Move vp6_filter_diag4() x86 SIMD code from inline ASM to YASM. This should

help in fixing the Win64 fate failures.

Originally committed as revision 24922 to svn://svn.ffmpeg.org/ffmpeg/trunk

7 files changed, 178 insertions, 270 deletions

diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile
index 0f0b8a24be..43119fcc6f 100644
--- a/libavcodec/x86/Makefile
+++ b/libavcodec/x86/Makefile
@@ -31,10 +31,9 @@ MMX-OBJS-$(CONFIG_VP3_DECODER)         += x86/vp3dsp_mmx.o              \
 MMX-OBJS-$(CONFIG_VP5_DECODER)         += x86/vp3dsp_mmx.o              \
                                           x86/vp3dsp_sse2.o             \
                                           x86/vp56dsp_init.o
+YASM-OBJS-$(CONFIG_VP6_DECODER)        += x86/vp56dsp.o
 MMX-OBJS-$(CONFIG_VP6_DECODER)         += x86/vp3dsp_mmx.o              \
                                           x86/vp3dsp_sse2.o             \
-                                          x86/vp6dsp_mmx.o              \
-                                          x86/vp6dsp_sse2.o             \
                                           x86/vp56dsp_init.o
 YASM-OBJS-$(CONFIG_VP8_DECODER)        += x86/vp8dsp.o
 MMX-OBJS-$(CONFIG_VP8_DECODER)         += x86/vp8dsp-init.o
diff --git a/libavcodec/x86/vp56dsp.asm b/libavcodec/x86/vp56dsp.asm
new file mode 100644
index 0000000000..217e63a078
--- /dev/null
+++ b/libavcodec/x86/vp56dsp.asm
@@ -0,0 +1,170 @@
+;******************************************************************************
+;* MMX/SSE2-optimized functions for the VP6 decoder
+;* Copyright (C) 2009  Sebastien Lucas <sebastien.lucas@gmail.com>
+;* Copyright (C) 2009  Zuxy Meng <zuxy.meng@gmail.com>
+;*
+;* This file is part of FFmpeg.
+;*
+;* FFmpeg 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.
+;*
+;* FFmpeg 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 FFmpeg; if not, write to the Free Software
+;* 51, Inc., Foundation Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+;******************************************************************************
+
+%include "x86inc.asm"
+%include "x86util.asm"
+
+cextern pw_64
+
+SECTION .text
+
+%macro DIAG4_MMX 6
+    movq          m0, [%1+%2]
+    movq          m1, [%1+%3]
+    movq          m3, m0
+    movq          m4, m1
+    punpcklbw     m0, m7
+    punpcklbw     m1, m7
+    punpckhbw     m3, m7
+    punpckhbw     m4, m7
+    pmullw        m0, [rsp+8*11] ; src[x-8 ] * biweight [0]
+    pmullw        m1, [rsp+8*12] ; src[x   ] * biweight [1]
+    pmullw        m3, [rsp+8*11] ; src[x-8 ] * biweight [0]
+    pmullw        m4, [rsp+8*12] ; src[x   ] * biweight [1]
+    paddw         m0, m1
+    paddw         m3, m4
+    movq          m1, [%1+%4]
+    movq          m2, [%1+%5]
+    movq          m4, m1
+    movq          m5, m2
+    punpcklbw     m1, m7
+    punpcklbw     m2, m7
+    punpcklbw     m4, m7
+    punpcklbw     m5, m7
+    pmullw        m1, [rsp+8*13] ; src[x+8 ] * biweight [2]
+    pmullw        m2, [rsp+8*14] ; src[x+16] * biweight [3]
+    pmullw        m4, [rsp+8*13] ; src[x+8 ] * biweight [2]
+    pmullw        m5, [rsp+8*14] ; src[x+16] * biweight [3]
+    paddw         m1, m2
+    paddw         m4, m5
+    paddsw        m0, m1
+    paddsw        m3, m4
+    paddsw        m0, m6         ; Add 64
+    paddsw        m3, m6         ; Add 64
+    psraw         m0, 7
+    psraw         m3, 7
+    packuswb      m0, m3
+    movq        [%6], m0
+%endmacro
+
+%macro DIAG4_SSE2 6
+    movq          m0, [%1+%2]
+    movq          m1, [%1+%3]
+    punpcklbw     m0, m7
+    punpcklbw     m1, m7
+    pmullw        m0, m4         ; src[x-8 ] * biweight [0]
+    pmullw        m1, m5         ; src[x   ] * biweight [1]
+    paddw         m0, m1
+    movq          m1, [%1+%4]
+    movq          m2, [%1+%5]
+    punpcklbw     m1, m7
+    punpcklbw     m2, m7
+    pmullw        m1, m6         ; src[x+8 ] * biweight [2]
+    pmullw        m2, m3         ; src[x+16] * biweight [3]
+    paddw         m1, m2
+    paddsw        m0, m1
+    paddsw        m0, [pw_64]    ; Add 64
+    psraw         m0, 7
+    packuswb      m0, m0
+    movq        [%6], m0
+%endmacro
+
+%macro SPLAT4REGS_MMX 0
+    movq         m5, m3
+    punpcklwd    m3, m3
+    movq         m4, m3
+    punpckldq    m3, m3
+    punpckhdq    m4, m4
+    punpckhwd    m5, m5
+    movq         m6, m5
+    punpckhdq    m6, m6
+    punpckldq    m5, m5
+    movq [rsp+8*11], m3
+    movq [rsp+8*12], m4
+    movq [rsp+8*13], m5
+    movq [rsp+8*14], m6
+%endmacro
+
+%macro SPLAT4REGS_SSE2 0
+    pshuflw      m4, m3, 0x0
+    pshuflw      m5, m3, 0x55
+    pshuflw      m6, m3, 0xAA
+    pshuflw      m3, m3, 0xFF
+    punpcklqdq   m4, m4
+    punpcklqdq   m5, m5
+    punpcklqdq   m6, m6
+    punpcklqdq   m3, m3
+%endmacro
+
+%macro vp6_filter_diag4 2
+; void ff_vp6_filter_diag4_<opt>(uint8_t *dst, uint8_t *src, int stride,
+;                                const int16_t h_weight[4], const int16_t v_weights[4])
+cglobal vp6_filter_diag4_%1, 5, 7, %2
+    mov          r5, rsp         ; backup stack pointer
+    and         rsp, ~(mmsize-1) ; align stack
+%ifidn %1, sse2
+    sub         rsp, 8*11
+%else
+    sub         rsp, 8*15
+    movq         m6, [pw_64]
+%endif
+
+    sub          r1, r2
+
+    pxor         m7, m7
+    movq         m3, [r3]
+    SPLAT4REGS
+
+    mov          r3, rsp
+    mov          r6, 11
+.nextrow
+    DIAG4        r1, -1, 0, 1, 2, r3
+    add          r3, 8
+    add          r1, r2
+    dec          r6
+    jnz .nextrow
+
+    movq         m3, [r4]
+    SPLAT4REGS
+
+    lea          r3, [rsp+8]
+    mov          r6, 8
+.nextcol
+    DIAG4        r3, -8, 0, 8, 16, r0
+    add          r3, 8
+    add          r0, r2
+    dec          r6
+    jnz .nextcol
+
+    mov         rsp, r5          ; restore stack pointer
+    RET
+%endmacro
+
+INIT_MMX
+%define DIAG4      DIAG4_MMX
+%define SPLAT4REGS SPLAT4REGS_MMX
+vp6_filter_diag4 mmx,  0
+
+INIT_XMM
+%define DIAG4      DIAG4_SSE2
+%define SPLAT4REGS SPLAT4REGS_SSE2
+vp6_filter_diag4 sse2, 8
diff --git a/libavcodec/x86/vp56dsp_init.c b/libavcodec/x86/vp56dsp_init.c
index 8b2eb68594..5120ed231d 100644
--- a/libavcodec/x86/vp56dsp_init.c
+++ b/libavcodec/x86/vp56dsp_init.c
@@ -23,11 +23,15 @@
 #include "libavutil/x86_cpu.h"
 #include "libavcodec/dsputil.h"
 #include "libavcodec/vp56dsp.h"
-#include "vp6dsp_mmx.h"
-#include "vp6dsp_sse2.h"
+
+void ff_vp6_filter_diag4_mmx(uint8_t *dst, uint8_t *src, int stride,
+                             const int16_t *h_weights,const int16_t *v_weights);
+void ff_vp6_filter_diag4_sse2(uint8_t *dst, uint8_t *src, int stride,
+                              const int16_t *h_weights,const int16_t *v_weights);
 
 av_cold void ff_vp56dsp_init_x86(VP56DSPContext* c, enum CodecID codec)
 {
+#if HAVE_YASM
     int mm_flags = mm_support();
 
     if (CONFIG_VP6_DECODER && codec == CODEC_ID_VP6) {
@@ -39,4 +43,5 @@ av_cold void ff_vp56dsp_init_x86(VP56DSPContext* c, enum CodecID codec)
             c->vp6_filter_diag4 = ff_vp6_filter_diag4_sse2;
         }
     }
+#endif
 }
diff --git a/libavcodec/x86/vp6dsp_mmx.c b/libavcodec/x86/vp6dsp_mmx.c
deleted file mode 100644
index 905b3a7f0d..0000000000
--- a/libavcodec/x86/vp6dsp_mmx.c
+++ /dev/null
@@ -1,108 +0,0 @@
-/**
- * @file
- * MMX-optimized functions for the VP6 decoder
- *
- * Copyright (C) 2009  Sebastien Lucas <sebastien.lucas@gmail.com>
- *
- * This file is part of FFmpeg.
- *
- * FFmpeg 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.
- *
- * FFmpeg 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 FFmpeg; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-#include "libavutil/x86_cpu.h"
-#include "libavcodec/dsputil.h"
-#include "dsputil_mmx.h"
-#include "vp6dsp_mmx.h"
-
-
-#define DIAG4_MMX(in1,in2,in3,in4)                                            \
-    "movq  "#in1"(%0), %%mm0             \n\t"                                \
-    "movq  "#in2"(%0), %%mm1             \n\t"                                \
-    "movq       %%mm0, %%mm3             \n\t"                                \
-    "movq       %%mm1, %%mm4             \n\t"                                \
-    "punpcklbw  %%mm7, %%mm0             \n\t"                                \
-    "punpcklbw  %%mm7, %%mm1             \n\t"                                \
-    "punpckhbw  %%mm7, %%mm3             \n\t"                                \
-    "punpckhbw  %%mm7, %%mm4             \n\t"                                \
-    "pmullw     0(%2), %%mm0             \n\t" /* src[x-8 ] * biweight [0] */ \
-    "pmullw     8(%2), %%mm1             \n\t" /* src[x   ] * biweight [1] */ \
-    "pmullw     0(%2), %%mm3             \n\t" /* src[x-8 ] * biweight [0] */ \
-    "pmullw     8(%2), %%mm4             \n\t" /* src[x   ] * biweight [1] */ \
-    "paddw      %%mm1, %%mm0             \n\t"                                \
-    "paddw      %%mm4, %%mm3             \n\t"                                \
-    "movq  "#in3"(%0), %%mm1             \n\t"                                \
-    "movq  "#in4"(%0), %%mm2             \n\t"                                \
-    "movq       %%mm1, %%mm4             \n\t"                                \
-    "movq       %%mm2, %%mm5             \n\t"                                \
-    "punpcklbw  %%mm7, %%mm1             \n\t"                                \
-    "punpcklbw  %%mm7, %%mm2             \n\t"                                \
-    "punpckhbw  %%mm7, %%mm4             \n\t"                                \
-    "punpckhbw  %%mm7, %%mm5             \n\t"                                \
-    "pmullw    16(%2), %%mm1             \n\t" /* src[x+8 ] * biweight [2] */ \
-    "pmullw    24(%2), %%mm2             \n\t" /* src[x+16] * biweight [3] */ \
-    "pmullw    16(%2), %%mm4             \n\t" /* src[x+8 ] * biweight [2] */ \
-    "pmullw    24(%2), %%mm5             \n\t" /* src[x+16] * biweight [3] */ \
-    "paddw      %%mm2, %%mm1             \n\t"                                \
-    "paddw      %%mm5, %%mm4             \n\t"                                \
-    "paddsw     %%mm1, %%mm0             \n\t"                                \
-    "paddsw     %%mm4, %%mm3             \n\t"                                \
-    "paddsw     %%mm6, %%mm0             \n\t" /* Add 64 */                   \
-    "paddsw     %%mm6, %%mm3             \n\t" /* Add 64 */                   \
-    "psraw         $7, %%mm0             \n\t"                                \
-    "psraw         $7, %%mm3             \n\t"                                \
-    "packuswb   %%mm3, %%mm0             \n\t"                                \
-    "movq       %%mm0,  (%1)             \n\t"
-
-void ff_vp6_filter_diag4_mmx(uint8_t *dst, uint8_t *src, int stride,
-                             const int16_t *h_weights, const int16_t *v_weights)
-{
-    uint8_t tmp[8*11], *t = tmp;
-    int16_t weights[4*4];
-    int i;
-    src -= stride;
-
-    for (i=0; i<4*4; i++)
-        weights[i] = h_weights[i>>2];
-
-    __asm__ volatile(
-    "pxor %%mm7, %%mm7                   \n\t"
-    "movq "MANGLE(ff_pw_64)", %%mm6      \n\t"
-    "1:                                  \n\t"
-    DIAG4_MMX(-1,0,1,2)
-    "add  $8, %1                         \n\t"
-    "add  %3, %0                         \n\t"
-    "decl %4                             \n\t"
-    "jnz 1b                              \n\t"
-    : "+r"(src), "+r"(t)
-    : "r"(weights), "r"((x86_reg)stride), "r"(11)
-    : "memory");
-
-    t = tmp + 8;
-    for (i=0; i<4*4; i++)
-        weights[i] = v_weights[i>>2];
-
-    __asm__ volatile(
-    "pxor %%mm7, %%mm7                   \n\t"
-    "movq "MANGLE(ff_pw_64)", %%mm6      \n\t"
-    "1:                                  \n\t"
-    DIAG4_MMX(-8,0,8,16)
-    "add  $8, %0                         \n\t"
-    "add  %3, %1                         \n\t"
-    "decl %4                             \n\t"
-    "jnz 1b                              \n\t"
-    : "+r"(t), "+r"(dst)
-    : "r"(weights), "r"((x86_reg)stride), "r"(8)
-    : "memory");
-}
diff --git a/libavcodec/x86/vp6dsp_mmx.h b/libavcodec/x86/vp6dsp_mmx.h
deleted file mode 100644
index 743bc4361d..0000000000
--- a/libavcodec/x86/vp6dsp_mmx.h
+++ /dev/null
@@ -1,30 +0,0 @@
-/*
- * vp6dsp MMX function declarations
- * Copyright (c) 2009  Sebastien Lucas <sebastien.lucas@gmail.com>
- *
- * This file is part of FFmpeg.
- *
- * FFmpeg 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.
- *
- * FFmpeg 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 FFmpeg; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-#ifndef AVCODEC_X86_VP6DSP_MMX_H
-#define AVCODEC_X86_VP6DSP_MMX_H
-
-#include <stdint.h>
-
-void ff_vp6_filter_diag4_mmx(uint8_t *dst, uint8_t *src, int stride,
-                             const int16_t *h_weights,const int16_t *v_weights);
-
-#endif /* AVCODEC_X86_VP6DSP_MMX_H */
diff --git a/libavcodec/x86/vp6dsp_sse2.c b/libavcodec/x86/vp6dsp_sse2.c
deleted file mode 100644
index bfd733aa7f..0000000000
--- a/libavcodec/x86/vp6dsp_sse2.c
+++ /dev/null
@@ -1,98 +0,0 @@
-/**
- * @file
- * SSE2-optimized functions for the VP6 decoder
- *
- * Copyright (C) 2009  Zuxy Meng <zuxy.meng@gmail.com>
- *
- * This file is part of FFmpeg.
- *
- * FFmpeg 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.
- *
- * FFmpeg 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 FFmpeg; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-#include "libavutil/x86_cpu.h"
-#include "libavcodec/dsputil.h"
-#include "dsputil_mmx.h"
-#include "vp6dsp_sse2.h"
-
-#define DIAG4_SSE2(in1,in2,in3,in4)                                           \
-    "movq  "#in1"(%0), %%xmm0            \n\t"                                \
-    "movq  "#in2"(%0), %%xmm1            \n\t"                                \
-    "punpcklbw %%xmm7, %%xmm0            \n\t"                                \
-    "punpcklbw %%xmm7, %%xmm1            \n\t"                                \
-    "pmullw    %%xmm4, %%xmm0            \n\t" /* src[x-8 ] * biweight [0] */ \
-    "pmullw    %%xmm5, %%xmm1            \n\t" /* src[x   ] * biweight [1] */ \
-    "paddw     %%xmm1, %%xmm0            \n\t"                                \
-    "movq  "#in3"(%0), %%xmm1            \n\t"                                \
-    "movq  "#in4"(%0), %%xmm2            \n\t"                                \
-    "punpcklbw %%xmm7, %%xmm1            \n\t"                                \
-    "punpcklbw %%xmm7, %%xmm2            \n\t"                                \
-    "pmullw    %%xmm6, %%xmm1            \n\t" /* src[x+8 ] * biweight [2] */ \
-    "pmullw    %%xmm3, %%xmm2            \n\t" /* src[x+16] * biweight [3] */ \
-    "paddw     %%xmm2, %%xmm1            \n\t"                                \
-    "paddsw     %%xmm1, %%xmm0           \n\t"                                \
-    "paddsw "MANGLE(ff_pw_64)", %%xmm0   \n\t" /* Add 64 */                   \
-    "psraw         $7, %%xmm0            \n\t"                                \
-    "packuswb  %%xmm0, %%xmm0            \n\t"                                \
-    "movq      %%xmm0,   (%1)            \n\t"                                \
-
-void ff_vp6_filter_diag4_sse2(uint8_t *dst, uint8_t *src, int stride,
-                              const int16_t *h_weights,const int16_t *v_weights)
-{
-    uint8_t tmp[8*11], *t = tmp;
-    src -= stride;
-
-    __asm__ volatile(
-    "pxor           %%xmm7, %%xmm7       \n\t"
-    "movq               %4, %%xmm3       \n\t"
-    "pshuflw    $0, %%xmm3, %%xmm4       \n\t"
-    "punpcklqdq     %%xmm4, %%xmm4       \n\t"
-    "pshuflw   $85, %%xmm3, %%xmm5       \n\t"
-    "punpcklqdq     %%xmm5, %%xmm5       \n\t"
-    "pshuflw  $170, %%xmm3, %%xmm6       \n\t"
-    "punpcklqdq     %%xmm6, %%xmm6       \n\t"
-    "pshuflw  $255, %%xmm3, %%xmm3       \n\t"
-    "punpcklqdq     %%xmm3, %%xmm3       \n\t"
-    "1:                                  \n\t"
-    DIAG4_SSE2(-1,0,1,2)
-    "add  $8, %1                         \n\t"
-    "add  %2, %0                         \n\t"
-    "decl %3                             \n\t"
-    "jnz 1b                              \n\t"
-    : "+r"(src), "+r"(t)
-    : "g"((x86_reg)stride), "r"(11), "m"(*(const int64_t*)h_weights)
-    : "memory");
-
-    t = tmp + 8;
-
-    __asm__ volatile(
-    "movq               %4, %%xmm3       \n\t"
-    "pshuflw    $0, %%xmm3, %%xmm4       \n\t"
-    "punpcklqdq     %%xmm4, %%xmm4       \n\t"
-    "pshuflw   $85, %%xmm3, %%xmm5       \n\t"
-    "punpcklqdq     %%xmm5, %%xmm5       \n\t"
-    "pshuflw  $170, %%xmm3, %%xmm6       \n\t"
-    "punpcklqdq     %%xmm6, %%xmm6       \n\t"
-    "pshuflw  $255, %%xmm3, %%xmm3       \n\t"
-    "punpcklqdq     %%xmm3, %%xmm3       \n\t"
-    "1:                                  \n\t"
-    DIAG4_SSE2(-8,0,8,16)
-    "add  $8, %0                         \n\t"
-    "add  %2, %1                         \n\t"
-    "decl %3                             \n\t"
-    "jnz 1b                              \n\t"
-    : "+r"(t), "+r"(dst)
-    : "g"((x86_reg)stride), "r"(8), "m"(*(const int64_t*)v_weights)
-    : "memory");
-}
diff --git a/libavcodec/x86/vp6dsp_sse2.h b/libavcodec/x86/vp6dsp_sse2.h
deleted file mode 100644
index a30089a3e5..0000000000
--- a/libavcodec/x86/vp6dsp_sse2.h
+++ /dev/null
@@ -1,30 +0,0 @@
-/*
- * vp6dsp SSE2 function declarations
- * Copyright (c) 2009  Zuxy Meng <zuxy.meng@gmail.com>
- *
- * This file is part of FFmpeg.
- *
- * FFmpeg 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.
- *
- * FFmpeg 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 FFmpeg; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-#ifndef AVCODEC_X86_VP6DSP_SSE2_H
-#define AVCODEC_X86_VP6DSP_SSE2_H
-
-#include <stdint.h>
-
-void ff_vp6_filter_diag4_sse2(uint8_t *dst, uint8_t *src, int stride,
-                             const int16_t *h_weights,const int16_t *v_weights);
-
-#endif /* AVCODEC_X86_VP6DSP_SSE2_H */