a52_internal.h 2006-06-05 02:23:02.000000000 +0200
@@ -103,18 +107,34 @@
@@ -107,18 +107,34 @@
#define DELTA_BIT_NONE (2)
#define DELTA_BIT_RESERVED (3)
@ -52,7 +52,7 @@
+void imdct_do_512 (sample_t * data, sample_t * delay, sample_t bias);
--- liba52/bitstream.c 2006-06-12 15:05:07.000000000 +0200
+++ liba52/bitstream.c 2006-06-05 02:23:02.000000000 +0200
@@ -31,6 +35,10 @@
@@ -35,6 +35,10 @@
#define BUFFER_SIZE 4096
@ -63,7 +63,7 @@
void a52_bitstream_set_ptr (a52_state_t * state, uint8_t * buf)
{
int align;
@@ -38,6 +46,9 @@
@@ -42,6 +46,9 @@
align = (long)buf & 3;
state->buffer_start = (uint32_t *) (buf - align);
state->bits_left = 0;
@ -75,7 +75,7 @@
--- liba52/bitstream.h 2006-06-12 15:05:07.000000000 +0200
+++ liba52/bitstream.h 2006-06-05 02:23:02.000000000 +0200
@@ -21,6 +25,42 @@
@@ -25,6 +25,42 @@
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
@ -118,7 +118,7 @@
/* (stolen from the kernel) */
#ifdef WORDS_BIGENDIAN
@@ -28,7 +74,7 @@
@@ -32,7 +74,7 @@
#else
@ -127,7 +127,7 @@
# define swab32(x) __i386_swab32(x)
static inline const uint32_t __i386_swab32(uint32_t x)
@@ -39,19 +85,34 @@
@@ -43,19 +85,34 @@
# else
@ -166,7 +166,7 @@
uint32_t result;
if (num_bits < state->bits_left) {
@@ -61,10 +122,29 @@
@@ -65,10 +122,29 @@
}
return a52_bitstream_get_bh (state, num_bits);
@ -196,7 +196,7 @@
int32_t result;
if (num_bits < state->bits_left) {
@@ -74,4 +154,5 @@
@@ -78,4 +154,5 @@
}
return a52_bitstream_get_bh_2 (state, num_bits);
@ -204,7 +204,7 @@
}
--- liba52/downmix.c 2006-06-12 15:17:53.000000000 +0200
+++ liba52/downmix.c 2006-06-05 02:23:02.000000000 +0200
@@ -19,18 +23,46 @@
@@ -23,18 +23,46 @@
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
@ -251,7 +251,7 @@
int a52_downmix_init (int input, int flags, sample_t * level,
sample_t clev, sample_t slev)
{
@@ -447,7 +479,7 @@
@@ -451,7 +479,7 @@
samples[i] = 0;
}
@ -260,7 +260,7 @@
sample_t clev, sample_t slev)
{
switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) {
@@ -559,7 +591,7 @@
@@ -563,7 +591,7 @@
break;
case CONVERT (A52_3F2R, A52_2F1R):
@ -269,7 +269,7 @@
move2to1 (samples + 768, samples + 512, bias);
break;
@@ -583,12 +615,12 @@
@@ -587,12 +615,12 @@
break;
case CONVERT (A52_3F1R, A52_3F2R):
@ -284,7 +284,7 @@
{
switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) {
@@ -653,3 +685,1104 @@
@@ -657,3 +685,1104 @@
goto mix_31to21;
}
}
@ -292,7 +292,7 @@
+#if defined(ARCH_X86) || defined(ARCH_X86_64)
+static void mix2to1_SSE (sample_t * dest, sample_t * src, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movlps %2, %%xmm7 \n\t"
+ "shufps $0x00, %%xmm7, %%xmm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -315,7 +315,7 @@
+
+static void mix3to1_SSE (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movlps %1, %%xmm7 \n\t"
+ "shufps $0x00, %%xmm7, %%xmm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -336,7 +336,7 @@
+
+static void mix4to1_SSE (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movlps %1, %%xmm7 \n\t"
+ "shufps $0x00, %%xmm7, %%xmm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -358,7 +358,7 @@
+
+static void mix5to1_SSE (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movlps %1, %%xmm7 \n\t"
+ "shufps $0x00, %%xmm7, %%xmm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -381,7 +381,7 @@
+
+static void mix3to2_SSE (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movlps %1, %%xmm7 \n\t"
+ "shufps $0x00, %%xmm7, %%xmm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -404,7 +404,7 @@
+
+static void mix21to2_SSE (sample_t * left, sample_t * right, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movlps %2, %%xmm7 \n\t"
+ "shufps $0x00, %%xmm7, %%xmm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -427,7 +427,7 @@
+
+static void mix21toS_SSE (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movlps %1, %%xmm7 \n\t"
+ "shufps $0x00, %%xmm7, %%xmm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -451,7 +451,7 @@
+
+static void mix31to2_SSE (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movlps %1, %%xmm7 \n\t"
+ "shufps $0x00, %%xmm7, %%xmm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -475,7 +475,7 @@
+
+static void mix31toS_SSE (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movlps %1, %%xmm7 \n\t"
+ "shufps $0x00, %%xmm7, %%xmm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -501,7 +501,7 @@
+
+static void mix22toS_SSE (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movlps %1, %%xmm7 \n\t"
+ "shufps $0x00, %%xmm7, %%xmm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -526,7 +526,7 @@
+
+static void mix32to2_SSE (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movlps %1, %%xmm7 \n\t"
+ "shufps $0x00, %%xmm7, %%xmm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -550,7 +550,7 @@
+
+static void mix32toS_SSE (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movlps %1, %%xmm7 \n\t"
+ "shufps $0x00, %%xmm7, %%xmm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -577,7 +577,7 @@
+
+static void move2to1_SSE (sample_t * src, sample_t * dest, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movlps %2, %%xmm7 \n\t"
+ "shufps $0x00, %%xmm7, %%xmm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -600,7 +600,7 @@
+
+static void zero_MMX(sample_t * samples)
+{
+ asm volatile(
+ __asm__ volatile(
+ "mov $-1024, %%"REG_S" \n\t"
+ "pxor %%mm0, %%mm0 \n\t"
+ ASMALIGN(4)
@ -826,7 +826,7 @@
+
+static void mix2to1_3dnow (sample_t * dest, sample_t * src, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movd %2, %%mm7 \n\t"
+ "punpckldq %2, %%mm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -857,7 +857,7 @@
+
+static void mix3to1_3dnow (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movd %1, %%mm7 \n\t"
+ "punpckldq %1, %%mm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -884,7 +884,7 @@
+
+static void mix4to1_3dnow (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movd %1, %%mm7 \n\t"
+ "punpckldq %1, %%mm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -913,7 +913,7 @@
+
+static void mix5to1_3dnow (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movd %1, %%mm7 \n\t"
+ "punpckldq %1, %%mm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -944,7 +944,7 @@
+
+static void mix3to2_3dnow (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movd %1, %%mm7 \n\t"
+ "punpckldq %1, %%mm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -975,7 +975,7 @@
+
+static void mix21to2_3dnow (sample_t * left, sample_t * right, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movd %2, %%mm7 \n\t"
+ "punpckldq %2, %%mm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -1006,7 +1006,7 @@
+
+static void mix21toS_3dnow (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movd %1, %%mm7 \n\t"
+ "punpckldq %1, %%mm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -1039,7 +1039,7 @@
+
+static void mix31to2_3dnow (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movd %1, %%mm7 \n\t"
+ "punpckldq %1, %%mm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -1072,7 +1072,7 @@
+
+static void mix31toS_3dnow (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movd %1, %%mm7 \n\t"
+ "punpckldq %1, %%mm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -1109,7 +1109,7 @@
+
+static void mix22toS_3dnow (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movd %1, %%mm7 \n\t"
+ "punpckldq %1, %%mm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -1144,7 +1144,7 @@
+
+static void mix32to2_3dnow (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movd %1, %%mm7 \n\t"
+ "punpckldq %1, %%mm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -1178,7 +1178,7 @@
+/* todo: should be optimized better */
+static void mix32toS_3dnow (sample_t * samples, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "mov $-1024, %%"REG_S" \n\t"
+ ASMALIGN(4)
+ "1: \n\t"
@ -1217,7 +1217,7 @@
+
+static void move2to1_3dnow (sample_t * src, sample_t * dest, sample_t bias)
+{
+ asm volatile(
+ __asm__ volatile(
+ "movd %2, %%mm7 \n\t"
+ "punpckldq %2, %%mm7 \n\t"
+ "mov $-1024, %%"REG_S" \n\t"
@ -1385,13 +1385,13 @@
+ memcpy (samples + 1024, samples + 768, 256 * sizeof (sample_t));
+ break;
+ }
+ __asm __volatile("femms":::"memory");
+ __asm__ volatile("femms":::"memory");
+}
+
+#endif // ARCH_X86 || ARCH_X86_64
--- liba52/imdct.c 2008-02-19 00:18:33.000000000 +0100
+++ liba52/imdct.c 2008-02-19 00:16:40.000000000 +0100
@@ -22,6 +26,11 @@
@@ -26,6 +26,11 @@
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
@ -1403,7 +1403,7 @@
*/
#include "config.h"
@@ -39,12 +48,49 @@
@@ -43,12 +48,49 @@
#include "a52.h"
#include "a52_internal.h"
#include "mm_accel.h"
@ -1453,7 +1453,7 @@
static uint8_t fftorder[] = {
0,128, 64,192, 32,160,224, 96, 16,144, 80,208,240,112, 48,176,
8,136, 72,200, 40,168,232,104,248,120, 56,184, 24,152,216, 88,
@@ -56,6 +102,40 @@
@@ -60,6 +102,40 @@
6,134, 70,198, 38,166,230,102,246,118, 54,182, 22,150,214, 86
};
@ -1494,7 +1494,7 @@
/* Root values for IFFT */
static sample_t roots16[3];
static sample_t roots32[7];
@@ -241,7 +321,7 @@
@@ -245,7 +321,7 @@
ifft_pass (buf, roots128 - 32, 32);
}
@ -1503,7 +1503,7 @@
{
int i, k;
sample_t t_r, t_i, a_r, a_i, b_r, b_i, w_1, w_2;
@@ -285,6 +365,701 @@
@@ -289,6 +365,701 @@
}
}
@ -1892,7 +1892,7 @@
+
+ /* Pre IFFT complex multiply plus IFFT cmplx conjugate */
+ /* Bit reversed shuffling */
+ asm volatile(
+ __asm__ volatile(
+ "xor %%"REG_S", %%"REG_S" \n\t"
+ "lea "MANGLE(bit_reverse_512)", %%"REG_a"\n\t"
+ "mov $1008, %%"REG_D" \n\t"
@ -1952,7 +1952,7 @@
+
+ /* 1. iteration */
+ // Note w[0][0]={1,0}
+ asm volatile(
+ __asm__ volatile(
+ "xorps %%xmm1, %%xmm1 \n\t"
+ "xorps %%xmm2, %%xmm2 \n\t"
+ "mov %0, %%"REG_S" \n\t"
@ -1974,7 +1974,7 @@
+
+ /* 2. iteration */
+ // Note w[1]={{1,0}, {0,-1}}
+ asm volatile(
+ __asm__ volatile(
+ "movaps "MANGLE(ps111_1)", %%xmm7\n\t" // 1,1,1,-1
+ "mov %0, %%"REG_S" \n\t"
+ ASMALIGN(4)
@ -2002,7 +2002,7 @@
+ Note sseW2+32={0,0,-sqrt(2),-sqrt(2))
+ Note sseW2+48={1,-1,sqrt(2),-sqrt(2))
+*/
+ asm volatile(
+ __asm__ volatile(
+ "movaps 48+"MANGLE(sseW2)", %%xmm6\n\t"
+ "movaps 16+"MANGLE(sseW2)", %%xmm7\n\t"
+ "xorps %%xmm5, %%xmm5 \n\t"
@ -2047,7 +2047,7 @@
+ two_m_plus_one = two_m<<1;
+ two_m_plus_one_shl3 = (two_m_plus_one<<3);
+ buf_offset = buf+128;
+ asm volatile(
+ __asm__ volatile(
+ "mov %0, %%"REG_S" \n\t"
+ ASMALIGN(4)
+ "1: \n\t"
@ -2079,7 +2079,7 @@
+ }
+
+ /* Post IFFT complex multiply plus IFFT complex conjugate*/
+ asm volatile(
+ __asm__ volatile(
+ "mov $-1024, %%"REG_S" \n\t"
+ ASMALIGN(4)
+ "1: \n\t"
@ -2102,7 +2102,7 @@
+ window_ptr = a52_imdct_window;
+
+ /* Window and convert to real valued signal */
+ asm volatile(
+ __asm__ volatile(
+ "xor %%"REG_D", %%"REG_D" \n\t" // 0
+ "xor %%"REG_S", %%"REG_S" \n\t" // 0
+ "movss %3, %%xmm2 \n\t" // bias
@ -2129,7 +2129,7 @@
+ delay_ptr+=128;
+// window_ptr+=128;
+
+ asm volatile(
+ __asm__ volatile(
+ "mov $1024, %%"REG_D" \n\t" // 512
+ "xor %%"REG_S", %%"REG_S" \n\t" // 0
+ "movss %3, %%xmm2 \n\t" // bias
@ -2158,7 +2158,7 @@
+ /* The trailing edge of the window goes into the delay line */
+ delay_ptr = delay;
+
+ asm volatile(
+ __asm__ volatile(
+ "xor %%"REG_D", %%"REG_D" \n\t" // 0
+ "xor %%"REG_S", %%"REG_S" \n\t" // 0
+ ASMALIGN(4)
@ -2180,7 +2180,7 @@
+ delay_ptr+=128;
+// window_ptr-=128;
+
+ asm volatile(
+ __asm__ volatile(
+ "mov $1024, %%"REG_D" \n\t" // 1024
+ "xor %%"REG_S", %%"REG_S" \n\t" // 0
+ ASMALIGN(4)
@ -2205,7 +2205,7 @@
void a52_imdct_256(sample_t * data, sample_t * delay, sample_t bias)
{
int i, k;
@@ -364,7 +1145,7 @@
@@ -368,7 +1145,7 @@
void a52_imdct_init (uint32_t mm_accel)
{
@ -2214,7 +2214,7 @@
double sum;
/* compute imdct window - kaiser-bessel derived window, alpha = 5.0 */
@@ -416,6 +1197,99 @@
@@ -420,6 +1197,99 @@
post2[i].real = cos ((M_PI / 128) * (i + 0.5));
post2[i].imag = sin ((M_PI / 128) * (i + 0.5));
}
@ -2314,7 +2314,7 @@
#ifdef LIBA52_DJBFFT
if (mm_accel & MM_ACCEL_DJBFFT) {
@@ -426,7 +1300,5 @@
@@ -430,7 +1300,5 @@
#endif
{
fprintf (stderr, "No accelerated IMDCT transform found\n");
@ -2324,7 +2324,7 @@
}
--- include/mm_accel.h 2006-06-12 15:05:00.000000000 +0200
+++ liba52/mm_accel.h 2006-06-05 02:23:04.000000000 +0200
@@ -30,7 +34,12 @@
@@ -34,7 +34,12 @@
/* x86 accelerations */
#define MM_ACCEL_X86_MMX 0x80000000
#define MM_ACCEL_X86_3DNOW 0x40000000

22
liba52/resample_mmx.c
View File

@ -38,7 +38,7 @@ static uint64_t attribute_used __attribute__((aligned(8))) wm1100= 0xFFFFFFFF000
static int a52_resample_MONO_to_5_MMX(float * _f, int16_t * s16){
int32_t * f = (int32_t *) _f;
asm volatile(
__asm__ volatile(
"mov $-512, %%"REG_S" \n\t"
"movq "MANGLE(magicF2W)", %%mm7 \n\t"
"movq "MANGLE(wm1100)", %%mm3 \n\t"
@ -77,7 +77,7 @@ static int a52_resample_STEREO_to_2_MMX(float * _f, int16_t * s16){
int32_t * f = (int32_t *) _f;
/* benchmark scores are 0.3% better with SSE but we would need to set bias=0 and premultiply it
#ifdef HAVE_SSE
asm volatile(
__asm__ volatile(
"mov $-1024, %%"REG_S" \n\t"
"1: \n\t"
"cvtps2pi (%1, %%"REG_S"), %%mm0\n\t"
@ -93,7 +93,7 @@ static int a52_resample_STEREO_to_2_MMX(float * _f, int16_t * s16){
:: "r" (s16+512), "r" (f+256)
:"%"REG_S, "memory"
);*/
asm volatile(
__asm__ volatile(
"mov $-1024, %%"REG_S" \n\t"
"movq "MANGLE(magicF2W)", %%mm7 \n\t"
"1: \n\t"
@ -123,7 +123,7 @@ static int a52_resample_STEREO_to_2_MMX(float * _f, int16_t * s16){
static int a52_resample_3F_to_5_MMX(float * _f, int16_t * s16){
int32_t * f = (int32_t *) _f;
asm volatile(
__asm__ volatile(
"mov $-1024, %%"REG_S" \n\t"
"movq "MANGLE(magicF2W)", %%mm7 \n\t"
"pxor %%mm6, %%mm6 \n\t"
@ -177,7 +177,7 @@ static int a52_resample_3F_to_5_MMX(float * _f, int16_t * s16){
static int a52_resample_2F_2R_to_4_MMX(float * _f, int16_t * s16){
int32_t * f = (int32_t *) _f;
asm volatile(
__asm__ volatile(
"mov $-1024, %%"REG_S" \n\t"
"movq "MANGLE(magicF2W)", %%mm7 \n\t"
"1: \n\t"
@ -228,7 +228,7 @@ static int a52_resample_2F_2R_to_4_MMX(float * _f, int16_t * s16){
static int a52_resample_3F_2R_to_5_MMX(float * _f, int16_t * s16){
int32_t * f = (int32_t *) _f;
asm volatile(
__asm__ volatile(
"mov $-1024, %%"REG_S" \n\t"
"movq "MANGLE(magicF2W)", %%mm7 \n\t"
"1: \n\t"
@ -287,7 +287,7 @@ static int a52_resample_3F_2R_to_5_MMX(float * _f, int16_t * s16){
static int a52_resample_MONO_LFE_to_6_MMX(float * _f, int16_t * s16){
int32_t * f = (int32_t *) _f;
asm volatile(
__asm__ volatile(
"mov $-1024, %%"REG_S" \n\t"
"movq "MANGLE(magicF2W)", %%mm7 \n\t"
"pxor %%mm6, %%mm6 \n\t"
@ -327,7 +327,7 @@ static int a52_resample_MONO_LFE_to_6_MMX(float * _f, int16_t * s16){
static int a52_resample_STEREO_LFE_to_6_MMX(float * _f, int16_t * s16){
int32_t * f = (int32_t *) _f;
asm volatile(
__asm__ volatile(
"mov $-1024, %%"REG_S" \n\t"
"movq "MANGLE(magicF2W)", %%mm7 \n\t"
"pxor %%mm6, %%mm6 \n\t"
@ -365,7 +365,7 @@ static int a52_resample_STEREO_LFE_to_6_MMX(float * _f, int16_t * s16){
static int a52_resample_3F_LFE_to_6_MMX(float * _f, int16_t * s16){
int32_t * f = (int32_t *) _f;
asm volatile(
__asm__ volatile(
"mov $-1024, %%"REG_S" \n\t"
"movq "MANGLE(magicF2W)", %%mm7 \n\t"
"pxor %%mm6, %%mm6 \n\t"
@ -405,7 +405,7 @@ static int a52_resample_3F_LFE_to_6_MMX(float * _f, int16_t * s16){
static int a52_resample_2F_2R_LFE_to_6_MMX(float * _f, int16_t * s16){
int32_t * f = (int32_t *) _f;
asm volatile(
__asm__ volatile(
"mov $-1024, %%"REG_S" \n\t"
"movq "MANGLE(magicF2W)", %%mm7 \n\t"
// "pxor %%mm6, %%mm6 \n\t"
@ -451,7 +451,7 @@ static int a52_resample_2F_2R_LFE_to_6_MMX(float * _f, int16_t * s16){
static int a52_resample_3F_2R_LFE_to_6_MMX(float * _f, int16_t * s16){
int32_t * f = (int32_t *) _f;
asm volatile(
__asm__ volatile(
"mov $-1024, %%"REG_S" \n\t"
"movq "MANGLE(magicF2W)", %%mm7 \n\t"
// "pxor %%mm6, %%mm6 \n\t"

2
liba52/test.c
View File

@ -46,7 +46,7 @@ void mp_msg( int x, const char *format, ... ) // stub for cpudetect.c
static inline long long rdtsc()
{
long long l;
asm volatile( "rdtsc\n\t"
__asm__ volatile("rdtsc\n\t"
: "=A" (l)
);
// printf("%d\n", int(l/1000));

4
libmpcodecs/dec_video.c
View File

@ -374,10 +374,10 @@ void *decode_video(sh_video_t *sh_video, unsigned char *start, int in_size,
// some codecs are broken, and doesn't restore MMX state :(
// it happens usually with broken/damaged files.
if (gCpuCaps.has3DNow) {
__asm __volatile ("femms\n\t":::"memory");
__asm__ __volatile ("femms\n\t":::"memory");
}
else if (gCpuCaps.hasMMX) {
__asm __volatile ("emms\n\t":::"memory");
__asm__ __volatile ("emms\n\t":::"memory");
}
#endif

6
libmpcodecs/pullup.c
View File

@ -14,7 +14,7 @@
static int diff_y_mmx(unsigned char *a, unsigned char *b, int s)
{
int ret;
asm volatile (
__asm__ volatile (
"movl $4, %%ecx \n\t"
"pxor %%mm4, %%mm4 \n\t"
"pxor %%mm7, %%mm7 \n\t"
@ -61,7 +61,7 @@ static int diff_y_mmx(unsigned char *a, unsigned char *b, int s)
static int licomb_y_mmx(unsigned char *a, unsigned char *b, int s)
{
int ret;
asm volatile (
__asm__ volatile (
"movl $4, %%ecx \n\t"
"pxor %%mm6, %%mm6 \n\t"
"pxor %%mm7, %%mm7 \n\t"
@ -150,7 +150,7 @@ static int licomb_y_mmx(unsigned char *a, unsigned char *b, int s)
static int var_y_mmx(unsigned char *a, unsigned char *b, int s)
{
int ret;
asm volatile (
__asm__ volatile (
"movl $3, %%ecx \n\t"
"pxor %%mm4, %%mm4 \n\t"
"pxor %%mm7, %%mm7 \n\t"

2
libmpcodecs/vf_decimate.c
View File

@ -23,7 +23,7 @@ struct vf_priv_s {
static int diff_MMX(unsigned char *old, unsigned char *new, int os, int ns)
{
volatile short out[4];
asm (
__asm__ (
"movl $8, %%ecx \n\t"
"pxor %%mm4, %%mm4 \n\t"
"pxor %%mm7, %%mm7 \n\t"

2
libmpcodecs/vf_divtc.c
View File

@ -37,7 +37,7 @@ struct vf_priv_s
static int diff_MMX(unsigned char *old, unsigned char *new, int os, int ns)
{
volatile short out[4];
asm (
__asm__ (
"movl $8, %%ecx \n\t"
"pxor %%mm4, %%mm4 \n\t"
"pxor %%mm7, %%mm7 \n\t"

4
libmpcodecs/vf_eq.c
View File

@ -44,7 +44,7 @@ static void process_MMX(unsigned char *dest, int dstride, unsigned char *src, in
contvec[0] = contvec[1] = contvec[2] = contvec[3] = contrast;
while (h--) {
asm volatile (
__asm__ volatile (
"movq (%5), %%mm3 \n\t"
"movq (%6), %%mm4 \n\t"
"pxor %%mm0, %%mm0 \n\t"
@ -82,7 +82,7 @@ static void process_MMX(unsigned char *dest, int dstride, unsigned char *src, in
src += sstep;
dest += dstep;
}
asm volatile ( "emms \n\t" ::: "memory" );
__asm__ volatile ( "emms \n\t" ::: "memory" );
}
#endif

4
libmpcodecs/vf_eq2.c
View File

@ -130,7 +130,7 @@ void affine_1d_MMX (eq2_param_t *par, unsigned char *dst, unsigned char *src,
dstep = dstride - w;
while (h-- > 0) {
asm volatile (
__asm__ volatile (
"movq (%5), %%mm3 \n\t"
"movq (%6), %%mm4 \n\t"
"pxor %%mm0, %%mm0 \n\t"
@ -170,7 +170,7 @@ void affine_1d_MMX (eq2_param_t *par, unsigned char *dst, unsigned char *src,
dst += dstep;
}
asm volatile ( "emms \n\t" ::: "memory" );
__asm__ volatile ( "emms \n\t" ::: "memory" );
}
#endif

22
libmpcodecs/vf_filmdint.c
View File

@ -368,13 +368,13 @@ block_metrics_faster_c(unsigned char *a, unsigned char *b, int as, int bs,
#define MEQ(X,Y) ((X).even == (Y).even && (X).odd == (Y).odd && (X).temp == (Y).temp && (X).noise == (Y).noise)
#define BLOCK_METRICS_TEMPLATE() \
asm volatile("pxor %mm7, %mm7\n\t" /* The result is colleted in mm7 */ \
__asm__ volatile("pxor %mm7, %mm7\n\t" /* The result is colleted in mm7 */ \
"pxor %mm6, %mm6\n\t" /* Temp to stay at 0 */ \
); \
a -= as; \
b -= bs; \
do { \
asm volatile( \
__asm__ volatile( \
"movq (%0,%2), %%mm0\n\t" \
"movq (%1,%3), %%mm1\n\t" /* mm1 = even */ \
PSADBW(%%mm1, %%mm0, %%mm4, %%mm6) \
@ -439,7 +439,7 @@ block_metrics_3dnow(unsigned char *a, unsigned char *b, int as, int bs,
static const unsigned long long ones = 0x0101010101010101ull;
BLOCK_METRICS_TEMPLATE();
asm volatile("movq %%mm7, %0\n\temms" : "=m" (tm));
__asm__ volatile("movq %%mm7, %0\n\temms" : "=m" (tm));
get_block_stats(&tm, p, s);
#endif
return tm;
@ -471,7 +471,7 @@ block_metrics_mmx2(unsigned char *a, unsigned char *b, int as, int bs,
#ifdef DEBUG
struct frame_stats ts = *s;
#endif
asm volatile("prefetcht0 (%0,%2)\n\t"
__asm__ volatile("prefetcht0 (%0,%2)\n\t"
"prefetcht0 (%1,%3)\n\t" :
: "r" (a), "r" (b),
"r" (prefetch_line * as), "r" (prefetch_line * bs));
@ -479,7 +479,7 @@ block_metrics_mmx2(unsigned char *a, unsigned char *b, int as, int bs,
BLOCK_METRICS_TEMPLATE();
s->num_blocks++;
asm volatile(
__asm__ volatile(
"movq %3, %%mm0\n\t"
"movq %%mm7, %%mm1\n\t"
"psubusw %%mm0, %%mm1\n\t"
@ -525,7 +525,7 @@ block_metrics_mmx2(unsigned char *a, unsigned char *b, int as, int bs,
s->interlaced_high += interlaced >> 16;
s->interlaced_low += interlaced;
} else {
asm volatile(
__asm__ volatile(
"pcmpeqw %%mm0, %%mm0\n\t" /* -1 */
"psubw %%mm0, %%mm4\n\t"
"psubw %%mm0, %%mm5\n\t"
@ -539,7 +539,7 @@ block_metrics_mmx2(unsigned char *a, unsigned char *b, int as, int bs,
: "=m" (s->tiny), "=m" (s->low), "=m" (s->high)
);
asm volatile(
__asm__ volatile(
"pshufw $0, %2, %%mm0\n\t"
"psubusw %%mm7, %%mm0\n\t"
"pcmpeqw %%mm6, %%mm0\n\t" /* 0 if below sad_thres */
@ -556,7 +556,7 @@ block_metrics_mmx2(unsigned char *a, unsigned char *b, int as, int bs,
);
}
asm volatile(
__asm__ volatile(
"movq %%mm7, (%1)\n\t"
PMAXUW((%0), %%mm7)
"movq %%mm7, (%0)\n\t"
@ -597,7 +597,7 @@ dint_copy_line_mmx2(unsigned char *dst, unsigned char *a, long bos,
#else
unsigned long len = (w+7) >> 3;
int ret;
asm volatile (
__asm__ volatile (
"pxor %%mm6, %%mm6 \n\t" /* deinterlaced pixel counter */
"movd %0, %%mm7 \n\t"
"punpcklbw %%mm7, %%mm7 \n\t"
@ -607,7 +607,7 @@ dint_copy_line_mmx2(unsigned char *dst, unsigned char *a, long bos,
: "rm" (t)
);
do {
asm volatile (
__asm__ volatile (
"movq (%0), %%mm0\n\t"
"movq (%0,%3,2), %%mm1\n\t"
"movq %%mm0, (%2)\n\t"
@ -639,7 +639,7 @@ dint_copy_line_mmx2(unsigned char *dst, unsigned char *a, long bos,
dst += 8;
} while (--len);
asm volatile ("pxor %%mm7, %%mm7 \n\t"
__asm__ volatile ("pxor %%mm7, %%mm7 \n\t"
"psadbw %%mm6, %%mm7 \n\t"
"movd %%mm7, %0 \n\t"
"emms \n\t"

16
libmpcodecs/vf_fspp.c
View File

@ -187,7 +187,7 @@ static void store_slice_mmx(uint8_t *dst, int16_t *src, long dst_stride, long sr
width = (width+7)&~7;
dst_stride-=width;
//src_stride=(src_stride-width)*2;
asm volatile(
__asm__ volatile(
"mov %5, %%"REG_d" \n\t"
"mov %6, %%"REG_S" \n\t"
"mov %7, %%"REG_D" \n\t"
@ -255,7 +255,7 @@ static void store_slice2_mmx(uint8_t *dst, int16_t *src, long dst_stride, long s
width = (width+7)&~7;
dst_stride-=width;
//src_stride=(src_stride-width)*2;
asm volatile(
__asm__ volatile(
"mov %5, %%"REG_d" \n\t"
"mov %6, %%"REG_S" \n\t"
"mov %7, %%"REG_D" \n\t"
@ -318,7 +318,7 @@ static void store_slice2_mmx(uint8_t *dst, int16_t *src, long dst_stride, long s
static void mul_thrmat_mmx(struct vf_priv_s *p, int q)
{
uint64_t *adr=&p->threshold_mtx_noq[0];
asm volatile(
__asm__ volatile(
"movd %0, %%mm7 \n\t"
"add $8*8*2, %%"REG_D" \n\t"
"movq 0*8(%%"REG_S"), %%mm0 \n\t"
@ -558,10 +558,10 @@ static int put_image(struct vf_instance_s* vf, mp_image_t *mpi, double pts)
}
#ifdef HAVE_MMX
if(gCpuCaps.hasMMX) asm volatile ("emms\n\t");
if(gCpuCaps.hasMMX) __asm__ volatile ("emms\n\t");
#endif
#ifdef HAVE_MMX2
if(gCpuCaps.hasMMX2) asm volatile ("sfence\n\t");
if(gCpuCaps.hasMMX2) __asm__ volatile ("sfence\n\t");
#endif
return vf_next_put_image(vf,dmpi, pts);
}
@ -868,7 +868,7 @@ static void column_fidct_c(int16_t* thr_adr, DCTELEM *data, DCTELEM *output, int
static void column_fidct_mmx(int16_t* thr_adr, DCTELEM *data, DCTELEM *output, int cnt)
{
uint64_t __attribute__((aligned(8))) temps[4];
asm volatile(
__asm__ volatile(
ASMALIGN(4)
"1: \n\t"
"movq "DCTSIZE_S"*0*2(%%"REG_S"), %%mm1 \n\t"
@ -1669,7 +1669,7 @@ static void row_idct_mmx (DCTELEM* workspace,
int16_t* output_adr, int output_stride, int cnt)
{
uint64_t __attribute__((aligned(8))) temps[4];
asm volatile(
__asm__ volatile(
"lea (%%"REG_a",%%"REG_a",2), %%"REG_d" \n\t"
"1: \n\t"
"movq "DCTSIZE_S"*0*2(%%"REG_S"), %%mm0 \n\t"
@ -1935,7 +1935,7 @@ static void row_fdct_c(DCTELEM *data, const uint8_t *pixels, int line_size, int
static void row_fdct_mmx(DCTELEM *data, const uint8_t *pixels, int line_size, int cnt)
{
uint64_t __attribute__((aligned(8))) temps[4];
asm volatile(
__asm__ volatile(
"lea (%%"REG_a",%%"REG_a",2), %%"REG_d" \n\t"
"6: \n\t"
"movd (%%"REG_S"), %%mm0 \n\t"

4
libmpcodecs/vf_halfpack.c
View File

@ -37,7 +37,7 @@ static void halfpack_MMX(unsigned char *dst, unsigned char *src[3],
vinc = srcstride[2] - w/2;
for (h/=2; h; h--) {
asm (
__asm__ (
"pxor %%mm0, %%mm0 \n\t"
ASMALIGN(4)
"1: \n\t"
@ -99,7 +99,7 @@ static void halfpack_MMX(unsigned char *dst, unsigned char *src[3],
v += vinc;
dst += dstinc;
}
asm volatile ( "emms \n\t" ::: "memory" );
__asm__ volatile ( "emms \n\t" ::: "memory" );
}
#endif

6
libmpcodecs/vf_ilpack.c
View File

@ -61,7 +61,7 @@ static void pack_li_1_C(unsigned char *dst, unsigned char *y,
static void pack_nn_MMX(unsigned char *dst, unsigned char *y,
unsigned char *u, unsigned char *v, int w)
{
asm volatile (""
__asm__ volatile (""
ASMALIGN(4)
"1: \n\t"
"movq (%0), %%mm1 \n\t"
@ -91,7 +91,7 @@ static void pack_nn_MMX(unsigned char *dst, unsigned char *y,
static void pack_li_0_MMX(unsigned char *dst, unsigned char *y,
unsigned char *u, unsigned char *v, int w, int us, int vs)
{
asm volatile (""
__asm__ volatile (""
"push %%"REG_BP" \n\t"
#ifdef ARCH_X86_64
"mov %6, %%"REG_BP" \n\t"
@ -199,7 +199,7 @@ static void pack_li_0_MMX(unsigned char *dst, unsigned char *y,
static void pack_li_1_MMX(unsigned char *dst, unsigned char *y,
unsigned char *u, unsigned char *v, int w, int us, int vs)
{
asm volatile (""
__asm__ volatile (""
"push %%"REG_BP" \n\t"
#ifdef ARCH_X86_64
"mov %6, %%"REG_BP" \n\t"

4
libmpcodecs/vf_ivtc.c
View File

@ -46,7 +46,7 @@ static void block_diffs_MMX(struct metrics *m, unsigned char *old, unsigned char
int i;
short out[24]; // output buffer for the partial metrics from the mmx code
asm (
__asm__ (
"movl $4, %%ecx \n\t"
"pxor %%mm4, %%mm4 \n\t" // 4 even difference sums
"pxor %%mm5, %%mm5 \n\t" // 4 odd difference sums
@ -105,7 +105,7 @@ static void block_diffs_MMX(struct metrics *m, unsigned char *old, unsigned char
m->o = out[4]+out[5]+out[6]+out[7];
m->d = m->e + m->o;
asm (
__asm__ (
// First loop to measure first four columns
"movl $4, %%ecx \n\t"
"pxor %%mm4, %%mm4 \n\t" // Past spacial noise

10
libmpcodecs/vf_noise.c
View File

@ -150,7 +150,7 @@ static inline void lineNoise_MMX(uint8_t *dst, uint8_t *src, int8_t *noise, int
long mmx_len= len&(~7);
noise+=shift;
asm volatile(
__asm__ volatile(
"mov %3, %%"REG_a" \n\t"
"pcmpeqb %%mm7, %%mm7 \n\t"
"psllw $15, %%mm7 \n\t"
@ -179,7 +179,7 @@ static inline void lineNoise_MMX2(uint8_t *dst, uint8_t *src, int8_t *noise, int
long mmx_len= len&(~7);
noise+=shift;
asm volatile(
__asm__ volatile(
"mov %3, %%"REG_a" \n\t"
"pcmpeqb %%mm7, %%mm7 \n\t"
"psllw $15, %%mm7 \n\t"
@ -220,7 +220,7 @@ static inline void lineNoise_C(uint8_t *dst, uint8_t *src, int8_t *noise, int le
static inline void lineNoiseAvg_MMX(uint8_t *dst, uint8_t *src, int len, int8_t **shift){
long mmx_len= len&(~7);
asm volatile(
__asm__ volatile(
"mov %5, %%"REG_a" \n\t"
ASMALIGN(4)
"1: \n\t"
@ -357,10 +357,10 @@ static int put_image(struct vf_instance_s* vf, mp_image_t *mpi, double pts){
vf_clone_mpi_attributes(dmpi, mpi);
#ifdef HAVE_MMX
if(gCpuCaps.hasMMX) asm volatile ("emms\n\t");
if(gCpuCaps.hasMMX) __asm__ volatile ("emms\n\t");
#endif
#ifdef HAVE_MMX2
if(gCpuCaps.hasMMX2) asm volatile ("sfence\n\t");
if(gCpuCaps.hasMMX2) __asm__ volatile ("sfence\n\t");
#endif
return vf_next_put_image(vf,dmpi, pts);

6
libmpcodecs/vf_pp7.c
View File

@ -157,7 +157,7 @@ static void dctB_c(DCTELEM *dst, DCTELEM *src){
#ifdef HAVE_MMX
static void dctB_mmx(DCTELEM *dst, DCTELEM *src){
asm volatile (
__asm__ volatile (
"movq (%0), %%mm0 \n\t"
"movq 1*4*2(%0), %%mm1 \n\t"
"paddw 6*4*2(%0), %%mm0 \n\t"
@ -398,10 +398,10 @@ static int put_image(struct vf_instance_s* vf, mp_image_t *mpi, double pts){
}
#ifdef HAVE_MMX
if(gCpuCaps.hasMMX) asm volatile ("emms\n\t");
if(gCpuCaps.hasMMX) __asm__ volatile ("emms\n\t");
#endif
#ifdef HAVE_MMX2
if(gCpuCaps.hasMMX2) asm volatile ("sfence\n\t");
if(gCpuCaps.hasMMX2) __asm__ volatile ("sfence\n\t");
#endif
return vf_next_put_image(vf,dmpi, pts);

2
libmpcodecs/vf_sab.c
View File

@ -140,7 +140,7 @@ static int config(struct vf_instance_s* vf,
unsigned int flags, unsigned int outfmt){
int sw, sh;
//asm volatile("emms\n\t");
//__asm__ volatile("emms\n\t");
allocStuff(&vf->priv->luma, width, height);
getSubSampleFactors(&sw, &sh, outfmt);

2
libmpcodecs/vf_scale.c
View File

@ -558,7 +558,7 @@ void sws_getFlagsAndFilterFromCmdLine(int *flags, SwsFilter **srcFilterParam, Sw
#ifdef ARCH_X86
if(gCpuCaps.hasMMX)
asm volatile("emms\n\t"::: "memory"); //FIXME this shouldnt be required but it IS (even for non mmx versions)
__asm__ volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)
#endif
if(firstTime)
{

10
libmpcodecs/vf_spp.c
View File

@ -153,7 +153,7 @@ static void hardthresh_mmx(DCTELEM dst[64], DCTELEM src[64], int qp, uint8_t *pe
threshold1= qp*((1<<4) - bias) - 1;
asm volatile(
__asm__ volatile(
#define REQUANT_CORE(dst0, dst1, dst2, dst3, src0, src1, src2, src3) \
"movq " #src0 ", %%mm0 \n\t"\
"movq " #src1 ", %%mm1 \n\t"\
@ -221,7 +221,7 @@ static void softthresh_mmx(DCTELEM dst[64], DCTELEM src[64], int qp, uint8_t *pe
threshold1= qp*((1<<4) - bias) - 1;
asm volatile(
__asm__ volatile(
#undef REQUANT_CORE
#define REQUANT_CORE(dst0, dst1, dst2, dst3, src0, src1, src2, src3) \
"movq " #src0 ", %%mm0 \n\t"\
@ -334,7 +334,7 @@ static void store_slice_mmx(uint8_t *dst, int16_t *src, int dst_stride, int src_
for(y=0; y<height; y++){
uint8_t *dst1= dst;
int16_t *src1= src;
asm volatile(
__asm__ volatile(
"movq (%3), %%mm3 \n\t"
"movq (%3), %%mm4 \n\t"
"movd %4, %%mm2 \n\t"
@ -500,10 +500,10 @@ static int put_image(struct vf_instance_s* vf, mp_image_t *mpi, double pts){
}
#ifdef HAVE_MMX
if(gCpuCaps.hasMMX) asm volatile ("emms\n\t");
if(gCpuCaps.hasMMX) __asm__ volatile ("emms\n\t");
#endif
#ifdef HAVE_MMX2
if(gCpuCaps.hasMMX2) asm volatile ("sfence\n\t");
if(gCpuCaps.hasMMX2) __asm__ volatile ("sfence\n\t");
#endif
return vf_next_put_image(vf,dmpi, pts);

16
libmpcodecs/vf_tfields.c
View File

@ -58,7 +58,7 @@ static void qpel_li_3DNOW(unsigned char *d, unsigned char *s, int w, int h, int
s += ss;
}
for (i=h-1; i; i--) {
asm volatile(
__asm__ volatile(
"1: \n\t"
"movq (%%"REG_S"), %%mm0 \n\t"
"movq (%%"REG_S",%%"REG_a"), %%mm1 \n\t"
@ -78,7 +78,7 @@ static void qpel_li_3DNOW(unsigned char *d, unsigned char *s, int w, int h, int
s += ss;
}
if (!up) fast_memcpy(d, s, w);
asm volatile("emms \n\t" : : : "memory");
__asm__ volatile("emms \n\t" : : : "memory");
}
#endif
@ -94,7 +94,7 @@ static void qpel_li_MMX2(unsigned char *d, unsigned char *s, int w, int h, int d
s += ss;
}
for (i=h-1; i; i--) {
asm volatile(
__asm__ volatile(
"pxor %%mm7, %%mm7 \n\t"
"2: \n\t"
"movq (%%"REG_S"), %%mm0 \n\t"
@ -115,7 +115,7 @@ static void qpel_li_MMX2(unsigned char *d, unsigned char *s, int w, int h, int d
s += ss;
}
if (!up) fast_memcpy(d, s, w);
asm volatile("emms \n\t" : : : "memory");
__asm__ volatile("emms \n\t" : : : "memory");
}
#endif
@ -131,7 +131,7 @@ static void qpel_li_MMX(unsigned char *d, unsigned char *s, int w, int h, int ds
s += ss;
}
for (i=h-1; i; i--) {
asm volatile(
__asm__ volatile(
"pxor %%mm7, %%mm7 \n\t"
"3: \n\t"
"movq (%%"REG_S"), %%mm0 \n\t"
@ -165,7 +165,7 @@ static void qpel_li_MMX(unsigned char *d, unsigned char *s, int w, int h, int ds
s += ss;
}
if (!up) fast_memcpy(d, s, w);
asm volatile("emms \n\t" : : : "memory");
__asm__ volatile("emms \n\t" : : : "memory");
}
static void qpel_4tap_MMX(unsigned char *d, unsigned char *s, int w, int h, int ds, int ss, int up)
@ -185,7 +185,7 @@ static void qpel_4tap_MMX(unsigned char *d, unsigned char *s, int w, int h, int
d[j] = (s[j+ssd] + 3*s[j])>>2;
d += ds; s += ss;
for (i=h-3; i; i--) {
asm volatile(
__asm__ volatile(
"pxor %%mm0, %%mm0 \n\t"
"movq (%%"REG_d"), %%mm4 \n\t"
"movq 8(%%"REG_d"), %%mm5 \n\t"
@ -245,7 +245,7 @@ static void qpel_4tap_MMX(unsigned char *d, unsigned char *s, int w, int h, int
d[j] = (s[j+ssd] + 3*s[j])>>2;
d += ds; s += ss;
if (!up) fast_memcpy(d, s, w);
asm volatile("emms \n\t" : : : "memory");
__asm__ volatile("emms \n\t" : : : "memory");
}
#endif

4
libmpcodecs/vf_unsharp.c
View File

@ -194,11 +194,11 @@ static int put_image( struct vf_instance_s* vf, mp_image_t *mpi, double pts) {
#ifdef HAVE_MMX
if(gCpuCaps.hasMMX)
asm volatile ("emms\n\t");
__asm__ volatile ("emms\n\t");
#endif
#ifdef HAVE_MMX2
if(gCpuCaps.hasMMX2)
asm volatile ("sfence\n\t");
__asm__ volatile ("sfence\n\t");
#endif
return vf_next_put_image( vf, dmpi, pts);

4
libmpcodecs/vf_uspp.c
View File

@ -292,10 +292,10 @@ static int put_image(struct vf_instance_s* vf, mp_image_t *mpi, double pts){
}
#ifdef HAVE_MMX
if(gCpuCaps.hasMMX) asm volatile ("emms\n\t");
if(gCpuCaps.hasMMX) __asm__ volatile ("emms\n\t");
#endif
#ifdef HAVE_MMX2
if(gCpuCaps.hasMMX2) asm volatile ("sfence\n\t");
if(gCpuCaps.hasMMX2) __asm__ volatile ("sfence\n\t");
#endif
return vf_next_put_image(vf,dmpi, pts);

6
libmpcodecs/vf_yadif.c
View File

@ -142,7 +142,7 @@ static void filter_line_mmx2(struct vf_priv_s *p, uint8_t *dst, uint8_t *prev, u
#define FILTER\
for(x=0; x<w; x+=4){\
asm volatile(\
__asm__ volatile(\
"pxor %%mm7, %%mm7 \n\t"\
LOAD4("(%[cur],%[mrefs])", %%mm0) /* c = cur[x-refs] */\
LOAD4("(%[cur],%[prefs])", %%mm1) /* e = cur[x+refs] */\
@ -259,7 +259,7 @@ static void filter_line_mmx2(struct vf_priv_s *p, uint8_t *dst, uint8_t *prev, u
[pb1] "m"(pb_1),\
[mode] "g"(mode)\
);\
asm volatile("movd %%mm1, %0" :"=m"(*dst));\
__asm__ volatile("movd %%mm1, %0" :"=m"(*dst));\
dst += 4;\
prev+= 4;\
cur += 4;\
@ -370,7 +370,7 @@ static void filter(struct vf_priv_s *p, uint8_t *dst[3], int dst_stride[3], int
}
}
#if defined(HAVE_MMX) && defined(NAMED_ASM_ARGS)
if(gCpuCaps.hasMMX2) asm volatile("emms \n\t" : : : "memory");
if(gCpuCaps.hasMMX2) __asm__ volatile("emms \n\t" : : : "memory");
#endif
}

2
libvo/aclib_template.c
View File

@ -256,7 +256,7 @@ static void * RENAME(fast_memcpy)(void * to, const void * from, size_t len)
// printf(" %d %d\n", (int)from&1023, (int)to&1023);
// Pure Assembly cuz gcc is a bit unpredictable ;)
if(i>=BLOCK_SIZE/64)
asm volatile(
__asm__ volatile(
"xor %%"REG_a", %%"REG_a" \n\t"
ASMALIGN(4)
"1: \n\t"

42
libvo/osd_template.c
View File

@ -33,7 +33,7 @@ static inline void RENAME(vo_draw_alpha_yv12)(int w,int h, unsigned char* src, u
w=w>>1;
#endif
#ifdef HAVE_MMX
asm volatile(
__asm__ volatile(
"pcmpeqb %%mm5, %%mm5\n\t" // F..F
"movq %%mm5, %%mm4\n\t"
"movq %%mm5, %%mm7\n\t"
@ -44,13 +44,13 @@ static inline void RENAME(vo_draw_alpha_yv12)(int w,int h, unsigned char* src, u
for(y=0;y<h;y++){
register int x;
#ifdef HAVE_MMX
asm volatile(
__asm__ volatile(
PREFETCHW" %0\n\t"
PREFETCH" %1\n\t"
PREFETCH" %2\n\t"
::"m"(*dstbase),"m"(*srca),"m"(*src):"memory");
for(x=0;x<w;x+=8){
asm volatile(
__asm__ volatile(
"movl %1, %%eax\n\t"
"orl 4%1, %%eax\n\t"
" jz 1f\n\t"
@ -92,7 +92,7 @@ static inline void RENAME(vo_draw_alpha_yv12)(int w,int h, unsigned char* src, u
dstbase+=dststride;
}
#ifdef HAVE_MMX
asm volatile(EMMS:::"memory");
__asm__ volatile(EMMS:::"memory");
#endif
return;
}
@ -103,7 +103,7 @@ static inline void RENAME(vo_draw_alpha_yuy2)(int w,int h, unsigned char* src, u
w=w>>1;
#endif
#ifdef HAVE_MMX
asm volatile(
__asm__ volatile(
"pxor %%mm7, %%mm7\n\t"
"pcmpeqb %%mm5, %%mm5\n\t" // F..F
"movq %%mm5, %%mm6\n\t"
@ -115,13 +115,13 @@ static inline void RENAME(vo_draw_alpha_yuy2)(int w,int h, unsigned char* src, u
for(y=0;y<h;y++){
register int x;
#ifdef HAVE_MMX
asm volatile(
__asm__ volatile(
PREFETCHW" %0\n\t"
PREFETCH" %1\n\t"
PREFETCH" %2\n\t"
::"m"(*dstbase),"m"(*srca),"m"(*src));
for(x=0;x<w;x+=4){
asm volatile(
__asm__ volatile(
"movl %1, %%eax\n\t"
"orl %%eax, %%eax\n\t"
" jz 1f\n\t"
@ -164,7 +164,7 @@ static inline void RENAME(vo_draw_alpha_yuy2)(int w,int h, unsigned char* src, u
dstbase+=dststride;
}
#ifdef HAVE_MMX
asm volatile(EMMS:::"memory");
__asm__ volatile(EMMS:::"memory");
#endif
return;
}
@ -196,7 +196,7 @@ static inline void RENAME(vo_draw_alpha_uyvy)(int w,int h, unsigned char* src, u
static inline void RENAME(vo_draw_alpha_rgb24)(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
int y;
#ifdef HAVE_MMX
asm volatile(
__asm__ volatile(
"pxor %%mm7, %%mm7\n\t"
"pcmpeqb %%mm6, %%mm6\n\t" // F..F
::);
@ -206,14 +206,14 @@ static inline void RENAME(vo_draw_alpha_rgb24)(int w,int h, unsigned char* src,
register int x;
#if defined(ARCH_X86) && (!defined(ARCH_X86_64) || defined(HAVE_MMX))
#ifdef HAVE_MMX
asm volatile(
__asm__ volatile(
PREFETCHW" %0\n\t"
PREFETCH" %1\n\t"
PREFETCH" %2\n\t"
::"m"(*dst),"m"(*srca),"m"(*src):"memory");
for(x=0;x<w;x+=2){
if(srca[x] || srca[x+1])
asm volatile(
__asm__ volatile(
PREFETCHW" 32%0\n\t"
PREFETCH" 32%1\n\t"
PREFETCH" 32%2\n\t"
@ -250,7 +250,7 @@ static inline void RENAME(vo_draw_alpha_rgb24)(int w,int h, unsigned char* src,
#else /* HAVE_MMX */
for(x=0;x<w;x++){
if(srca[x]){
asm volatile(
__asm__ volatile(
"movzbl (%0), %%ecx\n\t"
"movzbl 1(%0), %%eax\n\t"
@ -296,7 +296,7 @@ static inline void RENAME(vo_draw_alpha_rgb24)(int w,int h, unsigned char* src,
dstbase+=dststride;
}
#ifdef HAVE_MMX
asm volatile(EMMS:::"memory");
__asm__ volatile(EMMS:::"memory");
#endif
return;
}
@ -308,12 +308,12 @@ static inline void RENAME(vo_draw_alpha_rgb32)(int w,int h, unsigned char* src,
#endif
#ifdef HAVE_MMX
#ifdef HAVE_3DNOW
asm volatile(
__asm__ volatile(
"pxor %%mm7, %%mm7\n\t"
"pcmpeqb %%mm6, %%mm6\n\t" // F..F
::);
#else /* HAVE_3DNOW */
asm volatile(
__asm__ volatile(
"pxor %%mm7, %%mm7\n\t"
"pcmpeqb %%mm5, %%mm5\n\t" // F..F
"movq %%mm5, %%mm4\n\t"
@ -327,14 +327,14 @@ static inline void RENAME(vo_draw_alpha_rgb32)(int w,int h, unsigned char* src,
#if defined(ARCH_X86) && (!defined(ARCH_X86_64) || defined(HAVE_MMX))
#ifdef HAVE_MMX
#ifdef HAVE_3DNOW
asm volatile(
__asm__ volatile(
PREFETCHW" %0\n\t"
PREFETCH" %1\n\t"
PREFETCH" %2\n\t"
::"m"(*dstbase),"m"(*srca),"m"(*src):"memory");
for(x=0;x<w;x+=2){
if(srca[x] || srca[x+1])
asm volatile(
__asm__ volatile(
PREFETCHW" 32%0\n\t"
PREFETCH" 32%1\n\t"
PREFETCH" 32%2\n\t"
@ -362,13 +362,13 @@ static inline void RENAME(vo_draw_alpha_rgb32)(int w,int h, unsigned char* src,
:: "m" (dstbase[4*x]), "m" (srca[x]), "m" (src[x]));
}
#else //this is faster for intels crap
asm volatile(
__asm__ volatile(
PREFETCHW" %0\n\t"
PREFETCH" %1\n\t"
PREFETCH" %2\n\t"
::"m"(*dstbase),"m"(*srca),"m"(*src):"memory");
for(x=0;x<w;x+=4){
asm volatile(
__asm__ volatile(
"movl %1, %%eax\n\t"
"orl %%eax, %%eax\n\t"
" jz 1f\n\t"
@ -417,7 +417,7 @@ static inline void RENAME(vo_draw_alpha_rgb32)(int w,int h, unsigned char* src,
#else /* HAVE_MMX */
for(x=0;x<w;x++){
if(srca[x]){
asm volatile(
__asm__ volatile(
"movzbl (%0), %%ecx\n\t"
"movzbl 1(%0), %%eax\n\t"
"movzbl 2(%0), %%edx\n\t"
@ -461,7 +461,7 @@ static inline void RENAME(vo_draw_alpha_rgb32)(int w,int h, unsigned char* src,
dstbase+=dststride;
}
#ifdef HAVE_MMX
asm volatile(EMMS:::"memory");
__asm__ volatile(EMMS:::"memory");
#endif
return;
}

2
loader/win32.c
View File

@ -4569,7 +4569,7 @@ static INT WINAPI expMessageBoxA(HWND hWnd, LPCSTR text, LPCSTR title, UINT type
void exp_EH_prolog(void *dest);
//! just a dummy function that acts a container for the asm section
void exp_EH_prolog_dummy(void) {
asm volatile (
__asm__ volatile (
// take care, this "function" may not change flags or
// registers besides eax (which is also why we can't use
// exp_EH_prolog_dummy directly)

2
loader/wine/winnt.h
View File

@ -693,7 +693,7 @@ typedef HANDLE *PHANDLE;
#else /* __GNUC__ */
# define ASM_GLOBAL_FUNC(name,code) \
void asm_dummy_##name(void) { \
asm( ".align 4\n\t" \
__asm__( ".align 4\n\t" \
".globl " ASM_NAME(#name) "\n\t" \
".type " ASM_NAME(#name) ",@function\n" \
ASM_NAME(#name) ":\n\t" \

2
mp3lib/dct64_3dnow.c
View File

@ -18,7 +18,7 @@ static float attribute_used plus_1f = 1.0;
void dct64_MMX_3dnow(short *a,short *b,real *c)
{
char tmp[256];
__asm __volatile(
__asm__ __volatile(
" movl %2,%%eax\n\t"
" leal 128+%3,%%edx\n\t"

2
mp3lib/dct64_k7.c
View File

@ -18,7 +18,7 @@ static float attribute_used plus_1f = 1.0;
void dct64_MMX_3dnowex(short *a,short *b,real *c)
{
char tmp[256];
__asm __volatile(
__asm__ __volatile(
" movl %2,%%eax\n\t"
" leal 128+%3,%%edx\n\t"

2
mp3lib/dct64_mmx.c
View File

@ -9,7 +9,7 @@
void dct64_MMX(short *a,short *b,real *c)
{
char tmp[256];
__asm __volatile(
__asm__ __volatile(
" movl %2,%%eax\n\t"
/* Phase 1*/
" flds (%%eax)\n\t"

20
mp3lib/dct64_sse.c
View File

@ -32,7 +32,7 @@ void dct64_sse(short *out0,short *out1,real *c)
for (i = 0; i < 0x20 / 2; i += 4)
{
asm(
__asm__(
"movaps %2, %%xmm3\n\t"
"shufps $27, %%xmm3, %%xmm3\n\t"
"movaps %3, %%xmm1\n\t"
@ -57,7 +57,7 @@ void dct64_sse(short *out0,short *out1,real *c)
for (i = 0; i < 0x20; i += 0x10)
{
asm(
__asm__(
"movaps %4, %%xmm1\n\t"
"movaps %5, %%xmm3\n\t"
"movaps %6, %%xmm4\n\t"
@ -86,7 +86,7 @@ void dct64_sse(short *out0,short *out1,real *c)
{
real *costab = costab_mmx + 16;
asm(
__asm__(
"movaps %4, %%xmm0\n\t"
"movaps %5, %%xmm1\n\t"
"movaps %8, %%xmm4\n\t"
@ -116,7 +116,7 @@ void dct64_sse(short *out0,short *out1,real *c)
real *costab = costab_mmx + 24;
int i;
asm(
__asm__(
"movaps %0, %%xmm0\n\t"
"shufps $27, %%xmm0, %%xmm0\n\t"
"movaps %1, %%xmm5\n\t"
@ -127,7 +127,7 @@ void dct64_sse(short *out0,short *out1,real *c)
for (i = 0; i < 0x20; i += 8)
{
asm(
__asm__(
"movaps %2, %%xmm2\n\t"
"movaps %3, %%xmm3\n\t"
"movaps %%xmm2, %%xmm4\n\t"
@ -150,7 +150,7 @@ void dct64_sse(short *out0,short *out1,real *c)
{
int i;
asm(
__asm__(
"movss %0, %%xmm1\n\t"
"movss %1, %%xmm0\n\t"
"movaps %%xmm1, %%xmm3\n\t"
@ -166,7 +166,7 @@ void dct64_sse(short *out0,short *out1,real *c)
for (i = 0; i < 0x20; i += 8)
{
asm(
__asm__(
"movaps %2, %%xmm3\n\t"
"movaps %%xmm3, %%xmm4\n\t"
"shufps $20, %%xmm4, %%xmm4\n\t"
@ -190,7 +190,7 @@ void dct64_sse(short *out0,short *out1,real *c)
{
int i;
asm(
__asm__(
"movss %0, %%xmm0\n\t"
"movaps %%xmm1, %%xmm2\n\t"
"movaps %%xmm0, %%xmm7\n\t"
@ -204,7 +204,7 @@ void dct64_sse(short *out0,short *out1,real *c)
for (i = 0x8; i < 0x20; i += 8)
{
asm volatile (
__asm__ volatile (
"movaps %2, %%xmm1\n\t"
"movaps %%xmm1, %%xmm3\n\t"
"shufps $224, %%xmm3, %%xmm3\n\t"
@ -285,7 +285,7 @@ void dct64_sse(short *out0,short *out1,real *c)
To do saturation efficiently in x86 we can use fist(t)(p),
pf2iw, or packssdw. We use fist(p) here.
*/
asm(
__asm__(
"flds %0\n\t"
"flds (%2)\n\t"
"fadds 4(%2)\n\t"

2
mp3lib/decode_i586.c
View File

@ -41,7 +41,7 @@ int synth_1to1_pent(real *bandPtr, int channel, short *samples)
{
real tmp[3];
register int retval;
__asm __volatile(
__asm__ __volatile(
" movl %%ebp,"MANGLE(saved_ebp)"\n\t"
" movl %1,%%eax\n\t"/*bandPtr*/
" movl %3,%%esi\n\t"

2
mp3lib/decode_mmx.c
View File

@ -217,7 +217,7 @@ int synth_1to1_MMX(real *bandPtr, int channel, short *samples)
dct64_MMX_func(a, b, bandPtr);
window = mp3lib_decwins + 16 - bo1;
//printf("DEBUG: channel %d, bo %d, off %d\n", channel, bo, 16 - bo1);
__asm __volatile(
__asm__ __volatile(
ASMALIGN(4)
".L03:\n\t"
"movq (%1),%%mm0\n\t"

9
vidix/AsmMacros.h
View File

@ -99,11 +99,6 @@ void outl(U16_t, U32_t);
#else /* not _MINIX and _ACK */
# if defined(__STDC__) && (__STDC__ == 1)
# ifndef NCR
# define asm __asm
# endif
# endif
# ifdef SVR4
# include <sys/types.h>
# ifndef __USLC__
@ -115,8 +110,8 @@ void outl(U16_t, U32_t);
#else
# include "../common/scoasm.h"
#endif
#define intr_disable() asm("cli")
#define intr_enable() asm("sti")
#define intr_disable() __asm__("cli")
#define intr_enable() __asm__("sti")
#endif /* _MINIX and _ACK */
#endif /* __GNUC__ */

2
vidix/sysdep/pci_os2.c
View File

@ -62,7 +62,7 @@ static __inline__ int enable_os_io(void)
/* Calling callgate with function 13 sets IOPL for the program */
asm volatile ("movl $13,%%ebx;.byte 0xff,0x1d;.long _callgate"
__asm__ volatile ("movl $13,%%ebx;.byte 0xff,0x1d;.long _callgate"
: /*no outputs */
: /*no inputs */
: "eax","ebx","ecx","edx","cc");