[FFmpeg-devel] [PATCH 3/6] avcodec/vp8dsp: add VP7 idct and loop filter

Peter Ross pross at xvid.org
Thu Feb 6 14:19:55 CET 2014


Signed-off-by: Peter Ross <pross at xvid.org>
---
 libavcodec/arm/vp8dsp.h            |   4 +-
 libavcodec/arm/vp8dsp_init_arm.c   |   6 +-
 libavcodec/arm/vp8dsp_init_armv6.c |   4 +-
 libavcodec/arm/vp8dsp_init_neon.c  |   4 +-
 libavcodec/vp8.c                   |   2 +-
 libavcodec/vp8dsp.c                | 273 ++++++++++++++++++++++++++-----------
 libavcodec/vp8dsp.h                |   6 +-
 libavcodec/x86/vp8dsp_init.c       |  20 ++-
 8 files changed, 229 insertions(+), 90 deletions(-)

diff --git a/libavcodec/arm/vp8dsp.h b/libavcodec/arm/vp8dsp.h
index 6041ef1..786741e 100644
--- a/libavcodec/arm/vp8dsp.h
+++ b/libavcodec/arm/vp8dsp.h
@@ -21,8 +21,8 @@
 
 #include "libavcodec/vp8dsp.h"
 
-void ff_vp8dsp_init_armv6(VP8DSPContext *dsp);
-void ff_vp8dsp_init_neon(VP8DSPContext *dsp);
+void ff_vp8dsp_init_armv6(VP8DSPContext *dsp, int vp7);
+void ff_vp8dsp_init_neon(VP8DSPContext *dsp, int vp7);
 
 #define VP8_LF_Y(hv, inner, opt)                                             \
     void ff_vp8_##hv##_loop_filter16##inner##_##opt(uint8_t *dst,            \
diff --git a/libavcodec/arm/vp8dsp_init_arm.c b/libavcodec/arm/vp8dsp_init_arm.c
index d360ae3..b684e9b 100644
--- a/libavcodec/arm/vp8dsp_init_arm.c
+++ b/libavcodec/arm/vp8dsp_init_arm.c
@@ -23,12 +23,12 @@
 #include "libavcodec/vp8dsp.h"
 #include "vp8dsp.h"
 
-av_cold void ff_vp8dsp_init_arm(VP8DSPContext *dsp)
+av_cold void ff_vp8dsp_init_arm(VP8DSPContext *dsp, int vp7)
 {
     int cpu_flags = av_get_cpu_flags();
 
     if (have_armv6(cpu_flags))
-        ff_vp8dsp_init_armv6(dsp);
+        ff_vp8dsp_init_armv6(dsp, vp7);
     if (have_neon(cpu_flags))
-        ff_vp8dsp_init_neon(dsp);
+        ff_vp8dsp_init_neon(dsp, vp7);
 }
diff --git a/libavcodec/arm/vp8dsp_init_armv6.c b/libavcodec/arm/vp8dsp_init_armv6.c
index 563268e..d8197e7 100644
--- a/libavcodec/arm/vp8dsp_init_armv6.c
+++ b/libavcodec/arm/vp8dsp_init_armv6.c
@@ -40,8 +40,9 @@ VP8_BILIN(16, armv6);
 VP8_BILIN(8,  armv6);
 VP8_BILIN(4,  armv6);
 
-av_cold void ff_vp8dsp_init_armv6(VP8DSPContext *dsp)
+av_cold void ff_vp8dsp_init_armv6(VP8DSPContext *dsp, int vp7)
 {
+    if (!vp7) {
     dsp->vp8_luma_dc_wht    = ff_vp8_luma_dc_wht_armv6;
     dsp->vp8_luma_dc_wht_dc = ff_vp8_luma_dc_wht_dc_armv6;
 
@@ -62,6 +63,7 @@ av_cold void ff_vp8dsp_init_armv6(VP8DSPContext *dsp)
 
     dsp->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter16_simple_armv6;
     dsp->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter16_simple_armv6;
+    }
 
     dsp->put_vp8_epel_pixels_tab[0][0][0] = ff_put_vp8_pixels16_armv6;
     dsp->put_vp8_epel_pixels_tab[0][0][2] = ff_put_vp8_epel16_h6_armv6;
diff --git a/libavcodec/arm/vp8dsp_init_neon.c b/libavcodec/arm/vp8dsp_init_neon.c
index ae045a6..400ef2d 100644
--- a/libavcodec/arm/vp8dsp_init_neon.c
+++ b/libavcodec/arm/vp8dsp_init_neon.c
@@ -39,8 +39,9 @@ VP8_BILIN(16, neon);
 VP8_BILIN(8,  neon);
 VP8_BILIN(4,  neon);
 
-av_cold void ff_vp8dsp_init_neon(VP8DSPContext *dsp)
+av_cold void ff_vp8dsp_init_neon(VP8DSPContext *dsp, int vp7)
 {
+    if (!vp7) {
     dsp->vp8_luma_dc_wht    = ff_vp8_luma_dc_wht_neon;
 
     dsp->vp8_idct_add       = ff_vp8_idct_add_neon;
@@ -60,6 +61,7 @@ av_cold void ff_vp8dsp_init_neon(VP8DSPContext *dsp)
 
     dsp->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter16_simple_neon;
     dsp->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter16_simple_neon;
+    }
 
     dsp->put_vp8_epel_pixels_tab[0][0][0] = ff_put_vp8_pixels16_neon;
     dsp->put_vp8_epel_pixels_tab[0][0][2] = ff_put_vp8_epel16_h6_neon;
diff --git a/libavcodec/vp8.c b/libavcodec/vp8.c
index cf38a32..312090a 100644
--- a/libavcodec/vp8.c
+++ b/libavcodec/vp8.c
@@ -2010,7 +2010,7 @@ av_cold int ff_vp8_decode_init(AVCodecContext *avctx)
 
     ff_videodsp_init(&s->vdsp, 8);
     ff_h264_pred_init(&s->hpc, AV_CODEC_ID_VP8, 8, 1);
-    ff_vp8dsp_init(&s->vp8dsp);
+    ff_vp8dsp_init(&s->vp8dsp, avctx->codec->id == AV_CODEC_ID_VP7);
 
     if ((ret = vp8_init_frames(s)) < 0) {
         ff_vp8_decode_free(avctx);
diff --git a/libavcodec/vp8dsp.c b/libavcodec/vp8dsp.c
index ba267fd..4bf7f5f 100644
--- a/libavcodec/vp8dsp.c
+++ b/libavcodec/vp8dsp.c
@@ -1,6 +1,7 @@
 /*
  * Copyright (C) 2010 David Conrad
  * Copyright (C) 2010 Ronald S. Bultje
+ * Copyright (C) 2014 Peter Ross
  *
  * This file is part of FFmpeg.
  *
@@ -28,6 +29,92 @@
 #include "vp8dsp.h"
 #include "libavutil/common.h"
 
+static const int16_t vp7_dct_matrix[4][4] =
+{
+    { 23170,  30274,  23170,  12540 },
+    { 23170,  12540, -23170, -30274 },
+    { 23170, -12540, -23170,  30274 },
+    { 23170, -30274,  23170, -12540 }
+};
+
+static void vp7_luma_dc_wht_c(int16_t block[4][4][16], int16_t dc[16])
+{
+    int i, j, k;
+    short temp[4][4];
+    int sum;
+
+    for (i = 0; i < 4; i++) {
+        for (j = 0; j < 4; j++) {
+            sum = 0;
+            for (k = 0; k < 4; k++)
+                sum += dc[i*4 + k] * vp7_dct_matrix[j][k];
+            temp[i][j] = (sum >> 14);
+        }
+    }
+
+    for (i = 0; i < 4; i++) {
+        for (j = 0; j < 4; j++) {
+            sum = 0;
+            for (k = 0; k < 4; k++)
+                sum += vp7_dct_matrix[i][k] * temp[k][j];
+            block[i][j][0] = (sum + 0x20000) >> 18;
+        }
+    }
+}
+
+static void vp7_luma_dc_wht_dc_c(int16_t block[4][4][16], int16_t dc[16])
+{
+    int i, val = (23170 * (23170 * dc[0] >> 14) + 0x20000) >> 18;
+    dc[0] = 0;
+
+    for (i = 0; i < 4; i++) {
+        block[i][0][0] = val;
+        block[i][1][0] = val;
+        block[i][2][0] = val;
+        block[i][3][0] = val;
+    }
+}
+
+static void vp7_idct_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
+{
+    int i, j, k;
+    short temp[4][4];
+    int sum;
+
+    for (i = 0; i < 4; i++) {
+        for (j = 0; j < 4; j++) {
+            sum = 0;
+            for (k = 0; k < 4; k++)
+                sum += block[i*4 + k] * vp7_dct_matrix[j][k];
+            temp[i][j] = (sum >> 14);
+        }
+    }
+
+    for (i = 0; i < 4; i++) {
+        for (j = 0; j < 4; j++) {
+            sum = 0;
+            for (k = 0; k < 4; k++)
+                sum += vp7_dct_matrix[i][k] * temp[k][j];
+            dst[j] = av_clip_uint8(dst[j] + ((sum + 0x20000) >> 18));
+        }
+        dst += stride;
+    }
+}
+
+static void vp7_idct_dc_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
+{
+    int i, dc = (23170 * (23170 * block[0] >> 14) + 0x20000) >> 18;
+    block[0] = 0;
+
+    for (i = 0; i < 4; i++) {
+        dst[0] = av_clip_uint8(dst[0] + dc);
+        dst[1] = av_clip_uint8(dst[1] + dc);
+        dst[2] = av_clip_uint8(dst[2] + dc);
+        dst[3] = av_clip_uint8(dst[3] + dc);
+        dst += stride;
+    }
+}
+
 // TODO: Maybe add dequant
 static void vp8_luma_dc_wht_c(int16_t block[4][4][16], int16_t dc[16])
 {
@@ -127,21 +214,25 @@ static void vp8_idct_dc_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
     }
 }
 
-static void vp8_idct_dc_add4uv_c(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride)
-{
-    vp8_idct_dc_add_c(dst+stride*0+0, block[0], stride);
-    vp8_idct_dc_add_c(dst+stride*0+4, block[1], stride);
-    vp8_idct_dc_add_c(dst+stride*4+0, block[2], stride);
-    vp8_idct_dc_add_c(dst+stride*4+4, block[3], stride);
+#define MK_IDCT_DC_ADD4_C(name) \
+static void name ## _idct_dc_add4uv_c(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride)\
+{\
+    name ## _idct_dc_add_c(dst+stride*0+0, block[0], stride);\
+    name ## _idct_dc_add_c(dst+stride*0+4, block[1], stride);\
+    name ## _idct_dc_add_c(dst+stride*4+0, block[2], stride);\
+    name ## _idct_dc_add_c(dst+stride*4+4, block[3], stride);\
+}\
+\
+static void name ## _idct_dc_add4y_c(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride)\
+{\
+    name ## _idct_dc_add_c(dst+ 0, block[0], stride);\
+    name ## _idct_dc_add_c(dst+ 4, block[1], stride);\
+    name ## _idct_dc_add_c(dst+ 8, block[2], stride);\
+    name ## _idct_dc_add_c(dst+12, block[3], stride);\
 }
 
-static void vp8_idct_dc_add4y_c(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride)
-{
-    vp8_idct_dc_add_c(dst+ 0, block[0], stride);
-    vp8_idct_dc_add_c(dst+ 4, block[1], stride);
-    vp8_idct_dc_add_c(dst+ 8, block[2], stride);
-    vp8_idct_dc_add_c(dst+12, block[3], stride);
-}
+MK_IDCT_DC_ADD4_C(vp7)
+MK_IDCT_DC_ADD4_C(vp8)
 
 // because I like only having two parameters to pass functions...
 #define LOAD_PIXELS\
@@ -156,7 +247,7 @@ static void vp8_idct_dc_add4y_c(uint8_t *dst, int16_t block[4][16], ptrdiff_t st
 
 #define clip_int8(n) (cm[n+0x80]-0x80)
 
-static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride, int is4tap)
+static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride, int is4tap, int vpn)
 {
     LOAD_PIXELS
     int a, f1, f2;
@@ -172,7 +263,11 @@ static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride, int is4
     // We deviate from the spec here with c(a+3) >> 3
     // since that's what libvpx does.
     f1 = FFMIN(a+4, 127) >> 3;
-    f2 = FFMIN(a+3, 127) >> 3;
+
+    if (vpn == 7)
+        f2 = f1 - ((a & 7) == 4);
+    else
+        f2 = FFMIN(a+3, 127) >> 3;
 
     // Despite what the spec says, we do need to clamp here to
     // be bitexact with libvpx.
@@ -187,7 +282,13 @@ static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride, int is4
     }
 }
 
-static av_always_inline int simple_limit(uint8_t *p, ptrdiff_t stride, int flim)
+static av_always_inline int vp7_simple_limit(uint8_t *p, ptrdiff_t stride, int flim)
+{
+    LOAD_PIXELS
+    return FFABS(p0-q0) <= flim;
+}
+
+static av_always_inline int vp8_simple_limit(uint8_t *p, ptrdiff_t stride, int flim)
 {
     LOAD_PIXELS
     return 2*FFABS(p0-q0) + (FFABS(p1-q1) >> 1) <= flim;
@@ -197,14 +298,18 @@ static av_always_inline int simple_limit(uint8_t *p, ptrdiff_t stride, int flim)
  * E - limit at the macroblock edge
  * I - limit for interior difference
  */
-static av_always_inline int normal_limit(uint8_t *p, ptrdiff_t stride, int E, int I)
-{
-    LOAD_PIXELS
-    return simple_limit(p, stride, E)
-        && FFABS(p3-p2) <= I && FFABS(p2-p1) <= I && FFABS(p1-p0) <= I
-        && FFABS(q3-q2) <= I && FFABS(q2-q1) <= I && FFABS(q1-q0) <= I;
+#define NORMAL_LIMIT(vpn) \
+static av_always_inline int vp ## vpn ## _normal_limit(uint8_t *p, ptrdiff_t stride, int E, int I)\
+{\
+    LOAD_PIXELS\
+    return vp ## vpn ## _simple_limit(p, stride, E)\
+        && FFABS(p3-p2) <= I && FFABS(p2-p1) <= I && FFABS(p1-p0) <= I\
+        && FFABS(q3-q2) <= I && FFABS(q2-q1) <= I && FFABS(q1-q0) <= I;\
 }
 
+NORMAL_LIMIT(7)
+NORMAL_LIMIT(8)
+
 // high edge variance
 static av_always_inline int hev(uint8_t *p, ptrdiff_t stride, int thresh)
 {
@@ -234,74 +339,84 @@ static av_always_inline void filter_mbedge(uint8_t *p, ptrdiff_t stride)
     p[ 2*stride] = cm[q2 - a2];
 }
 
-#define LOOP_FILTER(dir, size, stridea, strideb, maybe_inline) \
-static maybe_inline void vp8_ ## dir ## _loop_filter ## size ## _c(uint8_t *dst, ptrdiff_t stride,\
+#define LOOP_FILTER(vpn, dir, size, stridea, strideb, maybe_inline) \
+static maybe_inline void vp ## vpn ## _ ## dir ## _loop_filter ## size ## _c(uint8_t *dst, ptrdiff_t stride,\
                                      int flim_E, int flim_I, int hev_thresh)\
 {\
     int i;\
 \
     for (i = 0; i < size; i++)\
-        if (normal_limit(dst+i*stridea, strideb, flim_E, flim_I)) {\
+        if (vp ## vpn ## _normal_limit(dst+i*stridea, strideb, flim_E, flim_I)) {\
             if (hev(dst+i*stridea, strideb, hev_thresh))\
-                filter_common(dst+i*stridea, strideb, 1);\
+                filter_common(dst+i*stridea, strideb, 1, vpn);\
             else\
                 filter_mbedge(dst+i*stridea, strideb);\
         }\
 }\
 \
-static maybe_inline void vp8_ ## dir ## _loop_filter ## size ## _inner_c(uint8_t *dst, ptrdiff_t stride,\
+static maybe_inline void vp ## vpn ## _ ## dir ## _loop_filter ## size ## _inner_c(uint8_t *dst, ptrdiff_t stride,\
                                       int flim_E, int flim_I, int hev_thresh)\
 {\
     int i;\
 \
     for (i = 0; i < size; i++)\
-        if (normal_limit(dst+i*stridea, strideb, flim_E, flim_I)) {\
+        if (vp ## vpn ## _normal_limit(dst+i*stridea, strideb, flim_E, flim_I)) {\
             int hv = hev(dst+i*stridea, strideb, hev_thresh);\
             if (hv) \
-                filter_common(dst+i*stridea, strideb, 1);\
+                filter_common(dst+i*stridea, strideb, 1, vpn);\
             else \
-                filter_common(dst+i*stridea, strideb, 0);\
+                filter_common(dst+i*stridea, strideb, 0, vpn);\
         }\
 }
 
-LOOP_FILTER(v, 16, 1, stride,)
-LOOP_FILTER(h, 16, stride, 1,)
+LOOP_FILTER(7, v, 16, 1, stride,)
+LOOP_FILTER(7, h, 16, stride, 1,)
+
+LOOP_FILTER(8, v, 16, 1, stride,)
+LOOP_FILTER(8, h, 16, stride, 1,)
 
-#define UV_LOOP_FILTER(dir, stridea, strideb) \
-LOOP_FILTER(dir, 8, stridea, strideb, av_always_inline) \
-static void vp8_ ## dir ## _loop_filter8uv_c(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,\
+#define UV_LOOP_FILTER(vpn, dir, stridea, strideb) \
+LOOP_FILTER(vpn, dir, 8, stridea, strideb, av_always_inline) \
+static void vp ## vpn ## _ ## dir ## _loop_filter8uv_c(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,\
                                       int fE, int fI, int hev_thresh)\
 {\
-  vp8_ ## dir ## _loop_filter8_c(dstU, stride, fE, fI, hev_thresh);\
-  vp8_ ## dir ## _loop_filter8_c(dstV, stride, fE, fI, hev_thresh);\
+  vp ## vpn ## _ ## dir ## _loop_filter8_c(dstU, stride, fE, fI, hev_thresh);\
+  vp ## vpn ## _ ## dir ## _loop_filter8_c(dstV, stride, fE, fI, hev_thresh);\
 }\
-static void vp8_ ## dir ## _loop_filter8uv_inner_c(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,\
+static void vp ## vpn ## _ ## dir ## _loop_filter8uv_inner_c(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,\
                                       int fE, int fI, int hev_thresh)\
 {\
-  vp8_ ## dir ## _loop_filter8_inner_c(dstU, stride, fE, fI, hev_thresh);\
-  vp8_ ## dir ## _loop_filter8_inner_c(dstV, stride, fE, fI, hev_thresh);\
+  vp ## vpn ## _ ## dir ## _loop_filter8_inner_c(dstU, stride, fE, fI, hev_thresh);\
+  vp ## vpn ## _ ## dir ## _loop_filter8_inner_c(dstV, stride, fE, fI, hev_thresh);\
 }
 
-UV_LOOP_FILTER(v, 1, stride)
-UV_LOOP_FILTER(h, stride, 1)
+UV_LOOP_FILTER(7, v, 1, stride)
+UV_LOOP_FILTER(7, h, stride, 1)
 
-static void vp8_v_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim)
-{
-    int i;
+UV_LOOP_FILTER(8, v, 1, stride)
+UV_LOOP_FILTER(8, h, stride, 1)
 
-    for (i = 0; i < 16; i++)
-        if (simple_limit(dst+i, stride, flim))
-            filter_common(dst+i, stride, 1);
+#define LOOP_FILTER_SIMPLE(vpn) \
+static void vp ## vpn ## _v_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim)\
+{\
+    int i;\
+\
+    for (i = 0; i < 16; i++)\
+        if (vp ## vpn ## _simple_limit(dst+i, stride, flim))\
+            filter_common(dst+i, stride, 1, vpn);\
+}\
+\
+static void vp ## vpn ## _h_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim)\
+{\
+    int i;\
+\
+    for (i = 0; i < 16; i++)\
+        if (vp ## vpn ## _simple_limit(dst+i*stride, 1, flim))\
+            filter_common(dst+i*stride, 1, 1, vpn);\
 }
 
-static void vp8_h_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim)
-{
-    int i;
-
-    for (i = 0; i < 16; i++)
-        if (simple_limit(dst+i*stride, 1, flim))
-            filter_common(dst+i*stride, 1, 1);
-}
+LOOP_FILTER_SIMPLE(7)
+LOOP_FILTER_SIMPLE(8)
 
 static const uint8_t subpel_filters[7][6] = {
     { 0,   6, 123,  12,   1,   0 },
@@ -491,27 +606,29 @@ VP8_BILINEAR(4)
     dsp->put_vp8_bilinear_pixels_tab[IDX][2][1] = put_vp8_bilinear ## SIZE ## _hv_c; \
     dsp->put_vp8_bilinear_pixels_tab[IDX][2][2] = put_vp8_bilinear ## SIZE ## _hv_c
 
-av_cold void ff_vp8dsp_init(VP8DSPContext *dsp)
+av_cold void ff_vp8dsp_init(VP8DSPContext *dsp, int vp7)
 {
-    dsp->vp8_luma_dc_wht    = vp8_luma_dc_wht_c;
-    dsp->vp8_luma_dc_wht_dc = vp8_luma_dc_wht_dc_c;
-    dsp->vp8_idct_add       = vp8_idct_add_c;
-    dsp->vp8_idct_dc_add    = vp8_idct_dc_add_c;
-    dsp->vp8_idct_dc_add4y  = vp8_idct_dc_add4y_c;
-    dsp->vp8_idct_dc_add4uv = vp8_idct_dc_add4uv_c;
-
-    dsp->vp8_v_loop_filter16y = vp8_v_loop_filter16_c;
-    dsp->vp8_h_loop_filter16y = vp8_h_loop_filter16_c;
-    dsp->vp8_v_loop_filter8uv = vp8_v_loop_filter8uv_c;
-    dsp->vp8_h_loop_filter8uv = vp8_h_loop_filter8uv_c;
-
-    dsp->vp8_v_loop_filter16y_inner = vp8_v_loop_filter16_inner_c;
-    dsp->vp8_h_loop_filter16y_inner = vp8_h_loop_filter16_inner_c;
-    dsp->vp8_v_loop_filter8uv_inner = vp8_v_loop_filter8uv_inner_c;
-    dsp->vp8_h_loop_filter8uv_inner = vp8_h_loop_filter8uv_inner_c;
-
-    dsp->vp8_v_loop_filter_simple = vp8_v_loop_filter_simple_c;
-    dsp->vp8_h_loop_filter_simple = vp8_h_loop_filter_simple_c;
+#define VPX(f) vp7 ? vp7 ## _ ## f : vp8 ## _ ## f
+
+    dsp->vp8_luma_dc_wht    = VPX(luma_dc_wht_c);
+    dsp->vp8_luma_dc_wht_dc = VPX(luma_dc_wht_dc_c);
+    dsp->vp8_idct_add       = VPX(idct_add_c);
+    dsp->vp8_idct_dc_add    = VPX(idct_dc_add_c);
+    dsp->vp8_idct_dc_add4y  = VPX(idct_dc_add4y_c);
+    dsp->vp8_idct_dc_add4uv = VPX(idct_dc_add4uv_c);
+
+    dsp->vp8_v_loop_filter16y = VPX(v_loop_filter16_c);
+    dsp->vp8_h_loop_filter16y = VPX(h_loop_filter16_c);
+    dsp->vp8_v_loop_filter8uv = VPX(v_loop_filter8uv_c);
+    dsp->vp8_h_loop_filter8uv = VPX(h_loop_filter8uv_c);
+
+    dsp->vp8_v_loop_filter16y_inner = VPX(v_loop_filter16_inner_c);
+    dsp->vp8_h_loop_filter16y_inner = VPX(h_loop_filter16_inner_c);
+    dsp->vp8_v_loop_filter8uv_inner = VPX(v_loop_filter8uv_inner_c);
+    dsp->vp8_h_loop_filter8uv_inner = VPX(h_loop_filter8uv_inner_c);
+
+    dsp->vp8_v_loop_filter_simple = VPX(v_loop_filter_simple_c);
+    dsp->vp8_h_loop_filter_simple = VPX(h_loop_filter_simple_c);
 
     VP8_MC_FUNC(0, 16);
     VP8_MC_FUNC(1, 8);
@@ -522,9 +639,9 @@ av_cold void ff_vp8dsp_init(VP8DSPContext *dsp)
     VP8_BILINEAR_MC_FUNC(2, 4);
 
     if (ARCH_ARM)
-        ff_vp8dsp_init_arm(dsp);
+        ff_vp8dsp_init_arm(dsp, vp7);
     if (ARCH_PPC)
         ff_vp8dsp_init_ppc(dsp);
     if (ARCH_X86)
-        ff_vp8dsp_init_x86(dsp);
+        ff_vp8dsp_init_x86(dsp, vp7);
 }
diff --git a/libavcodec/vp8dsp.h b/libavcodec/vp8dsp.h
index 58ec65a..02ac778 100644
--- a/libavcodec/vp8dsp.h
+++ b/libavcodec/vp8dsp.h
@@ -88,9 +88,9 @@ void ff_put_vp8_pixels8_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
 void ff_put_vp8_pixels4_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
                           int h, int x, int y);
 
-void ff_vp8dsp_init(VP8DSPContext *c);
-void ff_vp8dsp_init_x86(VP8DSPContext *c);
-void ff_vp8dsp_init_arm(VP8DSPContext *c);
+void ff_vp8dsp_init(VP8DSPContext *c, int vp7);
+void ff_vp8dsp_init_x86(VP8DSPContext *c, int vp7);
+void ff_vp8dsp_init_arm(VP8DSPContext *c, int vp7);
 void ff_vp8dsp_init_ppc(VP8DSPContext *c);
 
 #endif /* AVCODEC_VP8DSP_H */
diff --git a/libavcodec/x86/vp8dsp_init.c b/libavcodec/x86/vp8dsp_init.c
index 982e17d..d9ce977 100644
--- a/libavcodec/x86/vp8dsp_init.c
+++ b/libavcodec/x86/vp8dsp_init.c
@@ -315,18 +315,22 @@ DECLARE_LOOP_FILTER(sse4)
     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)
+av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c, int vp7)
 {
 #if HAVE_YASM
     int cpu_flags = av_get_cpu_flags();
 
     if (EXTERNAL_MMX(cpu_flags)) {
+        if (!vp7) {
         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
+        if (!vp7) {
         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
@@ -334,6 +338,7 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
         c->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_mmx;
 
 #if ARCH_X86_32
+        if (!vp7) {
         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;
 
@@ -346,6 +351,7 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
         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
     }
 
@@ -360,6 +366,7 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
         VP8_BILINEAR_MC_FUNC(0, 16, mmxext);
         VP8_BILINEAR_MC_FUNC(1,  8, mmxext);
 
+        if (!vp7) {
         c->vp8_v_loop_filter_simple   = ff_vp8_v_loop_filter_simple_mmxext;
         c->vp8_h_loop_filter_simple   = ff_vp8_h_loop_filter_simple_mmxext;
 
@@ -372,12 +379,15 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
         c->vp8_h_loop_filter16y       = ff_vp8_h_loop_filter16y_mbedge_mmxext;
         c->vp8_v_loop_filter8uv       = ff_vp8_v_loop_filter8uv_mbedge_mmxext;
         c->vp8_h_loop_filter8uv       = ff_vp8_h_loop_filter8uv_mbedge_mmxext;
+        }
 #endif
     }
 
     if (EXTERNAL_SSE(cpu_flags)) {
+        if (!vp7) {
         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;
     }
@@ -388,6 +398,7 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
         VP8_BILINEAR_MC_FUNC(0, 16, sse2);
         VP8_BILINEAR_MC_FUNC(1, 8, sse2);
 
+        if (!vp7) {
         c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_sse2;
 
         c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_sse2;
@@ -395,9 +406,11 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
 
         c->vp8_v_loop_filter16y       = ff_vp8_v_loop_filter16y_mbedge_sse2;
         c->vp8_v_loop_filter8uv       = ff_vp8_v_loop_filter8uv_mbedge_sse2;
+        }
     }
 
     if (EXTERNAL_SSE2(cpu_flags)) {
+        if (!vp7) {
         c->vp8_idct_dc_add4y          = ff_vp8_idct_dc_add4y_sse2;
 
         c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_sse2;
@@ -407,6 +420,7 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
 
         c->vp8_h_loop_filter16y       = ff_vp8_h_loop_filter16y_mbedge_sse2;
         c->vp8_h_loop_filter8uv       = ff_vp8_h_loop_filter8uv_mbedge_sse2;
+        }
     }
 
     if (EXTERNAL_SSSE3(cpu_flags)) {
@@ -417,6 +431,7 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
         VP8_BILINEAR_MC_FUNC(1, 8, ssse3);
         VP8_BILINEAR_MC_FUNC(2, 4, ssse3);
 
+        if (!vp7) {
         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;
 
@@ -429,14 +444,17 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
         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;
+        }
     }
 
     if (EXTERNAL_SSE4(cpu_flags)) {
+        if (!vp7) {
         c->vp8_idct_dc_add                  = ff_vp8_idct_dc_add_sse4;
 
         c->vp8_h_loop_filter_simple   = ff_vp8_h_loop_filter_simple_sse4;
         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 /* HAVE_YASM */
 }
-- 
1.8.3.2

-- Peter
(A907 E02F A6E5 0CD2 34CD 20D2 6760 79C5 AC40 DD6B)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20140207/600efa54/attachment.asc>


More information about the ffmpeg-devel mailing list