Chiebot-Mirror
/
FFmpeg
mirror of https://github.com/FFmpeg/FFmpeg.git
8
0
Fork 0
Code Issues Projects Releases Wiki Activity

avcodec/la: Add LSX optimization for h264 qpel.

Browse Source 
./configure --disable-lasx
ffmpeg -i 1_h264_1080p_30fps_3Mbps.mp4 -f rawvideo -y /dev/null -an
before: 214fps
after:  274fps

Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
pull/389/head
yuanhecai 2 years ago committed by Michael Niedermayer
parent 8815a7719e
commit f6077cc666
No known key found for this signature in database
GPG Key ID: B18E8928B3948D64
7 changed files with 2561 additions and 559 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 2
      libavcodec/loongarch/Makefile
  2. 1686
      libavcodec/loongarch/h264qpel.S
  3. 74
      libavcodec/loongarch/h264qpel_init_loongarch.c
  4. 401
      libavcodec/loongarch/h264qpel_lasx.c
  5. 158
      libavcodec/loongarch/h264qpel_lasx.h
  6. 312
      libavcodec/loongarch/h264qpel_loongarch.h
  7. 487
      libavcodec/loongarch/h264qpel_lsx.c

2
libavcodec/loongarch/Makefile
Unescape View File

@ -31,5 +31,7 @@ LSX-OBJS-$(CONFIG_HEVC_DECODER) += loongarch/hevcdsp_lsx.o \
LSX-OBJS-$(CONFIG_H264DSP) += loongarch/h264idct.o \
loongarch/h264idct_loongarch.o \
loongarch/h264dsp.o
LSX-OBJS-$(CONFIG_H264QPEL) += loongarch/h264qpel.o \
loongarch/h264qpel_lsx.o
LSX-OBJS-$(CONFIG_H264CHROMA) += loongarch/h264chroma.o
LSX-OBJS-$(CONFIG_H264PRED) += loongarch/h264intrapred.o

1686
libavcodec/loongarch/h264qpel.S
View File

File diff suppressed because it is too large Load Diff

74
libavcodec/loongarch/h264qpel_init_loongarch.c
Unescape View File

@ -19,7 +19,7 @@
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "h264qpel_lasx.h"
#include "h264qpel_loongarch.h"
#include "libavutil/attributes.h"
#include "libavutil/loongarch/cpu.h"
#include "libavcodec/h264qpel.h"
@ -27,6 +27,77 @@
av_cold void ff_h264qpel_init_loongarch(H264QpelContext *c, int bit_depth)
{
int cpu_flags = av_get_cpu_flags();
if (have_lsx(cpu_flags)) {
if (8 == bit_depth) {
c->put_h264_qpel_pixels_tab[0][0] = ff_put_h264_qpel16_mc00_lsx;
c->put_h264_qpel_pixels_tab[0][1] = ff_put_h264_qpel16_mc10_lsx;
c->put_h264_qpel_pixels_tab[0][2] = ff_put_h264_qpel16_mc20_lsx;
c->put_h264_qpel_pixels_tab[0][3] = ff_put_h264_qpel16_mc30_lsx;
c->put_h264_qpel_pixels_tab[0][4] = ff_put_h264_qpel16_mc01_lsx;
c->put_h264_qpel_pixels_tab[0][5] = ff_put_h264_qpel16_mc11_lsx;
c->put_h264_qpel_pixels_tab[0][6] = ff_put_h264_qpel16_mc21_lsx;
c->put_h264_qpel_pixels_tab[0][7] = ff_put_h264_qpel16_mc31_lsx;
c->put_h264_qpel_pixels_tab[0][8] = ff_put_h264_qpel16_mc02_lsx;
c->put_h264_qpel_pixels_tab[0][9] = ff_put_h264_qpel16_mc12_lsx;
c->put_h264_qpel_pixels_tab[0][10] = ff_put_h264_qpel16_mc22_lsx;
c->put_h264_qpel_pixels_tab[0][11] = ff_put_h264_qpel16_mc32_lsx;
c->put_h264_qpel_pixels_tab[0][12] = ff_put_h264_qpel16_mc03_lsx;
c->put_h264_qpel_pixels_tab[0][13] = ff_put_h264_qpel16_mc13_lsx;
c->put_h264_qpel_pixels_tab[0][14] = ff_put_h264_qpel16_mc23_lsx;
c->put_h264_qpel_pixels_tab[0][15] = ff_put_h264_qpel16_mc33_lsx;
c->avg_h264_qpel_pixels_tab[0][0] = ff_avg_h264_qpel16_mc00_lsx;
c->avg_h264_qpel_pixels_tab[0][1] = ff_avg_h264_qpel16_mc10_lsx;
c->avg_h264_qpel_pixels_tab[0][2] = ff_avg_h264_qpel16_mc20_lsx;
c->avg_h264_qpel_pixels_tab[0][3] = ff_avg_h264_qpel16_mc30_lsx;
c->avg_h264_qpel_pixels_tab[0][4] = ff_avg_h264_qpel16_mc01_lsx;
c->avg_h264_qpel_pixels_tab[0][5] = ff_avg_h264_qpel16_mc11_lsx;
c->avg_h264_qpel_pixels_tab[0][6] = ff_avg_h264_qpel16_mc21_lsx;
c->avg_h264_qpel_pixels_tab[0][7] = ff_avg_h264_qpel16_mc31_lsx;
c->avg_h264_qpel_pixels_tab[0][8] = ff_avg_h264_qpel16_mc02_lsx;
c->avg_h264_qpel_pixels_tab[0][9] = ff_avg_h264_qpel16_mc12_lsx;
c->avg_h264_qpel_pixels_tab[0][10] = ff_avg_h264_qpel16_mc22_lsx;
c->avg_h264_qpel_pixels_tab[0][11] = ff_avg_h264_qpel16_mc32_lsx;
c->avg_h264_qpel_pixels_tab[0][12] = ff_avg_h264_qpel16_mc03_lsx;
c->avg_h264_qpel_pixels_tab[0][13] = ff_avg_h264_qpel16_mc13_lsx;
c->avg_h264_qpel_pixels_tab[0][14] = ff_avg_h264_qpel16_mc23_lsx;
c->avg_h264_qpel_pixels_tab[0][15] = ff_avg_h264_qpel16_mc33_lsx;
c->put_h264_qpel_pixels_tab[1][0] = ff_put_h264_qpel8_mc00_lsx;
c->put_h264_qpel_pixels_tab[1][1] = ff_put_h264_qpel8_mc10_lsx;
c->put_h264_qpel_pixels_tab[1][2] = ff_put_h264_qpel8_mc20_lsx;
c->put_h264_qpel_pixels_tab[1][3] = ff_put_h264_qpel8_mc30_lsx;
c->put_h264_qpel_pixels_tab[1][4] = ff_put_h264_qpel8_mc01_lsx;
c->put_h264_qpel_pixels_tab[1][5] = ff_put_h264_qpel8_mc11_lsx;
c->put_h264_qpel_pixels_tab[1][6] = ff_put_h264_qpel8_mc21_lsx;
c->put_h264_qpel_pixels_tab[1][7] = ff_put_h264_qpel8_mc31_lsx;
c->put_h264_qpel_pixels_tab[1][8] = ff_put_h264_qpel8_mc02_lsx;
c->put_h264_qpel_pixels_tab[1][9] = ff_put_h264_qpel8_mc12_lsx;
c->put_h264_qpel_pixels_tab[1][10] = ff_put_h264_qpel8_mc22_lsx;
c->put_h264_qpel_pixels_tab[1][11] = ff_put_h264_qpel8_mc32_lsx;
c->put_h264_qpel_pixels_tab[1][12] = ff_put_h264_qpel8_mc03_lsx;
c->put_h264_qpel_pixels_tab[1][13] = ff_put_h264_qpel8_mc13_lsx;
c->put_h264_qpel_pixels_tab[1][14] = ff_put_h264_qpel8_mc23_lsx;
c->put_h264_qpel_pixels_tab[1][15] = ff_put_h264_qpel8_mc33_lsx;
c->avg_h264_qpel_pixels_tab[1][0] = ff_avg_h264_qpel8_mc00_lsx;
c->avg_h264_qpel_pixels_tab[1][1] = ff_avg_h264_qpel8_mc10_lsx;
c->avg_h264_qpel_pixels_tab[1][2] = ff_avg_h264_qpel8_mc20_lsx;
c->avg_h264_qpel_pixels_tab[1][3] = ff_avg_h264_qpel8_mc30_lsx;
c->avg_h264_qpel_pixels_tab[1][5] = ff_avg_h264_qpel8_mc11_lsx;
c->avg_h264_qpel_pixels_tab[1][6] = ff_avg_h264_qpel8_mc21_lsx;
c->avg_h264_qpel_pixels_tab[1][7] = ff_avg_h264_qpel8_mc31_lsx;
c->avg_h264_qpel_pixels_tab[1][8] = ff_avg_h264_qpel8_mc02_lsx;
c->avg_h264_qpel_pixels_tab[1][9] = ff_avg_h264_qpel8_mc12_lsx;
c->avg_h264_qpel_pixels_tab[1][10] = ff_avg_h264_qpel8_mc22_lsx;
c->avg_h264_qpel_pixels_tab[1][11] = ff_avg_h264_qpel8_mc32_lsx;
c->avg_h264_qpel_pixels_tab[1][13] = ff_avg_h264_qpel8_mc13_lsx;
c->avg_h264_qpel_pixels_tab[1][14] = ff_avg_h264_qpel8_mc23_lsx;
c->avg_h264_qpel_pixels_tab[1][15] = ff_avg_h264_qpel8_mc33_lsx;
}
}
#if HAVE_LASX
if (have_lasx(cpu_flags)) {
if (8 == bit_depth) {
c->put_h264_qpel_pixels_tab[0][0] = ff_put_h264_qpel16_mc00_lasx;
@ -95,4 +166,5 @@ av_cold void ff_h264qpel_init_loongarch(H264QpelContext *c, int bit_depth)
c->avg_h264_qpel_pixels_tab[1][15] = ff_avg_h264_qpel8_mc33_lasx;
}
}
#endif
}

401
libavcodec/loongarch/h264qpel_lasx.c
Unescape View File

@ -21,7 +21,7 @@
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "h264qpel_lasx.h"
#include "h264qpel_loongarch.h"
#include "libavutil/loongarch/loongson_intrinsics.h"
#include "libavutil/attributes.h"
@ -418,157 +418,6 @@ avg_pixels8_8_lsx(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)
);
}
/* avg_pixels8_8_lsx : dst = avg(src, dst)
* put_pixels8_l2_8_lsx: dst = avg(src, half) , half stride is 8.
* avg_pixels8_l2_8_lsx: dst = avg(avg(src, half), dst) , half stride is 8.*/
static av_always_inline void
put_pixels8_l2_8_lsx(uint8_t *dst, const uint8_t *src, const uint8_t *half,
ptrdiff_t dstStride, ptrdiff_t srcStride)
{
ptrdiff_t stride_2, stride_3, stride_4;
__asm__ volatile (
/* h0~h7 */
"slli.d %[stride_2], %[srcStride], 1 \n\t"
"add.d %[stride_3], %[stride_2], %[srcStride] \n\t"
"slli.d %[stride_4], %[stride_2], 1 \n\t"
"vld $vr0, %[src], 0 \n\t"
"vldx $vr1, %[src], %[srcStride] \n\t"
"vldx $vr2, %[src], %[stride_2] \n\t"
"vldx $vr3, %[src], %[stride_3] \n\t"
"add.d %[src], %[src], %[stride_4] \n\t"
"vld $vr4, %[src], 0 \n\t"
"vldx $vr5, %[src], %[srcStride] \n\t"
"vldx $vr6, %[src], %[stride_2] \n\t"
"vldx $vr7, %[src], %[stride_3] \n\t"
"vld $vr8, %[half], 0x00 \n\t"
"vld $vr9, %[half], 0x08 \n\t"
"vld $vr10, %[half], 0x10 \n\t"
"vld $vr11, %[half], 0x18 \n\t"
"vld $vr12, %[half], 0x20 \n\t"
"vld $vr13, %[half], 0x28 \n\t"
"vld $vr14, %[half], 0x30 \n\t"
"vld $vr15, %[half], 0x38 \n\t"
"vavgr.bu $vr0, $vr8, $vr0 \n\t"
"vavgr.bu $vr1, $vr9, $vr1 \n\t"
"vavgr.bu $vr2, $vr10, $vr2 \n\t"
"vavgr.bu $vr3, $vr11, $vr3 \n\t"
"vavgr.bu $vr4, $vr12, $vr4 \n\t"
"vavgr.bu $vr5, $vr13, $vr5 \n\t"
"vavgr.bu $vr6, $vr14, $vr6 \n\t"
"vavgr.bu $vr7, $vr15, $vr7 \n\t"
"vstelm.d $vr0, %[dst], 0, 0 \n\t"
"add.d %[dst], %[dst], %[dstStride] \n\t"
"vstelm.d $vr1, %[dst], 0, 0 \n\t"
"add.d %[dst], %[dst], %[dstStride] \n\t"
"vstelm.d $vr2, %[dst], 0, 0 \n\t"
"add.d %[dst], %[dst], %[dstStride] \n\t"
"vstelm.d $vr3, %[dst], 0, 0 \n\t"
"add.d %[dst], %[dst], %[dstStride] \n\t"
"vstelm.d $vr4, %[dst], 0, 0 \n\t"
"add.d %[dst], %[dst], %[dstStride] \n\t"
"vstelm.d $vr5, %[dst], 0, 0 \n\t"
"add.d %[dst], %[dst], %[dstStride] \n\t"
"vstelm.d $vr6, %[dst], 0, 0 \n\t"
"add.d %[dst], %[dst], %[dstStride] \n\t"
"vstelm.d $vr7, %[dst], 0, 0 \n\t"
: [dst]"+&r"(dst), [half]"+&r"(half), [src]"+&r"(src),
[stride_2]"=&r"(stride_2), [stride_3]"=&r"(stride_3),
[stride_4]"=&r"(stride_4)
: [srcStride]"r"(srcStride), [dstStride]"r"(dstStride)
: "memory"
);
}
/* avg_pixels8_8_lsx : dst = avg(src, dst)
* put_pixels8_l2_8_lsx: dst = avg(src, half) , half stride is 8.
* avg_pixels8_l2_8_lsx: dst = avg(avg(src, half), dst) , half stride is 8.*/
static av_always_inline void
avg_pixels8_l2_8_lsx(uint8_t *dst, const uint8_t *src, const uint8_t *half,
ptrdiff_t dstStride, ptrdiff_t srcStride)
{
uint8_t *tmp = dst;
ptrdiff_t stride_2, stride_3, stride_4;
__asm__ volatile (
/* h0~h7 */
"slli.d %[stride_2], %[srcStride], 1 \n\t"
"add.d %[stride_3], %[stride_2], %[srcStride] \n\t"
"slli.d %[stride_4], %[stride_2], 1 \n\t"
"vld $vr0, %[src], 0 \n\t"
"vldx $vr1, %[src], %[srcStride] \n\t"
"vldx $vr2, %[src], %[stride_2] \n\t"
"vldx $vr3, %[src], %[stride_3] \n\t"
"add.d %[src], %[src], %[stride_4] \n\t"
"vld $vr4, %[src], 0 \n\t"
"vldx $vr5, %[src], %[srcStride] \n\t"
"vldx $vr6, %[src], %[stride_2] \n\t"
"vldx $vr7, %[src], %[stride_3] \n\t"
"vld $vr8, %[half], 0x00 \n\t"
"vld $vr9, %[half], 0x08 \n\t"
"vld $vr10, %[half], 0x10 \n\t"
"vld $vr11, %[half], 0x18 \n\t"
"vld $vr12, %[half], 0x20 \n\t"
"vld $vr13, %[half], 0x28 \n\t"
"vld $vr14, %[half], 0x30 \n\t"
"vld $vr15, %[half], 0x38 \n\t"
"vavgr.bu $vr0, $vr8, $vr0 \n\t"
"vavgr.bu $vr1, $vr9, $vr1 \n\t"
"vavgr.bu $vr2, $vr10, $vr2 \n\t"
"vavgr.bu $vr3, $vr11, $vr3 \n\t"
"vavgr.bu $vr4, $vr12, $vr4 \n\t"
"vavgr.bu $vr5, $vr13, $vr5 \n\t"
"vavgr.bu $vr6, $vr14, $vr6 \n\t"
"vavgr.bu $vr7, $vr15, $vr7 \n\t"
"slli.d %[stride_2], %[dstStride], 1 \n\t"
"add.d %[stride_3], %[stride_2], %[dstStride] \n\t"
"slli.d %[stride_4], %[stride_2], 1 \n\t"
"vld $vr8, %[tmp], 0 \n\t"
"vldx $vr9, %[tmp], %[dstStride] \n\t"
"vldx $vr10, %[tmp], %[stride_2] \n\t"
"vldx $vr11, %[tmp], %[stride_3] \n\t"
"add.d %[tmp], %[tmp], %[stride_4] \n\t"
"vld $vr12, %[tmp], 0 \n\t"
"vldx $vr13, %[tmp], %[dstStride] \n\t"
"vldx $vr14, %[tmp], %[stride_2] \n\t"
"vldx $vr15, %[tmp], %[stride_3] \n\t"
"vavgr.bu $vr0, $vr8, $vr0 \n\t"
"vavgr.bu $vr1, $vr9, $vr1 \n\t"
"vavgr.bu $vr2, $vr10, $vr2 \n\t"
"vavgr.bu $vr3, $vr11, $vr3 \n\t"
"vavgr.bu $vr4, $vr12, $vr4 \n\t"
"vavgr.bu $vr5, $vr13, $vr5 \n\t"
"vavgr.bu $vr6, $vr14, $vr6 \n\t"
"vavgr.bu $vr7, $vr15, $vr7 \n\t"
"vstelm.d $vr0, %[dst], 0, 0 \n\t"
"add.d %[dst], %[dst], %[dstStride] \n\t"
"vstelm.d $vr1, %[dst], 0, 0 \n\t"
"add.d %[dst], %[dst], %[dstStride] \n\t"
"vstelm.d $vr2, %[dst], 0, 0 \n\t"
"add.d %[dst], %[dst], %[dstStride] \n\t"
"vstelm.d $vr3, %[dst], 0, 0 \n\t"
"add.d %[dst], %[dst], %[dstStride] \n\t"
"vstelm.d $vr4, %[dst], 0, 0 \n\t"
"add.d %[dst], %[dst], %[dstStride] \n\t"
"vstelm.d $vr5, %[dst], 0, 0 \n\t"
"add.d %[dst], %[dst], %[dstStride] \n\t"
"vstelm.d $vr6, %[dst], 0, 0 \n\t"
"add.d %[dst], %[dst], %[dstStride] \n\t"
"vstelm.d $vr7, %[dst], 0, 0 \n\t"
: [dst]"+&r"(dst), [tmp]"+&r"(tmp), [half]"+&r"(half),
[src]"+&r"(src), [stride_2]"=&r"(stride_2),
[stride_3]"=&r"(stride_3), [stride_4]"=&r"(stride_4)
: [dstStride]"r"(dstStride), [srcStride]"r"(srcStride)
: "memory"
);
}
/* put_pixels16_8_lsx: dst = src */
static av_always_inline void
put_pixels16_8_lsx(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)
@ -729,254 +578,6 @@ avg_pixels16_8_lsx(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)
);
}
/* avg_pixels16_8_lsx : dst = avg(src, dst)
* put_pixels16_l2_8_lsx: dst = avg(src, half) , half stride is 8.
* avg_pixels16_l2_8_lsx: dst = avg(avg(src, half), dst) , half stride is 8.*/
static av_always_inline void
put_pixels16_l2_8_lsx(uint8_t *dst, const uint8_t *src, uint8_t *half,
ptrdiff_t dstStride, ptrdiff_t srcStride)
{
ptrdiff_t stride_2, stride_3, stride_4;
ptrdiff_t dstride_2, dstride_3, dstride_4;
__asm__ volatile (
"slli.d %[stride_2], %[srcStride], 1 \n\t"
"add.d %[stride_3], %[stride_2], %[srcStride] \n\t"
"slli.d %[stride_4], %[stride_2], 1 \n\t"
"slli.d %[dstride_2], %[dstStride], 1 \n\t"
"add.d %[dstride_3], %[dstride_2], %[dstStride] \n\t"
"slli.d %[dstride_4], %[dstride_2], 1 \n\t"
/* h0~h7 */
"vld $vr0, %[src], 0 \n\t"
"vldx $vr1, %[src], %[srcStride] \n\t"
"vldx $vr2, %[src], %[stride_2] \n\t"
"vldx $vr3, %[src], %[stride_3] \n\t"
"add.d %[src], %[src], %[stride_4] \n\t"
"vld $vr4, %[src], 0 \n\t"
"vldx $vr5, %[src], %[srcStride] \n\t"
"vldx $vr6, %[src], %[stride_2] \n\t"
"vldx $vr7, %[src], %[stride_3] \n\t"
"add.d %[src], %[src], %[stride_4] \n\t"
"vld $vr8, %[half], 0x00 \n\t"
"vld $vr9, %[half], 0x10 \n\t"
"vld $vr10, %[half], 0x20 \n\t"
"vld $vr11, %[half], 0x30 \n\t"
"vld $vr12, %[half], 0x40 \n\t"
"vld $vr13, %[half], 0x50 \n\t"
"vld $vr14, %[half], 0x60 \n\t"
"vld $vr15, %[half], 0x70 \n\t"
"vavgr.bu $vr0, $vr8, $vr0 \n\t"
"vavgr.bu $vr1, $vr9, $vr1 \n\t"
"vavgr.bu $vr2, $vr10, $vr2 \n\t"
"vavgr.bu $vr3, $vr11, $vr3 \n\t"
"vavgr.bu $vr4, $vr12, $vr4 \n\t"
"vavgr.bu $vr5, $vr13, $vr5 \n\t"
"vavgr.bu $vr6, $vr14, $vr6 \n\t"
"vavgr.bu $vr7, $vr15, $vr7 \n\t"
"vst $vr0, %[dst], 0 \n\t"
"vstx $vr1, %[dst], %[dstStride] \n\t"
"vstx $vr2, %[dst], %[dstride_2] \n\t"
"vstx $vr3, %[dst], %[dstride_3] \n\t"
"add.d %[dst], %[dst], %[dstride_4] \n\t"
"vst $vr4, %[dst], 0 \n\t"
"vstx $vr5, %[dst], %[dstStride] \n\t"
"vstx $vr6, %[dst], %[dstride_2] \n\t"
"vstx $vr7, %[dst], %[dstride_3] \n\t"
"add.d %[dst], %[dst], %[dstride_4] \n\t"
/* h8~h15 */
"vld $vr0, %[src], 0 \n\t"
"vldx $vr1, %[src], %[srcStride] \n\t"
"vldx $vr2, %[src], %[stride_2] \n\t"
"vldx $vr3, %[src], %[stride_3] \n\t"
"add.d %[src], %[src], %[stride_4] \n\t"
"vld $vr4, %[src], 0 \n\t"
"vldx $vr5, %[src], %[srcStride] \n\t"
"vldx $vr6, %[src], %[stride_2] \n\t"
"vldx $vr7, %[src], %[stride_3] \n\t"
"vld $vr8, %[half], 0x80 \n\t"
"vld $vr9, %[half], 0x90 \n\t"
"vld $vr10, %[half], 0xa0 \n\t"
"vld $vr11, %[half], 0xb0 \n\t"
"vld $vr12, %[half], 0xc0 \n\t"
"vld $vr13, %[half], 0xd0 \n\t"
"vld $vr14, %[half], 0xe0 \n\t"
"vld $vr15, %[half], 0xf0 \n\t"
"vavgr.bu $vr0, $vr8, $vr0 \n\t"
"vavgr.bu $vr1, $vr9, $vr1 \n\t"
"vavgr.bu $vr2, $vr10, $vr2 \n\t"
"vavgr.bu $vr3, $vr11, $vr3 \n\t"
"vavgr.bu $vr4, $vr12, $vr4 \n\t"
"vavgr.bu $vr5, $vr13, $vr5 \n\t"
"vavgr.bu $vr6, $vr14, $vr6 \n\t"
"vavgr.bu $vr7, $vr15, $vr7 \n\t"
"vst $vr0, %[dst], 0 \n\t"
"vstx $vr1, %[dst], %[dstStride] \n\t"
"vstx $vr2, %[dst], %[dstride_2] \n\t"
"vstx $vr3, %[dst], %[dstride_3] \n\t"
"add.d %[dst], %[dst], %[dstride_4] \n\t"
"vst $vr4, %[dst], 0 \n\t"
"vstx $vr5, %[dst], %[dstStride] \n\t"
"vstx $vr6, %[dst], %[dstride_2] \n\t"
"vstx $vr7, %[dst], %[dstride_3] \n\t"
: [dst]"+&r"(dst), [half]"+&r"(half), [src]"+&r"(src),
[stride_2]"=&r"(stride_2), [stride_3]"=&r"(stride_3),
[stride_4]"=&r"(stride_4), [dstride_2]"=&r"(dstride_2),
[dstride_3]"=&r"(dstride_3), [dstride_4]"=&r"(dstride_4)
: [dstStride]"r"(dstStride), [srcStride]"r"(srcStride)
: "memory"
);
}
/* avg_pixels16_8_lsx : dst = avg(src, dst)
* put_pixels16_l2_8_lsx: dst = avg(src, half) , half stride is 8.
* avg_pixels16_l2_8_lsx: dst = avg(avg(src, half), dst) , half stride is 8.*/
static av_always_inline void
avg_pixels16_l2_8_lsx(uint8_t *dst, const uint8_t *src, uint8_t *half,
ptrdiff_t dstStride, ptrdiff_t srcStride)
{
uint8_t *tmp = dst;
ptrdiff_t stride_2, stride_3, stride_4;
ptrdiff_t dstride_2, dstride_3, dstride_4;
__asm__ volatile (
"slli.d %[stride_2], %[srcStride], 1 \n\t"
"add.d %[stride_3], %[stride_2], %[srcStride] \n\t"
"slli.d %[stride_4], %[stride_2], 1 \n\t"
"slli.d %[dstride_2], %[dstStride], 1 \n\t"
"add.d %[dstride_3], %[dstride_2], %[dstStride] \n\t"
"slli.d %[dstride_4], %[dstride_2], 1 \n\t"
/* h0~h7 */
"vld $vr0, %[src], 0 \n\t"
"vldx $vr1, %[src], %[srcStride] \n\t"
"vldx $vr2, %[src], %[stride_2] \n\t"
"vldx $vr3, %[src], %[stride_3] \n\t"
"add.d %[src], %[src], %[stride_4] \n\t"
"vld $vr4, %[src], 0 \n\t"
"vldx $vr5, %[src], %[srcStride] \n\t"
"vldx $vr6, %[src], %[stride_2] \n\t"
"vldx $vr7, %[src], %[stride_3] \n\t"
"add.d %[src], %[src], %[stride_4] \n\t"
"vld $vr8, %[half], 0x00 \n\t"
"vld $vr9, %[half], 0x10 \n\t"
"vld $vr10, %[half], 0x20 \n\t"
"vld $vr11, %[half], 0x30 \n\t"
"vld $vr12, %[half], 0x40 \n\t"
"vld $vr13, %[half], 0x50 \n\t"
"vld $vr14, %[half], 0x60 \n\t"
"vld $vr15, %[half], 0x70 \n\t"
"vavgr.bu $vr0, $vr8, $vr0 \n\t"
"vavgr.bu $vr1, $vr9, $vr1 \n\t"
"vavgr.bu $vr2, $vr10, $vr2 \n\t"
"vavgr.bu $vr3, $vr11, $vr3 \n\t"
"vavgr.bu $vr4, $vr12, $vr4 \n\t"
"vavgr.bu $vr5, $vr13, $vr5 \n\t"
"vavgr.bu $vr6, $vr14, $vr6 \n\t"
"vavgr.bu $vr7, $vr15, $vr7 \n\t"
"vld $vr8, %[tmp], 0 \n\t"
"vldx $vr9, %[tmp], %[dstStride] \n\t"
"vldx $vr10, %[tmp], %[dstride_2] \n\t"
"vldx $vr11, %[tmp], %[dstride_3] \n\t"
"add.d %[tmp], %[tmp], %[dstride_4] \n\t"
"vld $vr12, %[tmp], 0 \n\t"
"vldx $vr13, %[tmp], %[dstStride] \n\t"
"vldx $vr14, %[tmp], %[dstride_2] \n\t"
"vldx $vr15, %[tmp], %[dstride_3] \n\t"
"add.d %[tmp], %[tmp], %[dstride_4] \n\t"
"vavgr.bu $vr0, $vr8, $vr0 \n\t"
"vavgr.bu $vr1, $vr9, $vr1 \n\t"
"vavgr.bu $vr2, $vr10, $vr2 \n\t"
"vavgr.bu $vr3, $vr11, $vr3 \n\t"
"vavgr.bu $vr4, $vr12, $vr4 \n\t"
"vavgr.bu $vr5, $vr13, $vr5 \n\t"
"vavgr.bu $vr6, $vr14, $vr6 \n\t"
"vavgr.bu $vr7, $vr15, $vr7 \n\t"
"vst $vr0, %[dst], 0 \n\t"
"vstx $vr1, %[dst], %[dstStride] \n\t"
"vstx $vr2, %[dst], %[dstride_2] \n\t"
"vstx $vr3, %[dst], %[dstride_3] \n\t"
"add.d %[dst], %[dst], %[dstride_4] \n\t"
"vst $vr4, %[dst], 0 \n\t"
"vstx $vr5, %[dst], %[dstStride] \n\t"
"vstx $vr6, %[dst], %[dstride_2] \n\t"
"vstx $vr7, %[dst], %[dstride_3] \n\t"
"add.d %[dst], %[dst], %[dstride_4] \n\t"
/* h8~h15 */
"vld $vr0, %[src], 0 \n\t"
"vldx $vr1, %[src], %[srcStride] \n\t"
"vldx $vr2, %[src], %[stride_2] \n\t"
"vldx $vr3, %[src], %[stride_3] \n\t"
"add.d %[src], %[src], %[stride_4] \n\t"
"vld $vr4, %[src], 0 \n\t"
"vldx $vr5, %[src], %[srcStride] \n\t"
"vldx $vr6, %[src], %[stride_2] \n\t"
"vldx $vr7, %[src], %[stride_3] \n\t"
"vld $vr8, %[half], 0x80 \n\t"
"vld $vr9, %[half], 0x90 \n\t"
"vld $vr10, %[half], 0xa0 \n\t"
"vld $vr11, %[half], 0xb0 \n\t"
"vld $vr12, %[half], 0xc0 \n\t"
"vld $vr13, %[half], 0xd0 \n\t"
"vld $vr14, %[half], 0xe0 \n\t"
"vld $vr15, %[half], 0xf0 \n\t"
"vavgr.bu $vr0, $vr8, $vr0 \n\t"
"vavgr.bu $vr1, $vr9, $vr1 \n\t"
"vavgr.bu $vr2, $vr10, $vr2 \n\t"
"vavgr.bu $vr3, $vr11, $vr3 \n\t"
"vavgr.bu $vr4, $vr12, $vr4 \n\t"
"vavgr.bu $vr5, $vr13, $vr5 \n\t"
"vavgr.bu $vr6, $vr14, $vr6 \n\t"
"vavgr.bu $vr7, $vr15, $vr7 \n\t"
"vld $vr8, %[tmp], 0 \n\t"
"vldx $vr9, %[tmp], %[dstStride] \n\t"
"vldx $vr10, %[tmp], %[dstride_2] \n\t"
"vldx $vr11, %[tmp], %[dstride_3] \n\t"
"add.d %[tmp], %[tmp], %[dstride_4] \n\t"
"vld $vr12, %[tmp], 0 \n\t"
"vldx $vr13, %[tmp], %[dstStride] \n\t"
"vldx $vr14, %[tmp], %[dstride_2] \n\t"
"vldx $vr15, %[tmp], %[dstride_3] \n\t"
"vavgr.bu $vr0, $vr8, $vr0 \n\t"
"vavgr.bu $vr1, $vr9, $vr1 \n\t"
"vavgr.bu $vr2, $vr10, $vr2 \n\t"
"vavgr.bu $vr3, $vr11, $vr3 \n\t"
"vavgr.bu $vr4, $vr12, $vr4 \n\t"
"vavgr.bu $vr5, $vr13, $vr5 \n\t"
"vavgr.bu $vr6, $vr14, $vr6 \n\t"
"vavgr.bu $vr7, $vr15, $vr7 \n\t"
"vst $vr0, %[dst], 0 \n\t"
"vstx $vr1, %[dst], %[dstStride] \n\t"
"vstx $vr2, %[dst], %[dstride_2] \n\t"
"vstx $vr3, %[dst], %[dstride_3] \n\t"
"add.d %[dst], %[dst], %[dstride_4] \n\t"
"vst $vr4, %[dst], 0 \n\t"
"vstx $vr5, %[dst], %[dstStride] \n\t"
"vstx $vr6, %[dst], %[dstride_2] \n\t"
"vstx $vr7, %[dst], %[dstride_3] \n\t"
: [dst]"+&r"(dst), [tmp]"+&r"(tmp), [half]"+&r"(half), [src]"+&r"(src),
[stride_2]"=&r"(stride_2), [stride_3]"=&r"(stride_3),
[stride_4]"=&r"(stride_4), [dstride_2]"=&r"(dstride_2),
[dstride_3]"=&r"(dstride_3), [dstride_4]"=&r"(dstride_4)
: [dstStride]"r"(dstStride), [srcStride]"r"(srcStride)
: "memory"
);
}
#define QPEL8_H_LOWPASS(out_v) \
src00 = __lasx_xvld(src, - 2); \
src += srcStride; \

158
libavcodec/loongarch/h264qpel_lasx.h
Unescape View File

@ -1,158 +0,0 @@
/*
* Copyright (c) 2020 Loongson Technology Corporation Limited
* Contributed by Shiyou Yin <yinshiyou-hf@loongson.cn>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#ifndef AVCODEC_LOONGARCH_H264QPEL_LASX_H
#define AVCODEC_LOONGARCH_H264QPEL_LASX_H
#include <stdint.h>
#include <stddef.h>
#include "libavcodec/h264.h"
void ff_h264_h_lpf_luma_inter_lasx(uint8_t *src, int stride,
int alpha, int beta, int8_t *tc0);
void ff_h264_v_lpf_luma_inter_lasx(uint8_t *src, int stride,
int alpha, int beta, int8_t *tc0);
void ff_put_h264_qpel16_mc00_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc10_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc20_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc30_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc01_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc11_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc21_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc31_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc02_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc12_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc32_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc22_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc03_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc13_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc23_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc33_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc00_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc10_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc20_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc30_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc01_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc11_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc21_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc31_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc02_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc12_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc22_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc32_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc03_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc13_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc23_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc33_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel8_mc00_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc10_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc20_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc30_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc01_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc11_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc21_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc31_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc02_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc12_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc22_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc32_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc03_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc13_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc23_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc33_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel8_mc00_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc10_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc20_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc30_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc11_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc21_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc31_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc02_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc12_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc22_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc32_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc13_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc23_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc33_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
#endif // #ifndef AVCODEC_LOONGARCH_H264QPEL_LASX_H

312
libavcodec/loongarch/h264qpel_loongarch.h
Unescape View File

@ -0,0 +1,312 @@
/*
* Copyright (c) 2023 Loongson Technology Corporation Limited
* Contributed by Shiyou Yin <yinshiyou-hf@loongson.cn>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#ifndef AVCODEC_LOONGARCH_H264QPEL_LOONGARCH_H
#define AVCODEC_LOONGARCH_H264QPEL_LOONGARCH_H
#include <stdint.h>
#include <stddef.h>
#include "libavcodec/h264.h"
#include "config.h"
void put_h264_qpel8_hv_lowpass_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dstStride, ptrdiff_t srcStride);
void put_h264_qpel8_h_lowpass_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dstStride, ptrdiff_t srcStride);
void put_h264_qpel8_v_lowpass_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dstStride, ptrdiff_t srcStride);
void put_pixels16_l2_8_lsx(uint8_t *dst, const uint8_t *src, uint8_t *half,
ptrdiff_t dstStride, ptrdiff_t srcStride);
void put_pixels8_l2_8_lsx(uint8_t *dst, const uint8_t *src, const uint8_t *half,
ptrdiff_t dstStride, ptrdiff_t srcStride);
void avg_h264_qpel8_h_lowpass_lsx(uint8_t *dst, const uint8_t *src, int dstStride,
int srcStride);
void avg_h264_qpel8_v_lowpass_lsx(uint8_t *dst, uint8_t *src, int dstStride,
int srcStride);
void avg_pixels16_l2_8_lsx(uint8_t *dst, const uint8_t *src, uint8_t *half,
ptrdiff_t dstStride, ptrdiff_t srcStride);
void avg_h264_qpel8_hv_lowpass_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dstStride, ptrdiff_t srcStride);
void avg_pixels8_l2_8_lsx(uint8_t *dst, const uint8_t *src, const uint8_t *half,
ptrdiff_t dstStride, ptrdiff_t srcStride);
void ff_put_h264_qpel16_mc00_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc10_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc20_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc30_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc01_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc11_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc13_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc31_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc33_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc03_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc02_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc22_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc21_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel16_mc12_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel16_mc32_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel16_mc23_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel16_mc00_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc10_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc30_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc33_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc11_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc31_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc13_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc20_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc02_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel16_mc03_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel16_mc23_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel16_mc21_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel16_mc01_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel16_mc32_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel16_mc12_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel16_mc22_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc03_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc00_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc01_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc30_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc10_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc33_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc13_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc31_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc11_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc32_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc21_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc23_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc12_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc02_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc22_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc20_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel8_mc00_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel8_mc10_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel8_mc20_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel8_mc30_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel8_mc11_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel8_mc21_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel8_mc31_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel8_mc02_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel8_mc12_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel8_mc22_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel8_mc32_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel8_mc13_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel8_mc23_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel8_mc33_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
#if HAVE_LASX
void ff_h264_h_lpf_luma_inter_lasx(uint8_t *src, int stride,
int alpha, int beta, int8_t *tc0);
void ff_h264_v_lpf_luma_inter_lasx(uint8_t *src, int stride,
int alpha, int beta, int8_t *tc0);
void ff_put_h264_qpel16_mc00_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc10_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc20_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc30_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc01_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc11_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc21_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc31_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc02_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc12_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc32_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc22_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc03_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc13_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc23_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel16_mc33_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc00_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc10_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc20_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc30_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc01_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc11_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc21_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc31_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc02_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc12_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc22_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc32_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc03_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc13_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc23_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel16_mc33_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_put_h264_qpel8_mc00_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc10_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc20_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc30_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc01_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc11_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc21_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc31_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc02_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc12_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc22_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc32_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc03_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc13_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc23_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_put_h264_qpel8_mc33_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride);
void ff_avg_h264_qpel8_mc00_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc10_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc20_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc30_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc11_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc21_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc31_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc02_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc12_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc22_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc32_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc13_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc23_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
void ff_avg_h264_qpel8_mc33_lasx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dst_stride);
#endif
#endif // #ifndef AVCODEC_LOONGARCH_H264QPEL_LOONGARCH_H

487
libavcodec/loongarch/h264qpel_lsx.c
Unescape View File

@ -0,0 +1,487 @@
/*
* Loongson LSX optimized h264qpel
*
* Copyright (c) 2023 Loongson Technology Corporation Limited
* Contributed by Hecai Yuan <yuanhecai@loongson.cn>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "h264qpel_loongarch.h"
#include "libavutil/loongarch/loongson_intrinsics.h"
#include "libavutil/attributes.h"
static void put_h264_qpel16_hv_lowpass_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dstStride, ptrdiff_t srcStride)
{
put_h264_qpel8_hv_lowpass_lsx(dst, src, dstStride, srcStride);
put_h264_qpel8_hv_lowpass_lsx(dst + 8, src + 8, dstStride, srcStride);
src += srcStride << 3;
dst += dstStride << 3;
put_h264_qpel8_hv_lowpass_lsx(dst, src, dstStride, srcStride);
put_h264_qpel8_hv_lowpass_lsx(dst + 8, src + 8, dstStride, srcStride);
}
void ff_put_h264_qpel16_mc22_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
put_h264_qpel16_hv_lowpass_lsx(dst, src, stride, stride);
}
static void put_h264_qpel16_h_lowpass_lsx(uint8_t *dst, const uint8_t *src,
int dstStride, int srcStride)
{
put_h264_qpel8_h_lowpass_lsx(dst, src, dstStride, srcStride);
put_h264_qpel8_h_lowpass_lsx(dst+8, src+8, dstStride, srcStride);
src += srcStride << 3;
dst += dstStride << 3;
put_h264_qpel8_h_lowpass_lsx(dst, src, dstStride, srcStride);
put_h264_qpel8_h_lowpass_lsx(dst+8, src+8, dstStride, srcStride);
}
static void put_h264_qpel16_v_lowpass_lsx(uint8_t *dst, const uint8_t *src,
int dstStride, int srcStride)
{
put_h264_qpel8_v_lowpass_lsx(dst, (uint8_t*)src, dstStride, srcStride);
put_h264_qpel8_v_lowpass_lsx(dst+8, (uint8_t*)src+8, dstStride, srcStride);
src += 8*srcStride;
dst += 8*dstStride;
put_h264_qpel8_v_lowpass_lsx(dst, (uint8_t*)src, dstStride, srcStride);
put_h264_qpel8_v_lowpass_lsx(dst+8, (uint8_t*)src+8, dstStride, srcStride);
}
void ff_put_h264_qpel16_mc21_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t temp[512];
uint8_t *const halfH = temp;
uint8_t *const halfHV = temp + 256;
put_h264_qpel16_h_lowpass_lsx(halfH, src, 16, stride);
put_h264_qpel16_hv_lowpass_lsx(halfHV, src, 16, stride);
put_pixels16_l2_8_lsx(dst, halfH, halfHV, stride, 16);
}
void ff_put_h264_qpel16_mc12_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t temp[512];
uint8_t *const halfHV = temp;
uint8_t *const halfH = temp + 256;
put_h264_qpel16_hv_lowpass_lsx(halfHV, src, 16, stride);
put_h264_qpel16_v_lowpass_lsx(halfH, src, 16, stride);
put_pixels16_l2_8_lsx(dst, halfH, halfHV, stride, 16);
}
void ff_put_h264_qpel16_mc32_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t temp[512];
uint8_t *const halfHV = temp;
uint8_t *const halfH = temp + 256;
put_h264_qpel16_hv_lowpass_lsx(halfHV, src, 16, stride);
put_h264_qpel16_v_lowpass_lsx(halfH, src + 1, 16, stride);
put_pixels16_l2_8_lsx(dst, halfH, halfHV, stride, 16);
}
void ff_put_h264_qpel16_mc23_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t temp[512];
uint8_t *const halfH = temp;
uint8_t *const halfHV = temp + 256;
put_h264_qpel16_h_lowpass_lsx(halfH, src + stride, 16, stride);
put_h264_qpel16_hv_lowpass_lsx(halfHV, src, 16, stride);
put_pixels16_l2_8_lsx(dst, halfH, halfHV, stride, 16);
}
static void avg_h264_qpel16_v_lowpass_lsx(uint8_t *dst, const uint8_t *src,
int dstStride, int srcStride)
{
avg_h264_qpel8_v_lowpass_lsx(dst, (uint8_t*)src, dstStride, srcStride);
avg_h264_qpel8_v_lowpass_lsx(dst+8, (uint8_t*)src+8, dstStride, srcStride);
src += 8*srcStride;
dst += 8*dstStride;
avg_h264_qpel8_v_lowpass_lsx(dst, (uint8_t*)src, dstStride, srcStride);
avg_h264_qpel8_v_lowpass_lsx(dst+8, (uint8_t*)src+8, dstStride, srcStride);
}
void ff_avg_h264_qpel16_mc02_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
avg_h264_qpel16_v_lowpass_lsx(dst, src, stride, stride);
}
void ff_avg_h264_qpel16_mc03_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t half[256];
put_h264_qpel16_v_lowpass_lsx(half, src, 16, stride);
avg_pixels16_l2_8_lsx(dst, src + stride, half, stride, stride);
}
void ff_avg_h264_qpel16_mc23_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t temp[512];
uint8_t *const halfH = temp;
uint8_t *const halfHV = temp + 256;
put_h264_qpel16_h_lowpass_lsx(halfH, src + stride, 16, stride);
put_h264_qpel16_hv_lowpass_lsx(halfHV, src, 16, stride);
avg_pixels16_l2_8_lsx(dst, halfH, halfHV, stride, 16);
}
void ff_avg_h264_qpel16_mc21_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t temp[512];
uint8_t *const halfH = temp;
uint8_t *const halfHV = temp + 256;
put_h264_qpel16_h_lowpass_lsx(halfH, src, 16, stride);
put_h264_qpel16_hv_lowpass_lsx(halfHV, src, 16, stride);
avg_pixels16_l2_8_lsx(dst, halfH, halfHV, stride, 16);
}
void ff_avg_h264_qpel16_mc01_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t half[256];
put_h264_qpel16_v_lowpass_lsx(half, src, 16, stride);
avg_pixels16_l2_8_lsx(dst, src, half, stride, stride);
}
void ff_avg_h264_qpel16_mc32_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t temp[512];
uint8_t *const halfHV = temp;
uint8_t *const halfH = temp + 256;
put_h264_qpel16_hv_lowpass_lsx(halfHV, src, 16, stride);
put_h264_qpel16_v_lowpass_lsx(halfH, src + 1, 16, stride);
avg_pixels16_l2_8_lsx(dst, halfH, halfHV, stride, 16);
}
void ff_avg_h264_qpel16_mc12_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t temp[512];
uint8_t *const halfHV = temp;
uint8_t *const halfH = temp + 256;
put_h264_qpel16_hv_lowpass_lsx(halfHV, src, 16, stride);
put_h264_qpel16_v_lowpass_lsx(halfH, src, 16, stride);
avg_pixels16_l2_8_lsx(dst, halfH, halfHV, stride, 16);
}
static void avg_h264_qpel16_hv_lowpass_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t dstStride, ptrdiff_t srcStride)
{
avg_h264_qpel8_hv_lowpass_lsx(dst, src, dstStride, srcStride);
avg_h264_qpel8_hv_lowpass_lsx(dst + 8, src + 8, dstStride, srcStride);
src += srcStride << 3;
dst += dstStride << 3;
avg_h264_qpel8_hv_lowpass_lsx(dst, src, dstStride, srcStride);
avg_h264_qpel8_hv_lowpass_lsx(dst + 8, src + 8, dstStride, srcStride);
}
void ff_avg_h264_qpel16_mc22_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
avg_h264_qpel16_hv_lowpass_lsx(dst, src, stride, stride);
}
void ff_put_h264_qpel8_mc03_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t half[64];
put_h264_qpel8_v_lowpass_lsx(half, (uint8_t*)src, 8, stride);
put_pixels8_l2_8_lsx(dst, src + stride, half, stride, stride);
}
void ff_put_h264_qpel8_mc01_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t half[64];
put_h264_qpel8_v_lowpass_lsx(half, (uint8_t*)src, 8, stride);
put_pixels8_l2_8_lsx(dst, src, half, stride, stride);
}
void ff_put_h264_qpel8_mc30_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t half[64];
put_h264_qpel8_h_lowpass_lsx(half, src, 8, stride);
put_pixels8_l2_8_lsx(dst, src+1, half, stride, stride);
}
void ff_put_h264_qpel8_mc10_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t half[64];
put_h264_qpel8_h_lowpass_lsx(half, src, 8, stride);
put_pixels8_l2_8_lsx(dst, src, half, stride, stride);
}
void ff_put_h264_qpel8_mc33_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t halfH[64];
uint8_t halfV[64];
put_h264_qpel8_h_lowpass_lsx(halfH, src + stride, 8, stride);
put_h264_qpel8_v_lowpass_lsx(halfV, (uint8_t*)src + 1, 8, stride);
put_pixels8_l2_8_lsx(dst, halfH, halfV, stride, 8);
}
void ff_put_h264_qpel8_mc13_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t halfH[64];
uint8_t halfV[64];
put_h264_qpel8_h_lowpass_lsx(halfH, src + stride, 8, stride);
put_h264_qpel8_v_lowpass_lsx(halfV, (uint8_t*)src, 8, stride);
put_pixels8_l2_8_lsx(dst, halfH, halfV, stride, 8);
}
void ff_put_h264_qpel8_mc31_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t halfH[64];
uint8_t halfV[64];
put_h264_qpel8_h_lowpass_lsx(halfH, src, 8, stride);
put_h264_qpel8_v_lowpass_lsx(halfV, (uint8_t*)src + 1, 8, stride);
put_pixels8_l2_8_lsx(dst, halfH, halfV, stride, 8);
}
void ff_put_h264_qpel8_mc11_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t halfH[64];
uint8_t halfV[64];
put_h264_qpel8_h_lowpass_lsx(halfH, src, 8, stride);
put_h264_qpel8_v_lowpass_lsx(halfV, (uint8_t*)src, 8, stride);
put_pixels8_l2_8_lsx(dst, halfH, halfV, stride, 8);
}
void ff_put_h264_qpel8_mc32_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t temp[128];
uint8_t *const halfHV = temp;
uint8_t *const halfH = temp + 64;
put_h264_qpel8_hv_lowpass_lsx(halfHV, src, 8, stride);
put_h264_qpel8_v_lowpass_lsx(halfH, (uint8_t*)src + 1, 8, stride);
put_pixels8_l2_8_lsx(dst, halfH, halfHV, stride, 8);
}
void ff_put_h264_qpel8_mc21_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t temp[128];
uint8_t *const halfH = temp;
uint8_t *const halfHV = temp + 64;
put_h264_qpel8_h_lowpass_lsx(halfH, src, 8, stride);
put_h264_qpel8_hv_lowpass_lsx(halfHV, src, 8, stride);
put_pixels8_l2_8_lsx(dst, halfH, halfHV, stride, 8);
}
void ff_put_h264_qpel8_mc23_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t temp[128];
uint8_t *const halfH = temp;
uint8_t *const halfHV = temp + 64;
put_h264_qpel8_h_lowpass_lsx(halfH, src + stride, 8, stride);
put_h264_qpel8_hv_lowpass_lsx(halfHV, src, 8, stride);
put_pixels8_l2_8_lsx(dst, halfH, halfHV, stride, 8);
}
void ff_put_h264_qpel8_mc12_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t temp[128];
uint8_t *const halfHV = temp;
uint8_t *const halfH = temp + 64;
put_h264_qpel8_hv_lowpass_lsx(halfHV, src, 8, stride);
put_h264_qpel8_v_lowpass_lsx(halfH, (uint8_t*)src, 8, stride);
put_pixels8_l2_8_lsx(dst, halfH, halfHV, stride, 8);
}
void ff_put_h264_qpel8_mc02_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
put_h264_qpel8_v_lowpass_lsx(dst, (uint8_t*)src, stride, stride);
}
void ff_put_h264_qpel8_mc22_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
put_h264_qpel8_hv_lowpass_lsx(dst, src, stride, stride);
}
void ff_put_h264_qpel8_mc20_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
put_h264_qpel8_h_lowpass_lsx(dst, src, stride, stride);
}
void ff_avg_h264_qpel8_mc10_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t half[64];
put_h264_qpel8_h_lowpass_lsx(half, src, 8, stride);
avg_pixels8_l2_8_lsx(dst, src, half, stride, stride);
}
void ff_avg_h264_qpel8_mc20_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
avg_h264_qpel8_h_lowpass_lsx(dst, src, stride, stride);
}
void ff_avg_h264_qpel8_mc30_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t half[64];
put_h264_qpel8_h_lowpass_lsx(half, src, 8, stride);
avg_pixels8_l2_8_lsx(dst, src+1, half, stride, stride);
}
void ff_avg_h264_qpel8_mc11_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t halfH[64];
uint8_t halfV[64];
put_h264_qpel8_h_lowpass_lsx(halfH, src, 8, stride);
put_h264_qpel8_v_lowpass_lsx(halfV, (uint8_t*)src, 8, stride);
avg_pixels8_l2_8_lsx(dst, halfH, halfV, stride, 8);
}
void ff_avg_h264_qpel8_mc21_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t temp[128];
uint8_t *const halfH = temp;
uint8_t *const halfHV = temp + 64;
put_h264_qpel8_h_lowpass_lsx(halfH, src, 8, stride);
put_h264_qpel8_hv_lowpass_lsx(halfHV, src, 8, stride);
avg_pixels8_l2_8_lsx(dst, halfH, halfHV, stride, 8);
}
void ff_avg_h264_qpel8_mc31_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t halfH[64];
uint8_t halfV[64];
put_h264_qpel8_h_lowpass_lsx(halfH, src, 8, stride);
put_h264_qpel8_v_lowpass_lsx(halfV, (uint8_t*)src + 1, 8, stride);
avg_pixels8_l2_8_lsx(dst, halfH, halfV, stride, 8);
}
void ff_avg_h264_qpel8_mc02_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
avg_h264_qpel8_v_lowpass_lsx(dst, (uint8_t*)src, stride, stride);
}
void ff_avg_h264_qpel8_mc12_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t temp[128];
uint8_t *const halfHV = temp;
uint8_t *const halfH = temp + 64;
put_h264_qpel8_hv_lowpass_lsx(halfHV, src, 8, stride);
put_h264_qpel8_v_lowpass_lsx(halfH, (uint8_t*)src, 8, stride);
avg_pixels8_l2_8_lsx(dst, halfH, halfHV, stride, 8);
}
void ff_avg_h264_qpel8_mc22_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
avg_h264_qpel8_hv_lowpass_lsx(dst, src, stride, stride);
}
void ff_avg_h264_qpel8_mc32_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t temp[128];
uint8_t *const halfHV = temp;
uint8_t *const halfH = temp + 64;
put_h264_qpel8_hv_lowpass_lsx(halfHV, src, 8, stride);
put_h264_qpel8_v_lowpass_lsx(halfH, (uint8_t*)src + 1, 8, stride);
avg_pixels8_l2_8_lsx(dst, halfH, halfHV, stride, 8);
}
void ff_avg_h264_qpel8_mc13_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t halfH[64];
uint8_t halfV[64];
put_h264_qpel8_h_lowpass_lsx(halfH, src + stride, 8, stride);
put_h264_qpel8_v_lowpass_lsx(halfV, (uint8_t*)src, 8, stride);
avg_pixels8_l2_8_lsx(dst, halfH, halfV, stride, 8);
}
void ff_avg_h264_qpel8_mc23_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t temp[128];
uint8_t *const halfH = temp;
uint8_t *const halfHV = temp + 64;
put_h264_qpel8_h_lowpass_lsx(halfH, src + stride, 8, stride);
put_h264_qpel8_hv_lowpass_lsx(halfHV, src, 8, stride);
avg_pixels8_l2_8_lsx(dst, halfH, halfHV, stride, 8);
}
void ff_avg_h264_qpel8_mc33_lsx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
uint8_t halfH[64];
uint8_t halfV[64];
put_h264_qpel8_h_lowpass_lsx(halfH, src + stride, 8, stride);
put_h264_qpel8_v_lowpass_lsx(halfV, (uint8_t*)src + 1, 8, stride);
avg_pixels8_l2_8_lsx(dst, halfH, halfV, stride, 8);
}
Write Preview
Loading…
Cancel
Save