falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
Browse Source

Update 8-bit H.264 IDCT function names to reflect bit-depth. 

Signed-off-by: Ronald S. Bultje <rbultje@google.com>
tags/n0.8
Daniel Kang Ronald S. Bultje 14 years ago
parent
836f47d34b
commit
348493db60
3 changed files with 57 additions and 72 deletions
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. +0
    -1
      libavcodec/h264dsp.h
  2. +19
    -19
      libavcodec/x86/h264_idct.asm
  3. +38
    -52
      libavcodec/x86/h264dsp_mmx.c

+ 0
- 1
libavcodec/h264dsp.h View File

@@ -66,7 +66,6 @@ typedef struct H264DSPContext{
void (*h264_idct_dc_add)(uint8_t *dst/*align 4*/, DCTELEM *block/*align 16*/, int stride);
void (*h264_idct8_dc_add)(uint8_t *dst/*align 8*/, DCTELEM *block/*align 16*/, int stride);

void (*h264_dct)(DCTELEM block[4][4]);
void (*h264_idct_add16)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
void (*h264_idct8_add4)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
void (*h264_idct_add8)(uint8_t **dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);


+ 19
- 19
libavcodec/x86/h264_idct.asm View File

@@ -73,7 +73,7 @@ SECTION .text

INIT_MMX
; ff_h264_idct_add_mmx(uint8_t *dst, int16_t *block, int stride)
cglobal h264_idct_add_mmx, 3, 3, 0
cglobal h264_idct_add_8_mmx, 3, 3, 0
IDCT4_ADD r0, r1, r2
RET

@@ -125,7 +125,7 @@ cglobal h264_idct_add_mmx, 3, 3, 0
SUMSUB_BA w, 0, 4
SUMSUB_BA w, 3, 2
SUMSUB_BA w, 1, 5
SWAP 7, 6, 4, 5, 2, 3, 1, 0 ; 70315246 -> 01234567
SWAP 7, 6, 4, 5, 2, 3, 1, 0 ; 70315246 -> 01234567
%endmacro

%macro IDCT8_1D_FULL 1
@@ -177,7 +177,7 @@ cglobal h264_idct_add_mmx, 3, 3, 0

INIT_MMX
; ff_h264_idct8_add_mmx(uint8_t *dst, int16_t *block, int stride)
cglobal h264_idct8_add_mmx, 3, 4, 0
cglobal h264_idct8_add_8_mmx, 3, 4, 0
%assign pad 128+4-(stack_offset&7)
SUB rsp, pad

@@ -237,7 +237,7 @@ cglobal h264_idct8_add_mmx, 3, 4, 0

INIT_XMM
; ff_h264_idct8_add_sse2(uint8_t *dst, int16_t *block, int stride)
cglobal h264_idct8_add_sse2, 3, 4, 10
cglobal h264_idct8_add_8_sse2, 3, 4, 10
IDCT8_ADD_SSE r0, r1, r2, r3
RET

@@ -261,7 +261,7 @@ cglobal h264_idct8_add_sse2, 3, 4, 10
packuswb m1, m1
%endmacro

%macro DC_ADD_MMX2_OP 3-4
%macro DC_ADD_MMX2_OP 4
%1 m2, [%2 ]
%1 m3, [%2+%3 ]
%1 m4, [%2+%3*2]
@@ -282,13 +282,13 @@ cglobal h264_idct8_add_sse2, 3, 4, 10

INIT_MMX
; ff_h264_idct_dc_add_mmx2(uint8_t *dst, int16_t *block, int stride)
cglobal h264_idct_dc_add_mmx2, 3, 3, 0
cglobal h264_idct_dc_add_8_mmx2, 3, 3, 0
DC_ADD_MMX2_INIT r1, r2
DC_ADD_MMX2_OP movh, r0, r2, r1
RET

; ff_h264_idct8_dc_add_mmx2(uint8_t *dst, int16_t *block, int stride)
cglobal h264_idct8_dc_add_mmx2, 3, 3, 0
cglobal h264_idct8_dc_add_8_mmx2, 3, 3, 0
DC_ADD_MMX2_INIT r1, r2
DC_ADD_MMX2_OP mova, r0, r2, r1
lea r0, [r0+r2*4]
@@ -297,7 +297,7 @@ cglobal h264_idct8_dc_add_mmx2, 3, 3, 0

; ff_h264_idct_add16_mmx(uint8_t *dst, const int *block_offset,
; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
cglobal h264_idct_add16_mmx, 5, 7, 0
cglobal h264_idct_add16_8_mmx, 5, 7, 0
xor r5, r5
%ifdef PIC
lea r11, [scan8_mem]
@@ -319,7 +319,7 @@ cglobal h264_idct_add16_mmx, 5, 7, 0

; ff_h264_idct8_add4_mmx(uint8_t *dst, const int *block_offset,
; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
cglobal h264_idct8_add4_mmx, 5, 7, 0
cglobal h264_idct8_add4_8_mmx, 5, 7, 0
%assign pad 128+4-(stack_offset&7)
SUB rsp, pad

@@ -351,7 +351,7 @@ cglobal h264_idct8_add4_mmx, 5, 7, 0

; ff_h264_idct_add16_mmx2(uint8_t *dst, const int *block_offset,
; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
cglobal h264_idct_add16_mmx2, 5, 7, 0
cglobal h264_idct_add16_8_mmx2, 5, 7, 0
xor r5, r5
%ifdef PIC
lea r11, [scan8_mem]
@@ -398,7 +398,7 @@ cglobal h264_idct_add16_mmx2, 5, 7, 0

; ff_h264_idct_add16intra_mmx(uint8_t *dst, const int *block_offset,
; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
cglobal h264_idct_add16intra_mmx, 5, 7, 0
cglobal h264_idct_add16intra_8_mmx, 5, 7, 0
xor r5, r5
%ifdef PIC
lea r11, [scan8_mem]
@@ -421,7 +421,7 @@ cglobal h264_idct_add16intra_mmx, 5, 7, 0

; ff_h264_idct_add16intra_mmx2(uint8_t *dst, const int *block_offset,
; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
cglobal h264_idct_add16intra_mmx2, 5, 7, 0
cglobal h264_idct_add16intra_8_mmx2, 5, 7, 0
xor r5, r5
%ifdef PIC
lea r11, [scan8_mem]
@@ -466,7 +466,7 @@ cglobal h264_idct_add16intra_mmx2, 5, 7, 0

; ff_h264_idct8_add4_mmx2(uint8_t *dst, const int *block_offset,
; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
cglobal h264_idct8_add4_mmx2, 5, 7, 0
cglobal h264_idct8_add4_8_mmx2, 5, 7, 0
%assign pad 128+4-(stack_offset&7)
SUB rsp, pad

@@ -529,7 +529,7 @@ cglobal h264_idct8_add4_mmx2, 5, 7, 0
INIT_XMM
; ff_h264_idct8_add4_sse2(uint8_t *dst, const int *block_offset,
; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
cglobal h264_idct8_add4_sse2, 5, 7, 10
cglobal h264_idct8_add4_8_sse2, 5, 7, 10
xor r5, r5
%ifdef PIC
lea r11, [scan8_mem]
@@ -607,7 +607,7 @@ h264_idct_add8_mmx_plane:

; ff_h264_idct_add8_mmx(uint8_t **dest, const int *block_offset,
; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
cglobal h264_idct_add8_mmx, 5, 7, 0
cglobal h264_idct_add8_8_mmx, 5, 7, 0
mov r5, 16
add r2, 512
%ifdef PIC
@@ -668,7 +668,7 @@ h264_idct_add8_mmx2_plane

; ff_h264_idct_add8_mmx2(uint8_t **dest, const int *block_offset,
; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
cglobal h264_idct_add8_mmx2, 5, 7, 0
cglobal h264_idct_add8_8_mmx2, 5, 7, 0
mov r5, 16
add r2, 512
%ifdef ARCH_X86_64
@@ -744,7 +744,7 @@ x264_add8x4_idct_sse2:

; ff_h264_idct_add16_sse2(uint8_t *dst, const int *block_offset,
; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
cglobal h264_idct_add16_sse2, 5, 5, 8
cglobal h264_idct_add16_8_sse2, 5, 5, 8
%ifdef ARCH_X86_64
mov r10, r0
%endif
@@ -791,7 +791,7 @@ cglobal h264_idct_add16_sse2, 5, 5, 8

; ff_h264_idct_add16intra_sse2(uint8_t *dst, const int *block_offset,
; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
cglobal h264_idct_add16intra_sse2, 5, 7, 8
cglobal h264_idct_add16intra_8_sse2, 5, 7, 8
%ifdef ARCH_X86_64
mov r10, r0
%endif
@@ -840,7 +840,7 @@ cglobal h264_idct_add16intra_sse2, 5, 7, 8

; ff_h264_idct_add8_sse2(uint8_t **dest, const int *block_offset,
; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
cglobal h264_idct_add8_sse2, 5, 7, 8
cglobal h264_idct_add8_8_sse2, 5, 7, 8
add r2, 512
%ifdef ARCH_X86_64
mov r10, r0


+ 38
- 52
libavcodec/x86/h264dsp_mmx.c View File

@@ -30,9 +30,14 @@ DECLARE_ALIGNED(8, static const uint64_t, ff_pb_3_1 ) = 0x0103010301030103ULL;
#define IDCT_ADD_FUNC(NUM, DEPTH, OPT) \
void ff_h264_idct ## NUM ## _add_ ## DEPTH ## _ ## OPT (uint8_t *dst, int16_t *block, int stride);

IDCT_ADD_FUNC(, 8, mmx)
IDCT_ADD_FUNC(, 10, sse2)
IDCT_ADD_FUNC(_dc, 8, mmx2)
IDCT_ADD_FUNC(_dc, 10, mmx2)
IDCT_ADD_FUNC(8_dc, 8, mmx2)
IDCT_ADD_FUNC(8_dc, 10, sse2)
IDCT_ADD_FUNC(8, 8, mmx)
IDCT_ADD_FUNC(8, 8, sse2)
IDCT_ADD_FUNC(8, 10, sse2)
#if HAVE_AVX
IDCT_ADD_FUNC(, 10, avx)
@@ -46,9 +51,18 @@ void ff_h264_idct ## NUM ## _add ## REP ## _ ## DEPTH ## _ ## OPT \
(uint8_t *dst, const int *block_offset, \
DCTELEM *block, int stride, const uint8_t nnzc[6*8]);

IDCT_ADD_REP_FUNC(8, 4, 8, mmx)
IDCT_ADD_REP_FUNC(8, 4, 8, mmx2)
IDCT_ADD_REP_FUNC(8, 4, 8, sse2)
IDCT_ADD_REP_FUNC(8, 4, 10, sse2)
IDCT_ADD_REP_FUNC(8, 4, 10, avx)
IDCT_ADD_REP_FUNC(, 16, 8, mmx)
IDCT_ADD_REP_FUNC(, 16, 8, mmx2)
IDCT_ADD_REP_FUNC(, 16, 8, sse2)
IDCT_ADD_REP_FUNC(, 16, 10, sse2)
IDCT_ADD_REP_FUNC(, 16intra, 8, mmx)
IDCT_ADD_REP_FUNC(, 16intra, 8, mmx2)
IDCT_ADD_REP_FUNC(, 16intra, 8, sse2)
IDCT_ADD_REP_FUNC(, 16intra, 10, sse2)
#if HAVE_AVX
IDCT_ADD_REP_FUNC(, 16, 10, avx)
@@ -60,42 +74,14 @@ IDCT_ADD_REP_FUNC(, 16intra, 10, avx)
void ff_h264_idct ## NUM ## _add ## REP ## _ ## DEPTH ## _ ## OPT \
(uint8_t **dst, const int *block_offset, \
DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
IDCT_ADD_REP_FUNC2(, 8, 8, mmx)
IDCT_ADD_REP_FUNC2(, 8, 8, mmx2)
IDCT_ADD_REP_FUNC2(, 8, 8, sse2)
IDCT_ADD_REP_FUNC2(, 8, 10, sse2)
#if HAVE_AVX
IDCT_ADD_REP_FUNC2(, 8, 10, avx)
#endif

void ff_h264_idct_add_mmx (uint8_t *dst, int16_t *block, int stride);
void ff_h264_idct8_add_mmx (uint8_t *dst, int16_t *block, int stride);
void ff_h264_idct8_add_sse2 (uint8_t *dst, int16_t *block, int stride);
void ff_h264_idct_dc_add_mmx2 (uint8_t *dst, int16_t *block, int stride);
void ff_h264_idct8_dc_add_mmx2(uint8_t *dst, int16_t *block, int stride);

void ff_h264_idct_add16_mmx (uint8_t *dst, const int *block_offset,
DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
void ff_h264_idct8_add4_mmx (uint8_t *dst, const int *block_offset,
DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
void ff_h264_idct_add16_mmx2 (uint8_t *dst, const int *block_offset,
DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
void ff_h264_idct_add16intra_mmx (uint8_t *dst, const int *block_offset,
DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
void ff_h264_idct_add16intra_mmx2(uint8_t *dst, const int *block_offset,
DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
void ff_h264_idct8_add4_mmx2 (uint8_t *dst, const int *block_offset,
DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
void ff_h264_idct8_add4_sse2 (uint8_t *dst, const int *block_offset,
DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
void ff_h264_idct_add8_mmx (uint8_t **dest, const int *block_offset,
DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
void ff_h264_idct_add8_mmx2 (uint8_t **dest, const int *block_offset,
DCTELEM *block, int stride, const uint8_t nnzc[6*8]);

void ff_h264_idct_add16_sse2 (uint8_t *dst, const int *block_offset, DCTELEM *block,
int stride, const uint8_t nnzc[6*8]);
void ff_h264_idct_add16intra_sse2(uint8_t *dst, const int *block_offset, DCTELEM *block,
int stride, const uint8_t nnzc[6*8]);
void ff_h264_idct_add8_sse2 (uint8_t **dest, const int *block_offset, DCTELEM *block,
int stride, const uint8_t nnzc[6*8]);
void ff_h264_luma_dc_dequant_idct_mmx (DCTELEM *output, DCTELEM *input, int qmul);
void ff_h264_luma_dc_dequant_idct_sse2(DCTELEM *output, DCTELEM *input, int qmul);

@@ -350,24 +336,24 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth)
}
#if HAVE_YASM
if (mm_flags & AV_CPU_FLAG_MMX) {
c->h264_idct_dc_add=
c->h264_idct_add= ff_h264_idct_add_mmx;
c->h264_idct8_dc_add=
c->h264_idct8_add= ff_h264_idct8_add_mmx;
c->h264_idct_add16 = ff_h264_idct_add16_mmx;
c->h264_idct8_add4 = ff_h264_idct8_add4_mmx;
c->h264_idct_add8 = ff_h264_idct_add8_mmx;
c->h264_idct_add16intra= ff_h264_idct_add16intra_mmx;
c->h264_idct_dc_add =
c->h264_idct_add = ff_h264_idct_add_8_mmx;
c->h264_idct8_dc_add =
c->h264_idct8_add = ff_h264_idct8_add_8_mmx;
c->h264_idct_add16 = ff_h264_idct_add16_8_mmx;
c->h264_idct8_add4 = ff_h264_idct8_add4_8_mmx;
c->h264_idct_add8 = ff_h264_idct_add8_8_mmx;
c->h264_idct_add16intra = ff_h264_idct_add16intra_8_mmx;
c->h264_luma_dc_dequant_idct= ff_h264_luma_dc_dequant_idct_mmx;

if (mm_flags & AV_CPU_FLAG_MMX2) {
c->h264_idct_dc_add= ff_h264_idct_dc_add_mmx2;
c->h264_idct8_dc_add= ff_h264_idct8_dc_add_mmx2;
c->h264_idct_add16 = ff_h264_idct_add16_mmx2;
c->h264_idct8_add4 = ff_h264_idct8_add4_mmx2;
c->h264_idct_add8 = ff_h264_idct_add8_mmx2;
c->h264_idct_add16intra= ff_h264_idct_add16intra_mmx2;
c->h264_idct_dc_add = ff_h264_idct_dc_add_8_mmx2;
c->h264_idct8_dc_add = ff_h264_idct8_dc_add_8_mmx2;
c->h264_idct_add16 = ff_h264_idct_add16_8_mmx2;
c->h264_idct8_add4 = ff_h264_idct8_add4_8_mmx2;
c->h264_idct_add8 = ff_h264_idct_add8_8_mmx2;
c->h264_idct_add16intra= ff_h264_idct_add16intra_8_mmx2;

c->h264_v_loop_filter_chroma= ff_deblock_v_chroma_8_mmxext;
c->h264_h_loop_filter_chroma= ff_deblock_h_chroma_8_mmxext;
@@ -398,8 +384,12 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth)
c->biweight_h264_pixels_tab[7]= ff_h264_biweight_4x2_mmx2;

if (mm_flags&AV_CPU_FLAG_SSE2) {
c->h264_idct8_add = ff_h264_idct8_add_sse2;
c->h264_idct8_add4= ff_h264_idct8_add4_sse2;
c->h264_idct8_add = ff_h264_idct8_add_8_sse2;

c->h264_idct_add16 = ff_h264_idct_add16_8_sse2;
c->h264_idct8_add4 = ff_h264_idct8_add4_8_sse2;
c->h264_idct_add8 = ff_h264_idct_add8_8_sse2;
c->h264_idct_add16intra = ff_h264_idct_add16intra_8_sse2;
c->h264_luma_dc_dequant_idct= ff_h264_luma_dc_dequant_idct_sse2;

c->weight_h264_pixels_tab[0]= ff_h264_weight_16x16_sse2;
@@ -420,10 +410,6 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth)
c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_8_sse2;
c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_8_sse2;
#endif

c->h264_idct_add16 = ff_h264_idct_add16_sse2;
c->h264_idct_add8 = ff_h264_idct_add8_sse2;
c->h264_idct_add16intra = ff_h264_idct_add16intra_sse2;
}
if (mm_flags&AV_CPU_FLAG_SSSE3) {
c->biweight_h264_pixels_tab[0]= ff_h264_biweight_16x16_ssse3;


Write Preview
Loading…
Cancel
Save