virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-12-12 19:18:44 +02:00
Code Issues Releases Activity
4b6b75c752
FFmpeg/libavcodec/mips/vp3dsp_idct_mmi.c

771 lines
47 KiB
C
Raw Normal View History

avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
/*
* Copyright (c) 2018 gxw <guxiwei-hf@loongson.cn>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "vp3dsp_mips.h"
#include "libavutil/intreadwrite.h"
#include "libavutil/mips/mmiutils.h"
#include "libavutil/common.h"
#include "libavcodec/rnd_avg.h"
#define LOAD_CONST(dst, value) \
"li %[tmp1], "#value" \n\t" \
"dmtc1 %[tmp1], "#dst" \n\t" \
"pshufh "#dst", "#dst", %[ftmp10] \n\t"
static void idct_row_mmi(int16_t *input)
{
double ftmp[23];
uint64_t tmp[2];
__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>
2021-05-28 04:04:39 +02:00
"pxor %[ftmp10], %[ftmp10], %[ftmp10] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
LOAD_CONST(%[csth_1], 1)
"li %[tmp0], 0x02 \n\t"
"1: \n\t"
/* Load input */
"ldc1 %[ftmp0], 0x00(%[input]) \n\t"
"ldc1 %[ftmp1], 0x10(%[input]) \n\t"
"ldc1 %[ftmp2], 0x20(%[input]) \n\t"
"ldc1 %[ftmp3], 0x30(%[input]) \n\t"
"ldc1 %[ftmp4], 0x40(%[input]) \n\t"
"ldc1 %[ftmp5], 0x50(%[input]) \n\t"
"ldc1 %[ftmp6], 0x60(%[input]) \n\t"
"ldc1 %[ftmp7], 0x70(%[input]) \n\t"
LOAD_CONST(%[ftmp8], 64277)
LOAD_CONST(%[ftmp9], 12785)
"pmulhh %[A], %[ftmp9], %[ftmp7] \n\t"
"pcmpgth %[C], %[ftmp10], %[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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[C], %[csth_1] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[B], %[ftmp1], %[mask] \n\t"
"pmulhuh %[B], %[ftmp8], %[B] \n\t"
"pmullh %[B], %[B], %[mask] \n\t"
"paddh %[A], %[A], %[B] \n\t"
"paddh %[A], %[A], %[C] \n\t"
"pcmpgth %[D], %[ftmp10], %[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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[D], %[csth_1] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[ftmp7], %[ftmp7], %[mask] \n\t"
"pmulhuh %[B], %[ftmp8], %[ftmp7] \n\t"
"pmullh %[B], %[B], %[mask] \n\t"
"pmulhh %[C], %[ftmp9], %[ftmp1] \n\t"
"psubh %[B], %[C], %[B] \n\t"
"psubh %[B], %[B], %[D] \n\t"
LOAD_CONST(%[ftmp8], 54491)
LOAD_CONST(%[ftmp9], 36410)
"pcmpgth %[Ad], %[ftmp10], %[ftmp5] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Ad], %[csth_1] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[ftmp1], %[ftmp5], %[mask] \n\t"
"pmulhuh %[C], %[ftmp9], %[ftmp1] \n\t"
"pmullh %[C], %[C], %[mask] \n\t"
"pcmpgth %[Bd], %[ftmp10], %[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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Bd], %[csth_1] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[D], %[ftmp3], %[mask] \n\t"
"pmulhuh %[D], %[ftmp8], %[D] \n\t"
"pmullh %[D], %[D], %[mask] \n\t"
"paddh %[C], %[C], %[D] \n\t"
"paddh %[C], %[C], %[Ad] \n\t"
"paddh %[C], %[C], %[Bd] \n\t"
"pcmpgth %[Bd], %[ftmp10], %[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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Bd], %[csth_1] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[ftmp1], %[ftmp3], %[mask] \n\t"
"pmulhuh %[D], %[ftmp9], %[ftmp1] \n\t"
"pmullh %[D], %[D], %[mask] \n\t"
"pcmpgth %[Ed], %[ftmp10], %[ftmp5] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Ed], %[csth_1] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[Ad], %[ftmp5], %[mask] \n\t"
"pmulhuh %[Ad], %[ftmp8], %[Ad] \n\t"
"pmullh %[Ad], %[Ad], %[mask] \n\t"
"psubh %[D], %[Ad], %[D] \n\t"
"paddh %[D], %[D], %[Ed] \n\t"
"psubh %[D], %[D], %[Bd] \n\t"
LOAD_CONST(%[ftmp8], 46341)
"psubh %[Ad], %[A], %[C] \n\t"
"pcmpgth %[Bd], %[ftmp10], %[Ad] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Bd], %[csth_1] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[Ad], %[Ad], %[mask] \n\t"
"pmulhuh %[Ad], %[ftmp8], %[Ad] \n\t"
"pmullh %[Ad], %[Ad], %[mask] \n\t"
"paddh %[Ad], %[Ad], %[Bd] \n\t"
"psubh %[Bd], %[B], %[D] \n\t"
"pcmpgth %[Cd], %[ftmp10], %[Bd] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Cd], %[csth_1] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[Bd], %[Bd], %[mask] \n\t"
"pmulhuh %[Bd], %[ftmp8], %[Bd] \n\t"
"pmullh %[Bd], %[Bd], %[mask] \n\t"
"paddh %[Bd], %[Bd], %[Cd] \n\t"
"paddh %[Cd], %[A], %[C] \n\t"
"paddh %[Dd], %[B], %[D] \n\t"
"paddh %[A], %[ftmp0], %[ftmp4] \n\t"
"pcmpgth %[B], %[ftmp10], %[A] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[B], %[csth_1] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[A], %[A], %[mask] \n\t"
"pmulhuh %[A], %[ftmp8], %[A] \n\t"
"pmullh %[A], %[A], %[mask] \n\t"
"paddh %[A], %[A], %[B] \n\t"
"psubh %[B], %[ftmp0], %[ftmp4] \n\t"
"pcmpgth %[C], %[ftmp10], %[B] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[C], %[csth_1] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[B], %[B], %[mask] \n\t"
"pmulhuh %[B], %[ftmp8], %[B] \n\t"
"pmullh %[B], %[B], %[mask] \n\t"
"paddh %[B], %[B], %[C] \n\t"
LOAD_CONST(%[ftmp8], 60547)
LOAD_CONST(%[ftmp9], 25080)
"pmulhh %[C], %[ftmp9], %[ftmp6] \n\t"
"pcmpgth %[D], %[ftmp10], %[ftmp2] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[D], %[csth_1] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[Ed], %[ftmp2], %[mask] \n\t"
"pmulhuh %[Ed], %[ftmp8], %[Ed] \n\t"
"pmullh %[Ed], %[Ed], %[mask] \n\t"
"paddh %[C], %[C], %[Ed] \n\t"
"paddh %[C], %[C], %[D] \n\t"
"pcmpgth %[Ed], %[ftmp10], %[ftmp6] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Ed], %[csth_1] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[ftmp6], %[ftmp6], %[mask] \n\t"
"pmulhuh %[D], %[ftmp8], %[ftmp6] \n\t"
"pmullh %[D], %[D], %[mask] \n\t"
"pmulhh %[Gd], %[ftmp9], %[ftmp2] \n\t"
"psubh %[D], %[Gd], %[D] \n\t"
"psubh %[D], %[D], %[Ed] \n\t"
"psubh %[Ed], %[A], %[C] \n\t"
"paddh %[Gd], %[A], %[C] \n\t"
"paddh %[A], %[B], %[Ad] \n\t"
"psubh %[C], %[B], %[Ad] \n\t"
"psubh %[B], %[Bd], %[D] \n\t"
"paddh %[D], %[Bd], %[D] \n\t"
/* Final sequence of operations over-write original inputs */
"paddh %[ftmp0], %[Gd], %[Cd] \n\t"
"paddh %[ftmp1], %[A], %[D] \n\t"
"psubh %[ftmp2], %[A], %[D] \n\t"
"paddh %[ftmp3], %[Ed], %[Dd] \n\t"
"psubh %[ftmp4], %[Ed], %[Dd] \n\t"
"paddh %[ftmp5], %[C], %[B] \n\t"
"psubh %[ftmp6], %[C], %[B] \n\t"
"psubh %[ftmp7], %[Gd], %[Cd] \n\t"
"sdc1 %[ftmp0], 0x00(%[input]) \n\t"
"sdc1 %[ftmp1], 0x10(%[input]) \n\t"
"sdc1 %[ftmp2], 0x20(%[input]) \n\t"
"sdc1 %[ftmp3], 0x30(%[input]) \n\t"
"sdc1 %[ftmp4], 0x40(%[input]) \n\t"
"sdc1 %[ftmp5], 0x50(%[input]) \n\t"
"sdc1 %[ftmp6], 0x60(%[input]) \n\t"
"sdc1 %[ftmp7], 0x70(%[input]) \n\t"
PTR_ADDU "%[tmp0], %[tmp0], -0x01 \n\t"
PTR_ADDIU "%[input], %[input], 0x08 \n\t"
"bnez %[tmp0], 1b \n\t"
: [input]"+&r"(input), [tmp0]"=&r"(tmp[0]), [tmp1]"=&r"(tmp[1]),
[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]), [mask]"=&f"(ftmp[11]),
[A]"=&f"(ftmp[12]), [B]"=&f"(ftmp[13]), [C]"=&f"(ftmp[14]),
[D]"=&f"(ftmp[15]), [Ad]"=&f"(ftmp[16]), [Bd]"=&f"(ftmp[17]),
[Cd]"=&f"(ftmp[18]), [Dd]"=&f"(ftmp[19]), [Ed]"=&f"(ftmp[20]),
[Gd]"=&f"(ftmp[21]), [csth_1]"=&f"(ftmp[22])
:
: "memory"
);
}
static void idct_column_true_mmi(uint8_t *dst, int stride, int16_t *input)
{
uint8_t temp_value[8];
double ftmp[23];
uint64_t tmp[2];
for (int i = 0; i < 8; ++i)
temp_value[i] = av_clip_uint8(128 + ((46341 * input[i << 3] + (8 << 16)) >> 20));
__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>
2021-05-28 04:04:39 +02:00
"pxor %[ftmp10], %[ftmp10], %[ftmp10] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"li %[tmp0], 0x02 \n\t"
"1: \n\t"
"ldc1 %[ftmp0], 0x00(%[input]) \n\t"
"ldc1 %[ftmp4], 0x08(%[input]) \n\t"
"ldc1 %[ftmp1], 0x10(%[input]) \n\t"
"ldc1 %[ftmp5], 0x18(%[input]) \n\t"
"ldc1 %[ftmp2], 0x20(%[input]) \n\t"
"ldc1 %[ftmp6], 0x28(%[input]) \n\t"
"ldc1 %[ftmp3], 0x30(%[input]) \n\t"
"ldc1 %[ftmp7], 0x38(%[input]) \n\t"
TRANSPOSE_4H(%[ftmp0], %[ftmp1], %[ftmp2], %[ftmp3],
%[A], %[B], %[C], %[D])
TRANSPOSE_4H(%[ftmp4], %[ftmp5], %[ftmp6], %[ftmp7],
%[A], %[B], %[C], %[D])
LOAD_CONST(%[ftmp8], 64277)
LOAD_CONST(%[ftmp9], 12785)
LOAD_CONST(%[Gd], 1)
"pmulhh %[A], %[ftmp9], %[ftmp7] \n\t"
"pcmpgth %[C], %[ftmp10], %[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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[C], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[B], %[ftmp1], %[mask] \n\t"
"pmulhuh %[B], %[ftmp8], %[B] \n\t"
"pmullh %[B], %[B], %[mask] \n\t"
"paddh %[A], %[A], %[B] \n\t"
"paddh %[A], %[A], %[C] \n\t"
"pcmpgth %[D], %[ftmp10], %[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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[D], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[Ad], %[ftmp7], %[mask] \n\t"
"pmulhuh %[B], %[ftmp8], %[Ad] \n\t"
"pmullh %[B], %[B], %[mask] \n\t"
"pmulhh %[C], %[ftmp9], %[ftmp1] \n\t"
"psubh %[B], %[C], %[B] \n\t"
"psubh %[B], %[B], %[D] \n\t"
LOAD_CONST(%[ftmp8], 54491)
LOAD_CONST(%[ftmp9], 36410)
"pcmpgth %[Ad], %[ftmp10], %[ftmp5] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Ad], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[Cd], %[ftmp5], %[mask] \n\t"
"pmulhuh %[C], %[ftmp9], %[Cd] \n\t"
"pmullh %[C], %[C], %[mask] \n\t"
"pcmpgth %[Bd], %[ftmp10], %[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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Bd], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[D], %[ftmp3], %[mask] \n\t"
"pmulhuh %[D], %[ftmp8], %[D] \n\t"
"pmullh %[D], %[D], %[mask] \n\t"
"paddh %[C], %[C], %[D] \n\t"
"paddh %[C], %[C], %[Ad] \n\t"
"paddh %[C], %[C], %[Bd] \n\t"
"pcmpgth %[Bd], %[ftmp10], %[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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Bd], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[Cd], %[ftmp3], %[mask] \n\t"
"pmulhuh %[D], %[ftmp9], %[Cd] \n\t"
"pmullh %[D], %[D], %[mask] \n\t"
"pcmpgth %[Ed], %[ftmp10], %[ftmp5] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Ed], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[Ad], %[ftmp5], %[mask] \n\t"
"pmulhuh %[Ad], %[ftmp8], %[Ad] \n\t"
"pmullh %[Ad], %[Ad], %[mask] \n\t"
"psubh %[D], %[Ad], %[D] \n\t"
"paddh %[D], %[D], %[Ed] \n\t"
"psubh %[D], %[D], %[Bd] \n\t"
LOAD_CONST(%[ftmp8], 46341)
"psubh %[Ad], %[A], %[C] \n\t"
"pcmpgth %[Bd], %[ftmp10], %[Ad] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Bd], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[Ad], %[Ad], %[mask] \n\t"
"pmulhuh %[Ad], %[ftmp8], %[Ad] \n\t"
"pmullh %[Ad], %[Ad], %[mask] \n\t"
"paddh %[Ad], %[Ad], %[Bd] \n\t"
"psubh %[Bd], %[B], %[D] \n\t"
"pcmpgth %[Cd], %[ftmp10], %[Bd] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Cd], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[Bd], %[Bd], %[mask] \n\t"
"pmulhuh %[Bd], %[ftmp8], %[Bd] \n\t"
"pmullh %[Bd], %[Bd], %[mask] \n\t"
"paddh %[Bd], %[Bd], %[Cd] \n\t"
"paddh %[Cd], %[A], %[C] \n\t"
"paddh %[Dd], %[B], %[D] \n\t"
LOAD_CONST(%[Ed], 2056)
"paddh %[A], %[ftmp0], %[ftmp4] \n\t"
"pcmpgth %[B], %[ftmp10], %[A] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[B], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[A], %[A], %[mask] \n\t"
"pmulhuh %[A], %[ftmp8], %[A] \n\t"
"pmullh %[A], %[A], %[mask] \n\t"
"paddh %[A], %[A], %[B] \n\t"
"paddh %[A], %[A], %[Ed] \n\t"
"psubh %[B], %[ftmp0], %[ftmp4] \n\t"
"pcmpgth %[C], %[ftmp10], %[B] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[C], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[B], %[B], %[mask] \n\t"
"pmulhuh %[B], %[ftmp8], %[B] \n\t"
"pmullh %[B], %[B], %[mask] \n\t"
"paddh %[B], %[B], %[C] \n\t"
"paddh %[B], %[B], %[Ed] \n\t"
LOAD_CONST(%[ftmp8], 60547)
LOAD_CONST(%[ftmp9], 25080)
"pmulhh %[C], %[ftmp9], %[ftmp6] \n\t"
"pcmpgth %[D], %[ftmp10], %[ftmp2] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[D], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[Ed], %[ftmp2], %[mask] \n\t"
"pmulhuh %[Ed], %[ftmp8], %[Ed] \n\t"
"pmullh %[Ed], %[Ed], %[mask] \n\t"
"paddh %[C], %[C], %[Ed] \n\t"
"paddh %[C], %[C], %[D] \n\t"
"pcmpgth %[Ed], %[ftmp10], %[ftmp6] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Ed], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[D], %[ftmp6], %[mask] \n\t"
"pmulhuh %[D], %[ftmp8], %[D] \n\t"
"pmullh %[D], %[D], %[mask] \n\t"
"pmulhh %[Gd], %[ftmp9], %[ftmp2] \n\t"
"psubh %[D], %[Gd], %[D] \n\t"
"psubh %[D], %[D], %[Ed] \n\t"
"psubh %[Ed], %[A], %[C] \n\t"
"paddh %[Gd], %[A], %[C] \n\t"
"paddh %[A], %[B], %[Ad] \n\t"
"psubh %[C], %[B], %[Ad] \n\t"
"psubh %[B], %[Bd], %[D] \n\t"
"paddh %[D], %[Bd], %[D] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[ftmp1], %[ftmp2] \n\t"
"por %[mask], %[mask], %[ftmp3] \n\t"
"por %[mask], %[mask], %[ftmp4] \n\t"
"por %[mask], %[mask], %[ftmp5] \n\t"
"por %[mask], %[mask], %[ftmp6] \n\t"
"por %[mask], %[mask], %[ftmp7] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pcmpeqh %[mask], %[mask], %[ftmp10] \n\t"
"packushb %[mask], %[mask], %[ftmp10] \n\t"
"li %[tmp1], 0x04 \n\t"
"dmtc1 %[tmp1], %[ftmp8] \n\t"
"paddh %[ftmp0], %[Gd], %[Cd] \n\t"
"psrah %[ftmp0], %[ftmp0], %[ftmp8] \n\t"
"paddh %[ftmp1], %[A], %[D] \n\t"
"psrah %[ftmp1], %[ftmp1], %[ftmp8] \n\t"
"psubh %[ftmp2], %[A], %[D] \n\t"
"psrah %[ftmp2], %[ftmp2], %[ftmp8] \n\t"
"paddh %[ftmp3], %[Ed], %[Dd] \n\t"
"psrah %[ftmp3], %[ftmp3], %[ftmp8] \n\t"
"psubh %[ftmp4], %[Ed], %[Dd] \n\t"
"psrah %[ftmp4], %[ftmp4], %[ftmp8] \n\t"
"paddh %[ftmp5], %[C], %[B] \n\t"
"psrah %[ftmp5], %[ftmp5], %[ftmp8] \n\t"
"psubh %[ftmp6], %[C], %[B] \n\t"
"psrah %[ftmp6], %[ftmp6], %[ftmp8] \n\t"
"psubh %[ftmp7], %[Gd], %[Cd] \n\t"
"psrah %[ftmp7], %[ftmp7], %[ftmp8] \n\t"
"pmaxsh %[ftmp0], %[ftmp0], %[ftmp10] \n\t"
"packushb %[ftmp0], %[ftmp0], %[ftmp10] \n\t"
"pmaxsh %[ftmp1], %[ftmp1], %[ftmp10] \n\t"
"packushb %[ftmp1], %[ftmp1], %[ftmp10] \n\t"
"pmaxsh %[ftmp2], %[ftmp2], %[ftmp10] \n\t"
"packushb %[ftmp2], %[ftmp2], %[ftmp10] \n\t"
"pmaxsh %[ftmp3], %[ftmp3], %[ftmp10] \n\t"
"packushb %[ftmp3], %[ftmp3], %[ftmp10] \n\t"
"pmaxsh %[ftmp4], %[ftmp4], %[ftmp10] \n\t"
"packushb %[ftmp4], %[ftmp4], %[ftmp10] \n\t"
"pmaxsh %[ftmp5], %[ftmp5], %[ftmp10] \n\t"
"packushb %[ftmp5], %[ftmp5], %[ftmp10] \n\t"
"pmaxsh %[ftmp6], %[ftmp6], %[ftmp10] \n\t"
"packushb %[ftmp6], %[ftmp6], %[ftmp10] \n\t"
"pmaxsh %[ftmp7], %[ftmp7], %[ftmp10] \n\t"
"packushb %[ftmp7], %[ftmp7], %[ftmp10] \n\t"
"lwc1 %[Ed], 0x00(%[temp_value]) \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>
2021-05-28 04:04:39 +02:00
"pand %[Ed], %[Ed], %[mask] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"paddb %[ftmp0], %[ftmp0], %[Ed] \n\t"
"paddb %[ftmp1], %[ftmp1], %[Ed] \n\t"
"paddb %[ftmp2], %[ftmp2], %[Ed] \n\t"
"paddb %[ftmp3], %[ftmp3], %[Ed] \n\t"
"paddb %[ftmp4], %[ftmp4], %[Ed] \n\t"
"paddb %[ftmp5], %[ftmp5], %[Ed] \n\t"
"paddb %[ftmp6], %[ftmp6], %[Ed] \n\t"
"paddb %[ftmp7], %[ftmp7], %[Ed] \n\t"
"swc1 %[ftmp0], 0x00(%[dst]) \n\t"
PTR_ADDU "%[tmp1], %[dst], %[stride] \n\t"
"swc1 %[ftmp1], 0x00(%[tmp1]) \n\t"
PTR_ADDU "%[tmp1], %[tmp1], %[stride] \n\t"
"swc1 %[ftmp2], 0x00(%[tmp1]) \n\t"
PTR_ADDU "%[tmp1], %[tmp1], %[stride] \n\t"
"swc1 %[ftmp3], 0x00(%[tmp1]) \n\t"
PTR_ADDU "%[tmp1], %[tmp1], %[stride] \n\t"
"swc1 %[ftmp4], 0x00(%[tmp1]) \n\t"
PTR_ADDU "%[tmp1], %[tmp1], %[stride] \n\t"
"swc1 %[ftmp5], 0x00(%[tmp1]) \n\t"
PTR_ADDU "%[tmp1], %[tmp1], %[stride] \n\t"
"swc1 %[ftmp6], 0x00(%[tmp1]) \n\t"
PTR_ADDU "%[tmp1], %[tmp1], %[stride] \n\t"
"swc1 %[ftmp7], 0x00(%[tmp1]) \n\t"
PTR_ADDIU "%[dst], %[dst], 0x04 \n\t"
PTR_ADDIU "%[input], %[input], 0x40 \n\t"
PTR_ADDIU "%[temp_value], %[temp_value], 0x04 \n\t"
PTR_ADDIU "%[tmp0], %[tmp0], -0x01 \n\t"
"bnez %[tmp0], 1b \n\t"
: [dst]"+&r"(dst), [tmp0]"=&r"(tmp[0]), [tmp1]"=&r"(tmp[1]),
[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]), [mask]"=&f"(ftmp[11]),
[A]"=&f"(ftmp[12]), [B]"=&f"(ftmp[13]), [C]"=&f"(ftmp[14]),
[D]"=&f"(ftmp[15]), [Ad]"=&f"(ftmp[16]), [Bd]"=&f"(ftmp[17]),
[Cd]"=&f"(ftmp[18]), [Dd]"=&f"(ftmp[19]), [Ed]"=&f"(ftmp[20]),
[Gd]"=&f"(ftmp[21]), [input]"+&r"(input)
: [stride]"r"(stride), [temp_value]"r"(temp_value)
: "memory"
);
}
static void idct_column_false_mmi(uint8_t *dst, int stride, int16_t *input)
{
int16_t temp_value[8];
double ftmp[23];
uint64_t tmp[2];
for (int i = 0; i < 8; ++i)
temp_value[i] = (46341 * input[i << 3] + (8 << 16)) >> 20;
__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>
2021-05-28 04:04:39 +02:00
"pxor %[ftmp10], %[ftmp10], %[ftmp10] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"li %[tmp0], 0x02 \n\t"
"1: \n\t"
"ldc1 %[ftmp0], 0x00(%[input]) \n\t"
"ldc1 %[ftmp4], 0x08(%[input]) \n\t"
"ldc1 %[ftmp1], 0x10(%[input]) \n\t"
"ldc1 %[ftmp5], 0x18(%[input]) \n\t"
"ldc1 %[ftmp2], 0x20(%[input]) \n\t"
"ldc1 %[ftmp6], 0x28(%[input]) \n\t"
"ldc1 %[ftmp3], 0x30(%[input]) \n\t"
"ldc1 %[ftmp7], 0x38(%[input]) \n\t"
TRANSPOSE_4H(%[ftmp0], %[ftmp1], %[ftmp2], %[ftmp3],
%[A], %[B], %[C], %[D])
TRANSPOSE_4H(%[ftmp4], %[ftmp5], %[ftmp6], %[ftmp7],
%[A], %[B], %[C], %[D])
LOAD_CONST(%[ftmp8], 64277)
LOAD_CONST(%[ftmp9], 12785)
LOAD_CONST(%[Gd], 1)
"pmulhh %[A], %[ftmp9], %[ftmp7] \n\t"
"pcmpgth %[C], %[ftmp10], %[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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[C], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[B], %[ftmp1], %[mask] \n\t"
"pmulhuh %[B], %[ftmp8], %[B] \n\t"
"pmullh %[B], %[B], %[mask] \n\t"
"paddh %[A], %[A], %[B] \n\t"
"paddh %[A], %[A], %[C] \n\t"
"pcmpgth %[D], %[ftmp10], %[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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[D], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[Ad], %[ftmp7], %[mask] \n\t"
"pmulhuh %[B], %[ftmp8], %[Ad] \n\t"
"pmullh %[B], %[B], %[mask] \n\t"
"pmulhh %[C], %[ftmp9], %[ftmp1] \n\t"
"psubh %[B], %[C], %[B] \n\t"
"psubh %[B], %[B], %[D] \n\t"
LOAD_CONST(%[ftmp8], 54491)
LOAD_CONST(%[ftmp9], 36410)
"pcmpgth %[Ad], %[ftmp10], %[ftmp5] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Ad], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[Cd], %[ftmp5], %[mask] \n\t"
"pmulhuh %[C], %[ftmp9], %[Cd] \n\t"
"pmullh %[C], %[C], %[mask] \n\t"
"pcmpgth %[Bd], %[ftmp10], %[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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Bd], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[D], %[ftmp3], %[mask] \n\t"
"pmulhuh %[D], %[ftmp8], %[D] \n\t"
"pmullh %[D], %[D], %[mask] \n\t"
"paddh %[C], %[C], %[D] \n\t"
"paddh %[C], %[C], %[Ad] \n\t"
"paddh %[C], %[C], %[Bd] \n\t"
"pcmpgth %[Bd], %[ftmp10], %[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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Bd], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[Cd], %[ftmp3], %[mask] \n\t"
"pmulhuh %[D], %[ftmp9], %[Cd] \n\t"
"pmullh %[D], %[D], %[mask] \n\t"
"pcmpgth %[Ed], %[ftmp10], %[ftmp5] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Ed], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[Ad], %[ftmp5], %[mask] \n\t"
"pmulhuh %[Ad], %[ftmp8], %[Ad] \n\t"
"pmullh %[Ad], %[Ad], %[mask] \n\t"
"psubh %[D], %[Ad], %[D] \n\t"
"paddh %[D], %[D], %[Ed] \n\t"
"psubh %[D], %[D], %[Bd] \n\t"
LOAD_CONST(%[ftmp8], 46341)
"psubh %[Ad], %[A], %[C] \n\t"
"pcmpgth %[Bd], %[ftmp10], %[Ad] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Bd], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[Ad], %[Ad], %[mask] \n\t"
"pmulhuh %[Ad], %[ftmp8], %[Ad] \n\t"
"pmullh %[Ad], %[Ad], %[mask] \n\t"
"paddh %[Ad], %[Ad], %[Bd] \n\t"
"psubh %[Bd], %[B], %[D] \n\t"
"pcmpgth %[Cd], %[ftmp10], %[Bd] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Cd], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[Bd], %[Bd], %[mask] \n\t"
"pmulhuh %[Bd], %[ftmp8], %[Bd] \n\t"
"pmullh %[Bd], %[Bd], %[mask] \n\t"
"paddh %[Bd], %[Bd], %[Cd] \n\t"
"paddh %[Cd], %[A], %[C] \n\t"
"paddh %[Dd], %[B], %[D] \n\t"
LOAD_CONST(%[Ed], 8)
"paddh %[A], %[ftmp0], %[ftmp4] \n\t"
"pcmpgth %[B], %[ftmp10], %[A] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[B], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[A], %[A], %[mask] \n\t"
"pmulhuh %[A], %[ftmp8], %[A] \n\t"
"pmullh %[A], %[A], %[mask] \n\t"
"paddh %[A], %[A], %[B] \n\t"
"paddh %[A], %[A], %[Ed] \n\t"
"psubh %[B], %[ftmp0], %[ftmp4] \n\t"
"pcmpgth %[C], %[ftmp10], %[B] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[C], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[B], %[B], %[mask] \n\t"
"pmulhuh %[B], %[ftmp8], %[B] \n\t"
"pmullh %[B], %[B], %[mask] \n\t"
"paddh %[B], %[B], %[C] \n\t"
"paddh %[B], %[B], %[Ed] \n\t"
LOAD_CONST(%[ftmp8], 60547)
LOAD_CONST(%[ftmp9], 25080)
"pmulhh %[C], %[ftmp9], %[ftmp6] \n\t"
"pcmpgth %[D], %[ftmp10], %[ftmp2] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[D], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[Ed], %[ftmp2], %[mask] \n\t"
"pmulhuh %[Ed], %[ftmp8], %[Ed] \n\t"
"pmullh %[Ed], %[Ed], %[mask] \n\t"
"paddh %[C], %[C], %[Ed] \n\t"
"paddh %[C], %[C], %[D] \n\t"
"pcmpgth %[Ed], %[ftmp10], %[ftmp6] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[Ed], %[Gd] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pmullh %[D], %[ftmp6], %[mask] \n\t"
"pmulhuh %[D], %[ftmp8], %[D] \n\t"
"pmullh %[D], %[D], %[mask] \n\t"
"pmulhh %[Gd], %[ftmp9], %[ftmp2] \n\t"
"psubh %[D], %[Gd], %[D] \n\t"
"psubh %[D], %[D], %[Ed] \n\t"
"psubh %[Ed], %[A], %[C] \n\t"
"paddh %[Gd], %[A], %[C] \n\t"
"paddh %[A], %[B], %[Ad] \n\t"
"psubh %[C], %[B], %[Ad] \n\t"
"psubh %[B], %[Bd], %[D] \n\t"
"paddh %[D], %[Bd], %[D] \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>
2021-05-28 04:04:39 +02:00
"por %[mask], %[ftmp1], %[ftmp2] \n\t"
"por %[mask], %[mask], %[ftmp3] \n\t"
"por %[mask], %[mask], %[ftmp4] \n\t"
"por %[mask], %[mask], %[ftmp5] \n\t"
"por %[mask], %[mask], %[ftmp6] \n\t"
"por %[mask], %[mask], %[ftmp7] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"pcmpeqh %[mask], %[mask], %[ftmp10] \n\t"
"li %[tmp1], 0x04 \n\t"
"dmtc1 %[tmp1], %[ftmp8] \n\t"
"paddh %[ftmp0], %[Gd], %[Cd] \n\t"
"psrah %[ftmp0], %[ftmp0], %[ftmp8] \n\t"
"paddh %[ftmp1], %[A], %[D] \n\t"
"psrah %[ftmp1], %[ftmp1], %[ftmp8] \n\t"
"psubh %[ftmp2], %[A], %[D] \n\t"
"psrah %[ftmp2], %[ftmp2], %[ftmp8] \n\t"
"paddh %[ftmp3], %[Ed], %[Dd] \n\t"
"psrah %[ftmp3], %[ftmp3], %[ftmp8] \n\t"
"psubh %[ftmp4], %[Ed], %[Dd] \n\t"
"psrah %[ftmp4], %[ftmp4], %[ftmp8] \n\t"
"paddh %[ftmp5], %[C], %[B] \n\t"
"psrah %[ftmp5], %[ftmp5], %[ftmp8] \n\t"
"psubh %[ftmp6], %[C], %[B] \n\t"
"psrah %[ftmp6], %[ftmp6], %[ftmp8] \n\t"
"psubh %[ftmp7], %[Gd], %[Cd] \n\t"
"psrah %[ftmp7], %[ftmp7], %[ftmp8] \n\t"
/* Load from dst */
"lwc1 %[A], 0x00(%[dst]) \n\t"
PTR_ADDU "%[tmp1], %[dst], %[stride] \n\t"
"lwc1 %[B], 0x00(%[tmp1]) \n\t"
PTR_ADDU "%[tmp1], %[tmp1], %[stride] \n\t"
"lwc1 %[C], 0x00(%[tmp1]) \n\t"
PTR_ADDU "%[tmp1], %[tmp1], %[stride] \n\t"
"lwc1 %[D], 0x00(%[tmp1]) \n\t"
PTR_ADDU "%[tmp1], %[tmp1], %[stride] \n\t"
"lwc1 %[Ad], 0x00(%[tmp1]) \n\t"
PTR_ADDU "%[tmp1], %[tmp1], %[stride] \n\t"
"lwc1 %[Bd], 0x00(%[tmp1]) \n\t"
PTR_ADDU "%[tmp1], %[tmp1], %[stride] \n\t"
"lwc1 %[Cd], 0x00(%[tmp1]) \n\t"
PTR_ADDU "%[tmp1], %[tmp1], %[stride] \n\t"
"lwc1 %[Dd], 0x00(%[tmp1]) \n\t"
"punpcklbh %[A], %[A], %[ftmp10] \n\t"
"punpcklbh %[B], %[B], %[ftmp10] \n\t"
"punpcklbh %[C], %[C], %[ftmp10] \n\t"
"punpcklbh %[D], %[D], %[ftmp10] \n\t"
"punpcklbh %[Ad], %[Ad], %[ftmp10] \n\t"
"punpcklbh %[Bd], %[Bd], %[ftmp10] \n\t"
"punpcklbh %[Cd], %[Cd], %[ftmp10] \n\t"
"punpcklbh %[Dd], %[Dd], %[ftmp10] \n\t"
"ldc1 %[Ed], 0x00(%[temp_value]) \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>
2021-05-28 04:04:39 +02:00
"pand %[Ed], %[Ed], %[mask] \n\t"
"pnor %[mask], %[mask], %[mask] \n\t"
"pand %[ftmp0], %[ftmp0], %[mask] \n\t"
"pand %[ftmp1], %[ftmp1], %[mask] \n\t"
"pand %[ftmp2], %[ftmp2], %[mask] \n\t"
"pand %[ftmp3], %[ftmp3], %[mask] \n\t"
"pand %[ftmp4], %[ftmp4], %[mask] \n\t"
"pand %[ftmp5], %[ftmp5], %[mask] \n\t"
"pand %[ftmp6], %[ftmp6], %[mask] \n\t"
"pand %[ftmp7], %[ftmp7], %[mask] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"paddh %[ftmp0], %[ftmp0], %[A] \n\t"
"paddh %[ftmp1], %[ftmp1], %[B] \n\t"
"paddh %[ftmp2], %[ftmp2], %[C] \n\t"
"paddh %[ftmp3], %[ftmp3], %[D] \n\t"
"paddh %[ftmp4], %[ftmp4], %[Ad] \n\t"
"paddh %[ftmp5], %[ftmp5], %[Bd] \n\t"
"paddh %[ftmp6], %[ftmp6], %[Cd] \n\t"
"paddh %[ftmp7], %[ftmp7], %[Dd] \n\t"
"paddh %[ftmp0], %[ftmp0], %[Ed] \n\t"
"paddh %[ftmp1], %[ftmp1], %[Ed] \n\t"
"paddh %[ftmp2], %[ftmp2], %[Ed] \n\t"
"paddh %[ftmp3], %[ftmp3], %[Ed] \n\t"
"paddh %[ftmp4], %[ftmp4], %[Ed] \n\t"
"paddh %[ftmp5], %[ftmp5], %[Ed] \n\t"
"paddh %[ftmp6], %[ftmp6], %[Ed] \n\t"
"paddh %[ftmp7], %[ftmp7], %[Ed] \n\t"
"pmaxsh %[ftmp0], %[ftmp0], %[ftmp10] \n\t"
"packushb %[ftmp0], %[ftmp0], %[ftmp10] \n\t"
"pmaxsh %[ftmp1], %[ftmp1], %[ftmp10] \n\t"
"packushb %[ftmp1], %[ftmp1], %[ftmp10] \n\t"
"pmaxsh %[ftmp2], %[ftmp2], %[ftmp10] \n\t"
"packushb %[ftmp2], %[ftmp2], %[ftmp10] \n\t"
"pmaxsh %[ftmp3], %[ftmp3], %[ftmp10] \n\t"
"packushb %[ftmp3], %[ftmp3], %[ftmp10] \n\t"
"pmaxsh %[ftmp4], %[ftmp4], %[ftmp10] \n\t"
"packushb %[ftmp4], %[ftmp4], %[ftmp10] \n\t"
"pmaxsh %[ftmp5], %[ftmp5], %[ftmp10] \n\t"
"packushb %[ftmp5], %[ftmp5], %[ftmp10] \n\t"
"pmaxsh %[ftmp6], %[ftmp6], %[ftmp10] \n\t"
"packushb %[ftmp6], %[ftmp6], %[ftmp10] \n\t"
"pmaxsh %[ftmp7], %[ftmp7], %[ftmp10] \n\t"
"packushb %[ftmp7], %[ftmp7], %[ftmp10] \n\t"
"swc1 %[ftmp0], 0x00(%[dst]) \n\t"
PTR_ADDU "%[tmp1], %[dst], %[stride] \n\t"
"swc1 %[ftmp1], 0x00(%[tmp1]) \n\t"
PTR_ADDU "%[tmp1], %[tmp1], %[stride] \n\t"
"swc1 %[ftmp2], 0x00(%[tmp1]) \n\t"
PTR_ADDU "%[tmp1], %[tmp1], %[stride] \n\t"
"swc1 %[ftmp3], 0x00(%[tmp1]) \n\t"
PTR_ADDU "%[tmp1], %[tmp1], %[stride] \n\t"
"swc1 %[ftmp4], 0x00(%[tmp1]) \n\t"
PTR_ADDU "%[tmp1], %[tmp1], %[stride] \n\t"
"swc1 %[ftmp5], 0x00(%[tmp1]) \n\t"
PTR_ADDU "%[tmp1], %[tmp1], %[stride] \n\t"
"swc1 %[ftmp6], 0x00(%[tmp1]) \n\t"
PTR_ADDU "%[tmp1], %[tmp1], %[stride] \n\t"
"swc1 %[ftmp7], 0x00(%[tmp1]) \n\t"
PTR_ADDIU "%[dst], %[dst], 0x04 \n\t"
PTR_ADDIU "%[input], %[input], 0x40 \n\t"
PTR_ADDIU "%[temp_value], %[temp_value], 0x08 \n\t"
PTR_ADDIU "%[tmp0], %[tmp0], -0x01 \n\t"
"bnez %[tmp0], 1b \n\t"
: [dst]"+&r"(dst), [tmp0]"=&r"(tmp[0]), [tmp1]"=&r"(tmp[1]),
[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]), [mask]"=&f"(ftmp[11]),
[A]"=&f"(ftmp[12]), [B]"=&f"(ftmp[13]), [C]"=&f"(ftmp[14]),
[D]"=&f"(ftmp[15]), [Ad]"=&f"(ftmp[16]), [Bd]"=&f"(ftmp[17]),
[Cd]"=&f"(ftmp[18]), [Dd]"=&f"(ftmp[19]), [Ed]"=&f"(ftmp[20]),
[Gd]"=&f"(ftmp[21]), [input]"+&r"(input)
: [stride]"r"(stride), [temp_value]"r"(temp_value)
: "memory"
);
}
static void idct_mmi(uint8_t *dst, int stride, int16_t *input, int type)
{
idct_row_mmi(input);
if (type == 1)
idct_column_true_mmi(dst, stride, input);
else
idct_column_false_mmi(dst, stride, input);
}
void ff_vp3_idct_put_mmi(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
{
idct_mmi(dest, line_size, block, 1);
memset(block, 0, sizeof(*block) << 6);
}
void ff_vp3_idct_add_mmi(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
{
idct_mmi(dest, line_size, block, 2);
memset(block, 0, sizeof(*block) << 6);
}
void ff_vp3_idct_dc_add_mmi(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
{
int dc = (block[0] + 15) >> 5;
double ftmp[7];
uint64_t tmp;
__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>
2021-05-28 04:04:39 +02:00
"pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"mtc1 %[dc], %[ftmp5] \n\t"
"pshufh %[ftmp5], %[ftmp5], %[ftmp0] \n\t"
"li %[tmp0], 0x08 \n\t"
"1: \n\t"
"ldc1 %[ftmp1], 0x00(%[dest]) \n\t"
"punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t"
"punpckhbh %[ftmp3], %[ftmp1], %[ftmp0] \n\t"
"paddh %[ftmp4], %[ftmp2], %[ftmp5] \n\t"
"paddh %[ftmp6], %[ftmp3], %[ftmp5] \n\t"
"packushb %[ftmp4], %[ftmp4], %[ftmp0] \n\t"
"packushb %[ftmp6], %[ftmp6], %[ftmp0] \n\t"
"swc1 %[ftmp4], 0x00(%[dest]) \n\t"
"swc1 %[ftmp6], 0x04(%[dest]) \n\t"
PTR_ADDU "%[dest], %[dest], %[line_size] \n\t"
PTR_ADDIU "%[tmp0], %[tmp0], -0x01 \n\t"
"bnez %[tmp0], 1b \n\t"
: [dest]"+&r"(dest), [block]"+&r"(block), [tmp0]"=&r"(tmp),
[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])
: [line_size]"r"(line_size), [dc]"r"(dc)
: "memory"
);
block[0] = 0;
}
void ff_put_no_rnd_pixels_l2_mmi(uint8_t *dst, const uint8_t *src1,
const uint8_t *src2, ptrdiff_t stride, int h)
{
if (h == 8) {
double ftmp[6];
uint64_t tmp[2];
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>
2021-07-23 07:53:42 +02:00
DECLARE_VAR_ALL64;
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
__asm__ volatile (
"li %[tmp0], 0x08 \n\t"
"li %[tmp1], 0xfefefefe \n\t"
"dmtc1 %[tmp1], %[ftmp4] \n\t"
"punpcklwd %[ftmp4], %[ftmp4], %[ftmp4] \n\t"
"li %[tmp1], 0x01 \n\t"
"dmtc1 %[tmp1], %[ftmp5] \n\t"
"1: \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>
2021-07-23 07:53:42 +02:00
MMI_ULDC1(%[ftmp1], %[src1], 0x0)
MMI_ULDC1(%[ftmp2], %[src2], 0x0)
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>
2021-05-28 04:04:39 +02:00
"pxor %[ftmp3], %[ftmp1], %[ftmp2] \n\t"
"pand %[ftmp3], %[ftmp3], %[ftmp4] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"psrlw %[ftmp3], %[ftmp3], %[ftmp5] \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>
2021-05-28 04:04:39 +02:00
"pand %[ftmp6], %[ftmp1], %[ftmp2] \n\t"
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
"paddw %[ftmp3], %[ftmp3], %[ftmp6] \n\t"
"sdc1 %[ftmp3], 0x00(%[dst]) \n\t"
PTR_ADDU "%[src1], %[src1], %[stride] \n\t"
PTR_ADDU "%[src2], %[src2], %[stride] \n\t"
PTR_ADDU "%[dst], %[dst], %[stride] \n\t"
PTR_ADDIU "%[tmp0], %[tmp0], -0x01 \n\t"
"bnez %[tmp0], 1b \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>
2021-07-23 07:53:42 +02:00
: RESTRICT_ASM_ALL64
[dst]"+&r"(dst), [src1]"+&r"(src1), [src2]"+&r"(src2),
avcodec/mips: [loongson] optimize theora decoding with mmi. Optimize theora decoding with mmi in functions: 1. ff_vp3_idct_add_mmi 2. ff_vp3_idct_put_mmi 3. ff_vp3_idct_dc_add_mmi 4. ff_put_no_rnd_pixels_l2_mmi Theora decoding speed improved about 32%(from 88fps to 116fps, Tested on loongson 3A3000). Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2019-02-12 12:55:48 +02:00
[ftmp1]"=&f"(ftmp[0]), [ftmp2]"=&f"(ftmp[1]), [ftmp3]"=&f"(ftmp[2]),
[ftmp4]"=&f"(ftmp[3]), [ftmp5]"=&f"(ftmp[4]), [ftmp6]"=&f"(ftmp[5]),
[tmp0]"=&r"(tmp[0]), [tmp1]"=&r"(tmp[1])
: [stride]"r"(stride)
: "memory"
);
} else {
int i;
for (i = 0; i < h; i++) {
uint32_t a, b;
a = AV_RN32(&src1[i * stride]);
b = AV_RN32(&src2[i * stride]);
AV_WN32A(&dst[i * stride], no_rnd_avg32(a, b));
a = AV_RN32(&src1[i * stride + 4]);
b = AV_RN32(&src2[i * stride + 4]);
AV_WN32A(&dst[i * stride + 4], no_rnd_avg32(a, b));
}
}
}
Reference in New Issue Copy Permalink