[SCM] libav/experimental: x86: dsputil: Coalesce all init files
siretart at users.alioth.debian.org
siretart at users.alioth.debian.org
Sun Aug 10 16:03:52 UTC 2014
The following commit has been merged in the experimental branch:
commit c23ce454b3e33634a188d6facfd2b7182af5af93
Author: Diego Biurrun <diego at biurrun.de>
Date: Sat Jan 4 15:23:15 2014 +0100
x86: dsputil: Coalesce all init files
This makes the init files match the structure of the dsputil split.
diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile
index 18bb012..c8804bb 100644
--- a/libavcodec/x86/Makefile
+++ b/libavcodec/x86/Makefile
@@ -6,8 +6,7 @@ OBJS-$(CONFIG_AUDIODSP) += x86/audiodsp_init.o
OBJS-$(CONFIG_BLOCKDSP) += x86/blockdsp.o
OBJS-$(CONFIG_BSWAPDSP) += x86/bswapdsp_init.o
OBJS-$(CONFIG_DCT) += x86/dct_init.o
-OBJS-$(CONFIG_DSPUTIL) += x86/dsputil_init.o \
- x86/motion_est.o
+OBJS-$(CONFIG_DSPUTIL) += x86/dsputil_init.o
OBJS-$(CONFIG_FDCTDSP) += x86/fdctdsp_init.o
OBJS-$(CONFIG_FFT) += x86/fft_init.o
OBJS-$(CONFIG_H263DSP) += x86/h263dsp_init.o
diff --git a/libavcodec/x86/dsputil_init.c b/libavcodec/x86/dsputil_init.c
index d867c7c..1303757 100644
--- a/libavcodec/x86/dsputil_init.c
+++ b/libavcodec/x86/dsputil_init.c
@@ -28,7 +28,6 @@
#include "libavutil/x86/cpu.h"
#include "libavcodec/dsputil.h"
#include "libavcodec/mpegvideo.h"
-#include "dsputil_x86.h"
#if HAVE_INLINE_ASM
@@ -797,6 +796,419 @@ DCT_SAD_FUNC(ssse3)
#undef HSUM
#undef DCT_SAD
+
+DECLARE_ASM_CONST(8, uint64_t, round_tab)[3] = {
+ 0x0000000000000000ULL,
+ 0x0001000100010001ULL,
+ 0x0002000200020002ULL,
+};
+
+DECLARE_ASM_CONST(8, uint64_t, bone) = 0x0101010101010101LL;
+
+static inline void sad8_1_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
+{
+ x86_reg len = -(stride * h);
+ __asm__ volatile (
+ ".p2align 4 \n\t"
+ "1: \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, %%"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"
+ "por %%mm1, %%mm3 \n\t"
+ "movq %%mm0, %%mm1 \n\t"
+ "movq %%mm3, %%mm2 \n\t"
+ "punpcklbw %%mm7, %%mm0 \n\t"
+ "punpckhbw %%mm7, %%mm1 \n\t"
+ "punpcklbw %%mm7, %%mm3 \n\t"
+ "punpckhbw %%mm7, %%mm2 \n\t"
+ "paddw %%mm1, %%mm0 \n\t"
+ "paddw %%mm3, %%mm2 \n\t"
+ "paddw %%mm2, %%mm0 \n\t"
+ "paddw %%mm0, %%mm6 \n\t"
+ "add %3, %%"REG_a" \n\t"
+ " js 1b \n\t"
+ : "+a" (len)
+ : "r" (blk1 - len), "r" (blk2 - len), "r" ((x86_reg) stride));
+}
+
+static inline void sad8_1_mmxext(uint8_t *blk1, uint8_t *blk2,
+ int stride, int h)
+{
+ __asm__ volatile (
+ ".p2align 4 \n\t"
+ "1: \n\t"
+ "movq (%1), %%mm0 \n\t"
+ "movq (%1, %3), %%mm1 \n\t"
+ "psadbw (%2), %%mm0 \n\t"
+ "psadbw (%2, %3), %%mm1 \n\t"
+ "paddw %%mm0, %%mm6 \n\t"
+ "paddw %%mm1, %%mm6 \n\t"
+ "lea (%1,%3,2), %1 \n\t"
+ "lea (%2,%3,2), %2 \n\t"
+ "sub $2, %0 \n\t"
+ " jg 1b \n\t"
+ : "+r" (h), "+r" (blk1), "+r" (blk2)
+ : "r" ((x86_reg) stride));
+}
+
+static int sad16_sse2(MpegEncContext *v, uint8_t *blk2, uint8_t *blk1,
+ int stride, int h)
+{
+ int ret;
+ __asm__ volatile (
+ "pxor %%xmm2, %%xmm2 \n\t"
+ ".p2align 4 \n\t"
+ "1: \n\t"
+ "movdqu (%1), %%xmm0 \n\t"
+ "movdqu (%1, %4), %%xmm1 \n\t"
+ "psadbw (%2), %%xmm0 \n\t"
+ "psadbw (%2, %4), %%xmm1 \n\t"
+ "paddw %%xmm0, %%xmm2 \n\t"
+ "paddw %%xmm1, %%xmm2 \n\t"
+ "lea (%1,%4,2), %1 \n\t"
+ "lea (%2,%4,2), %2 \n\t"
+ "sub $2, %0 \n\t"
+ " jg 1b \n\t"
+ "movhlps %%xmm2, %%xmm0 \n\t"
+ "paddw %%xmm0, %%xmm2 \n\t"
+ "movd %%xmm2, %3 \n\t"
+ : "+r" (h), "+r" (blk1), "+r" (blk2), "=r" (ret)
+ : "r" ((x86_reg) stride));
+ return ret;
+}
+
+static inline void sad8_x2a_mmxext(uint8_t *blk1, uint8_t *blk2,
+ int stride, int h)
+{
+ __asm__ volatile (
+ ".p2align 4 \n\t"
+ "1: \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"
+ "psadbw (%2), %%mm0 \n\t"
+ "psadbw (%2, %3), %%mm1 \n\t"
+ "paddw %%mm0, %%mm6 \n\t"
+ "paddw %%mm1, %%mm6 \n\t"
+ "lea (%1,%3,2), %1 \n\t"
+ "lea (%2,%3,2), %2 \n\t"
+ "sub $2, %0 \n\t"
+ " jg 1b \n\t"
+ : "+r" (h), "+r" (blk1), "+r" (blk2)
+ : "r" ((x86_reg) stride));
+}
+
+static inline void sad8_y2a_mmxext(uint8_t *blk1, uint8_t *blk2,
+ int stride, int h)
+{
+ __asm__ volatile (
+ "movq (%1), %%mm0 \n\t"
+ "add %3, %1 \n\t"
+ ".p2align 4 \n\t"
+ "1: \n\t"
+ "movq (%1), %%mm1 \n\t"
+ "movq (%1, %3), %%mm2 \n\t"
+ "pavgb %%mm1, %%mm0 \n\t"
+ "pavgb %%mm2, %%mm1 \n\t"
+ "psadbw (%2), %%mm0 \n\t"
+ "psadbw (%2, %3), %%mm1 \n\t"
+ "paddw %%mm0, %%mm6 \n\t"
+ "paddw %%mm1, %%mm6 \n\t"
+ "movq %%mm2, %%mm0 \n\t"
+ "lea (%1,%3,2), %1 \n\t"
+ "lea (%2,%3,2), %2 \n\t"
+ "sub $2, %0 \n\t"
+ " jg 1b \n\t"
+ : "+r" (h), "+r" (blk1), "+r" (blk2)
+ : "r" ((x86_reg) stride));
+}
+
+static inline void sad8_4_mmxext(uint8_t *blk1, uint8_t *blk2,
+ int stride, int h)
+{
+ __asm__ volatile (
+ "movq "MANGLE(bone)", %%mm5 \n\t"
+ "movq (%1), %%mm0 \n\t"
+ "pavgb 1(%1), %%mm0 \n\t"
+ "add %3, %1 \n\t"
+ ".p2align 4 \n\t"
+ "1: \n\t"
+ "movq (%1), %%mm1 \n\t"
+ "movq (%1,%3), %%mm2 \n\t"
+ "pavgb 1(%1), %%mm1 \n\t"
+ "pavgb 1(%1,%3), %%mm2 \n\t"
+ "psubusb %%mm5, %%mm1 \n\t"
+ "pavgb %%mm1, %%mm0 \n\t"
+ "pavgb %%mm2, %%mm1 \n\t"
+ "psadbw (%2), %%mm0 \n\t"
+ "psadbw (%2,%3), %%mm1 \n\t"
+ "paddw %%mm0, %%mm6 \n\t"
+ "paddw %%mm1, %%mm6 \n\t"
+ "movq %%mm2, %%mm0 \n\t"
+ "lea (%1,%3,2), %1 \n\t"
+ "lea (%2,%3,2), %2 \n\t"
+ "sub $2, %0 \n\t"
+ " jg 1b \n\t"
+ : "+r" (h), "+r" (blk1), "+r" (blk2)
+ : "r" ((x86_reg) stride));
+}
+
+static inline void sad8_2_mmx(uint8_t *blk1a, uint8_t *blk1b, uint8_t *blk2,
+ int stride, int h)
+{
+ x86_reg len = -(stride * h);
+ __asm__ volatile (
+ ".p2align 4 \n\t"
+ "1: \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, %%"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"
+ "psrlw $1, %%mm3 \n\t"
+ "packuswb %%mm3, %%mm1 \n\t"
+ "psubusb %%mm1, %%mm4 \n\t"
+ "psubusb %%mm2, %%mm1 \n\t"
+ "por %%mm4, %%mm1 \n\t"
+ "movq %%mm1, %%mm0 \n\t"
+ "punpcklbw %%mm7, %%mm0 \n\t"
+ "punpckhbw %%mm7, %%mm1 \n\t"
+ "paddw %%mm1, %%mm0 \n\t"
+ "paddw %%mm0, %%mm6 \n\t"
+ "add %4, %%"REG_a" \n\t"
+ " js 1b \n\t"
+ : "+a" (len)
+ : "r" (blk1a - len), "r" (blk1b - len), "r" (blk2 - len),
+ "r" ((x86_reg) stride));
+}
+
+static inline void sad8_4_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
+{
+ x86_reg len = -(stride * h);
+ __asm__ volatile (
+ "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"
+ "punpckhbw %%mm7, %%mm1 \n\t"
+ "punpcklbw %%mm7, %%mm2 \n\t"
+ "punpckhbw %%mm7, %%mm3 \n\t"
+ "paddw %%mm2, %%mm0 \n\t"
+ "paddw %%mm3, %%mm1 \n\t"
+ ".p2align 4 \n\t"
+ "1: \n\t"
+ "movq (%2, %%"REG_a"), %%mm2 \n\t"
+ "movq 1(%2, %%"REG_a"), %%mm4 \n\t"
+ "movq %%mm2, %%mm3 \n\t"
+ "movq %%mm4, %%mm5 \n\t"
+ "punpcklbw %%mm7, %%mm2 \n\t"
+ "punpckhbw %%mm7, %%mm3 \n\t"
+ "punpcklbw %%mm7, %%mm4 \n\t"
+ "punpckhbw %%mm7, %%mm5 \n\t"
+ "paddw %%mm4, %%mm2 \n\t"
+ "paddw %%mm5, %%mm3 \n\t"
+ "movq 16+"MANGLE(round_tab)", %%mm5 \n\t"
+ "paddw %%mm2, %%mm0 \n\t"
+ "paddw %%mm3, %%mm1 \n\t"
+ "paddw %%mm5, %%mm0 \n\t"
+ "paddw %%mm5, %%mm1 \n\t"
+ "movq (%3, %%"REG_a"), %%mm4 \n\t"
+ "movq (%3, %%"REG_a"), %%mm5 \n\t"
+ "psrlw $2, %%mm0 \n\t"
+ "psrlw $2, %%mm1 \n\t"
+ "packuswb %%mm1, %%mm0 \n\t"
+ "psubusb %%mm0, %%mm4 \n\t"
+ "psubusb %%mm5, %%mm0 \n\t"
+ "por %%mm4, %%mm0 \n\t"
+ "movq %%mm0, %%mm4 \n\t"
+ "punpcklbw %%mm7, %%mm0 \n\t"
+ "punpckhbw %%mm7, %%mm4 \n\t"
+ "paddw %%mm0, %%mm6 \n\t"
+ "paddw %%mm4, %%mm6 \n\t"
+ "movq %%mm2, %%mm0 \n\t"
+ "movq %%mm3, %%mm1 \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" ((x86_reg) stride));
+}
+
+static inline int sum_mmx(void)
+{
+ int ret;
+ __asm__ volatile (
+ "movq %%mm6, %%mm0 \n\t"
+ "psrlq $32, %%mm6 \n\t"
+ "paddw %%mm0, %%mm6 \n\t"
+ "movq %%mm6, %%mm0 \n\t"
+ "psrlq $16, %%mm6 \n\t"
+ "paddw %%mm0, %%mm6 \n\t"
+ "movd %%mm6, %0 \n\t"
+ : "=r" (ret));
+ return ret & 0xFFFF;
+}
+
+static inline int sum_mmxext(void)
+{
+ int ret;
+ __asm__ volatile (
+ "movd %%mm6, %0 \n\t"
+ : "=r" (ret));
+ return ret;
+}
+
+static inline void sad8_x2a_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
+{
+ sad8_2_mmx(blk1, blk1 + 1, blk2, stride, h);
+}
+
+static inline void sad8_y2a_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
+{
+ sad8_2_mmx(blk1, blk1 + stride, blk2, stride, h);
+}
+
+#define PIX_SAD(suf) \
+static int sad8_ ## suf(MpegEncContext *v, uint8_t *blk2, \
+ uint8_t *blk1, int stride, int h) \
+{ \
+ assert(h == 8); \
+ __asm__ volatile ( \
+ "pxor %%mm7, %%mm7 \n\t" \
+ "pxor %%mm6, %%mm6 \n\t" \
+ :); \
+ \
+ sad8_1_ ## suf(blk1, blk2, stride, 8); \
+ \
+ return sum_ ## suf(); \
+} \
+ \
+static int sad8_x2_ ## suf(MpegEncContext *v, uint8_t *blk2, \
+ uint8_t *blk1, int stride, int h) \
+{ \
+ assert(h == 8); \
+ __asm__ volatile ( \
+ "pxor %%mm7, %%mm7 \n\t" \
+ "pxor %%mm6, %%mm6 \n\t" \
+ "movq %0, %%mm5 \n\t" \
+ :: "m" (round_tab[1])); \
+ \
+ sad8_x2a_ ## suf(blk1, blk2, stride, 8); \
+ \
+ return sum_ ## suf(); \
+} \
+ \
+static int sad8_y2_ ## suf(MpegEncContext *v, uint8_t *blk2, \
+ uint8_t *blk1, int stride, int h) \
+{ \
+ assert(h == 8); \
+ __asm__ volatile ( \
+ "pxor %%mm7, %%mm7 \n\t" \
+ "pxor %%mm6, %%mm6 \n\t" \
+ "movq %0, %%mm5 \n\t" \
+ :: "m" (round_tab[1])); \
+ \
+ sad8_y2a_ ## suf(blk1, blk2, stride, 8); \
+ \
+ return sum_ ## suf(); \
+} \
+ \
+static int sad8_xy2_ ## suf(MpegEncContext *v, uint8_t *blk2, \
+ uint8_t *blk1, int stride, int h) \
+{ \
+ assert(h == 8); \
+ __asm__ volatile ( \
+ "pxor %%mm7, %%mm7 \n\t" \
+ "pxor %%mm6, %%mm6 \n\t" \
+ ::); \
+ \
+ sad8_4_ ## suf(blk1, blk2, stride, 8); \
+ \
+ return sum_ ## suf(); \
+} \
+ \
+static int sad16_ ## suf(MpegEncContext *v, uint8_t *blk2, \
+ uint8_t *blk1, int stride, int h) \
+{ \
+ __asm__ volatile ( \
+ "pxor %%mm7, %%mm7 \n\t" \
+ "pxor %%mm6, %%mm6 \n\t" \
+ :); \
+ \
+ sad8_1_ ## suf(blk1, blk2, stride, h); \
+ sad8_1_ ## suf(blk1 + 8, blk2 + 8, stride, h); \
+ \
+ return sum_ ## suf(); \
+} \
+ \
+static int sad16_x2_ ## suf(MpegEncContext *v, uint8_t *blk2, \
+ uint8_t *blk1, int stride, int h) \
+{ \
+ __asm__ volatile ( \
+ "pxor %%mm7, %%mm7 \n\t" \
+ "pxor %%mm6, %%mm6 \n\t" \
+ "movq %0, %%mm5 \n\t" \
+ :: "m" (round_tab[1])); \
+ \
+ sad8_x2a_ ## suf(blk1, blk2, stride, h); \
+ sad8_x2a_ ## suf(blk1 + 8, blk2 + 8, stride, h); \
+ \
+ return sum_ ## suf(); \
+} \
+ \
+static int sad16_y2_ ## suf(MpegEncContext *v, uint8_t *blk2, \
+ uint8_t *blk1, int stride, int h) \
+{ \
+ __asm__ volatile ( \
+ "pxor %%mm7, %%mm7 \n\t" \
+ "pxor %%mm6, %%mm6 \n\t" \
+ "movq %0, %%mm5 \n\t" \
+ :: "m" (round_tab[1])); \
+ \
+ sad8_y2a_ ## suf(blk1, blk2, stride, h); \
+ sad8_y2a_ ## suf(blk1 + 8, blk2 + 8, stride, h); \
+ \
+ return sum_ ## suf(); \
+} \
+ \
+static int sad16_xy2_ ## suf(MpegEncContext *v, uint8_t *blk2, \
+ uint8_t *blk1, int stride, int h) \
+{ \
+ __asm__ volatile ( \
+ "pxor %%mm7, %%mm7 \n\t" \
+ "pxor %%mm6, %%mm6 \n\t" \
+ ::); \
+ \
+ sad8_4_ ## suf(blk1, blk2, stride, h); \
+ sad8_4_ ## suf(blk1 + 8, blk2 + 8, stride, h); \
+ \
+ return sum_ ## suf(); \
+} \
+
+PIX_SAD(mmx)
+PIX_SAD(mmxext)
+
#endif /* HAVE_INLINE_ASM */
int ff_sse16_sse2(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,
@@ -821,22 +1233,49 @@ av_cold void ff_dsputil_init_x86(DSPContext *c, AVCodecContext *avctx)
if (INLINE_MMX(cpu_flags)) {
c->sum_abs_dctelem = sum_abs_dctelem_mmx;
+ c->pix_abs[0][0] = sad16_mmx;
+ c->pix_abs[0][1] = sad16_x2_mmx;
+ c->pix_abs[0][2] = sad16_y2_mmx;
+ c->pix_abs[0][3] = sad16_xy2_mmx;
+ c->pix_abs[1][0] = sad8_mmx;
+ c->pix_abs[1][1] = sad8_x2_mmx;
+ c->pix_abs[1][2] = sad8_y2_mmx;
+ c->pix_abs[1][3] = sad8_xy2_mmx;
+
+ c->sad[0] = sad16_mmx;
+ c->sad[1] = sad8_mmx;
+
c->sse[0] = sse16_mmx;
c->sse[1] = sse8_mmx;
c->vsad[4] = vsad_intra16_mmx;
c->nsse[0] = nsse16_mmx;
c->nsse[1] = nsse8_mmx;
+
if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
- c->vsad[0] = vsad16_mmx;
+ c->vsad[0] = vsad16_mmx;
}
}
if (INLINE_MMXEXT(cpu_flags)) {
c->sum_abs_dctelem = sum_abs_dctelem_mmxext;
- c->vsad[4] = vsad_intra16_mmxext;
+
+ c->vsad[4] = vsad_intra16_mmxext;
+
+ c->pix_abs[0][0] = sad16_mmxext;
+ c->pix_abs[1][0] = sad8_mmxext;
+
+ c->sad[0] = sad16_mmxext;
+ c->sad[1] = sad8_mmxext;
if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
+ c->pix_abs[0][1] = sad16_x2_mmxext;
+ c->pix_abs[0][2] = sad16_y2_mmxext;
+ c->pix_abs[0][3] = sad16_xy2_mmxext;
+ c->pix_abs[1][1] = sad8_x2_mmxext;
+ c->pix_abs[1][2] = sad8_y2_mmxext;
+ c->pix_abs[1][3] = sad8_xy2_mmxext;
+
c->vsad[0] = vsad16_mmxext;
}
}
@@ -845,6 +1284,10 @@ av_cold void ff_dsputil_init_x86(DSPContext *c, AVCodecContext *avctx)
c->sum_abs_dctelem = sum_abs_dctelem_sse2;
}
+ if (INLINE_SSE2(cpu_flags) && !(cpu_flags & AV_CPU_FLAG_3DNOW)) {
+ c->sad[0] = sad16_sse2;
+ }
+
#if HAVE_SSSE3_INLINE
if (INLINE_SSSE3(cpu_flags)) {
c->sum_abs_dctelem = sum_abs_dctelem_ssse3;
@@ -875,6 +1318,4 @@ av_cold void ff_dsputil_init_x86(DSPContext *c, AVCodecContext *avctx)
c->hadamard8_diff[0] = ff_hadamard8_diff16_ssse3;
c->hadamard8_diff[1] = ff_hadamard8_diff_ssse3;
}
-
- ff_dsputil_init_pix_mmx(c, avctx);
}
diff --git a/libavcodec/x86/dsputil_x86.h b/libavcodec/x86/dsputil_x86.h
deleted file mode 100644
index 8787de9..0000000
--- a/libavcodec/x86/dsputil_x86.h
+++ /dev/null
@@ -1,32 +0,0 @@
-/*
- * MMX optimized DSP utils
- * Copyright (c) 2007 Aurelien Jacobs <aurel at gnuage.org>
- *
- * This file is part of Libav.
- *
- * Libav is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Lesser General Public
- * License as published by the Free Software Foundation; either
- * version 2.1 of the License, or (at your option) any later version.
- *
- * Libav is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- * Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-#ifndef AVCODEC_X86_DSPUTIL_X86_H
-#define AVCODEC_X86_DSPUTIL_X86_H
-
-#include <stdint.h>
-
-#include "libavcodec/avcodec.h"
-#include "libavcodec/dsputil.h"
-
-void ff_dsputil_init_pix_mmx(DSPContext *c, AVCodecContext *avctx);
-
-#endif /* AVCODEC_X86_DSPUTIL_X86_H */
diff --git a/libavcodec/x86/motion_est.c b/libavcodec/x86/motion_est.c
deleted file mode 100644
index 2b34a28..0000000
--- a/libavcodec/x86/motion_est.c
+++ /dev/null
@@ -1,487 +0,0 @@
-/*
- * MMX optimized motion estimation
- * Copyright (c) 2001 Fabrice Bellard
- * Copyright (c) 2002-2004 Michael Niedermayer
- *
- * mostly by Michael Niedermayer <michaelni at gmx.at>
- *
- * This file is part of Libav.
- *
- * Libav is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Lesser General Public
- * License as published by the Free Software Foundation; either
- * version 2.1 of the License, or (at your option) any later version.
- *
- * Libav is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- * Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-#include "libavutil/attributes.h"
-#include "libavutil/internal.h"
-#include "libavutil/mem.h"
-#include "libavutil/x86/asm.h"
-#include "libavutil/x86/cpu.h"
-#include "libavcodec/mpegvideo.h"
-#include "dsputil_x86.h"
-
-#if HAVE_INLINE_ASM
-
-DECLARE_ASM_CONST(8, uint64_t, round_tab)[3] = {
- 0x0000000000000000ULL,
- 0x0001000100010001ULL,
- 0x0002000200020002ULL,
-};
-
-DECLARE_ASM_CONST(8, uint64_t, bone) = 0x0101010101010101LL;
-
-static inline void sad8_1_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
-{
- x86_reg len = -(stride * h);
- __asm__ volatile (
- ".p2align 4 \n\t"
- "1: \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, %%"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"
- "por %%mm1, %%mm3 \n\t"
- "movq %%mm0, %%mm1 \n\t"
- "movq %%mm3, %%mm2 \n\t"
- "punpcklbw %%mm7, %%mm0 \n\t"
- "punpckhbw %%mm7, %%mm1 \n\t"
- "punpcklbw %%mm7, %%mm3 \n\t"
- "punpckhbw %%mm7, %%mm2 \n\t"
- "paddw %%mm1, %%mm0 \n\t"
- "paddw %%mm3, %%mm2 \n\t"
- "paddw %%mm2, %%mm0 \n\t"
- "paddw %%mm0, %%mm6 \n\t"
- "add %3, %%"REG_a" \n\t"
- " js 1b \n\t"
- : "+a" (len)
- : "r" (blk1 - len), "r" (blk2 - len), "r" ((x86_reg) stride));
-}
-
-static inline void sad8_1_mmxext(uint8_t *blk1, uint8_t *blk2,
- int stride, int h)
-{
- __asm__ volatile (
- ".p2align 4 \n\t"
- "1: \n\t"
- "movq (%1), %%mm0 \n\t"
- "movq (%1, %3), %%mm1 \n\t"
- "psadbw (%2), %%mm0 \n\t"
- "psadbw (%2, %3), %%mm1 \n\t"
- "paddw %%mm0, %%mm6 \n\t"
- "paddw %%mm1, %%mm6 \n\t"
- "lea (%1,%3,2), %1 \n\t"
- "lea (%2,%3,2), %2 \n\t"
- "sub $2, %0 \n\t"
- " jg 1b \n\t"
- : "+r" (h), "+r" (blk1), "+r" (blk2)
- : "r" ((x86_reg) stride));
-}
-
-static int sad16_sse2(MpegEncContext *v, uint8_t *blk2, uint8_t *blk1,
- int stride, int h)
-{
- int ret;
- __asm__ volatile (
- "pxor %%xmm2, %%xmm2 \n\t"
- ".p2align 4 \n\t"
- "1: \n\t"
- "movdqu (%1), %%xmm0 \n\t"
- "movdqu (%1, %4), %%xmm1 \n\t"
- "psadbw (%2), %%xmm0 \n\t"
- "psadbw (%2, %4), %%xmm1 \n\t"
- "paddw %%xmm0, %%xmm2 \n\t"
- "paddw %%xmm1, %%xmm2 \n\t"
- "lea (%1,%4,2), %1 \n\t"
- "lea (%2,%4,2), %2 \n\t"
- "sub $2, %0 \n\t"
- " jg 1b \n\t"
- "movhlps %%xmm2, %%xmm0 \n\t"
- "paddw %%xmm0, %%xmm2 \n\t"
- "movd %%xmm2, %3 \n\t"
- : "+r" (h), "+r" (blk1), "+r" (blk2), "=r" (ret)
- : "r" ((x86_reg) stride));
- return ret;
-}
-
-static inline void sad8_x2a_mmxext(uint8_t *blk1, uint8_t *blk2,
- int stride, int h)
-{
- __asm__ volatile (
- ".p2align 4 \n\t"
- "1: \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"
- "psadbw (%2), %%mm0 \n\t"
- "psadbw (%2, %3), %%mm1 \n\t"
- "paddw %%mm0, %%mm6 \n\t"
- "paddw %%mm1, %%mm6 \n\t"
- "lea (%1,%3,2), %1 \n\t"
- "lea (%2,%3,2), %2 \n\t"
- "sub $2, %0 \n\t"
- " jg 1b \n\t"
- : "+r" (h), "+r" (blk1), "+r" (blk2)
- : "r" ((x86_reg) stride));
-}
-
-static inline void sad8_y2a_mmxext(uint8_t *blk1, uint8_t *blk2,
- int stride, int h)
-{
- __asm__ volatile (
- "movq (%1), %%mm0 \n\t"
- "add %3, %1 \n\t"
- ".p2align 4 \n\t"
- "1: \n\t"
- "movq (%1), %%mm1 \n\t"
- "movq (%1, %3), %%mm2 \n\t"
- "pavgb %%mm1, %%mm0 \n\t"
- "pavgb %%mm2, %%mm1 \n\t"
- "psadbw (%2), %%mm0 \n\t"
- "psadbw (%2, %3), %%mm1 \n\t"
- "paddw %%mm0, %%mm6 \n\t"
- "paddw %%mm1, %%mm6 \n\t"
- "movq %%mm2, %%mm0 \n\t"
- "lea (%1,%3,2), %1 \n\t"
- "lea (%2,%3,2), %2 \n\t"
- "sub $2, %0 \n\t"
- " jg 1b \n\t"
- : "+r" (h), "+r" (blk1), "+r" (blk2)
- : "r" ((x86_reg) stride));
-}
-
-static inline void sad8_4_mmxext(uint8_t *blk1, uint8_t *blk2,
- int stride, int h)
-{
- __asm__ volatile (
- "movq "MANGLE(bone)", %%mm5 \n\t"
- "movq (%1), %%mm0 \n\t"
- "pavgb 1(%1), %%mm0 \n\t"
- "add %3, %1 \n\t"
- ".p2align 4 \n\t"
- "1: \n\t"
- "movq (%1), %%mm1 \n\t"
- "movq (%1,%3), %%mm2 \n\t"
- "pavgb 1(%1), %%mm1 \n\t"
- "pavgb 1(%1,%3), %%mm2 \n\t"
- "psubusb %%mm5, %%mm1 \n\t"
- "pavgb %%mm1, %%mm0 \n\t"
- "pavgb %%mm2, %%mm1 \n\t"
- "psadbw (%2), %%mm0 \n\t"
- "psadbw (%2,%3), %%mm1 \n\t"
- "paddw %%mm0, %%mm6 \n\t"
- "paddw %%mm1, %%mm6 \n\t"
- "movq %%mm2, %%mm0 \n\t"
- "lea (%1,%3,2), %1 \n\t"
- "lea (%2,%3,2), %2 \n\t"
- "sub $2, %0 \n\t"
- " jg 1b \n\t"
- : "+r" (h), "+r" (blk1), "+r" (blk2)
- : "r" ((x86_reg) stride));
-}
-
-static inline void sad8_2_mmx(uint8_t *blk1a, uint8_t *blk1b, uint8_t *blk2,
- int stride, int h)
-{
- x86_reg len = -(stride * h);
- __asm__ volatile (
- ".p2align 4 \n\t"
- "1: \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, %%"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"
- "psrlw $1, %%mm3 \n\t"
- "packuswb %%mm3, %%mm1 \n\t"
- "psubusb %%mm1, %%mm4 \n\t"
- "psubusb %%mm2, %%mm1 \n\t"
- "por %%mm4, %%mm1 \n\t"
- "movq %%mm1, %%mm0 \n\t"
- "punpcklbw %%mm7, %%mm0 \n\t"
- "punpckhbw %%mm7, %%mm1 \n\t"
- "paddw %%mm1, %%mm0 \n\t"
- "paddw %%mm0, %%mm6 \n\t"
- "add %4, %%"REG_a" \n\t"
- " js 1b \n\t"
- : "+a" (len)
- : "r" (blk1a - len), "r" (blk1b - len), "r" (blk2 - len),
- "r" ((x86_reg) stride));
-}
-
-static inline void sad8_4_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
-{
- x86_reg len = -(stride * h);
- __asm__ volatile (
- "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"
- "punpckhbw %%mm7, %%mm1 \n\t"
- "punpcklbw %%mm7, %%mm2 \n\t"
- "punpckhbw %%mm7, %%mm3 \n\t"
- "paddw %%mm2, %%mm0 \n\t"
- "paddw %%mm3, %%mm1 \n\t"
- ".p2align 4 \n\t"
- "1: \n\t"
- "movq (%2, %%"REG_a"), %%mm2 \n\t"
- "movq 1(%2, %%"REG_a"), %%mm4 \n\t"
- "movq %%mm2, %%mm3 \n\t"
- "movq %%mm4, %%mm5 \n\t"
- "punpcklbw %%mm7, %%mm2 \n\t"
- "punpckhbw %%mm7, %%mm3 \n\t"
- "punpcklbw %%mm7, %%mm4 \n\t"
- "punpckhbw %%mm7, %%mm5 \n\t"
- "paddw %%mm4, %%mm2 \n\t"
- "paddw %%mm5, %%mm3 \n\t"
- "movq 16+"MANGLE(round_tab)", %%mm5 \n\t"
- "paddw %%mm2, %%mm0 \n\t"
- "paddw %%mm3, %%mm1 \n\t"
- "paddw %%mm5, %%mm0 \n\t"
- "paddw %%mm5, %%mm1 \n\t"
- "movq (%3, %%"REG_a"), %%mm4 \n\t"
- "movq (%3, %%"REG_a"), %%mm5 \n\t"
- "psrlw $2, %%mm0 \n\t"
- "psrlw $2, %%mm1 \n\t"
- "packuswb %%mm1, %%mm0 \n\t"
- "psubusb %%mm0, %%mm4 \n\t"
- "psubusb %%mm5, %%mm0 \n\t"
- "por %%mm4, %%mm0 \n\t"
- "movq %%mm0, %%mm4 \n\t"
- "punpcklbw %%mm7, %%mm0 \n\t"
- "punpckhbw %%mm7, %%mm4 \n\t"
- "paddw %%mm0, %%mm6 \n\t"
- "paddw %%mm4, %%mm6 \n\t"
- "movq %%mm2, %%mm0 \n\t"
- "movq %%mm3, %%mm1 \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" ((x86_reg) stride));
-}
-
-static inline int sum_mmx(void)
-{
- int ret;
- __asm__ volatile (
- "movq %%mm6, %%mm0 \n\t"
- "psrlq $32, %%mm6 \n\t"
- "paddw %%mm0, %%mm6 \n\t"
- "movq %%mm6, %%mm0 \n\t"
- "psrlq $16, %%mm6 \n\t"
- "paddw %%mm0, %%mm6 \n\t"
- "movd %%mm6, %0 \n\t"
- : "=r" (ret));
- return ret & 0xFFFF;
-}
-
-static inline int sum_mmxext(void)
-{
- int ret;
- __asm__ volatile (
- "movd %%mm6, %0 \n\t"
- : "=r" (ret));
- return ret;
-}
-
-static inline void sad8_x2a_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
-{
- sad8_2_mmx(blk1, blk1 + 1, blk2, stride, h);
-}
-
-static inline void sad8_y2a_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
-{
- sad8_2_mmx(blk1, blk1 + stride, blk2, stride, h);
-}
-
-#define PIX_SAD(suf) \
-static int sad8_ ## suf(MpegEncContext *v, uint8_t *blk2, \
- uint8_t *blk1, int stride, int h) \
-{ \
- assert(h == 8); \
- __asm__ volatile ( \
- "pxor %%mm7, %%mm7 \n\t" \
- "pxor %%mm6, %%mm6 \n\t" \
- :); \
- \
- sad8_1_ ## suf(blk1, blk2, stride, 8); \
- \
- return sum_ ## suf(); \
-} \
- \
-static int sad8_x2_ ## suf(MpegEncContext *v, uint8_t *blk2, \
- uint8_t *blk1, int stride, int h) \
-{ \
- assert(h == 8); \
- __asm__ volatile ( \
- "pxor %%mm7, %%mm7 \n\t" \
- "pxor %%mm6, %%mm6 \n\t" \
- "movq %0, %%mm5 \n\t" \
- :: "m" (round_tab[1])); \
- \
- sad8_x2a_ ## suf(blk1, blk2, stride, 8); \
- \
- return sum_ ## suf(); \
-} \
- \
-static int sad8_y2_ ## suf(MpegEncContext *v, uint8_t *blk2, \
- uint8_t *blk1, int stride, int h) \
-{ \
- assert(h == 8); \
- __asm__ volatile ( \
- "pxor %%mm7, %%mm7 \n\t" \
- "pxor %%mm6, %%mm6 \n\t" \
- "movq %0, %%mm5 \n\t" \
- :: "m" (round_tab[1])); \
- \
- sad8_y2a_ ## suf(blk1, blk2, stride, 8); \
- \
- return sum_ ## suf(); \
-} \
- \
-static int sad8_xy2_ ## suf(MpegEncContext *v, uint8_t *blk2, \
- uint8_t *blk1, int stride, int h) \
-{ \
- assert(h == 8); \
- __asm__ volatile ( \
- "pxor %%mm7, %%mm7 \n\t" \
- "pxor %%mm6, %%mm6 \n\t" \
- ::); \
- \
- sad8_4_ ## suf(blk1, blk2, stride, 8); \
- \
- return sum_ ## suf(); \
-} \
- \
-static int sad16_ ## suf(MpegEncContext *v, uint8_t *blk2, \
- uint8_t *blk1, int stride, int h) \
-{ \
- __asm__ volatile ( \
- "pxor %%mm7, %%mm7 \n\t" \
- "pxor %%mm6, %%mm6 \n\t" \
- :); \
- \
- sad8_1_ ## suf(blk1, blk2, stride, h); \
- sad8_1_ ## suf(blk1 + 8, blk2 + 8, stride, h); \
- \
- return sum_ ## suf(); \
-} \
- \
-static int sad16_x2_ ## suf(MpegEncContext *v, uint8_t *blk2, \
- uint8_t *blk1, int stride, int h) \
-{ \
- __asm__ volatile ( \
- "pxor %%mm7, %%mm7 \n\t" \
- "pxor %%mm6, %%mm6 \n\t" \
- "movq %0, %%mm5 \n\t" \
- :: "m" (round_tab[1])); \
- \
- sad8_x2a_ ## suf(blk1, blk2, stride, h); \
- sad8_x2a_ ## suf(blk1 + 8, blk2 + 8, stride, h); \
- \
- return sum_ ## suf(); \
-} \
- \
-static int sad16_y2_ ## suf(MpegEncContext *v, uint8_t *blk2, \
- uint8_t *blk1, int stride, int h) \
-{ \
- __asm__ volatile ( \
- "pxor %%mm7, %%mm7 \n\t" \
- "pxor %%mm6, %%mm6 \n\t" \
- "movq %0, %%mm5 \n\t" \
- :: "m" (round_tab[1])); \
- \
- sad8_y2a_ ## suf(blk1, blk2, stride, h); \
- sad8_y2a_ ## suf(blk1 + 8, blk2 + 8, stride, h); \
- \
- return sum_ ## suf(); \
-} \
- \
-static int sad16_xy2_ ## suf(MpegEncContext *v, uint8_t *blk2, \
- uint8_t *blk1, int stride, int h) \
-{ \
- __asm__ volatile ( \
- "pxor %%mm7, %%mm7 \n\t" \
- "pxor %%mm6, %%mm6 \n\t" \
- ::); \
- \
- sad8_4_ ## suf(blk1, blk2, stride, h); \
- sad8_4_ ## suf(blk1 + 8, blk2 + 8, stride, h); \
- \
- return sum_ ## suf(); \
-} \
-
-PIX_SAD(mmx)
-PIX_SAD(mmxext)
-
-#endif /* HAVE_INLINE_ASM */
-
-av_cold void ff_dsputil_init_pix_mmx(DSPContext *c, AVCodecContext *avctx)
-{
-#if HAVE_INLINE_ASM
- int cpu_flags = av_get_cpu_flags();
-
- if (INLINE_MMX(cpu_flags)) {
- c->pix_abs[0][0] = sad16_mmx;
- c->pix_abs[0][1] = sad16_x2_mmx;
- c->pix_abs[0][2] = sad16_y2_mmx;
- c->pix_abs[0][3] = sad16_xy2_mmx;
- c->pix_abs[1][0] = sad8_mmx;
- c->pix_abs[1][1] = sad8_x2_mmx;
- c->pix_abs[1][2] = sad8_y2_mmx;
- c->pix_abs[1][3] = sad8_xy2_mmx;
-
- c->sad[0] = sad16_mmx;
- c->sad[1] = sad8_mmx;
- }
- if (INLINE_MMXEXT(cpu_flags)) {
- c->pix_abs[0][0] = sad16_mmxext;
- c->pix_abs[1][0] = sad8_mmxext;
-
- c->sad[0] = sad16_mmxext;
- c->sad[1] = sad8_mmxext;
-
- if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
- c->pix_abs[0][1] = sad16_x2_mmxext;
- c->pix_abs[0][2] = sad16_y2_mmxext;
- c->pix_abs[0][3] = sad16_xy2_mmxext;
- c->pix_abs[1][1] = sad8_x2_mmxext;
- c->pix_abs[1][2] = sad8_y2_mmxext;
- c->pix_abs[1][3] = sad8_xy2_mmxext;
- }
- }
- if (INLINE_SSE2(cpu_flags) && !(cpu_flags & AV_CPU_FLAG_3DNOW)) {
- c->sad[0] = sad16_sse2;
- }
-#endif /* HAVE_INLINE_ASM */
-}
--
Libav/FFmpeg packaging
More information about the pkg-multimedia-commits
mailing list