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

x86: mmx2 ---> mmxext in asm constructs

tags/n1.1
Diego Biurrun 13 years ago
parent
da39cac8de
commit
26301caaa1
31 changed files with 317 additions and 312 deletions
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. +2
    -2
      libavcodec/x86/ac3dsp.asm
  2. +2
    -2
      libavcodec/x86/ac3dsp_init.c
  3. +4
    -4
      libavcodec/x86/dsputil.asm
  4. +28
    -28
      libavcodec/x86/dsputil_mmx.c
  5. +1
    -1
      libavcodec/x86/dsputilenc.asm
  6. +3
    -3
      libavcodec/x86/dsputilenc_mmx.c
  7. +7
    -7
      libavcodec/x86/h264_chromamc.asm
  8. +2
    -2
      libavcodec/x86/h264_chromamc_10bit.asm
  9. +10
    -10
      libavcodec/x86/h264_deblock.asm
  10. +2
    -2
      libavcodec/x86/h264_deblock_10bit.asm
  11. +24
    -22
      libavcodec/x86/h264_idct.asm
  12. +1
    -1
      libavcodec/x86/h264_idct_10bit.asm
  13. +13
    -13
      libavcodec/x86/h264_intrapred.asm
  14. +10
    -10
      libavcodec/x86/h264_intrapred_10bit.asm
  15. +40
    -40
      libavcodec/x86/h264_intrapred_init.c
  16. +6
    -6
      libavcodec/x86/h264_weight.asm
  17. +54
    -53
      libavcodec/x86/h264dsp_init.c
  18. +1
    -1
      libavcodec/x86/pngdsp.asm
  19. +3
    -3
      libavcodec/x86/pngdsp_init.c
  20. +3
    -3
      libavcodec/x86/rv34dsp.asm
  21. +5
    -5
      libavcodec/x86/rv34dsp_init.c
  22. +2
    -2
      libavcodec/x86/rv40dsp.asm
  23. +15
    -15
      libavcodec/x86/rv40dsp_init.c
  24. +3
    -3
      libavcodec/x86/vc1dsp_init.c
  25. +2
    -2
      libavcodec/x86/vp3dsp.asm
  26. +9
    -7
      libavcodec/x86/vp3dsp_init.c
  27. +15
    -15
      libavcodec/x86/vp8dsp.asm
  28. +43
    -43
      libavcodec/x86/vp8dsp_init.c
  29. +1
    -1
      libavutil/x86/x86util.asm
  30. +2
    -2
      libswscale/x86/output.asm
  31. +4
    -4
      libswscale/x86/swscale.c

+ 2
- 2
libavcodec/x86/ac3dsp.asm View File

@@ -97,7 +97,7 @@ AC3_EXPONENT_MIN
por %1, %2
pshuflw %2, %1, q0001
por %1, %2
%elif cpuflag(mmx2)
%elif cpuflag(mmxext)
pshufw %2, %1, q0032
por %1, %2
pshufw %2, %1, q0001
@@ -153,7 +153,7 @@ cglobal ac3_max_msb_abs_int16, 2,2,5, src, len
INIT_MMX mmx
%define ABS2 ABS2_MMX
AC3_MAX_MSB_ABS_INT16 or_abs
INIT_MMX mmx2
INIT_MMX mmxext
%define ABS2 ABS2_MMXEXT
AC3_MAX_MSB_ABS_INT16 min_max
INIT_XMM sse2


+ 2
- 2
libavcodec/x86/ac3dsp_init.c View File

@@ -31,7 +31,7 @@ extern void ff_ac3_exponent_min_mmxext(uint8_t *exp, int num_reuse_blocks, int n
extern void ff_ac3_exponent_min_sse2 (uint8_t *exp, int num_reuse_blocks, int nb_coefs);

extern int ff_ac3_max_msb_abs_int16_mmx (const int16_t *src, int len);
extern int ff_ac3_max_msb_abs_int16_mmx2 (const int16_t *src, int len);
extern int ff_ac3_max_msb_abs_int16_mmxext(const int16_t *src, int len);
extern int ff_ac3_max_msb_abs_int16_sse2 (const int16_t *src, int len);
extern int ff_ac3_max_msb_abs_int16_ssse3(const int16_t *src, int len);

@@ -182,7 +182,7 @@ av_cold void ff_ac3dsp_init_x86(AC3DSPContext *c, int bit_exact)
}
if (EXTERNAL_MMXEXT(mm_flags)) {
c->ac3_exponent_min = ff_ac3_exponent_min_mmxext;
c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_mmx2;
c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_mmxext;
}
if (EXTERNAL_SSE(mm_flags)) {
c->float_to_fixed24 = ff_float_to_fixed24_sse;


+ 4
- 4
libavcodec/x86/dsputil.asm View File

@@ -108,7 +108,7 @@ cglobal scalarproduct_and_madd_int16_%1, 4,4,8, v1, v2, v3, order, mul
%endmacro

INIT_MMX
SCALARPRODUCT mmx2
SCALARPRODUCT mmxext
INIT_XMM
SCALARPRODUCT sse2

@@ -327,8 +327,8 @@ APPLY_WINDOW_INT16 ssse3_atom, 0, 1
APPLY_WINDOW_INT16 ssse3, 0, 1


; void add_hfyu_median_prediction_mmx2(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top)
cglobal add_hfyu_median_prediction_mmx2, 6,6,0, dst, top, diff, w, left, left_top
; void add_hfyu_median_prediction_mmxext(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top)
cglobal add_hfyu_median_prediction_mmxext, 6,6,0, dst, top, diff, w, left, left_top
movq mm0, [topq]
movq mm2, mm0
movd mm4, [left_topq]
@@ -804,7 +804,7 @@ ALIGN 128
mov valh, vall
%if %1 >= 8
movd mm0, vald
%if cpuflag(mmx2)
%if cpuflag(mmxext)
pshufw mm0, mm0, 0
%else ; mmx
punpcklwd mm0, mm0


+ 28
- 28
libavcodec/x86/dsputil_mmx.c View File

@@ -2045,21 +2045,21 @@ PREFETCH(prefetch_3dnow, prefetch)

void ff_put_h264_chroma_mc8_rnd_mmx (uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);
void ff_avg_h264_chroma_mc8_rnd_mmx2 (uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);
void ff_avg_h264_chroma_mc8_rnd_mmxext(uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);
void ff_avg_h264_chroma_mc8_rnd_3dnow(uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);

void ff_put_h264_chroma_mc4_mmx (uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);
void ff_avg_h264_chroma_mc4_mmx2 (uint8_t *dst, uint8_t *src,
void ff_avg_h264_chroma_mc4_mmxext (uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);
void ff_avg_h264_chroma_mc4_3dnow (uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);

void ff_put_h264_chroma_mc2_mmx2 (uint8_t *dst, uint8_t *src,
void ff_put_h264_chroma_mc2_mmxext (uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);
void ff_avg_h264_chroma_mc2_mmx2 (uint8_t *dst, uint8_t *src,
void ff_avg_h264_chroma_mc2_mmxext (uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);

void ff_put_h264_chroma_mc8_rnd_ssse3(uint8_t *dst, uint8_t *src,
@@ -2077,10 +2077,10 @@ void ff_ ## OP ## _h264_chroma_mc ## NUM ## _ ## DEPTH ## _ ## OPT \
(uint8_t *dst, uint8_t *src, \
int stride, int h, int x, int y);

CHROMA_MC(put, 2, 10, mmx2)
CHROMA_MC(avg, 2, 10, mmx2)
CHROMA_MC(put, 4, 10, mmx2)
CHROMA_MC(avg, 4, 10, mmx2)
CHROMA_MC(put, 2, 10, mmxext)
CHROMA_MC(avg, 2, 10, mmxext)
CHROMA_MC(put, 4, 10, mmxext)
CHROMA_MC(avg, 4, 10, mmxext)
CHROMA_MC(put, 8, 10, sse2)
CHROMA_MC(avg, 8, 10, sse2)
CHROMA_MC(put, 8, 10, avx)
@@ -2283,13 +2283,13 @@ static void vector_clipf_sse(float *dst, const float *src,

#endif /* HAVE_INLINE_ASM */

int32_t ff_scalarproduct_int16_mmx2(const int16_t *v1, const int16_t *v2,
int order);
int32_t ff_scalarproduct_int16_mmxext(const int16_t *v1, const int16_t *v2,
int order);
int32_t ff_scalarproduct_int16_sse2(const int16_t *v1, const int16_t *v2,
int order);
int32_t ff_scalarproduct_and_madd_int16_mmx2(int16_t *v1, const int16_t *v2,
const int16_t *v3,
int order, int mul);
int32_t ff_scalarproduct_and_madd_int16_mmxext(int16_t *v1, const int16_t *v2,
const int16_t *v3,
int order, int mul);
int32_t ff_scalarproduct_and_madd_int16_sse2(int16_t *v1, const int16_t *v2,
const int16_t *v3,
int order, int mul);
@@ -2313,9 +2313,9 @@ void ff_apply_window_int16_ssse3_atom(int16_t *output, const int16_t *input,
void ff_bswap32_buf_ssse3(uint32_t *dst, const uint32_t *src, int w);
void ff_bswap32_buf_sse2(uint32_t *dst, const uint32_t *src, int w);

void ff_add_hfyu_median_prediction_mmx2(uint8_t *dst, const uint8_t *top,
const uint8_t *diff, int w,
int *left, int *left_top);
void ff_add_hfyu_median_prediction_mmxext(uint8_t *dst, const uint8_t *top,
const uint8_t *diff, int w,
int *left, int *left_top);
int ff_add_hfyu_left_prediction_ssse3(uint8_t *dst, const uint8_t *src,
int w, int left);
int ff_add_hfyu_left_prediction_sse4(uint8_t *dst, const uint8_t *src,
@@ -2548,24 +2548,24 @@ static void dsputil_init_mmxext(DSPContext *c, AVCodecContext *avctx,

#if HAVE_YASM
if (!high_bit_depth && CONFIG_H264CHROMA) {
c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_rnd_mmx2;
c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_mmx2;
c->avg_h264_chroma_pixels_tab[2] = ff_avg_h264_chroma_mc2_mmx2;
c->put_h264_chroma_pixels_tab[2] = ff_put_h264_chroma_mc2_mmx2;
c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_rnd_mmxext;
c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_mmxext;
c->avg_h264_chroma_pixels_tab[2] = ff_avg_h264_chroma_mc2_mmxext;
c->put_h264_chroma_pixels_tab[2] = ff_put_h264_chroma_mc2_mmxext;
}
if (bit_depth == 10 && CONFIG_H264CHROMA) {
c->put_h264_chroma_pixels_tab[2] = ff_put_h264_chroma_mc2_10_mmx2;
c->avg_h264_chroma_pixels_tab[2] = ff_avg_h264_chroma_mc2_10_mmx2;
c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_10_mmx2;
c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_10_mmx2;
c->put_h264_chroma_pixels_tab[2] = ff_put_h264_chroma_mc2_10_mmxext;
c->avg_h264_chroma_pixels_tab[2] = ff_avg_h264_chroma_mc2_10_mmxext;
c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_10_mmxext;
c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_10_mmxext;
}

/* slower than cmov version on AMD */
if (!(mm_flags & AV_CPU_FLAG_3DNOW))
c->add_hfyu_median_prediction = ff_add_hfyu_median_prediction_mmx2;
c->add_hfyu_median_prediction = ff_add_hfyu_median_prediction_mmxext;

c->scalarproduct_int16 = ff_scalarproduct_int16_mmx2;
c->scalarproduct_and_madd_int16 = ff_scalarproduct_and_madd_int16_mmx2;
c->scalarproduct_int16 = ff_scalarproduct_int16_mmxext;
c->scalarproduct_and_madd_int16 = ff_scalarproduct_and_madd_int16_mmxext;

if (avctx->flags & CODEC_FLAG_BITEXACT) {
c->apply_window_int16 = ff_apply_window_int16_mmxext_ba;


+ 1
- 1
libavcodec/x86/dsputilenc.asm View File

@@ -265,7 +265,7 @@ HADAMARD8_DIFF_MMX mmx

%define ABS1 ABS1_MMXEXT
%define HSUM HSUM_MMXEXT
HADAMARD8_DIFF_MMX mmx2
HADAMARD8_DIFF_MMX mmxext

INIT_XMM
%define ABS2 ABS2_MMXEXT


+ 3
- 3
libavcodec/x86/dsputilenc_mmx.c View File

@@ -1104,7 +1104,7 @@ int ff_hadamard8_diff16_##cpu(void *s, uint8_t *src1, uint8_t *src2, \
int stride, int h);

hadamard_func(mmx)
hadamard_func(mmx2)
hadamard_func(mmxext)
hadamard_func(sse2)
hadamard_func(ssse3)

@@ -1195,8 +1195,8 @@ void ff_dsputilenc_init_mmx(DSPContext* c, AVCodecContext *avctx)
c->hadamard8_diff[1] = ff_hadamard8_diff_mmx;

if (EXTERNAL_MMXEXT(mm_flags)) {
c->hadamard8_diff[0] = ff_hadamard8_diff16_mmx2;
c->hadamard8_diff[1] = ff_hadamard8_diff_mmx2;
c->hadamard8_diff[0] = ff_hadamard8_diff16_mmxext;
c->hadamard8_diff[1] = ff_hadamard8_diff_mmxext;
}

if (EXTERNAL_SSE2(mm_flags)) {


+ 7
- 7
libavcodec/x86/h264_chromamc.asm View File

@@ -442,17 +442,17 @@ chroma_mc8_mmx_func put, vc1, nornd_mmx
chroma_mc8_mmx_func put, rv40, mmx
chroma_mc4_mmx_func put, h264, mmx
chroma_mc4_mmx_func put, rv40, mmx
chroma_mc2_mmx_func put, h264, mmx2
chroma_mc2_mmx_func put, h264, mmxext

%define CHROMAMC_AVG DIRECT_AVG
%define CHROMAMC_AVG4 COPY_AVG
%define PAVG pavgb
chroma_mc8_mmx_func avg, h264, rnd_mmx2
chroma_mc8_mmx_func avg, vc1, nornd_mmx2
chroma_mc8_mmx_func avg, rv40, mmx2
chroma_mc4_mmx_func avg, h264, mmx2
chroma_mc4_mmx_func avg, rv40, mmx2
chroma_mc2_mmx_func avg, h264, mmx2
chroma_mc8_mmx_func avg, h264, rnd_mmxext
chroma_mc8_mmx_func avg, vc1, nornd_mmxext
chroma_mc8_mmx_func avg, rv40, mmxext
chroma_mc4_mmx_func avg, h264, mmxext
chroma_mc4_mmx_func avg, rv40, mmxext
chroma_mc2_mmx_func avg, h264, mmxext

%define PAVG pavgusb
chroma_mc8_mmx_func avg, h264, rnd_3dnow


+ 2
- 2
libavcodec/x86/h264_chromamc_10bit.asm View File

@@ -253,7 +253,7 @@ INIT_XMM sse2
CHROMA_MC8 put
INIT_XMM avx
CHROMA_MC8 put
INIT_MMX mmx2
INIT_MMX mmxext
CHROMA_MC4 put
CHROMA_MC2 put

@@ -262,6 +262,6 @@ INIT_XMM sse2
CHROMA_MC8 avg
INIT_XMM avx
CHROMA_MC8 avg
INIT_MMX mmx2
INIT_MMX mmxext
CHROMA_MC4 avg
CHROMA_MC2 avg

+ 10
- 10
libavcodec/x86/h264_deblock.asm View File

@@ -504,7 +504,7 @@ cglobal deblock_h_luma_8, 0,5
RET
%endmacro ; DEBLOCK_LUMA

INIT_MMX mmx2
INIT_MMX mmxext
DEBLOCK_LUMA v8, 8
INIT_XMM sse2
DEBLOCK_LUMA v, 16
@@ -783,11 +783,11 @@ DEBLOCK_LUMA_INTRA v
INIT_XMM avx
DEBLOCK_LUMA_INTRA v
%if ARCH_X86_64 == 0
INIT_MMX mmx2
INIT_MMX mmxext
DEBLOCK_LUMA_INTRA v8
%endif

INIT_MMX mmx2
INIT_MMX mmxext

%macro CHROMA_V_START 0
dec r2d ; alpha-1
@@ -818,7 +818,7 @@ cglobal deblock_v_chroma_8, 5,6
movq m1, [t5+r1]
movq m2, [r0]
movq m3, [r0+r1]
call ff_chroma_inter_body_mmx2
call ff_chroma_inter_body_mmxext
movq [t5+r1], m1
movq [r0], m2
RET
@@ -842,7 +842,7 @@ cglobal deblock_h_chroma_8, 5,7
TRANSPOSE4x8_LOAD bw, wd, dq, PASS8ROWS(t5, r0, r1, t6)
movq buf0, m0
movq buf1, m3
call ff_chroma_inter_body_mmx2
call ff_chroma_inter_body_mmxext
movq m0, buf0
movq m3, buf1
TRANSPOSE8x4B_STORE PASS8ROWS(t5, r0, r1, t6)
@@ -852,7 +852,7 @@ cglobal deblock_h_chroma_8, 5,7
RET

ALIGN 16
ff_chroma_inter_body_mmx2:
ff_chroma_inter_body_mmxext:
LOAD_MASK r2d, r3d
movd m6, [r4] ; tc0
punpcklbw m6, m6
@@ -885,7 +885,7 @@ cglobal deblock_v_chroma_intra_8, 4,5
movq m1, [t5+r1]
movq m2, [r0]
movq m3, [r0+r1]
call ff_chroma_intra_body_mmx2
call ff_chroma_intra_body_mmxext
movq [t5+r1], m1
movq [r0], m2
RET
@@ -896,12 +896,12 @@ cglobal deblock_v_chroma_intra_8, 4,5
cglobal deblock_h_chroma_intra_8, 4,6
CHROMA_H_START
TRANSPOSE4x8_LOAD bw, wd, dq, PASS8ROWS(t5, r0, r1, t6)
call ff_chroma_intra_body_mmx2
call ff_chroma_intra_body_mmxext
TRANSPOSE8x4B_STORE PASS8ROWS(t5, r0, r1, t6)
RET

ALIGN 16
ff_chroma_intra_body_mmx2:
ff_chroma_intra_body_mmxext:
LOAD_MASK r2d, r3d
movq m5, m1
movq m6, m2
@@ -1025,7 +1025,7 @@ ff_chroma_intra_body_mmx2:
jl %%.b_idx_loop
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
cglobal h264_loop_filter_strength, 9, 9, 0, bs, nnz, ref, mv, bidir, edges, \
step, mask_mv0, mask_mv1, field
%define b_idxq bidirq


+ 2
- 2
libavcodec/x86/h264_deblock_10bit.asm View File

@@ -791,7 +791,7 @@ cglobal deblock_h_luma_intra_10, 4,7,8*(mmsize/16)
%endmacro

%if ARCH_X86_64 == 0
INIT_MMX mmx2
INIT_MMX mmxext
DEBLOCK_LUMA
DEBLOCK_LUMA_INTRA
INIT_XMM sse2
@@ -906,7 +906,7 @@ cglobal deblock_v_chroma_intra_10, 4,6-(mmsize/16),8*(mmsize/16)
%endmacro

%if ARCH_X86_64 == 0
INIT_MMX mmx2
INIT_MMX mmxext
DEBLOCK_CHROMA
%endif
INIT_XMM sse2


+ 24
- 22
libavcodec/x86/h264_idct.asm View File

@@ -286,14 +286,14 @@ cglobal h264_idct8_add_8_sse2, 3, 4, 10
%endmacro

INIT_MMX
; ff_h264_idct_dc_add_mmx2(uint8_t *dst, int16_t *block, int stride)
cglobal h264_idct_dc_add_8_mmx2, 3, 3, 0
; ff_h264_idct_dc_add_mmxext(uint8_t *dst, int16_t *block, int stride)
cglobal h264_idct_dc_add_8_mmxext, 3, 3, 0
DC_ADD_MMXEXT_INIT r1, r2
DC_ADD_MMXEXT_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_8_mmx2, 3, 3, 0
; ff_h264_idct8_dc_add_mmxext(uint8_t *dst, int16_t *block, int stride)
cglobal h264_idct8_dc_add_8_mmxext, 3, 3, 0
DC_ADD_MMXEXT_INIT r1, r2
DC_ADD_MMXEXT_OP mova, r0, r2, r1
lea r0, [r0+r2*4]
@@ -354,9 +354,9 @@ cglobal h264_idct8_add4_8_mmx, 5, 7 + npicregs, 0, dst, block_offset, block, str
ADD rsp, pad
RET

; 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_8_mmx2, 5, 8 + npicregs, 0, dst1, block_offset, block, stride, nnzc, cntr, coeff, dst2, picreg
; ff_h264_idct_add16_mmxext(uint8_t *dst, const int *block_offset,
; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
cglobal h264_idct_add16_8_mmxext, 5, 8 + npicregs, 0, dst1, block_offset, block, stride, nnzc, cntr, coeff, dst2, picreg
xor r5, r5
%ifdef PIC
lea picregq, [scan8_mem]
@@ -421,9 +421,10 @@ cglobal h264_idct_add16intra_8_mmx, 5, 7 + npicregs, 0, dst, block_offset, block
jl .nextblock
REP_RET

; 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_8_mmx2, 5, 8 + npicregs, 0, dst1, block_offset, block, stride, nnzc, cntr, coeff, dst2, picreg
; ff_h264_idct_add16intra_mmxext(uint8_t *dst, const int *block_offset,
; DCTELEM *block, int stride,
; const uint8_t nnzc[6*8])
cglobal h264_idct_add16intra_8_mmxext, 5, 8 + npicregs, 0, dst1, block_offset, block, stride, nnzc, cntr, coeff, dst2, picreg
xor r5, r5
%ifdef PIC
lea picregq, [scan8_mem]
@@ -463,9 +464,10 @@ cglobal h264_idct_add16intra_8_mmx2, 5, 8 + npicregs, 0, dst1, block_offset, blo
jl .nextblock
REP_RET

; 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_8_mmx2, 5, 8 + npicregs, 0, dst1, block_offset, block, stride, nnzc, cntr, coeff, dst2, picreg
; ff_h264_idct8_add4_mmxext(uint8_t *dst, const int *block_offset,
; DCTELEM *block, int stride,
; const uint8_t nnzc[6*8])
cglobal h264_idct8_add4_8_mmxext, 5, 8 + npicregs, 0, dst1, block_offset, block, stride, nnzc, cntr, coeff, dst2, picreg
%assign pad 128+4-(stack_offset&7)
SUB rsp, pad

@@ -620,7 +622,7 @@ cglobal h264_idct_add8_8_mmx, 5, 8 + npicregs, 0, dst1, block_offset, block, str
call h264_idct_add8_mmx_plane
RET

h264_idct_add8_mmx2_plane:
h264_idct_add8_mmxext_plane:
.nextblock:
movzx r6, byte [scan8+r5]
movzx r6, byte [r4+r6]
@@ -661,9 +663,9 @@ h264_idct_add8_mmx2_plane:
jnz .nextblock
rep ret

; 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_8_mmx2, 5, 8 + npicregs, 0, dst1, block_offset, block, stride, nnzc, cntr, coeff, dst2, picreg
; ff_h264_idct_add8_mmxext(uint8_t **dest, const int *block_offset,
; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
cglobal h264_idct_add8_8_mmxext, 5, 8 + npicregs, 0, dst1, block_offset, block, stride, nnzc, cntr, coeff, dst2, picreg
mov r5, 16
add r2, 512
%if ARCH_X86_64
@@ -672,7 +674,7 @@ cglobal h264_idct_add8_8_mmx2, 5, 8 + npicregs, 0, dst1, block_offset, block, st
%ifdef PIC
lea picregq, [scan8_mem]
%endif
call h264_idct_add8_mmx2_plane
call h264_idct_add8_mmxext_plane
mov r5, 32
add r2, 384
%if ARCH_X86_64
@@ -680,12 +682,12 @@ cglobal h264_idct_add8_8_mmx2, 5, 8 + npicregs, 0, dst1, block_offset, block, st
%else
add r0mp, gprsize
%endif
call h264_idct_add8_mmx2_plane
call h264_idct_add8_mmxext_plane
RET

INIT_MMX
; r0 = uint8_t *dst, r2 = int16_t *block, r3 = int stride, r6=clobbered
h264_idct_dc_add8_mmx2:
h264_idct_dc_add8_mmxext:
movd m0, [r2 ] ; 0 0 X D
punpcklwd m0, [r2+32] ; x X d D
paddsw m0, [pw_32]
@@ -779,7 +781,7 @@ cglobal h264_idct_add16_8_sse2, 5, 5 + ARCH_X86_64, 8
%else
add r0, r0m
%endif
call h264_idct_dc_add8_mmx2
call h264_idct_dc_add8_mmxext
.cycle%1end:
%if %1 < 7
add r2, 64
@@ -828,7 +830,7 @@ cglobal h264_idct_add16intra_8_sse2, 5, 7 + ARCH_X86_64, 8
mov r0, [r0]
add r0, dword [r1+(%1&1)*8+64*(1+(%1>>1))]
%endif
call h264_idct_dc_add8_mmx2
call h264_idct_dc_add8_mmxext
.cycle%1end:
%if %1 == 1
add r2, 384+64


+ 1
- 1
libavcodec/x86/h264_idct_10bit.asm View File

@@ -178,7 +178,7 @@ IDCT_ADD16_10
mova [%1+%3 ], m4
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
cglobal h264_idct_dc_add_10,3,3
movd m0, [r1]
paddd m0, [pd_32]


+ 13
- 13
libavcodec/x86/h264_intrapred.asm View File

@@ -120,7 +120,7 @@ cglobal pred16x16_horizontal_8, 2,3

INIT_MMX mmx
PRED16x16_H
INIT_MMX mmx2
INIT_MMX mmxext
PRED16x16_H
INIT_XMM ssse3
PRED16x16_H
@@ -180,7 +180,7 @@ cglobal pred16x16_dc_8, 2,7
REP_RET
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
PRED16x16_DC
INIT_XMM sse2
PRED16x16_DC
@@ -229,7 +229,7 @@ cglobal pred16x16_tm_vp8_8, 2,5

INIT_MMX mmx
PRED16x16_TM
INIT_MMX mmx2
INIT_MMX mmxext
PRED16x16_TM

INIT_XMM sse2
@@ -309,14 +309,14 @@ cglobal pred16x16_plane_%1_8, 2,9,7
movhlps m1, m0
%endif
paddw m0, m1
%if cpuflag(mmx2)
%if cpuflag(mmxext)
PSHUFLW m1, m0, 0xE
%elif cpuflag(mmx)
mova m1, m0
psrlq m1, 32
%endif
paddw m0, m1
%if cpuflag(mmx2)
%if cpuflag(mmxext)
PSHUFLW m1, m0, 0x1
%elif cpuflag(mmx)
mova m1, m0
@@ -536,7 +536,7 @@ INIT_MMX mmx
H264_PRED16x16_PLANE h264
H264_PRED16x16_PLANE rv40
H264_PRED16x16_PLANE svq3
INIT_MMX mmx2
INIT_MMX mmxext
H264_PRED16x16_PLANE h264
H264_PRED16x16_PLANE rv40
H264_PRED16x16_PLANE svq3
@@ -582,7 +582,7 @@ cglobal pred8x8_plane_8, 2,9,7
paddw m0, m1

%if notcpuflag(ssse3)
%if cpuflag(mmx2)
%if cpuflag(mmxext)
PSHUFLW m1, m0, 0xE
%elif cpuflag(mmx)
mova m1, m0
@@ -591,7 +591,7 @@ cglobal pred8x8_plane_8, 2,9,7
paddw m0, m1
%endif ; !ssse3

%if cpuflag(mmx2)
%if cpuflag(mmxext)
PSHUFLW m1, m0, 0x1
%elif cpuflag(mmx)
mova m1, m0
@@ -716,7 +716,7 @@ ALIGN 16

INIT_MMX mmx
H264_PRED8x8_PLANE
INIT_MMX mmx2
INIT_MMX mmxext
H264_PRED8x8_PLANE
INIT_XMM sse2
H264_PRED8x8_PLANE
@@ -763,7 +763,7 @@ cglobal pred8x8_horizontal_8, 2,3

INIT_MMX mmx
PRED8x8_H
INIT_MMX mmx2
INIT_MMX mmxext
PRED8x8_H
INIT_MMX ssse3
PRED8x8_H
@@ -941,7 +941,7 @@ cglobal pred8x8_tm_vp8_8, 2,6

INIT_MMX mmx
PRED8x8_TM
INIT_MMX mmx2
INIT_MMX mmxext
PRED8x8_TM

INIT_XMM sse2
@@ -2442,7 +2442,7 @@ cglobal pred4x4_tm_vp8_8, 3,6
sub r3d, r4d
movd mm2, r1d
movd mm4, r3d
%if cpuflag(mmx2)
%if cpuflag(mmxext)
pshufw mm2, mm2, 0
pshufw mm4, mm4, 0
%else
@@ -2465,7 +2465,7 @@ cglobal pred4x4_tm_vp8_8, 3,6

INIT_MMX mmx
PRED4x4_TM
INIT_MMX mmx2
INIT_MMX mmxext
PRED4x4_TM

INIT_XMM ssse3


+ 10
- 10
libavcodec/x86/h264_intrapred_10bit.asm View File

@@ -182,7 +182,7 @@ PRED4x4_HD
HADDD %1, %2
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
cglobal pred4x4_dc_10, 3, 3
sub r0, r2
lea r1, [r0+r2*2]
@@ -261,7 +261,7 @@ PRED4x4_VL
;-----------------------------------------------------------------------------
; void pred4x4_horizontal_up(pixel *src, const pixel *topright, int stride)
;-----------------------------------------------------------------------------
INIT_MMX mmx2
INIT_MMX mmxext
cglobal pred4x4_horizontal_up_10, 3, 3
sub r0, r2
lea r1, [r0+r2*2]
@@ -410,7 +410,7 @@ cglobal pred8x8_dc_10, 2, 6
RET
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
PRED8x8_DC pshufw
INIT_XMM sse2
PRED8x8_DC pshuflw
@@ -524,7 +524,7 @@ cglobal pred8x8l_128_dc_10, 4, 4
RET
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
PRED8x8L_128_DC
INIT_XMM sse2
PRED8x8L_128_DC
@@ -1007,7 +1007,7 @@ cglobal pred16x16_vertical_10, 2, 3
REP_RET
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
PRED16x16_VERTICAL
INIT_XMM sse2
PRED16x16_VERTICAL
@@ -1031,7 +1031,7 @@ cglobal pred16x16_horizontal_10, 2, 3
REP_RET
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
PRED16x16_HORIZONTAL
INIT_XMM sse2
PRED16x16_HORIZONTAL
@@ -1077,7 +1077,7 @@ cglobal pred16x16_dc_10, 2, 6
REP_RET
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
PRED16x16_DC
INIT_XMM sse2
PRED16x16_DC
@@ -1109,7 +1109,7 @@ cglobal pred16x16_top_dc_10, 2, 3
REP_RET
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
PRED16x16_TOP_DC
INIT_XMM sse2
PRED16x16_TOP_DC
@@ -1146,7 +1146,7 @@ cglobal pred16x16_left_dc_10, 2, 6
REP_RET
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
PRED16x16_LEFT_DC
INIT_XMM sse2
PRED16x16_LEFT_DC
@@ -1167,7 +1167,7 @@ cglobal pred16x16_128_dc_10, 2,3
REP_RET
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
PRED16x16_128_DC
INIT_XMM sse2
PRED16x16_128_DC

+ 40
- 40
libavcodec/x86/h264_intrapred_init.c View File

@@ -27,7 +27,7 @@ void ff_pred4x4_ ## TYPE ## _ ## DEPTH ## _ ## OPT (uint8_t *src, \
const uint8_t *topright, \
ptrdiff_t stride);

PRED4x4(dc, 10, mmx2)
PRED4x4(dc, 10, mmxext)
PRED4x4(down_left, 10, sse2)
PRED4x4(down_left, 10, avx)
PRED4x4(down_right, 10, sse2)
@@ -38,7 +38,7 @@ PRED4x4(vertical_left, 10, avx)
PRED4x4(vertical_right, 10, sse2)
PRED4x4(vertical_right, 10, ssse3)
PRED4x4(vertical_right, 10, avx)
PRED4x4(horizontal_up, 10, mmx2)
PRED4x4(horizontal_up, 10, mmxext)
PRED4x4(horizontal_down, 10, sse2)
PRED4x4(horizontal_down, 10, ssse3)
PRED4x4(horizontal_down, 10, avx)
@@ -47,7 +47,7 @@ PRED4x4(horizontal_down, 10, avx)
void ff_pred8x8_ ## TYPE ## _ ## DEPTH ## _ ## OPT (uint8_t *src, \
ptrdiff_t stride);

PRED8x8(dc, 10, mmx2)
PRED8x8(dc, 10, mmxext)
PRED8x8(dc, 10, sse2)
PRED8x8(top_dc, 10, sse2)
PRED8x8(plane, 10, sse2)
@@ -62,7 +62,7 @@ void ff_pred8x8l_ ## TYPE ## _ ## DEPTH ## _ ## OPT (uint8_t *src, \

PRED8x8L(dc, 10, sse2)
PRED8x8L(dc, 10, avx)
PRED8x8L(128_dc, 10, mmx2)
PRED8x8L(128_dc, 10, mmxext)
PRED8x8L(128_dc, 10, sse2)
PRED8x8L(top_dc, 10, sse2)
PRED8x8L(top_dc, 10, avx)
@@ -88,42 +88,42 @@ PRED8x8L(horizontal_up, 10, avx)
void ff_pred16x16_ ## TYPE ## _ ## DEPTH ## _ ## OPT (uint8_t *src, \
ptrdiff_t stride);

PRED16x16(dc, 10, mmx2)
PRED16x16(dc, 10, mmxext)
PRED16x16(dc, 10, sse2)
PRED16x16(top_dc, 10, mmx2)
PRED16x16(top_dc, 10, mmxext)
PRED16x16(top_dc, 10, sse2)
PRED16x16(128_dc, 10, mmx2)
PRED16x16(128_dc, 10, mmxext)
PRED16x16(128_dc, 10, sse2)
PRED16x16(left_dc, 10, mmx2)
PRED16x16(left_dc, 10, mmxext)
PRED16x16(left_dc, 10, sse2)
PRED16x16(vertical, 10, mmx2)
PRED16x16(vertical, 10, mmxext)
PRED16x16(vertical, 10, sse2)
PRED16x16(horizontal, 10, mmx2)
PRED16x16(horizontal, 10, mmxext)
PRED16x16(horizontal, 10, sse2)

/* 8-bit versions */
PRED16x16(vertical, 8, mmx)
PRED16x16(vertical, 8, sse)
PRED16x16(horizontal, 8, mmx)
PRED16x16(horizontal, 8, mmx2)
PRED16x16(horizontal, 8, mmxext)
PRED16x16(horizontal, 8, ssse3)
PRED16x16(dc, 8, mmx2)
PRED16x16(dc, 8, mmxext)
PRED16x16(dc, 8, sse2)
PRED16x16(dc, 8, ssse3)
PRED16x16(plane_h264, 8, mmx)
PRED16x16(plane_h264, 8, mmx2)
PRED16x16(plane_h264, 8, mmxext)
PRED16x16(plane_h264, 8, sse2)
PRED16x16(plane_h264, 8, ssse3)
PRED16x16(plane_rv40, 8, mmx)
PRED16x16(plane_rv40, 8, mmx2)
PRED16x16(plane_rv40, 8, mmxext)
PRED16x16(plane_rv40, 8, sse2)
PRED16x16(plane_rv40, 8, ssse3)
PRED16x16(plane_svq3, 8, mmx)
PRED16x16(plane_svq3, 8, mmx2)
PRED16x16(plane_svq3, 8, mmxext)
PRED16x16(plane_svq3, 8, sse2)
PRED16x16(plane_svq3, 8, ssse3)
PRED16x16(tm_vp8, 8, mmx)
PRED16x16(tm_vp8, 8, mmx2)
PRED16x16(tm_vp8, 8, mmxext)
PRED16x16(tm_vp8, 8, sse2)

PRED8x8(top_dc, 8, mmxext)
@@ -131,14 +131,14 @@ PRED8x8(dc_rv40, 8, mmxext)
PRED8x8(dc, 8, mmxext)
PRED8x8(vertical, 8, mmx)
PRED8x8(horizontal, 8, mmx)
PRED8x8(horizontal, 8, mmx2)
PRED8x8(horizontal, 8, mmxext)
PRED8x8(horizontal, 8, ssse3)
PRED8x8(plane, 8, mmx)
PRED8x8(plane, 8, mmx2)
PRED8x8(plane, 8, mmxext)
PRED8x8(plane, 8, sse2)
PRED8x8(plane, 8, ssse3)
PRED8x8(tm_vp8, 8, mmx)
PRED8x8(tm_vp8, 8, mmx2)
PRED8x8(tm_vp8, 8, mmxext)
PRED8x8(tm_vp8, 8, sse2)
PRED8x8(tm_vp8, 8, ssse3)

@@ -175,7 +175,7 @@ PRED4x4(vertical_right, 8, mmxext)
PRED4x4(horizontal_up, 8, mmxext)
PRED4x4(horizontal_down, 8, mmxext)
PRED4x4(tm_vp8, 8, mmx)
PRED4x4(tm_vp8, 8, mmx2)
PRED4x4(tm_vp8, 8, mmxext)
PRED4x4(tm_vp8, 8, ssse3)
PRED4x4(vertical_vp8, 8, mmxext)

@@ -210,10 +210,10 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
}

if (EXTERNAL_MMXEXT(mm_flags)) {
h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_8_mmx2;
h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_8_mmx2;
h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_8_mmxext;
h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_8_mmxext;
if (chroma_format_idc == 1)
h->pred8x8[HOR_PRED8x8 ] = ff_pred8x8_horizontal_8_mmx2;
h->pred8x8[HOR_PRED8x8 ] = ff_pred8x8_horizontal_8_mmxext;
h->pred8x8l [TOP_DC_PRED ] = ff_pred8x8l_top_dc_8_mmxext;
h->pred8x8l [DC_PRED ] = ff_pred8x8l_dc_8_mmxext;
h->pred8x8l [HOR_PRED ] = ff_pred8x8l_horizontal_8_mmxext;
@@ -243,20 +243,20 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
}
}
if (codec_id == AV_CODEC_ID_VP8) {
h->pred16x16[PLANE_PRED8x8 ] = ff_pred16x16_tm_vp8_8_mmx2;
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_mmx2;
h->pred4x4 [TM_VP8_PRED ] = ff_pred4x4_tm_vp8_8_mmx2;
h->pred8x8 [PLANE_PRED8x8 ] = ff_pred8x8_tm_vp8_8_mmxext;
h->pred4x4 [TM_VP8_PRED ] = ff_pred4x4_tm_vp8_8_mmxext;
h->pred4x4 [VERT_PRED ] = ff_pred4x4_vertical_vp8_8_mmxext;
} else {
if (chroma_format_idc == 1)
h->pred8x8 [PLANE_PRED8x8] = ff_pred8x8_plane_8_mmx2;
h->pred8x8 [PLANE_PRED8x8] = ff_pred8x8_plane_8_mmxext;
if (codec_id == AV_CODEC_ID_SVQ3) {
h->pred16x16[PLANE_PRED8x8 ] = ff_pred16x16_plane_svq3_8_mmx2;
h->pred16x16[PLANE_PRED8x8 ] = ff_pred16x16_plane_svq3_8_mmxext;
} else if (codec_id == AV_CODEC_ID_RV40) {
h->pred16x16[PLANE_PRED8x8 ] = ff_pred16x16_plane_rv40_8_mmx2;
h->pred16x16[PLANE_PRED8x8 ] = ff_pred16x16_plane_rv40_8_mmxext;
} else {
h->pred16x16[PLANE_PRED8x8 ] = ff_pred16x16_plane_h264_8_mmx2;
h->pred16x16[PLANE_PRED8x8 ] = ff_pred16x16_plane_h264_8_mmxext;
}
}
}
@@ -320,20 +320,20 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
}
} else if (bit_depth == 10) {
if (EXTERNAL_MMXEXT(mm_flags)) {
h->pred4x4[DC_PRED ] = ff_pred4x4_dc_10_mmx2;
h->pred4x4[HOR_UP_PRED ] = ff_pred4x4_horizontal_up_10_mmx2;
h->pred4x4[DC_PRED ] = ff_pred4x4_dc_10_mmxext;
h->pred4x4[HOR_UP_PRED ] = ff_pred4x4_horizontal_up_10_mmxext;

if (chroma_format_idc == 1)
h->pred8x8[DC_PRED8x8 ] = ff_pred8x8_dc_10_mmx2;
h->pred8x8[DC_PRED8x8 ] = ff_pred8x8_dc_10_mmxext;

h->pred8x8l[DC_128_PRED ] = ff_pred8x8l_128_dc_10_mmx2;
h->pred8x8l[DC_128_PRED ] = ff_pred8x8l_128_dc_10_mmxext;

h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_10_mmx2;
h->pred16x16[TOP_DC_PRED8x8 ] = ff_pred16x16_top_dc_10_mmx2;
h->pred16x16[DC_128_PRED8x8 ] = ff_pred16x16_128_dc_10_mmx2;
h->pred16x16[LEFT_DC_PRED8x8 ] = ff_pred16x16_left_dc_10_mmx2;
h->pred16x16[VERT_PRED8x8 ] = ff_pred16x16_vertical_10_mmx2;
h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_10_mmx2;
h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_10_mmxext;
h->pred16x16[TOP_DC_PRED8x8 ] = ff_pred16x16_top_dc_10_mmxext;
h->pred16x16[DC_128_PRED8x8 ] = ff_pred16x16_128_dc_10_mmxext;
h->pred16x16[LEFT_DC_PRED8x8 ] = ff_pred16x16_left_dc_10_mmxext;
h->pred16x16[VERT_PRED8x8 ] = ff_pred16x16_vertical_10_mmxext;
h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_10_mmxext;
}
if (EXTERNAL_SSE2(mm_flags)) {
h->pred4x4[DIAG_DOWN_LEFT_PRED ] = ff_pred4x4_down_left_10_sse2;


+ 6
- 6
libavcodec/x86/h264_weight.asm View File

@@ -71,7 +71,7 @@ SECTION .text
%endmacro

INIT_MMX
cglobal h264_weight_16_mmx2, 6, 6, 0
cglobal h264_weight_16_mmxext, 6, 6, 0
WEIGHT_SETUP
.nextrow:
WEIGHT_OP 0, 4
@@ -96,7 +96,7 @@ cglobal h264_weight_%1_%3, 6, 6, %2
%endmacro

INIT_MMX
WEIGHT_FUNC_MM 8, 0, mmx2
WEIGHT_FUNC_MM 8, 0, mmxext
INIT_XMM
WEIGHT_FUNC_MM 16, 8, sse2

@@ -121,7 +121,7 @@ cglobal h264_weight_%1_%3, 6, 6, %2
%endmacro

INIT_MMX
WEIGHT_FUNC_HALF_MM 4, 0, mmx2
WEIGHT_FUNC_HALF_MM 4, 0, mmxext
INIT_XMM
WEIGHT_FUNC_HALF_MM 8, 8, sse2

@@ -175,7 +175,7 @@ WEIGHT_FUNC_HALF_MM 8, 8, sse2
%endmacro

INIT_MMX
cglobal h264_biweight_16_mmx2, 7, 8, 0
cglobal h264_biweight_16_mmxext, 7, 8, 0
BIWEIGHT_SETUP
movifnidn r3d, r3m
.nextrow:
@@ -210,7 +210,7 @@ cglobal h264_biweight_%1_%3, 7, 8, %2
%endmacro

INIT_MMX
BIWEIGHT_FUNC_MM 8, 0, mmx2
BIWEIGHT_FUNC_MM 8, 0, mmxext
INIT_XMM
BIWEIGHT_FUNC_MM 16, 8, sse2

@@ -239,7 +239,7 @@ cglobal h264_biweight_%1_%3, 7, 8, %2
%endmacro

INIT_MMX
BIWEIGHT_FUNC_HALF_MM 4, 0, mmx2
BIWEIGHT_FUNC_HALF_MM 4, 0, mmxext
INIT_XMM
BIWEIGHT_FUNC_HALF_MM 8, 8, sse2



+ 54
- 53
libavcodec/x86/h264dsp_init.c View File

@@ -33,9 +33,9 @@ void ff_h264_idct ## NUM ## _add_ ## DEPTH ## _ ## OPT(uint8_t *dst, \

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(_dc, 8, mmxext)
IDCT_ADD_FUNC(_dc, 10, mmxext)
IDCT_ADD_FUNC(8_dc, 8, mmxext)
IDCT_ADD_FUNC(8_dc, 10, sse2)
IDCT_ADD_FUNC(8, 8, mmx)
IDCT_ADD_FUNC(8, 8, sse2)
@@ -51,16 +51,16 @@ void ff_h264_idct ## NUM ## _add ## REP ## _ ## DEPTH ## _ ## OPT \
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, mmxext)
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, mmxext)
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, mmxext)
IDCT_ADD_REP_FUNC(, 16intra, 8, sse2)
IDCT_ADD_REP_FUNC(, 16intra, 10, sse2)
IDCT_ADD_REP_FUNC(, 16, 10, avx)
@@ -73,7 +73,7 @@ void ff_h264_idct ## NUM ## _add ## REP ## _ ## DEPTH ## _ ## OPT \
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, mmxext)
IDCT_ADD_REP_FUNC2(, 8, 8, sse2)
IDCT_ADD_REP_FUNC2(, 8, 10, sse2)
IDCT_ADD_REP_FUNC2(, 8, 10, avx)
@@ -84,10 +84,11 @@ void ff_h264_luma_dc_dequant_idct_sse2(DCTELEM *output, DCTELEM *input, int qmul
/***********************************/
/* deblocking */

void ff_h264_loop_filter_strength_mmx2(int16_t bS[2][4][4], uint8_t nnz[40],
int8_t ref[2][40], int16_t mv[2][40][2],
int bidir, int edges, int step,
int mask_mv0, int mask_mv1, int field);
void ff_h264_loop_filter_strength_mmxext(int16_t bS[2][4][4], uint8_t nnz[40],
int8_t ref[2][40],
int16_t mv[2][40][2],
int bidir, int edges, int step,
int mask_mv0, int mask_mv1, int field);

#define LF_FUNC(DIR, TYPE, DEPTH, OPT) \
void ff_deblock_ ## DIR ## _ ## TYPE ## _ ## DEPTH ## _ ## OPT(uint8_t *pix, \
@@ -102,12 +103,12 @@ void ff_deblock_ ## DIR ## _ ## TYPE ## _ ## DEPTH ## _ ## OPT(uint8_t *pix, \
int beta);

#define LF_FUNCS(type, depth) \
LF_FUNC(h, chroma, depth, mmx2) \
LF_IFUNC(h, chroma_intra, depth, mmx2) \
LF_FUNC(v, chroma, depth, mmx2) \
LF_IFUNC(v, chroma_intra, depth, mmx2) \
LF_FUNC(h, luma, depth, mmx2) \
LF_IFUNC(h, luma_intra, depth, mmx2) \
LF_FUNC(h, chroma, depth, mmxext) \
LF_IFUNC(h, chroma_intra, depth, mmxext) \
LF_FUNC(v, chroma, depth, mmxext) \
LF_IFUNC(v, chroma_intra, depth, mmxext) \
LF_FUNC(h, luma, depth, mmxext) \
LF_IFUNC(h, luma_intra, depth, mmxext) \
LF_FUNC(h, luma, depth, sse2) \
LF_IFUNC(h, luma_intra, depth, sse2) \
LF_FUNC(v, luma, depth, sse2) \
@@ -129,26 +130,26 @@ LF_FUNCS(uint8_t, 8)
LF_FUNCS(uint16_t, 10)

#if ARCH_X86_32
LF_FUNC(v8, luma, 8, mmx2)
LF_FUNC(v8, luma, 8, mmxext)
static void ff_deblock_v_luma_8_mmxext(uint8_t *pix, int stride, int alpha,
int beta, int8_t *tc0)
{
if ((tc0[0] & tc0[1]) >= 0)
ff_deblock_v8_luma_8_mmx2(pix + 0, stride, alpha, beta, tc0);
ff_deblock_v8_luma_8_mmxext(pix + 0, stride, alpha, beta, tc0);
if ((tc0[2] & tc0[3]) >= 0)
ff_deblock_v8_luma_8_mmx2(pix + 8, stride, alpha, beta, tc0 + 2);
ff_deblock_v8_luma_8_mmxext(pix + 8, stride, alpha, beta, tc0 + 2);
}
LF_IFUNC(v8, luma_intra, 8, mmx2)
LF_IFUNC(v8, luma_intra, 8, mmxext)
static void ff_deblock_v_luma_intra_8_mmxext(uint8_t *pix, int stride,
int alpha, int beta)
{
ff_deblock_v8_luma_intra_8_mmx2(pix + 0, stride, alpha, beta);
ff_deblock_v8_luma_intra_8_mmx2(pix + 8, stride, alpha, beta);
ff_deblock_v8_luma_intra_8_mmxext(pix + 0, stride, alpha, beta);
ff_deblock_v8_luma_intra_8_mmxext(pix + 8, stride, alpha, beta);
}
#endif /* ARCH_X86_32 */

LF_FUNC(v, luma, 10, mmx2)
LF_IFUNC(v, luma_intra, 10, mmx2)
LF_FUNC(v, luma, 10, mmxext)
LF_IFUNC(v, luma_intra, 10, mmxext)

/***********************************/
/* weighted prediction */
@@ -165,8 +166,8 @@ void ff_h264_biweight_ ## W ## _ ## OPT(uint8_t *dst, uint8_t *src, \
int weights, int offset);

#define H264_BIWEIGHT_MMX(W) \
H264_WEIGHT(W, mmx2) \
H264_BIWEIGHT(W, mmx2)
H264_WEIGHT(W, mmxext) \
H264_BIWEIGHT(W, mmxext)

#define H264_BIWEIGHT_MMX_SSE(W) \
H264_BIWEIGHT_MMX(W) \
@@ -212,7 +213,7 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
int mm_flags = av_get_cpu_flags();

if (chroma_format_idc == 1 && EXTERNAL_MMXEXT(mm_flags))
c->h264_loop_filter_strength = ff_h264_loop_filter_strength_mmx2;
c->h264_loop_filter_strength = ff_h264_loop_filter_strength_mmxext;

if (bit_depth == 8) {
if (EXTERNAL_MMX(mm_flags)) {
@@ -230,33 +231,33 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
c->h264_luma_dc_dequant_idct = ff_h264_luma_dc_dequant_idct_mmx;

if (EXTERNAL_MMXEXT(mm_flags)) {
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_dc_add = ff_h264_idct_dc_add_8_mmxext;
c->h264_idct8_dc_add = ff_h264_idct8_dc_add_8_mmxext;
c->h264_idct_add16 = ff_h264_idct_add16_8_mmxext;
c->h264_idct8_add4 = ff_h264_idct8_add4_8_mmxext;
if (chroma_format_idc == 1)
c->h264_idct_add8 = ff_h264_idct_add8_8_mmx2;
c->h264_idct_add16intra = ff_h264_idct_add16intra_8_mmx2;
c->h264_idct_add8 = ff_h264_idct_add8_8_mmxext;
c->h264_idct_add16intra = ff_h264_idct_add16intra_8_mmxext;

c->h264_v_loop_filter_chroma = ff_deblock_v_chroma_8_mmx2;
c->h264_v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_8_mmx2;
c->h264_v_loop_filter_chroma = ff_deblock_v_chroma_8_mmxext;
c->h264_v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_8_mmxext;
if (chroma_format_idc == 1) {
c->h264_h_loop_filter_chroma = ff_deblock_h_chroma_8_mmx2;
c->h264_h_loop_filter_chroma_intra = ff_deblock_h_chroma_intra_8_mmx2;
c->h264_h_loop_filter_chroma = ff_deblock_h_chroma_8_mmxext;
c->h264_h_loop_filter_chroma_intra = ff_deblock_h_chroma_intra_8_mmxext;
}
#if ARCH_X86_32
c->h264_v_loop_filter_luma = ff_deblock_v_luma_8_mmxext;
c->h264_h_loop_filter_luma = ff_deblock_h_luma_8_mmx2;
c->h264_h_loop_filter_luma = ff_deblock_h_luma_8_mmxext;
c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_8_mmxext;
c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_8_mmx2;
c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_8_mmxext;
#endif /* ARCH_X86_32 */
c->weight_h264_pixels_tab[0] = ff_h264_weight_16_mmx2;
c->weight_h264_pixels_tab[1] = ff_h264_weight_8_mmx2;
c->weight_h264_pixels_tab[2] = ff_h264_weight_4_mmx2;
c->weight_h264_pixels_tab[0] = ff_h264_weight_16_mmxext;
c->weight_h264_pixels_tab[1] = ff_h264_weight_8_mmxext;
c->weight_h264_pixels_tab[2] = ff_h264_weight_4_mmxext;

c->biweight_h264_pixels_tab[0] = ff_h264_biweight_16_mmx2;
c->biweight_h264_pixels_tab[1] = ff_h264_biweight_8_mmx2;
c->biweight_h264_pixels_tab[2] = ff_h264_biweight_4_mmx2;
c->biweight_h264_pixels_tab[0] = ff_h264_biweight_16_mmxext;
c->biweight_h264_pixels_tab[1] = ff_h264_biweight_8_mmxext;
c->biweight_h264_pixels_tab[2] = ff_h264_biweight_4_mmxext;

if (EXTERNAL_SSE2(mm_flags)) {
c->h264_idct8_add = ff_h264_idct8_add_8_sse2;
@@ -297,14 +298,14 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
if (EXTERNAL_MMX(mm_flags)) {
if (EXTERNAL_MMXEXT(mm_flags)) {
#if ARCH_X86_32
c->h264_v_loop_filter_chroma = ff_deblock_v_chroma_10_mmx2;
c->h264_v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_10_mmx2;
c->h264_v_loop_filter_luma = ff_deblock_v_luma_10_mmx2;
c->h264_h_loop_filter_luma = ff_deblock_h_luma_10_mmx2;
c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_10_mmx2;
c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_10_mmx2;
c->h264_v_loop_filter_chroma = ff_deblock_v_chroma_10_mmxext;
c->h264_v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_10_mmxext;
c->h264_v_loop_filter_luma = ff_deblock_v_luma_10_mmxext;
c->h264_h_loop_filter_luma = ff_deblock_h_luma_10_mmxext;
c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_10_mmxext;
c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_10_mmxext;
#endif /* ARCH_X86_32 */
c->h264_idct_dc_add = ff_h264_idct_dc_add_10_mmx2;
c->h264_idct_dc_add = ff_h264_idct_dc_add_10_mmxext;
if (EXTERNAL_SSE2(mm_flags)) {
c->h264_idct_add = ff_h264_idct_add_10_sse2;
c->h264_idct8_dc_add = ff_h264_idct8_dc_add_10_sse2;


+ 1
- 1
libavcodec/x86/pngdsp.asm View File

@@ -166,7 +166,7 @@ cglobal add_png_paeth_prediction, 5, 7, %1, dst, src, top, w, bpp, end, cntr
RET
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
ADD_PAETH_PRED_FN 0

INIT_MMX ssse3


+ 3
- 3
libavcodec/x86/pngdsp_init.c View File

@@ -23,8 +23,8 @@
#include "libavutil/x86/cpu.h"
#include "libavcodec/pngdsp.h"

void ff_add_png_paeth_prediction_mmx2 (uint8_t *dst, uint8_t *src,
uint8_t *top, int w, int bpp);
void ff_add_png_paeth_prediction_mmxext(uint8_t *dst, uint8_t *src,
uint8_t *top, int w, int bpp);
void ff_add_png_paeth_prediction_ssse3(uint8_t *dst, uint8_t *src,
uint8_t *top, int w, int bpp);
void ff_add_bytes_l2_mmx (uint8_t *dst, uint8_t *src1,
@@ -41,7 +41,7 @@ void ff_pngdsp_init_x86(PNGDSPContext *dsp)
dsp->add_bytes_l2 = ff_add_bytes_l2_mmx;
#endif
if (EXTERNAL_MMXEXT(flags))
dsp->add_paeth_prediction = ff_add_png_paeth_prediction_mmx2;
dsp->add_paeth_prediction = ff_add_png_paeth_prediction_mmxext;
if (EXTERNAL_SSE2(flags))
dsp->add_bytes_l2 = ff_add_bytes_l2_sse2;
if (EXTERNAL_SSSE3(flags))


+ 3
- 3
libavcodec/x86/rv34dsp.asm View File

@@ -57,7 +57,7 @@ cglobal rv34_idct_%1, 1, 2, 0
REP_RET
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
%define IDCT_DC IDCT_DC_ROUND
rv34_idct dc
%define IDCT_DC IDCT_DC_NOROUND
@@ -133,7 +133,7 @@ cglobal rv34_idct_dc_add, 3, 3
mova mm5, [pd_512] ; 0x200
%endmacro

; ff_rv34_idct_add_mmx2(uint8_t *dst, ptrdiff_t stride, DCTELEM *block);
; ff_rv34_idct_add_mmxext(uint8_t *dst, ptrdiff_t stride, DCTELEM *block);
%macro COL_TRANSFORM 4
pshufw mm3, %2, 0xDD ; col. 1,3,1,3
pshufw %2, %2, 0x88 ; col. 0,2,0,2
@@ -154,7 +154,7 @@ cglobal rv34_idct_dc_add, 3, 3
packuswb %2, %2
movd %1, %2
%endmacro
INIT_MMX mmx2
INIT_MMX mmxext
cglobal rv34_idct_add, 3,3,0, d, s, b
ROW_TRANSFORM bq
COL_TRANSFORM [dq], mm0, [pw_col_coeffs+ 0], [pw_col_coeffs+ 8]


+ 5
- 5
libavcodec/x86/rv34dsp_init.c View File

@@ -25,11 +25,11 @@
#include "libavcodec/dsputil.h"
#include "libavcodec/rv34dsp.h"

void ff_rv34_idct_dc_mmx2(DCTELEM *block);
void ff_rv34_idct_dc_noround_mmx2(DCTELEM *block);
void ff_rv34_idct_dc_mmxext(DCTELEM *block);
void ff_rv34_idct_dc_noround_mmxext(DCTELEM *block);
void ff_rv34_idct_dc_add_mmx(uint8_t *dst, ptrdiff_t stride, int dc);
void ff_rv34_idct_dc_add_sse4(uint8_t *dst, ptrdiff_t stride, int dc);
void ff_rv34_idct_add_mmx2(uint8_t *dst, ptrdiff_t stride, DCTELEM *block);
void ff_rv34_idct_add_mmxext(uint8_t *dst, ptrdiff_t stride, DCTELEM *block);

av_cold void ff_rv34dsp_init_x86(RV34DSPContext* c, DSPContext *dsp)
{
@@ -38,8 +38,8 @@ av_cold void ff_rv34dsp_init_x86(RV34DSPContext* c, DSPContext *dsp)
if (EXTERNAL_MMX(mm_flags))
c->rv34_idct_dc_add = ff_rv34_idct_dc_add_mmx;
if (EXTERNAL_MMXEXT(mm_flags)) {
c->rv34_inv_transform_dc = ff_rv34_idct_dc_noround_mmx2;
c->rv34_idct_add = ff_rv34_idct_add_mmx2;
c->rv34_inv_transform_dc = ff_rv34_idct_dc_noround_mmxext;
c->rv34_idct_add = ff_rv34_idct_add_mmxext;
}
if (EXTERNAL_SSE4(mm_flags))
c->rv34_idct_dc_add = ff_rv34_idct_dc_add_sse4;


+ 2
- 2
libavcodec/x86/rv40dsp.asm View File

@@ -240,7 +240,7 @@ INIT_MMX mmx
FILTER_V put
FILTER_H put

INIT_MMX mmx2
INIT_MMX mmxext
FILTER_V avg
FILTER_H avg

@@ -486,7 +486,7 @@ cglobal rv40_weight_func_%1_%2, 6, 7, 8
REP_RET
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
RV40_WEIGHT rnd, 8, 3
RV40_WEIGHT rnd, 16, 4
RV40_WEIGHT nornd, 8, 3


+ 15
- 15
libavcodec/x86/rv40dsp_init.c View File

@@ -34,15 +34,15 @@
#if HAVE_YASM
void ff_put_rv40_chroma_mc8_mmx (uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);
void ff_avg_rv40_chroma_mc8_mmx2 (uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);
void ff_avg_rv40_chroma_mc8_mmxext(uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);
void ff_avg_rv40_chroma_mc8_3dnow(uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);

void ff_put_rv40_chroma_mc4_mmx (uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);
void ff_avg_rv40_chroma_mc4_mmx2 (uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);
void ff_avg_rv40_chroma_mc4_mmxext(uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);
void ff_avg_rv40_chroma_mc4_3dnow(uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);

@@ -55,7 +55,7 @@ void ff_rv40_weight_func_nornd_16_##opt(uint8_t *dst, uint8_t *src1, uint8_t *sr
int w1, int w2, ptrdiff_t stride); \
void ff_rv40_weight_func_nornd_8_##opt (uint8_t *dst, uint8_t *src1, uint8_t *src2, \
int w1, int w2, ptrdiff_t stride);
DECLARE_WEIGHT(mmx2)
DECLARE_WEIGHT(mmxext)
DECLARE_WEIGHT(sse2)
DECLARE_WEIGHT(ssse3)

@@ -150,9 +150,9 @@ QPEL_MC_DECL(avg_, _sse2)

QPEL_MC_DECL(put_, _mmx)

#define ff_put_rv40_qpel_h_mmx2 ff_put_rv40_qpel_h_mmx
#define ff_put_rv40_qpel_v_mmx2 ff_put_rv40_qpel_v_mmx
QPEL_MC_DECL(avg_, _mmx2)
#define ff_put_rv40_qpel_h_mmxext ff_put_rv40_qpel_h_mmx
#define ff_put_rv40_qpel_v_mmxext ff_put_rv40_qpel_v_mmx
QPEL_MC_DECL(avg_, _mmxext)

#define ff_put_rv40_qpel_h_3dnow ff_put_rv40_qpel_h_mmx
#define ff_put_rv40_qpel_v_3dnow ff_put_rv40_qpel_v_mmx
@@ -206,14 +206,14 @@ void ff_rv40dsp_init_x86(RV34DSPContext *c, DSPContext *dsp)
#endif
}
if (EXTERNAL_MMXEXT(mm_flags)) {
c->avg_chroma_pixels_tab[0] = ff_avg_rv40_chroma_mc8_mmx2;
c->avg_chroma_pixels_tab[1] = ff_avg_rv40_chroma_mc4_mmx2;
c->rv40_weight_pixels_tab[0][0] = ff_rv40_weight_func_rnd_16_mmx2;
c->rv40_weight_pixels_tab[0][1] = ff_rv40_weight_func_rnd_8_mmx2;
c->rv40_weight_pixels_tab[1][0] = ff_rv40_weight_func_nornd_16_mmx2;
c->rv40_weight_pixels_tab[1][1] = ff_rv40_weight_func_nornd_8_mmx2;
c->avg_chroma_pixels_tab[0] = ff_avg_rv40_chroma_mc8_mmxext;
c->avg_chroma_pixels_tab[1] = ff_avg_rv40_chroma_mc4_mmxext;
c->rv40_weight_pixels_tab[0][0] = ff_rv40_weight_func_rnd_16_mmxext;
c->rv40_weight_pixels_tab[0][1] = ff_rv40_weight_func_rnd_8_mmxext;
c->rv40_weight_pixels_tab[1][0] = ff_rv40_weight_func_nornd_16_mmxext;
c->rv40_weight_pixels_tab[1][1] = ff_rv40_weight_func_nornd_8_mmxext;
#if ARCH_X86_32
QPEL_MC_SET(avg_, _mmx2)
QPEL_MC_SET(avg_, _mmxext)
#endif
} else if (EXTERNAL_AMD3DNOW(mm_flags)) {
c->avg_chroma_pixels_tab[0] = ff_avg_rv40_chroma_mc8_3dnow;


+ 3
- 3
libavcodec/x86/vc1dsp_init.c View File

@@ -64,8 +64,8 @@ static void vc1_h_loop_filter16_sse4(uint8_t *src, int stride, int pq)

void ff_put_vc1_chroma_mc8_nornd_mmx (uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);
void ff_avg_vc1_chroma_mc8_nornd_mmx2 (uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);
void ff_avg_vc1_chroma_mc8_nornd_mmxext(uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);
void ff_avg_vc1_chroma_mc8_nornd_3dnow(uint8_t *dst, uint8_t *src,
int stride, int h, int x, int y);
void ff_put_vc1_chroma_mc8_nornd_ssse3(uint8_t *dst, uint8_t *src,
@@ -99,7 +99,7 @@ av_cold void ff_vc1dsp_init_x86(VC1DSPContext *dsp)

if (mm_flags & AV_CPU_FLAG_MMXEXT) {
ASSIGN_LF(mmxext);
dsp->avg_no_rnd_vc1_chroma_pixels_tab[0] = ff_avg_vc1_chroma_mc8_nornd_mmx2;
dsp->avg_no_rnd_vc1_chroma_pixels_tab[0] = ff_avg_vc1_chroma_mc8_nornd_mmxext;
} else if (mm_flags & AV_CPU_FLAG_3DNOW) {
dsp->avg_no_rnd_vc1_chroma_pixels_tab[0] = ff_avg_vc1_chroma_mc8_nornd_3dnow;
}


+ 2
- 2
libavcodec/x86/vp3dsp.asm View File

@@ -101,7 +101,7 @@ SECTION .text
mov [r0+r3 -1], r2w
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
cglobal vp3_v_loop_filter, 3, 4
%if ARCH_X86_64
movsxd r1, r1d
@@ -633,7 +633,7 @@ vp3_idct_funcs
movq [r0+r3 ], m5
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
cglobal vp3_idct_dc_add, 3, 4
%if ARCH_X86_64
movsxd r1, r1d


+ 9
- 7
libavcodec/x86/vp3dsp_init.c View File

@@ -31,11 +31,13 @@ void ff_vp3_idct_add_mmx(uint8_t *dest, int line_size, DCTELEM *block);
void ff_vp3_idct_put_sse2(uint8_t *dest, int line_size, DCTELEM *block);
void ff_vp3_idct_add_sse2(uint8_t *dest, int line_size, DCTELEM *block);

void ff_vp3_idct_dc_add_mmx2(uint8_t *dest, int line_size,
const DCTELEM *block);
void ff_vp3_idct_dc_add_mmxext(uint8_t *dest, int line_size,
const DCTELEM *block);

void ff_vp3_v_loop_filter_mmx2(uint8_t *src, int stride, int *bounding_values);
void ff_vp3_h_loop_filter_mmx2(uint8_t *src, int stride, int *bounding_values);
void ff_vp3_v_loop_filter_mmxext(uint8_t *src, int stride,
int *bounding_values);
void ff_vp3_h_loop_filter_mmxext(uint8_t *src, int stride,
int *bounding_values);

av_cold void ff_vp3dsp_init_x86(VP3DSPContext *c, int flags)
{
@@ -50,11 +52,11 @@ av_cold void ff_vp3dsp_init_x86(VP3DSPContext *c, int flags)
#endif

if (EXTERNAL_MMXEXT(cpuflags)) {
c->idct_dc_add = ff_vp3_idct_dc_add_mmx2;
c->idct_dc_add = ff_vp3_idct_dc_add_mmxext;

if (!(flags & CODEC_FLAG_BITEXACT)) {
c->v_loop_filter = ff_vp3_v_loop_filter_mmx2;
c->h_loop_filter = ff_vp3_h_loop_filter_mmx2;
c->v_loop_filter = ff_vp3_v_loop_filter_mmxext;
c->h_loop_filter = ff_vp3_h_loop_filter_mmxext;
}
}



+ 15
- 15
libavcodec/x86/vp8dsp.asm View File

@@ -338,7 +338,7 @@ INIT_XMM ssse3
FILTER_SSSE3 8

; 4x4 block, H-only 4-tap filter
INIT_MMX mmx2
INIT_MMX mmxext
cglobal put_vp8_epel4_h4, 6, 6 + npicregs, 0, dst, dststride, src, srcstride, height, mx, picreg
shl mxd, 4
%ifdef PIC
@@ -386,7 +386,7 @@ cglobal put_vp8_epel4_h4, 6, 6 + npicregs, 0, dst, dststride, src, srcstride, he
REP_RET

; 4x4 block, H-only 6-tap filter
INIT_MMX mmx2
INIT_MMX mmxext
cglobal put_vp8_epel4_h6, 6, 6 + npicregs, 0, dst, dststride, src, srcstride, height, mx, picreg
lea mxd, [mxq*3]
%ifdef PIC
@@ -673,7 +673,7 @@ cglobal put_vp8_epel%1_v6, 7, 7, 8, dst, dststride, src, srcstride, height, picr
REP_RET
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
FILTER_V 4
INIT_XMM sse2
FILTER_V 8
@@ -769,7 +769,7 @@ cglobal put_vp8_bilinear%1_h, 6, 6 + npicregs, 7, dst, dststride, src, srcstride
REP_RET
%endmacro

INIT_MMX mmx2
INIT_MMX mmxext
FILTER_BILINEAR 4
INIT_XMM sse2
FILTER_BILINEAR 8
@@ -1611,7 +1611,7 @@ cglobal vp8_%1_loop_filter_simple, 3, %2, 8, dst, stride, flim, cntr
INIT_MMX mmx
SIMPLE_LOOPFILTER v, 4
SIMPLE_LOOPFILTER h, 5
INIT_MMX mmx2
INIT_MMX mmxext
SIMPLE_LOOPFILTER v, 4
SIMPLE_LOOPFILTER h, 5
%endif
@@ -1835,7 +1835,7 @@ cglobal vp8_%1_loop_filter16y_inner, 5, 5, 13, dst, stride, flimE, flimI, hevthr
psubusb m6, m5 ; q2-q1
por m6, m4 ; abs(q2-q1)

%if notcpuflag(mmx2)
%if notcpuflag(mmxext)
mova m4, m_flimI
pxor m3, m3
psubusb m0, m4
@@ -1875,7 +1875,7 @@ cglobal vp8_%1_loop_filter16y_inner, 5, 5, 13, dst, stride, flimE, flimI, hevthr
psubusb m1, m3 ; p1-p0
psubusb m6, m2 ; p0-p1
por m1, m6 ; abs(p1-p0)
%if notcpuflag(mmx2)
%if notcpuflag(mmxext)
mova m6, m1
psubusb m1, m4
psubusb m6, m_hevthr
@@ -1906,7 +1906,7 @@ cglobal vp8_%1_loop_filter16y_inner, 5, 5, 13, dst, stride, flimE, flimI, hevthr
psubusb m1, m5 ; q0-q1
psubusb m7, m4 ; q1-q0
por m1, m7 ; abs(q1-q0)
%if notcpuflag(mmx2)
%if notcpuflag(mmxext)
mova m7, m1
psubusb m1, m6
psubusb m7, m_hevthr
@@ -2014,14 +2014,14 @@ cglobal vp8_%1_loop_filter16y_inner, 5, 5, 13, dst, stride, flimE, flimI, hevthr
%else
mova m6, m_maskres
%endif
%if notcpuflag(mmx2)
%if notcpuflag(mmxext)
mova m7, [pb_1]
%else ; mmxext/sse2
pxor m7, m7
%endif
pand m0, m6
pand m1, m6
%if notcpuflag(mmx2)
%if notcpuflag(mmxext)
paddusb m0, m7
pand m1, [pb_FE]
pandn m7, m0
@@ -2097,7 +2097,7 @@ INNER_LOOPFILTER h, 16
INNER_LOOPFILTER v, 8
INNER_LOOPFILTER h, 8

INIT_MMX mmx2
INIT_MMX mmxext
INNER_LOOPFILTER v, 16
INNER_LOOPFILTER h, 16
INNER_LOOPFILTER v, 8
@@ -2343,7 +2343,7 @@ cglobal vp8_%1_loop_filter16y_mbedge, 5, 5, 15, dst1, stride, flimE, flimI, hevt
psubusb m6, m5 ; q2-q1
por m6, m4 ; abs(q2-q1)

%if notcpuflag(mmx2)
%if notcpuflag(mmxext)
mova m4, m_flimI
pxor m3, m3
psubusb m0, m4
@@ -2383,7 +2383,7 @@ cglobal vp8_%1_loop_filter16y_mbedge, 5, 5, 15, dst1, stride, flimE, flimI, hevt
psubusb m1, m3 ; p1-p0
psubusb m6, m2 ; p0-p1
por m1, m6 ; abs(p1-p0)
%if notcpuflag(mmx2)
%if notcpuflag(mmxext)
mova m6, m1
psubusb m1, m4
psubusb m6, m_hevthr
@@ -2414,7 +2414,7 @@ cglobal vp8_%1_loop_filter16y_mbedge, 5, 5, 15, dst1, stride, flimE, flimI, hevt
psubusb m1, m5 ; q0-q1
psubusb m7, m4 ; q1-q0
por m1, m7 ; abs(q1-q0)
%if notcpuflag(mmx2)
%if notcpuflag(mmxext)
mova m7, m1
psubusb m1, m6
psubusb m7, m_hevthr
@@ -2755,7 +2755,7 @@ MBEDGE_LOOPFILTER h, 16
MBEDGE_LOOPFILTER v, 8
MBEDGE_LOOPFILTER h, 8

INIT_MMX mmx2
INIT_MMX mmxext
MBEDGE_LOOPFILTER v, 16
MBEDGE_LOOPFILTER h, 16
MBEDGE_LOOPFILTER v, 8


+ 43
- 43
libavcodec/x86/vp8dsp_init.c View File

@@ -30,16 +30,16 @@
/*
* MC functions
*/
extern void ff_put_vp8_epel4_h4_mmx2 (uint8_t *dst, ptrdiff_t dststride,
extern void ff_put_vp8_epel4_h4_mmxext(uint8_t *dst, ptrdiff_t dststride,
uint8_t *src, ptrdiff_t srcstride,
int height, int mx, int my);
extern void ff_put_vp8_epel4_h6_mmx2 (uint8_t *dst, ptrdiff_t dststride,
extern void ff_put_vp8_epel4_h6_mmxext(uint8_t *dst, ptrdiff_t dststride,
uint8_t *src, ptrdiff_t srcstride,
int height, int mx, int my);
extern void ff_put_vp8_epel4_v4_mmx2 (uint8_t *dst, ptrdiff_t dststride,
extern void ff_put_vp8_epel4_v4_mmxext(uint8_t *dst, ptrdiff_t dststride,
uint8_t *src, ptrdiff_t srcstride,
int height, int mx, int my);
extern void ff_put_vp8_epel4_v6_mmx2 (uint8_t *dst, ptrdiff_t dststride,
extern void ff_put_vp8_epel4_v6_mmxext(uint8_t *dst, ptrdiff_t dststride,
uint8_t *src, ptrdiff_t srcstride,
int height, int mx, int my);

@@ -81,7 +81,7 @@ extern void ff_put_vp8_epel8_v6_ssse3 (uint8_t *dst, ptrdiff_t dststride,
uint8_t *src, ptrdiff_t srcstride,
int height, int mx, int my);

extern void ff_put_vp8_bilinear4_h_mmx2 (uint8_t *dst, ptrdiff_t dststride,
extern void ff_put_vp8_bilinear4_h_mmxext(uint8_t *dst, ptrdiff_t dststride,
uint8_t *src, ptrdiff_t srcstride,
int height, int mx, int my);
extern void ff_put_vp8_bilinear8_h_sse2 (uint8_t *dst, ptrdiff_t dststride,
@@ -94,7 +94,7 @@ extern void ff_put_vp8_bilinear8_h_ssse3 (uint8_t *dst, ptrdiff_t dststride,
uint8_t *src, ptrdiff_t srcstride,
int height, int mx, int my);

extern void ff_put_vp8_bilinear4_v_mmx2 (uint8_t *dst, ptrdiff_t dststride,
extern void ff_put_vp8_bilinear4_v_mmxext(uint8_t *dst, ptrdiff_t dststride,
uint8_t *src, ptrdiff_t srcstride,
int height, int mx, int my);
extern void ff_put_vp8_bilinear8_v_sse2 (uint8_t *dst, ptrdiff_t dststride,
@@ -140,16 +140,16 @@ static void ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \
}

#if ARCH_X86_32
TAP_W8 (mmx2, epel, h4)
TAP_W8 (mmx2, epel, h6)
TAP_W16(mmx2, epel, h6)
TAP_W8 (mmx2, epel, v4)
TAP_W8 (mmx2, epel, v6)
TAP_W16(mmx2, epel, v6)
TAP_W8 (mmx2, bilinear, h)
TAP_W16(mmx2, bilinear, h)
TAP_W8 (mmx2, bilinear, v)
TAP_W16(mmx2, bilinear, v)
TAP_W8 (mmxext, epel, h4)
TAP_W8 (mmxext, epel, h6)
TAP_W16(mmxext, epel, h6)
TAP_W8 (mmxext, epel, v4)
TAP_W8 (mmxext, epel, v6)
TAP_W16(mmxext, epel, v6)
TAP_W8 (mmxext, bilinear, h)
TAP_W16(mmxext, bilinear, h)
TAP_W8 (mmxext, bilinear, v)
TAP_W16(mmxext, bilinear, v)
#endif

TAP_W16(sse2, epel, h6)
@@ -178,13 +178,13 @@ static void ff_put_vp8_epel ## SIZE ## _h ## TAPNUMX ## v ## TAPNUMY ## _ ## OPT

#if ARCH_X86_32
#define HVTAPMMX(x, y) \
HVTAP(mmx2, 8, x, y, 4, 8) \
HVTAP(mmx2, 8, x, y, 8, 16)
HVTAP(mmxext, 8, x, y, 4, 8) \
HVTAP(mmxext, 8, x, y, 8, 16)

HVTAP(mmx2, 8, 6, 6, 16, 16)
HVTAP(mmxext, 8, 6, 6, 16, 16)
#else
#define HVTAPMMX(x, y) \
HVTAP(mmx2, 8, x, y, 4, 8)
HVTAP(mmxext, 8, x, y, 4, 8)
#endif

HVTAPMMX(4, 4)
@@ -219,10 +219,10 @@ static void ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT( \
dst, dststride, tmp, SIZE, height, mx, my); \
}

HVBILIN(mmx2, 8, 4, 8)
HVBILIN(mmxext, 8, 4, 8)
#if ARCH_X86_32
HVBILIN(mmx2, 8, 8, 16)
HVBILIN(mmx2, 8, 16, 16)
HVBILIN(mmxext, 8, 8, 16)
HVBILIN(mmxext, 8, 16, 16)
#endif
HVBILIN(sse2, 8, 8, 16)
HVBILIN(sse2, 8, 16, 16)
@@ -284,7 +284,7 @@ extern void ff_vp8_h_loop_filter8uv_mbedge_ ## NAME(uint8_t *dstU, \
int e, int i, int hvt);

DECLARE_LOOP_FILTER(mmx)
DECLARE_LOOP_FILTER(mmx2)
DECLARE_LOOP_FILTER(mmxext)
DECLARE_LOOP_FILTER(sse2)
DECLARE_LOOP_FILTER(ssse3)
DECLARE_LOOP_FILTER(sse4)
@@ -352,26 +352,26 @@ 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 (mm_flags & AV_CPU_FLAG_MMXEXT) {
VP8_MC_FUNC(2, 4, mmx2);
VP8_BILINEAR_MC_FUNC(2, 4, mmx2);
VP8_MC_FUNC(2, 4, mmxext);
VP8_BILINEAR_MC_FUNC(2, 4, mmxext);
#if ARCH_X86_32
VP8_LUMA_MC_FUNC(0, 16, mmx2);
VP8_MC_FUNC(1, 8, mmx2);
VP8_BILINEAR_MC_FUNC(0, 16, mmx2);
VP8_BILINEAR_MC_FUNC(1, 8, mmx2);
c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_mmx2;
c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_mmx2;
c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_mmx2;
c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_mmx2;
c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_mmx2;
c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_mmx2;
c->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16y_mbedge_mmx2;
c->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16y_mbedge_mmx2;
c->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_mbedge_mmx2;
c->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mbedge_mmx2;
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;
c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_mmxext;
c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_mmxext;
c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_mmxext;
c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_mmxext;
c->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16y_mbedge_mmxext;
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
}



+ 1
- 1
libavutil/x86/x86util.asm View File

@@ -555,7 +555,7 @@
%if mmsize == 16
pshuflw %1, %2, (%3)*0x55
punpcklqdq %1, %1
%elif cpuflag(mmx2)
%elif cpuflag(mmxext)
pshufw %1, %2, (%3)*0x55
%else
%ifnidn %1, %2


+ 2
- 2
libswscale/x86/output.asm View File

@@ -247,7 +247,7 @@ cglobal yuv2planeX_%1, %3, 8, %2, filter, fltsize, src, dst, w, dither, offset
%endmacro

%if ARCH_X86_32
INIT_MMX mmx2
INIT_MMX mmxext
yuv2planeX_fn 8, 0, 7
yuv2planeX_fn 9, 0, 5
yuv2planeX_fn 10, 0, 5
@@ -388,7 +388,7 @@ INIT_MMX mmx
yuv2plane1_fn 8, 0, 5
yuv2plane1_fn 16, 0, 3

INIT_MMX mmx2
INIT_MMX mmxext
yuv2plane1_fn 9, 0, 3
yuv2plane1_fn 10, 0, 3
%endif


+ 4
- 4
libswscale/x86/swscale.c View File

@@ -250,7 +250,7 @@ extern void ff_yuv2planeX_ ## size ## _ ## opt(const int16_t *filter, int filter
VSCALEX_FUNC(10, opt)

#if ARCH_X86_32
VSCALEX_FUNCS(mmx2);
VSCALEX_FUNCS(mmxext);
#endif
VSCALEX_FUNCS(sse2);
VSCALEX_FUNCS(sse4);
@@ -267,7 +267,7 @@ extern void ff_yuv2plane1_ ## size ## _ ## opt(const int16_t *src, uint8_t *dst,
VSCALE_FUNC(16, opt1)

#if ARCH_X86_32
VSCALE_FUNCS(mmx, mmx2);
VSCALE_FUNCS(mmx, mmxext);
#endif
VSCALE_FUNCS(sse2, sse2);
VSCALE_FUNC(16, sse4);
@@ -360,7 +360,7 @@ switch(c->dstBpc){ \
if (EXTERNAL_MMX(cpu_flags)) {
ASSIGN_MMX_SCALE_FUNC(c->hyScale, c->hLumFilterSize, mmx, mmx);
ASSIGN_MMX_SCALE_FUNC(c->hcScale, c->hChrFilterSize, mmx, mmx);
ASSIGN_VSCALE_FUNC(c->yuv2plane1, mmx, mmx2, cpu_flags & AV_CPU_FLAG_MMXEXT);
ASSIGN_VSCALE_FUNC(c->yuv2plane1, mmx, mmxext, cpu_flags & AV_CPU_FLAG_MMXEXT);

switch (c->srcFormat) {
case AV_PIX_FMT_Y400A:
@@ -393,7 +393,7 @@ switch(c->dstBpc){ \
}
}
if (EXTERNAL_MMXEXT(cpu_flags)) {
ASSIGN_VSCALEX_FUNC(c->yuv2planeX, mmx2, , 1);
ASSIGN_VSCALEX_FUNC(c->yuv2planeX, mmxext, , 1);
}
#endif /* ARCH_X86_32 */
#define ASSIGN_SSE_SCALE_FUNC(hscalefn, filtersize, opt1, opt2) \


Write Preview
Loading…
Cancel
Save