Commit d0e132ba authored by Clément Bœsch's avatar Clément Bœsch

Merge commit '1bd890ad'

* commit '1bd890ad':
  hevc: Separate adding residual to prediction from IDCT

This commit should be a noop but isn't because of the following renames:

- transform_add  → add_residual
- transform_skip → dequant
- idct_4x4_luma  → transform_4x4_luma
Merged-by: 's avatarClément Bœsch <cboesch@gopro.com>
parents c778a965 1bd890ad
...@@ -97,7 +97,7 @@ function ff_hevc_idct_32x32_dc_neon_8, export=1 ...@@ -97,7 +97,7 @@ function ff_hevc_idct_32x32_dc_neon_8, export=1
bx lr bx lr
endfunc endfunc
function ff_hevc_transform_add_4x4_neon_8, export=1 function ff_hevc_add_residual_4x4_neon_8, export=1
vldm r1, {q0-q1} vldm r1, {q0-q1}
vld1.32 d4[0], [r0], r2 vld1.32 d4[0], [r0], r2
vld1.32 d4[1], [r0], r2 vld1.32 d4[1], [r0], r2
...@@ -117,7 +117,7 @@ function ff_hevc_transform_add_4x4_neon_8, export=1 ...@@ -117,7 +117,7 @@ function ff_hevc_transform_add_4x4_neon_8, export=1
bx lr bx lr
endfunc endfunc
function ff_hevc_transform_add_8x8_neon_8, export=1 function ff_hevc_add_residual_8x8_neon_8, export=1
mov r3, #8 mov r3, #8
1: subs r3, #1 1: subs r3, #1
vld1.16 {q0}, [r1]! vld1.16 {q0}, [r1]!
...@@ -130,7 +130,7 @@ function ff_hevc_transform_add_8x8_neon_8, export=1 ...@@ -130,7 +130,7 @@ function ff_hevc_transform_add_8x8_neon_8, export=1
bx lr bx lr
endfunc endfunc
function ff_hevc_transform_add_16x16_neon_8, export=1 function ff_hevc_add_residual_16x16_neon_8, export=1
mov r3, #16 mov r3, #16
1: subs r3, #1 1: subs r3, #1
vld1.16 {q0, q1}, [r1]! vld1.16 {q0, q1}, [r1]!
...@@ -146,7 +146,7 @@ function ff_hevc_transform_add_16x16_neon_8, export=1 ...@@ -146,7 +146,7 @@ function ff_hevc_transform_add_16x16_neon_8, export=1
bx lr bx lr
endfunc endfunc
function ff_hevc_transform_add_32x32_neon_8, export=1 function ff_hevc_add_residual_32x32_neon_8, export=1
mov r3, #32 mov r3, #32
1: subs r3, #1 1: subs r3, #1
vldm r1!, {q0-q3} vldm r1!, {q0-q3}
......
...@@ -34,14 +34,14 @@ void ff_hevc_idct_8x8_dc_neon_8(int16_t *coeffs); ...@@ -34,14 +34,14 @@ void ff_hevc_idct_8x8_dc_neon_8(int16_t *coeffs);
void ff_hevc_idct_16x16_dc_neon_8(int16_t *coeffs); void ff_hevc_idct_16x16_dc_neon_8(int16_t *coeffs);
void ff_hevc_idct_32x32_dc_neon_8(int16_t *coeffs); void ff_hevc_idct_32x32_dc_neon_8(int16_t *coeffs);
void ff_hevc_transform_luma_4x4_neon_8(int16_t *coeffs); void ff_hevc_transform_luma_4x4_neon_8(int16_t *coeffs);
void ff_hevc_transform_add_4x4_neon_8(uint8_t *_dst, int16_t *coeffs, void ff_hevc_add_residual_4x4_neon_8(uint8_t *_dst, int16_t *coeffs,
ptrdiff_t stride); ptrdiff_t stride);
void ff_hevc_transform_add_8x8_neon_8(uint8_t *_dst, int16_t *coeffs, void ff_hevc_add_residual_8x8_neon_8(uint8_t *_dst, int16_t *coeffs,
ptrdiff_t stride); ptrdiff_t stride);
void ff_hevc_transform_add_16x16_neon_8(uint8_t *_dst, int16_t *coeffs, void ff_hevc_add_residual_16x16_neon_8(uint8_t *_dst, int16_t *coeffs,
ptrdiff_t stride); ptrdiff_t stride);
void ff_hevc_transform_add_32x32_neon_8(uint8_t *_dst, int16_t *coeffs, void ff_hevc_add_residual_32x32_neon_8(uint8_t *_dst, int16_t *coeffs,
ptrdiff_t stride); ptrdiff_t stride);
#define PUT_PIXELS(name) \ #define PUT_PIXELS(name) \
void name(int16_t *dst, uint8_t *src, \ void name(int16_t *dst, uint8_t *src, \
...@@ -156,11 +156,11 @@ av_cold void ff_hevcdsp_init_neon(HEVCDSPContext *c, const int bit_depth) ...@@ -156,11 +156,11 @@ av_cold void ff_hevcdsp_init_neon(HEVCDSPContext *c, const int bit_depth)
c->idct_dc[1] = ff_hevc_idct_8x8_dc_neon_8; c->idct_dc[1] = ff_hevc_idct_8x8_dc_neon_8;
c->idct_dc[2] = ff_hevc_idct_16x16_dc_neon_8; c->idct_dc[2] = ff_hevc_idct_16x16_dc_neon_8;
c->idct_dc[3] = ff_hevc_idct_32x32_dc_neon_8; c->idct_dc[3] = ff_hevc_idct_32x32_dc_neon_8;
c->transform_add[0] = ff_hevc_transform_add_4x4_neon_8; c->add_residual[0] = ff_hevc_add_residual_4x4_neon_8;
c->transform_add[1] = ff_hevc_transform_add_8x8_neon_8; c->add_residual[1] = ff_hevc_add_residual_8x8_neon_8;
c->transform_add[2] = ff_hevc_transform_add_16x16_neon_8; c->add_residual[2] = ff_hevc_add_residual_16x16_neon_8;
c->transform_add[3] = ff_hevc_transform_add_32x32_neon_8; c->add_residual[3] = ff_hevc_add_residual_32x32_neon_8;
c->idct_4x4_luma = ff_hevc_transform_luma_4x4_neon_8; c->transform_4x4_luma = ff_hevc_transform_luma_4x4_neon_8;
put_hevc_qpel_neon[1][0] = ff_hevc_put_qpel_v1_neon_8; put_hevc_qpel_neon[1][0] = ff_hevc_put_qpel_v1_neon_8;
put_hevc_qpel_neon[2][0] = ff_hevc_put_qpel_v2_neon_8; put_hevc_qpel_neon[2][0] = ff_hevc_put_qpel_v2_neon_8;
put_hevc_qpel_neon[3][0] = ff_hevc_put_qpel_v3_neon_8; put_hevc_qpel_neon[3][0] = ff_hevc_put_qpel_v3_neon_8;
......
...@@ -1052,7 +1052,7 @@ static int hls_transform_unit(HEVCContext *s, int x0, int y0, ...@@ -1052,7 +1052,7 @@ static int hls_transform_unit(HEVCContext *s, int x0, int y0,
for (i = 0; i < (size * size); i++) { for (i = 0; i < (size * size); i++) {
coeffs[i] = ((lc->tu.res_scale_val * coeffs_y[i]) >> 3); coeffs[i] = ((lc->tu.res_scale_val * coeffs_y[i]) >> 3);
} }
s->hevcdsp.transform_add[log2_trafo_size_c-2](dst, coeffs, stride); s->hevcdsp.add_residual[log2_trafo_size_c-2](dst, coeffs, stride);
} }
} }
...@@ -1081,7 +1081,7 @@ static int hls_transform_unit(HEVCContext *s, int x0, int y0, ...@@ -1081,7 +1081,7 @@ static int hls_transform_unit(HEVCContext *s, int x0, int y0,
for (i = 0; i < (size * size); i++) { for (i = 0; i < (size * size); i++) {
coeffs[i] = ((lc->tu.res_scale_val * coeffs_y[i]) >> 3); coeffs[i] = ((lc->tu.res_scale_val * coeffs_y[i]) >> 3);
} }
s->hevcdsp.transform_add[log2_trafo_size_c-2](dst, coeffs, stride); s->hevcdsp.add_residual[log2_trafo_size_c-2](dst, coeffs, stride);
} }
} }
} else if (s->ps.sps->chroma_format_idc && blk_idx == 3) { } else if (s->ps.sps->chroma_format_idc && blk_idx == 3) {
......
...@@ -1476,7 +1476,7 @@ void ff_hevc_hls_residual_coding(HEVCContext *s, int x0, int y0, ...@@ -1476,7 +1476,7 @@ void ff_hevc_hls_residual_coding(HEVCContext *s, int x0, int y0,
FFSWAP(int16_t, coeffs[i], coeffs[16 - i - 1]); FFSWAP(int16_t, coeffs[i], coeffs[16 - i - 1]);
} }
s->hevcdsp.transform_skip(coeffs, log2_trafo_size); s->hevcdsp.dequant(coeffs, log2_trafo_size);
if (explicit_rdpcm_flag || (s->ps.sps->implicit_rdpcm_enabled_flag && if (explicit_rdpcm_flag || (s->ps.sps->implicit_rdpcm_enabled_flag &&
lc->cu.pred_mode == MODE_INTRA && lc->cu.pred_mode == MODE_INTRA &&
...@@ -1486,7 +1486,7 @@ void ff_hevc_hls_residual_coding(HEVCContext *s, int x0, int y0, ...@@ -1486,7 +1486,7 @@ void ff_hevc_hls_residual_coding(HEVCContext *s, int x0, int y0,
s->hevcdsp.transform_rdpcm(coeffs, log2_trafo_size, mode); s->hevcdsp.transform_rdpcm(coeffs, log2_trafo_size, mode);
} }
} else if (lc->cu.pred_mode == MODE_INTRA && c_idx == 0 && log2_trafo_size == 2) { } else if (lc->cu.pred_mode == MODE_INTRA && c_idx == 0 && log2_trafo_size == 2) {
s->hevcdsp.idct_4x4_luma(coeffs); s->hevcdsp.transform_4x4_luma(coeffs);
} else { } else {
int max_xy = FFMAX(last_significant_coeff_x, last_significant_coeff_y); int max_xy = FFMAX(last_significant_coeff_x, last_significant_coeff_y);
if (max_xy == 0) if (max_xy == 0)
...@@ -1510,7 +1510,7 @@ void ff_hevc_hls_residual_coding(HEVCContext *s, int x0, int y0, ...@@ -1510,7 +1510,7 @@ void ff_hevc_hls_residual_coding(HEVCContext *s, int x0, int y0,
coeffs[i] = coeffs[i] + ((lc->tu.res_scale_val * coeffs_y[i]) >> 3); coeffs[i] = coeffs[i] + ((lc->tu.res_scale_val * coeffs_y[i]) >> 3);
} }
} }
s->hevcdsp.transform_add[log2_trafo_size-2](dst, coeffs, stride); s->hevcdsp.add_residual[log2_trafo_size-2](dst, coeffs, stride);
} }
void ff_hevc_hls_mvd_coding(HEVCContext *s, int x0, int y0, int log2_cb_size) void ff_hevc_hls_mvd_coding(HEVCContext *s, int x0, int y0, int log2_cb_size)
......
...@@ -195,13 +195,13 @@ void ff_hevc_dsp_init(HEVCDSPContext *hevcdsp, int bit_depth) ...@@ -195,13 +195,13 @@ void ff_hevc_dsp_init(HEVCDSPContext *hevcdsp, int bit_depth)
#define HEVC_DSP(depth) \ #define HEVC_DSP(depth) \
hevcdsp->put_pcm = FUNC(put_pcm, depth); \ hevcdsp->put_pcm = FUNC(put_pcm, depth); \
hevcdsp->transform_add[0] = FUNC(transform_add4x4, depth); \ hevcdsp->add_residual[0] = FUNC(add_residual4x4, depth); \
hevcdsp->transform_add[1] = FUNC(transform_add8x8, depth); \ hevcdsp->add_residual[1] = FUNC(add_residual8x8, depth); \
hevcdsp->transform_add[2] = FUNC(transform_add16x16, depth); \ hevcdsp->add_residual[2] = FUNC(add_residual16x16, depth); \
hevcdsp->transform_add[3] = FUNC(transform_add32x32, depth); \ hevcdsp->add_residual[3] = FUNC(add_residual32x32, depth); \
hevcdsp->transform_skip = FUNC(transform_skip, depth); \ hevcdsp->dequant = FUNC(dequant, depth); \
hevcdsp->transform_rdpcm = FUNC(transform_rdpcm, depth); \ hevcdsp->transform_rdpcm = FUNC(transform_rdpcm, depth); \
hevcdsp->idct_4x4_luma = FUNC(transform_4x4_luma, depth); \ hevcdsp->transform_4x4_luma = FUNC(transform_4x4_luma, depth); \
hevcdsp->idct[0] = FUNC(idct_4x4, depth); \ hevcdsp->idct[0] = FUNC(idct_4x4, depth); \
hevcdsp->idct[1] = FUNC(idct_8x8, depth); \ hevcdsp->idct[1] = FUNC(idct_8x8, depth); \
hevcdsp->idct[2] = FUNC(idct_16x16, depth); \ hevcdsp->idct[2] = FUNC(idct_16x16, depth); \
......
...@@ -46,13 +46,13 @@ typedef struct HEVCDSPContext { ...@@ -46,13 +46,13 @@ typedef struct HEVCDSPContext {
void (*put_pcm)(uint8_t *_dst, ptrdiff_t _stride, int width, int height, void (*put_pcm)(uint8_t *_dst, ptrdiff_t _stride, int width, int height,
struct GetBitContext *gb, int pcm_bit_depth); struct GetBitContext *gb, int pcm_bit_depth);
void (*transform_add[4])(uint8_t *_dst, int16_t *coeffs, ptrdiff_t _stride); void (*add_residual[4])(uint8_t *_dst, int16_t *coeffs, ptrdiff_t _stride);
void (*transform_skip)(int16_t *coeffs, int16_t log2_size); void (*dequant)(int16_t *coeffs, int16_t log2_size);
void (*transform_rdpcm)(int16_t *coeffs, int16_t log2_size, int mode); void (*transform_rdpcm)(int16_t *coeffs, int16_t log2_size, int mode);
void (*idct_4x4_luma)(int16_t *coeffs); void (*transform_4x4_luma)(int16_t *coeffs);
void (*idct[4])(int16_t *coeffs, int col_limit); void (*idct[4])(int16_t *coeffs, int col_limit);
......
...@@ -42,8 +42,8 @@ static void FUNC(put_pcm)(uint8_t *_dst, ptrdiff_t stride, int width, int height ...@@ -42,8 +42,8 @@ static void FUNC(put_pcm)(uint8_t *_dst, ptrdiff_t stride, int width, int height
} }
} }
static av_always_inline void FUNC(transquant_bypass)(uint8_t *_dst, int16_t *coeffs, static av_always_inline void FUNC(add_residual)(uint8_t *_dst, int16_t *res,
ptrdiff_t stride, int size) ptrdiff_t stride, int size)
{ {
int x, y; int x, y;
pixel *dst = (pixel *)_dst; pixel *dst = (pixel *)_dst;
...@@ -52,35 +52,35 @@ static av_always_inline void FUNC(transquant_bypass)(uint8_t *_dst, int16_t *coe ...@@ -52,35 +52,35 @@ static av_always_inline void FUNC(transquant_bypass)(uint8_t *_dst, int16_t *coe
for (y = 0; y < size; y++) { for (y = 0; y < size; y++) {
for (x = 0; x < size; x++) { for (x = 0; x < size; x++) {
dst[x] = av_clip_pixel(dst[x] + *coeffs); dst[x] = av_clip_pixel(dst[x] + *res);
coeffs++; res++;
} }
dst += stride; dst += stride;
} }
} }
static void FUNC(transform_add4x4)(uint8_t *_dst, int16_t *coeffs, static void FUNC(add_residual4x4)(uint8_t *_dst, int16_t *res,
ptrdiff_t stride) ptrdiff_t stride)
{ {
FUNC(transquant_bypass)(_dst, coeffs, stride, 4); FUNC(add_residual)(_dst, res, stride, 4);
} }
static void FUNC(transform_add8x8)(uint8_t *_dst, int16_t *coeffs, static void FUNC(add_residual8x8)(uint8_t *_dst, int16_t *res,
ptrdiff_t stride) ptrdiff_t stride)
{ {
FUNC(transquant_bypass)(_dst, coeffs, stride, 8); FUNC(add_residual)(_dst, res, stride, 8);
} }
static void FUNC(transform_add16x16)(uint8_t *_dst, int16_t *coeffs, static void FUNC(add_residual16x16)(uint8_t *_dst, int16_t *res,
ptrdiff_t stride) ptrdiff_t stride)
{ {
FUNC(transquant_bypass)(_dst, coeffs, stride, 16); FUNC(add_residual)(_dst, res, stride, 16);
} }
static void FUNC(transform_add32x32)(uint8_t *_dst, int16_t *coeffs, static void FUNC(add_residual32x32)(uint8_t *_dst, int16_t *res,
ptrdiff_t stride) ptrdiff_t stride)
{ {
FUNC(transquant_bypass)(_dst, coeffs, stride, 32); FUNC(add_residual)(_dst, res, stride, 32);
} }
...@@ -106,13 +106,11 @@ static void FUNC(transform_rdpcm)(int16_t *_coeffs, int16_t log2_size, int mode) ...@@ -106,13 +106,11 @@ static void FUNC(transform_rdpcm)(int16_t *_coeffs, int16_t log2_size, int mode)
} }
} }
static void FUNC(transform_skip)(int16_t *_coeffs, int16_t log2_size) static void FUNC(dequant)(int16_t *coeffs, int16_t log2_size)
{ {
int shift = 15 - BIT_DEPTH - log2_size; int shift = 15 - BIT_DEPTH - log2_size;
int x, y; int x, y;
int size = 1 << log2_size; int size = 1 << log2_size;
int16_t *coeffs = _coeffs;
if (shift > 0) { if (shift > 0) {
int offset = 1 << (shift - 1); int offset = 1 << (shift - 1);
...@@ -134,8 +132,6 @@ static void FUNC(transform_skip)(int16_t *_coeffs, int16_t log2_size) ...@@ -134,8 +132,6 @@ static void FUNC(transform_skip)(int16_t *_coeffs, int16_t log2_size)
#define SET(dst, x) (dst) = (x) #define SET(dst, x) (dst) = (x)
#define SCALE(dst, x) (dst) = av_clip_int16(((x) + add) >> shift) #define SCALE(dst, x) (dst) = av_clip_int16(((x) + add) >> shift)
#define ADD_AND_SCALE(dst, x) \
(dst) = av_clip_pixel((dst) + av_clip_int16(((x) + add) >> shift))
#define TR_4x4_LUMA(dst, src, step, assign) \ #define TR_4x4_LUMA(dst, src, step, assign) \
do { \ do { \
...@@ -299,7 +295,6 @@ IDCT_DC(32) ...@@ -299,7 +295,6 @@ IDCT_DC(32)
#undef SET #undef SET
#undef SCALE #undef SCALE
#undef ADD_AND_SCALE
static void FUNC(sao_band_filter)(uint8_t *_dst, uint8_t *_src, static void FUNC(sao_band_filter)(uint8_t *_dst, uint8_t *_src,
ptrdiff_t stride_dst, ptrdiff_t stride_src, ptrdiff_t stride_dst, ptrdiff_t stride_src,
......
...@@ -437,11 +437,11 @@ static av_cold void hevc_dsp_init_msa(HEVCDSPContext *c, ...@@ -437,11 +437,11 @@ static av_cold void hevc_dsp_init_msa(HEVCDSPContext *c,
c->idct_dc[1] = ff_hevc_idct_dc_8x8_msa; c->idct_dc[1] = ff_hevc_idct_dc_8x8_msa;
c->idct_dc[2] = ff_hevc_idct_dc_16x16_msa; c->idct_dc[2] = ff_hevc_idct_dc_16x16_msa;
c->idct_dc[3] = ff_hevc_idct_dc_32x32_msa; c->idct_dc[3] = ff_hevc_idct_dc_32x32_msa;
c->transform_add[0] = ff_hevc_addblk_4x4_msa; c->add_residual[0] = ff_hevc_addblk_4x4_msa;
c->transform_add[1] = ff_hevc_addblk_8x8_msa; c->add_residual[1] = ff_hevc_addblk_8x8_msa;
c->transform_add[2] = ff_hevc_addblk_16x16_msa; c->add_residual[2] = ff_hevc_addblk_16x16_msa;
c->transform_add[3] = ff_hevc_addblk_32x32_msa; c->add_residual[3] = ff_hevc_addblk_32x32_msa;
c->idct_4x4_luma = ff_hevc_idct_luma_4x4_msa; c->transform_4x4_luma = ff_hevc_idct_luma_4x4_msa;
} }
} }
#endif // #if HAVE_MSA #endif // #if HAVE_MSA
......
; /* ; /*
; * Provide SIMD optimizations for transform_add functions for HEVC decoding ; * Provide SIMD optimizations for add_residual functions for HEVC decoding
; * Copyright (c) 2014 Pierre-Edouard LEPERE ; * Copyright (c) 2014 Pierre-Edouard LEPERE
; * ; *
; * This file is part of FFmpeg. ; * This file is part of FFmpeg.
...@@ -52,7 +52,7 @@ cextern pw_1023 ...@@ -52,7 +52,7 @@ cextern pw_1023
INIT_MMX mmxext INIT_MMX mmxext
; void ff_hevc_tranform_add_8_mmxext(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride) ; void ff_hevc_tranform_add_8_mmxext(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride)
cglobal hevc_transform_add4_8, 3, 4, 6 cglobal hevc_add_residual4_8, 3, 4, 6
TR_ADD_MMX_4_8 TR_ADD_MMX_4_8
add r1, 16 add r1, 16
lea r0, [r0+r2*2] lea r0, [r0+r2*2]
...@@ -135,8 +135,8 @@ cglobal hevc_transform_add4_8, 3, 4, 6 ...@@ -135,8 +135,8 @@ cglobal hevc_transform_add4_8, 3, 4, 6
%macro TRANSFORM_ADD_8 0 %macro TRANSFORM_ADD_8 0
; void ff_hevc_transform_add8_8_<opt>(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride) ; void ff_hevc_add_residual8_8_<opt>(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride)
cglobal hevc_transform_add8_8, 3, 4, 8 cglobal hevc_add_residual8_8, 3, 4, 8
lea r3, [r2*3] lea r3, [r2*3]
TR_ADD_SSE_8_8 TR_ADD_SSE_8_8
add r1, 64 add r1, 64
...@@ -144,8 +144,8 @@ cglobal hevc_transform_add8_8, 3, 4, 8 ...@@ -144,8 +144,8 @@ cglobal hevc_transform_add8_8, 3, 4, 8
TR_ADD_SSE_8_8 TR_ADD_SSE_8_8
RET RET
; void ff_hevc_transform_add16_8_<opt>(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride) ; void ff_hevc_add_residual16_8_<opt>(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride)
cglobal hevc_transform_add16_8, 3, 4, 7 cglobal hevc_add_residual16_8, 3, 4, 7
pxor m0, m0 pxor m0, m0
lea r3, [r2*3] lea r3, [r2*3]
TR_ADD_SSE_16_32_8 0, r0, r0+r2 TR_ADD_SSE_16_32_8 0, r0, r0+r2
...@@ -158,8 +158,8 @@ cglobal hevc_transform_add16_8, 3, 4, 7 ...@@ -158,8 +158,8 @@ cglobal hevc_transform_add16_8, 3, 4, 7
%endrep %endrep
RET RET
; void ff_hevc_transform_add32_8_<opt>(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride) ; void ff_hevc_add_residual32_8_<opt>(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride)
cglobal hevc_transform_add32_8, 3, 4, 7 cglobal hevc_add_residual32_8, 3, 4, 7
pxor m0, m0 pxor m0, m0
TR_ADD_SSE_16_32_8 0, r0, r0+16 TR_ADD_SSE_16_32_8 0, r0, r0+16
TR_ADD_SSE_16_32_8 64, r0+r2, r0+r2+16 TR_ADD_SSE_16_32_8 64, r0+r2, r0+r2+16
...@@ -179,8 +179,8 @@ TRANSFORM_ADD_8 ...@@ -179,8 +179,8 @@ TRANSFORM_ADD_8
%if HAVE_AVX2_EXTERNAL %if HAVE_AVX2_EXTERNAL
INIT_YMM avx2 INIT_YMM avx2
; void ff_hevc_transform_add32_8_avx2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride) ; void ff_hevc_add_residual32_8_avx2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride)
cglobal hevc_transform_add32_8, 3, 4, 7 cglobal hevc_add_residual32_8, 3, 4, 7
pxor m0, m0 pxor m0, m0
lea r3, [r2*3] lea r3, [r2*3]
TR_ADD_SSE_16_32_8 0, r0, r0+r2 TR_ADD_SSE_16_32_8 0, r0, r0+r2
...@@ -195,7 +195,7 @@ cglobal hevc_transform_add32_8, 3, 4, 7 ...@@ -195,7 +195,7 @@ cglobal hevc_transform_add32_8, 3, 4, 7
%endif %endif
;----------------------------------------------------------------------------- ;-----------------------------------------------------------------------------
; void ff_hevc_transform_add_10(pixel *dst, int16_t *block, int stride) ; void ff_hevc_add_residual_10(pixel *dst, int16_t *block, int stride)
;----------------------------------------------------------------------------- ;-----------------------------------------------------------------------------
%macro TR_ADD_SSE_8_10 4 %macro TR_ADD_SSE_8_10 4
mova m0, [%4] mova m0, [%4]
...@@ -310,7 +310,7 @@ cglobal hevc_transform_add32_8, 3, 4, 7 ...@@ -310,7 +310,7 @@ cglobal hevc_transform_add32_8, 3, 4, 7
INIT_MMX mmxext INIT_MMX mmxext
cglobal hevc_transform_add4_10,3,4, 6 cglobal hevc_add_residual4_10,3,4, 6
pxor m2, m2 pxor m2, m2
mova m3, [max_pixels_10] mova m3, [max_pixels_10]
TR_ADD_MMX4_10 r0, r2, r1 TR_ADD_MMX4_10 r0, r2, r1
...@@ -320,10 +320,10 @@ cglobal hevc_transform_add4_10,3,4, 6 ...@@ -320,10 +320,10 @@ cglobal hevc_transform_add4_10,3,4, 6
RET RET
;----------------------------------------------------------------------------- ;-----------------------------------------------------------------------------
; void ff_hevc_transform_add_10(pixel *dst, int16_t *block, int stride) ; void ff_hevc_add_residual_10(pixel *dst, int16_t *block, int stride)
;----------------------------------------------------------------------------- ;-----------------------------------------------------------------------------
INIT_XMM sse2 INIT_XMM sse2
cglobal hevc_transform_add8_10,3,4,6 cglobal hevc_add_residual8_10,3,4,6
pxor m4, m4 pxor m4, m4
mova m5, [max_pixels_10] mova m5, [max_pixels_10]
lea r3, [r2*3] lea r3, [r2*3]
...@@ -334,7 +334,7 @@ cglobal hevc_transform_add8_10,3,4,6 ...@@ -334,7 +334,7 @@ cglobal hevc_transform_add8_10,3,4,6
TR_ADD_SSE_8_10 r0, r2, r3, r1 TR_ADD_SSE_8_10 r0, r2, r3, r1
RET RET
cglobal hevc_transform_add16_10,3,4,6 cglobal hevc_add_residual16_10,3,4,6
pxor m4, m4 pxor m4, m4
mova m5, [max_pixels_10] mova m5, [max_pixels_10]
...@@ -346,7 +346,7 @@ cglobal hevc_transform_add16_10,3,4,6 ...@@ -346,7 +346,7 @@ cglobal hevc_transform_add16_10,3,4,6
%endrep %endrep
RET RET
cglobal hevc_transform_add32_10,3,4,6 cglobal hevc_add_residual32_10,3,4,6
pxor m4, m4 pxor m4, m4
mova m5, [max_pixels_10] mova m5, [max_pixels_10]
...@@ -361,7 +361,7 @@ cglobal hevc_transform_add32_10,3,4,6 ...@@ -361,7 +361,7 @@ cglobal hevc_transform_add32_10,3,4,6
%if HAVE_AVX2_EXTERNAL %if HAVE_AVX2_EXTERNAL
INIT_YMM avx2 INIT_YMM avx2
cglobal hevc_transform_add16_10,3,4,6 cglobal hevc_add_residual16_10,3,4,6
pxor m4, m4 pxor m4, m4
mova m5, [max_pixels_10] mova m5, [max_pixels_10]
lea r3, [r2*3] lea r3, [r2*3]
...@@ -374,7 +374,7 @@ cglobal hevc_transform_add16_10,3,4,6 ...@@ -374,7 +374,7 @@ cglobal hevc_transform_add16_10,3,4,6
%endrep %endrep
RET RET
cglobal hevc_transform_add32_10,3,4,6 cglobal hevc_add_residual32_10,3,4,6
pxor m4, m4 pxor m4, m4
mova m5, [max_pixels_10] mova m5, [max_pixels_10]
......
...@@ -239,23 +239,23 @@ WEIGHTING_PROTOTYPES(12, sse4); ...@@ -239,23 +239,23 @@ WEIGHTING_PROTOTYPES(12, sse4);
/////////////////////////////////////////////////////////////////////////////// ///////////////////////////////////////////////////////////////////////////////
// TRANSFORM_ADD // TRANSFORM_ADD
/////////////////////////////////////////////////////////////////////////////// ///////////////////////////////////////////////////////////////////////////////
void ff_hevc_transform_add4_8_mmxext(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); void ff_hevc_add_residual4_8_mmxext(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
void ff_hevc_transform_add8_8_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); void ff_hevc_add_residual8_8_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
void ff_hevc_transform_add16_8_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); void ff_hevc_add_residual16_8_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
void ff_hevc_transform_add32_8_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); void ff_hevc_add_residual32_8_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
void ff_hevc_transform_add8_8_avx(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); void ff_hevc_add_residual8_8_avx(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
void ff_hevc_transform_add16_8_avx(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); void ff_hevc_add_residual16_8_avx(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
void ff_hevc_transform_add32_8_avx(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); void ff_hevc_add_residual32_8_avx(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
void ff_hevc_transform_add32_8_avx2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); void ff_hevc_add_residual32_8_avx2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
void ff_hevc_transform_add4_10_mmxext(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); void ff_hevc_add_residual4_10_mmxext(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
void ff_hevc_transform_add8_10_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); void ff_hevc_add_residual8_10_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
void ff_hevc_transform_add16_10_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); void ff_hevc_add_residual16_10_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
void ff_hevc_transform_add32_10_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); void ff_hevc_add_residual32_10_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
void ff_hevc_transform_add16_10_avx2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); void ff_hevc_add_residual16_10_avx2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
void ff_hevc_transform_add32_10_avx2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); void ff_hevc_add_residual32_10_avx2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride);
#endif // AVCODEC_X86_HEVCDSP_H #endif // AVCODEC_X86_HEVCDSP_H
...@@ -700,7 +700,7 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) ...@@ -700,7 +700,7 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
if (EXTERNAL_MMXEXT(cpu_flags)) { if (EXTERNAL_MMXEXT(cpu_flags)) {
c->idct_dc[0] = ff_hevc_idct4x4_dc_8_mmxext; c->idct_dc[0] = ff_hevc_idct4x4_dc_8_mmxext;
c->idct_dc[1] = ff_hevc_idct8x8_dc_8_mmxext; c->idct_dc[1] = ff_hevc_idct8x8_dc_8_mmxext;
c->transform_add[0] = ff_hevc_transform_add4_8_mmxext; c->add_residual[0] = ff_hevc_add_residual4_8_mmxext;
} }
if (EXTERNAL_SSE2(cpu_flags)) { if (EXTERNAL_SSE2(cpu_flags)) {
c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_8_sse2; c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_8_sse2;
...@@ -716,9 +716,9 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) ...@@ -716,9 +716,9 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
c->idct_dc[2] = ff_hevc_idct16x16_dc_8_sse2; c->idct_dc[2] = ff_hevc_idct16x16_dc_8_sse2;
c->idct_dc[3] = ff_hevc_idct32x32_dc_8_sse2; c->idct_dc[3] = ff_hevc_idct32x32_dc_8_sse2;
c->transform_add[1] = ff_hevc_transform_add8_8_sse2; c->add_residual[1] = ff_hevc_add_residual8_8_sse2;
c->transform_add[2] = ff_hevc_transform_add16_8_sse2; c->add_residual[2] = ff_hevc_add_residual16_8_sse2;
c->transform_add[3] = ff_hevc_transform_add32_8_sse2; c->add_residual[3] = ff_hevc_add_residual32_8_sse2;
} }
if (EXTERNAL_SSSE3(cpu_flags)) { if (EXTERNAL_SSSE3(cpu_flags)) {
if(ARCH_X86_64) { if(ARCH_X86_64) {
...@@ -748,9 +748,9 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) ...@@ -748,9 +748,9 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
} }
SAO_BAND_INIT(8, avx); SAO_BAND_INIT(8, avx);
c->transform_add[1] = ff_hevc_transform_add8_8_avx; c->add_residual[1] = ff_hevc_add_residual8_8_avx;
c->transform_add[2] = ff_hevc_transform_add16_8_avx; c->add_residual[2] = ff_hevc_add_residual16_8_avx;
c->transform_add[3] = ff_hevc_transform_add32_8_avx; c->add_residual[3] = ff_hevc_add_residual32_8_avx;
} }
if (EXTERNAL_AVX2(cpu_flags)) { if (EXTERNAL_AVX2(cpu_flags)) {
c->sao_band_filter[0] = ff_hevc_sao_band_filter_8_8_avx2; c->sao_band_filter[0] = ff_hevc_sao_band_filter_8_8_avx2;
...@@ -850,11 +850,11 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) ...@@ -850,11 +850,11 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
c->sao_edge_filter[3] = ff_hevc_sao_edge_filter_48_8_avx2; c->sao_edge_filter[3] = ff_hevc_sao_edge_filter_48_8_avx2;
c->sao_edge_filter[4] = ff_hevc_sao_edge_filter_64_8_avx2; c->sao_edge_filter[4] = ff_hevc_sao_edge_filter_64_8_avx2;
c->transform_add[3] = ff_hevc_transform_add32_8_avx2; c->add_residual[3] = ff_hevc_add_residual32_8_avx2;
} }
} else if (bit_depth == 10) { } else if (bit_depth == 10) {
if (EXTERNAL_MMXEXT(cpu_flags)) { if (EXTERNAL_MMXEXT(cpu_flags)) {
c->transform_add[0] = ff_hevc_transform_add4_10_mmxext; c->add_residual[0] = ff_hevc_add_residual4_10_mmxext;
c->idct_dc[0] = ff_hevc_idct4x4_dc_10_mmxext; c->idct_dc[0] = ff_hevc_idct4x4_dc_10_mmxext;
c->idct_dc[1] = ff_hevc_idct8x8_dc_10_mmxext; c->idct_dc[1] = ff_hevc_idct8x8_dc_10_mmxext;
} }
...@@ -872,9 +872,9 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) ...@@ -872,9 +872,9 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
c->idct_dc[2] = ff_hevc_idct16x16_dc_10_sse2; c->idct_dc[2] = ff_hevc_idct16x16_dc_10_sse2;
c->idct_dc[3] = ff_hevc_idct32x32_dc_10_sse2; c->idct_dc[3] = ff_hevc_idct32x32_dc_10_sse2;
c->transform_add[1] = ff_hevc_transform_add8_10_sse2; c->add_residual[1] = ff_hevc_add_residual8_10_sse2;
c->transform_add[2] = ff_hevc_transform_add16_10_sse2; c->add_residual[2] = ff_hevc_add_residual16_10_sse2;
c->transform_add[3] = ff_hevc_transform_add32_10_sse2; c->add_residual[3] = ff_hevc_add_residual32_10_sse2;
} }
if (EXTERNAL_SSSE3(cpu_flags) && ARCH_X86_64) { if (EXTERNAL_SSSE3(cpu_flags) && ARCH_X86_64) {
c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_10_ssse3; c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_10_ssse3;
...@@ -1053,8 +1053,8 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) ...@@ -1053,8 +1053,8 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
SAO_BAND_INIT(10, avx2); SAO_BAND_INIT(10, avx2);
SAO_EDGE_INIT(10, avx2); SAO_EDGE_INIT(10, avx2);
c->transform_add[2] = ff_hevc_transform_add16_10_avx2; c->add_residual[2] = ff_hevc_add_residual16_10_avx2;
c->transform_add[3] = ff_hevc_transform_add32_10_avx2; c->add_residual[3] = ff_hevc_add_residual32_10_avx2;
} }
} else if (bit_depth == 12) { } else if (bit_depth == 12) {
......
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