MrBesen/ffmpeg
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

avcodec/mips: MSA (MIPS-SIMD-Arch) optimizations for H263 lpf functions

Browse Source 
This patch adds MSA (MIPS-SIMD-Arch) optimizations for H263 lpf functions in new file h263dsp_msa.c

Signed-off-by: Shivraj Patil <shivraj.patil@imgtec.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
This commit is contained in:
Shivraj Patil 2015-06-14 23:26:22 +05:30 committed by Michael Niedermayer
parent bb42a7d4d4
commit 63eaf529bc
6 changed files with 241 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
libavcodec/h263dsp.c
View File

@ -121,4 +121,6 @@ av_cold void ff_h263dsp_init(H263DSPContext *ctx)
if (ARCH_X86)
ff_h263dsp_init_x86(ctx);
if (ARCH_MIPS)
ff_h263dsp_init_mips(ctx);
}

1
libavcodec/h263dsp.h
View File

@ -30,5 +30,6 @@ typedef struct H263DSPContext {
void ff_h263dsp_init(H263DSPContext *ctx);
void ff_h263dsp_init_x86(H263DSPContext *ctx);
void ff_h263dsp_init_mips(H263DSPContext *ctx);
#endif /* AVCODEC_H263DSP_H */

2
libavcodec/mips/Makefile
View File

@ -24,6 +24,7 @@ OBJS-$(CONFIG_H264DSP) += mips/h264dsp_init_mips.o
OBJS-$(CONFIG_H264QPEL) += mips/h264qpel_init_mips.o
OBJS-$(CONFIG_H264CHROMA) += mips/h264chroma_init_mips.o
OBJS-$(CONFIG_H264PRED) += mips/h264pred_init_mips.o
OBJS-$(CONFIG_H263DSP) += mips/h263dsp_init_mips.o
MSA-OBJS-$(CONFIG_HEVC_DECODER) += mips/hevcdsp_msa.o \
mips/hevc_mc_uni_msa.o \
mips/hevc_mc_uniw_msa.o \
@ -37,5 +38,6 @@ MSA-OBJS-$(CONFIG_H264DSP) += mips/h264dsp_msa.o \
MSA-OBJS-$(CONFIG_H264QPEL) += mips/h264qpel_msa.o
MSA-OBJS-$(CONFIG_H264CHROMA) += mips/h264chroma_msa.o
MSA-OBJS-$(CONFIG_H264PRED) += mips/h264pred_msa.o
MSA-OBJS-$(CONFIG_H263DSP) += mips/h263dsp_msa.o
LOONGSON3-OBJS-$(CONFIG_H264DSP) += mips/h264dsp_mmi.o
LOONGSON3-OBJS-$(CONFIG_H264CHROMA) += mips/h264chroma_mmi.o

36
libavcodec/mips/h263dsp_init_mips.c Normal file
View File

@ -0,0 +1,36 @@
/*
* Copyright (c) 2015 Manojkumar Bhosale (Manojkumar.Bhosale@imgtec.com)
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "h263dsp_mips.h"
#if HAVE_MSA
static av_cold void h263dsp_init_msa(H263DSPContext *c)
{
c->h263_h_loop_filter = ff_h263_h_loop_filter_msa;
c->h263_v_loop_filter = ff_h263_v_loop_filter_msa;
}
#endif // #if HAVE_MSA
av_cold void ff_h263dsp_init_mips(H263DSPContext *c)
{
#if HAVE_MSA
h263dsp_init_msa(c);
#endif // #if HAVE_MSA
}

36
libavcodec/mips/h263dsp_mips.h Normal file
View File

@ -0,0 +1,36 @@
/*
* Copyright (c) 2015 Manojkumar Bhosale (Manojkumar.Bhosale@imgtec.com)
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#ifndef AVCODEC_MIPS_H263DSP_MIPS_H
#define AVCODEC_MIPS_H263DSP_MIPS_H
#include "libavcodec/mpegvideo.h"
void ff_h263_h_loop_filter_msa(uint8_t *src, int stride, int q_scale);
void ff_h263_v_loop_filter_msa(uint8_t *src, int stride, int q_scale);
void ff_dct_unquantize_mpeg2_inter_msa(MpegEncContext *s, int16_t *block,
int32_t index, int32_t q_scale);
void ff_dct_unquantize_h263_inter_msa(MpegEncContext *s, int16_t *block,
int32_t index, int32_t q_scale);
void ff_dct_unquantize_h263_intra_msa(MpegEncContext *s, int16_t *block,
int32_t index, int32_t q_scale);
int ff_pix_sum_msa(uint8_t *pix, int line_size);
#endif // #ifndef AVCODEC_MIPS_H263DSP_MIPS_H

164
libavcodec/mips/h263dsp_msa.c Normal file
View File

@ -0,0 +1,164 @@
/*
* Copyright (c) 2015 Manojkumar Bhosale (Manojkumar.Bhosale@imgtec.com)
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "libavutil/mips/generic_macros_msa.h"
#include "h263dsp_mips.h"
static const uint8_t h263_loop_filter_strength_msa[32] = {
0, 1, 1, 2, 2, 3, 3, 4, 4, 4, 5, 5, 6, 6, 7, 7,
7, 8, 8, 8, 9, 9, 9, 10, 10, 10, 11, 11, 11, 12, 12, 12
};
static void h263_h_loop_filter_msa(uint8_t *src, int32_t stride, int32_t qscale)
{
int32_t strength = h263_loop_filter_strength_msa[qscale];
v16u8 in0, in1, in2, in3, in4, in5, in6, in7;
v8i16 temp0, temp1, temp2;
v8i16 diff0, diff2, diff4, diff6, diff8;
v8i16 d0, a_d0, str_x2, str;
src -= 2;
LD_UB8(src, stride, in0, in1, in2, in3, in4, in5, in6, in7);
TRANSPOSE8x4_UB_UB(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in3, in2, in1);
temp0 = (v8i16) __msa_ilvr_b((v16i8) in0, (v16i8) in1);
a_d0 = __msa_hsub_u_h((v16u8) temp0, (v16u8) temp0);
temp2 = (v8i16) __msa_ilvr_b((v16i8) in2, (v16i8) in3);
temp2 = __msa_hsub_u_h((v16u8) temp2, (v16u8) temp2);
temp2 <<= 2;
diff0 = a_d0 + temp2;
diff2 = -(-diff0 >> 3);
str_x2 = __msa_fill_h(-(strength << 1));
temp0 = (str_x2 <= diff2);
diff2 = (v8i16) __msa_bmz_v((v16u8) diff2, (v16u8) temp0, (v16u8) temp0);
temp2 = str_x2 - diff2;
str = __msa_fill_h(-strength);
temp0 = (diff2 < str);
diff2 = (v8i16) __msa_bmnz_v((v16u8) diff2, (v16u8) temp2, (v16u8) temp0);
diff4 = diff0 >> 3;
str_x2 = __msa_fill_h(strength << 1);
temp0 = (diff4 <= str_x2);
diff4 = (v8i16) __msa_bmz_v((v16u8) diff4, (v16u8) temp0, (v16u8) temp0);
temp2 = str_x2 - diff4;
str = __msa_fill_h(strength);
temp0 = (str < diff4);
diff4 = (v8i16) __msa_bmnz_v((v16u8) diff4, (v16u8) temp2, (v16u8) temp0);
temp0 = __msa_clti_s_h(diff0, 0);
d0 = (v8i16) __msa_bmnz_v((v16u8) diff4, (v16u8) diff2, (v16u8) temp0);
diff2 = -diff2 >> 1;
diff4 >>= 1;
diff8 = (v8i16) __msa_bmnz_v((v16u8) diff4, (v16u8) diff2, (v16u8) temp0);
diff6 = (-a_d0) >> 2;
diff6 = -(diff6);
temp2 = -diff8;
temp0 = (diff6 < temp2);
diff6 = (v8i16) __msa_bmnz_v((v16u8) diff6, (v16u8) temp2, (v16u8) temp0);
diff2 = a_d0 >> 2;
temp0 = (diff2 <= diff8);
diff2 = (v8i16) __msa_bmz_v((v16u8) diff2, (v16u8) diff8, (v16u8) temp0);
temp0 = __msa_clti_s_h(a_d0, 0);
diff6 = (v8i16) __msa_bmz_v((v16u8) diff6, (v16u8) diff2, (v16u8) temp0);
PCKEV_B2_SH(a_d0, diff6, a_d0, d0, diff6, d0);
in0 = (v16u8) ((v16i8) in0 - (v16i8) diff6);
in1 = (v16u8) ((v16i8) in1 + (v16i8) diff6);
in3 = __msa_xori_b(in3, 128);
in3 = (v16u8) __msa_adds_s_b((v16i8) in3, (v16i8) d0);
in3 = __msa_xori_b(in3, 128);
in2 = __msa_subsus_u_b(in2, (v16i8) d0);
ILVR_B2_SH(in3, in0, in1, in2, temp0, temp1);
in0 = (v16u8) __msa_ilvr_h(temp1, temp0);
in3 = (v16u8) __msa_ilvl_h(temp1, temp0);
ST4x4_UB(in0, in0, 0, 1, 2, 3, src, stride);
src += 4 * stride;
ST4x4_UB(in3, in3, 0, 1, 2, 3, src, stride);
src += 4 * stride;
}
static void h263_v_loop_filter_msa(uint8_t *src, int32_t stride, int32_t qscale)
{
int32_t strength = h263_loop_filter_strength_msa[qscale];
uint64_t res0, res1, res2, res3;
v16u8 in0, in1, in2, in3;
v8i16 temp0, temp2, diff0, diff2, diff4, diff6, diff8;
v8i16 d0, a_d0, str_x2, str;
src -= 2 * stride;
LD_UB4(src, stride, in0, in3, in2, in1);
temp0 = (v8i16) __msa_ilvr_b((v16i8) in0, (v16i8) in1);
a_d0 = __msa_hsub_u_h((v16u8) temp0, (v16u8) temp0);
temp2 = (v8i16) __msa_ilvr_b((v16i8) in2, (v16i8) in3);
temp2 = __msa_hsub_u_h((v16u8) temp2, (v16u8) temp2);
temp2 <<= 2;
diff0 = a_d0 + temp2;
diff2 = -(-diff0 >> 3);
str_x2 = __msa_fill_h(-(strength << 1));
temp0 = (str_x2 <= diff2);
diff2 = (v8i16) __msa_bmz_v((v16u8) diff2, (v16u8) temp0, (v16u8) temp0);
temp2 = str_x2 - diff2;
str = __msa_fill_h(-strength);
temp0 = (diff2 < str);
diff2 = (v8i16) __msa_bmnz_v((v16u8) diff2, (v16u8) temp2, (v16u8) temp0);
diff4 = diff0 >> 3;
str_x2 = __msa_fill_h(strength << 1);
temp0 = (diff4 <= str_x2);
diff4 = (v8i16) __msa_bmz_v((v16u8) diff4, (v16u8) temp0, (v16u8) temp0);
temp2 = str_x2 - diff4;
str = __msa_fill_h(strength);
temp0 = (str < diff4);
diff4 = (v8i16) __msa_bmnz_v((v16u8) diff4, (v16u8) temp2, (v16u8) temp0);
temp0 = __msa_clti_s_h(diff0, 0);
d0 = (v8i16) __msa_bmnz_v((v16u8) diff4, (v16u8) diff2, (v16u8) temp0);
diff2 = -diff2 >> 1;
diff4 >>= 1;
diff8 = (v8i16) __msa_bmnz_v((v16u8) diff4, (v16u8) diff2, (v16u8) temp0);
diff6 = (-a_d0) >> 2;
diff6 = -(diff6);
temp2 = -diff8;
temp0 = (diff6 < temp2);
diff6 = (v8i16) __msa_bmnz_v((v16u8) diff6, (v16u8) temp2, (v16u8) temp0);
diff2 = a_d0 >> 2;
temp0 = (diff2 <= diff8);
diff2 = (v8i16) __msa_bmz_v((v16u8) diff2, (v16u8) diff8, (v16u8) temp0);
temp0 = __msa_clti_s_h(a_d0, 0);
diff6 = (v8i16) __msa_bmz_v((v16u8) diff6, (v16u8) diff2, (v16u8) temp0);
PCKEV_B2_SH(a_d0, diff6, a_d0, d0, diff6, d0);
in0 = (v16u8) ((v16i8) in0 - (v16i8) diff6);
in1 = (v16u8) ((v16i8) in1 + (v16i8) diff6);
in3 = __msa_xori_b(in3, 128);
in3 = (v16u8) __msa_adds_s_b((v16i8) in3, (v16i8) d0);
in3 = __msa_xori_b(in3, 128);
in2 = __msa_subsus_u_b(in2, (v16i8) d0);
res0 = __msa_copy_u_d((v2i64) in0, 0);
res1 = __msa_copy_u_d((v2i64) in3, 0);
res2 = __msa_copy_u_d((v2i64) in2, 0);
res3 = __msa_copy_u_d((v2i64) in1, 0);
SD4(res0, res1, res2, res3, src, stride);
}
void ff_h263_h_loop_filter_msa(uint8_t *src, int32_t stride, int32_t q_scale)
{
h263_h_loop_filter_msa(src, stride, q_scale);
}
void ff_h263_v_loop_filter_msa(uint8_t *src, int32_t stride, int32_t q_scale)
{
h263_v_loop_filter_msa(src, stride, q_scale);
}