virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-12-18 03:19:31 +02:00
Code Issues Releases Activity
f2da650258
FFmpeg/libavcodec/mips/vp9_intra_msa.c

535 lines
22 KiB
C
Raw Normal View History

avcodec/mips: MSA (MIPS-SIMD-Arch) optimizations for VP9 intra functions Signed-off-by: Shivraj Patil <shivraj.patil@imgtec.com> Reviewed-by: "Ronald S. Bultje" <rsbultje@gmail.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2015-07-23 14:55:07 +02:00
/*
* Copyright (c) 2015 Shivraj Patil (Shivraj.Patil@imgtec.com)
*
* 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
*/
vp9: re-split the decoder/format/dsp interface header files. The advantage here is that the internal software decoder interface is not exposed to the DSP functions or the hardware accelerations.
2017-03-27 22:47:46 +02:00
#include "libavcodec/vp9dsp.h"
avcodec/mips: MSA (MIPS-SIMD-Arch) optimizations for VP9 intra functions Signed-off-by: Shivraj Patil <shivraj.patil@imgtec.com> Reviewed-by: "Ronald S. Bultje" <rsbultje@gmail.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2015-07-23 14:55:07 +02:00
#include "libavutil/mips/generic_macros_msa.h"
#include "vp9dsp_mips.h"
#define IPRED_SUBS_UH2_UH(in0, in1, out0, out1) \
{ \
out0 = __msa_subs_u_h(out0, in0); \
out1 = __msa_subs_u_h(out1, in1); \
}
void ff_vert_16x16_msa(uint8_t *dst, ptrdiff_t dst_stride, const uint8_t *left,
const uint8_t *src)
{
uint32_t row;
v16u8 src0;
src0 = LD_UB(src);
for (row = 16; row--;) {
ST_UB(src0, dst);
dst += dst_stride;
}
}
void ff_vert_32x32_msa(uint8_t *dst, ptrdiff_t dst_stride, const uint8_t *left,
const uint8_t *src)
{
uint32_t row;
v16u8 src1, src2;
src1 = LD_UB(src);
src2 = LD_UB(src + 16);
for (row = 32; row--;) {
ST_UB2(src1, src2, dst, 16);
dst += dst_stride;
}
}
void ff_hor_16x16_msa(uint8_t *dst, ptrdiff_t dst_stride, const uint8_t *src,
const uint8_t *top)
{
uint32_t row, inp;
v16u8 src0, src1, src2, src3;
src += 12;
for (row = 4; row--;) {
inp = LW(src);
src -= 4;
src0 = (v16u8) __msa_fill_b(inp >> 24);
src1 = (v16u8) __msa_fill_b(inp >> 16);
src2 = (v16u8) __msa_fill_b(inp >> 8);
src3 = (v16u8) __msa_fill_b(inp);
ST_UB4(src0, src1, src2, src3, dst, dst_stride);
dst += (4 * dst_stride);
}
}
void ff_hor_32x32_msa(uint8_t *dst, ptrdiff_t dst_stride, const uint8_t *src,
const uint8_t *top)
{
uint32_t row, inp;
v16u8 src0, src1, src2, src3;
src += 28;
for (row = 8; row--;) {
inp = LW(src);
src -= 4;
src0 = (v16u8) __msa_fill_b(inp >> 24);
src1 = (v16u8) __msa_fill_b(inp >> 16);
src2 = (v16u8) __msa_fill_b(inp >> 8);
src3 = (v16u8) __msa_fill_b(inp);
ST_UB2(src0, src0, dst, 16);
dst += dst_stride;
ST_UB2(src1, src1, dst, 16);
dst += dst_stride;
ST_UB2(src2, src2, dst, 16);
dst += dst_stride;
ST_UB2(src3, src3, dst, 16);
dst += dst_stride;
}
}
void ff_dc_4x4_msa(uint8_t *dst, ptrdiff_t dst_stride, const uint8_t *src_left,
const uint8_t *src_top)
{
uint32_t val0, val1;
v16i8 store, src = { 0 };
v8u16 sum_h;
v4u32 sum_w;
v2u64 sum_d;
val0 = LW(src_top);
val1 = LW(src_left);
INSERT_W2_SB(val0, val1, src);
sum_h = __msa_hadd_u_h((v16u8) src, (v16u8) src);
sum_w = __msa_hadd_u_w(sum_h, sum_h);
sum_d = __msa_hadd_u_d(sum_w, sum_w);
sum_w = (v4u32) __msa_srari_w((v4i32) sum_d, 3);
store = __msa_splati_b((v16i8) sum_w, 0);
val0 = __msa_copy_u_w((v4i32) store, 0);
SW4(val0, val0, val0, val0, dst, dst_stride);
}
#define INTRA_DC_TL_4x4(dir) \
void ff_dc_##dir##_4x4_msa(uint8_t *dst, ptrdiff_t dst_stride, \
const uint8_t *left, \
const uint8_t *top) \
{ \
uint32_t val0; \
v16i8 store, data = { 0 }; \
v8u16 sum_h; \
v4u32 sum_w; \
\
val0 = LW(dir); \
data = (v16i8) __msa_insert_w((v4i32) data, 0, val0); \
sum_h = __msa_hadd_u_h((v16u8) data, (v16u8) data); \
sum_w = __msa_hadd_u_w(sum_h, sum_h); \
sum_w = (v4u32) __msa_srari_w((v4i32) sum_w, 2); \
store = __msa_splati_b((v16i8) sum_w, 0); \
val0 = __msa_copy_u_w((v4i32) store, 0); \
\
SW4(val0, val0, val0, val0, dst, dst_stride); \
}
INTRA_DC_TL_4x4(top);
INTRA_DC_TL_4x4(left);
void ff_dc_8x8_msa(uint8_t *dst, ptrdiff_t dst_stride, const uint8_t *src_left,
const uint8_t *src_top)
{
uint64_t val0, val1;
v16i8 store;
v16u8 src = { 0 };
v8u16 sum_h;
v4u32 sum_w;
v2u64 sum_d;
val0 = LD(src_top);
val1 = LD(src_left);
INSERT_D2_UB(val0, val1, src);
sum_h = __msa_hadd_u_h(src, src);
sum_w = __msa_hadd_u_w(sum_h, sum_h);
sum_d = __msa_hadd_u_d(sum_w, sum_w);
sum_w = (v4u32) __msa_pckev_w((v4i32) sum_d, (v4i32) sum_d);
sum_d = __msa_hadd_u_d(sum_w, sum_w);
sum_w = (v4u32) __msa_srari_w((v4i32) sum_d, 4);
store = __msa_splati_b((v16i8) sum_w, 0);
val0 = __msa_copy_u_d((v2i64) store, 0);
SD4(val0, val0, val0, val0, dst, dst_stride);
dst += (4 * dst_stride);
SD4(val0, val0, val0, val0, dst, dst_stride);
}
#define INTRA_DC_TL_8x8(dir) \
void ff_dc_##dir##_8x8_msa(uint8_t *dst, ptrdiff_t dst_stride, \
const uint8_t *left, \
const uint8_t *top) \
{ \
uint64_t val0; \
v16i8 store; \
v16u8 data = { 0 }; \
v8u16 sum_h; \
v4u32 sum_w; \
v2u64 sum_d; \
\
val0 = LD(dir); \
data = (v16u8) __msa_insert_d((v2i64) data, 0, val0); \
sum_h = __msa_hadd_u_h(data, data); \
sum_w = __msa_hadd_u_w(sum_h, sum_h); \
sum_d = __msa_hadd_u_d(sum_w, sum_w); \
sum_w = (v4u32) __msa_srari_w((v4i32) sum_d, 3); \
store = __msa_splati_b((v16i8) sum_w, 0); \
val0 = __msa_copy_u_d((v2i64) store, 0); \
\
SD4(val0, val0, val0, val0, dst, dst_stride); \
dst += (4 * dst_stride); \
SD4(val0, val0, val0, val0, dst, dst_stride); \
}
INTRA_DC_TL_8x8(top);
INTRA_DC_TL_8x8(left);
void ff_dc_16x16_msa(uint8_t *dst, ptrdiff_t dst_stride,
const uint8_t *src_left, const uint8_t *src_top)
{
v16u8 top, left, out;
v8u16 sum_h, sum_top, sum_left;
v4u32 sum_w;
v2u64 sum_d;
top = LD_UB(src_top);
left = LD_UB(src_left);
HADD_UB2_UH(top, left, sum_top, sum_left);
sum_h = sum_top + sum_left;
sum_w = __msa_hadd_u_w(sum_h, sum_h);
sum_d = __msa_hadd_u_d(sum_w, sum_w);
sum_w = (v4u32) __msa_pckev_w((v4i32) sum_d, (v4i32) sum_d);
sum_d = __msa_hadd_u_d(sum_w, sum_w);
sum_w = (v4u32) __msa_srari_w((v4i32) sum_d, 5);
out = (v16u8) __msa_splati_b((v16i8) sum_w, 0);
ST_UB8(out, out, out, out, out, out, out, out, dst, dst_stride);
dst += (8 * dst_stride);
ST_UB8(out, out, out, out, out, out, out, out, dst, dst_stride);
}
#define INTRA_DC_TL_16x16(dir) \
void ff_dc_##dir##_16x16_msa(uint8_t *dst, ptrdiff_t dst_stride, \
const uint8_t *left, \
const uint8_t *top) \
{ \
v16u8 data, out; \
v8u16 sum_h; \
v4u32 sum_w; \
v2u64 sum_d; \
\
data = LD_UB(dir); \
sum_h = __msa_hadd_u_h(data, data); \
sum_w = __msa_hadd_u_w(sum_h, sum_h); \
sum_d = __msa_hadd_u_d(sum_w, sum_w); \
sum_w = (v4u32) __msa_pckev_w((v4i32) sum_d, (v4i32) sum_d); \
sum_d = __msa_hadd_u_d(sum_w, sum_w); \
sum_w = (v4u32) __msa_srari_w((v4i32) sum_d, 4); \
out = (v16u8) __msa_splati_b((v16i8) sum_w, 0); \
\
ST_UB8(out, out, out, out, out, out, out, out, dst, dst_stride); \
dst += (8 * dst_stride); \
ST_UB8(out, out, out, out, out, out, out, out, dst, dst_stride); \
}
INTRA_DC_TL_16x16(top);
INTRA_DC_TL_16x16(left);
void ff_dc_32x32_msa(uint8_t *dst, ptrdiff_t dst_stride,
const uint8_t *src_left, const uint8_t *src_top)
{
uint32_t row;
v16u8 top0, top1, left0, left1, out;
v8u16 sum_h, sum_top0, sum_top1, sum_left0, sum_left1;
v4u32 sum_w;
v2u64 sum_d;
LD_UB2(src_top, 16, top0, top1);
LD_UB2(src_left, 16, left0, left1);
HADD_UB2_UH(top0, top1, sum_top0, sum_top1);
HADD_UB2_UH(left0, left1, sum_left0, sum_left1);
sum_h = sum_top0 + sum_top1;
sum_h += sum_left0 + sum_left1;
sum_w = __msa_hadd_u_w(sum_h, sum_h);
sum_d = __msa_hadd_u_d(sum_w, sum_w);
sum_w = (v4u32) __msa_pckev_w((v4i32) sum_d, (v4i32) sum_d);
sum_d = __msa_hadd_u_d(sum_w, sum_w);
sum_w = (v4u32) __msa_srari_w((v4i32) sum_d, 6);
out = (v16u8) __msa_splati_b((v16i8) sum_w, 0);
for (row = 16; row--;)
{
ST_UB2(out, out, dst, 16);
dst += dst_stride;
ST_UB2(out, out, dst, 16);
dst += dst_stride;
}
}
#define INTRA_DC_TL_32x32(dir) \
void ff_dc_##dir##_32x32_msa(uint8_t *dst, ptrdiff_t dst_stride, \
const uint8_t *left, \
const uint8_t *top) \
{ \
uint32_t row; \
v16u8 data0, data1, out; \
v8u16 sum_h, sum_data0, sum_data1; \
v4u32 sum_w; \
v2u64 sum_d; \
\
LD_UB2(dir, 16, data0, data1); \
HADD_UB2_UH(data0, data1, sum_data0, sum_data1); \
sum_h = sum_data0 + sum_data1; \
sum_w = __msa_hadd_u_w(sum_h, sum_h); \
sum_d = __msa_hadd_u_d(sum_w, sum_w); \
sum_w = (v4u32) __msa_pckev_w((v4i32) sum_d, (v4i32) sum_d); \
sum_d = __msa_hadd_u_d(sum_w, sum_w); \
sum_w = (v4u32) __msa_srari_w((v4i32) sum_d, 5); \
out = (v16u8) __msa_splati_b((v16i8) sum_w, 0); \
\
for (row = 16; row--;) \
{ \
ST_UB2(out, out, dst, 16); \
dst += dst_stride; \
ST_UB2(out, out, dst, 16); \
dst += dst_stride; \
} \
}
INTRA_DC_TL_32x32(top);
INTRA_DC_TL_32x32(left);
#define INTRA_PREDICT_VALDC_16X16_MSA(val) \
void ff_dc_##val##_16x16_msa(uint8_t *dst, ptrdiff_t dst_stride, \
const uint8_t *left, const uint8_t *top) \
{ \
v16u8 out = (v16u8) __msa_ldi_b(val); \
\
ST_UB8(out, out, out, out, out, out, out, out, dst, dst_stride); \
dst += (8 * dst_stride); \
ST_UB8(out, out, out, out, out, out, out, out, dst, dst_stride); \
}
INTRA_PREDICT_VALDC_16X16_MSA(127);
INTRA_PREDICT_VALDC_16X16_MSA(128);
INTRA_PREDICT_VALDC_16X16_MSA(129);
#define INTRA_PREDICT_VALDC_32X32_MSA(val) \
void ff_dc_##val##_32x32_msa(uint8_t *dst, ptrdiff_t dst_stride, \
const uint8_t *left, const uint8_t *top) \
{ \
uint32_t row; \
v16u8 out = (v16u8) __msa_ldi_b(val); \
\
for (row = 16; row--;) \
{ \
ST_UB2(out, out, dst, 16); \
dst += dst_stride; \
ST_UB2(out, out, dst, 16); \
dst += dst_stride; \
} \
}
INTRA_PREDICT_VALDC_32X32_MSA(127);
INTRA_PREDICT_VALDC_32X32_MSA(128);
INTRA_PREDICT_VALDC_32X32_MSA(129);
void ff_tm_4x4_msa(uint8_t *dst, ptrdiff_t dst_stride,
const uint8_t *src_left, const uint8_t *src_top_ptr)
{
uint32_t left;
uint8_t top_left = src_top_ptr[-1];
v16i8 src_top, src_left0, src_left1, src_left2, src_left3, tmp0, tmp1;
v16u8 src0, src1, src2, src3;
v8u16 src_top_left, vec0, vec1, vec2, vec3;
src_top_left = (v8u16) __msa_fill_h(top_left);
src_top = LD_SB(src_top_ptr);
left = LW(src_left);
src_left0 = __msa_fill_b(left >> 24);
src_left1 = __msa_fill_b(left >> 16);
src_left2 = __msa_fill_b(left >> 8);
src_left3 = __msa_fill_b(left);
ILVR_B4_UB(src_left0, src_top, src_left1, src_top, src_left2, src_top,
src_left3, src_top, src0, src1, src2, src3);
HADD_UB4_UH(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, vec0, vec1);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, vec2, vec3);
SAT_UH4_UH(vec0, vec1, vec2, vec3, 7);
PCKEV_B2_SB(vec1, vec0, vec3, vec2, tmp0, tmp1);
avutil/mips: refactor msa load and store macros. Replace STnxm_UB and LDnxm_SH with new macros ST_{H/W/D}{1/2/4/8}. The old macros are difficult to use because they don't follow the same parameter passing rules. Changing details as following: 1. remove LD4x4_SH. 2. replace ST2x4_UB with ST_H4. 3. replace ST4x2_UB with ST_W2. 4. replace ST4x4_UB with ST_W4. 5. replace ST4x8_UB with ST_W8. 6. replace ST6x4_UB with ST_W2 and ST_H2. 7. replace ST8x1_UB with ST_D1. 8. replace ST8x2_UB with ST_D2. 9. replace ST8x4_UB with ST_D4. 10. replace ST8x8_UB with ST_D8. 11. replace ST12x4_UB with ST_D4 and ST_W4. Examples of new macro: ST_H4(in, idx0, idx1, idx2, idx3, pdst, stride) ST_H4 store four half-word elements in vector 'in' to pdst with stride. About the macro name: 1) 'ST' means store operation. 2) 'H/W/D' means type of vector element is 'half-word/word/double-word'. 3) Number '1/2/4/8' means how many elements will be stored. About the macro parameter: 1) 'in0, in1...' 128-bits vector. 2) 'idx0, idx1...' elements index. 3) 'pdst' destination pointer to store to 4) 'stride' stride of each store operation. Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-07-17 11:35:00 +02:00
ST_W2(tmp0, 0, 2, dst, dst_stride);
ST_W2(tmp1, 0, 2, dst + 2 * dst_stride, dst_stride);
avcodec/mips: MSA (MIPS-SIMD-Arch) optimizations for VP9 intra functions Signed-off-by: Shivraj Patil <shivraj.patil@imgtec.com> Reviewed-by: "Ronald S. Bultje" <rsbultje@gmail.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2015-07-23 14:55:07 +02:00
}
void ff_tm_8x8_msa(uint8_t *dst, ptrdiff_t dst_stride,
const uint8_t *src_left, const uint8_t *src_top_ptr)
{
uint8_t top_left = src_top_ptr[-1];
uint32_t loop_cnt, left;
v16i8 src_top, src_left0, src_left1, src_left2, src_left3, tmp0, tmp1;
v8u16 src_top_left, vec0, vec1, vec2, vec3;
v16u8 src0, src1, src2, src3;
src_top = LD_SB(src_top_ptr);
src_top_left = (v8u16) __msa_fill_h(top_left);
src_left += 4;
for (loop_cnt = 2; loop_cnt--;) {
left = LW(src_left);
src_left0 = __msa_fill_b(left >> 24);
src_left1 = __msa_fill_b(left >> 16);
src_left2 = __msa_fill_b(left >> 8);
src_left3 = __msa_fill_b(left);
src_left -= 4;
ILVR_B4_UB(src_left0, src_top, src_left1, src_top, src_left2, src_top,
src_left3, src_top, src0, src1, src2, src3);
HADD_UB4_UH(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, vec0, vec1);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, vec2, vec3);
SAT_UH4_UH(vec0, vec1, vec2, vec3, 7);
PCKEV_B2_SB(vec1, vec0, vec3, vec2, tmp0, tmp1);
avutil/mips: refactor msa load and store macros. Replace STnxm_UB and LDnxm_SH with new macros ST_{H/W/D}{1/2/4/8}. The old macros are difficult to use because they don't follow the same parameter passing rules. Changing details as following: 1. remove LD4x4_SH. 2. replace ST2x4_UB with ST_H4. 3. replace ST4x2_UB with ST_W2. 4. replace ST4x4_UB with ST_W4. 5. replace ST4x8_UB with ST_W8. 6. replace ST6x4_UB with ST_W2 and ST_H2. 7. replace ST8x1_UB with ST_D1. 8. replace ST8x2_UB with ST_D2. 9. replace ST8x4_UB with ST_D4. 10. replace ST8x8_UB with ST_D8. 11. replace ST12x4_UB with ST_D4 and ST_W4. Examples of new macro: ST_H4(in, idx0, idx1, idx2, idx3, pdst, stride) ST_H4 store four half-word elements in vector 'in' to pdst with stride. About the macro name: 1) 'ST' means store operation. 2) 'H/W/D' means type of vector element is 'half-word/word/double-word'. 3) Number '1/2/4/8' means how many elements will be stored. About the macro parameter: 1) 'in0, in1...' 128-bits vector. 2) 'idx0, idx1...' elements index. 3) 'pdst' destination pointer to store to 4) 'stride' stride of each store operation. Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-07-17 11:35:00 +02:00
ST_D4(tmp0, tmp1, 0, 1, 0, 1, dst, dst_stride);
avcodec/mips: MSA (MIPS-SIMD-Arch) optimizations for VP9 intra functions Signed-off-by: Shivraj Patil <shivraj.patil@imgtec.com> Reviewed-by: "Ronald S. Bultje" <rsbultje@gmail.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2015-07-23 14:55:07 +02:00
dst += (4 * dst_stride);
}
}
void ff_tm_16x16_msa(uint8_t *dst, ptrdiff_t dst_stride,
const uint8_t *src_left, const uint8_t *src_top_ptr)
{
uint8_t top_left = src_top_ptr[-1];
uint32_t loop_cnt, left;
v16i8 src_top, src_left0, src_left1, src_left2, src_left3;
v8u16 src_top_left, res_r, res_l;
src_top = LD_SB(src_top_ptr);
src_top_left = (v8u16) __msa_fill_h(top_left);
src_left += 12;
for (loop_cnt = 4; loop_cnt--;) {
left = LW(src_left);
src_left0 = __msa_fill_b(left >> 24);
src_left1 = __msa_fill_b(left >> 16);
src_left2 = __msa_fill_b(left >> 8);
src_left3 = __msa_fill_b(left);
src_left -= 4;
ILVRL_B2_UH(src_left0, src_top, res_r, res_l);
HADD_UB2_UH(res_r, res_l, res_r, res_l);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r, res_l);
SAT_UH2_UH(res_r, res_l, 7);
PCKEV_ST_SB(res_r, res_l, dst);
dst += dst_stride;
ILVRL_B2_UH(src_left1, src_top, res_r, res_l);
HADD_UB2_UH(res_r, res_l, res_r, res_l);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r, res_l);
SAT_UH2_UH(res_r, res_l, 7);
PCKEV_ST_SB(res_r, res_l, dst);
dst += dst_stride;
ILVRL_B2_UH(src_left2, src_top, res_r, res_l);
HADD_UB2_UH(res_r, res_l, res_r, res_l);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r, res_l);
SAT_UH2_UH(res_r, res_l, 7);
PCKEV_ST_SB(res_r, res_l, dst);
dst += dst_stride;
ILVRL_B2_UH(src_left3, src_top, res_r, res_l);
HADD_UB2_UH(res_r, res_l, res_r, res_l);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r, res_l);
SAT_UH2_UH(res_r, res_l, 7);
PCKEV_ST_SB(res_r, res_l, dst);
dst += dst_stride;
}
}
void ff_tm_32x32_msa(uint8_t *dst, ptrdiff_t dst_stride,
const uint8_t *src_left, const uint8_t *src_top_ptr)
{
uint8_t top_left = src_top_ptr[-1];
uint32_t loop_cnt, left;
v16i8 src_top0, src_top1, src_left0, src_left1, src_left2, src_left3;
v8u16 src_top_left, res_r0, res_r1, res_l0, res_l1;
src_top0 = LD_SB(src_top_ptr);
src_top1 = LD_SB(src_top_ptr + 16);
src_top_left = (v8u16) __msa_fill_h(top_left);
src_left += 28;
for (loop_cnt = 8; loop_cnt--;) {
left = LW(src_left);
src_left0 = __msa_fill_b(left >> 24);
src_left1 = __msa_fill_b(left >> 16);
src_left2 = __msa_fill_b(left >> 8);
src_left3 = __msa_fill_b(left);
src_left -= 4;
ILVR_B2_UH(src_left0, src_top0, src_left0, src_top1, res_r0, res_r1);
ILVL_B2_UH(src_left0, src_top0, src_left0, src_top1, res_l0, res_l1);
HADD_UB4_UH(res_r0, res_l0, res_r1, res_l1, res_r0, res_l0, res_r1,
res_l1);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r0, res_l0);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r1, res_l1);
SAT_UH4_UH(res_r0, res_l0, res_r1, res_l1, 7);
PCKEV_ST_SB(res_r0, res_l0, dst);
PCKEV_ST_SB(res_r1, res_l1, dst + 16);
dst += dst_stride;
ILVR_B2_UH(src_left1, src_top0, src_left1, src_top1, res_r0, res_r1);
ILVL_B2_UH(src_left1, src_top0, src_left1, src_top1, res_l0, res_l1);
HADD_UB4_UH(res_r0, res_l0, res_r1, res_l1, res_r0, res_l0, res_r1,
res_l1);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r0, res_l0);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r1, res_l1);
SAT_UH4_UH(res_r0, res_l0, res_r1, res_l1, 7);
PCKEV_ST_SB(res_r0, res_l0, dst);
PCKEV_ST_SB(res_r1, res_l1, dst + 16);
dst += dst_stride;
ILVR_B2_UH(src_left2, src_top0, src_left2, src_top1, res_r0, res_r1);
ILVL_B2_UH(src_left2, src_top0, src_left2, src_top1, res_l0, res_l1);
HADD_UB4_UH(res_r0, res_l0, res_r1, res_l1, res_r0, res_l0, res_r1,
res_l1);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r0, res_l0);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r1, res_l1);
SAT_UH4_UH(res_r0, res_l0, res_r1, res_l1, 7);
PCKEV_ST_SB(res_r0, res_l0, dst);
PCKEV_ST_SB(res_r1, res_l1, dst + 16);
dst += dst_stride;
ILVR_B2_UH(src_left3, src_top0, src_left3, src_top1, res_r0, res_r1);
ILVL_B2_UH(src_left3, src_top0, src_left3, src_top1, res_l0, res_l1);
HADD_UB4_UH(res_r0, res_l0, res_r1, res_l1, res_r0, res_l0, res_r1,
res_l1);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r0, res_l0);
IPRED_SUBS_UH2_UH(src_top_left, src_top_left, res_r1, res_l1);
SAT_UH4_UH(res_r0, res_l0, res_r1, res_l1, 7);
PCKEV_ST_SB(res_r0, res_l0, dst);
PCKEV_ST_SB(res_r1, res_l1, dst + 16);
dst += dst_stride;
}
}
Reference in New Issue Copy Permalink