Chiebot-Mirror
/
FFmpeg
mirror of https://github.com/FFmpeg/FFmpeg.git
8
0
Fork 0
Code Issues Projects Releases Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
112860 Commits
37 Branches
400 Tags
465 MiB
Tag: Branch: Tree: 5f87a68cf7
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from '5f87a68cf7'
${ noResults }
FFmpeg/libavcodec/mips/vp8dsp_mmi.c

3440 lines
154 KiB
Raw Normal View History Unescape

avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
* Loongson SIMD optimized vp8dsp
*
* Copyright (c) 2016 Loongson Technology Corporation Limited
* Copyright (c) 2016 Zhou Xiaoyong <zhouxiaoyong@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 "vp8dsp_mips.h"
#include "constants.h"
Include attributes.h directly Some files currently rely on libavutil/cpu.h to include it for them; yet said file won't use include it any more after the currently deprecated functions are removed, so include attributes.h directly. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
4 years ago
#include "libavutil/attributes.h"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
#include "libavutil/mips/mmiutils.h"
lavu/mem: move the DECLARE_ALIGNED macro family to mem_internal on next+1 bump They are not properly namespaced and not intended for public use.
5 years ago
#include "libavutil/mem_internal.h"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
#define DECLARE_DOUBLE_1 double db_1
#define DECLARE_DOUBLE_2 double db_2
#define DECLARE_UINT32_T uint32_t it_1
#define RESTRICT_ASM_DOUBLE_1 [db_1]"=&f"(db_1)
#define RESTRICT_ASM_DOUBLE_2 [db_2]"=&f"(db_2)
#define RESTRICT_ASM_UINT32_T [it_1]"=&r"(it_1)
#define MMI_PCMPGTUB(dst, src1, src2) \
"pcmpeqb %[db_1], "#src1", "#src2" \n\t" \
"pmaxub %[db_2], "#src1", "#src2" \n\t" \
"pcmpeqb %[db_2], %[db_2], "#src1" \n\t" \
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor "#dst", %[db_2], %[db_1] \n\t"
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
#define MMI_BTOH(dst_l, dst_r, src) \
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[db_1], %[db_1], %[db_1] \n\t" \
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
"pcmpgtb %[db_2], %[db_1], "#src" \n\t" \
"punpcklbh "#dst_r", "#src", %[db_2] \n\t" \
"punpckhbh "#dst_l", "#src", %[db_2] \n\t"
#define MMI_VP8_LOOP_FILTER \
/* Calculation of hev */ \
"dmtc1 %[thresh], %[ftmp3] \n\t" \
"punpcklbh %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \
"punpcklhw %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \
"punpcklwd %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \
"pasubub %[ftmp0], %[p1], %[p0] \n\t" \
"pasubub %[ftmp1], %[q1], %[q0] \n\t" \
"pmaxub %[ftmp0], %[ftmp0], %[ftmp1] \n\t" \
MMI_PCMPGTUB(%[hev], %[ftmp0], %[ftmp3]) \
/* Calculation of mask */ \
"pasubub %[ftmp1], %[p0], %[q0] \n\t" \
"paddusb %[ftmp1], %[ftmp1], %[ftmp1] \n\t" \
"pasubub %[ftmp2], %[p1], %[q1] \n\t" \
"li %[tmp0], 0x09 \n\t" \
"dmtc1 %[tmp0], %[ftmp3] \n\t" \
PSRLB_MMI(%[ftmp2], %[ftmp3], %[ftmp4], %[ftmp5], %[ftmp2]) \
"paddusb %[ftmp1], %[ftmp1], %[ftmp2] \n\t" \
"dmtc1 %[e], %[ftmp3] \n\t" \
"punpcklbh %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \
"punpcklhw %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \
"punpcklwd %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \
MMI_PCMPGTUB(%[mask], %[ftmp1], %[ftmp3]) \
"pmaxub %[mask], %[mask], %[ftmp0] \n\t" \
"pasubub %[ftmp1], %[p3], %[p2] \n\t" \
"pasubub %[ftmp2], %[p2], %[p1] \n\t" \
"pmaxub %[ftmp1], %[ftmp1], %[ftmp2] \n\t" \
"pmaxub %[mask], %[mask], %[ftmp1] \n\t" \
"pasubub %[ftmp1], %[q3], %[q2] \n\t" \
"pasubub %[ftmp2], %[q2], %[q1] \n\t" \
"pmaxub %[ftmp1], %[ftmp1], %[ftmp2] \n\t" \
"pmaxub %[mask], %[mask], %[ftmp1] \n\t" \
"dmtc1 %[i], %[ftmp3] \n\t" \
"punpcklbh %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \
"punpcklhw %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \
"punpcklwd %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \
MMI_PCMPGTUB(%[mask], %[mask], %[ftmp3]) \
"pcmpeqw %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[mask], %[mask], %[ftmp3] \n\t" \
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
/* VP8_MBFILTER */ \
"li %[tmp0], 0x80808080 \n\t" \
"dmtc1 %[tmp0], %[ftmp7] \n\t" \
"punpcklwd %[ftmp7], %[ftmp7], %[ftmp7] \n\t" \
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[p2], %[p2], %[ftmp7] \n\t" \
"pxor %[p1], %[p1], %[ftmp7] \n\t" \
"pxor %[p0], %[p0], %[ftmp7] \n\t" \
"pxor %[q0], %[q0], %[ftmp7] \n\t" \
"pxor %[q1], %[q1], %[ftmp7] \n\t" \
"pxor %[q2], %[q2], %[ftmp7] \n\t" \
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
"psubsb %[ftmp4], %[p1], %[q1] \n\t" \
"psubb %[ftmp5], %[q0], %[p0] \n\t" \
MMI_BTOH(%[ftmp1], %[ftmp0], %[ftmp5]) \
MMI_BTOH(%[ftmp3], %[ftmp2], %[ftmp4]) \
/* Right part */ \
"paddh %[ftmp5], %[ftmp0], %[ftmp0] \n\t" \
"paddh %[ftmp0], %[ftmp0], %[ftmp5] \n\t" \
"paddh %[ftmp0], %[ftmp2], %[ftmp0] \n\t" \
/* Left part */ \
"paddh %[ftmp5], %[ftmp1], %[ftmp1] \n\t" \
"paddh %[ftmp1], %[ftmp1], %[ftmp5] \n\t" \
"paddh %[ftmp1], %[ftmp3], %[ftmp1] \n\t" \
/* Combine left and right part */ \
"packsshb %[ftmp1], %[ftmp0], %[ftmp1] \n\t" \
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pand %[ftmp1], %[ftmp1], %[mask] \n\t" \
"pand %[ftmp2], %[ftmp1], %[hev] \n\t" \
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
"li %[tmp0], 0x04040404 \n\t" \
"dmtc1 %[tmp0], %[ftmp0] \n\t" \
"punpcklwd %[ftmp0], %[ftmp0], %[ftmp0] \n\t" \
"paddsb %[ftmp3], %[ftmp2], %[ftmp0] \n\t" \
"li %[tmp0], 0x0B \n\t" \
"dmtc1 %[tmp0], %[ftmp4] \n\t" \
PSRAB_MMI(%[ftmp3], %[ftmp4], %[ftmp5], %[ftmp6], %[ftmp3]) \
"li %[tmp0], 0x03030303 \n\t" \
"dmtc1 %[tmp0], %[ftmp0] \n\t" \
"punpcklwd %[ftmp0], %[ftmp0], %[ftmp0] \n\t" \
"paddsb %[ftmp4], %[ftmp2], %[ftmp0] \n\t" \
"li %[tmp0], 0x0B \n\t" \
"dmtc1 %[tmp0], %[ftmp2] \n\t" \
PSRAB_MMI(%[ftmp4], %[ftmp2], %[ftmp5], %[ftmp6], %[ftmp4]) \
"psubsb %[q0], %[q0], %[ftmp3] \n\t" \
"paddsb %[p0], %[p0], %[ftmp4] \n\t" \
/* filt_val &= ~hev */ \
"pcmpeqw %[ftmp0], %[ftmp0], %[ftmp0] \n\t" \
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[hev], %[hev], %[ftmp0] \n\t" \
"pand %[ftmp1], %[ftmp1], %[hev] \n\t" \
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
MMI_BTOH(%[ftmp5], %[ftmp6], %[ftmp1]) \
"li %[tmp0], 0x07 \n\t" \
"dmtc1 %[tmp0], %[ftmp2] \n\t" \
"li %[tmp0], 0x001b001b \n\t" \
"dmtc1 %[tmp0], %[ftmp1] \n\t" \
"punpcklwd %[ftmp1], %[ftmp1], %[ftmp1] \n\t" \
"li %[tmp0], 0x003f003f \n\t" \
"dmtc1 %[tmp0], %[ftmp0] \n\t" \
"punpcklwd %[ftmp0], %[ftmp0], %[ftmp0] \n\t" \
/* Right part */ \
"pmullh %[ftmp3], %[ftmp6], %[ftmp1] \n\t" \
"paddh %[ftmp3], %[ftmp3], %[ftmp0] \n\t" \
"psrah %[ftmp3], %[ftmp3], %[ftmp2] \n\t" \
/* Left part */ \
"pmullh %[ftmp4], %[ftmp5], %[ftmp1] \n\t" \
"paddh %[ftmp4], %[ftmp4], %[ftmp0] \n\t" \
"psrah %[ftmp4], %[ftmp4], %[ftmp2] \n\t" \
/* Combine left and right part */ \
"packsshb %[ftmp4], %[ftmp3], %[ftmp4] \n\t" \
"psubsb %[q0], %[q0], %[ftmp4] \n\t" \
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[q0], %[q0], %[ftmp7] \n\t" \
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
"paddsb %[p0], %[p0], %[ftmp4] \n\t" \
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[p0], %[p0], %[ftmp7] \n\t" \
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
"li %[tmp0], 0x00120012 \n\t" \
"dmtc1 %[tmp0], %[ftmp1] \n\t" \
"punpcklwd %[ftmp1], %[ftmp1], %[ftmp1] \n\t" \
/* Right part */ \
"pmullh %[ftmp3], %[ftmp6], %[ftmp1] \n\t" \
"paddh %[ftmp3], %[ftmp3], %[ftmp0] \n\t" \
"psrah %[ftmp3], %[ftmp3], %[ftmp2] \n\t" \
/* Left part */ \
"pmullh %[ftmp4], %[ftmp5], %[ftmp1] \n\t" \
"paddh %[ftmp4], %[ftmp4], %[ftmp0] \n\t" \
"psrah %[ftmp4], %[ftmp4], %[ftmp2] \n\t" \
/* Combine left and right part */ \
"packsshb %[ftmp4], %[ftmp3], %[ftmp4] \n\t" \
"psubsb %[q1], %[q1], %[ftmp4] \n\t" \
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[q1], %[q1], %[ftmp7] \n\t" \
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
"paddsb %[p1], %[p1], %[ftmp4] \n\t" \
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[p1], %[p1], %[ftmp7] \n\t" \
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
"li %[tmp0], 0x03 \n\t" \
"dmtc1 %[tmp0], %[ftmp1] \n\t" \
/* Right part */ \
"psllh %[ftmp3], %[ftmp6], %[ftmp1] \n\t" \
"paddh %[ftmp3], %[ftmp3], %[ftmp6] \n\t" \
"paddh %[ftmp3], %[ftmp3], %[ftmp0] \n\t" \
"psrah %[ftmp3], %[ftmp3], %[ftmp2] \n\t" \
/* Left part */ \
"psllh %[ftmp4], %[ftmp5], %[ftmp1] \n\t" \
"paddh %[ftmp4], %[ftmp4], %[ftmp5] \n\t" \
"paddh %[ftmp4], %[ftmp4], %[ftmp0] \n\t" \
"psrah %[ftmp4], %[ftmp4], %[ftmp2] \n\t" \
/* Combine left and right part */ \
"packsshb %[ftmp4], %[ftmp3], %[ftmp4] \n\t" \
"psubsb %[q2], %[q2], %[ftmp4] \n\t" \
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[q2], %[q2], %[ftmp7] \n\t" \
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
"paddsb %[p2], %[p2], %[ftmp4] \n\t" \
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[p2], %[p2], %[ftmp7] \n\t"
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
#define PUT_VP8_EPEL4_H6_MMI(src, dst) \
MMI_ULWC1(%[ftmp1], src, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp3], %[ftmp2], %[filter2] \n\t" \
\
MMI_ULWC1(%[ftmp1], src, -0x01) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter1] \n\t" \
"psubsh %[ftmp3], %[ftmp3], %[ftmp2] \n\t" \
\
MMI_ULWC1(%[ftmp1], src, -0x02) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter0] \n\t" \
"paddsh %[ftmp5], %[ftmp3], %[ftmp2] \n\t" \
\
MMI_ULWC1(%[ftmp1], src, 0x01) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp3], %[ftmp2], %[filter3] \n\t" \
\
MMI_ULWC1(%[ftmp1], src, 0x02) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter4] \n\t" \
"psubsh %[ftmp3], %[ftmp3], %[ftmp2] \n\t" \
\
MMI_ULWC1(%[ftmp1], src, 0x03) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter5] \n\t" \
"paddsh %[ftmp3], %[ftmp3], %[ftmp2] \n\t" \
\
"paddsh %[ftmp3], %[ftmp3], %[ftmp5] \n\t" \
"paddsh %[ftmp3], %[ftmp3], %[ff_pw_64] \n\t" \
"psrah %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
"packushb %[ftmp1], %[ftmp3], %[ftmp0] \n\t" \
\
MMI_SWC1(%[ftmp1], dst, 0x00)
#define PUT_VP8_EPEL4_H4_MMI(src, dst) \
MMI_ULWC1(%[ftmp1], src, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp3], %[ftmp2], %[filter2] \n\t" \
\
MMI_ULWC1(%[ftmp1], src, -0x01) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter1] \n\t" \
"psubsh %[ftmp5], %[ftmp3], %[ftmp2] \n\t" \
\
MMI_ULWC1(%[ftmp1], src, 0x01) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp3], %[ftmp2], %[filter3] \n\t" \
\
MMI_ULWC1(%[ftmp1], src, 0x02) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter4] \n\t" \
"psubh %[ftmp3], %[ftmp3], %[ftmp2] \n\t" \
\
"paddsh %[ftmp3], %[ftmp3], %[ftmp5] \n\t" \
\
"paddsh %[ftmp3], %[ftmp3], %[ff_pw_64] \n\t" \
"psrah %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
\
"packushb %[ftmp1], %[ftmp3], %[ftmp0] \n\t" \
MMI_SWC1(%[ftmp1], dst, 0x00)
#define PUT_VP8_EPEL4_V6_MMI(src, src1, dst, srcstride) \
MMI_ULWC1(%[ftmp1], src, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp3], %[ftmp2], %[filter2] \n\t" \
\
PTR_SUBU ""#src1", "#src", "#srcstride" \n\t" \
MMI_ULWC1(%[ftmp1], src1, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter1] \n\t" \
"psubsh %[ftmp3], %[ftmp3], %[ftmp2] \n\t" \
\
PTR_SUBU ""#src1", "#src1", "#srcstride" \n\t" \
MMI_ULWC1(%[ftmp1], src1, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter0] \n\t" \
"paddsh %[ftmp5], %[ftmp3], %[ftmp2] \n\t" \
\
PTR_ADDU ""#src1", "#src", "#srcstride" \n\t" \
MMI_ULWC1(%[ftmp1], src1, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp3], %[ftmp2], %[filter3] \n\t" \
\
PTR_ADDU ""#src1", "#src1", "#srcstride" \n\t" \
MMI_ULWC1(%[ftmp1], src1, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter4] \n\t" \
"psubsh %[ftmp3], %[ftmp3], %[ftmp2] \n\t" \
\
PTR_ADDU ""#src1", "#src1", "#srcstride" \n\t" \
MMI_ULWC1(%[ftmp1], src1, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter5] \n\t" \
"paddsh %[ftmp3], %[ftmp3], %[ftmp2] \n\t" \
\
"paddsh %[ftmp3], %[ftmp3], %[ftmp5] \n\t" \
\
"paddsh %[ftmp3], %[ftmp3], %[ff_pw_64] \n\t" \
"psrah %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
"packushb %[ftmp1], %[ftmp3], %[ftmp0] \n\t" \
\
MMI_SWC1(%[ftmp1], dst, 0x00)
#define PUT_VP8_EPEL4_V4_MMI(src, src1, dst, srcstride) \
MMI_ULWC1(%[ftmp1], src, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp3], %[ftmp2], %[filter2] \n\t" \
\
PTR_SUBU ""#src1", "#src", "#srcstride" \n\t" \
MMI_ULWC1(%[ftmp1], src1, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter1] \n\t" \
"psubsh %[ftmp5], %[ftmp3], %[ftmp2] \n\t" \
\
PTR_ADDU ""#src1", "#src", "#srcstride" \n\t" \
MMI_ULWC1(%[ftmp1], src1, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp3], %[ftmp2], %[filter3] \n\t" \
\
PTR_ADDU ""#src1", "#src1", "#srcstride" \n\t" \
MMI_ULWC1(%[ftmp1], src1, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter4] \n\t" \
"psubsh %[ftmp3], %[ftmp3], %[ftmp2] \n\t" \
\
"paddsh %[ftmp3], %[ftmp3], %[ftmp5] \n\t" \
\
"paddsh %[ftmp3], %[ftmp3], %[ff_pw_64] \n\t" \
"psrah %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
"packushb %[ftmp1], %[ftmp3], %[ftmp0] \n\t" \
\
MMI_SWC1(%[ftmp1], dst, 0x00)
#define PUT_VP8_EPEL8_H6_MMI(src, dst) \
MMI_ULDC1(%[ftmp1], src, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp5], %[ftmp2], %[filter2] \n\t" \
"pmullh %[ftmp6], %[ftmp3], %[filter2] \n\t" \
\
MMI_ULDC1(%[ftmp1], src, -0x01) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter1] \n\t" \
"pmullh %[ftmp3], %[ftmp3], %[filter1] \n\t" \
"psubsh %[ftmp5], %[ftmp5], %[ftmp2] \n\t" \
"psubsh %[ftmp6], %[ftmp6], %[ftmp3] \n\t" \
\
MMI_ULDC1(%[ftmp1], src, -0x02) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter0] \n\t" \
"pmullh %[ftmp3], %[ftmp3], %[filter0] \n\t" \
"paddsh %[ftmp7], %[ftmp5], %[ftmp2] \n\t" \
"paddsh %[ftmp8], %[ftmp6], %[ftmp3] \n\t" \
\
MMI_ULDC1(%[ftmp1], src, 0x01) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp5], %[ftmp2], %[filter3] \n\t" \
"pmullh %[ftmp6], %[ftmp3], %[filter3] \n\t" \
\
MMI_ULDC1(%[ftmp1], src, 0x02) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter4] \n\t" \
"pmullh %[ftmp3], %[ftmp3], %[filter4] \n\t" \
"psubsh %[ftmp5], %[ftmp5], %[ftmp2] \n\t" \
"psubsh %[ftmp6], %[ftmp6], %[ftmp3] \n\t" \
\
MMI_ULDC1(%[ftmp1], src, 0x03) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter5] \n\t" \
"pmullh %[ftmp3], %[ftmp3], %[filter5] \n\t" \
"paddsh %[ftmp5], %[ftmp5], %[ftmp2] \n\t" \
"paddsh %[ftmp6], %[ftmp6], %[ftmp3] \n\t" \
\
"paddsh %[ftmp5], %[ftmp5], %[ftmp7] \n\t" \
"paddsh %[ftmp6], %[ftmp6], %[ftmp8] \n\t" \
\
"paddsh %[ftmp5], %[ftmp5], %[ff_pw_64] \n\t" \
"paddsh %[ftmp6], %[ftmp6], %[ff_pw_64] \n\t" \
"psrah %[ftmp5], %[ftmp5], %[ftmp4] \n\t" \
"psrah %[ftmp6], %[ftmp6], %[ftmp4] \n\t" \
"packushb %[ftmp1], %[ftmp5], %[ftmp6] \n\t" \
\
MMI_SDC1(%[ftmp1], dst, 0x00)
#define PUT_VP8_EPEL8_H4_MMI(src, dst) \
MMI_ULDC1(%[ftmp1], src, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp5], %[ftmp2], %[filter2] \n\t" \
"pmullh %[ftmp6], %[ftmp3], %[filter2] \n\t" \
\
MMI_ULDC1(%[ftmp1], src, -0x01) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter1] \n\t" \
"pmullh %[ftmp3], %[ftmp3], %[filter1] \n\t" \
"psubsh %[ftmp7], %[ftmp5], %[ftmp2] \n\t" \
"psubsh %[ftmp8], %[ftmp6], %[ftmp3] \n\t" \
\
MMI_ULDC1(%[ftmp1], src, 0x01) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp5], %[ftmp2], %[filter3] \n\t" \
"pmullh %[ftmp6], %[ftmp3], %[filter3] \n\t" \
\
MMI_ULDC1(%[ftmp1], src, 0x02) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter4] \n\t" \
"pmullh %[ftmp3], %[ftmp3], %[filter4] \n\t" \
"psubsh %[ftmp5], %[ftmp5], %[ftmp2] \n\t" \
"psubsh %[ftmp6], %[ftmp6], %[ftmp3] \n\t" \
\
"paddsh %[ftmp5], %[ftmp5], %[ftmp7] \n\t" \
"paddsh %[ftmp6], %[ftmp6], %[ftmp8] \n\t" \
\
"paddsh %[ftmp5], %[ftmp5], %[ff_pw_64] \n\t" \
"paddsh %[ftmp6], %[ftmp6], %[ff_pw_64] \n\t" \
"psrah %[ftmp5], %[ftmp5], %[ftmp4] \n\t" \
"psrah %[ftmp6], %[ftmp6], %[ftmp4] \n\t" \
\
"packushb %[ftmp1], %[ftmp5], %[ftmp6] \n\t" \
MMI_SDC1(%[ftmp1], dst, 0x00)
#define PUT_VP8_EPEL8_V6_MMI(src, src1, dst, srcstride) \
MMI_ULDC1(%[ftmp1], src, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp5], %[ftmp2], %[filter2] \n\t" \
"pmullh %[ftmp6], %[ftmp3], %[filter2] \n\t" \
\
PTR_SUBU ""#src1", "#src", "#srcstride" \n\t" \
MMI_ULDC1(%[ftmp1], src1, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter1] \n\t" \
"pmullh %[ftmp3], %[ftmp3], %[filter1] \n\t" \
"psubsh %[ftmp5], %[ftmp5], %[ftmp2] \n\t" \
"psubsh %[ftmp6], %[ftmp6], %[ftmp3] \n\t" \
\
PTR_SUBU ""#src1", "#src1", "#srcstride" \n\t" \
MMI_ULDC1(%[ftmp1], src1, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter0] \n\t" \
"pmullh %[ftmp3], %[ftmp3], %[filter0] \n\t" \
"paddsh %[ftmp7], %[ftmp5], %[ftmp2] \n\t" \
"paddsh %[ftmp8], %[ftmp6], %[ftmp3] \n\t" \
\
PTR_ADDU ""#src1", "#src", "#srcstride" \n\t" \
MMI_ULDC1(%[ftmp1], src1, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp5], %[ftmp2], %[filter3] \n\t" \
"pmullh %[ftmp6], %[ftmp3], %[filter3] \n\t" \
\
PTR_ADDU ""#src1", "#src1", "#srcstride" \n\t" \
MMI_ULDC1(%[ftmp1], src1, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter4] \n\t" \
"pmullh %[ftmp3], %[ftmp3], %[filter4] \n\t" \
"psubsh %[ftmp5], %[ftmp5], %[ftmp2] \n\t" \
"psubsh %[ftmp6], %[ftmp6], %[ftmp3] \n\t" \
\
PTR_ADDU ""#src1", "#src1", "#srcstride" \n\t" \
MMI_ULDC1(%[ftmp1], src1, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter5] \n\t" \
"pmullh %[ftmp3], %[ftmp3], %[filter5] \n\t" \
"paddsh %[ftmp5], %[ftmp5], %[ftmp2] \n\t" \
"paddsh %[ftmp6], %[ftmp6], %[ftmp3] \n\t" \
\
"paddsh %[ftmp5], %[ftmp5], %[ftmp7] \n\t" \
"paddsh %[ftmp6], %[ftmp6], %[ftmp8] \n\t" \
\
"paddsh %[ftmp5], %[ftmp5], %[ff_pw_64] \n\t" \
"paddsh %[ftmp6], %[ftmp6], %[ff_pw_64] \n\t" \
"psrah %[ftmp5], %[ftmp5], %[ftmp4] \n\t" \
"psrah %[ftmp6], %[ftmp6], %[ftmp4] \n\t" \
"packushb %[ftmp1], %[ftmp5], %[ftmp6] \n\t" \
\
MMI_SDC1(%[ftmp1], dst, 0x00)
#define PUT_VP8_EPEL8_V4_MMI(src, src1, dst, srcstride) \
MMI_ULDC1(%[ftmp1], src, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp5], %[ftmp2], %[filter2] \n\t" \
"pmullh %[ftmp6], %[ftmp3], %[filter2] \n\t" \
\
PTR_SUBU ""#src1", "#src", "#srcstride" \n\t" \
MMI_ULDC1(%[ftmp1], src1, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter1] \n\t" \
"pmullh %[ftmp3], %[ftmp3], %[filter1] \n\t" \
"psubsh %[ftmp7], %[ftmp5], %[ftmp2] \n\t" \
"psubsh %[ftmp8], %[ftmp6], %[ftmp3] \n\t" \
\
PTR_ADDU ""#src1", "#src", "#srcstride" \n\t" \
MMI_ULDC1(%[ftmp1], src1, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp5], %[ftmp2], %[filter3] \n\t" \
"pmullh %[ftmp6], %[ftmp3], %[filter3] \n\t" \
\
PTR_ADDU ""#src1", "#src1", "#srcstride" \n\t" \
MMI_ULDC1(%[ftmp1], src1, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[filter4] \n\t" \
"pmullh %[ftmp3], %[ftmp3], %[filter4] \n\t" \
"psubsh %[ftmp5], %[ftmp5], %[ftmp2] \n\t" \
"psubsh %[ftmp6], %[ftmp6], %[ftmp3] \n\t" \
\
"paddsh %[ftmp5], %[ftmp5], %[ftmp7] \n\t" \
"paddsh %[ftmp6], %[ftmp6], %[ftmp8] \n\t" \
\
"paddsh %[ftmp5], %[ftmp5], %[ff_pw_64] \n\t" \
"paddsh %[ftmp6], %[ftmp6], %[ff_pw_64] \n\t" \
"psrah %[ftmp5], %[ftmp5], %[ftmp4] \n\t" \
"psrah %[ftmp6], %[ftmp6], %[ftmp4] \n\t" \
"packushb %[ftmp1], %[ftmp5], %[ftmp6] \n\t" \
\
MMI_SDC1(%[ftmp1], dst, 0x00)
#define PUT_VP8_BILINEAR8_H_MMI(src, dst) \
MMI_ULDC1(%[ftmp1], src, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp5], %[ftmp2], %[a] \n\t" \
"pmullh %[ftmp6], %[ftmp3], %[a] \n\t" \
\
MMI_ULDC1(%[ftmp1], src, 0x01) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[b] \n\t" \
"pmullh %[ftmp3], %[ftmp3], %[b] \n\t" \
"paddsh %[ftmp5], %[ftmp5], %[ftmp2] \n\t" \
"paddsh %[ftmp6], %[ftmp6], %[ftmp3] \n\t" \
\
"paddsh %[ftmp5], %[ftmp5], %[ff_pw_4] \n\t" \
"paddsh %[ftmp6], %[ftmp6], %[ff_pw_4] \n\t" \
"psrah %[ftmp5], %[ftmp5], %[ftmp4] \n\t" \
"psrah %[ftmp6], %[ftmp6], %[ftmp4] \n\t" \
\
"packushb %[ftmp1], %[ftmp5], %[ftmp6] \n\t" \
MMI_SDC1(%[ftmp1], dst, 0x00)
#define PUT_VP8_BILINEAR4_H_MMI(src, dst) \
MMI_ULWC1(%[ftmp1], src, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp3], %[ftmp2], %[a] \n\t" \
\
MMI_ULWC1(%[ftmp1], src, 0x01) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[b] \n\t" \
"paddsh %[ftmp3], %[ftmp3], %[ftmp2] \n\t" \
\
"paddsh %[ftmp3], %[ftmp3], %[ff_pw_4] \n\t" \
"psrah %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
\
"packushb %[ftmp1], %[ftmp3], %[ftmp0] \n\t" \
MMI_SWC1(%[ftmp1], dst, 0x00)
#define PUT_VP8_BILINEAR8_V_MMI(src, src1, dst, sstride) \
MMI_ULDC1(%[ftmp1], src, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp5], %[ftmp2], %[c] \n\t" \
"pmullh %[ftmp6], %[ftmp3], %[c] \n\t" \
\
PTR_ADDU ""#src1", "#src", "#sstride" \n\t" \
MMI_ULDC1(%[ftmp1], src1, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[d] \n\t" \
"pmullh %[ftmp3], %[ftmp3], %[d] \n\t" \
"paddsh %[ftmp5], %[ftmp5], %[ftmp2] \n\t" \
"paddsh %[ftmp6], %[ftmp6], %[ftmp3] \n\t" \
\
"paddsh %[ftmp5], %[ftmp5], %[ff_pw_4] \n\t" \
"paddsh %[ftmp6], %[ftmp6], %[ff_pw_4] \n\t" \
"psrah %[ftmp5], %[ftmp5], %[ftmp4] \n\t" \
"psrah %[ftmp6], %[ftmp6], %[ftmp4] \n\t" \
\
"packushb %[ftmp1], %[ftmp5], %[ftmp6] \n\t" \
MMI_SDC1(%[ftmp1], dst, 0x00)
#define PUT_VP8_BILINEAR4_V_MMI(src, src1, dst, sstride) \
MMI_ULWC1(%[ftmp1], src, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp3], %[ftmp2], %[c] \n\t" \
\
PTR_ADDU ""#src1", "#src", "#sstride" \n\t" \
MMI_ULWC1(%[ftmp1], src1, 0x00) \
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \
"pmullh %[ftmp2], %[ftmp2], %[d] \n\t" \
"paddsh %[ftmp3], %[ftmp3], %[ftmp2] \n\t" \
\
"paddsh %[ftmp3], %[ftmp3], %[ff_pw_4] \n\t" \
"psrah %[ftmp3], %[ftmp3], %[ftmp4] \n\t" \
\
"packushb %[ftmp1], %[ftmp3], %[ftmp0] \n\t" \
MMI_SWC1(%[ftmp1], dst, 0x00)
DECLARE_ALIGNED(8, static const uint64_t, fourtap_subpel_filters[7][6]) = {
{0x0000000000000000, 0x0006000600060006, 0x007b007b007b007b,
0x000c000c000c000c, 0x0001000100010001, 0x0000000000000000},
{0x0002000200020002, 0x000b000b000b000b, 0x006c006c006c006c,
0x0024002400240024, 0x0008000800080008, 0x0001000100010001},
{0x0000000000000000, 0x0009000900090009, 0x005d005d005d005d,
0x0032003200320032, 0x0006000600060006, 0x0000000000000000},
{0x0003000300030003, 0x0010001000100010, 0x004d004d004d004d,
0x004d004d004d004d, 0x0010001000100010, 0x0003000300030003},
{0x0000000000000000, 0x0006000600060006, 0x0032003200320032,
0x005d005d005d005d, 0x0009000900090009, 0x0000000000000000},
{0x0001000100010001, 0x0008000800080008, 0x0024002400240024,
0x006c006c006c006c, 0x000b000b000b000b, 0x0002000200020002},
{0x0000000000000000, 0x0001000100010001, 0x000c000c000c000c,
0x007b007b007b007b, 0x0006000600060006, 0x0000000000000000}
};
#if 0
#define FILTER_6TAP(src, F, stride) \
cm[(F[2] * src[x + 0 * stride] - F[1] * src[x - 1 * stride] + \
F[0] * src[x - 2 * stride] + F[3] * src[x + 1 * stride] - \
F[4] * src[x + 2 * stride] + F[5] * src[x + 3 * stride] + 64) >> 7]
#define FILTER_4TAP(src, F, stride) \
cm[(F[2] * src[x + 0 * stride] - F[1] * src[x - 1 * stride] + \
F[3] * src[x + 1 * stride] - F[4] * src[x + 2 * stride] + 64) >> 7]
static const uint8_t subpel_filters[7][6] = {
{ 0, 6, 123, 12, 1, 0 },
{ 2, 11, 108, 36, 8, 1 },
{ 0, 9, 93, 50, 6, 0 },
{ 3, 16, 77, 77, 16, 3 },
{ 0, 6, 50, 93, 9, 0 },
{ 1, 8, 36, 108, 11, 2 },
{ 0, 1, 12, 123, 6, 0 },
};
#define MUL_20091(a) ((((a) * 20091) >> 16) + (a))
#define MUL_35468(a) (((a) * 35468) >> 16)
#endif
#define clip_int8(n) (cm[(n) + 0x80] - 0x80)
static av_always_inline void vp8_filter_common_is4tap(uint8_t *p,
ptrdiff_t stride)
{
int av_unused p1 = p[-2 * stride];
int av_unused p0 = p[-1 * stride];
int av_unused q0 = p[ 0 * stride];
int av_unused q1 = p[ 1 * stride];
int a, f1, f2;
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
a = 3 * (q0 - p0);
a += clip_int8(p1 - q1);
a = clip_int8(a);
// We deviate from the spec here with c(a+3) >> 3
// since that's what libvpx does.
f1 = FFMIN(a + 4, 127) >> 3;
f2 = FFMIN(a + 3, 127) >> 3;
// Despite what the spec says, we do need to clamp here to
// be bitexact with libvpx.
p[-1 * stride] = cm[p0 + f2];
p[ 0 * stride] = cm[q0 - f1];
}
static av_always_inline void vp8_filter_common_isnot4tap(uint8_t *p,
ptrdiff_t stride)
{
int av_unused p1 = p[-2 * stride];
int av_unused p0 = p[-1 * stride];
int av_unused q0 = p[ 0 * stride];
int av_unused q1 = p[ 1 * stride];
int a, f1, f2;
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
a = 3 * (q0 - p0);
a = clip_int8(a);
// We deviate from the spec here with c(a+3) >> 3
// since that's what libvpx does.
f1 = FFMIN(a + 4, 127) >> 3;
f2 = FFMIN(a + 3, 127) >> 3;
// Despite what the spec says, we do need to clamp here to
// be bitexact with libvpx.
p[-1 * stride] = cm[p0 + f2];
p[ 0 * stride] = cm[q0 - f1];
a = (f1 + 1) >> 1;
p[-2 * stride] = cm[p1 + a];
p[ 1 * stride] = cm[q1 - a];
}
static av_always_inline int vp8_simple_limit(uint8_t *p, ptrdiff_t stride,
int flim)
{
int av_unused p1 = p[-2 * stride];
int av_unused p0 = p[-1 * stride];
int av_unused q0 = p[ 0 * stride];
int av_unused q1 = p[ 1 * stride];
return 2 * FFABS(p0 - q0) + (FFABS(p1 - q1) >> 1) <= flim;
}
static av_always_inline int hev(uint8_t *p, ptrdiff_t stride, int thresh)
{
int av_unused p1 = p[-2 * stride];
int av_unused p0 = p[-1 * stride];
int av_unused q0 = p[ 0 * stride];
int av_unused q1 = p[ 1 * stride];
return FFABS(p1 - p0) > thresh || FFABS(q1 - q0) > thresh;
}
static av_always_inline void filter_mbedge(uint8_t *p, ptrdiff_t stride)
{
int a0, a1, a2, w;
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int av_unused p2 = p[-3 * stride];
int av_unused p1 = p[-2 * stride];
int av_unused p0 = p[-1 * stride];
int av_unused q0 = p[ 0 * stride];
int av_unused q1 = p[ 1 * stride];
int av_unused q2 = p[ 2 * stride];
w = clip_int8(p1 - q1);
w = clip_int8(w + 3 * (q0 - p0));
a0 = (27 * w + 63) >> 7;
a1 = (18 * w + 63) >> 7;
a2 = (9 * w + 63) >> 7;
p[-3 * stride] = cm[p2 + a2];
p[-2 * stride] = cm[p1 + a1];
p[-1 * stride] = cm[p0 + a0];
p[ 0 * stride] = cm[q0 - a0];
p[ 1 * stride] = cm[q1 - a1];
p[ 2 * stride] = cm[q2 - a2];
}
static av_always_inline int vp8_normal_limit(uint8_t *p, ptrdiff_t stride,
int E, int I)
{
int av_unused p3 = p[-4 * stride];
int av_unused p2 = p[-3 * stride];
int av_unused p1 = p[-2 * stride];
int av_unused p0 = p[-1 * stride];
int av_unused q0 = p[ 0 * stride];
int av_unused q1 = p[ 1 * stride];
int av_unused q2 = p[ 2 * stride];
int av_unused q3 = p[ 3 * stride];
return vp8_simple_limit(p, stride, E) &&
FFABS(p3 - p2) <= I && FFABS(p2 - p1) <= I &&
FFABS(p1 - p0) <= I && FFABS(q3 - q2) <= I &&
FFABS(q2 - q1) <= I && FFABS(q1 - q0) <= I;
}
static av_always_inline void vp8_v_loop_filter8_mmi(uint8_t *dst,
ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh)
{
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
double ftmp[18];
uint32_t tmp[1];
DECLARE_DOUBLE_1;
DECLARE_DOUBLE_2;
DECLARE_UINT32_T;
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
DECLARE_VAR_ALL64;
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
__asm__ volatile(
/* Get data from dst */
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_ULDC1(%[q0], %[dst], 0x0)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_SUBU "%[tmp0], %[dst], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_ULDC1(%[p0], %[tmp0], 0x0)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_SUBU "%[tmp0], %[tmp0], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_ULDC1(%[p1], %[tmp0], 0x0)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_SUBU "%[tmp0], %[tmp0], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_ULDC1(%[p2], %[tmp0], 0x0)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_SUBU "%[tmp0], %[tmp0], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_ULDC1(%[p3], %[tmp0], 0x0)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[tmp0], %[dst], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_ULDC1(%[q1], %[tmp0], 0x0)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[tmp0], %[tmp0], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_ULDC1(%[q2], %[tmp0], 0x0)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[tmp0], %[tmp0], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_ULDC1(%[q3], %[tmp0], 0x0)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
MMI_VP8_LOOP_FILTER
/* Move to dst */
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_USDC1(%[q0], %[dst], 0x0)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_SUBU "%[tmp0], %[dst], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_USDC1(%[p0], %[tmp0], 0x0)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_SUBU "%[tmp0], %[tmp0], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_USDC1(%[p1], %[tmp0], 0x0)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_SUBU "%[tmp0], %[tmp0], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_USDC1(%[p2], %[tmp0], 0x0)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[tmp0], %[dst], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_USDC1(%[q1], %[tmp0], 0x0)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[tmp0], %[tmp0], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_USDC1(%[q2], %[tmp0], 0x0)
: RESTRICT_ASM_ALL64
[p3]"=&f"(ftmp[0]), [p2]"=&f"(ftmp[1]),
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
[p1]"=&f"(ftmp[2]), [p0]"=&f"(ftmp[3]),
[q0]"=&f"(ftmp[4]), [q1]"=&f"(ftmp[5]),
[q2]"=&f"(ftmp[6]), [q3]"=&f"(ftmp[7]),
[ftmp0]"=&f"(ftmp[8]), [ftmp1]"=&f"(ftmp[9]),
[ftmp2]"=&f"(ftmp[10]), [ftmp3]"=&f"(ftmp[11]),
[hev]"=&f"(ftmp[12]), [mask]"=&f"(ftmp[13]),
[ftmp4]"=&f"(ftmp[14]), [ftmp5]"=&f"(ftmp[15]),
[ftmp6]"=&f"(ftmp[16]), [ftmp7]"=&f"(ftmp[17]),
[dst]"+&r"(dst), [tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_DOUBLE_1, RESTRICT_ASM_DOUBLE_2,
RESTRICT_ASM_UINT32_T
: [e]"r"((mips_reg)flim_E), [thresh]"r"((mips_reg)hev_thresh),
[i]"r"((mips_reg)flim_I), [stride]"r"((mips_reg)stride)
: "memory"
);
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
}
static av_always_inline void vp8_v_loop_filter8_inner_mmi(uint8_t *dst,
ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh)
{
int i;
for (i = 0; i < 8; i++)
if (vp8_normal_limit(dst + i * 1, stride, flim_E, flim_I)) {
int hv = hev(dst + i * 1, stride, hev_thresh);
if (hv)
vp8_filter_common_is4tap(dst + i * 1, stride);
else
vp8_filter_common_isnot4tap(dst + i * 1, stride);
}
}
static av_always_inline void vp8_h_loop_filter8_mmi(uint8_t *dst,
ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh)
{
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
double ftmp[18];
uint32_t tmp[1];
DECLARE_DOUBLE_1;
DECLARE_DOUBLE_2;
DECLARE_UINT32_T;
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
DECLARE_VAR_ALL64;
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
__asm__ volatile(
/* Get data from dst */
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_ULDC1(%[p3], %[dst], -0x04)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[tmp0], %[dst], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_ULDC1(%[p2], %[tmp0], -0x04)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[tmp0], %[tmp0], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_ULDC1(%[p1], %[tmp0], -0x04)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[tmp0], %[tmp0], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_ULDC1(%[p0], %[tmp0], -0x04)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[tmp0], %[tmp0], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_ULDC1(%[q0], %[tmp0], -0x04)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[tmp0], %[tmp0], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_ULDC1(%[q1], %[tmp0], -0x04)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[tmp0], %[tmp0], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_ULDC1(%[q2], %[tmp0], -0x04)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[tmp0], %[tmp0], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_ULDC1(%[q3], %[tmp0], -0x04)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
/* Matrix transpose */
avutil/mips: [loongson] simplify macro TRANSPOSE_4H and TRANSPOSE_8B Simplify macro TRANSPOSE_4H in mmiutils.h and add TRANSPOSE_8B as a common macro. Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
TRANSPOSE_8B(%[p3], %[p2], %[p1], %[p0],
%[q0], %[q1], %[q2], %[q3],
%[ftmp1], %[ftmp2], %[ftmp3], %[ftmp4])
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
MMI_VP8_LOOP_FILTER
/* Matrix transpose */
avutil/mips: [loongson] simplify macro TRANSPOSE_4H and TRANSPOSE_8B Simplify macro TRANSPOSE_4H in mmiutils.h and add TRANSPOSE_8B as a common macro. Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
TRANSPOSE_8B(%[p3], %[p2], %[p1], %[p0],
%[q0], %[q1], %[q2], %[q3],
%[ftmp1], %[ftmp2], %[ftmp3], %[ftmp4])
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
/* Move to dst */
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_USDC1(%[p3], %[dst], -0x04)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[dst], %[dst], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_USDC1(%[p2], %[dst], -0x04)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[dst], %[dst], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_USDC1(%[p1], %[dst], -0x04)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[dst], %[dst], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_USDC1(%[p0], %[dst], -0x04)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[dst], %[dst], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_USDC1(%[q0], %[dst], -0x04)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[dst], %[dst], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_USDC1(%[q1], %[dst], -0x04)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[dst], %[dst], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_USDC1(%[q2], %[dst], -0x04)
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
PTR_ADDU "%[dst], %[dst], %[stride] \n\t"
avcodec/mips: Use MMI marcos to replace Loongson3 instructions Loongson3's extention instructions (prefixed with gs) are widely used in our MMI codebase. However, these instructions are not avilable on Loongson-2E/F while MMI code should work on these processors. Previously we introduced mmiutils marcos to provide backward compactbility but newly commited code didn't follow that. In this patch I revised the codebase and converted all these instructions into MMI marcos to get Loongson2 supproted again. Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
3 years ago
MMI_USDC1(%[q3], %[dst], -0x04)
: RESTRICT_ASM_ALL64
[p3]"=&f"(ftmp[0]), [p2]"=&f"(ftmp[1]),
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
[p1]"=&f"(ftmp[2]), [p0]"=&f"(ftmp[3]),
[q0]"=&f"(ftmp[4]), [q1]"=&f"(ftmp[5]),
[q2]"=&f"(ftmp[6]), [q3]"=&f"(ftmp[7]),
[ftmp0]"=&f"(ftmp[8]), [ftmp1]"=&f"(ftmp[9]),
[ftmp2]"=&f"(ftmp[10]), [ftmp3]"=&f"(ftmp[11]),
[hev]"=&f"(ftmp[12]), [mask]"=&f"(ftmp[13]),
[ftmp4]"=&f"(ftmp[14]), [ftmp5]"=&f"(ftmp[15]),
[ftmp6]"=&f"(ftmp[16]), [ftmp7]"=&f"(ftmp[17]),
[dst]"+&r"(dst), [tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_DOUBLE_1, RESTRICT_ASM_DOUBLE_2,
RESTRICT_ASM_UINT32_T
: [e]"r"((mips_reg)flim_E), [thresh]"r"((mips_reg)hev_thresh),
[i]"r"((mips_reg)flim_I), [stride]"r"((mips_reg)stride)
: "memory"
);
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
}
static av_always_inline void vp8_h_loop_filter8_inner_mmi(uint8_t *dst,
ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh)
{
int i;
for (i = 0; i < 8; i++)
if (vp8_normal_limit(dst + i * stride, 1, flim_E, flim_I)) {
int hv = hev(dst + i * stride, 1, hev_thresh);
if (hv)
vp8_filter_common_is4tap(dst + i * stride, 1);
else
vp8_filter_common_isnot4tap(dst + i * stride, 1);
}
}
void ff_vp8_luma_dc_wht_mmi(int16_t block[4][4][16], int16_t dc[16])
{
#if 1
double ftmp[8];
DECLARE_VAR_ALL64;
__asm__ volatile (
MMI_LDC1(%[ftmp0], %[dc], 0x00)
MMI_LDC1(%[ftmp1], %[dc], 0x08)
MMI_LDC1(%[ftmp2], %[dc], 0x10)
MMI_LDC1(%[ftmp3], %[dc], 0x18)
"paddsh %[ftmp4], %[ftmp0], %[ftmp3] \n\t"
"psubsh %[ftmp5], %[ftmp0], %[ftmp3] \n\t"
"paddsh %[ftmp6], %[ftmp1], %[ftmp2] \n\t"
"psubsh %[ftmp7], %[ftmp1], %[ftmp2] \n\t"
"paddsh %[ftmp0], %[ftmp4], %[ftmp6] \n\t"
"paddsh %[ftmp1], %[ftmp5], %[ftmp7] \n\t"
"psubsh %[ftmp2], %[ftmp4], %[ftmp6] \n\t"
"psubsh %[ftmp3], %[ftmp5], %[ftmp7] \n\t"
MMI_SDC1(%[ftmp0], %[dc], 0x00)
MMI_SDC1(%[ftmp1], %[dc], 0x08)
MMI_SDC1(%[ftmp2], %[dc], 0x10)
MMI_SDC1(%[ftmp3], %[dc], 0x18)
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
[ftmp6]"=&f"(ftmp[6]),
RESTRICT_ASM_ALL64
[ftmp7]"=&f"(ftmp[7])
: [dc]"r"((uint8_t*)dc)
: "memory"
);
block[0][0][0] = (dc[0] + dc[3] + 3 + dc[1] + dc[2]) >> 3;
block[0][1][0] = (dc[0] - dc[3] + 3 + dc[1] - dc[2]) >> 3;
block[0][2][0] = (dc[0] + dc[3] + 3 - dc[1] - dc[2]) >> 3;
block[0][3][0] = (dc[0] - dc[3] + 3 - dc[1] + dc[2]) >> 3;
block[1][0][0] = (dc[4] + dc[7] + 3 + dc[5] + dc[6]) >> 3;
block[1][1][0] = (dc[4] - dc[7] + 3 + dc[5] - dc[6]) >> 3;
block[1][2][0] = (dc[4] + dc[7] + 3 - dc[5] - dc[6]) >> 3;
block[1][3][0] = (dc[4] - dc[7] + 3 - dc[5] + dc[6]) >> 3;
block[2][0][0] = (dc[8] + dc[11] + 3 + dc[9] + dc[10]) >> 3;
block[2][1][0] = (dc[8] - dc[11] + 3 + dc[9] - dc[10]) >> 3;
block[2][2][0] = (dc[8] + dc[11] + 3 - dc[9] - dc[10]) >> 3;
block[2][3][0] = (dc[8] - dc[11] + 3 - dc[9] + dc[10]) >> 3;
block[3][0][0] = (dc[12] + dc[15] + 3 + dc[13] + dc[14]) >> 3;
block[3][1][0] = (dc[12] - dc[15] + 3 + dc[13] - dc[14]) >> 3;
block[3][2][0] = (dc[12] + dc[15] + 3 - dc[13] - dc[14]) >> 3;
block[3][3][0] = (dc[12] - dc[15] + 3 - dc[13] + dc[14]) >> 3;
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
MMI_SDC1(%[ftmp0], %[dc], 0x00)
MMI_SDC1(%[ftmp0], %[dc], 0x08)
MMI_SDC1(%[ftmp0], %[dc], 0x10)
MMI_SDC1(%[ftmp0], %[dc], 0x18)
: RESTRICT_ASM_ALL64
[ftmp0]"=&f"(ftmp[0])
: [dc]"r"((uint8_t *)dc)
: "memory"
);
#else
int t00, t01, t02, t03, t10, t11, t12, t13, t20, t21, t22, t23, t30, t31, t32, t33;
t00 = dc[0] + dc[12];
t10 = dc[1] + dc[13];
t20 = dc[2] + dc[14];
t30 = dc[3] + dc[15];
t03 = dc[0] - dc[12];
t13 = dc[1] - dc[13];
t23 = dc[2] - dc[14];
t33 = dc[3] - dc[15];
t01 = dc[4] + dc[ 8];
t11 = dc[5] + dc[ 9];
t21 = dc[6] + dc[10];
t31 = dc[7] + dc[11];
t02 = dc[4] - dc[ 8];
t12 = dc[5] - dc[ 9];
t22 = dc[6] - dc[10];
t32 = dc[7] - dc[11];
dc[ 0] = t00 + t01;
dc[ 1] = t10 + t11;
dc[ 2] = t20 + t21;
dc[ 3] = t30 + t31;
dc[ 4] = t03 + t02;
dc[ 5] = t13 + t12;
dc[ 6] = t23 + t22;
dc[ 7] = t33 + t32;
dc[ 8] = t00 - t01;
dc[ 9] = t10 - t11;
dc[10] = t20 - t21;
dc[11] = t30 - t31;
dc[12] = t03 - t02;
dc[13] = t13 - t12;
dc[14] = t23 - t22;
dc[15] = t33 - t32;
block[0][0][0] = (dc[0] + dc[3] + 3 + dc[1] + dc[2]) >> 3;
block[0][1][0] = (dc[0] - dc[3] + 3 + dc[1] - dc[2]) >> 3;
block[0][2][0] = (dc[0] + dc[3] + 3 - dc[1] - dc[2]) >> 3;
block[0][3][0] = (dc[0] - dc[3] + 3 - dc[1] + dc[2]) >> 3;
block[1][0][0] = (dc[4] + dc[7] + 3 + dc[5] + dc[6]) >> 3;
block[1][1][0] = (dc[4] - dc[7] + 3 + dc[5] - dc[6]) >> 3;
block[1][2][0] = (dc[4] + dc[7] + 3 - dc[5] - dc[6]) >> 3;
block[1][3][0] = (dc[4] - dc[7] + 3 - dc[5] + dc[6]) >> 3;
block[2][0][0] = (dc[8] + dc[11] + 3 + dc[9] + dc[10]) >> 3;
block[2][1][0] = (dc[8] - dc[11] + 3 + dc[9] - dc[10]) >> 3;
block[2][2][0] = (dc[8] + dc[11] + 3 - dc[9] - dc[10]) >> 3;
block[2][3][0] = (dc[8] - dc[11] + 3 - dc[9] + dc[10]) >> 3;
block[3][0][0] = (dc[12] + dc[15] + 3 + dc[13] + dc[14]) >> 3;
block[3][1][0] = (dc[12] - dc[15] + 3 + dc[13] - dc[14]) >> 3;
block[3][2][0] = (dc[12] + dc[15] + 3 - dc[13] - dc[14]) >> 3;
block[3][3][0] = (dc[12] - dc[15] + 3 - dc[13] + dc[14]) >> 3;
AV_ZERO64(dc + 0);
AV_ZERO64(dc + 4);
AV_ZERO64(dc + 8);
AV_ZERO64(dc + 12);
#endif
}
void ff_vp8_luma_dc_wht_dc_mmi(int16_t block[4][4][16], int16_t dc[16])
{
int val = (dc[0] + 3) >> 3;
dc[0] = 0;
block[0][0][0] = val;
block[0][1][0] = val;
block[0][2][0] = val;
block[0][3][0] = val;
block[1][0][0] = val;
block[1][1][0] = val;
block[1][2][0] = val;
block[1][3][0] = val;
block[2][0][0] = val;
block[2][1][0] = val;
block[2][2][0] = val;
block[2][3][0] = val;
block[3][0][0] = val;
block[3][1][0] = val;
block[3][2][0] = val;
block[3][3][0] = val;
}
void ff_vp8_idct_add_mmi(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
{
#if 1
double ftmp[12];
uint32_t tmp[1];
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union av_intfloat64 ff_ph_4e7b_u;
union av_intfloat64 ff_ph_22a3_u;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
DECLARE_VAR_LOW32;
DECLARE_VAR_ALL64;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
ff_ph_4e7b_u.i = 0x4e7b4e7b4e7b4e7bULL;
ff_ph_22a3_u.i = 0x22a322a322a322a3ULL;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
MMI_LDC1(%[ftmp1], %[block], 0x00)
MMI_LDC1(%[ftmp2], %[block], 0x08)
MMI_LDC1(%[ftmp3], %[block], 0x10)
MMI_LDC1(%[ftmp4], %[block], 0x18)
"li %[tmp0], 0x02 \n\t"
"mtc1 %[tmp0], %[ftmp11] \n\t"
// block[0...3] + block[8...11]
"paddh %[ftmp5], %[ftmp1], %[ftmp3] \n\t"
// block[0...3] - block[8...11]
"psubh %[ftmp6], %[ftmp1], %[ftmp3] \n\t"
// MUL_35468(block[12...15])
"psllh %[ftmp9], %[ftmp4], %[ftmp11] \n\t"
"pmulhh %[ftmp7], %[ftmp9], %[ff_ph_22a3] \n\t"
// MUL_35468(block[4...7])
"psllh %[ftmp9], %[ftmp2], %[ftmp11] \n\t"
"pmulhh %[ftmp8], %[ftmp9], %[ff_ph_22a3] \n\t"
// MUL_20091(block[4...7]
"pmulhh %[ftmp9], %[ftmp2], %[ff_ph_4e7b] \n\t"
"paddh %[ftmp9], %[ftmp9], %[ftmp2] \n\t"
// MUL_20091(block[12...15])
"pmulhh %[ftmp10], %[ftmp4], %[ff_ph_4e7b] \n\t"
"paddh %[ftmp10], %[ftmp10], %[ftmp4] \n\t"
// tmp[0 4 8 12]
"paddh %[ftmp1], %[ftmp5], %[ftmp7] \n\t"
"paddh %[ftmp1], %[ftmp1], %[ftmp9] \n\t"
// tmp[1 5 9 13]
"paddh %[ftmp2], %[ftmp6], %[ftmp8] \n\t"
"psubh %[ftmp2], %[ftmp2], %[ftmp10] \n\t"
// tmp[2 6 10 14]
"psubh %[ftmp3], %[ftmp6], %[ftmp8] \n\t"
"paddh %[ftmp3], %[ftmp3], %[ftmp10] \n\t"
// tmp[3 7 11 15]
"psubh %[ftmp4], %[ftmp5], %[ftmp7] \n\t"
"psubh %[ftmp4], %[ftmp4], %[ftmp9] \n\t"
MMI_SDC1(%[ftmp0], %[block], 0x00)
MMI_SDC1(%[ftmp0], %[block], 0x08)
MMI_SDC1(%[ftmp0], %[block], 0x10)
MMI_SDC1(%[ftmp0], %[block], 0x18)
TRANSPOSE_4H(%[ftmp1], %[ftmp2], %[ftmp3], %[ftmp4],
avutil/mips: [loongson] simplify macro TRANSPOSE_4H and TRANSPOSE_8B Simplify macro TRANSPOSE_4H in mmiutils.h and add TRANSPOSE_8B as a common macro. Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
%[ftmp5], %[ftmp6], %[ftmp7], %[ftmp8])
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
// t[0 4 8 12]
"paddh %[ftmp5], %[ftmp1], %[ftmp3] \n\t"
// t[1 5 9 13]
"psubh %[ftmp6], %[ftmp1], %[ftmp3] \n\t"
// t[2 6 10 14]
"psllh %[ftmp9], %[ftmp2], %[ftmp11] \n\t"
"pmulhh %[ftmp9], %[ftmp9], %[ff_ph_22a3] \n\t"
"psubh %[ftmp7], %[ftmp9], %[ftmp4] \n\t"
"pmulhh %[ftmp10], %[ftmp4], %[ff_ph_4e7b] \n\t"
"psubh %[ftmp7], %[ftmp7], %[ftmp10] \n\t"
// t[3 7 11 15]
"psllh %[ftmp9], %[ftmp4], %[ftmp11] \n\t"
"pmulhh %[ftmp9], %[ftmp9], %[ff_ph_22a3] \n\t"
"paddh %[ftmp8], %[ftmp9], %[ftmp2] \n\t"
"pmulhh %[ftmp10], %[ftmp2], %[ff_ph_4e7b] \n\t"
"paddh %[ftmp8], %[ftmp8], %[ftmp10] \n\t"
"li %[tmp0], 0x03 \n\t"
"mtc1 %[tmp0], %[ftmp11] \n\t"
"paddh %[ftmp1], %[ftmp5], %[ftmp8] \n\t"
"paddh %[ftmp1], %[ftmp1], %[ff_pw_4] \n\t"
"psrah %[ftmp1], %[ftmp1], %[ftmp11] \n\t"
"paddh %[ftmp2], %[ftmp6], %[ftmp7] \n\t"
"paddh %[ftmp2], %[ftmp2], %[ff_pw_4] \n\t"
"psrah %[ftmp2], %[ftmp2], %[ftmp11] \n\t"
"psubh %[ftmp3], %[ftmp6], %[ftmp7] \n\t"
"paddh %[ftmp3], %[ftmp3], %[ff_pw_4] \n\t"
"psrah %[ftmp3], %[ftmp3], %[ftmp11] \n\t"
"psubh %[ftmp4], %[ftmp5], %[ftmp8] \n\t"
"paddh %[ftmp4], %[ftmp4], %[ff_pw_4] \n\t"
"psrah %[ftmp4], %[ftmp4], %[ftmp11] \n\t"
TRANSPOSE_4H(%[ftmp1], %[ftmp2], %[ftmp3], %[ftmp4],
avutil/mips: [loongson] simplify macro TRANSPOSE_4H and TRANSPOSE_8B Simplify macro TRANSPOSE_4H in mmiutils.h and add TRANSPOSE_8B as a common macro. Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
%[ftmp5], %[ftmp6], %[ftmp7], %[ftmp8])
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
MMI_LWC1(%[ftmp5], %[dst0], 0x00)
MMI_LWC1(%[ftmp6], %[dst1], 0x00)
MMI_LWC1(%[ftmp7], %[dst2], 0x00)
MMI_LWC1(%[ftmp8], %[dst3], 0x00)
"punpcklbh %[ftmp5], %[ftmp5], %[ftmp0] \n\t"
"punpcklbh %[ftmp6], %[ftmp6], %[ftmp0] \n\t"
"punpcklbh %[ftmp7], %[ftmp7], %[ftmp0] \n\t"
"punpcklbh %[ftmp8], %[ftmp8], %[ftmp0] \n\t"
"paddh %[ftmp1], %[ftmp1], %[ftmp5] \n\t"
"paddh %[ftmp2], %[ftmp2], %[ftmp6] \n\t"
"paddh %[ftmp3], %[ftmp3], %[ftmp7] \n\t"
"paddh %[ftmp4], %[ftmp4], %[ftmp8] \n\t"
"packushb %[ftmp1], %[ftmp1], %[ftmp0] \n\t"
"packushb %[ftmp2], %[ftmp2], %[ftmp0] \n\t"
"packushb %[ftmp3], %[ftmp3], %[ftmp0] \n\t"
"packushb %[ftmp4], %[ftmp4], %[ftmp0] \n\t"
MMI_SWC1(%[ftmp1], %[dst0], 0x00)
MMI_SWC1(%[ftmp2], %[dst1], 0x00)
MMI_SWC1(%[ftmp3], %[dst2], 0x00)
MMI_SWC1(%[ftmp4], %[dst3], 0x00)
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
[ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
[ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]),
[ftmp10]"=&f"(ftmp[10]), [ftmp11]"=&f"(ftmp[11]),
RESTRICT_ASM_LOW32
RESTRICT_ASM_ALL64
[tmp0]"=&r"(tmp[0])
: [dst0]"r"(dst), [dst1]"r"(dst+stride),
[dst2]"r"(dst+2*stride), [dst3]"r"(dst+3*stride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[block]"r"(block), [ff_pw_4]"f"(ff_pw_4.f),
[ff_ph_4e7b]"f"(ff_ph_4e7b_u.f), [ff_ph_22a3]"f"(ff_ph_22a3_u.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
int i, t0, t1, t2, t3;
int16_t tmp[16];
for (i = 0; i < 4; i++) {
t0 = block[0 + i] + block[8 + i];
t1 = block[0 + i] - block[8 + i];
t2 = MUL_35468(block[4 + i]) - MUL_20091(block[12 + i]);
t3 = MUL_20091(block[4 + i]) + MUL_35468(block[12 + i]);
block[ 0 + i] = 0;
block[ 4 + i] = 0;
block[ 8 + i] = 0;
block[12 + i] = 0;
tmp[i * 4 + 0] = t0 + t3;
tmp[i * 4 + 1] = t1 + t2;
tmp[i * 4 + 2] = t1 - t2;
tmp[i * 4 + 3] = t0 - t3;
}
for (i = 0; i < 4; i++) {
t0 = tmp[0 + i] + tmp[8 + i];
t1 = tmp[0 + i] - tmp[8 + i];
t2 = MUL_35468(tmp[4 + i]) - MUL_20091(tmp[12 + i]);
t3 = MUL_20091(tmp[4 + i]) + MUL_35468(tmp[12 + i]);
dst[0] = av_clip_uint8(dst[0] + ((t0 + t3 + 4) >> 3));
dst[1] = av_clip_uint8(dst[1] + ((t1 + t2 + 4) >> 3));
dst[2] = av_clip_uint8(dst[2] + ((t1 - t2 + 4) >> 3));
dst[3] = av_clip_uint8(dst[3] + ((t0 - t3 + 4) >> 3));
dst += stride;
}
#endif
}
void ff_vp8_idct_dc_add_mmi(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
{
#if 1
int dc = (block[0] + 4) >> 3;
double ftmp[6];
DECLARE_VAR_LOW32;
block[0] = 0;
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"mtc1 %[dc], %[ftmp5] \n\t"
MMI_LWC1(%[ftmp1], %[dst0], 0x00)
MMI_LWC1(%[ftmp2], %[dst1], 0x00)
MMI_LWC1(%[ftmp3], %[dst2], 0x00)
MMI_LWC1(%[ftmp4], %[dst3], 0x00)
"pshufh %[ftmp5], %[ftmp5], %[ftmp0] \n\t"
"punpcklbh %[ftmp1], %[ftmp1], %[ftmp0] \n\t"
"punpcklbh %[ftmp2], %[ftmp2], %[ftmp0] \n\t"
"punpcklbh %[ftmp3], %[ftmp3], %[ftmp0] \n\t"
"punpcklbh %[ftmp4], %[ftmp4], %[ftmp0] \n\t"
"paddsh %[ftmp1], %[ftmp1], %[ftmp5] \n\t"
"paddsh %[ftmp2], %[ftmp2], %[ftmp5] \n\t"
"paddsh %[ftmp3], %[ftmp3], %[ftmp5] \n\t"
"paddsh %[ftmp4], %[ftmp4], %[ftmp5] \n\t"
"packushb %[ftmp1], %[ftmp1], %[ftmp0] \n\t"
"packushb %[ftmp2], %[ftmp2], %[ftmp0] \n\t"
"packushb %[ftmp3], %[ftmp3], %[ftmp0] \n\t"
"packushb %[ftmp4], %[ftmp4], %[ftmp0] \n\t"
MMI_SWC1(%[ftmp1], %[dst0], 0x00)
MMI_SWC1(%[ftmp2], %[dst1], 0x00)
MMI_SWC1(%[ftmp3], %[dst2], 0x00)
MMI_SWC1(%[ftmp4], %[dst3], 0x00)
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]),
RESTRICT_ASM_LOW32
[ftmp5]"=&f"(ftmp[5])
: [dst0]"r"(dst), [dst1]"r"(dst+stride),
[dst2]"r"(dst+2*stride), [dst3]"r"(dst+3*stride),
[dc]"r"(dc)
: "memory"
);
#else
int i, dc = (block[0] + 4) >> 3;
block[0] = 0;
for (i = 0; i < 4; i++) {
dst[0] = av_clip_uint8(dst[0] + dc);
dst[1] = av_clip_uint8(dst[1] + dc);
dst[2] = av_clip_uint8(dst[2] + dc);
dst[3] = av_clip_uint8(dst[3] + dc);
dst += stride;
}
#endif
}
void ff_vp8_idct_dc_add4y_mmi(uint8_t *dst, int16_t block[4][16],
ptrdiff_t stride)
{
ff_vp8_idct_dc_add_mmi(dst + 0, block[0], stride);
ff_vp8_idct_dc_add_mmi(dst + 4, block[1], stride);
ff_vp8_idct_dc_add_mmi(dst + 8, block[2], stride);
ff_vp8_idct_dc_add_mmi(dst + 12, block[3], stride);
}
void ff_vp8_idct_dc_add4uv_mmi(uint8_t *dst, int16_t block[4][16],
ptrdiff_t stride)
{
ff_vp8_idct_dc_add_mmi(dst + stride * 0 + 0, block[0], stride);
ff_vp8_idct_dc_add_mmi(dst + stride * 0 + 4, block[1], stride);
ff_vp8_idct_dc_add_mmi(dst + stride * 4 + 0, block[2], stride);
ff_vp8_idct_dc_add_mmi(dst + stride * 4 + 4, block[3], stride);
}
// loop filter applied to edges between macroblocks
void ff_vp8_v_loop_filter16_mmi(uint8_t *dst, ptrdiff_t stride, int flim_E,
int flim_I, int hev_thresh)
{
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
vp8_v_loop_filter8_mmi(dst, stride, flim_E, flim_I, hev_thresh);
vp8_v_loop_filter8_mmi(dst + 8, stride, flim_E, flim_I, hev_thresh);
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
}
void ff_vp8_h_loop_filter16_mmi(uint8_t *dst, ptrdiff_t stride, int flim_E,
int flim_I, int hev_thresh)
{
avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. Optimize vp8 loop filter with mmi, four functions optimized: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Vp8 decoding speed improved about 50%(from 73fps to 110fps, Tested on loongson 3A3000). Signed-off-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
6 years ago
vp8_h_loop_filter8_mmi(dst, stride, flim_E, flim_I, hev_thresh);
vp8_h_loop_filter8_mmi(dst + 8 * stride, stride, flim_E, flim_I,
hev_thresh);
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
}
void ff_vp8_v_loop_filter8uv_mmi(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,
int flim_E, int flim_I, int hev_thresh)
{
vp8_v_loop_filter8_mmi(dstU, stride, flim_E, flim_I, hev_thresh);
vp8_v_loop_filter8_mmi(dstV, stride, flim_E, flim_I, hev_thresh);
}
void ff_vp8_h_loop_filter8uv_mmi(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,
int flim_E, int flim_I, int hev_thresh)
{
vp8_h_loop_filter8_mmi(dstU, stride, flim_E, flim_I, hev_thresh);
vp8_h_loop_filter8_mmi(dstV, stride, flim_E, flim_I, hev_thresh);
}
// loop filter applied to inner macroblock edges
void ff_vp8_v_loop_filter16_inner_mmi(uint8_t *dst, ptrdiff_t stride,
int flim_E, int flim_I, int hev_thresh)
{
int i;
for (i = 0; i < 16; i++)
if (vp8_normal_limit(dst + i * 1, stride, flim_E, flim_I)) {
int hv = hev(dst + i * 1, stride, hev_thresh);
if (hv)
vp8_filter_common_is4tap(dst + i * 1, stride);
else
vp8_filter_common_isnot4tap(dst + i * 1, stride);
}
}
void ff_vp8_h_loop_filter16_inner_mmi(uint8_t *dst, ptrdiff_t stride,
int flim_E, int flim_I, int hev_thresh)
{
int i;
for (i = 0; i < 16; i++)
if (vp8_normal_limit(dst + i * stride, 1, flim_E, flim_I)) {
int hv = hev(dst + i * stride, 1, hev_thresh);
if (hv)
vp8_filter_common_is4tap(dst + i * stride, 1);
else
vp8_filter_common_isnot4tap(dst + i * stride, 1);
}
}
void ff_vp8_v_loop_filter8uv_inner_mmi(uint8_t *dstU, uint8_t *dstV,
ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh)
{
vp8_v_loop_filter8_inner_mmi(dstU, stride, flim_E, flim_I, hev_thresh);
vp8_v_loop_filter8_inner_mmi(dstV, stride, flim_E, flim_I, hev_thresh);
}
void ff_vp8_h_loop_filter8uv_inner_mmi(uint8_t *dstU, uint8_t *dstV,
ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh)
{
vp8_h_loop_filter8_inner_mmi(dstU, stride, flim_E, flim_I, hev_thresh);
vp8_h_loop_filter8_inner_mmi(dstV, stride, flim_E, flim_I, hev_thresh);
}
void ff_vp8_v_loop_filter_simple_mmi(uint8_t *dst, ptrdiff_t stride, int flim)
{
int i;
for (i = 0; i < 16; i++)
if (vp8_simple_limit(dst + i, stride, flim))
vp8_filter_common_is4tap(dst + i, stride);
}
void ff_vp8_h_loop_filter_simple_mmi(uint8_t *dst, ptrdiff_t stride, int flim)
{
int i;
for (i = 0; i < 16; i++)
if (vp8_simple_limit(dst + i * stride, 1, flim))
vp8_filter_common_is4tap(dst + i * stride, 1);
}
void ff_put_vp8_pixels16_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int x, int y)
{
#if 1
double ftmp[2];
uint64_t tmp[2];
mips_reg addr[2];
DECLARE_VAR_ALL64;
__asm__ volatile (
"1: \n\t"
PTR_ADDU "%[addr0], %[src], %[srcstride] \n\t"
MMI_ULDC1(%[ftmp0], %[src], 0x00)
"ldl %[tmp0], 0x0f(%[src]) \n\t"
"ldr %[tmp0], 0x08(%[src]) \n\t"
MMI_ULDC1(%[ftmp1], %[addr0], 0x00)
"ldl %[tmp1], 0x0f(%[addr0]) \n\t"
"ldr %[tmp1], 0x08(%[addr0]) \n\t"
PTR_ADDU "%[addr1], %[dst], %[dststride] \n\t"
MMI_SDC1(%[ftmp0], %[dst], 0x00)
"sdl %[tmp0], 0x0f(%[dst]) \n\t"
"sdr %[tmp0], 0x08(%[dst]) \n\t"
"addiu %[h], %[h], -0x02 \n\t"
MMI_SDC1(%[ftmp1], %[addr1], 0x00)
PTR_ADDU "%[src], %[addr0], %[srcstride] \n\t"
"sdl %[tmp1], 0x0f(%[addr1]) \n\t"
"sdr %[tmp1], 0x08(%[addr1]) \n\t"
PTR_ADDU "%[dst], %[addr1], %[dststride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[tmp0]"=&r"(tmp[0]), [tmp1]"=&r"(tmp[1]),
RESTRICT_ASM_ALL64
[addr0]"=&r"(addr[0]), [addr1]"=&r"(addr[1]),
[dst]"+&r"(dst), [src]"+&r"(src),
[h]"+&r"(h)
: [dststride]"r"((mips_reg)dststride),
[srcstride]"r"((mips_reg)srcstride)
: "memory"
);
#else
int i;
for (i = 0; i < h; i++, dst += dststride, src += srcstride)
memcpy(dst, src, 16);
#endif
}
void ff_put_vp8_pixels8_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int x, int y)
{
#if 1
double ftmp[1];
uint64_t tmp[1];
mips_reg addr[2];
DECLARE_VAR_ALL64;
__asm__ volatile (
"1: \n\t"
PTR_ADDU "%[addr0], %[src], %[srcstride] \n\t"
MMI_ULDC1(%[ftmp0], %[src], 0x00)
"ldl %[tmp0], 0x07(%[addr0]) \n\t"
"ldr %[tmp0], 0x00(%[addr0]) \n\t"
PTR_ADDU "%[addr1], %[dst], %[dststride] \n\t"
MMI_SDC1(%[ftmp0], %[dst], 0x00)
"addiu %[h], %[h], -0x02 \n\t"
"sdl %[tmp0], 0x07(%[addr1]) \n\t"
"sdr %[tmp0], 0x00(%[addr1]) \n\t"
PTR_ADDU "%[src], %[addr0], %[srcstride] \n\t"
PTR_ADDU "%[dst], %[addr1], %[dststride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_ALL64
[addr0]"=&r"(addr[0]), [addr1]"=&r"(addr[1]),
[dst]"+&r"(dst), [src]"+&r"(src),
[h]"+&r"(h)
: [dststride]"r"((mips_reg)dststride),
[srcstride]"r"((mips_reg)srcstride)
: "memory"
);
#else
int i;
for (i = 0; i < h; i++, dst += dststride, src += srcstride)
memcpy(dst, src, 8);
#endif
}
void ff_put_vp8_pixels4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int x, int y)
{
#if 1
double ftmp[1];
uint64_t tmp[1];
mips_reg addr[2];
DECLARE_VAR_LOW32;
__asm__ volatile (
"1: \n\t"
PTR_ADDU "%[addr0], %[src], %[srcstride] \n\t"
MMI_LWC1(%[ftmp0], %[src], 0x00)
"lwl %[tmp0], 0x03(%[addr0]) \n\t"
"lwr %[tmp0], 0x00(%[addr0]) \n\t"
PTR_ADDU "%[addr1], %[dst], %[dststride] \n\t"
MMI_SWC1(%[ftmp0], %[dst], 0x00)
"addiu %[h], %[h], -0x02 \n\t"
"swl %[tmp0], 0x03(%[addr1]) \n\t"
"swr %[tmp0], 0x00(%[addr1]) \n\t"
PTR_ADDU "%[src], %[addr0], %[srcstride] \n\t"
PTR_ADDU "%[dst], %[addr1], %[dststride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_LOW32
[addr0]"=&r"(addr[0]), [addr1]"=&r"(addr[1]),
[dst]"+&r"(dst), [src]"+&r"(src),
[h]"+&r"(h)
: [dststride]"r"((mips_reg)dststride),
[srcstride]"r"((mips_reg)srcstride)
: "memory"
);
#else
int i;
for (i = 0; i < h; i++, dst += dststride, src += srcstride)
memcpy(dst, src, 4);
#endif
}
void ff_put_vp8_epel16_h4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
const uint64_t *filter = fourtap_subpel_filters[mx - 1];
double ftmp[9];
uint32_t tmp[1];
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union av_intfloat64 filter1;
union av_intfloat64 filter2;
union av_intfloat64 filter3;
union av_intfloat64 filter4;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
mips_reg src1, dst1;
DECLARE_VAR_ALL64;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
filter1.i = filter[1];
filter2.i = filter[2];
filter3.i = filter[3];
filter4.i = filter[4];
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
dst[0] = cm[(filter[2] * src[0] - filter[1] * src[-1] + filter[3] * src[1] - filter[4] * src[2] + 64) >> 7];
dst[1] = cm[(filter[2] * src[1] - filter[1] * src[ 0] + filter[3] * src[2] - filter[4] * src[3] + 64) >> 7];
dst[2] = cm[(filter[2] * src[2] - filter[1] * src[ 1] + filter[3] * src[3] - filter[4] * src[4] + 64) >> 7];
dst[3] = cm[(filter[2] * src[3] - filter[1] * src[ 2] + filter[3] * src[4] - filter[4] * src[5] + 64) >> 7];
dst[4] = cm[(filter[2] * src[4] - filter[1] * src[ 3] + filter[3] * src[5] - filter[4] * src[6] + 64) >> 7];
dst[5] = cm[(filter[2] * src[5] - filter[1] * src[ 4] + filter[3] * src[6] - filter[4] * src[7] + 64) >> 7];
dst[6] = cm[(filter[2] * src[6] - filter[1] * src[ 5] + filter[3] * src[7] - filter[4] * src[8] + 64) >> 7];
dst[7] = cm[(filter[2] * src[7] - filter[1] * src[ 6] + filter[3] * src[8] - filter[4] * src[9] + 64) >> 7];
dst[ 8] = cm[(filter[2] * src[ 8] - filter[1] * src[ 7] + filter[3] * src[ 9] - filter[4] * src[10] + 64) >> 7];
dst[ 9] = cm[(filter[2] * src[ 9] - filter[1] * src[ 8] + filter[3] * src[10] - filter[4] * src[11] + 64) >> 7];
dst[10] = cm[(filter[2] * src[10] - filter[1] * src[ 9] + filter[3] * src[11] - filter[4] * src[12] + 64) >> 7];
dst[11] = cm[(filter[2] * src[11] - filter[1] * src[10] + filter[3] * src[12] - filter[4] * src[13] + 64) >> 7];
dst[12] = cm[(filter[2] * src[12] - filter[1] * src[11] + filter[3] * src[13] - filter[4] * src[14] + 64) >> 7];
dst[13] = cm[(filter[2] * src[13] - filter[1] * src[12] + filter[3] * src[14] - filter[4] * src[15] + 64) >> 7];
dst[14] = cm[(filter[2] * src[14] - filter[1] * src[13] + filter[3] * src[15] - filter[4] * src[16] + 64) >> 7];
dst[15] = cm[(filter[2] * src[15] - filter[1] * src[14] + filter[3] * src[16] - filter[4] * src[17] + 64) >> 7];
*/
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"li %[tmp0], 0x07 \n\t"
"mtc1 %[tmp0], %[ftmp4] \n\t"
"1: \n\t"
// 0 - 7
PUT_VP8_EPEL8_H4_MMI(%[src], %[dst])
PTR_ADDIU "%[src1], %[src], 0x08 \n\t"
PTR_ADDIU "%[dst1], %[dst], 0x08 \n\t"
// 8 - 15
PUT_VP8_EPEL8_H4_MMI(%[src1], %[dst1])
"addiu %[h], %[h], -0x01 \n\t"
PTR_ADDU "%[src], %[src], %[srcstride] \n\t"
PTR_ADDU "%[dst], %[dst], %[dststride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
[ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
[ftmp8]"=&f"(ftmp[8]),
[tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_ALL64
[dst1]"=&r"(dst1), [src1]"=&r"(src1),
[h]"+&r"(h),
[dst]"+&r"(dst), [src]"+&r"(src)
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
: [ff_pw_64]"f"(ff_pw_64.f),
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
[srcstride]"r"((mips_reg)srcstride),
[dststride]"r"((mips_reg)dststride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[filter1]"f"(filter1.f), [filter2]"f"(filter2.f),
[filter3]"f"(filter3.f), [filter4]"f"(filter4.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
const uint8_t *filter = subpel_filters[mx - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < 16; x++)
dst[x] = FILTER_4TAP(src, filter, 1);
dst += dststride;
src += srcstride;
}
#endif
}
void ff_put_vp8_epel8_h4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
const uint64_t *filter = fourtap_subpel_filters[mx - 1];
double ftmp[9];
uint32_t tmp[1];
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union av_intfloat64 filter1;
union av_intfloat64 filter2;
union av_intfloat64 filter3;
union av_intfloat64 filter4;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
DECLARE_VAR_ALL64;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
filter1.i = filter[1];
filter2.i = filter[2];
filter3.i = filter[3];
filter4.i = filter[4];
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
dst[0] = cm[(filter[2] * src[0] - filter[1] * src[-1] + filter[3] * src[1] - filter[4] * src[2] + 64) >> 7];
dst[1] = cm[(filter[2] * src[1] - filter[1] * src[ 0] + filter[3] * src[2] - filter[4] * src[3] + 64) >> 7];
dst[2] = cm[(filter[2] * src[2] - filter[1] * src[ 1] + filter[3] * src[3] - filter[4] * src[4] + 64) >> 7];
dst[3] = cm[(filter[2] * src[3] - filter[1] * src[ 2] + filter[3] * src[4] - filter[4] * src[5] + 64) >> 7];
dst[4] = cm[(filter[2] * src[4] - filter[1] * src[ 3] + filter[3] * src[5] - filter[4] * src[6] + 64) >> 7];
dst[5] = cm[(filter[2] * src[5] - filter[1] * src[ 4] + filter[3] * src[6] - filter[4] * src[7] + 64) >> 7];
dst[6] = cm[(filter[2] * src[6] - filter[1] * src[ 5] + filter[3] * src[7] - filter[4] * src[8] + 64) >> 7];
dst[7] = cm[(filter[2] * src[7] - filter[1] * src[ 6] + filter[3] * src[8] - filter[4] * src[9] + 64) >> 7];
*/
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"li %[tmp0], 0x07 \n\t"
"mtc1 %[tmp0], %[ftmp4] \n\t"
"1: \n\t"
PUT_VP8_EPEL8_H4_MMI(%[src], %[dst])
"addiu %[h], %[h], -0x01 \n\t"
PTR_ADDU "%[src], %[src], %[srcstride] \n\t"
PTR_ADDU "%[dst], %[dst], %[dststride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
[ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
[ftmp8]"=&f"(ftmp[8]),
[tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_ALL64
[h]"+&r"(h),
[dst]"+&r"(dst), [src]"+&r"(src)
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
: [ff_pw_64]"f"(ff_pw_64.f),
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
[srcstride]"r"((mips_reg)srcstride),
[dststride]"r"((mips_reg)dststride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[filter1]"f"(filter1.f), [filter2]"f"(filter2.f),
[filter3]"f"(filter3.f), [filter4]"f"(filter4.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
const uint8_t *filter = subpel_filters[mx - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < 8; x++)
dst[x] = FILTER_4TAP(src, filter, 1);
dst += dststride;
src += srcstride;
}
#endif
}
void ff_put_vp8_epel4_h4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
const uint64_t *filter = fourtap_subpel_filters[mx - 1];
double ftmp[6];
uint32_t tmp[1];
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union av_intfloat64 filter1;
union av_intfloat64 filter2;
union av_intfloat64 filter3;
union av_intfloat64 filter4;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
DECLARE_VAR_LOW32;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
filter1.i = filter[1];
filter2.i = filter[2];
filter3.i = filter[3];
filter4.i = filter[4];
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
dst[0] = cm[(filter[2] * src[0] - filter[1] * src[-1] + filter[3] * src[1] - filter[4] * src[2] + 64) >> 7];
dst[1] = cm[(filter[2] * src[1] - filter[1] * src[ 0] + filter[3] * src[2] - filter[4] * src[3] + 64) >> 7];
dst[2] = cm[(filter[2] * src[2] - filter[1] * src[ 1] + filter[3] * src[3] - filter[4] * src[4] + 64) >> 7];
dst[3] = cm[(filter[2] * src[3] - filter[1] * src[ 2] + filter[3] * src[4] - filter[4] * src[5] + 64) >> 7];
*/
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"li %[tmp0], 0x07 \n\t"
"mtc1 %[tmp0], %[ftmp4] \n\t"
"1: \n\t"
PUT_VP8_EPEL4_H4_MMI(%[src], %[dst])
"addiu %[h], %[h], -0x01 \n\t"
PTR_ADDU "%[src], %[src], %[srcstride] \n\t"
PTR_ADDU "%[dst], %[dst], %[dststride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
[tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_LOW32
[h]"+&r"(h),
[dst]"+&r"(dst), [src]"+&r"(src)
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
: [ff_pw_64]"f"(ff_pw_64.f),
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
[srcstride]"r"((mips_reg)srcstride),
[dststride]"r"((mips_reg)dststride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[filter1]"f"(filter1.f), [filter2]"f"(filter2.f),
[filter3]"f"(filter3.f), [filter4]"f"(filter4.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
const uint8_t *filter = subpel_filters[mx - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < 4; x++)
dst[x] = FILTER_4TAP(src, filter, 1);
dst += dststride;
src += srcstride;
}
#endif
}
void ff_put_vp8_epel16_h6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
const uint64_t *filter = fourtap_subpel_filters[mx - 1];
double ftmp[9];
uint32_t tmp[1];
mips_reg src1, dst1;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union av_intfloat64 filter0;
union av_intfloat64 filter1;
union av_intfloat64 filter2;
union av_intfloat64 filter3;
union av_intfloat64 filter4;
union av_intfloat64 filter5;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
DECLARE_VAR_ALL64;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
filter0.i = filter[0];
filter1.i = filter[1];
filter2.i = filter[2];
filter3.i = filter[3];
filter4.i = filter[4];
filter5.i = filter[5];
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
dst[ 0] = cm[(filter[2]*src[ 0] - filter[1]*src[-1] + filter[0]*src[-2] + filter[3]*src[ 1] - filter[4]*src[ 2] + filter[5]*src[ 3] + 64) >> 7];
dst[ 1] = cm[(filter[2]*src[ 1] - filter[1]*src[ 0] + filter[0]*src[-1] + filter[3]*src[ 2] - filter[4]*src[ 3] + filter[5]*src[ 4] + 64) >> 7];
dst[ 2] = cm[(filter[2]*src[ 2] - filter[1]*src[ 1] + filter[0]*src[ 0] + filter[3]*src[ 3] - filter[4]*src[ 4] + filter[5]*src[ 5] + 64) >> 7];
dst[ 3] = cm[(filter[2]*src[ 3] - filter[1]*src[ 2] + filter[0]*src[ 1] + filter[3]*src[ 4] - filter[4]*src[ 5] + filter[5]*src[ 6] + 64) >> 7];
dst[ 4] = cm[(filter[2]*src[ 4] - filter[1]*src[ 3] + filter[0]*src[ 2] + filter[3]*src[ 5] - filter[4]*src[ 6] + filter[5]*src[ 7] + 64) >> 7];
dst[ 5] = cm[(filter[2]*src[ 5] - filter[1]*src[ 4] + filter[0]*src[ 3] + filter[3]*src[ 6] - filter[4]*src[ 7] + filter[5]*src[ 8] + 64) >> 7];
dst[ 6] = cm[(filter[2]*src[ 6] - filter[1]*src[ 5] + filter[0]*src[ 4] + filter[3]*src[ 7] - filter[4]*src[ 8] + filter[5]*src[ 9] + 64) >> 7];
dst[ 7] = cm[(filter[2]*src[ 7] - filter[1]*src[ 6] + filter[0]*src[ 5] + filter[3]*src[ 8] - filter[4]*src[ 9] + filter[5]*src[10] + 64) >> 7];
dst[ 8] = cm[(filter[2]*src[ 8] - filter[1]*src[ 7] + filter[0]*src[ 6] + filter[3]*src[ 9] - filter[4]*src[10] + filter[5]*src[11] + 64) >> 7];
dst[ 9] = cm[(filter[2]*src[ 9] - filter[1]*src[ 8] + filter[0]*src[ 7] + filter[3]*src[10] - filter[4]*src[11] + filter[5]*src[12] + 64) >> 7];
dst[10] = cm[(filter[2]*src[10] - filter[1]*src[ 9] + filter[0]*src[ 8] + filter[3]*src[11] - filter[4]*src[12] + filter[5]*src[13] + 64) >> 7];
dst[11] = cm[(filter[2]*src[11] - filter[1]*src[10] + filter[0]*src[ 9] + filter[3]*src[12] - filter[4]*src[13] + filter[5]*src[14] + 64) >> 7];
dst[12] = cm[(filter[2]*src[12] - filter[1]*src[11] + filter[0]*src[10] + filter[3]*src[13] - filter[4]*src[14] + filter[5]*src[15] + 64) >> 7];
dst[13] = cm[(filter[2]*src[13] - filter[1]*src[12] + filter[0]*src[11] + filter[3]*src[14] - filter[4]*src[15] + filter[5]*src[16] + 64) >> 7];
dst[14] = cm[(filter[2]*src[14] - filter[1]*src[13] + filter[0]*src[12] + filter[3]*src[15] - filter[4]*src[16] + filter[5]*src[17] + 64) >> 7];
dst[15] = cm[(filter[2]*src[15] - filter[1]*src[14] + filter[0]*src[13] + filter[3]*src[16] - filter[4]*src[17] + filter[5]*src[18] + 64) >> 7];
*/
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"li %[tmp0], 0x07 \n\t"
"mtc1 %[tmp0], %[ftmp4] \n\t"
"1: \n\t"
// 0 - 7
PUT_VP8_EPEL8_H6_MMI(%[src], %[dst])
PTR_ADDIU "%[src1], %[src], 0x08 \n\t"
PTR_ADDIU "%[dst1], %[dst], 0x08 \n\t"
// 8 - 15
PUT_VP8_EPEL8_H6_MMI(%[src1], %[dst1])
"addiu %[h], %[h], -0x01 \n\t"
PTR_ADDU "%[src], %[src], %[srcstride] \n\t"
PTR_ADDU "%[dst], %[dst], %[dststride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
[ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
[ftmp8]"=&f"(ftmp[8]),
[tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_ALL64
[dst1]"=&r"(dst1), [src1]"=&r"(src1),
[h]"+&r"(h),
[dst]"+&r"(dst), [src]"+&r"(src)
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
: [ff_pw_64]"f"(ff_pw_64.f),
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
[srcstride]"r"((mips_reg)srcstride),
[dststride]"r"((mips_reg)dststride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[filter0]"f"(filter0.f), [filter1]"f"(filter1.f),
[filter2]"f"(filter2.f), [filter3]"f"(filter3.f),
[filter4]"f"(filter4.f), [filter5]"f"(filter5.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
const uint8_t *filter = subpel_filters[mx - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < 16; x++)
dst[x] = FILTER_6TAP(src, filter, 1);
dst += dststride;
src += srcstride;
}
#endif
}
void ff_put_vp8_epel8_h6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
const uint64_t *filter = fourtap_subpel_filters[mx - 1];
double ftmp[9];
uint32_t tmp[1];
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union av_intfloat64 filter0;
union av_intfloat64 filter1;
union av_intfloat64 filter2;
union av_intfloat64 filter3;
union av_intfloat64 filter4;
union av_intfloat64 filter5;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
DECLARE_VAR_ALL64;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
filter0.i = filter[0];
filter1.i = filter[1];
filter2.i = filter[2];
filter3.i = filter[3];
filter4.i = filter[4];
filter5.i = filter[5];
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
dst[0] = cm[(filter[2]*src[0] - filter[1]*src[-1] + filter[0]*src[-2] + filter[3]*src[1] - filter[4]*src[2] + filter[5]*src[ 3] + 64) >> 7];
dst[1] = cm[(filter[2]*src[1] - filter[1]*src[ 0] + filter[0]*src[-1] + filter[3]*src[2] - filter[4]*src[3] + filter[5]*src[ 4] + 64) >> 7];
dst[2] = cm[(filter[2]*src[2] - filter[1]*src[ 1] + filter[0]*src[ 0] + filter[3]*src[3] - filter[4]*src[4] + filter[5]*src[ 5] + 64) >> 7];
dst[3] = cm[(filter[2]*src[3] - filter[1]*src[ 2] + filter[0]*src[ 1] + filter[3]*src[4] - filter[4]*src[5] + filter[5]*src[ 6] + 64) >> 7];
dst[4] = cm[(filter[2]*src[4] - filter[1]*src[ 3] + filter[0]*src[ 2] + filter[3]*src[5] - filter[4]*src[6] + filter[5]*src[ 7] + 64) >> 7];
dst[5] = cm[(filter[2]*src[5] - filter[1]*src[ 4] + filter[0]*src[ 3] + filter[3]*src[6] - filter[4]*src[7] + filter[5]*src[ 8] + 64) >> 7];
dst[6] = cm[(filter[2]*src[6] - filter[1]*src[ 5] + filter[0]*src[ 4] + filter[3]*src[7] - filter[4]*src[8] + filter[5]*src[ 9] + 64) >> 7];
dst[7] = cm[(filter[2]*src[7] - filter[1]*src[ 6] + filter[0]*src[ 5] + filter[3]*src[8] - filter[4]*src[9] + filter[5]*src[10] + 64) >> 7];
*/
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"li %[tmp0], 0x07 \n\t"
"mtc1 %[tmp0], %[ftmp4] \n\t"
"1: \n\t"
PUT_VP8_EPEL8_H6_MMI(%[src], %[dst])
"addiu %[h], %[h], -0x01 \n\t"
PTR_ADDU "%[src], %[src], %[srcstride] \n\t"
PTR_ADDU "%[dst], %[dst], %[dststride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
[ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
[ftmp8]"=&f"(ftmp[8]),
[tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_ALL64
[h]"+&r"(h),
[dst]"+&r"(dst), [src]"+&r"(src)
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
: [ff_pw_64]"f"(ff_pw_64.f),
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
[srcstride]"r"((mips_reg)srcstride),
[dststride]"r"((mips_reg)dststride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[filter0]"f"(filter0.f), [filter1]"f"(filter1.f),
[filter2]"f"(filter2.f), [filter3]"f"(filter3.f),
[filter4]"f"(filter4.f), [filter5]"f"(filter5.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
const uint8_t *filter = subpel_filters[mx - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < 8; x++)
dst[x] = FILTER_6TAP(src, filter, 1);
dst += dststride;
src += srcstride;
}
#endif
}
void ff_put_vp8_epel4_h6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
const uint64_t *filter = fourtap_subpel_filters[mx - 1];
double ftmp[6];
uint32_t tmp[1];
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union av_intfloat64 filter0;
union av_intfloat64 filter1;
union av_intfloat64 filter2;
union av_intfloat64 filter3;
union av_intfloat64 filter4;
union av_intfloat64 filter5;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
DECLARE_VAR_LOW32;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
filter0.i = filter[0];
filter1.i = filter[1];
filter2.i = filter[2];
filter3.i = filter[3];
filter4.i = filter[4];
filter5.i = filter[5];
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
dst[0] = cm[(filter[2]*src[0] - filter[1]*src[-1] + filter[0]*src[-2] + filter[3]*src[1] - filter[4]*src[2] + filter[5]*src[ 3] + 64) >> 7];
dst[1] = cm[(filter[2]*src[1] - filter[1]*src[ 0] + filter[0]*src[-1] + filter[3]*src[2] - filter[4]*src[3] + filter[5]*src[ 4] + 64) >> 7];
dst[2] = cm[(filter[2]*src[2] - filter[1]*src[ 1] + filter[0]*src[ 0] + filter[3]*src[3] - filter[4]*src[4] + filter[5]*src[ 5] + 64) >> 7];
dst[3] = cm[(filter[2]*src[3] - filter[1]*src[ 2] + filter[0]*src[ 1] + filter[3]*src[4] - filter[4]*src[5] + filter[5]*src[ 6] + 64) >> 7];
*/
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"li %[tmp0], 0x07 \n\t"
"mtc1 %[tmp0], %[ftmp4] \n\t"
"1: \n\t"
PUT_VP8_EPEL4_H6_MMI(%[src], %[dst])
"addiu %[h], %[h], -0x01 \n\t"
PTR_ADDU "%[src], %[src], %[srcstride] \n\t"
PTR_ADDU "%[dst], %[dst], %[dststride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
[tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_LOW32
[h]"+&r"(h),
[dst]"+&r"(dst), [src]"+&r"(src)
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
: [ff_pw_64]"f"(ff_pw_64.f),
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
[srcstride]"r"((mips_reg)srcstride),
[dststride]"r"((mips_reg)dststride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[filter0]"f"(filter0.f), [filter1]"f"(filter1.f),
[filter2]"f"(filter2.f), [filter3]"f"(filter3.f),
[filter4]"f"(filter4.f), [filter5]"f"(filter5.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
const uint8_t *filter = subpel_filters[mx - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < 4; x++)
dst[x] = FILTER_6TAP(src, filter, 1);
dst += dststride;
src += srcstride;
}
#endif
}
void ff_put_vp8_epel16_v4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
const uint64_t *filter = fourtap_subpel_filters[my - 1];
double ftmp[9];
uint32_t tmp[1];
mips_reg src0, src1, dst0;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union av_intfloat64 filter1;
union av_intfloat64 filter2;
union av_intfloat64 filter3;
union av_intfloat64 filter4;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
DECLARE_VAR_ALL64;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
filter1.i = filter[1];
filter2.i = filter[2];
filter3.i = filter[3];
filter4.i = filter[4];
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
dst[0] = cm[(filter[2] * src[0] - filter[1] * src[ -srcstride] + filter[3] * src[ srcstride] - filter[4] * src[ 2*srcstride] + 64) >> 7];
dst[1] = cm[(filter[2] * src[1] - filter[1] * src[1-srcstride] + filter[3] * src[1+srcstride] - filter[4] * src[1+2*srcstride] + 64) >> 7];
dst[2] = cm[(filter[2] * src[2] - filter[1] * src[2-srcstride] + filter[3] * src[2+srcstride] - filter[4] * src[2+2*srcstride] + 64) >> 7];
dst[3] = cm[(filter[2] * src[3] - filter[1] * src[3-srcstride] + filter[3] * src[3+srcstride] - filter[4] * src[3+2*srcstride] + 64) >> 7];
dst[4] = cm[(filter[2] * src[4] - filter[1] * src[4-srcstride] + filter[3] * src[4+srcstride] - filter[4] * src[4+2*srcstride] + 64) >> 7];
dst[5] = cm[(filter[2] * src[5] - filter[1] * src[5-srcstride] + filter[3] * src[5+srcstride] - filter[4] * src[5+2*srcstride] + 64) >> 7];
dst[6] = cm[(filter[2] * src[6] - filter[1] * src[6-srcstride] + filter[3] * src[6+srcstride] - filter[4] * src[6+2*srcstride] + 64) >> 7];
dst[7] = cm[(filter[2] * src[7] - filter[1] * src[7-srcstride] + filter[3] * src[7+srcstride] - filter[4] * src[7+2*srcstride] + 64) >> 7];
dst[ 8] = cm[(filter[2] * src[ 8] - filter[1] * src[ 8-srcstride] + filter[3] * src[ 8+srcstride] - filter[4] * src[ 8+2*srcstride] + 64) >> 7];
dst[ 9] = cm[(filter[2] * src[ 9] - filter[1] * src[ 9-srcstride] + filter[3] * src[ 9+srcstride] - filter[4] * src[ 9+2*srcstride] + 64) >> 7];
dst[10] = cm[(filter[2] * src[10] - filter[1] * src[10-srcstride] + filter[3] * src[10+srcstride] - filter[4] * src[10+2*srcstride] + 64) >> 7];
dst[11] = cm[(filter[2] * src[11] - filter[1] * src[11-srcstride] + filter[3] * src[11+srcstride] - filter[4] * src[11+2*srcstride] + 64) >> 7];
dst[12] = cm[(filter[2] * src[12] - filter[1] * src[12-srcstride] + filter[3] * src[12+srcstride] - filter[4] * src[12+2*srcstride] + 64) >> 7];
dst[13] = cm[(filter[2] * src[13] - filter[1] * src[13-srcstride] + filter[3] * src[13+srcstride] - filter[4] * src[13+2*srcstride] + 64) >> 7];
dst[14] = cm[(filter[2] * src[14] - filter[1] * src[14-srcstride] + filter[3] * src[14+srcstride] - filter[4] * src[14+2*srcstride] + 64) >> 7];
dst[15] = cm[(filter[2] * src[15] - filter[1] * src[15-srcstride] + filter[3] * src[15+srcstride] - filter[4] * src[15+2*srcstride] + 64) >> 7];
*/
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"li %[tmp0], 0x07 \n\t"
"mtc1 %[tmp0], %[ftmp4] \n\t"
"1: \n\t"
// 0 - 7
PUT_VP8_EPEL8_V4_MMI(%[src], %[src1], %[dst], %[srcstride])
PTR_ADDIU "%[src0], %[src], 0x08 \n\t"
PTR_ADDIU "%[dst0], %[dst], 0x08 \n\t"
// 8 - 15
PUT_VP8_EPEL8_V4_MMI(%[src0], %[src1], %[dst], %[srcstride])
"addiu %[h], %[h], -0x01 \n\t"
PTR_ADDU "%[src], %[src], %[srcstride] \n\t"
PTR_ADDU "%[dst], %[dst], %[dststride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
[ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
[ftmp8]"=&f"(ftmp[8]),
[tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_ALL64
[src0]"=&r"(src0), [dst0]"=&r"(dst0),
[src1]"=&r"(src1),
[h]"+&r"(h),
[dst]"+&r"(dst), [src]"+&r"(src)
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
: [ff_pw_64]"f"(ff_pw_64.f),
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
[srcstride]"r"((mips_reg)srcstride),
[dststride]"r"((mips_reg)dststride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[filter1]"f"(filter1.f), [filter2]"f"(filter2.f),
[filter3]"f"(filter3.f), [filter4]"f"(filter4.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
const uint8_t *filter = subpel_filters[my - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < 16; x++)
dst[x] = FILTER_4TAP(src, filter, srcstride);
dst += dststride;
src += srcstride;
}
#endif
}
void ff_put_vp8_epel8_v4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
const uint64_t *filter = fourtap_subpel_filters[my - 1];
double ftmp[9];
uint32_t tmp[1];
mips_reg src1;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union av_intfloat64 filter1;
union av_intfloat64 filter2;
union av_intfloat64 filter3;
union av_intfloat64 filter4;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
DECLARE_VAR_ALL64;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
filter1.i = filter[1];
filter2.i = filter[2];
filter3.i = filter[3];
filter4.i = filter[4];
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
dst[0] = cm[(filter[2] * src[0] - filter[1] * src[ -srcstride] + filter[3] * src[ srcstride] - filter[4] * src[ 2*srcstride] + 64) >> 7];
dst[1] = cm[(filter[2] * src[1] - filter[1] * src[1-srcstride] + filter[3] * src[1+srcstride] - filter[4] * src[1+2*srcstride] + 64) >> 7];
dst[2] = cm[(filter[2] * src[2] - filter[1] * src[2-srcstride] + filter[3] * src[2+srcstride] - filter[4] * src[2+2*srcstride] + 64) >> 7];
dst[3] = cm[(filter[2] * src[3] - filter[1] * src[3-srcstride] + filter[3] * src[3+srcstride] - filter[4] * src[3+2*srcstride] + 64) >> 7];
dst[4] = cm[(filter[2] * src[4] - filter[1] * src[4-srcstride] + filter[3] * src[4+srcstride] - filter[4] * src[4+2*srcstride] + 64) >> 7];
dst[5] = cm[(filter[2] * src[5] - filter[1] * src[5-srcstride] + filter[3] * src[5+srcstride] - filter[4] * src[5+2*srcstride] + 64) >> 7];
dst[6] = cm[(filter[2] * src[6] - filter[1] * src[6-srcstride] + filter[3] * src[6+srcstride] - filter[4] * src[6+2*srcstride] + 64) >> 7];
dst[7] = cm[(filter[2] * src[7] - filter[1] * src[7-srcstride] + filter[3] * src[7+srcstride] - filter[4] * src[7+2*srcstride] + 64) >> 7];
*/
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"li %[tmp0], 0x07 \n\t"
"mtc1 %[tmp0], %[ftmp4] \n\t"
"1: \n\t"
PUT_VP8_EPEL8_V4_MMI(%[src], %[src1], %[dst], %[srcstride])
"addiu %[h], %[h], -0x01 \n\t"
PTR_ADDU "%[src], %[src], %[srcstride] \n\t"
PTR_ADDU "%[dst], %[dst], %[dststride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
[ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
[ftmp8]"=&f"(ftmp[8]),
[tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_ALL64
[src1]"=&r"(src1),
[h]"+&r"(h),
[dst]"+&r"(dst), [src]"+&r"(src)
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
: [ff_pw_64]"f"(ff_pw_64.f),
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
[srcstride]"r"((mips_reg)srcstride),
[dststride]"r"((mips_reg)dststride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[filter1]"f"(filter1.f), [filter2]"f"(filter2.f),
[filter3]"f"(filter3.f), [filter4]"f"(filter4.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
const uint8_t *filter = subpel_filters[my - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < 8; x++)
dst[x] = FILTER_4TAP(src, filter, srcstride);
dst += dststride;
src += srcstride;
}
#endif
}
void ff_put_vp8_epel4_v4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
const uint64_t *filter = fourtap_subpel_filters[my - 1];
double ftmp[6];
uint32_t tmp[1];
mips_reg src1;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union av_intfloat64 filter1;
union av_intfloat64 filter2;
union av_intfloat64 filter3;
union av_intfloat64 filter4;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
DECLARE_VAR_LOW32;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
filter1.i = filter[1];
filter2.i = filter[2];
filter3.i = filter[3];
filter4.i = filter[4];
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
dst[0] = cm[(filter[2] * src[0] - filter[1] * src[ -srcstride] + filter[3] * src[ srcstride] - filter[4] * src[ 2*srcstride] + 64) >> 7];
dst[1] = cm[(filter[2] * src[1] - filter[1] * src[1-srcstride] + filter[3] * src[1+srcstride] - filter[4] * src[1+2*srcstride] + 64) >> 7];
dst[2] = cm[(filter[2] * src[2] - filter[1] * src[2-srcstride] + filter[3] * src[2+srcstride] - filter[4] * src[2+2*srcstride] + 64) >> 7];
dst[3] = cm[(filter[2] * src[3] - filter[1] * src[3-srcstride] + filter[3] * src[3+srcstride] - filter[4] * src[3+2*srcstride] + 64) >> 7];
*/
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"li %[tmp0], 0x07 \n\t"
"mtc1 %[tmp0], %[ftmp4] \n\t"
"1: \n\t"
PUT_VP8_EPEL4_V4_MMI(%[src], %[src1], %[dst], %[srcstride])
"addiu %[h], %[h], -0x01 \n\t"
PTR_ADDU "%[src], %[src], %[srcstride] \n\t"
PTR_ADDU "%[dst], %[dst], %[dststride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
[tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_LOW32
[src1]"=&r"(src1),
[h]"+&r"(h),
[dst]"+&r"(dst), [src]"+&r"(src)
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
: [ff_pw_64]"f"(ff_pw_64.f),
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
[srcstride]"r"((mips_reg)srcstride),
[dststride]"r"((mips_reg)dststride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[filter1]"f"(filter1.f), [filter2]"f"(filter2.f),
[filter3]"f"(filter3.f), [filter4]"f"(filter4.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
const uint8_t *filter = subpel_filters[my - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < 4; x++)
dst[x] = FILTER_4TAP(src, filter, srcstride);
dst += dststride;
src += srcstride;
}
#endif
}
void ff_put_vp8_epel16_v6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
const uint64_t *filter = fourtap_subpel_filters[my - 1];
double ftmp[9];
uint32_t tmp[1];
mips_reg src0, src1, dst0;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union av_intfloat64 filter0;
union av_intfloat64 filter1;
union av_intfloat64 filter2;
union av_intfloat64 filter3;
union av_intfloat64 filter4;
union av_intfloat64 filter5;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
DECLARE_VAR_ALL64;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
filter0.i = filter[0];
filter1.i = filter[1];
filter2.i = filter[2];
filter3.i = filter[3];
filter4.i = filter[4];
filter5.i = filter[5];
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
dst[0] = cm[(filter[2]*src[0] - filter[1]*src[0-srcstride] + filter[0]*src[0-2*srcstride] + filter[3]*src[0+srcstride] - filter[4]*src[0+2*srcstride] + filter[5]*src[0+3*srcstride] + 64) >> 7];
dst[1] = cm[(filter[2]*src[1] - filter[1]*src[1-srcstride] + filter[0]*src[1-2*srcstride] + filter[3]*src[1+srcstride] - filter[4]*src[1+2*srcstride] + filter[5]*src[1+3*srcstride] + 64) >> 7];
dst[2] = cm[(filter[2]*src[2] - filter[1]*src[2-srcstride] + filter[0]*src[2-2*srcstride] + filter[3]*src[2+srcstride] - filter[4]*src[2+2*srcstride] + filter[5]*src[2+3*srcstride] + 64) >> 7];
dst[3] = cm[(filter[2]*src[3] - filter[1]*src[3-srcstride] + filter[0]*src[3-2*srcstride] + filter[3]*src[3+srcstride] - filter[4]*src[3+2*srcstride] + filter[5]*src[3+3*srcstride] + 64) >> 7];
dst[4] = cm[(filter[2]*src[4] - filter[1]*src[4-srcstride] + filter[0]*src[4-2*srcstride] + filter[3]*src[4+srcstride] - filter[4]*src[4+2*srcstride] + filter[5]*src[4+3*srcstride] + 64) >> 7];
dst[5] = cm[(filter[2]*src[5] - filter[1]*src[5-srcstride] + filter[0]*src[5-2*srcstride] + filter[3]*src[5+srcstride] - filter[4]*src[5+2*srcstride] + filter[5]*src[5+3*srcstride] + 64) >> 7];
dst[6] = cm[(filter[2]*src[6] - filter[1]*src[6-srcstride] + filter[0]*src[6-2*srcstride] + filter[3]*src[6+srcstride] - filter[4]*src[6+2*srcstride] + filter[5]*src[6+3*srcstride] + 64) >> 7];
dst[7] = cm[(filter[2]*src[7] - filter[1]*src[7-srcstride] + filter[0]*src[7-2*srcstride] + filter[3]*src[7+srcstride] - filter[4]*src[7+2*srcstride] + filter[5]*src[7+3*srcstride] + 64) >> 7];
dst[ 8] = cm[(filter[2]*src[ 8] - filter[1]*src[ 8-srcstride] + filter[0]*src[ 8-2*srcstride] + filter[3]*src[ 8+srcstride] - filter[4]*src[ 8+2*srcstride] + filter[5]*src[ 8+3*srcstride] + 64) >> 7];
dst[ 9] = cm[(filter[2]*src[ 9] - filter[1]*src[ 9-srcstride] + filter[0]*src[ 9-2*srcstride] + filter[3]*src[ 9+srcstride] - filter[4]*src[ 9+2*srcstride] + filter[5]*src[ 9+3*srcstride] + 64) >> 7];
dst[10] = cm[(filter[2]*src[10] - filter[1]*src[10-srcstride] + filter[0]*src[10-2*srcstride] + filter[3]*src[10+srcstride] - filter[4]*src[10+2*srcstride] + filter[5]*src[10+3*srcstride] + 64) >> 7];
dst[11] = cm[(filter[2]*src[11] - filter[1]*src[11-srcstride] + filter[0]*src[11-2*srcstride] + filter[3]*src[11+srcstride] - filter[4]*src[11+2*srcstride] + filter[5]*src[11+3*srcstride] + 64) >> 7];
dst[12] = cm[(filter[2]*src[12] - filter[1]*src[12-srcstride] + filter[0]*src[12-2*srcstride] + filter[3]*src[12+srcstride] - filter[4]*src[12+2*srcstride] + filter[5]*src[12+3*srcstride] + 64) >> 7];
dst[13] = cm[(filter[2]*src[13] - filter[1]*src[13-srcstride] + filter[0]*src[13-2*srcstride] + filter[3]*src[13+srcstride] - filter[4]*src[13+2*srcstride] + filter[5]*src[13+3*srcstride] + 64) >> 7];
dst[14] = cm[(filter[2]*src[14] - filter[1]*src[14-srcstride] + filter[0]*src[14-2*srcstride] + filter[3]*src[14+srcstride] - filter[4]*src[14+2*srcstride] + filter[5]*src[14+3*srcstride] + 64) >> 7];
dst[15] = cm[(filter[2]*src[15] - filter[1]*src[15-srcstride] + filter[0]*src[15-2*srcstride] + filter[3]*src[15+srcstride] - filter[4]*src[15+2*srcstride] + filter[5]*src[15+3*srcstride] + 64) >> 7];
*/
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"li %[tmp0], 0x07 \n\t"
"mtc1 %[tmp0], %[ftmp4] \n\t"
"1: \n\t"
// 0 - 7
PUT_VP8_EPEL8_V6_MMI(%[src], %[src1], %[dst], %[srcstride])
PTR_ADDIU "%[src0], %[src], 0x08 \n\t"
PTR_ADDIU "%[dst0], %[dst], 0x08 \n\t"
// 8 - 15
PUT_VP8_EPEL8_V6_MMI(%[src0], %[src1], %[dst0], %[srcstride])
"addiu %[h], %[h], -0x01 \n\t"
PTR_ADDU "%[src], %[src], %[srcstride] \n\t"
PTR_ADDU "%[dst], %[dst], %[dststride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
[ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
[ftmp8]"=&f"(ftmp[8]),
[tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_ALL64
[src0]"=&r"(src0), [dst0]"=&r"(dst0),
[src1]"=&r"(src1),
[h]"+&r"(h),
[dst]"+&r"(dst), [src]"+&r"(src)
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
: [ff_pw_64]"f"(ff_pw_64.f),
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
[srcstride]"r"((mips_reg)srcstride),
[dststride]"r"((mips_reg)dststride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[filter0]"f"(filter0.f), [filter1]"f"(filter1.f),
[filter2]"f"(filter2.f), [filter3]"f"(filter3.f),
[filter4]"f"(filter4.f), [filter5]"f"(filter5.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
const uint8_t *filter = subpel_filters[my - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < 16; x++)
dst[x] = FILTER_6TAP(src, filter, srcstride);
dst += dststride;
src += srcstride;
}
#endif
}
void ff_put_vp8_epel8_v6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
const uint64_t *filter = fourtap_subpel_filters[my - 1];
double ftmp[9];
uint32_t tmp[1];
mips_reg src1;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union av_intfloat64 filter0;
union av_intfloat64 filter1;
union av_intfloat64 filter2;
union av_intfloat64 filter3;
union av_intfloat64 filter4;
union av_intfloat64 filter5;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
DECLARE_VAR_ALL64;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
filter0.i = filter[0];
filter1.i = filter[1];
filter2.i = filter[2];
filter3.i = filter[3];
filter4.i = filter[4];
filter5.i = filter[5];
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
dst[0] = cm[(filter[2]*src[0] - filter[1]*src[0-srcstride] + filter[0]*src[0-2*srcstride] + filter[3]*src[0+srcstride] - filter[4]*src[0+2*srcstride] + filter[5]*src[0+3*srcstride] + 64) >> 7];
dst[1] = cm[(filter[2]*src[1] - filter[1]*src[1-srcstride] + filter[0]*src[1-2*srcstride] + filter[3]*src[1+srcstride] - filter[4]*src[1+2*srcstride] + filter[5]*src[1+3*srcstride] + 64) >> 7];
dst[2] = cm[(filter[2]*src[2] - filter[1]*src[2-srcstride] + filter[0]*src[2-2*srcstride] + filter[3]*src[2+srcstride] - filter[4]*src[2+2*srcstride] + filter[5]*src[2+3*srcstride] + 64) >> 7];
dst[3] = cm[(filter[2]*src[3] - filter[1]*src[3-srcstride] + filter[0]*src[3-2*srcstride] + filter[3]*src[3+srcstride] - filter[4]*src[3+2*srcstride] + filter[5]*src[3+3*srcstride] + 64) >> 7];
dst[4] = cm[(filter[2]*src[4] - filter[1]*src[4-srcstride] + filter[0]*src[4-2*srcstride] + filter[3]*src[4+srcstride] - filter[4]*src[4+2*srcstride] + filter[5]*src[4+3*srcstride] + 64) >> 7];
dst[5] = cm[(filter[2]*src[5] - filter[1]*src[5-srcstride] + filter[0]*src[5-2*srcstride] + filter[3]*src[5+srcstride] - filter[4]*src[5+2*srcstride] + filter[5]*src[5+3*srcstride] + 64) >> 7];
dst[6] = cm[(filter[2]*src[6] - filter[1]*src[6-srcstride] + filter[0]*src[6-2*srcstride] + filter[3]*src[6+srcstride] - filter[4]*src[6+2*srcstride] + filter[5]*src[6+3*srcstride] + 64) >> 7];
dst[7] = cm[(filter[2]*src[7] - filter[1]*src[7-srcstride] + filter[0]*src[7-2*srcstride] + filter[3]*src[7+srcstride] - filter[4]*src[7+2*srcstride] + filter[5]*src[7+3*srcstride] + 64) >> 7];
*/
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"li %[tmp0], 0x07 \n\t"
"mtc1 %[tmp0], %[ftmp4] \n\t"
"1: \n\t"
PUT_VP8_EPEL8_V6_MMI(%[src], %[src1], %[dst], %[srcstride])
"addiu %[h], %[h], -0x01 \n\t"
PTR_ADDU "%[src], %[src], %[srcstride] \n\t"
PTR_ADDU "%[dst], %[dst], %[dststride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
[ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]),
[ftmp8]"=&f"(ftmp[8]),
[tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_ALL64
[src1]"=&r"(src1),
[h]"+&r"(h),
[dst]"+&r"(dst), [src]"+&r"(src)
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
: [ff_pw_64]"f"(ff_pw_64.f),
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
[srcstride]"r"((mips_reg)srcstride),
[dststride]"r"((mips_reg)dststride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[filter0]"f"(filter0.f), [filter1]"f"(filter1.f),
[filter2]"f"(filter2.f), [filter3]"f"(filter3.f),
[filter4]"f"(filter4.f), [filter5]"f"(filter5.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
const uint8_t *filter = subpel_filters[my - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < 8; x++)
dst[x] = FILTER_6TAP(src, filter, srcstride);
dst += dststride;
src += srcstride;
}
#endif
}
void ff_put_vp8_epel4_v6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
const uint64_t *filter = fourtap_subpel_filters[my - 1];
double ftmp[6];
uint32_t tmp[1];
mips_reg src1;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union av_intfloat64 filter0;
union av_intfloat64 filter1;
union av_intfloat64 filter2;
union av_intfloat64 filter3;
union av_intfloat64 filter4;
union av_intfloat64 filter5;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
DECLARE_VAR_LOW32;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
filter0.i = filter[0];
filter1.i = filter[1];
filter2.i = filter[2];
filter3.i = filter[3];
filter4.i = filter[4];
filter5.i = filter[5];
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
dst[0] = cm[(filter[2]*src[0] - filter[1]*src[0-srcstride] + filter[0]*src[0-2*srcstride] + filter[3]*src[0+srcstride] - filter[4]*src[0+2*srcstride] + filter[5]*src[0+3*srcstride] + 64) >> 7];
dst[1] = cm[(filter[2]*src[1] - filter[1]*src[1-srcstride] + filter[0]*src[1-2*srcstride] + filter[3]*src[1+srcstride] - filter[4]*src[1+2*srcstride] + filter[5]*src[1+3*srcstride] + 64) >> 7];
dst[2] = cm[(filter[2]*src[2] - filter[1]*src[2-srcstride] + filter[0]*src[2-2*srcstride] + filter[3]*src[2+srcstride] - filter[4]*src[2+2*srcstride] + filter[5]*src[2+3*srcstride] + 64) >> 7];
dst[3] = cm[(filter[2]*src[3] - filter[1]*src[3-srcstride] + filter[0]*src[3-2*srcstride] + filter[3]*src[3+srcstride] - filter[4]*src[3+2*srcstride] + filter[5]*src[3+3*srcstride] + 64) >> 7];
*/
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"li %[tmp0], 0x07 \n\t"
"mtc1 %[tmp0], %[ftmp4] \n\t"
"1: \n\t"
PUT_VP8_EPEL4_V6_MMI(%[src], %[src1], %[dst], %[srcstride])
"addiu %[h], %[h], -0x01 \n\t"
PTR_ADDU "%[src], %[src], %[srcstride] \n\t"
PTR_ADDU "%[dst], %[dst], %[dststride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
[tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_LOW32
[src1]"=&r"(src1),
[h]"+&r"(h),
[dst]"+&r"(dst), [src]"+&r"(src)
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
: [ff_pw_64]"f"(ff_pw_64.f),
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
[srcstride]"r"((mips_reg)srcstride),
[dststride]"r"((mips_reg)dststride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[filter0]"f"(filter0.f), [filter1]"f"(filter1.f),
[filter2]"f"(filter2.f), [filter3]"f"(filter3.f),
[filter4]"f"(filter4.f), [filter5]"f"(filter5.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
const uint8_t *filter = subpel_filters[my - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < 4; x++)
dst[x] = FILTER_6TAP(src, filter, srcstride);
dst += dststride;
src += srcstride;
}
#endif
}
void ff_put_vp8_epel16_h4v4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
DECLARE_ALIGNED(8, uint8_t, tmp_array[560]);
uint8_t *tmp = tmp_array;
src -= srcstride;
ff_put_vp8_epel16_h4_mmi(tmp, 16, src, srcstride, h + 3, mx, my);
tmp = tmp_array + 16;
ff_put_vp8_epel16_v4_mmi(dst, dststride, tmp, 16, h, mx, my);
#else
const uint8_t *filter = subpel_filters[mx - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
uint8_t tmp_array[560];
uint8_t *tmp = tmp_array;
src -= srcstride;
for (y = 0; y < h + 3; y++) {
for (x = 0; x < 16; x++)
tmp[x] = FILTER_4TAP(src, filter, 1);
tmp += 16;
src += srcstride;
}
tmp = tmp_array + 16;
filter = subpel_filters[my - 1];
for (y = 0; y < h; y++) {
for (x = 0; x < 16; x++)
dst[x] = FILTER_4TAP(tmp, filter, 16);
dst += dststride;
tmp += 16;
}
#endif
}
void ff_put_vp8_epel8_h4v4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
DECLARE_ALIGNED(8, uint8_t, tmp_array[152]);
uint8_t *tmp = tmp_array;
src -= srcstride;
ff_put_vp8_epel8_h4_mmi(tmp, 8, src, srcstride, h + 3, mx, my);
tmp = tmp_array + 8;
ff_put_vp8_epel8_v4_mmi(dst, dststride, tmp, 8, h, mx, my);
#else
const uint8_t *filter = subpel_filters[mx - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
uint8_t tmp_array[152];
uint8_t *tmp = tmp_array;
src -= srcstride;
for (y = 0; y < h + 3; y++) {
for (x = 0; x < 8; x++)
tmp[x] = FILTER_4TAP(src, filter, 1);
tmp += 8;
src += srcstride;
}
tmp = tmp_array + 8;
filter = subpel_filters[my - 1];
for (y = 0; y < h; y++) {
for (x = 0; x < 8; x++)
dst[x] = FILTER_4TAP(tmp, filter, 8);
dst += dststride;
tmp += 8;
}
#endif
}
void ff_put_vp8_epel4_h4v4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
DECLARE_ALIGNED(4, uint8_t, tmp_array[44]);
uint8_t *tmp = tmp_array;
src -= srcstride;
ff_put_vp8_epel4_h4_mmi(tmp, 4, src, srcstride, h + 3, mx, my);
tmp = tmp_array + 4;
ff_put_vp8_epel4_v4_mmi(dst, dststride, tmp, 4, h, mx, my);
#else
const uint8_t *filter = subpel_filters[mx - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
uint8_t tmp_array[44];
uint8_t *tmp = tmp_array;
src -= srcstride;
for (y = 0; y < h + 3; y++) {
for (x = 0; x < 4; x++)
tmp[x] = FILTER_4TAP(src, filter, 1);
tmp += 4;
src += srcstride;
}
tmp = tmp_array + 4;
filter = subpel_filters[my - 1];
for (y = 0; y < h; y++) {
for (x = 0; x < 4; x++)
dst[x] = FILTER_4TAP(tmp, filter, 4);
dst += dststride;
tmp += 4;
}
#endif
}
void ff_put_vp8_epel16_h4v6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
DECLARE_ALIGNED(8, uint8_t, tmp_array[592]);
uint8_t *tmp = tmp_array;
src -= 2 * srcstride;
ff_put_vp8_epel16_h4_mmi(tmp, 16, src, srcstride, h + 5, mx, my);
tmp = tmp_array + 32;
ff_put_vp8_epel16_v6_mmi(dst, dststride, tmp, 16, h, mx, my);
#else
const uint8_t *filter = subpel_filters[mx - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
uint8_t tmp_array[592];
uint8_t *tmp = tmp_array;
src -= 2 * srcstride;
for (y = 0; y < h + 5; y++) {
for (x = 0; x < 16; x++)
tmp[x] = FILTER_4TAP(src, filter, 1);
tmp += 16;
src += srcstride;
}
tmp = tmp_array + 32;
filter = subpel_filters[my - 1];
for (y = 0; y < h; y++) {
for (x = 0; x < 16; x++)
dst[x] = FILTER_6TAP(tmp, filter, 16);
dst += dststride;
tmp += 16;
}
#endif
}
void ff_put_vp8_epel8_h4v6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
DECLARE_ALIGNED(8, uint8_t, tmp_array[168]);
uint8_t *tmp = tmp_array;
src -= 2 * srcstride;
ff_put_vp8_epel8_h4_mmi(tmp, 8, src, srcstride, h + 5, mx, my);
tmp = tmp_array + 16;
ff_put_vp8_epel8_v6_mmi(dst, dststride, tmp, 8, h, mx, my);
#else
const uint8_t *filter = subpel_filters[mx - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
uint8_t tmp_array[168];
uint8_t *tmp = tmp_array;
src -= 2 * srcstride;
for (y = 0; y < h + 5; y++) {
for (x = 0; x < 8; x++)
tmp[x] = FILTER_4TAP(src, filter, 1);
tmp += 8;
src += srcstride;
}
tmp = tmp_array + 16;
filter = subpel_filters[my - 1];
for (y = 0; y < h; y++) {
for (x = 0; x < 8; x++)
dst[x] = FILTER_6TAP(tmp, filter, 8);
dst += dststride;
tmp += 8;
}
#endif
}
void ff_put_vp8_epel4_h4v6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
DECLARE_ALIGNED(4, uint8_t, tmp_array[52]);
uint8_t *tmp = tmp_array;
src -= 2 * srcstride;
ff_put_vp8_epel4_h4_mmi(tmp, 4, src, srcstride, h + 5, mx, my);
tmp = tmp_array + 8;
ff_put_vp8_epel4_v6_mmi(dst, dststride, tmp, 4, h, mx, my);
#else
const uint8_t *filter = subpel_filters[mx - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
uint8_t tmp_array[52];
uint8_t *tmp = tmp_array;
src -= 2 * srcstride;
for (y = 0; y < h + 5; y++) {
for (x = 0; x < 4; x++)
tmp[x] = FILTER_4TAP(src, filter, 1);
tmp += 4;
src += srcstride;
}
tmp = tmp_array + 8;
filter = subpel_filters[my - 1];
for (y = 0; y < h; y++) {
for (x = 0; x < 4; x++)
dst[x] = FILTER_6TAP(tmp, filter, 4);
dst += dststride;
tmp += 4;
}
#endif
}
void ff_put_vp8_epel16_h6v4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
DECLARE_ALIGNED(8, uint8_t, tmp_array[560]);
uint8_t *tmp = tmp_array;
src -= srcstride;
ff_put_vp8_epel16_h6_mmi(tmp, 16, src, srcstride, h + 3, mx, my);
tmp = tmp_array + 16;
ff_put_vp8_epel16_v4_mmi(dst, dststride, tmp, 16, h, mx, my);
#else
const uint8_t *filter = subpel_filters[mx - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
uint8_t tmp_array[560];
uint8_t *tmp = tmp_array;
src -= srcstride;
for (y = 0; y < h + 3; y++) {
for (x = 0; x < 16; x++)
tmp[x] = FILTER_6TAP(src, filter, 1);
tmp += 16;
src += srcstride;
}
tmp = tmp_array + 16;
filter = subpel_filters[my - 1];
for (y = 0; y < h; y++) {
for (x = 0; x < 16; x++)
dst[x] = FILTER_4TAP(tmp, filter, 16);
dst += dststride;
tmp += 16;
}
#endif
}
void ff_put_vp8_epel8_h6v4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
DECLARE_ALIGNED(8, uint8_t, tmp_array[152]);
uint8_t *tmp = tmp_array;
src -= srcstride;
ff_put_vp8_epel8_h6_mmi(tmp, 8, src, srcstride, h + 3, mx, my);
tmp = tmp_array + 8;
ff_put_vp8_epel8_v4_mmi(dst, dststride, tmp, 8, h, mx, my);
#else
const uint8_t *filter = subpel_filters[mx - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
uint8_t tmp_array[152];
uint8_t *tmp = tmp_array;
src -= srcstride;
for (y = 0; y < h + 3; y++) {
for (x = 0; x < 8; x++)
tmp[x] = FILTER_6TAP(src, filter, 1);
tmp += 8;
src += srcstride;
}
tmp = tmp_array + 8;
filter = subpel_filters[my - 1];
for (y = 0; y < h; y++) {
for (x = 0; x < 8; x++)
dst[x] = FILTER_4TAP(tmp, filter, 8);
dst += dststride;
tmp += 8;
}
#endif
}
void ff_put_vp8_epel4_h6v4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
DECLARE_ALIGNED(4, uint8_t, tmp_array[44]);
uint8_t *tmp = tmp_array;
src -= srcstride;
ff_put_vp8_epel4_h6_mmi(tmp, 4, src, srcstride, h + 3, mx, my);
tmp = tmp_array + 4;
ff_put_vp8_epel4_v4_mmi(dst, dststride, tmp, 4, h, mx, my);
#else
const uint8_t *filter = subpel_filters[mx - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
uint8_t tmp_array[44];
uint8_t *tmp = tmp_array;
src -= srcstride;
for (y = 0; y < h + 3; y++) {
for (x = 0; x < 4; x++)
tmp[x] = FILTER_6TAP(src, filter, 1);
tmp += 4;
src += srcstride;
}
tmp = tmp_array + 4;
filter = subpel_filters[my - 1];
for (y = 0; y < h; y++) {
for (x = 0; x < 4; x++)
dst[x] = FILTER_4TAP(tmp, filter, 4);
dst += dststride;
tmp += 4;
}
#endif
}
void ff_put_vp8_epel16_h6v6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
DECLARE_ALIGNED(8, uint8_t, tmp_array[592]);
uint8_t *tmp = tmp_array;
src -= 2 * srcstride;
ff_put_vp8_epel16_h6_mmi(tmp, 16, src, srcstride, h + 5, mx, my);
tmp = tmp_array + 32;
ff_put_vp8_epel16_v6_mmi(dst, dststride, tmp, 16, h, mx, my);
#else
const uint8_t *filter = subpel_filters[mx - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
uint8_t tmp_array[592];
uint8_t *tmp = tmp_array;
src -= 2 * srcstride;
for (y = 0; y < h + 5; y++) {
for (x = 0; x < 16; x++)
tmp[x] = FILTER_6TAP(src, filter, 1);
tmp += 16;
src += srcstride;
}
tmp = tmp_array + 32;
filter = subpel_filters[my - 1];
for (y = 0; y < h; y++) {
for (x = 0; x < 16; x++)
dst[x] = FILTER_6TAP(tmp, filter, 16);
dst += dststride;
tmp += 16;
}
#endif
}
void ff_put_vp8_epel8_h6v6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
DECLARE_ALIGNED(8, uint8_t, tmp_array[168]);
uint8_t *tmp = tmp_array;
src -= 2 * srcstride;
ff_put_vp8_epel8_h6_mmi(tmp, 8, src, srcstride, h + 5, mx, my);
tmp = tmp_array + 16;
ff_put_vp8_epel8_v6_mmi(dst, dststride, tmp, 8, h, mx, my);
#else
const uint8_t *filter = subpel_filters[mx - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
uint8_t tmp_array[168];
uint8_t *tmp = tmp_array;
src -= 2 * srcstride;
for (y = 0; y < h + 5; y++) {
for (x = 0; x < 8; x++)
tmp[x] = FILTER_6TAP(src, filter, 1);
tmp += 8;
src += srcstride;
}
tmp = tmp_array + 16;
filter = subpel_filters[my - 1];
for (y = 0; y < h; y++) {
for (x = 0; x < 8; x++)
dst[x] = FILTER_6TAP(tmp, filter, 8);
dst += dststride;
tmp += 8;
}
#endif
}
void ff_put_vp8_epel4_h6v6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src,
ptrdiff_t srcstride, int h, int mx, int my)
{
#if 1
DECLARE_ALIGNED(4, uint8_t, tmp_array[52]);
uint8_t *tmp = tmp_array;
src -= 2 * srcstride;
ff_put_vp8_epel4_h6_mmi(tmp, 4, src, srcstride, h + 5, mx, my);
tmp = tmp_array + 8;
ff_put_vp8_epel4_v6_mmi(dst, dststride, tmp, 4, h, mx, my);
#else
const uint8_t *filter = subpel_filters[mx - 1];
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
int x, y;
uint8_t tmp_array[52];
uint8_t *tmp = tmp_array;
src -= 2 * srcstride;
for (y = 0; y < h + 5; y++) {
for (x = 0; x < 4; x++)
tmp[x] = FILTER_6TAP(src, filter, 1);
tmp += 4;
src += srcstride;
}
tmp = tmp_array + 8;
filter = subpel_filters[my - 1];
for (y = 0; y < h; y++) {
for (x = 0; x < 4; x++)
dst[x] = FILTER_6TAP(tmp, filter, 4);
dst += dststride;
tmp += 4;
}
#endif
}
void ff_put_vp8_bilinear16_h_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src,
ptrdiff_t sstride, int h, int mx, int my)
{
#if 1
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union mmi_intfloat64 a, b;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
double ftmp[7];
uint32_t tmp[1];
mips_reg dst0, src0;
DECLARE_VAR_ALL64;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
a.i = 8 - mx;
b.i = mx;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
dst[0] = (a * src[0] + b * src[1] + 4) >> 3;
dst[1] = (a * src[1] + b * src[2] + 4) >> 3;
dst[2] = (a * src[2] + b * src[3] + 4) >> 3;
dst[3] = (a * src[3] + b * src[4] + 4) >> 3;
dst[4] = (a * src[4] + b * src[5] + 4) >> 3;
dst[5] = (a * src[5] + b * src[6] + 4) >> 3;
dst[6] = (a * src[6] + b * src[7] + 4) >> 3;
dst[7] = (a * src[7] + b * src[8] + 4) >> 3;
dst[ 8] = (a * src[ 8] + b * src[ 9] + 4) >> 3;
dst[ 9] = (a * src[ 9] + b * src[10] + 4) >> 3;
dst[10] = (a * src[10] + b * src[11] + 4) >> 3;
dst[11] = (a * src[11] + b * src[12] + 4) >> 3;
dst[12] = (a * src[12] + b * src[13] + 4) >> 3;
dst[13] = (a * src[13] + b * src[14] + 4) >> 3;
dst[14] = (a * src[14] + b * src[15] + 4) >> 3;
dst[15] = (a * src[15] + b * src[16] + 4) >> 3;
*/
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"li %[tmp0], 0x03 \n\t"
"mtc1 %[tmp0], %[ftmp4] \n\t"
"pshufh %[a], %[a], %[ftmp0] \n\t"
"pshufh %[b], %[b], %[ftmp0] \n\t"
"1: \n\t"
// 0 - 7
PUT_VP8_BILINEAR8_H_MMI(%[src], %[dst])
PTR_ADDIU "%[src0], %[src], 0x08 \n\t"
PTR_ADDIU "%[dst0], %[dst], 0x08 \n\t"
// 8 - 15
PUT_VP8_BILINEAR8_H_MMI(%[src0], %[dst0])
"addiu %[h], %[h], -0x01 \n\t"
PTR_ADDU "%[src], %[src], %[sstride] \n\t"
PTR_ADDU "%[dst], %[dst], %[dstride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
[ftmp6]"=&f"(ftmp[6]),
[tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_ALL64
[dst0]"=&r"(dst0), [src0]"=&r"(src0),
[h]"+&r"(h),
[dst]"+&r"(dst), [src]"+&r"(src),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[a]"+&f"(a.f), [b]"+&f"(b.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: [sstride]"r"((mips_reg)sstride),
[dstride]"r"((mips_reg)dstride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[ff_pw_4]"f"(ff_pw_4.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
int a = 8 - mx, b = mx;
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < 16; x++)
dst[x] = (a * src[x] + b * src[x + 1] + 4) >> 3;
dst += dstride;
src += sstride;
}
#endif
}
void ff_put_vp8_bilinear16_v_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src,
ptrdiff_t sstride, int h, int mx, int my)
{
#if 1
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union mmi_intfloat64 c, d;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
double ftmp[7];
uint32_t tmp[1];
mips_reg src0, src1, dst0;
DECLARE_VAR_ALL64;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
c.i = 8 - my;
d.i = my;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
dst[0] = (c * src[0] + d * src[ sstride] + 4) >> 3;
dst[1] = (c * src[1] + d * src[1 + sstride] + 4) >> 3;
dst[2] = (c * src[2] + d * src[2 + sstride] + 4) >> 3;
dst[3] = (c * src[3] + d * src[3 + sstride] + 4) >> 3;
dst[4] = (c * src[4] + d * src[4 + sstride] + 4) >> 3;
dst[5] = (c * src[5] + d * src[5 + sstride] + 4) >> 3;
dst[6] = (c * src[6] + d * src[6 + sstride] + 4) >> 3;
dst[7] = (c * src[7] + d * src[7 + sstride] + 4) >> 3;
*/
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"li %[tmp0], 0x03 \n\t"
"mtc1 %[tmp0], %[ftmp4] \n\t"
"pshufh %[c], %[c], %[ftmp0] \n\t"
"pshufh %[d], %[d], %[ftmp0] \n\t"
"1: \n\t"
// 0 - 7
PUT_VP8_BILINEAR8_V_MMI(%[src], %[src1], %[dst], %[sstride])
PTR_ADDIU "%[src0], %[src], 0x08 \n\t"
PTR_ADDIU "%[dst0], %[dst], 0x08 \n\t"
// 8 - 15
PUT_VP8_BILINEAR8_V_MMI(%[src0], %[src1], %[dst0], %[sstride])
"addiu %[h], %[h], -0x01 \n\t"
PTR_ADDU "%[src], %[src], %[sstride] \n\t"
PTR_ADDU "%[dst], %[dst], %[dstride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
[ftmp6]"=&f"(ftmp[6]),
[tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_ALL64
[src0]"=&r"(src0), [dst0]"=&r"(dst0),
[src1]"=&r"(src1),
[h]"+&r"(h),
[dst]"+&r"(dst), [src]"+&r"(src),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[c]"+&f"(c.f), [d]"+&f"(d.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: [sstride]"r"((mips_reg)sstride),
[dstride]"r"((mips_reg)dstride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[ff_pw_4]"f"(ff_pw_4.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
int c = 8 - my, d = my;
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < 16; x++)
dst[x] = (c * src[x] + d * src[x + sstride] + 4) >> 3;
dst += dstride;
src += sstride;
}
#endif
}
void ff_put_vp8_bilinear16_hv_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src,
ptrdiff_t sstride, int h, int mx, int my)
{
#if 1
DECLARE_ALIGNED(8, uint8_t, tmp_array[528]);
uint8_t *tmp = tmp_array;
ff_put_vp8_bilinear16_h_mmi(tmp, 16, src, sstride, h + 1, mx, my);
ff_put_vp8_bilinear16_v_mmi(dst, dstride, tmp, 16, h, mx, my);
#else
int a = 8 - mx, b = mx;
int c = 8 - my, d = my;
int x, y;
uint8_t tmp_array[528];
uint8_t *tmp = tmp_array;
for (y = 0; y < h + 1; y++) {
for (x = 0; x < 16; x++)
tmp[x] = (a * src[x] + b * src[x + 1] + 4) >> 3;
tmp += 16;
src += sstride;
}
tmp = tmp_array;
for (y = 0; y < h; y++) {
for (x = 0; x < 16; x++)
dst[x] = (c * tmp[x] + d * tmp[x + 16] + 4) >> 3;
dst += dstride;
tmp += 16;
}
#endif
}
void ff_put_vp8_bilinear8_h_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src,
ptrdiff_t sstride, int h, int mx, int my)
{
#if 1
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union mmi_intfloat64 a, b;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
double ftmp[7];
uint32_t tmp[1];
DECLARE_VAR_ALL64;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
a.i = 8 - mx;
b.i = mx;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
dst[0] = (a * src[0] + b * src[1] + 4) >> 3;
dst[1] = (a * src[1] + b * src[2] + 4) >> 3;
dst[2] = (a * src[2] + b * src[3] + 4) >> 3;
dst[3] = (a * src[3] + b * src[4] + 4) >> 3;
dst[4] = (a * src[4] + b * src[5] + 4) >> 3;
dst[5] = (a * src[5] + b * src[6] + 4) >> 3;
dst[6] = (a * src[6] + b * src[7] + 4) >> 3;
dst[7] = (a * src[7] + b * src[8] + 4) >> 3;
*/
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"li %[tmp0], 0x03 \n\t"
"mtc1 %[tmp0], %[ftmp4] \n\t"
"pshufh %[a], %[a], %[ftmp0] \n\t"
"pshufh %[b], %[b], %[ftmp0] \n\t"
"1: \n\t"
PUT_VP8_BILINEAR8_H_MMI(%[src], %[dst])
"addiu %[h], %[h], -0x01 \n\t"
PTR_ADDU "%[src], %[src], %[sstride] \n\t"
PTR_ADDU "%[dst], %[dst], %[dstride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
[ftmp6]"=&f"(ftmp[6]),
[tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_ALL64
[h]"+&r"(h),
[dst]"+&r"(dst), [src]"+&r"(src),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[a]"+&f"(a.f), [b]"+&f"(b.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: [sstride]"r"((mips_reg)sstride),
[dstride]"r"((mips_reg)dstride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[ff_pw_4]"f"(ff_pw_4.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
int a = 8 - mx, b = mx;
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < 8; x++)
dst[x] = (a * src[x] + b * src[x + 1] + 4) >> 3;
dst += dstride;
src += sstride;
}
#endif
}
void ff_put_vp8_bilinear8_v_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src,
ptrdiff_t sstride, int h, int mx, int my)
{
#if 1
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union mmi_intfloat64 c, d;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
double ftmp[7];
uint32_t tmp[1];
mips_reg src1;
DECLARE_VAR_ALL64;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
c.i = 8 - my;
d.i = my;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
dst[0] = (c * src[0] + d * src[ sstride] + 4) >> 3;
dst[1] = (c * src[1] + d * src[1 + sstride] + 4) >> 3;
dst[2] = (c * src[2] + d * src[2 + sstride] + 4) >> 3;
dst[3] = (c * src[3] + d * src[3 + sstride] + 4) >> 3;
dst[4] = (c * src[4] + d * src[4 + sstride] + 4) >> 3;
dst[5] = (c * src[5] + d * src[5 + sstride] + 4) >> 3;
dst[6] = (c * src[6] + d * src[6 + sstride] + 4) >> 3;
dst[7] = (c * src[7] + d * src[7 + sstride] + 4) >> 3;
*/
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"li %[tmp0], 0x03 \n\t"
"mtc1 %[tmp0], %[ftmp4] \n\t"
"pshufh %[c], %[c], %[ftmp0] \n\t"
"pshufh %[d], %[d], %[ftmp0] \n\t"
"1: \n\t"
PUT_VP8_BILINEAR8_V_MMI(%[src], %[src1], %[dst], %[sstride])
"addiu %[h], %[h], -0x01 \n\t"
PTR_ADDU "%[src], %[src], %[sstride] \n\t"
PTR_ADDU "%[dst], %[dst], %[dstride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]),
[ftmp6]"=&f"(ftmp[6]),
[tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_ALL64
[src1]"=&r"(src1),
[h]"+&r"(h),
[dst]"+&r"(dst), [src]"+&r"(src),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[c]"+&f"(c.f), [d]"+&f"(d.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: [sstride]"r"((mips_reg)sstride),
[dstride]"r"((mips_reg)dstride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[ff_pw_4]"f"(ff_pw_4.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
int c = 8 - my, d = my;
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < 8; x++)
dst[x] = (c * src[x] + d * src[x + sstride] + 4) >> 3;
dst += dstride;
src += sstride;
}
#endif
}
void ff_put_vp8_bilinear8_hv_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src,
ptrdiff_t sstride, int h, int mx, int my)
{
#if 1
DECLARE_ALIGNED(8, uint8_t, tmp_array[136]);
uint8_t *tmp = tmp_array;
ff_put_vp8_bilinear8_h_mmi(tmp, 8, src, sstride, h + 1, mx, my);
ff_put_vp8_bilinear8_v_mmi(dst, dstride, tmp, 8, h, mx, my);
#else
int a = 8 - mx, b = mx;
int c = 8 - my, d = my;
int x, y;
uint8_t tmp_array[136];
uint8_t *tmp = tmp_array;
for (y = 0; y < h + 1; y++) {
for (x = 0; x < 8; x++)
tmp[x] = (a * src[x] + b * src[x + 1] + 4) >> 3;
tmp += 8;
src += sstride;
}
tmp = tmp_array;
for (y = 0; y < h; y++) {
for (x = 0; x < 8; x++)
dst[x] = (c * tmp[x] + d * tmp[x + 8] + 4) >> 3;
dst += dstride;
tmp += 8;
}
#endif
}
void ff_put_vp8_bilinear4_h_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src,
ptrdiff_t sstride, int h, int mx, int my)
{
#if 1
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union mmi_intfloat64 a, b;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
double ftmp[5];
uint32_t tmp[1];
DECLARE_VAR_LOW32;
DECLARE_VAR_ALL64;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
a.i = 8 - mx;
b.i = mx;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
dst[0] = (a * src[0] + b * src[1] + 4) >> 3;
dst[1] = (a * src[1] + b * src[2] + 4) >> 3;
dst[2] = (a * src[2] + b * src[3] + 4) >> 3;
dst[3] = (a * src[3] + b * src[4] + 4) >> 3;
*/
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"li %[tmp0], 0x03 \n\t"
"mtc1 %[tmp0], %[ftmp4] \n\t"
"pshufh %[a], %[a], %[ftmp0] \n\t"
"pshufh %[b], %[b], %[ftmp0] \n\t"
"1: \n\t"
PUT_VP8_BILINEAR4_H_MMI(%[src], %[dst])
"addiu %[h], %[h], -0x01 \n\t"
PTR_ADDU "%[src], %[src], %[sstride] \n\t"
PTR_ADDU "%[dst], %[dst], %[dstride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]),
[tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_LOW32
RESTRICT_ASM_ALL64
[h]"+&r"(h),
[dst]"+&r"(dst), [src]"+&r"(src),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[a]"+&f"(a.f), [b]"+&f"(b.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: [sstride]"r"((mips_reg)sstride),
[dstride]"r"((mips_reg)dstride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[ff_pw_4]"f"(ff_pw_4.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
int a = 8 - mx, b = mx;
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < 4; x++)
dst[x] = (a * src[x] + b * src[x + 1] + 4) >> 3;
dst += dstride;
src += sstride;
}
#endif
}
void ff_put_vp8_bilinear4_v_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src,
ptrdiff_t sstride, int h, int mx, int my)
{
#if 1
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
union mmi_intfloat64 c, d;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
double ftmp[7];
uint32_t tmp[1];
mips_reg src1;
DECLARE_VAR_LOW32;
DECLARE_VAR_ALL64;
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
c.i = 8 - my;
d.i = my;
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
/*
dst[0] = (c * src[0] + d * src[ sstride] + 4) >> 3;
dst[1] = (c * src[1] + d * src[1 + sstride] + 4) >> 3;
dst[2] = (c * src[2] + d * src[2 + sstride] + 4) >> 3;
dst[3] = (c * src[3] + d * src[3 + sstride] + 4) >> 3;
*/
__asm__ volatile (
libavcodec/mips: Fix specification of instruction name 1.'xor,or,and' to 'pxor,por,pand'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. 2.'dsrl,srl' to 'ssrld,ssrlw'. In the case of operating FPR, gcc supports both of them, clang only supports the second type. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
"li %[tmp0], 0x03 \n\t"
"mtc1 %[tmp0], %[ftmp4] \n\t"
"pshufh %[c], %[c], %[ftmp0] \n\t"
"pshufh %[d], %[d], %[ftmp0] \n\t"
"1: \n\t"
PUT_VP8_BILINEAR4_V_MMI(%[src], %[src1], %[dst], %[sstride])
"addiu %[h], %[h], -0x01 \n\t"
PTR_ADDU "%[src], %[src], %[sstride] \n\t"
PTR_ADDU "%[dst], %[dst], %[dstride] \n\t"
"bnez %[h], 1b \n\t"
: [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]),
[ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]),
[ftmp4]"=&f"(ftmp[4]),
[tmp0]"=&r"(tmp[0]),
RESTRICT_ASM_LOW32
RESTRICT_ASM_ALL64
[src1]"=&r"(src1),
[h]"+&r"(h),
[dst]"+&r"(dst), [src]"+&r"(src),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[c]"+&f"(c.f), [d]"+&f"(d.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: [sstride]"r"((mips_reg)sstride),
[dstride]"r"((mips_reg)dstride),
libavcodec/mips: Fix build errors reported by clang Clang is more strict on the type of asm operands, float or double type variable should use constraint 'f', integer variable should use constraint 'r'. Signed-off-by: Jin Bo <jinbo@loongson.cn> Reviewed-by: yinshiyou-hf@loongson.cn Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
4 years ago
[ff_pw_4]"f"(ff_pw_4.f)
avcodec/mips: version 1 of vp8dsp optimizations for loongson mmi Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
8 years ago
: "memory"
);
#else
int c = 8 - my, d = my;
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < 4; x++)
dst[x] = (c * src[x] + d * src[x + sstride] + 4) >> 3;
dst += dstride;
src += sstride;
}
#endif
}
void ff_put_vp8_bilinear4_hv_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src,
ptrdiff_t sstride, int h, int mx, int my)
{
#if 1
DECLARE_ALIGNED(4, uint8_t, tmp_array[36]);
uint8_t *tmp = tmp_array;
ff_put_vp8_bilinear4_h_mmi(tmp, 4, src, sstride, h + 1, mx, my);
ff_put_vp8_bilinear4_v_mmi(dst, dstride, tmp, 4, h, mx, my);
#else
int a = 8 - mx, b = mx;
int c = 8 - my, d = my;
int x, y;
uint8_t tmp_array[36];
uint8_t *tmp = tmp_array;
for (y = 0; y < h + 1; y++) {
for (x = 0; x < 4; x++)
tmp[x] = (a * src[x] + b * src[x + 1] + 4) >> 3;
tmp += 4;
src += sstride;
}
tmp = tmp_array;
for (y = 0; y < h; y++) {
for (x = 0; x < 4; x++)
dst[x] = (c * tmp[x] + d * tmp[x + 4] + 4) >> 3;
dst += dstride;
tmp += 4;
}
#endif
}