x86: dsputilenc: convert hf_noise*_mmx to yasm

Signed-off-by: Timothy Gu <timothygu99@gmail.com>
Several bugfixes by: Christophe Gisquet <christophe.gisquet@gmail.com>
See: [FFmpeg-devel] [WIP] [PATCH 4/4] x86: dsputilenc: convert hf_noise*_mmx to yasm
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

2 files changed, 87 insertions, 255 deletions

diff --git a/libavcodec/x86/dsputilenc.asm b/libavcodec/x86/dsputilenc.asm
index 0628550145..84cb7b363b 100644
--- a/libavcodec/x86/dsputilenc.asm
+++ b/libavcodec/x86/dsputilenc.asm
@@ -608,3 +608,80 @@ INIT_XMM sse2
 SUM_ABS_DCTELEM 7, 2
 INIT_XMM ssse3
 SUM_ABS_DCTELEM 6, 2
+
+;------------------------------------------------------------------------------
+; int ff_hf_noise*_mmx(uint8_t *pix1, int lsize, int h)
+;------------------------------------------------------------------------------
+; %1 = 8/16. %2-5=m#
+%macro HF_NOISE_PART1 5
+    mova      m%2, [pix1q]
+%if %1 == 8
+    mova      m%3, m%2
+    psllq     m%2, 8
+    psrlq     m%3, 8
+    psrlq     m%2, 8
+%else
+    mova      m%3, [pix1q+1]
+%endif
+    mova      m%4, m%2
+    mova      m%5, m%3
+    punpcklbw m%2, m7
+    punpcklbw m%3, m7
+    punpckhbw m%4, m7
+    punpckhbw m%5, m7
+    psubw     m%2, m%3
+    psubw     m%4, m%5
+%endmacro
+
+; %1-2 = m#
+%macro HF_NOISE_PART2 4
+    psubw     m%1, m%3
+    psubw     m%2, m%4
+    pxor       m3, m3
+    pxor       m1, m1
+    pcmpgtw    m3, m%1
+    pcmpgtw    m1, m%2
+    pxor      m%1, m3
+    pxor      m%2, m1
+    psubw     m%1, m3
+    psubw     m%2, m1
+    paddw     m%2, m%1
+    paddw      m6, m%2
+%endmacro
+
+; %1 = 8/16
+%macro HF_NOISE 1
+cglobal hf_noise%1, 3,3,0, pix1, lsize, h
+    movsxdifnidn lsizeq, lsized
+    sub        hd, 2
+    pxor       m7, m7
+    pxor       m6, m6
+    HF_NOISE_PART1 %1, 0, 1, 2, 3
+    add     pix1q, lsizeq
+    HF_NOISE_PART1 %1, 4, 1, 5, 3
+    HF_NOISE_PART2     0, 2, 4, 5
+    add     pix1q, lsizeq
+.loop:
+    HF_NOISE_PART1 %1, 0, 1, 2, 3
+    HF_NOISE_PART2     4, 5, 0, 2
+    add     pix1q, lsizeq
+    HF_NOISE_PART1 %1, 4, 1, 5, 3
+    HF_NOISE_PART2     0, 2, 4, 5
+    add     pix1q, lsizeq
+    sub        hd, 2
+        jne .loop
+
+    mova       m0, m6
+    punpcklwd  m0, m7
+    punpckhwd  m6, m7
+    paddd      m6, m0
+    mova       m0, m6
+    psrlq      m6, 32
+    paddd      m0, m6
+    movd      eax, m0   ; eax = result of hf_noise8;
+    REP_RET                 ; return eax;
+%endmacro
+
+INIT_MMX mmx
+HF_NOISE 8
+HF_NOISE 16
diff --git a/libavcodec/x86/dsputilenc_mmx.c b/libavcodec/x86/dsputilenc_mmx.c
index 13128d29ad..e18048681f 100644
--- a/libavcodec/x86/dsputilenc_mmx.c
+++ b/libavcodec/x86/dsputilenc_mmx.c
@@ -52,6 +52,8 @@ int ff_sse16_mmx(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,
                  int line_size, int h);
 int ff_sse16_sse2(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,
                   int line_size, int h);
+int ff_hf_noise8_mmx(uint8_t *pix1, int lsize, int h);
+int ff_hf_noise16_mmx(uint8_t *pix1, int lsize, int h);
 
 #define hadamard_func(cpu)                                              \
     int ff_hadamard8_diff_ ## cpu(MpegEncContext *s, uint8_t *src1,     \
@@ -64,255 +66,8 @@ hadamard_func(mmxext)
 hadamard_func(sse2)
 hadamard_func(ssse3)
 
-#if HAVE_INLINE_ASM
-
 #if HAVE_YASM
 
-static int hf_noise8_mmx(uint8_t *pix1, int line_size, int h)
-{
-    int tmp;
-
-    __asm__ volatile (
-        "movl %3, %%ecx\n"
-        "pxor %%mm7, %%mm7\n"
-        "pxor %%mm6, %%mm6\n"
-
-        "movq (%0), %%mm0\n"
-        "movq %%mm0, %%mm1\n"
-        "psllq $8, %%mm0\n"
-        "psrlq $8, %%mm1\n"
-        "psrlq $8, %%mm0\n"
-        "movq %%mm0, %%mm2\n"
-        "movq %%mm1, %%mm3\n"
-        "punpcklbw %%mm7, %%mm0\n"
-        "punpcklbw %%mm7, %%mm1\n"
-        "punpckhbw %%mm7, %%mm2\n"
-        "punpckhbw %%mm7, %%mm3\n"
-        "psubw %%mm1, %%mm0\n"
-        "psubw %%mm3, %%mm2\n"
-
-        "add %2, %0\n"
-
-        "movq (%0), %%mm4\n"
-        "movq %%mm4, %%mm1\n"
-        "psllq $8, %%mm4\n"
-        "psrlq $8, %%mm1\n"
-        "psrlq $8, %%mm4\n"
-        "movq %%mm4, %%mm5\n"
-        "movq %%mm1, %%mm3\n"
-        "punpcklbw %%mm7, %%mm4\n"
-        "punpcklbw %%mm7, %%mm1\n"
-        "punpckhbw %%mm7, %%mm5\n"
-        "punpckhbw %%mm7, %%mm3\n"
-        "psubw %%mm1, %%mm4\n"
-        "psubw %%mm3, %%mm5\n"
-        "psubw %%mm4, %%mm0\n"
-        "psubw %%mm5, %%mm2\n"
-        "pxor %%mm3, %%mm3\n"
-        "pxor %%mm1, %%mm1\n"
-        "pcmpgtw %%mm0, %%mm3\n\t"
-        "pcmpgtw %%mm2, %%mm1\n\t"
-        "pxor %%mm3, %%mm0\n"
-        "pxor %%mm1, %%mm2\n"
-        "psubw %%mm3, %%mm0\n"
-        "psubw %%mm1, %%mm2\n"
-        "paddw %%mm0, %%mm2\n"
-        "paddw %%mm2, %%mm6\n"
-
-        "add %2, %0\n"
-        "1:\n"
-
-        "movq (%0), %%mm0\n"
-        "movq %%mm0, %%mm1\n"
-        "psllq $8, %%mm0\n"
-        "psrlq $8, %%mm1\n"
-        "psrlq $8, %%mm0\n"
-        "movq %%mm0, %%mm2\n"
-        "movq %%mm1, %%mm3\n"
-        "punpcklbw %%mm7, %%mm0\n"
-        "punpcklbw %%mm7, %%mm1\n"
-        "punpckhbw %%mm7, %%mm2\n"
-        "punpckhbw %%mm7, %%mm3\n"
-        "psubw %%mm1, %%mm0\n"
-        "psubw %%mm3, %%mm2\n"
-        "psubw %%mm0, %%mm4\n"
-        "psubw %%mm2, %%mm5\n"
-        "pxor  %%mm3, %%mm3\n"
-        "pxor  %%mm1, %%mm1\n"
-        "pcmpgtw %%mm4, %%mm3\n\t"
-        "pcmpgtw %%mm5, %%mm1\n\t"
-        "pxor  %%mm3, %%mm4\n"
-        "pxor  %%mm1, %%mm5\n"
-        "psubw %%mm3, %%mm4\n"
-        "psubw %%mm1, %%mm5\n"
-        "paddw %%mm4, %%mm5\n"
-        "paddw %%mm5, %%mm6\n"
-
-        "add %2, %0\n"
-
-        "movq (%0), %%mm4\n"
-        "movq      %%mm4, %%mm1\n"
-        "psllq $8, %%mm4\n"
-        "psrlq $8, %%mm1\n"
-        "psrlq $8, %%mm4\n"
-        "movq      %%mm4, %%mm5\n"
-        "movq      %%mm1, %%mm3\n"
-        "punpcklbw %%mm7, %%mm4\n"
-        "punpcklbw %%mm7, %%mm1\n"
-        "punpckhbw %%mm7, %%mm5\n"
-        "punpckhbw %%mm7, %%mm3\n"
-        "psubw     %%mm1, %%mm4\n"
-        "psubw     %%mm3, %%mm5\n"
-        "psubw     %%mm4, %%mm0\n"
-        "psubw     %%mm5, %%mm2\n"
-        "pxor      %%mm3, %%mm3\n"
-        "pxor      %%mm1, %%mm1\n"
-        "pcmpgtw   %%mm0, %%mm3\n\t"
-        "pcmpgtw   %%mm2, %%mm1\n\t"
-        "pxor      %%mm3, %%mm0\n"
-        "pxor      %%mm1, %%mm2\n"
-        "psubw     %%mm3, %%mm0\n"
-        "psubw     %%mm1, %%mm2\n"
-        "paddw     %%mm0, %%mm2\n"
-        "paddw     %%mm2, %%mm6\n"
-
-        "add  %2, %0\n"
-        "subl $2, %%ecx\n"
-        " jnz 1b\n"
-
-        "movq      %%mm6, %%mm0\n"
-        "punpcklwd %%mm7, %%mm0\n"
-        "punpckhwd %%mm7, %%mm6\n"
-        "paddd     %%mm0, %%mm6\n"
-
-        "movq  %%mm6, %%mm0\n"
-        "psrlq $32,   %%mm6\n"
-        "paddd %%mm6, %%mm0\n"
-        "movd  %%mm0, %1\n"
-        : "+r" (pix1), "=r" (tmp)
-        : "r" ((x86_reg) line_size), "g" (h - 2)
-        : "%ecx");
-
-    return tmp;
-}
-
-static int hf_noise16_mmx(uint8_t *pix1, int line_size, int h)
-{
-    int tmp;
-    uint8_t *pix = pix1;
-
-    __asm__ volatile (
-        "movl %3, %%ecx\n"
-        "pxor %%mm7, %%mm7\n"
-        "pxor %%mm6, %%mm6\n"
-
-        "movq (%0), %%mm0\n"
-        "movq 1(%0), %%mm1\n"
-        "movq %%mm0, %%mm2\n"
-        "movq %%mm1, %%mm3\n"
-        "punpcklbw %%mm7, %%mm0\n"
-        "punpcklbw %%mm7, %%mm1\n"
-        "punpckhbw %%mm7, %%mm2\n"
-        "punpckhbw %%mm7, %%mm3\n"
-        "psubw %%mm1, %%mm0\n"
-        "psubw %%mm3, %%mm2\n"
-
-        "add %2, %0\n"
-
-        "movq (%0), %%mm4\n"
-        "movq 1(%0), %%mm1\n"
-        "movq %%mm4, %%mm5\n"
-        "movq %%mm1, %%mm3\n"
-        "punpcklbw %%mm7, %%mm4\n"
-        "punpcklbw %%mm7, %%mm1\n"
-        "punpckhbw %%mm7, %%mm5\n"
-        "punpckhbw %%mm7, %%mm3\n"
-        "psubw %%mm1, %%mm4\n"
-        "psubw %%mm3, %%mm5\n"
-        "psubw %%mm4, %%mm0\n"
-        "psubw %%mm5, %%mm2\n"
-        "pxor %%mm3, %%mm3\n"
-        "pxor %%mm1, %%mm1\n"
-        "pcmpgtw %%mm0, %%mm3\n\t"
-        "pcmpgtw %%mm2, %%mm1\n\t"
-        "pxor %%mm3, %%mm0\n"
-        "pxor %%mm1, %%mm2\n"
-        "psubw %%mm3, %%mm0\n"
-        "psubw %%mm1, %%mm2\n"
-        "paddw %%mm0, %%mm2\n"
-        "paddw %%mm2, %%mm6\n"
-
-        "add %2, %0\n"
-        "1:\n"
-
-        "movq (%0), %%mm0\n"
-        "movq 1(%0), %%mm1\n"
-        "movq %%mm0, %%mm2\n"
-        "movq %%mm1, %%mm3\n"
-        "punpcklbw %%mm7, %%mm0\n"
-        "punpcklbw %%mm7, %%mm1\n"
-        "punpckhbw %%mm7, %%mm2\n"
-        "punpckhbw %%mm7, %%mm3\n"
-        "psubw %%mm1, %%mm0\n"
-        "psubw %%mm3, %%mm2\n"
-        "psubw %%mm0, %%mm4\n"
-        "psubw %%mm2, %%mm5\n"
-        "pxor %%mm3, %%mm3\n"
-        "pxor %%mm1, %%mm1\n"
-        "pcmpgtw %%mm4, %%mm3\n\t"
-        "pcmpgtw %%mm5, %%mm1\n\t"
-        "pxor %%mm3, %%mm4\n"
-        "pxor %%mm1, %%mm5\n"
-        "psubw %%mm3, %%mm4\n"
-        "psubw %%mm1, %%mm5\n"
-        "paddw %%mm4, %%mm5\n"
-        "paddw %%mm5, %%mm6\n"
-
-        "add %2, %0\n"
-
-        "movq (%0), %%mm4\n"
-        "movq 1(%0), %%mm1\n"
-        "movq %%mm4, %%mm5\n"
-        "movq %%mm1, %%mm3\n"
-        "punpcklbw %%mm7, %%mm4\n"
-        "punpcklbw %%mm7, %%mm1\n"
-        "punpckhbw %%mm7, %%mm5\n"
-        "punpckhbw %%mm7, %%mm3\n"
-        "psubw %%mm1, %%mm4\n"
-        "psubw %%mm3, %%mm5\n"
-        "psubw %%mm4, %%mm0\n"
-        "psubw %%mm5, %%mm2\n"
-        "pxor %%mm3, %%mm3\n"
-        "pxor %%mm1, %%mm1\n"
-        "pcmpgtw %%mm0, %%mm3\n\t"
-        "pcmpgtw %%mm2, %%mm1\n\t"
-        "pxor %%mm3, %%mm0\n"
-        "pxor %%mm1, %%mm2\n"
-        "psubw %%mm3, %%mm0\n"
-        "psubw %%mm1, %%mm2\n"
-        "paddw %%mm0, %%mm2\n"
-        "paddw %%mm2, %%mm6\n"
-
-        "add %2, %0\n"
-        "subl $2, %%ecx\n"
-        " jnz 1b\n"
-
-        "movq %%mm6, %%mm0\n"
-        "punpcklwd %%mm7, %%mm0\n"
-        "punpckhwd %%mm7, %%mm6\n"
-        "paddd %%mm0, %%mm6\n"
-
-        "movq %%mm6, %%mm0\n"
-        "psrlq $32, %%mm6\n"
-        "paddd %%mm6, %%mm0\n"
-        "movd %%mm0, %1\n"
-        : "+r" (pix1), "=r" (tmp)
-        : "r" ((x86_reg) line_size), "g" (h - 2)
-        : "%ecx");
-
-    return tmp + hf_noise8_mmx(pix + 8, line_size, h);
-}
-
 static int nsse16_mmx(MpegEncContext *c, uint8_t *pix1, uint8_t *pix2,
                       int line_size, int h)
 {
@@ -322,8 +77,8 @@ static int nsse16_mmx(MpegEncContext *c, uint8_t *pix1, uint8_t *pix2,
         score1 = c->dsp.sse[0](c, pix1, pix2, line_size, h);
     else
         score1 = ff_sse16_mmx(c, pix1, pix2, line_size, h);
-    score2 = hf_noise16_mmx(pix1, line_size, h) -
-             hf_noise16_mmx(pix2, line_size, h);
+    score2 = ff_hf_noise16_mmx(pix1, line_size, h) + ff_hf_noise8_mmx(pix1+8, line_size, h)
+           - ff_hf_noise16_mmx(pix2, line_size, h) - ff_hf_noise8_mmx(pix2+8, line_size, h);
 
     if (c)
         return score1 + FFABS(score2) * c->avctx->nsse_weight;
@@ -335,8 +90,8 @@ static int nsse8_mmx(MpegEncContext *c, uint8_t *pix1, uint8_t *pix2,
                      int line_size, int h)
 {
     int score1 = ff_sse8_mmx(c, pix1, pix2, line_size, h);
-    int score2 = hf_noise8_mmx(pix1, line_size, h) -
-                 hf_noise8_mmx(pix2, line_size, h);
+    int score2 = ff_hf_noise8_mmx(pix1, line_size, h) -
+                 ff_hf_noise8_mmx(pix2, line_size, h);
 
     if (c)
         return score1 + FFABS(score2) * c->avctx->nsse_weight;
@@ -346,6 +101,8 @@ static int nsse8_mmx(MpegEncContext *c, uint8_t *pix1, uint8_t *pix2,
 
 #endif /* HAVE_YASM */
 
+#if HAVE_INLINE_ASM
+
 static int vsad_intra16_mmx(MpegEncContext *v, uint8_t *pix, uint8_t *dummy,
                             int line_size, int h)
 {
@@ -689,10 +446,6 @@ av_cold void ff_dsputilenc_init_mmx(DSPContext *c, AVCodecContext *avctx,
 
         c->vsad[4] = vsad_intra16_mmx;
 
-#if HAVE_YASM
-        c->nsse[0] = nsse16_mmx;
-        c->nsse[1] = nsse8_mmx;
-#endif /* HAVE_YASM */
         if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
             c->vsad[0]      = vsad16_mmx;
             c->try_8x8basis = try_8x8basis_mmx;
@@ -741,6 +494,8 @@ av_cold void ff_dsputilenc_init_mmx(DSPContext *c, AVCodecContext *avctx,
         c->sum_abs_dctelem   = ff_sum_abs_dctelem_mmx;
         c->sse[0]            = ff_sse16_mmx;
         c->sse[1]            = ff_sse8_mmx;
+        c->nsse[0]           = nsse16_mmx;
+        c->nsse[1]           = nsse8_mmx;
     }
 
     if (EXTERNAL_MMXEXT(cpu_flags)) {