x86: Replace checks for CPU extensions and flags by convenience macros

This separates code relying on inline from that relying on external
assembly and fixes instances where the coalesced check was incorrect.
This commit is contained in:
Diego Biurrun
2012-08-29 19:01:05 +02:00
parent 6a0200f24d
commit e0c6cce447
21 changed files with 133 additions and 156 deletions

View File

@ -46,6 +46,7 @@
#include "libavutil/opt.h"
#include "libavutil/pixdesc.h"
#include "libavutil/x86/asm.h"
#include "libavutil/x86/cpu.h"
#include "rgb2rgb.h"
#include "swscale.h"
#include "swscale_internal.h"
@ -473,7 +474,7 @@ static int initFilter(int16_t **outFilter, int32_t **filterPos,
filterAlign = 1;
}
if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) {
if (INLINE_MMX(cpu_flags)) {
// special case for unscaled vertical filtering
if (minFilterSize == 1 && filterAlign == 2)
filterAlign = 1;
@ -973,8 +974,7 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
FF_ALLOC_OR_GOTO(c, c->formatConvBuffer,
(FFALIGN(srcW, 16) * 2 * FFALIGN(c->srcBpc, 8) >> 3) + 16,
fail);
if (HAVE_MMXEXT && HAVE_INLINE_ASM && cpu_flags & AV_CPU_FLAG_MMXEXT &&
c->srcBpc == 8 && c->dstBpc <= 10) {
if (INLINE_MMXEXT(cpu_flags) && c->srcBpc == 8 && c->dstBpc <= 10) {
c->canMMX2BeUsed = (dstW >= srcW && (dstW & 31) == 0 &&
(srcW & 15) == 0) ? 1 : 0;
if (!c->canMMX2BeUsed && dstW >= srcW && (srcW & 15) == 0
@ -1004,7 +1004,7 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
c->chrXInc += 20;
}
// we don't use the x86 asm scaler if MMX is available
else if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) {
else if (INLINE_MMX(cpu_flags)) {
c->lumXInc = ((int64_t)(srcW - 2) << 16) / (dstW - 2) - 20;
c->chrXInc = ((int64_t)(c->chrSrcW - 2) << 16) / (c->chrDstW - 2) - 20;
}
@ -1050,8 +1050,7 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
} else
#endif /* HAVE_MMXEXT_INLINE */
{
const int filterAlign =
(HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) ? 4 :
const int filterAlign = INLINE_MMX(cpu_flags) ? 4 :
(HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) ? 8 :
1;
@ -1074,8 +1073,7 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
/* precalculate vertical scaler filter coefficients */
{
const int filterAlign =
(HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) ? 2 :
const int filterAlign = INLINE_MMX(cpu_flags) ? 2 :
(HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) ? 8 :
1;
@ -1208,11 +1206,11 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
#endif
sws_format_name(dstFormat));
if (HAVE_MMXEXT && cpu_flags & AV_CPU_FLAG_MMXEXT)
if (INLINE_MMXEXT(cpu_flags))
av_log(c, AV_LOG_INFO, "using MMX2\n");
else if (HAVE_AMD3DNOW && cpu_flags & AV_CPU_FLAG_3DNOW)
else if (INLINE_AMD3DNOW(cpu_flags))
av_log(c, AV_LOG_INFO, "using 3DNOW\n");
else if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX)
else if (INLINE_MMX(cpu_flags))
av_log(c, AV_LOG_INFO, "using MMX\n");
else if (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC)
av_log(c, AV_LOG_INFO, "using AltiVec\n");