On2 VP7 decoder

Further performance improvements and security fixes by
Vittorio Giovara, Luca Barbato and Diego Biurrun.

Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
Signed-off-by: Diego Biurrun <diego@biurrun.de>

3 files changed, 59 insertions, 32 deletions

diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile
index eb312a2dd1..910471aeb5 100644
--- a/libavcodec/x86/Makefile
+++ b/libavcodec/x86/Makefile
@@ -37,6 +37,7 @@ OBJS-$(CONFIG_TRUEHD_DECODER)          += x86/mlpdsp.o
 OBJS-$(CONFIG_VC1_DECODER)             += x86/vc1dsp_init.o
 OBJS-$(CONFIG_VORBIS_DECODER)          += x86/vorbisdsp_init.o
 OBJS-$(CONFIG_VP6_DECODER)             += x86/vp6dsp_init.o
+OBJS-$(CONFIG_VP7_DECODER)             += x86/vp8dsp_init.o
 OBJS-$(CONFIG_VP8_DECODER)             += x86/vp8dsp_init.o
 OBJS-$(CONFIG_VP9_DECODER)             += x86/vp9dsp_init.o
 
@@ -92,6 +93,8 @@ YASM-OBJS-$(CONFIG_RV40_DECODER)       += x86/rv34dsp.o                 \
 YASM-OBJS-$(CONFIG_VC1_DECODER)        += x86/vc1dsp.o
 YASM-OBJS-$(CONFIG_VORBIS_DECODER)     += x86/vorbisdsp.o
 YASM-OBJS-$(CONFIG_VP6_DECODER)        += x86/vp6dsp.o
+YASM-OBJS-$(CONFIG_VP7_DECODER)        += x86/vp8dsp.o                  \
+                                          x86/vp8dsp_loopfilter.o
 YASM-OBJS-$(CONFIG_VP8_DECODER)        += x86/vp8dsp.o                  \
                                           x86/vp8dsp_loopfilter.o
 YASM-OBJS-$(CONFIG_VP9_DECODER)        += x86/vp9dsp.o
diff --git a/libavcodec/x86/h264_intrapred_init.c b/libavcodec/x86/h264_intrapred_init.c
index f934256706..6bd33ed388 100644
--- a/libavcodec/x86/h264_intrapred_init.c
+++ b/libavcodec/x86/h264_intrapred_init.c
@@ -195,7 +195,7 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id,
                 h->pred8x8  [VERT_PRED8x8     ] = ff_pred8x8_vertical_8_mmx;
                 h->pred8x8  [HOR_PRED8x8      ] = ff_pred8x8_horizontal_8_mmx;
             }
-            if (codec_id == AV_CODEC_ID_VP8) {
+            if (codec_id == AV_CODEC_ID_VP7 || codec_id == AV_CODEC_ID_VP8) {
                 h->pred16x16[PLANE_PRED8x8    ] = ff_pred16x16_tm_vp8_8_mmx;
                 h->pred8x8  [PLANE_PRED8x8    ] = ff_pred8x8_tm_vp8_8_mmx;
                 h->pred4x4  [TM_VP8_PRED      ] = ff_pred4x4_tm_vp8_8_mmx;
@@ -231,7 +231,8 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id,
             h->pred4x4  [VERT_RIGHT_PRED        ] = ff_pred4x4_vertical_right_8_mmxext;
             h->pred4x4  [HOR_DOWN_PRED          ] = ff_pred4x4_horizontal_down_8_mmxext;
             h->pred4x4  [DC_PRED                ] = ff_pred4x4_dc_8_mmxext;
-            if (codec_id == AV_CODEC_ID_VP8 || codec_id == AV_CODEC_ID_H264) {
+            if (codec_id == AV_CODEC_ID_VP7 || codec_id == AV_CODEC_ID_VP8 ||
+                codec_id == AV_CODEC_ID_H264) {
                 h->pred4x4  [DIAG_DOWN_LEFT_PRED] = ff_pred4x4_down_left_8_mmxext;
             }
             if (codec_id == AV_CODEC_ID_SVQ3 || codec_id == AV_CODEC_ID_H264) {
@@ -246,7 +247,7 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id,
                     h->pred8x8[DC_PRED8x8       ] = ff_pred8x8_dc_8_mmxext;
                 }
             }
-            if (codec_id == AV_CODEC_ID_VP8) {
+            if (codec_id == AV_CODEC_ID_VP7 || codec_id == AV_CODEC_ID_VP8) {
                 h->pred16x16[PLANE_PRED8x8      ] = ff_pred16x16_tm_vp8_8_mmxext;
                 h->pred8x8  [DC_PRED8x8         ] = ff_pred8x8_dc_rv40_8_mmxext;
                 h->pred8x8  [PLANE_PRED8x8      ] = ff_pred8x8_tm_vp8_8_mmxext;
@@ -276,7 +277,7 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id,
             h->pred8x8l [VERT_RIGHT_PRED      ] = ff_pred8x8l_vertical_right_8_sse2;
             h->pred8x8l [VERT_LEFT_PRED       ] = ff_pred8x8l_vertical_left_8_sse2;
             h->pred8x8l [HOR_DOWN_PRED        ] = ff_pred8x8l_horizontal_down_8_sse2;
-            if (codec_id == AV_CODEC_ID_VP8) {
+            if (codec_id == AV_CODEC_ID_VP7 || codec_id == AV_CODEC_ID_VP8) {
                 h->pred16x16[PLANE_PRED8x8    ] = ff_pred16x16_tm_vp8_8_sse2;
                 h->pred8x8  [PLANE_PRED8x8    ] = ff_pred8x8_tm_vp8_8_sse2;
             } else {
@@ -307,7 +308,7 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id,
             h->pred8x8l [VERT_LEFT_PRED       ] = ff_pred8x8l_vertical_left_8_ssse3;
             h->pred8x8l [HOR_UP_PRED          ] = ff_pred8x8l_horizontal_up_8_ssse3;
             h->pred8x8l [HOR_DOWN_PRED        ] = ff_pred8x8l_horizontal_down_8_ssse3;
-            if (codec_id == AV_CODEC_ID_VP8) {
+            if (codec_id == AV_CODEC_ID_VP7 || codec_id == AV_CODEC_ID_VP8) {
                 h->pred8x8  [PLANE_PRED8x8    ] = ff_pred8x8_tm_vp8_8_ssse3;
                 h->pred4x4  [TM_VP8_PRED      ] = ff_pred4x4_tm_vp8_8_ssse3;
             } else {
diff --git a/libavcodec/x86/vp8dsp_init.c b/libavcodec/x86/vp8dsp_init.c
index 69460aa73f..b3b32b7314 100644
--- a/libavcodec/x86/vp8dsp_init.c
+++ b/libavcodec/x86/vp8dsp_init.c
@@ -315,25 +315,69 @@ 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_vp78dsp_init_x86(VP8DSPContext *c)
 {
 #if HAVE_YASM
     int cpu_flags = av_get_cpu_flags();
 
     if (EXTERNAL_MMX(cpu_flags)) {
-        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;
+    }
+
+    /* 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 (EXTERNAL_MMXEXT(cpu_flags)) {
+        VP8_MC_FUNC(2, 4, mmxext);
+        VP8_BILINEAR_MC_FUNC(2, 4, mmxext);
+#if ARCH_X86_32
+        VP8_LUMA_MC_FUNC(0, 16, mmxext);
+        VP8_MC_FUNC(1, 8, mmxext);
+        VP8_BILINEAR_MC_FUNC(0, 16, mmxext);
+        VP8_BILINEAR_MC_FUNC(1,  8, mmxext);
+#endif
+    }
+
+    if (EXTERNAL_SSE(cpu_flags)) {
+        c->put_vp8_epel_pixels_tab[0][0][0]     =
+        c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_sse;
+    }
+
+    if (EXTERNAL_SSE2(cpu_flags) && (cpu_flags & 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);
+    }
+
+    if (EXTERNAL_SSSE3(cpu_flags)) {
+        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);
+    }
+#endif /* HAVE_YASM */
+}
 
+av_cold void ff_vp8dsp_init_x86(VP8DSPContext *c)
+{
+#if HAVE_YASM
+    int cpu_flags = av_get_cpu_flags();
+
+    if (EXTERNAL_MMX(cpu_flags)) {
+        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->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;
 
@@ -352,14 +396,7 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
     /* 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 (EXTERNAL_MMXEXT(cpu_flags)) {
-        VP8_MC_FUNC(2, 4, mmxext);
-        VP8_BILINEAR_MC_FUNC(2, 4, mmxext);
 #if ARCH_X86_32
-        VP8_LUMA_MC_FUNC(0, 16, mmxext);
-        VP8_MC_FUNC(1, 8, mmxext);
-        VP8_BILINEAR_MC_FUNC(0, 16, mmxext);
-        VP8_BILINEAR_MC_FUNC(1,  8, mmxext);
-
         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;
 
@@ -378,16 +415,9 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
     if (EXTERNAL_SSE(cpu_flags)) {
         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 (EXTERNAL_SSE2(cpu_flags) && (cpu_flags & 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;
 
         c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_sse2;
@@ -410,13 +440,6 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
     }
 
     if (EXTERNAL_SSSE3(cpu_flags)) {
-        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;