Commit 721d57e6 authored by Diego Biurrun's avatar Diego Biurrun

vp56: Separate VP5 and VP6 dsp initialization

VP5 has no arch-specific optimizations (nor will it get some in the
future), so it makes no sense to try to share dsp init code with VP6.
parent 3fd22538
...@@ -22,13 +22,13 @@ ...@@ -22,13 +22,13 @@
#include "libavutil/attributes.h" #include "libavutil/attributes.h"
#include "libavutil/arm/cpu.h" #include "libavutil/arm/cpu.h"
#include "libavcodec/avcodec.h"
#include "libavcodec/vp56dsp.h" #include "libavcodec/vp56dsp.h"
void ff_vp6_edge_filter_hor_neon(uint8_t *yuv, ptrdiff_t stride, int t); void ff_vp6_edge_filter_hor_neon(uint8_t *yuv, ptrdiff_t stride, int t);
void ff_vp6_edge_filter_ver_neon(uint8_t *yuv, ptrdiff_t stride, int t); void ff_vp6_edge_filter_ver_neon(uint8_t *yuv, ptrdiff_t stride, int t);
av_cold void ff_vp6dsp_init_arm(VP56DSPContext *s, enum AVCodecID codec) av_cold void ff_vp6dsp_init_arm(VP56DSPContext *s)
{ {
int cpu_flags = av_get_cpu_flags(); int cpu_flags = av_get_cpu_flags();
......
...@@ -270,6 +270,7 @@ static av_cold int vp5_decode_init(AVCodecContext *avctx) ...@@ -270,6 +270,7 @@ static av_cold int vp5_decode_init(AVCodecContext *avctx)
if ((ret = ff_vp56_init(avctx, 1, 0)) < 0) if ((ret = ff_vp56_init(avctx, 1, 0)) < 0)
return ret; return ret;
ff_vp5dsp_init(&s->vp56dsp);
s->vp56_coord_div = vp5_coord_div; s->vp56_coord_div = vp5_coord_div;
s->parse_vector_adjustment = vp5_parse_vector_adjustment; s->parse_vector_adjustment = vp5_parse_vector_adjustment;
s->parse_coeff = vp5_parse_coeff; s->parse_coeff = vp5_parse_coeff;
......
...@@ -663,7 +663,6 @@ av_cold int ff_vp56_init(AVCodecContext *avctx, int flip, int has_alpha) ...@@ -663,7 +663,6 @@ av_cold int ff_vp56_init(AVCodecContext *avctx, int flip, int has_alpha)
ff_hpeldsp_init(&s->hdsp, avctx->flags); ff_hpeldsp_init(&s->hdsp, avctx->flags);
ff_videodsp_init(&s->vdsp, 8); ff_videodsp_init(&s->vdsp, 8);
ff_vp3dsp_init(&s->vp3dsp, avctx->flags); ff_vp3dsp_init(&s->vp3dsp, avctx->flags);
ff_vp56dsp_init(&s->vp56dsp, avctx->codec->id);
for (i = 0; i < 64; i++) { for (i = 0; i < 64; i++) {
#define TRANSPOSE(x) (x >> 3) | ((x & 7) << 3) #define TRANSPOSE(x) (x >> 3) | ((x & 7) << 3)
s->idct_scantable[i] = TRANSPOSE(ff_zigzag_direct[i]); s->idct_scantable[i] = TRANSPOSE(ff_zigzag_direct[i]);
......
...@@ -26,6 +26,7 @@ ...@@ -26,6 +26,7 @@
#ifndef AVCODEC_VP56_H #ifndef AVCODEC_VP56_H
#define AVCODEC_VP56_H #define AVCODEC_VP56_H
#include "avcodec.h"
#include "get_bits.h" #include "get_bits.h"
#include "hpeldsp.h" #include "hpeldsp.h"
#include "bytestream.h" #include "bytestream.h"
......
...@@ -26,6 +26,23 @@ ...@@ -26,6 +26,23 @@
#include "vp56dsp.h" #include "vp56dsp.h"
#include "libavutil/common.h" #include "libavutil/common.h"
#define VP56_EDGE_FILTER(pfx, suf, pix_inc, line_inc) \
static void pfx ## _edge_filter_ ## suf(uint8_t *yuv, ptrdiff_t stride, \
int t) \
{ \
int pix2_inc = 2 * pix_inc; \
int i, v; \
\
for (i=0; i<12; i++) { \
v = (yuv[-pix2_inc] + 3*(yuv[0]-yuv[-pix_inc]) - yuv[pix_inc] + 4)>>3;\
v = pfx##_adjust(v, t); \
yuv[-pix_inc] = av_clip_uint8(yuv[-pix_inc] + v); \
yuv[0] = av_clip_uint8(yuv[0] - v); \
yuv += line_inc; \
} \
}
#if CONFIG_VP5_DECODER
/* Gives very similar result than the vp6 version except in a few cases */ /* Gives very similar result than the vp6 version except in a few cases */
static int vp5_adjust(int v, int t) static int vp5_adjust(int v, int t)
{ {
...@@ -43,6 +60,18 @@ static int vp5_adjust(int v, int t) ...@@ -43,6 +60,18 @@ static int vp5_adjust(int v, int t)
return v; return v;
} }
VP56_EDGE_FILTER(vp5, hor, 1, stride)
VP56_EDGE_FILTER(vp5, ver, stride, 1)
av_cold void ff_vp5dsp_init(VP56DSPContext *s)
{
s->edge_filter_hor = vp5_edge_filter_hor;
s->edge_filter_ver = vp5_edge_filter_ver;
}
#endif /* CONFIG_VP5_DECODER */
#if CONFIG_VP6_DECODER
static int vp6_adjust(int v, int t) static int vp6_adjust(int v, int t)
{ {
int V = v, s = v >> 31; int V = v, s = v >> 31;
...@@ -56,44 +85,19 @@ static int vp6_adjust(int v, int t) ...@@ -56,44 +85,19 @@ static int vp6_adjust(int v, int t)
return V; return V;
} }
#define VP56_EDGE_FILTER(pfx, suf, pix_inc, line_inc) \
static void pfx ## _edge_filter_ ## suf(uint8_t *yuv, ptrdiff_t stride, \
int t) \
{ \
int pix2_inc = 2 * pix_inc; \
int i, v; \
\
for (i=0; i<12; i++) { \
v = (yuv[-pix2_inc] + 3*(yuv[0]-yuv[-pix_inc]) - yuv[pix_inc] + 4)>>3;\
v = pfx##_adjust(v, t); \
yuv[-pix_inc] = av_clip_uint8(yuv[-pix_inc] + v); \
yuv[0] = av_clip_uint8(yuv[0] - v); \
yuv += line_inc; \
} \
}
VP56_EDGE_FILTER(vp5, hor, 1, stride)
VP56_EDGE_FILTER(vp5, ver, stride, 1)
VP56_EDGE_FILTER(vp6, hor, 1, stride) VP56_EDGE_FILTER(vp6, hor, 1, stride)
VP56_EDGE_FILTER(vp6, ver, stride, 1) VP56_EDGE_FILTER(vp6, ver, stride, 1)
av_cold void ff_vp56dsp_init(VP56DSPContext *s, enum AVCodecID codec) av_cold void ff_vp6dsp_init(VP56DSPContext *s)
{ {
if (codec == AV_CODEC_ID_VP5) { s->edge_filter_hor = vp6_edge_filter_hor;
s->edge_filter_hor = vp5_edge_filter_hor; s->edge_filter_ver = vp6_edge_filter_ver;
s->edge_filter_ver = vp5_edge_filter_ver;
} else {
s->edge_filter_hor = vp6_edge_filter_hor;
s->edge_filter_ver = vp6_edge_filter_ver;
if (CONFIG_VP6_DECODER) { s->vp6_filter_diag4 = ff_vp6_filter_diag4_c;
s->vp6_filter_diag4 = ff_vp6_filter_diag4_c;
if (ARCH_ARM) if (ARCH_ARM)
ff_vp6dsp_init_arm(s, codec); ff_vp6dsp_init_arm(s);
if (ARCH_X86) if (ARCH_X86)
ff_vp6dsp_init_x86(s, codec); ff_vp6dsp_init_x86(s);
}
}
} }
#endif /* CONFIG_VP6_DECODER */
...@@ -24,8 +24,6 @@ ...@@ -24,8 +24,6 @@
#include <stddef.h> #include <stddef.h>
#include <stdint.h> #include <stdint.h>
#include "avcodec.h"
typedef struct VP56DSPContext { typedef struct VP56DSPContext {
void (*edge_filter_hor)(uint8_t *yuv, ptrdiff_t stride, int t); void (*edge_filter_hor)(uint8_t *yuv, ptrdiff_t stride, int t);
void (*edge_filter_ver)(uint8_t *yuv, ptrdiff_t stride, int t); void (*edge_filter_ver)(uint8_t *yuv, ptrdiff_t stride, int t);
...@@ -37,8 +35,10 @@ typedef struct VP56DSPContext { ...@@ -37,8 +35,10 @@ typedef struct VP56DSPContext {
void ff_vp6_filter_diag4_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride, void ff_vp6_filter_diag4_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
const int16_t *h_weights, const int16_t *v_weights); const int16_t *h_weights, const int16_t *v_weights);
void ff_vp56dsp_init(VP56DSPContext *s, enum AVCodecID codec); void ff_vp5dsp_init(VP56DSPContext *s);
void ff_vp6dsp_init_arm(VP56DSPContext *s, enum AVCodecID codec); void ff_vp6dsp_init(VP56DSPContext *s);
void ff_vp6dsp_init_x86(VP56DSPContext* c, enum AVCodecID codec);
void ff_vp6dsp_init_arm(VP56DSPContext *s);
void ff_vp6dsp_init_x86(VP56DSPContext *s);
#endif /* AVCODEC_VP56DSP_H */ #endif /* AVCODEC_VP56DSP_H */
...@@ -612,6 +612,7 @@ static av_cold int vp6_decode_init(AVCodecContext *avctx) ...@@ -612,6 +612,7 @@ static av_cold int vp6_decode_init(AVCodecContext *avctx)
if ((ret = ff_vp56_init(avctx, avctx->codec->id == AV_CODEC_ID_VP6, if ((ret = ff_vp56_init(avctx, avctx->codec->id == AV_CODEC_ID_VP6,
avctx->codec->id == AV_CODEC_ID_VP6A)) < 0) avctx->codec->id == AV_CODEC_ID_VP6A)) < 0)
return ret; return ret;
ff_vp6dsp_init(&s->vp56dsp);
s->vp56_coord_div = vp6_coord_div; s->vp56_coord_div = vp6_coord_div;
s->parse_vector_adjustment = vp6_parse_vector_adjustment; s->parse_vector_adjustment = vp6_parse_vector_adjustment;
......
...@@ -30,7 +30,7 @@ void ff_vp6_filter_diag4_mmx(uint8_t *dst, uint8_t *src, ptrdiff_t stride, ...@@ -30,7 +30,7 @@ void ff_vp6_filter_diag4_mmx(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
void ff_vp6_filter_diag4_sse2(uint8_t *dst, uint8_t *src, ptrdiff_t stride, void ff_vp6_filter_diag4_sse2(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
const int16_t *h_weights,const int16_t *v_weights); const int16_t *h_weights,const int16_t *v_weights);
av_cold void ff_vp6dsp_init_x86(VP56DSPContext* c, enum AVCodecID codec) av_cold void ff_vp6dsp_init_x86(VP56DSPContext *c)
{ {
int cpu_flags = av_get_cpu_flags(); int cpu_flags = av_get_cpu_flags();
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment