aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorAnton Khirnov <anton@khirnov.net>2013-12-20 13:00:22 +0100
committerDiego Biurrun <diego@biurrun.de>2013-12-20 17:16:11 +0100
commitdfc50ac85e9d68a771b556297b7c411650206f3b (patch)
treeea76c18b1527330eb5fd6d74b5d562b400f4eb6b
parentb254490bdabb21bd517c05b1a68717f9952ac8c4 (diff)
downloadffmpeg-dfc50ac85e9d68a771b556297b7c411650206f3b.tar.gz
x86: mpegvideo: move denoise_dct asm to mpegvideoenc
This function is encoding-only. Signed-off-by: Diego Biurrun <diego@biurrun.de>
-rw-r--r--libavcodec/x86/mpegvideo.c114
-rw-r--r--libavcodec/x86/mpegvideoenc.c120
2 files changed, 118 insertions, 116 deletions
diff --git a/libavcodec/x86/mpegvideo.c b/libavcodec/x86/mpegvideo.c
index 07fd1e538c..25b44e6768 100644
--- a/libavcodec/x86/mpegvideo.c
+++ b/libavcodec/x86/mpegvideo.c
@@ -443,116 +443,6 @@ __asm__ volatile(
);
}
-static void denoise_dct_mmx(MpegEncContext *s, int16_t *block){
- const int intra= s->mb_intra;
- int *sum= s->dct_error_sum[intra];
- uint16_t *offset= s->dct_offset[intra];
-
- s->dct_count[intra]++;
-
- __asm__ volatile(
- "pxor %%mm7, %%mm7 \n\t"
- "1: \n\t"
- "pxor %%mm0, %%mm0 \n\t"
- "pxor %%mm1, %%mm1 \n\t"
- "movq (%0), %%mm2 \n\t"
- "movq 8(%0), %%mm3 \n\t"
- "pcmpgtw %%mm2, %%mm0 \n\t"
- "pcmpgtw %%mm3, %%mm1 \n\t"
- "pxor %%mm0, %%mm2 \n\t"
- "pxor %%mm1, %%mm3 \n\t"
- "psubw %%mm0, %%mm2 \n\t"
- "psubw %%mm1, %%mm3 \n\t"
- "movq %%mm2, %%mm4 \n\t"
- "movq %%mm3, %%mm5 \n\t"
- "psubusw (%2), %%mm2 \n\t"
- "psubusw 8(%2), %%mm3 \n\t"
- "pxor %%mm0, %%mm2 \n\t"
- "pxor %%mm1, %%mm3 \n\t"
- "psubw %%mm0, %%mm2 \n\t"
- "psubw %%mm1, %%mm3 \n\t"
- "movq %%mm2, (%0) \n\t"
- "movq %%mm3, 8(%0) \n\t"
- "movq %%mm4, %%mm2 \n\t"
- "movq %%mm5, %%mm3 \n\t"
- "punpcklwd %%mm7, %%mm4 \n\t"
- "punpckhwd %%mm7, %%mm2 \n\t"
- "punpcklwd %%mm7, %%mm5 \n\t"
- "punpckhwd %%mm7, %%mm3 \n\t"
- "paddd (%1), %%mm4 \n\t"
- "paddd 8(%1), %%mm2 \n\t"
- "paddd 16(%1), %%mm5 \n\t"
- "paddd 24(%1), %%mm3 \n\t"
- "movq %%mm4, (%1) \n\t"
- "movq %%mm2, 8(%1) \n\t"
- "movq %%mm5, 16(%1) \n\t"
- "movq %%mm3, 24(%1) \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)
- );
-}
-
-static void denoise_dct_sse2(MpegEncContext *s, int16_t *block){
- const int intra= s->mb_intra;
- int *sum= s->dct_error_sum[intra];
- uint16_t *offset= s->dct_offset[intra];
-
- s->dct_count[intra]++;
-
- __asm__ volatile(
- "pxor %%xmm7, %%xmm7 \n\t"
- "1: \n\t"
- "pxor %%xmm0, %%xmm0 \n\t"
- "pxor %%xmm1, %%xmm1 \n\t"
- "movdqa (%0), %%xmm2 \n\t"
- "movdqa 16(%0), %%xmm3 \n\t"
- "pcmpgtw %%xmm2, %%xmm0 \n\t"
- "pcmpgtw %%xmm3, %%xmm1 \n\t"
- "pxor %%xmm0, %%xmm2 \n\t"
- "pxor %%xmm1, %%xmm3 \n\t"
- "psubw %%xmm0, %%xmm2 \n\t"
- "psubw %%xmm1, %%xmm3 \n\t"
- "movdqa %%xmm2, %%xmm4 \n\t"
- "movdqa %%xmm3, %%xmm5 \n\t"
- "psubusw (%2), %%xmm2 \n\t"
- "psubusw 16(%2), %%xmm3 \n\t"
- "pxor %%xmm0, %%xmm2 \n\t"
- "pxor %%xmm1, %%xmm3 \n\t"
- "psubw %%xmm0, %%xmm2 \n\t"
- "psubw %%xmm1, %%xmm3 \n\t"
- "movdqa %%xmm2, (%0) \n\t"
- "movdqa %%xmm3, 16(%0) \n\t"
- "movdqa %%xmm4, %%xmm6 \n\t"
- "movdqa %%xmm5, %%xmm0 \n\t"
- "punpcklwd %%xmm7, %%xmm4 \n\t"
- "punpckhwd %%xmm7, %%xmm6 \n\t"
- "punpcklwd %%xmm7, %%xmm5 \n\t"
- "punpckhwd %%xmm7, %%xmm0 \n\t"
- "paddd (%1), %%xmm4 \n\t"
- "paddd 16(%1), %%xmm6 \n\t"
- "paddd 32(%1), %%xmm5 \n\t"
- "paddd 48(%1), %%xmm0 \n\t"
- "movdqa %%xmm4, (%1) \n\t"
- "movdqa %%xmm6, 16(%1) \n\t"
- "movdqa %%xmm5, 32(%1) \n\t"
- "movdqa %%xmm0, 48(%1) \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)
- XMM_CLOBBERS_ONLY("%xmm0", "%xmm1", "%xmm2", "%xmm3",
- "%xmm4", "%xmm5", "%xmm6", "%xmm7")
- );
-}
-
#endif /* HAVE_INLINE_ASM */
av_cold void ff_MPV_common_init_x86(MpegEncContext *s)
@@ -568,10 +458,6 @@ av_cold void ff_MPV_common_init_x86(MpegEncContext *s)
if(!(s->flags & CODEC_FLAG_BITEXACT))
s->dct_unquantize_mpeg2_intra = dct_unquantize_mpeg2_intra_mmx;
s->dct_unquantize_mpeg2_inter = dct_unquantize_mpeg2_inter_mmx;
- s->denoise_dct = denoise_dct_mmx;
- }
- if (INLINE_SSE2(cpu_flags)) {
- s->denoise_dct = denoise_dct_sse2;
}
#endif /* HAVE_INLINE_ASM */
}
diff --git a/libavcodec/x86/mpegvideoenc.c b/libavcodec/x86/mpegvideoenc.c
index 3167ea602b..19ab83a7fa 100644
--- a/libavcodec/x86/mpegvideoenc.c
+++ b/libavcodec/x86/mpegvideoenc.c
@@ -82,6 +82,118 @@ DECLARE_ALIGNED(16, static uint16_t, inv_zigzag_direct16)[64];
#include "mpegvideoenc_template.c"
#endif /* HAVE_SSSE3_INLINE */
+#if HAVE_INLINE_ASM
+static void denoise_dct_mmx(MpegEncContext *s, int16_t *block){
+ const int intra= s->mb_intra;
+ int *sum= s->dct_error_sum[intra];
+ uint16_t *offset= s->dct_offset[intra];
+
+ s->dct_count[intra]++;
+
+ __asm__ volatile(
+ "pxor %%mm7, %%mm7 \n\t"
+ "1: \n\t"
+ "pxor %%mm0, %%mm0 \n\t"
+ "pxor %%mm1, %%mm1 \n\t"
+ "movq (%0), %%mm2 \n\t"
+ "movq 8(%0), %%mm3 \n\t"
+ "pcmpgtw %%mm2, %%mm0 \n\t"
+ "pcmpgtw %%mm3, %%mm1 \n\t"
+ "pxor %%mm0, %%mm2 \n\t"
+ "pxor %%mm1, %%mm3 \n\t"
+ "psubw %%mm0, %%mm2 \n\t"
+ "psubw %%mm1, %%mm3 \n\t"
+ "movq %%mm2, %%mm4 \n\t"
+ "movq %%mm3, %%mm5 \n\t"
+ "psubusw (%2), %%mm2 \n\t"
+ "psubusw 8(%2), %%mm3 \n\t"
+ "pxor %%mm0, %%mm2 \n\t"
+ "pxor %%mm1, %%mm3 \n\t"
+ "psubw %%mm0, %%mm2 \n\t"
+ "psubw %%mm1, %%mm3 \n\t"
+ "movq %%mm2, (%0) \n\t"
+ "movq %%mm3, 8(%0) \n\t"
+ "movq %%mm4, %%mm2 \n\t"
+ "movq %%mm5, %%mm3 \n\t"
+ "punpcklwd %%mm7, %%mm4 \n\t"
+ "punpckhwd %%mm7, %%mm2 \n\t"
+ "punpcklwd %%mm7, %%mm5 \n\t"
+ "punpckhwd %%mm7, %%mm3 \n\t"
+ "paddd (%1), %%mm4 \n\t"
+ "paddd 8(%1), %%mm2 \n\t"
+ "paddd 16(%1), %%mm5 \n\t"
+ "paddd 24(%1), %%mm3 \n\t"
+ "movq %%mm4, (%1) \n\t"
+ "movq %%mm2, 8(%1) \n\t"
+ "movq %%mm5, 16(%1) \n\t"
+ "movq %%mm3, 24(%1) \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)
+ );
+}
+
+static void denoise_dct_sse2(MpegEncContext *s, int16_t *block){
+ const int intra= s->mb_intra;
+ int *sum= s->dct_error_sum[intra];
+ uint16_t *offset= s->dct_offset[intra];
+
+ s->dct_count[intra]++;
+
+ __asm__ volatile(
+ "pxor %%xmm7, %%xmm7 \n\t"
+ "1: \n\t"
+ "pxor %%xmm0, %%xmm0 \n\t"
+ "pxor %%xmm1, %%xmm1 \n\t"
+ "movdqa (%0), %%xmm2 \n\t"
+ "movdqa 16(%0), %%xmm3 \n\t"
+ "pcmpgtw %%xmm2, %%xmm0 \n\t"
+ "pcmpgtw %%xmm3, %%xmm1 \n\t"
+ "pxor %%xmm0, %%xmm2 \n\t"
+ "pxor %%xmm1, %%xmm3 \n\t"
+ "psubw %%xmm0, %%xmm2 \n\t"
+ "psubw %%xmm1, %%xmm3 \n\t"
+ "movdqa %%xmm2, %%xmm4 \n\t"
+ "movdqa %%xmm3, %%xmm5 \n\t"
+ "psubusw (%2), %%xmm2 \n\t"
+ "psubusw 16(%2), %%xmm3 \n\t"
+ "pxor %%xmm0, %%xmm2 \n\t"
+ "pxor %%xmm1, %%xmm3 \n\t"
+ "psubw %%xmm0, %%xmm2 \n\t"
+ "psubw %%xmm1, %%xmm3 \n\t"
+ "movdqa %%xmm2, (%0) \n\t"
+ "movdqa %%xmm3, 16(%0) \n\t"
+ "movdqa %%xmm4, %%xmm6 \n\t"
+ "movdqa %%xmm5, %%xmm0 \n\t"
+ "punpcklwd %%xmm7, %%xmm4 \n\t"
+ "punpckhwd %%xmm7, %%xmm6 \n\t"
+ "punpcklwd %%xmm7, %%xmm5 \n\t"
+ "punpckhwd %%xmm7, %%xmm0 \n\t"
+ "paddd (%1), %%xmm4 \n\t"
+ "paddd 16(%1), %%xmm6 \n\t"
+ "paddd 32(%1), %%xmm5 \n\t"
+ "paddd 48(%1), %%xmm0 \n\t"
+ "movdqa %%xmm4, (%1) \n\t"
+ "movdqa %%xmm6, 16(%1) \n\t"
+ "movdqa %%xmm5, 32(%1) \n\t"
+ "movdqa %%xmm0, 48(%1) \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)
+ XMM_CLOBBERS_ONLY("%xmm0", "%xmm1", "%xmm2", "%xmm3",
+ "%xmm4", "%xmm5", "%xmm6", "%xmm7")
+ );
+}
+#endif /* HAVE_INLINE_ASM */
+
av_cold void ff_MPV_encode_init_x86(MpegEncContext *s)
{
const int dct_algo = s->avctx->dct_algo;
@@ -93,16 +205,20 @@ av_cold void ff_MPV_encode_init_x86(MpegEncContext *s)
if (dct_algo == FF_DCT_AUTO || dct_algo == FF_DCT_MMX) {
#if HAVE_MMX_INLINE
int cpu_flags = av_get_cpu_flags();
- if (INLINE_MMX(cpu_flags))
+ if (INLINE_MMX(cpu_flags)) {
s->dct_quantize = dct_quantize_MMX;
+ s->denoise_dct = denoise_dct_mmx;
+ }
#endif
#if HAVE_MMXEXT_INLINE
if (INLINE_MMXEXT(cpu_flags))
s->dct_quantize = dct_quantize_MMXEXT;
#endif
#if HAVE_SSE2_INLINE
- if (INLINE_SSE2(cpu_flags))
+ if (INLINE_SSE2(cpu_flags)) {
s->dct_quantize = dct_quantize_SSE2;
+ s->denoise_dct = denoise_dct_sse2;
+ }
#endif
#if HAVE_SSSE3_INLINE
if (INLINE_SSSE3(cpu_flags))