summaryrefslogtreecommitdiff
path: root/libavfilter/x86
diff options
context:
space:
mode:
authorMichael Niedermayer <michaelni@gmx.at>2012-07-02 21:19:41 +0200
committerMichael Niedermayer <michaelni@gmx.at>2012-07-02 21:19:41 +0200
commitb286383bd2b13fdf59aecdb23d8514323460483b (patch)
tree4f8ee9745d719e2918843d0d8ebcf053ac220ef4 /libavfilter/x86
parent7c84e7d33762a4bccc0002476a3b20e0b8f26fcc (diff)
parent5e745cefc0f89cf698c4cf0104182472fe0f603e (diff)
Merge remote-tracking branch 'qatar/master'
* qatar/master: asfdec: read attached pictures. apetag: reindent apetag: export attached covers as video streams. apetag: fix the amount of data read from binary tags. apetag: make sure avio_get_str() doesn't read more than it should. mov: read itunes cover art. snow: remove VLA in mc_block() intfloat: Don't use designated initializers in the public headers snow: remove a VLA. doc: Remind devs to check return values, especially for malloc() et al MS ATC Screen (aka MSS3) decoder vf_yadif: move x86 init code to x86/yadif.c vf_gradfun: move x86 init code to x86/gradfun.c roqvideo: Remove a totally unused dspcontext smacker: remove some unused code dsicin: remove dead assignment aacdec: remove dead assignment rl2: remove dead assignment proresenc: make a variable local to the loop where it is used alsdec: remove dead assignments Conflicts: Changelog doc/developer.texi libavcodec/allcodecs.c libavcodec/avcodec.h libavcodec/version.h libavfilter/gradfun.h libavfilter/x86/gradfun.c Merged-by: Michael Niedermayer <michaelni@gmx.at>
Diffstat (limited to 'libavfilter/x86')
-rw-r--r--libavfilter/x86/gradfun.c29
-rw-r--r--libavfilter/x86/yadif.c13
-rw-r--r--libavfilter/x86/yadif_template.c6
3 files changed, 37 insertions, 11 deletions
diff --git a/libavfilter/x86/gradfun.c b/libavfilter/x86/gradfun.c
index e892117d67..4a0c01209c 100644
--- a/libavfilter/x86/gradfun.c
+++ b/libavfilter/x86/gradfun.c
@@ -18,6 +18,7 @@
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
+#include "libavutil/attributes.h"
#include "libavutil/cpu.h"
#include "libavutil/x86_cpu.h"
#include "libavfilter/gradfun.h"
@@ -25,9 +26,9 @@
DECLARE_ALIGNED(16, static const uint16_t, pw_7f)[8] = {0x7F,0x7F,0x7F,0x7F,0x7F,0x7F,0x7F,0x7F};
DECLARE_ALIGNED(16, static const uint16_t, pw_ff)[8] = {0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF};
-void ff_gradfun_filter_line_mmx2(uint8_t *dst, const uint8_t *src, const uint16_t *dc, int width, int thresh, const uint16_t *dithers)
+#if HAVE_MMX2
+static void gradfun_filter_line_mmx2(uint8_t *dst, const uint8_t *src, const uint16_t *dc, int width, int thresh, const uint16_t *dithers)
{
-#if HAVE_MMX
intptr_t x;
if (width & 3) {
x = width & ~3;
@@ -70,12 +71,12 @@ void ff_gradfun_filter_line_mmx2(uint8_t *dst, const uint8_t *src, const uint16_
"rm"(thresh), "m"(*dithers), "m"(*pw_7f)
:"memory"
);
-#endif
}
+#endif
-void ff_gradfun_filter_line_ssse3(uint8_t *dst, const uint8_t *src, const uint16_t *dc, int width, int thresh, const uint16_t *dithers)
-{
#if HAVE_SSSE3
+static void gradfun_filter_line_ssse3(uint8_t *dst, const uint8_t *src, const uint16_t *dc, int width, int thresh, const uint16_t *dithers)
+{
intptr_t x;
if (width & 7) {
// could be 10% faster if I somehow eliminated this
@@ -117,12 +118,12 @@ void ff_gradfun_filter_line_ssse3(uint8_t *dst, const uint8_t *src, const uint16
"rm"(thresh), "m"(*dithers), "m"(*pw_7f)
:"memory"
);
-#endif // HAVE_SSSE3
}
+#endif // HAVE_SSSE3
-void ff_gradfun_blur_line_sse2(uint16_t *dc, uint16_t *buf, const uint16_t *buf1, const uint8_t *src, int src_linesize, int width)
-{
#if HAVE_SSE
+static void gradfun_blur_line_sse2(uint16_t *dc, uint16_t *buf, const uint16_t *buf1, const uint8_t *src, int src_linesize, int width)
+{
#define BLURV(load)\
intptr_t x = -2*width;\
__asm__ volatile(\
@@ -160,5 +161,17 @@ void ff_gradfun_blur_line_sse2(uint16_t *dc, uint16_t *buf, const uint16_t *buf1
} else {
BLURV("movdqa");
}
+}
#endif // HAVE_SSE
+
+av_cold void ff_gradfun_init_x86(GradFunContext *gf)
+{
+ int cpu_flags = av_get_cpu_flags();
+
+ if (HAVE_MMX2 && cpu_flags & AV_CPU_FLAG_MMX2)
+ gf->filter_line = gradfun_filter_line_mmx2;
+ if (HAVE_SSSE3 && cpu_flags & AV_CPU_FLAG_SSSE3)
+ gf->filter_line = gradfun_filter_line_ssse3;
+ if (HAVE_SSE && cpu_flags & AV_CPU_FLAG_SSE2)
+ gf->blur_line = gradfun_blur_line_sse2;
}
diff --git a/libavfilter/x86/yadif.c b/libavfilter/x86/yadif.c
index 2b55c00763..d98238809a 100644
--- a/libavfilter/x86/yadif.c
+++ b/libavfilter/x86/yadif.c
@@ -18,6 +18,7 @@
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
*/
+#include "libavutil/attributes.h"
#include "libavutil/cpu.h"
#include "libavutil/x86_cpu.h"
#include "libavcodec/x86/dsputil_mmx.h"
@@ -47,3 +48,15 @@ DECLARE_ASM_CONST(16, const xmm_reg, pw_1) = {0x0001000100010001ULL, 0x000100010
#define RENAME(a) a ## _mmx
#include "yadif_template.c"
#endif
+
+av_cold void ff_yadif_init_x86(YADIFContext *yadif)
+{
+ int cpu_flags = av_get_cpu_flags();
+
+ if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX)
+ yadif->filter_line = yadif_filter_line_mmx;
+ if (HAVE_SSE && cpu_flags & AV_CPU_FLAG_SSE2)
+ yadif->filter_line = yadif_filter_line_sse2;
+ if (HAVE_SSSE3 && cpu_flags & AV_CPU_FLAG_SSSE3)
+ yadif->filter_line = yadif_filter_line_ssse3;
+}
diff --git a/libavfilter/x86/yadif_template.c b/libavfilter/x86/yadif_template.c
index 71cf9d1358..06c8797e28 100644
--- a/libavfilter/x86/yadif_template.c
+++ b/libavfilter/x86/yadif_template.c
@@ -103,9 +103,9 @@
"por "MM"5, "MM"3 \n\t"\
MOVQ" "MM"3, "MM"1 \n\t"
-void RENAME(ff_yadif_filter_line)(uint8_t *dst,
- uint8_t *prev, uint8_t *cur, uint8_t *next,
- int w, int prefs, int mrefs, int parity, int mode)
+static void RENAME(yadif_filter_line)(uint8_t *dst, uint8_t *prev, uint8_t *cur,
+ uint8_t *next, int w, int prefs,
+ int mrefs, int parity, int mode)
{
uint8_t tmp[5*16];
uint8_t *tmpA= (uint8_t*)(((uint64_t)(tmp+15)) & ~15);