RepoMirrors
/
ffmpeg
mirror of https://git.ffmpeg.org/ffmpeg.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity
ffmpeg/libavcodec/x86/rv34dsp.asm

119 lines
3.0 KiB
NASM
Raw Normal View History

rv34: DC-only inverse transform When decoding coefficients, detect whether the block is DC-only, and take advantage of this knowledge to perform DC-only inverse transform. This is achieved by: - first, changing the 108x4 element modulo_three_table into a 108 element table (kind of base4), and accessing each value using mask and shifts. - then, checking low bits for 0 (as they represent the presence of higher frequency coefficients) Also provide x86 SIMD code for the DC-only inverse transform. Signed-off-by: Kostya Shishkov <kostya.shishkov@gmail.com>
2012-01-01 17:33:22 +00:00
;******************************************************************************
;* MMX/SSE2-optimized functions for the RV30 and RV40 decoders
;* Copyright (C) 2012 Christophe Gisquet <christophe.gisquet@gmail.com>
;*
;* This file is part of Libav.
;*
;* Libav 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.
;*
;* Libav 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 Libav; if not, write to the Free Software
;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
;******************************************************************************
%include "x86inc.asm"
%include "x86util.asm"
SECTION .text
%macro IDCT_DC_NOROUND 1
imul %1, 13*13*3
sar %1, 11
%endmacro
%macro IDCT_DC_ROUND 1
imul %1, 13*13
add %1, 0x200
sar %1, 10
%endmacro
rv34: Intra 16x16 handling Extract processing of intra 16x16 blocks from intra macroblock processing. Also implement a function performing inverse transform and block reconstruction for DC-only blocks in 1 pass instead of 2.
2012-01-02 23:22:11 +00:00
%macro rv34_idct 1
cglobal rv34_idct_%1_mmx2, 1, 2, 0
rv34: DC-only inverse transform When decoding coefficients, detect whether the block is DC-only, and take advantage of this knowledge to perform DC-only inverse transform. This is achieved by: - first, changing the 108x4 element modulo_three_table into a 108 element table (kind of base4), and accessing each value using mask and shifts. - then, checking low bits for 0 (as they represent the presence of higher frequency coefficients) Also provide x86 SIMD code for the DC-only inverse transform. Signed-off-by: Kostya Shishkov <kostya.shishkov@gmail.com>
2012-01-01 17:33:22 +00:00
movsx r1, word [r0]
IDCT_DC r1
Merge remote-tracking branch 'qatar/master' * qatar/master: rv34: add NEON rv34_idct_add rv34: 1-pass inter MB reconstruction add SMJPEG muxer avformat: split out common SMJPEG code pictordec: Use bytestream2 functions avconv: use avcodec_encode_audio2() pcmenc: use AVCodec.encode2() avcodec: bump minor version and add APIChanges for the new audio encoding API avcodec: Add avcodec_encode_audio2() as replacement for avcodec_encode_audio() avcodec: add a public function, avcodec_fill_audio_frame(). rv34: Intra 16x16 handling rv34: Inter/intra MB code split Conflicts: Changelog libavcodec/avcodec.h libavcodec/pictordec.c libavcodec/utils.c libavcodec/version.h libavcodec/x86/rv34dsp.asm libavformat/version.h Merged-by: Michael Niedermayer <michaelni@gmx.at>
2012-01-17 00:40:45 +00:00
movd m0, r1d
rv34: Intra 16x16 handling Extract processing of intra 16x16 blocks from intra macroblock processing. Also implement a function performing inverse transform and block reconstruction for DC-only blocks in 1 pass instead of 2.
2012-01-02 23:22:11 +00:00
pshufw m0, m0, 0
movq [r0+ 0], m0
rv34: 1-pass inter MB reconstruction Implement 1-pass inverse transform and reconstruction for inter blocks.
2012-01-03 19:38:29 +00:00
movq [r0+ 8], m0
rv34: Intra 16x16 handling Extract processing of intra 16x16 blocks from intra macroblock processing. Also implement a function performing inverse transform and block reconstruction for DC-only blocks in 1 pass instead of 2.
2012-01-02 23:22:11 +00:00
movq [r0+16], m0
rv34: 1-pass inter MB reconstruction Implement 1-pass inverse transform and reconstruction for inter blocks.
2012-01-03 19:38:29 +00:00
movq [r0+24], m0
rv34: DC-only inverse transform When decoding coefficients, detect whether the block is DC-only, and take advantage of this knowledge to perform DC-only inverse transform. This is achieved by: - first, changing the 108x4 element modulo_three_table into a 108 element table (kind of base4), and accessing each value using mask and shifts. - then, checking low bits for 0 (as they represent the presence of higher frequency coefficients) Also provide x86 SIMD code for the DC-only inverse transform. Signed-off-by: Kostya Shishkov <kostya.shishkov@gmail.com>
2012-01-01 17:33:22 +00:00
REP_RET
%endmacro
INIT_MMX
%define IDCT_DC IDCT_DC_ROUND
rv34: Intra 16x16 handling Extract processing of intra 16x16 blocks from intra macroblock processing. Also implement a function performing inverse transform and block reconstruction for DC-only blocks in 1 pass instead of 2.
2012-01-02 23:22:11 +00:00
rv34_idct dc
rv34: DC-only inverse transform When decoding coefficients, detect whether the block is DC-only, and take advantage of this knowledge to perform DC-only inverse transform. This is achieved by: - first, changing the 108x4 element modulo_three_table into a 108 element table (kind of base4), and accessing each value using mask and shifts. - then, checking low bits for 0 (as they represent the presence of higher frequency coefficients) Also provide x86 SIMD code for the DC-only inverse transform. Signed-off-by: Kostya Shishkov <kostya.shishkov@gmail.com>
2012-01-01 17:33:22 +00:00
%define IDCT_DC IDCT_DC_NOROUND
rv34: Intra 16x16 handling Extract processing of intra 16x16 blocks from intra macroblock processing. Also implement a function performing inverse transform and block reconstruction for DC-only blocks in 1 pass instead of 2.
2012-01-02 23:22:11 +00:00
rv34_idct dc_noround
; ff_rv34_idct_dc_add_mmx(uint8_t *dst, int stride, int dc);
cglobal rv34_idct_dc_add_mmx, 3, 3
; calculate DC
IDCT_DC_ROUND r2
pxor m1, m1
Use correct register size. Fixes compilation with NASM. Signed-off-by: Reimar Döffinger <Reimar.Doeffinger@gmx.de>
2012-01-17 07:41:04 +00:00
movd m0, r2d
rv34: Intra 16x16 handling Extract processing of intra 16x16 blocks from intra macroblock processing. Also implement a function performing inverse transform and block reconstruction for DC-only blocks in 1 pass instead of 2.
2012-01-02 23:22:11 +00:00
psubw m1, m0
packuswb m0, m0
packuswb m1, m1
punpcklbw m0, m0
punpcklbw m1, m1
punpcklwd m0, m0
punpcklwd m1, m1
; add DC
lea r2, [r0+r1*2]
movh m2, [r0]
movh m3, [r0+r1]
movh m4, [r2]
movh m5, [r2+r1]
paddusb m2, m0
paddusb m3, m0
paddusb m4, m0
paddusb m5, m0
psubusb m2, m1
psubusb m3, m1
psubusb m4, m1
psubusb m5, m1
movh [r0], m2
movh [r0+r1], m3
movh [r2], m4
movh [r2+r1], m5
RET
; ff_rv34_idct_dc_add_sse4(uint8_t *dst, int stride, int dc);
INIT_XMM
cglobal rv34_idct_dc_add_sse4, 3, 3, 6
; load data
IDCT_DC_ROUND r2
pxor m1, m1
; calculate DC
Use correct register size. Fixes compilation with NASM. Signed-off-by: Reimar Döffinger <Reimar.Doeffinger@gmx.de>
2012-01-17 07:41:04 +00:00
movd m0, r2d
rv34: Intra 16x16 handling Extract processing of intra 16x16 blocks from intra macroblock processing. Also implement a function performing inverse transform and block reconstruction for DC-only blocks in 1 pass instead of 2.
2012-01-02 23:22:11 +00:00
lea r2, [r0+r1*2]
movd m2, [r0]
movd m3, [r0+r1]
pshuflw m0, m0, 0
movd m4, [r2]
movd m5, [r2+r1]
punpcklqdq m0, m0
punpckldq m2, m3
punpckldq m4, m5
punpcklbw m2, m1
punpcklbw m4, m1
paddw m2, m0
paddw m4, m0
packuswb m2, m4
movd [r0], m2
pextrd [r0+r1], m2, 1
pextrd [r2], m2, 2
pextrd [r2+r1], m2, 3
RET