x86: avcodec: Consistently name all init files
authorDiego Biurrun <diego@biurrun.de>
Sat, 11 Aug 2012 21:07:27 +0000 (23:07 +0200)
committerDiego Biurrun <diego@biurrun.de>
Thu, 16 Aug 2012 09:05:38 +0000 (11:05 +0200)
libavcodec/x86/Makefile
libavcodec/x86/pngdsp-init.c [deleted file]
libavcodec/x86/pngdsp_init.c [new file with mode: 0644]
libavcodec/x86/proresdsp-init.c [deleted file]
libavcodec/x86/proresdsp_init.c [new file with mode: 0644]
libavcodec/x86/vp8dsp-init.c [deleted file]
libavcodec/x86/vp8dsp_init.c [new file with mode: 0644]

index 4d06685..ad13367 100644 (file)
@@ -23,15 +23,15 @@ MMX-OBJS-$(CONFIG_H264DSP)             += x86/h264dsp_init.o
 MMX-OBJS-$(CONFIG_H264PRED)            += x86/h264_intrapred_init.o
 MMX-OBJS-$(CONFIG_LPC)                 += x86/lpc_mmx.o
 MMX-OBJS-$(CONFIG_MPEGAUDIODSP)        += x86/mpegaudiodec_mmx.o
-MMX-OBJS-$(CONFIG_PNG_DECODER)         += x86/pngdsp-init.o
-MMX-OBJS-$(CONFIG_PRORES_DECODER)      += x86/proresdsp-init.o
+MMX-OBJS-$(CONFIG_PNG_DECODER)         += x86/pngdsp_init.o
+MMX-OBJS-$(CONFIG_PRORES_DECODER)      += x86/proresdsp_init.o
 MMX-OBJS-$(CONFIG_RV30_DECODER)        += x86/rv34dsp_init.o
 MMX-OBJS-$(CONFIG_RV40_DECODER)        += x86/rv34dsp_init.o            \
                                           x86/rv40dsp_init.o
 MMX-OBJS-$(CONFIG_VC1_DECODER)         += x86/vc1dsp_mmx.o
 MMX-OBJS-$(CONFIG_VP5_DECODER)         += x86/vp56dsp_init.o
 MMX-OBJS-$(CONFIG_VP6_DECODER)         += x86/vp56dsp_init.o
-MMX-OBJS-$(CONFIG_VP8_DECODER)         += x86/vp8dsp-init.o
+MMX-OBJS-$(CONFIG_VP8_DECODER)         += x86/vp8dsp_init.o
 
 YASM-OBJS-$(CONFIG_AAC_DECODER)        += x86/sbrdsp.o
 YASM-OBJS-$(CONFIG_AC3DSP)             += x86/ac3dsp.o
diff --git a/libavcodec/x86/pngdsp-init.c b/libavcodec/x86/pngdsp-init.c
deleted file mode 100644 (file)
index aa21847..0000000
+++ /dev/null
@@ -1,51 +0,0 @@
-/*
- * x86 PNG optimizations.
- * Copyright (c) 2008 Loren Merrit <lorenm@u.washington.edu>
- *
- * 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/common.h"
-#include "libavutil/cpu.h"
-#include "libavcodec/pngdsp.h"
-
-void ff_add_png_paeth_prediction_mmx2 (uint8_t *dst, uint8_t *src,
-                                       uint8_t *top, int w, int bpp);
-void ff_add_png_paeth_prediction_ssse3(uint8_t *dst, uint8_t *src,
-                                       uint8_t *top, int w, int bpp);
-void ff_add_bytes_l2_mmx (uint8_t *dst, uint8_t *src1,
-                          uint8_t *src2, int w);
-void ff_add_bytes_l2_sse2(uint8_t *dst, uint8_t *src1,
-                          uint8_t *src2, int w);
-
-void ff_pngdsp_init_x86(PNGDSPContext *dsp)
-{
-#if HAVE_YASM
-    int flags = av_get_cpu_flags();
-
-#if ARCH_X86_32
-    if (flags & AV_CPU_FLAG_MMX)
-        dsp->add_bytes_l2         = ff_add_bytes_l2_mmx;
-#endif
-    if (flags & AV_CPU_FLAG_MMXEXT)
-        dsp->add_paeth_prediction = ff_add_png_paeth_prediction_mmx2;
-    if (flags & AV_CPU_FLAG_SSE2)
-        dsp->add_bytes_l2         = ff_add_bytes_l2_sse2;
-    if (flags & AV_CPU_FLAG_SSSE3)
-        dsp->add_paeth_prediction = ff_add_png_paeth_prediction_ssse3;
-#endif
-}
diff --git a/libavcodec/x86/pngdsp_init.c b/libavcodec/x86/pngdsp_init.c
new file mode 100644 (file)
index 0000000..aa21847
--- /dev/null
@@ -0,0 +1,51 @@
+/*
+ * x86 PNG optimizations.
+ * Copyright (c) 2008 Loren Merrit <lorenm@u.washington.edu>
+ *
+ * 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/common.h"
+#include "libavutil/cpu.h"
+#include "libavcodec/pngdsp.h"
+
+void ff_add_png_paeth_prediction_mmx2 (uint8_t *dst, uint8_t *src,
+                                       uint8_t *top, int w, int bpp);
+void ff_add_png_paeth_prediction_ssse3(uint8_t *dst, uint8_t *src,
+                                       uint8_t *top, int w, int bpp);
+void ff_add_bytes_l2_mmx (uint8_t *dst, uint8_t *src1,
+                          uint8_t *src2, int w);
+void ff_add_bytes_l2_sse2(uint8_t *dst, uint8_t *src1,
+                          uint8_t *src2, int w);
+
+void ff_pngdsp_init_x86(PNGDSPContext *dsp)
+{
+#if HAVE_YASM
+    int flags = av_get_cpu_flags();
+
+#if ARCH_X86_32
+    if (flags & AV_CPU_FLAG_MMX)
+        dsp->add_bytes_l2         = ff_add_bytes_l2_mmx;
+#endif
+    if (flags & AV_CPU_FLAG_MMXEXT)
+        dsp->add_paeth_prediction = ff_add_png_paeth_prediction_mmx2;
+    if (flags & AV_CPU_FLAG_SSE2)
+        dsp->add_bytes_l2         = ff_add_bytes_l2_sse2;
+    if (flags & AV_CPU_FLAG_SSSE3)
+        dsp->add_paeth_prediction = ff_add_png_paeth_prediction_ssse3;
+#endif
+}
diff --git a/libavcodec/x86/proresdsp-init.c b/libavcodec/x86/proresdsp-init.c
deleted file mode 100644 (file)
index f202f9f..0000000
+++ /dev/null
@@ -1,54 +0,0 @@
-/*
- * Apple ProRes compatible decoder
- *
- * Copyright (c) 2010-2011 Maxim Poliakovski
- *
- * 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 "libavcodec/proresdsp.h"
-
-void ff_prores_idct_put_10_sse2(uint16_t *dst, int linesize,
-                                DCTELEM *block, const int16_t *qmat);
-void ff_prores_idct_put_10_sse4(uint16_t *dst, int linesize,
-                                DCTELEM *block, const int16_t *qmat);
-void ff_prores_idct_put_10_avx (uint16_t *dst, int linesize,
-                                DCTELEM *block, const int16_t *qmat);
-
-void ff_proresdsp_x86_init(ProresDSPContext *dsp)
-{
-#if ARCH_X86_64 && HAVE_YASM
-    int flags = av_get_cpu_flags();
-
-    if (flags & AV_CPU_FLAG_SSE2) {
-        dsp->idct_permutation_type = FF_TRANSPOSE_IDCT_PERM;
-        dsp->idct_put = ff_prores_idct_put_10_sse2;
-    }
-
-    if (flags & AV_CPU_FLAG_SSE4) {
-        dsp->idct_permutation_type = FF_TRANSPOSE_IDCT_PERM;
-        dsp->idct_put = ff_prores_idct_put_10_sse4;
-    }
-
-#if HAVE_AVX
-    if (flags & AV_CPU_FLAG_AVX) {
-        dsp->idct_permutation_type = FF_TRANSPOSE_IDCT_PERM;
-        dsp->idct_put = ff_prores_idct_put_10_avx;
-    }
-#endif /* HAVE_AVX */
-#endif /* ARCH_X86_64 && HAVE_YASM */
-}
diff --git a/libavcodec/x86/proresdsp_init.c b/libavcodec/x86/proresdsp_init.c
new file mode 100644 (file)
index 0000000..f202f9f
--- /dev/null
@@ -0,0 +1,54 @@
+/*
+ * Apple ProRes compatible decoder
+ *
+ * Copyright (c) 2010-2011 Maxim Poliakovski
+ *
+ * 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 "libavcodec/proresdsp.h"
+
+void ff_prores_idct_put_10_sse2(uint16_t *dst, int linesize,
+                                DCTELEM *block, const int16_t *qmat);
+void ff_prores_idct_put_10_sse4(uint16_t *dst, int linesize,
+                                DCTELEM *block, const int16_t *qmat);
+void ff_prores_idct_put_10_avx (uint16_t *dst, int linesize,
+                                DCTELEM *block, const int16_t *qmat);
+
+void ff_proresdsp_x86_init(ProresDSPContext *dsp)
+{
+#if ARCH_X86_64 && HAVE_YASM
+    int flags = av_get_cpu_flags();
+
+    if (flags & AV_CPU_FLAG_SSE2) {
+        dsp->idct_permutation_type = FF_TRANSPOSE_IDCT_PERM;
+        dsp->idct_put = ff_prores_idct_put_10_sse2;
+    }
+
+    if (flags & AV_CPU_FLAG_SSE4) {
+        dsp->idct_permutation_type = FF_TRANSPOSE_IDCT_PERM;
+        dsp->idct_put = ff_prores_idct_put_10_sse4;
+    }
+
+#if HAVE_AVX
+    if (flags & AV_CPU_FLAG_AVX) {
+        dsp->idct_permutation_type = FF_TRANSPOSE_IDCT_PERM;
+        dsp->idct_put = ff_prores_idct_put_10_avx;
+    }
+#endif /* HAVE_AVX */
+#endif /* ARCH_X86_64 && HAVE_YASM */
+}
diff --git a/libavcodec/x86/vp8dsp-init.c b/libavcodec/x86/vp8dsp-init.c
deleted file mode 100644 (file)
index 04cfecc..0000000
+++ /dev/null
@@ -1,450 +0,0 @@
-/*
- * VP8 DSP functions x86-optimized
- * Copyright (c) 2010 Ronald S. Bultje <rsbultje@gmail.com>
- * Copyright (c) 2010 Jason Garrett-Glaser <darkshikari@gmail.com>
- *
- * 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/cpu.h"
-#include "libavutil/mem.h"
-#include "libavutil/x86/asm.h"
-#include "libavcodec/vp8dsp.h"
-
-#if HAVE_YASM
-
-/*
- * MC functions
- */
-extern void ff_put_vp8_epel4_h4_mmx2  (uint8_t *dst, ptrdiff_t dststride,
-                                       uint8_t *src, ptrdiff_t srcstride,
-                                       int height, int mx, int my);
-extern void ff_put_vp8_epel4_h6_mmx2  (uint8_t *dst, ptrdiff_t dststride,
-                                       uint8_t *src, ptrdiff_t srcstride,
-                                       int height, int mx, int my);
-extern void ff_put_vp8_epel4_v4_mmx2  (uint8_t *dst, ptrdiff_t dststride,
-                                       uint8_t *src, ptrdiff_t srcstride,
-                                       int height, int mx, int my);
-extern void ff_put_vp8_epel4_v6_mmx2  (uint8_t *dst, ptrdiff_t dststride,
-                                       uint8_t *src, ptrdiff_t srcstride,
-                                       int height, int mx, int my);
-
-extern void ff_put_vp8_epel8_h4_sse2  (uint8_t *dst, ptrdiff_t dststride,
-                                       uint8_t *src, ptrdiff_t srcstride,
-                                       int height, int mx, int my);
-extern void ff_put_vp8_epel8_h6_sse2  (uint8_t *dst, ptrdiff_t dststride,
-                                       uint8_t *src, ptrdiff_t srcstride,
-                                       int height, int mx, int my);
-extern void ff_put_vp8_epel8_v4_sse2  (uint8_t *dst, ptrdiff_t dststride,
-                                       uint8_t *src, ptrdiff_t srcstride,
-                                       int height, int mx, int my);
-extern void ff_put_vp8_epel8_v6_sse2  (uint8_t *dst, ptrdiff_t dststride,
-                                       uint8_t *src, ptrdiff_t srcstride,
-                                       int height, int mx, int my);
-
-extern void ff_put_vp8_epel4_h4_ssse3 (uint8_t *dst, ptrdiff_t dststride,
-                                       uint8_t *src, ptrdiff_t srcstride,
-                                       int height, int mx, int my);
-extern void ff_put_vp8_epel4_h6_ssse3 (uint8_t *dst, ptrdiff_t dststride,
-                                       uint8_t *src, ptrdiff_t srcstride,
-                                       int height, int mx, int my);
-extern void ff_put_vp8_epel4_v4_ssse3 (uint8_t *dst, ptrdiff_t dststride,
-                                       uint8_t *src, ptrdiff_t srcstride,
-                                       int height, int mx, int my);
-extern void ff_put_vp8_epel4_v6_ssse3 (uint8_t *dst, ptrdiff_t dststride,
-                                       uint8_t *src, ptrdiff_t srcstride,
-                                       int height, int mx, int my);
-extern void ff_put_vp8_epel8_h4_ssse3 (uint8_t *dst, ptrdiff_t dststride,
-                                       uint8_t *src, ptrdiff_t srcstride,
-                                       int height, int mx, int my);
-extern void ff_put_vp8_epel8_h6_ssse3 (uint8_t *dst, ptrdiff_t dststride,
-                                       uint8_t *src, ptrdiff_t srcstride,
-                                       int height, int mx, int my);
-extern void ff_put_vp8_epel8_v4_ssse3 (uint8_t *dst, ptrdiff_t dststride,
-                                       uint8_t *src, ptrdiff_t srcstride,
-                                       int height, int mx, int my);
-extern void ff_put_vp8_epel8_v6_ssse3 (uint8_t *dst, ptrdiff_t dststride,
-                                       uint8_t *src, ptrdiff_t srcstride,
-                                       int height, int mx, int my);
-
-extern void ff_put_vp8_bilinear4_h_mmx2  (uint8_t *dst, ptrdiff_t dststride,
-                                          uint8_t *src, ptrdiff_t srcstride,
-                                          int height, int mx, int my);
-extern void ff_put_vp8_bilinear8_h_sse2  (uint8_t *dst, ptrdiff_t dststride,
-                                          uint8_t *src, ptrdiff_t srcstride,
-                                          int height, int mx, int my);
-extern void ff_put_vp8_bilinear4_h_ssse3 (uint8_t *dst, ptrdiff_t dststride,
-                                          uint8_t *src, ptrdiff_t srcstride,
-                                          int height, int mx, int my);
-extern void ff_put_vp8_bilinear8_h_ssse3 (uint8_t *dst, ptrdiff_t dststride,
-                                          uint8_t *src, ptrdiff_t srcstride,
-                                          int height, int mx, int my);
-
-extern void ff_put_vp8_bilinear4_v_mmx2  (uint8_t *dst, ptrdiff_t dststride,
-                                          uint8_t *src, ptrdiff_t srcstride,
-                                          int height, int mx, int my);
-extern void ff_put_vp8_bilinear8_v_sse2  (uint8_t *dst, ptrdiff_t dststride,
-                                          uint8_t *src, ptrdiff_t srcstride,
-                                          int height, int mx, int my);
-extern void ff_put_vp8_bilinear4_v_ssse3 (uint8_t *dst, ptrdiff_t dststride,
-                                          uint8_t *src, ptrdiff_t srcstride,
-                                          int height, int mx, int my);
-extern void ff_put_vp8_bilinear8_v_ssse3 (uint8_t *dst, ptrdiff_t dststride,
-                                          uint8_t *src, ptrdiff_t srcstride,
-                                          int height, int mx, int my);
-
-
-extern void ff_put_vp8_pixels8_mmx (uint8_t *dst, ptrdiff_t dststride,
-                                    uint8_t *src, ptrdiff_t srcstride,
-                                    int height, int mx, int my);
-extern void ff_put_vp8_pixels16_mmx(uint8_t *dst, ptrdiff_t dststride,
-                                    uint8_t *src, ptrdiff_t srcstride,
-                                    int height, int mx, int my);
-extern void ff_put_vp8_pixels16_sse(uint8_t *dst, ptrdiff_t dststride,
-                                    uint8_t *src, ptrdiff_t srcstride,
-                                    int height, int mx, int my);
-
-#define TAP_W16(OPT, FILTERTYPE, TAPTYPE) \
-static void ff_put_vp8_ ## FILTERTYPE ## 16_ ## TAPTYPE ## _ ## OPT( \
-    uint8_t *dst,  ptrdiff_t dststride, uint8_t *src, \
-    ptrdiff_t srcstride, int height, int mx, int my) \
-{ \
-    ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \
-        dst,     dststride, src,     srcstride, height, mx, my); \
-    ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \
-        dst + 8, dststride, src + 8, srcstride, height, mx, my); \
-}
-#define TAP_W8(OPT, FILTERTYPE, TAPTYPE) \
-static void ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \
-    uint8_t *dst,  ptrdiff_t dststride, uint8_t *src, \
-    ptrdiff_t srcstride, int height, int mx, int my) \
-{ \
-    ff_put_vp8_ ## FILTERTYPE ## 4_ ## TAPTYPE ## _ ## OPT( \
-        dst,     dststride, src,     srcstride, height, mx, my); \
-    ff_put_vp8_ ## FILTERTYPE ## 4_ ## TAPTYPE ## _ ## OPT( \
-        dst + 4, dststride, src + 4, srcstride, height, mx, my); \
-}
-
-#if ARCH_X86_32
-TAP_W8 (mmx2,  epel, h4)
-TAP_W8 (mmx2,  epel, h6)
-TAP_W16(mmx2,  epel, h6)
-TAP_W8 (mmx2,  epel, v4)
-TAP_W8 (mmx2,  epel, v6)
-TAP_W16(mmx2,  epel, v6)
-TAP_W8 (mmx2,  bilinear, h)
-TAP_W16(mmx2,  bilinear, h)
-TAP_W8 (mmx2,  bilinear, v)
-TAP_W16(mmx2,  bilinear, v)
-#endif
-
-TAP_W16(sse2,  epel, h6)
-TAP_W16(sse2,  epel, v6)
-TAP_W16(sse2,  bilinear, h)
-TAP_W16(sse2,  bilinear, v)
-
-TAP_W16(ssse3, epel, h6)
-TAP_W16(ssse3, epel, v6)
-TAP_W16(ssse3, bilinear, h)
-TAP_W16(ssse3, bilinear, v)
-
-#define HVTAP(OPT, ALIGN, TAPNUMX, TAPNUMY, SIZE, MAXHEIGHT) \
-static void ff_put_vp8_epel ## SIZE ## _h ## TAPNUMX ## v ## TAPNUMY ## _ ## OPT( \
-    uint8_t *dst, ptrdiff_t dststride, uint8_t *src, \
-    ptrdiff_t srcstride, int height, int mx, int my) \
-{ \
-    DECLARE_ALIGNED(ALIGN, uint8_t, tmp)[SIZE * (MAXHEIGHT + TAPNUMY - 1)]; \
-    uint8_t *tmpptr = tmp + SIZE * (TAPNUMY / 2 - 1); \
-    src -= srcstride * (TAPNUMY / 2 - 1); \
-    ff_put_vp8_epel ## SIZE ## _h ## TAPNUMX ## _ ## OPT( \
-        tmp, SIZE,      src,    srcstride, height + TAPNUMY - 1, mx, my); \
-    ff_put_vp8_epel ## SIZE ## _v ## TAPNUMY ## _ ## OPT( \
-        dst, dststride, tmpptr, SIZE,      height,               mx, my); \
-}
-
-#if ARCH_X86_32
-#define HVTAPMMX(x, y) \
-HVTAP(mmx2, 8, x, y,  4,  8) \
-HVTAP(mmx2, 8, x, y,  8, 16)
-
-HVTAP(mmx2, 8, 6, 6, 16, 16)
-#else
-#define HVTAPMMX(x, y) \
-HVTAP(mmx2, 8, x, y,  4,  8)
-#endif
-
-HVTAPMMX(4, 4)
-HVTAPMMX(4, 6)
-HVTAPMMX(6, 4)
-HVTAPMMX(6, 6)
-
-#define HVTAPSSE2(x, y, w) \
-HVTAP(sse2,  16, x, y, w, 16) \
-HVTAP(ssse3, 16, x, y, w, 16)
-
-HVTAPSSE2(4, 4, 8)
-HVTAPSSE2(4, 6, 8)
-HVTAPSSE2(6, 4, 8)
-HVTAPSSE2(6, 6, 8)
-HVTAPSSE2(6, 6, 16)
-
-HVTAP(ssse3, 16, 4, 4, 4, 8)
-HVTAP(ssse3, 16, 4, 6, 4, 8)
-HVTAP(ssse3, 16, 6, 4, 4, 8)
-HVTAP(ssse3, 16, 6, 6, 4, 8)
-
-#define HVBILIN(OPT, ALIGN, SIZE, MAXHEIGHT) \
-static void ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT( \
-    uint8_t *dst, ptrdiff_t dststride, uint8_t *src, \
-    ptrdiff_t srcstride, int height, int mx, int my) \
-{ \
-    DECLARE_ALIGNED(ALIGN, uint8_t, tmp)[SIZE * (MAXHEIGHT + 2)]; \
-    ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT( \
-        tmp, SIZE,      src, srcstride, height + 1, mx, my); \
-    ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT( \
-        dst, dststride, tmp, SIZE,      height,     mx, my); \
-}
-
-HVBILIN(mmx2,  8,  4,  8)
-#if ARCH_X86_32
-HVBILIN(mmx2,  8,  8, 16)
-HVBILIN(mmx2,  8, 16, 16)
-#endif
-HVBILIN(sse2,  8,  8, 16)
-HVBILIN(sse2,  8, 16, 16)
-HVBILIN(ssse3, 8,  4,  8)
-HVBILIN(ssse3, 8,  8, 16)
-HVBILIN(ssse3, 8, 16, 16)
-
-extern void ff_vp8_idct_dc_add_mmx(uint8_t *dst, DCTELEM block[16],
-                                   ptrdiff_t stride);
-extern void ff_vp8_idct_dc_add_sse4(uint8_t *dst, DCTELEM block[16],
-                                    ptrdiff_t stride);
-extern void ff_vp8_idct_dc_add4y_mmx(uint8_t *dst, DCTELEM block[4][16],
-                                      ptrdiff_t stride);
-extern void ff_vp8_idct_dc_add4y_sse2(uint8_t *dst, DCTELEM block[4][16],
-                                      ptrdiff_t stride);
-extern void ff_vp8_idct_dc_add4uv_mmx(uint8_t *dst, DCTELEM block[2][16],
-                                      ptrdiff_t stride);
-extern void ff_vp8_luma_dc_wht_mmx(DCTELEM block[4][4][16], DCTELEM dc[16]);
-extern void ff_vp8_luma_dc_wht_sse(DCTELEM block[4][4][16], DCTELEM dc[16]);
-extern void ff_vp8_idct_add_mmx(uint8_t *dst, DCTELEM block[16],
-                                ptrdiff_t stride);
-extern void ff_vp8_idct_add_sse(uint8_t *dst, DCTELEM block[16],
-                                ptrdiff_t stride);
-
-#define DECLARE_LOOP_FILTER(NAME)\
-extern void ff_vp8_v_loop_filter_simple_ ## NAME(uint8_t *dst, \
-                                                 ptrdiff_t stride, \
-                                                 int flim);\
-extern void ff_vp8_h_loop_filter_simple_ ## NAME(uint8_t *dst, \
-                                                 ptrdiff_t stride, \
-                                                 int flim);\
-extern void ff_vp8_v_loop_filter16y_inner_ ## NAME (uint8_t *dst, \
-                                                     ptrdiff_t stride,\
-                                                    int e, int i, int hvt);\
-extern void ff_vp8_h_loop_filter16y_inner_ ## NAME (uint8_t *dst, \
-                                                    ptrdiff_t stride,\
-                                                    int e, int i, int hvt);\
-extern void ff_vp8_v_loop_filter8uv_inner_ ## NAME (uint8_t *dstU, \
-                                                    uint8_t *dstV,\
-                                                    ptrdiff_t s, \
-                                                    int e, int i, int hvt);\
-extern void ff_vp8_h_loop_filter8uv_inner_ ## NAME (uint8_t *dstU, \
-                                                    uint8_t *dstV,\
-                                                    ptrdiff_t s, \
-                                                    int e, int i, int hvt);\
-extern void ff_vp8_v_loop_filter16y_mbedge_ ## NAME(uint8_t *dst, \
-                                                    ptrdiff_t stride,\
-                                                    int e, int i, int hvt);\
-extern void ff_vp8_h_loop_filter16y_mbedge_ ## NAME(uint8_t *dst, \
-                                                    ptrdiff_t stride,\
-                                                    int e, int i, int hvt);\
-extern void ff_vp8_v_loop_filter8uv_mbedge_ ## NAME(uint8_t *dstU, \
-                                                    uint8_t *dstV,\
-                                                    ptrdiff_t s, \
-                                                    int e, int i, int hvt);\
-extern void ff_vp8_h_loop_filter8uv_mbedge_ ## NAME(uint8_t *dstU, \
-                                                    uint8_t *dstV,\
-                                                    ptrdiff_t s, \
-                                                    int e, int i, int hvt);
-
-DECLARE_LOOP_FILTER(mmx)
-DECLARE_LOOP_FILTER(mmx2)
-DECLARE_LOOP_FILTER(sse2)
-DECLARE_LOOP_FILTER(ssse3)
-DECLARE_LOOP_FILTER(sse4)
-
-#endif
-
-#define VP8_LUMA_MC_FUNC(IDX, SIZE, OPT) \
-    c->put_vp8_epel_pixels_tab[IDX][0][2] = ff_put_vp8_epel ## SIZE ## _h6_ ## OPT; \
-    c->put_vp8_epel_pixels_tab[IDX][2][0] = ff_put_vp8_epel ## SIZE ## _v6_ ## OPT; \
-    c->put_vp8_epel_pixels_tab[IDX][2][2] = ff_put_vp8_epel ## SIZE ## _h6v6_ ## OPT
-
-#define VP8_MC_FUNC(IDX, SIZE, OPT) \
-    c->put_vp8_epel_pixels_tab[IDX][0][1] = ff_put_vp8_epel ## SIZE ## _h4_ ## OPT; \
-    c->put_vp8_epel_pixels_tab[IDX][1][0] = ff_put_vp8_epel ## SIZE ## _v4_ ## OPT; \
-    c->put_vp8_epel_pixels_tab[IDX][1][1] = ff_put_vp8_epel ## SIZE ## _h4v4_ ## OPT; \
-    c->put_vp8_epel_pixels_tab[IDX][1][2] = ff_put_vp8_epel ## SIZE ## _h6v4_ ## OPT; \
-    c->put_vp8_epel_pixels_tab[IDX][2][1] = ff_put_vp8_epel ## SIZE ## _h4v6_ ## OPT; \
-    VP8_LUMA_MC_FUNC(IDX, SIZE, OPT)
-
-#define VP8_BILINEAR_MC_FUNC(IDX, SIZE, OPT) \
-    c->put_vp8_bilinear_pixels_tab[IDX][0][1] = ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT; \
-    c->put_vp8_bilinear_pixels_tab[IDX][0][2] = ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT; \
-    c->put_vp8_bilinear_pixels_tab[IDX][1][0] = ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT; \
-    c->put_vp8_bilinear_pixels_tab[IDX][1][1] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \
-    c->put_vp8_bilinear_pixels_tab[IDX][1][2] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \
-    c->put_vp8_bilinear_pixels_tab[IDX][2][0] = ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT; \
-    c->put_vp8_bilinear_pixels_tab[IDX][2][1] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \
-    c->put_vp8_bilinear_pixels_tab[IDX][2][2] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT
-
-
-av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
-{
-#if HAVE_YASM
-    int mm_flags = av_get_cpu_flags();
-
-    if (mm_flags & AV_CPU_FLAG_MMX) {
-        c->vp8_idct_dc_add    = ff_vp8_idct_dc_add_mmx;
-        c->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_mmx;
-#if ARCH_X86_32
-        c->vp8_idct_dc_add4y  = ff_vp8_idct_dc_add4y_mmx;
-        c->vp8_idct_add       = ff_vp8_idct_add_mmx;
-        c->vp8_luma_dc_wht    = ff_vp8_luma_dc_wht_mmx;
-        c->put_vp8_epel_pixels_tab[0][0][0]     =
-        c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_mmx;
-#endif
-        c->put_vp8_epel_pixels_tab[1][0][0]     =
-        c->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_mmx;
-
-#if ARCH_X86_32
-        c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_mmx;
-        c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_mmx;
-
-        c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_mmx;
-        c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_mmx;
-        c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_mmx;
-        c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_mmx;
-
-        c->vp8_v_loop_filter16y       = ff_vp8_v_loop_filter16y_mbedge_mmx;
-        c->vp8_h_loop_filter16y       = ff_vp8_h_loop_filter16y_mbedge_mmx;
-        c->vp8_v_loop_filter8uv       = ff_vp8_v_loop_filter8uv_mbedge_mmx;
-        c->vp8_h_loop_filter8uv       = ff_vp8_h_loop_filter8uv_mbedge_mmx;
-#endif
-    }
-
-    /* note that 4-tap width=16 functions are missing because w=16
-     * is only used for luma, and luma is always a copy or sixtap. */
-    if (mm_flags & AV_CPU_FLAG_MMXEXT) {
-        VP8_MC_FUNC(2, 4, mmx2);
-        VP8_BILINEAR_MC_FUNC(2, 4, mmx2);
-#if ARCH_X86_32
-        VP8_LUMA_MC_FUNC(0, 16, mmx2);
-        VP8_MC_FUNC(1, 8, mmx2);
-        VP8_BILINEAR_MC_FUNC(0, 16, mmx2);
-        VP8_BILINEAR_MC_FUNC(1, 8, mmx2);
-
-        c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_mmx2;
-        c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_mmx2;
-
-        c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_mmx2;
-        c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_mmx2;
-        c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_mmx2;
-        c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_mmx2;
-
-        c->vp8_v_loop_filter16y       = ff_vp8_v_loop_filter16y_mbedge_mmx2;
-        c->vp8_h_loop_filter16y       = ff_vp8_h_loop_filter16y_mbedge_mmx2;
-        c->vp8_v_loop_filter8uv       = ff_vp8_v_loop_filter8uv_mbedge_mmx2;
-        c->vp8_h_loop_filter8uv       = ff_vp8_h_loop_filter8uv_mbedge_mmx2;
-#endif
-    }
-
-    if (mm_flags & AV_CPU_FLAG_SSE) {
-        c->vp8_idct_add                         = ff_vp8_idct_add_sse;
-        c->vp8_luma_dc_wht                      = ff_vp8_luma_dc_wht_sse;
-        c->put_vp8_epel_pixels_tab[0][0][0]     =
-        c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_sse;
-    }
-
-    if (mm_flags & (AV_CPU_FLAG_SSE2|AV_CPU_FLAG_SSE2SLOW)) {
-        VP8_LUMA_MC_FUNC(0, 16, sse2);
-        VP8_MC_FUNC(1, 8, sse2);
-        VP8_BILINEAR_MC_FUNC(0, 16, sse2);
-        VP8_BILINEAR_MC_FUNC(1, 8, sse2);
-
-        c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_sse2;
-
-#if ARCH_X86_64 || HAVE_ALIGNED_STACK
-        c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_sse2;
-        c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_sse2;
-
-        c->vp8_v_loop_filter16y       = ff_vp8_v_loop_filter16y_mbedge_sse2;
-        c->vp8_v_loop_filter8uv       = ff_vp8_v_loop_filter8uv_mbedge_sse2;
-#endif
-    }
-
-    if (mm_flags & AV_CPU_FLAG_SSE2) {
-        c->vp8_idct_dc_add4y          = ff_vp8_idct_dc_add4y_sse2;
-
-        c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_sse2;
-
-#if ARCH_X86_64 || HAVE_ALIGNED_STACK
-        c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_sse2;
-        c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_sse2;
-
-        c->vp8_h_loop_filter16y       = ff_vp8_h_loop_filter16y_mbedge_sse2;
-        c->vp8_h_loop_filter8uv       = ff_vp8_h_loop_filter8uv_mbedge_sse2;
-#endif
-    }
-
-    if (mm_flags & AV_CPU_FLAG_SSSE3) {
-        VP8_LUMA_MC_FUNC(0, 16, ssse3);
-        VP8_MC_FUNC(1, 8, ssse3);
-        VP8_MC_FUNC(2, 4, ssse3);
-        VP8_BILINEAR_MC_FUNC(0, 16, ssse3);
-        VP8_BILINEAR_MC_FUNC(1, 8, ssse3);
-        VP8_BILINEAR_MC_FUNC(2, 4, ssse3);
-
-        c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_ssse3;
-        c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_ssse3;
-
-#if ARCH_X86_64 || HAVE_ALIGNED_STACK
-        c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_ssse3;
-        c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_ssse3;
-        c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_ssse3;
-        c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_ssse3;
-
-        c->vp8_v_loop_filter16y       = ff_vp8_v_loop_filter16y_mbedge_ssse3;
-        c->vp8_h_loop_filter16y       = ff_vp8_h_loop_filter16y_mbedge_ssse3;
-        c->vp8_v_loop_filter8uv       = ff_vp8_v_loop_filter8uv_mbedge_ssse3;
-        c->vp8_h_loop_filter8uv       = ff_vp8_h_loop_filter8uv_mbedge_ssse3;
-#endif
-    }
-
-    if (mm_flags & AV_CPU_FLAG_SSE4) {
-        c->vp8_idct_dc_add                  = ff_vp8_idct_dc_add_sse4;
-
-        c->vp8_h_loop_filter_simple   = ff_vp8_h_loop_filter_simple_sse4;
-#if ARCH_X86_64 || HAVE_ALIGNED_STACK
-        c->vp8_h_loop_filter16y       = ff_vp8_h_loop_filter16y_mbedge_sse4;
-        c->vp8_h_loop_filter8uv       = ff_vp8_h_loop_filter8uv_mbedge_sse4;
-#endif
-    }
-#endif
-}
diff --git a/libavcodec/x86/vp8dsp_init.c b/libavcodec/x86/vp8dsp_init.c
new file mode 100644 (file)
index 0000000..04cfecc
--- /dev/null
@@ -0,0 +1,450 @@
+/*
+ * VP8 DSP functions x86-optimized
+ * Copyright (c) 2010 Ronald S. Bultje <rsbultje@gmail.com>
+ * Copyright (c) 2010 Jason Garrett-Glaser <darkshikari@gmail.com>
+ *
+ * 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/cpu.h"
+#include "libavutil/mem.h"
+#include "libavutil/x86/asm.h"
+#include "libavcodec/vp8dsp.h"
+
+#if HAVE_YASM
+
+/*
+ * MC functions
+ */
+extern void ff_put_vp8_epel4_h4_mmx2  (uint8_t *dst, ptrdiff_t dststride,
+                                       uint8_t *src, ptrdiff_t srcstride,
+                                       int height, int mx, int my);
+extern void ff_put_vp8_epel4_h6_mmx2  (uint8_t *dst, ptrdiff_t dststride,
+                                       uint8_t *src, ptrdiff_t srcstride,
+                                       int height, int mx, int my);
+extern void ff_put_vp8_epel4_v4_mmx2  (uint8_t *dst, ptrdiff_t dststride,
+                                       uint8_t *src, ptrdiff_t srcstride,
+                                       int height, int mx, int my);
+extern void ff_put_vp8_epel4_v6_mmx2  (uint8_t *dst, ptrdiff_t dststride,
+                                       uint8_t *src, ptrdiff_t srcstride,
+                                       int height, int mx, int my);
+
+extern void ff_put_vp8_epel8_h4_sse2  (uint8_t *dst, ptrdiff_t dststride,
+                                       uint8_t *src, ptrdiff_t srcstride,
+                                       int height, int mx, int my);
+extern void ff_put_vp8_epel8_h6_sse2  (uint8_t *dst, ptrdiff_t dststride,
+                                       uint8_t *src, ptrdiff_t srcstride,
+                                       int height, int mx, int my);
+extern void ff_put_vp8_epel8_v4_sse2  (uint8_t *dst, ptrdiff_t dststride,
+                                       uint8_t *src, ptrdiff_t srcstride,
+                                       int height, int mx, int my);
+extern void ff_put_vp8_epel8_v6_sse2  (uint8_t *dst, ptrdiff_t dststride,
+                                       uint8_t *src, ptrdiff_t srcstride,
+                                       int height, int mx, int my);
+
+extern void ff_put_vp8_epel4_h4_ssse3 (uint8_t *dst, ptrdiff_t dststride,
+                                       uint8_t *src, ptrdiff_t srcstride,
+                                       int height, int mx, int my);
+extern void ff_put_vp8_epel4_h6_ssse3 (uint8_t *dst, ptrdiff_t dststride,
+                                       uint8_t *src, ptrdiff_t srcstride,
+                                       int height, int mx, int my);
+extern void ff_put_vp8_epel4_v4_ssse3 (uint8_t *dst, ptrdiff_t dststride,
+                                       uint8_t *src, ptrdiff_t srcstride,
+                                       int height, int mx, int my);
+extern void ff_put_vp8_epel4_v6_ssse3 (uint8_t *dst, ptrdiff_t dststride,
+                                       uint8_t *src, ptrdiff_t srcstride,
+                                       int height, int mx, int my);
+extern void ff_put_vp8_epel8_h4_ssse3 (uint8_t *dst, ptrdiff_t dststride,
+                                       uint8_t *src, ptrdiff_t srcstride,
+                                       int height, int mx, int my);
+extern void ff_put_vp8_epel8_h6_ssse3 (uint8_t *dst, ptrdiff_t dststride,
+                                       uint8_t *src, ptrdiff_t srcstride,
+                                       int height, int mx, int my);
+extern void ff_put_vp8_epel8_v4_ssse3 (uint8_t *dst, ptrdiff_t dststride,
+                                       uint8_t *src, ptrdiff_t srcstride,
+                                       int height, int mx, int my);
+extern void ff_put_vp8_epel8_v6_ssse3 (uint8_t *dst, ptrdiff_t dststride,
+                                       uint8_t *src, ptrdiff_t srcstride,
+                                       int height, int mx, int my);
+
+extern void ff_put_vp8_bilinear4_h_mmx2  (uint8_t *dst, ptrdiff_t dststride,
+                                          uint8_t *src, ptrdiff_t srcstride,
+                                          int height, int mx, int my);
+extern void ff_put_vp8_bilinear8_h_sse2  (uint8_t *dst, ptrdiff_t dststride,
+                                          uint8_t *src, ptrdiff_t srcstride,
+                                          int height, int mx, int my);
+extern void ff_put_vp8_bilinear4_h_ssse3 (uint8_t *dst, ptrdiff_t dststride,
+                                          uint8_t *src, ptrdiff_t srcstride,
+                                          int height, int mx, int my);
+extern void ff_put_vp8_bilinear8_h_ssse3 (uint8_t *dst, ptrdiff_t dststride,
+                                          uint8_t *src, ptrdiff_t srcstride,
+                                          int height, int mx, int my);
+
+extern void ff_put_vp8_bilinear4_v_mmx2  (uint8_t *dst, ptrdiff_t dststride,
+                                          uint8_t *src, ptrdiff_t srcstride,
+                                          int height, int mx, int my);
+extern void ff_put_vp8_bilinear8_v_sse2  (uint8_t *dst, ptrdiff_t dststride,
+                                          uint8_t *src, ptrdiff_t srcstride,
+                                          int height, int mx, int my);
+extern void ff_put_vp8_bilinear4_v_ssse3 (uint8_t *dst, ptrdiff_t dststride,
+                                          uint8_t *src, ptrdiff_t srcstride,
+                                          int height, int mx, int my);
+extern void ff_put_vp8_bilinear8_v_ssse3 (uint8_t *dst, ptrdiff_t dststride,
+                                          uint8_t *src, ptrdiff_t srcstride,
+                                          int height, int mx, int my);
+
+
+extern void ff_put_vp8_pixels8_mmx (uint8_t *dst, ptrdiff_t dststride,
+                                    uint8_t *src, ptrdiff_t srcstride,
+                                    int height, int mx, int my);
+extern void ff_put_vp8_pixels16_mmx(uint8_t *dst, ptrdiff_t dststride,
+                                    uint8_t *src, ptrdiff_t srcstride,
+                                    int height, int mx, int my);
+extern void ff_put_vp8_pixels16_sse(uint8_t *dst, ptrdiff_t dststride,
+                                    uint8_t *src, ptrdiff_t srcstride,
+                                    int height, int mx, int my);
+
+#define TAP_W16(OPT, FILTERTYPE, TAPTYPE) \
+static void ff_put_vp8_ ## FILTERTYPE ## 16_ ## TAPTYPE ## _ ## OPT( \
+    uint8_t *dst,  ptrdiff_t dststride, uint8_t *src, \
+    ptrdiff_t srcstride, int height, int mx, int my) \
+{ \
+    ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \
+        dst,     dststride, src,     srcstride, height, mx, my); \
+    ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \
+        dst + 8, dststride, src + 8, srcstride, height, mx, my); \
+}
+#define TAP_W8(OPT, FILTERTYPE, TAPTYPE) \
+static void ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \
+    uint8_t *dst,  ptrdiff_t dststride, uint8_t *src, \
+    ptrdiff_t srcstride, int height, int mx, int my) \
+{ \
+    ff_put_vp8_ ## FILTERTYPE ## 4_ ## TAPTYPE ## _ ## OPT( \
+        dst,     dststride, src,     srcstride, height, mx, my); \
+    ff_put_vp8_ ## FILTERTYPE ## 4_ ## TAPTYPE ## _ ## OPT( \
+        dst + 4, dststride, src + 4, srcstride, height, mx, my); \
+}
+
+#if ARCH_X86_32
+TAP_W8 (mmx2,  epel, h4)
+TAP_W8 (mmx2,  epel, h6)
+TAP_W16(mmx2,  epel, h6)
+TAP_W8 (mmx2,  epel, v4)
+TAP_W8 (mmx2,  epel, v6)
+TAP_W16(mmx2,  epel, v6)
+TAP_W8 (mmx2,  bilinear, h)
+TAP_W16(mmx2,  bilinear, h)
+TAP_W8 (mmx2,  bilinear, v)
+TAP_W16(mmx2,  bilinear, v)
+#endif
+
+TAP_W16(sse2,  epel, h6)
+TAP_W16(sse2,  epel, v6)
+TAP_W16(sse2,  bilinear, h)
+TAP_W16(sse2,  bilinear, v)
+
+TAP_W16(ssse3, epel, h6)
+TAP_W16(ssse3, epel, v6)
+TAP_W16(ssse3, bilinear, h)
+TAP_W16(ssse3, bilinear, v)
+
+#define HVTAP(OPT, ALIGN, TAPNUMX, TAPNUMY, SIZE, MAXHEIGHT) \
+static void ff_put_vp8_epel ## SIZE ## _h ## TAPNUMX ## v ## TAPNUMY ## _ ## OPT( \
+    uint8_t *dst, ptrdiff_t dststride, uint8_t *src, \
+    ptrdiff_t srcstride, int height, int mx, int my) \
+{ \
+    DECLARE_ALIGNED(ALIGN, uint8_t, tmp)[SIZE * (MAXHEIGHT + TAPNUMY - 1)]; \
+    uint8_t *tmpptr = tmp + SIZE * (TAPNUMY / 2 - 1); \
+    src -= srcstride * (TAPNUMY / 2 - 1); \
+    ff_put_vp8_epel ## SIZE ## _h ## TAPNUMX ## _ ## OPT( \
+        tmp, SIZE,      src,    srcstride, height + TAPNUMY - 1, mx, my); \
+    ff_put_vp8_epel ## SIZE ## _v ## TAPNUMY ## _ ## OPT( \
+        dst, dststride, tmpptr, SIZE,      height,               mx, my); \
+}
+
+#if ARCH_X86_32
+#define HVTAPMMX(x, y) \
+HVTAP(mmx2, 8, x, y,  4,  8) \
+HVTAP(mmx2, 8, x, y,  8, 16)
+
+HVTAP(mmx2, 8, 6, 6, 16, 16)
+#else
+#define HVTAPMMX(x, y) \
+HVTAP(mmx2, 8, x, y,  4,  8)
+#endif
+
+HVTAPMMX(4, 4)
+HVTAPMMX(4, 6)
+HVTAPMMX(6, 4)
+HVTAPMMX(6, 6)
+
+#define HVTAPSSE2(x, y, w) \
+HVTAP(sse2,  16, x, y, w, 16) \
+HVTAP(ssse3, 16, x, y, w, 16)
+
+HVTAPSSE2(4, 4, 8)
+HVTAPSSE2(4, 6, 8)
+HVTAPSSE2(6, 4, 8)
+HVTAPSSE2(6, 6, 8)
+HVTAPSSE2(6, 6, 16)
+
+HVTAP(ssse3, 16, 4, 4, 4, 8)
+HVTAP(ssse3, 16, 4, 6, 4, 8)
+HVTAP(ssse3, 16, 6, 4, 4, 8)
+HVTAP(ssse3, 16, 6, 6, 4, 8)
+
+#define HVBILIN(OPT, ALIGN, SIZE, MAXHEIGHT) \
+static void ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT( \
+    uint8_t *dst, ptrdiff_t dststride, uint8_t *src, \
+    ptrdiff_t srcstride, int height, int mx, int my) \
+{ \
+    DECLARE_ALIGNED(ALIGN, uint8_t, tmp)[SIZE * (MAXHEIGHT + 2)]; \
+    ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT( \
+        tmp, SIZE,      src, srcstride, height + 1, mx, my); \
+    ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT( \
+        dst, dststride, tmp, SIZE,      height,     mx, my); \
+}
+
+HVBILIN(mmx2,  8,  4,  8)
+#if ARCH_X86_32
+HVBILIN(mmx2,  8,  8, 16)
+HVBILIN(mmx2,  8, 16, 16)
+#endif
+HVBILIN(sse2,  8,  8, 16)
+HVBILIN(sse2,  8, 16, 16)
+HVBILIN(ssse3, 8,  4,  8)
+HVBILIN(ssse3, 8,  8, 16)
+HVBILIN(ssse3, 8, 16, 16)
+
+extern void ff_vp8_idct_dc_add_mmx(uint8_t *dst, DCTELEM block[16],
+                                   ptrdiff_t stride);
+extern void ff_vp8_idct_dc_add_sse4(uint8_t *dst, DCTELEM block[16],
+                                    ptrdiff_t stride);
+extern void ff_vp8_idct_dc_add4y_mmx(uint8_t *dst, DCTELEM block[4][16],
+                                      ptrdiff_t stride);
+extern void ff_vp8_idct_dc_add4y_sse2(uint8_t *dst, DCTELEM block[4][16],
+                                      ptrdiff_t stride);
+extern void ff_vp8_idct_dc_add4uv_mmx(uint8_t *dst, DCTELEM block[2][16],
+                                      ptrdiff_t stride);
+extern void ff_vp8_luma_dc_wht_mmx(DCTELEM block[4][4][16], DCTELEM dc[16]);
+extern void ff_vp8_luma_dc_wht_sse(DCTELEM block[4][4][16], DCTELEM dc[16]);
+extern void ff_vp8_idct_add_mmx(uint8_t *dst, DCTELEM block[16],
+                                ptrdiff_t stride);
+extern void ff_vp8_idct_add_sse(uint8_t *dst, DCTELEM block[16],
+                                ptrdiff_t stride);
+
+#define DECLARE_LOOP_FILTER(NAME)\
+extern void ff_vp8_v_loop_filter_simple_ ## NAME(uint8_t *dst, \
+                                                 ptrdiff_t stride, \
+                                                 int flim);\
+extern void ff_vp8_h_loop_filter_simple_ ## NAME(uint8_t *dst, \
+                                                 ptrdiff_t stride, \
+                                                 int flim);\
+extern void ff_vp8_v_loop_filter16y_inner_ ## NAME (uint8_t *dst, \
+                                                     ptrdiff_t stride,\
+                                                    int e, int i, int hvt);\
+extern void ff_vp8_h_loop_filter16y_inner_ ## NAME (uint8_t *dst, \
+                                                    ptrdiff_t stride,\
+                                                    int e, int i, int hvt);\
+extern void ff_vp8_v_loop_filter8uv_inner_ ## NAME (uint8_t *dstU, \
+                                                    uint8_t *dstV,\
+                                                    ptrdiff_t s, \
+                                                    int e, int i, int hvt);\
+extern void ff_vp8_h_loop_filter8uv_inner_ ## NAME (uint8_t *dstU, \
+                                                    uint8_t *dstV,\
+                                                    ptrdiff_t s, \
+                                                    int e, int i, int hvt);\
+extern void ff_vp8_v_loop_filter16y_mbedge_ ## NAME(uint8_t *dst, \
+                                                    ptrdiff_t stride,\
+                                                    int e, int i, int hvt);\
+extern void ff_vp8_h_loop_filter16y_mbedge_ ## NAME(uint8_t *dst, \
+                                                    ptrdiff_t stride,\
+                                                    int e, int i, int hvt);\
+extern void ff_vp8_v_loop_filter8uv_mbedge_ ## NAME(uint8_t *dstU, \
+                                                    uint8_t *dstV,\
+                                                    ptrdiff_t s, \
+                                                    int e, int i, int hvt);\
+extern void ff_vp8_h_loop_filter8uv_mbedge_ ## NAME(uint8_t *dstU, \
+                                                    uint8_t *dstV,\
+                                                    ptrdiff_t s, \
+                                                    int e, int i, int hvt);
+
+DECLARE_LOOP_FILTER(mmx)
+DECLARE_LOOP_FILTER(mmx2)
+DECLARE_LOOP_FILTER(sse2)
+DECLARE_LOOP_FILTER(ssse3)
+DECLARE_LOOP_FILTER(sse4)
+
+#endif
+
+#define VP8_LUMA_MC_FUNC(IDX, SIZE, OPT) \
+    c->put_vp8_epel_pixels_tab[IDX][0][2] = ff_put_vp8_epel ## SIZE ## _h6_ ## OPT; \
+    c->put_vp8_epel_pixels_tab[IDX][2][0] = ff_put_vp8_epel ## SIZE ## _v6_ ## OPT; \
+    c->put_vp8_epel_pixels_tab[IDX][2][2] = ff_put_vp8_epel ## SIZE ## _h6v6_ ## OPT
+
+#define VP8_MC_FUNC(IDX, SIZE, OPT) \
+    c->put_vp8_epel_pixels_tab[IDX][0][1] = ff_put_vp8_epel ## SIZE ## _h4_ ## OPT; \
+    c->put_vp8_epel_pixels_tab[IDX][1][0] = ff_put_vp8_epel ## SIZE ## _v4_ ## OPT; \
+    c->put_vp8_epel_pixels_tab[IDX][1][1] = ff_put_vp8_epel ## SIZE ## _h4v4_ ## OPT; \
+    c->put_vp8_epel_pixels_tab[IDX][1][2] = ff_put_vp8_epel ## SIZE ## _h6v4_ ## OPT; \
+    c->put_vp8_epel_pixels_tab[IDX][2][1] = ff_put_vp8_epel ## SIZE ## _h4v6_ ## OPT; \
+    VP8_LUMA_MC_FUNC(IDX, SIZE, OPT)
+
+#define VP8_BILINEAR_MC_FUNC(IDX, SIZE, OPT) \
+    c->put_vp8_bilinear_pixels_tab[IDX][0][1] = ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT; \
+    c->put_vp8_bilinear_pixels_tab[IDX][0][2] = ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT; \
+    c->put_vp8_bilinear_pixels_tab[IDX][1][0] = ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT; \
+    c->put_vp8_bilinear_pixels_tab[IDX][1][1] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \
+    c->put_vp8_bilinear_pixels_tab[IDX][1][2] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \
+    c->put_vp8_bilinear_pixels_tab[IDX][2][0] = ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT; \
+    c->put_vp8_bilinear_pixels_tab[IDX][2][1] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \
+    c->put_vp8_bilinear_pixels_tab[IDX][2][2] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT
+
+
+av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
+{
+#if HAVE_YASM
+    int mm_flags = av_get_cpu_flags();
+
+    if (mm_flags & AV_CPU_FLAG_MMX) {
+        c->vp8_idct_dc_add    = ff_vp8_idct_dc_add_mmx;
+        c->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_mmx;
+#if ARCH_X86_32
+        c->vp8_idct_dc_add4y  = ff_vp8_idct_dc_add4y_mmx;
+        c->vp8_idct_add       = ff_vp8_idct_add_mmx;
+        c->vp8_luma_dc_wht    = ff_vp8_luma_dc_wht_mmx;
+        c->put_vp8_epel_pixels_tab[0][0][0]     =
+        c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_mmx;
+#endif
+        c->put_vp8_epel_pixels_tab[1][0][0]     =
+        c->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_mmx;
+
+#if ARCH_X86_32
+        c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_mmx;
+        c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_mmx;
+
+        c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_mmx;
+        c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_mmx;
+        c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_mmx;
+        c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_mmx;
+
+        c->vp8_v_loop_filter16y       = ff_vp8_v_loop_filter16y_mbedge_mmx;
+        c->vp8_h_loop_filter16y       = ff_vp8_h_loop_filter16y_mbedge_mmx;
+        c->vp8_v_loop_filter8uv       = ff_vp8_v_loop_filter8uv_mbedge_mmx;
+        c->vp8_h_loop_filter8uv       = ff_vp8_h_loop_filter8uv_mbedge_mmx;
+#endif
+    }
+
+    /* note that 4-tap width=16 functions are missing because w=16
+     * is only used for luma, and luma is always a copy or sixtap. */
+    if (mm_flags & AV_CPU_FLAG_MMXEXT) {
+        VP8_MC_FUNC(2, 4, mmx2);
+        VP8_BILINEAR_MC_FUNC(2, 4, mmx2);
+#if ARCH_X86_32
+        VP8_LUMA_MC_FUNC(0, 16, mmx2);
+        VP8_MC_FUNC(1, 8, mmx2);
+        VP8_BILINEAR_MC_FUNC(0, 16, mmx2);
+        VP8_BILINEAR_MC_FUNC(1, 8, mmx2);
+
+        c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_mmx2;
+        c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_mmx2;
+
+        c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_mmx2;
+        c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_mmx2;
+        c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_mmx2;
+        c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_mmx2;
+
+        c->vp8_v_loop_filter16y       = ff_vp8_v_loop_filter16y_mbedge_mmx2;
+        c->vp8_h_loop_filter16y       = ff_vp8_h_loop_filter16y_mbedge_mmx2;
+        c->vp8_v_loop_filter8uv       = ff_vp8_v_loop_filter8uv_mbedge_mmx2;
+        c->vp8_h_loop_filter8uv       = ff_vp8_h_loop_filter8uv_mbedge_mmx2;
+#endif
+    }
+
+    if (mm_flags & AV_CPU_FLAG_SSE) {
+        c->vp8_idct_add                         = ff_vp8_idct_add_sse;
+        c->vp8_luma_dc_wht                      = ff_vp8_luma_dc_wht_sse;
+        c->put_vp8_epel_pixels_tab[0][0][0]     =
+        c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_sse;
+    }
+
+    if (mm_flags & (AV_CPU_FLAG_SSE2|AV_CPU_FLAG_SSE2SLOW)) {
+        VP8_LUMA_MC_FUNC(0, 16, sse2);
+        VP8_MC_FUNC(1, 8, sse2);
+        VP8_BILINEAR_MC_FUNC(0, 16, sse2);
+        VP8_BILINEAR_MC_FUNC(1, 8, sse2);
+
+        c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_sse2;
+
+#if ARCH_X86_64 || HAVE_ALIGNED_STACK
+        c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_sse2;
+        c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_sse2;
+
+        c->vp8_v_loop_filter16y       = ff_vp8_v_loop_filter16y_mbedge_sse2;
+        c->vp8_v_loop_filter8uv       = ff_vp8_v_loop_filter8uv_mbedge_sse2;
+#endif
+    }
+
+    if (mm_flags & AV_CPU_FLAG_SSE2) {
+        c->vp8_idct_dc_add4y          = ff_vp8_idct_dc_add4y_sse2;
+
+        c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_sse2;
+
+#if ARCH_X86_64 || HAVE_ALIGNED_STACK
+        c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_sse2;
+        c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_sse2;
+
+        c->vp8_h_loop_filter16y       = ff_vp8_h_loop_filter16y_mbedge_sse2;
+        c->vp8_h_loop_filter8uv       = ff_vp8_h_loop_filter8uv_mbedge_sse2;
+#endif
+    }
+
+    if (mm_flags & AV_CPU_FLAG_SSSE3) {
+        VP8_LUMA_MC_FUNC(0, 16, ssse3);
+        VP8_MC_FUNC(1, 8, ssse3);
+        VP8_MC_FUNC(2, 4, ssse3);
+        VP8_BILINEAR_MC_FUNC(0, 16, ssse3);
+        VP8_BILINEAR_MC_FUNC(1, 8, ssse3);
+        VP8_BILINEAR_MC_FUNC(2, 4, ssse3);
+
+        c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_ssse3;
+        c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_ssse3;
+
+#if ARCH_X86_64 || HAVE_ALIGNED_STACK
+        c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_ssse3;
+        c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_ssse3;
+        c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_ssse3;
+        c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_ssse3;
+
+        c->vp8_v_loop_filter16y       = ff_vp8_v_loop_filter16y_mbedge_ssse3;
+        c->vp8_h_loop_filter16y       = ff_vp8_h_loop_filter16y_mbedge_ssse3;
+        c->vp8_v_loop_filter8uv       = ff_vp8_v_loop_filter8uv_mbedge_ssse3;
+        c->vp8_h_loop_filter8uv       = ff_vp8_h_loop_filter8uv_mbedge_ssse3;
+#endif
+    }
+
+    if (mm_flags & AV_CPU_FLAG_SSE4) {
+        c->vp8_idct_dc_add                  = ff_vp8_idct_dc_add_sse4;
+
+        c->vp8_h_loop_filter_simple   = ff_vp8_h_loop_filter_simple_sse4;
+#if ARCH_X86_64 || HAVE_ALIGNED_STACK
+        c->vp8_h_loop_filter16y       = ff_vp8_h_loop_filter16y_mbedge_sse4;
+        c->vp8_h_loop_filter8uv       = ff_vp8_h_loop_filter8uv_mbedge_sse4;
+#endif
+    }
+#endif
+}