提交 63ca0fe8 编写于 作者: S Seppo Tomperi 提交者: Michael Niedermayer

avcodec/hevcdsp: ARM NEON optimized qpel functions

uses comma as macro parameter separator
Signed-off-by: NMichael Niedermayer <michaelni@gmx.at>
上级 d1e2d59f
......@@ -138,7 +138,8 @@ NEON-OBJS-$(CONFIG_DCA_DECODER) += arm/dcadsp_neon.o \
arm/synth_filter_neon.o
NEON-OBJS-$(CONFIG_HEVC_DECODER) += arm/hevcdsp_init_neon.o \
arm/hevcdsp_deblock_neon.o \
arm/hevcdsp_idct_neon.o
arm/hevcdsp_idct_neon.o \
arm/hevcdsp_qpel_neon.o
NEON-OBJS-$(CONFIG_RV30_DECODER) += arm/rv34dsp_neon.o
NEON-OBJS-$(CONFIG_RV40_DECODER) += arm/rv34dsp_neon.o \
arm/rv40dsp_neon.o
......
......@@ -42,10 +42,110 @@ void ff_hevc_transform_add_16x16_neon_8(uint8_t *_dst, int16_t *coeffs,
void ff_hevc_transform_add_32x32_neon_8(uint8_t *_dst, int16_t *coeffs,
ptrdiff_t stride);
#define PUT_PIXELS(name) \
void name(int16_t *dst, uint8_t *src, \
ptrdiff_t srcstride, int height, \
intptr_t mx, intptr_t my, int width)
PUT_PIXELS(ff_hevc_put_pixels_w2_neon_8);
PUT_PIXELS(ff_hevc_put_pixels_w4_neon_8);
PUT_PIXELS(ff_hevc_put_pixels_w6_neon_8);
PUT_PIXELS(ff_hevc_put_pixels_w8_neon_8);
PUT_PIXELS(ff_hevc_put_pixels_w12_neon_8);
PUT_PIXELS(ff_hevc_put_pixels_w16_neon_8);
PUT_PIXELS(ff_hevc_put_pixels_w24_neon_8);
PUT_PIXELS(ff_hevc_put_pixels_w32_neon_8);
PUT_PIXELS(ff_hevc_put_pixels_w48_neon_8);
PUT_PIXELS(ff_hevc_put_pixels_w64_neon_8);
#undef PUT_PIXELS
static void (*put_hevc_qpel_neon[4][4])(int16_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride,
int height, int width);
static void (*put_hevc_qpel_uw_neon[4][4])(uint8_t *dst, ptrdiff_t dststride, uint8_t *_src, ptrdiff_t _srcstride,
int width, int height, int16_t* src2, ptrdiff_t src2stride);
void ff_hevc_put_qpel_neon_wrapper(int16_t *dst, uint8_t *src, ptrdiff_t srcstride,
int height, intptr_t mx, intptr_t my, int width);
void ff_hevc_put_qpel_uni_neon_wrapper(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride,
int height, intptr_t mx, intptr_t my, int width);
void ff_hevc_put_qpel_bi_neon_wrapper(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride,
int16_t *src2,
int height, intptr_t mx, intptr_t my, int width);
#define QPEL_FUNC(name) \
void name(int16_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride, \
int height, int width)
QPEL_FUNC(ff_hevc_put_qpel_v1_neon_8);
QPEL_FUNC(ff_hevc_put_qpel_v2_neon_8);
QPEL_FUNC(ff_hevc_put_qpel_v3_neon_8);
QPEL_FUNC(ff_hevc_put_qpel_h1_neon_8);
QPEL_FUNC(ff_hevc_put_qpel_h2_neon_8);
QPEL_FUNC(ff_hevc_put_qpel_h3_neon_8);
QPEL_FUNC(ff_hevc_put_qpel_h1v1_neon_8);
QPEL_FUNC(ff_hevc_put_qpel_h1v2_neon_8);
QPEL_FUNC(ff_hevc_put_qpel_h1v3_neon_8);
QPEL_FUNC(ff_hevc_put_qpel_h2v1_neon_8);
QPEL_FUNC(ff_hevc_put_qpel_h2v2_neon_8);
QPEL_FUNC(ff_hevc_put_qpel_h2v3_neon_8);
QPEL_FUNC(ff_hevc_put_qpel_h3v1_neon_8);
QPEL_FUNC(ff_hevc_put_qpel_h3v2_neon_8);
QPEL_FUNC(ff_hevc_put_qpel_h3v3_neon_8);
#undef QPEL_FUNC
#define QPEL_FUNC_UW_PIX(name) \
void name(uint8_t *dst, ptrdiff_t dststride, uint8_t *_src, ptrdiff_t _srcstride, \
int height, intptr_t mx, intptr_t my, int width);
QPEL_FUNC_UW_PIX(ff_hevc_put_qpel_uw_pixels_w4_neon_8);
QPEL_FUNC_UW_PIX(ff_hevc_put_qpel_uw_pixels_w8_neon_8);
QPEL_FUNC_UW_PIX(ff_hevc_put_qpel_uw_pixels_w16_neon_8);
QPEL_FUNC_UW_PIX(ff_hevc_put_qpel_uw_pixels_w24_neon_8);
QPEL_FUNC_UW_PIX(ff_hevc_put_qpel_uw_pixels_w32_neon_8);
QPEL_FUNC_UW_PIX(ff_hevc_put_qpel_uw_pixels_w48_neon_8);
QPEL_FUNC_UW_PIX(ff_hevc_put_qpel_uw_pixels_w64_neon_8);
#undef QPEL_FUNC_UW_PIX
#define QPEL_FUNC_UW(name) \
void name(uint8_t *dst, ptrdiff_t dststride, uint8_t *_src, ptrdiff_t _srcstride, \
int width, int height, int16_t* src2, ptrdiff_t src2stride);
QPEL_FUNC_UW(ff_hevc_put_qpel_uw_pixels_neon_8);
QPEL_FUNC_UW(ff_hevc_put_qpel_uw_v1_neon_8);
QPEL_FUNC_UW(ff_hevc_put_qpel_uw_v2_neon_8);
QPEL_FUNC_UW(ff_hevc_put_qpel_uw_v3_neon_8);
QPEL_FUNC_UW(ff_hevc_put_qpel_uw_h1_neon_8);
QPEL_FUNC_UW(ff_hevc_put_qpel_uw_h2_neon_8);
QPEL_FUNC_UW(ff_hevc_put_qpel_uw_h3_neon_8);
QPEL_FUNC_UW(ff_hevc_put_qpel_uw_h1v1_neon_8);
QPEL_FUNC_UW(ff_hevc_put_qpel_uw_h1v2_neon_8);
QPEL_FUNC_UW(ff_hevc_put_qpel_uw_h1v3_neon_8);
QPEL_FUNC_UW(ff_hevc_put_qpel_uw_h2v1_neon_8);
QPEL_FUNC_UW(ff_hevc_put_qpel_uw_h2v2_neon_8);
QPEL_FUNC_UW(ff_hevc_put_qpel_uw_h2v3_neon_8);
QPEL_FUNC_UW(ff_hevc_put_qpel_uw_h3v1_neon_8);
QPEL_FUNC_UW(ff_hevc_put_qpel_uw_h3v2_neon_8);
QPEL_FUNC_UW(ff_hevc_put_qpel_uw_h3v3_neon_8);
#undef QPEL_FUNC_UW
void ff_hevc_put_qpel_neon_wrapper(int16_t *dst, uint8_t *src, ptrdiff_t srcstride,
int height, intptr_t mx, intptr_t my, int width) {
put_hevc_qpel_neon[my][mx](dst, MAX_PB_SIZE, src, srcstride, height, width);
}
void ff_hevc_put_qpel_uni_neon_wrapper(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride,
int height, intptr_t mx, intptr_t my, int width) {
put_hevc_qpel_uw_neon[my][mx](dst, dststride, src, srcstride, width, height, NULL, 0);
}
void ff_hevc_put_qpel_bi_neon_wrapper(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride,
int16_t *src2,
int height, intptr_t mx, intptr_t my, int width) {
put_hevc_qpel_uw_neon[my][mx](dst, dststride, src, srcstride, width, height, src2, MAX_PB_SIZE);
}
static av_cold void hevcdsp_init_neon(HEVCDSPContext *c, const int bit_depth)
{
#if HAVE_NEON
if (bit_depth == 8) {
int x;
c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_neon;
c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_neon;
c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_neon;
......@@ -61,6 +161,65 @@ static av_cold void hevcdsp_init_neon(HEVCDSPContext *c, const int bit_depth)
c->transform_add[2] = ff_hevc_transform_add_16x16_neon_8;
c->transform_add[3] = ff_hevc_transform_add_32x32_neon_8;
c->idct_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[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[0][1] = ff_hevc_put_qpel_h1_neon_8;
put_hevc_qpel_neon[0][2] = ff_hevc_put_qpel_h2_neon_8;
put_hevc_qpel_neon[0][3] = ff_hevc_put_qpel_h3_neon_8;
put_hevc_qpel_neon[1][1] = ff_hevc_put_qpel_h1v1_neon_8;
put_hevc_qpel_neon[1][2] = ff_hevc_put_qpel_h2v1_neon_8;
put_hevc_qpel_neon[1][3] = ff_hevc_put_qpel_h3v1_neon_8;
put_hevc_qpel_neon[2][1] = ff_hevc_put_qpel_h1v2_neon_8;
put_hevc_qpel_neon[2][2] = ff_hevc_put_qpel_h2v2_neon_8;
put_hevc_qpel_neon[2][3] = ff_hevc_put_qpel_h3v2_neon_8;
put_hevc_qpel_neon[3][1] = ff_hevc_put_qpel_h1v3_neon_8;
put_hevc_qpel_neon[3][2] = ff_hevc_put_qpel_h2v3_neon_8;
put_hevc_qpel_neon[3][3] = ff_hevc_put_qpel_h3v3_neon_8;
put_hevc_qpel_uw_neon[1][0] = ff_hevc_put_qpel_uw_v1_neon_8;
put_hevc_qpel_uw_neon[2][0] = ff_hevc_put_qpel_uw_v2_neon_8;
put_hevc_qpel_uw_neon[3][0] = ff_hevc_put_qpel_uw_v3_neon_8;
put_hevc_qpel_uw_neon[0][1] = ff_hevc_put_qpel_uw_h1_neon_8;
put_hevc_qpel_uw_neon[0][2] = ff_hevc_put_qpel_uw_h2_neon_8;
put_hevc_qpel_uw_neon[0][3] = ff_hevc_put_qpel_uw_h3_neon_8;
put_hevc_qpel_uw_neon[1][1] = ff_hevc_put_qpel_uw_h1v1_neon_8;
put_hevc_qpel_uw_neon[1][2] = ff_hevc_put_qpel_uw_h2v1_neon_8;
put_hevc_qpel_uw_neon[1][3] = ff_hevc_put_qpel_uw_h3v1_neon_8;
put_hevc_qpel_uw_neon[2][1] = ff_hevc_put_qpel_uw_h1v2_neon_8;
put_hevc_qpel_uw_neon[2][2] = ff_hevc_put_qpel_uw_h2v2_neon_8;
put_hevc_qpel_uw_neon[2][3] = ff_hevc_put_qpel_uw_h3v2_neon_8;
put_hevc_qpel_uw_neon[3][1] = ff_hevc_put_qpel_uw_h1v3_neon_8;
put_hevc_qpel_uw_neon[3][2] = ff_hevc_put_qpel_uw_h2v3_neon_8;
put_hevc_qpel_uw_neon[3][3] = ff_hevc_put_qpel_uw_h3v3_neon_8;
for (x = 0; x < 10; x++) {
c->put_hevc_qpel[x][1][0] = ff_hevc_put_qpel_neon_wrapper;
c->put_hevc_qpel[x][0][1] = ff_hevc_put_qpel_neon_wrapper;
c->put_hevc_qpel[x][1][1] = ff_hevc_put_qpel_neon_wrapper;
c->put_hevc_qpel_uni[x][1][0] = ff_hevc_put_qpel_uni_neon_wrapper;
c->put_hevc_qpel_uni[x][0][1] = ff_hevc_put_qpel_uni_neon_wrapper;
c->put_hevc_qpel_uni[x][1][1] = ff_hevc_put_qpel_uni_neon_wrapper;
c->put_hevc_qpel_bi[x][1][0] = ff_hevc_put_qpel_bi_neon_wrapper;
c->put_hevc_qpel_bi[x][0][1] = ff_hevc_put_qpel_bi_neon_wrapper;
c->put_hevc_qpel_bi[x][1][1] = ff_hevc_put_qpel_bi_neon_wrapper;
}
c->put_hevc_qpel[0][0][0] = ff_hevc_put_pixels_w2_neon_8;
c->put_hevc_qpel[1][0][0] = ff_hevc_put_pixels_w4_neon_8;
c->put_hevc_qpel[2][0][0] = ff_hevc_put_pixels_w6_neon_8;
c->put_hevc_qpel[3][0][0] = ff_hevc_put_pixels_w8_neon_8;
c->put_hevc_qpel[4][0][0] = ff_hevc_put_pixels_w12_neon_8;
c->put_hevc_qpel[5][0][0] = ff_hevc_put_pixels_w16_neon_8;
c->put_hevc_qpel[6][0][0] = ff_hevc_put_pixels_w24_neon_8;
c->put_hevc_qpel[7][0][0] = ff_hevc_put_pixels_w32_neon_8;
c->put_hevc_qpel[8][0][0] = ff_hevc_put_pixels_w48_neon_8;
c->put_hevc_qpel[9][0][0] = ff_hevc_put_pixels_w64_neon_8;
c->put_hevc_qpel_uni[1][0][0] = ff_hevc_put_qpel_uw_pixels_w4_neon_8;
c->put_hevc_qpel_uni[3][0][0] = ff_hevc_put_qpel_uw_pixels_w8_neon_8;
c->put_hevc_qpel_uni[5][0][0] = ff_hevc_put_qpel_uw_pixels_w16_neon_8;
c->put_hevc_qpel_uni[6][0][0] = ff_hevc_put_qpel_uw_pixels_w24_neon_8;
c->put_hevc_qpel_uni[7][0][0] = ff_hevc_put_qpel_uw_pixels_w32_neon_8;
c->put_hevc_qpel_uni[8][0][0] = ff_hevc_put_qpel_uw_pixels_w48_neon_8;
c->put_hevc_qpel_uni[9][0][0] = ff_hevc_put_qpel_uw_pixels_w64_neon_8;
}
#endif // HAVE_NEON
}
......
此差异已折叠。
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册