generic-library/ffmpeg
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
b4af7d68fe
ffmpeg/libavcodec/x86/proresdsp.asm

67 lines
1.8 KiB
NASM
Raw Normal View History

prores: idct sse2/sse4 optimizations. ~3.0-3.5x as fast as original C version, 1.6x as fast overall.
2011-09-30 14:37:11 +02:00
;******************************************************************************
;* x86-SIMD-optimized IDCT for prores
proresdsp: Correct credits to point to the Author and not just the code this is based on. Also change Libav to FFmpeg Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2011-10-12 22:09:40 +02:00
;* this is identical to "simple" IDCT written by Michael Niedermayer
;* except for the clip range
prores: idct sse2/sse4 optimizations. ~3.0-3.5x as fast as original C version, 1.6x as fast overall.
2011-09-30 14:37:11 +02:00
;*
;* Copyright (c) 2011 Ronald S. Bultje <rsbultje@gmail.com>
;*
proresdsp: Correct credits to point to the Author and not just the code this is based on. Also change Libav to FFmpeg Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2011-10-12 22:09:40 +02:00
;* This file is part of FFmpeg.
prores: idct sse2/sse4 optimizations. ~3.0-3.5x as fast as original C version, 1.6x as fast overall.
2011-09-30 14:37:11 +02:00
;*
proresdsp: Correct credits to point to the Author and not just the code this is based on. Also change Libav to FFmpeg Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2011-10-12 22:09:40 +02:00
;* FFmpeg is free software; you can redistribute it and/or
prores: idct sse2/sse4 optimizations. ~3.0-3.5x as fast as original C version, 1.6x as fast overall.
2011-09-30 14:37:11 +02:00
;* 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.
;*
proresdsp: Correct credits to point to the Author and not just the code this is based on. Also change Libav to FFmpeg Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2011-10-12 22:09:40 +02:00
;* FFmpeg is distributed in the hope that it will be useful,
prores: idct sse2/sse4 optimizations. ~3.0-3.5x as fast as original C version, 1.6x as fast overall.
2011-09-30 14:37:11 +02:00
;* 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
proresdsp: Correct credits to point to the Author and not just the code this is based on. Also change Libav to FFmpeg Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2011-10-12 22:09:40 +02:00
;* License along with FFmpeg; if not, write to the Free Software
prores: idct sse2/sse4 optimizations. ~3.0-3.5x as fast as original C version, 1.6x as fast overall.
2011-09-30 14:37:11 +02:00
;* 51, Inc., Foundation Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
;******************************************************************************
x86: yasm: Use complete source path for macro helper %includes This is more consistent with the way we handle C #includes and it simplifies the build system.
2012-07-15 11:48:21 +02:00
%include "libavutil/x86/x86util.asm"
prores: idct sse2/sse4 optimizations. ~3.0-3.5x as fast as original C version, 1.6x as fast overall.
2011-09-30 14:37:11 +02:00
config.asm: change %ifdef directives to %if directives. This allows combining multiple conditionals in a single statement.
2012-01-23 10:45:58 +01:00
%if ARCH_X86_64
prores: idct sse2/sse4 optimizations. ~3.0-3.5x as fast as original C version, 1.6x as fast overall.
2011-09-30 14:37:11 +02:00
SECTION_RODATA
proresdsp: Optimize series of padds out Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2011-10-12 21:59:44 +02:00
pw_88: times 8 dw 0x2008
proresdsp: fix rounding Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2011-10-12 22:18:35 +02:00
cextern pw_1
prores: idct sse2/sse4 optimizations. ~3.0-3.5x as fast as original C version, 1.6x as fast overall.
2011-09-30 14:37:11 +02:00
cextern pw_4
cextern pw_1019
x86: simple_idct10_template: use const This avoid going through constants.c while still sharing them with proresdsp.asm Reviewed-by: "Ronald S. Bultje" <rsbultje@gmail.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2015-10-13 21:21:39 +02:00
; Below are defined in simple_idct10.asm built from selecting idctdsp
cextern w4_plus_w2
cextern w4_min_w2
cextern w4_plus_w6
cextern w4_min_w6
cextern w1_plus_w3
cextern w3_min_w1
cextern w7_plus_w3
cextern w3_min_w7
cextern w1_plus_w5
cextern w5_min_w1
cextern w5_plus_w7
cextern w7_min_w5
prores: idct sse2/sse4 optimizations. ~3.0-3.5x as fast as original C version, 1.6x as fast overall.
2011-09-30 14:37:11 +02:00
x86: prores: templatize 10 bits simple_idct This should be reused for a generic simple_idct10 function. Requires a bit of trickery to declare common constants in C. Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2015-10-12 19:37:44 +02:00
%include "libavcodec/x86/simple_idct10_template.asm"
prores: idct sse2/sse4 optimizations. ~3.0-3.5x as fast as original C version, 1.6x as fast overall.
2011-09-30 14:37:11 +02:00
x86inc: Add debug symbols indicating sizes of compiled functions Some debuggers/profilers use this metadata to determine which function a given instruction is in; without it they get can confused by local labels (if you haven't stripped those). On the other hand, some tools are still confused even with this metadata. e.g. this fixes `gdb`, but not `perf`. Currently only implemented for ELF.
2015-10-12 14:13:00 +02:00
SECTION .text
prores: idct sse2/sse4 optimizations. ~3.0-3.5x as fast as original C version, 1.6x as fast overall.
2011-09-30 14:37:11 +02:00
x86: simple_idct(_put): 10bits versions Modeled from the prores version. Clips to [0;1023] and is bitexact. Bitexactness requires to add offsets in different places compared to prores or C, and makes the function approximately 2% slower. For 16 frames of a DNxHD 4:2:2 10bits test sequence: C: 60861 decicycles in idct, 1048205 runs, 371 skips sse2: 27567 decicycles in idct, 1048216 runs, 360 skips avx: 26272 decicycles in idct, 1048171 runs, 405 skips The add version is not implemented, so the corresponding dsp function is set to NULL to make it clear in a code executing it. Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2015-10-12 19:37:47 +02:00
%macro idct_fn 0
x86: simple_idct10_template: fix overflow in pass When the input of a pass has 15 or 16 bits of precision (in particular the column pass), the addition of a bias to W4 may lead to overflows in the input to pmaddwd. This requires postponing the adding of the bias to after the first butterfly. To do so, the fact that m15, unused although zeroed, is exploited. In case the pass is safe, an address can be directly used, and the number of xmm regs can be decreased. Otherwise, the 32bits bias is loaded into it. Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2015-10-12 19:37:46 +02:00
cglobal prores_idct_put_10, 4, 4, 15
x86: simple_idct(_put): 10bits versions Modeled from the prores version. Clips to [0;1023] and is bitexact. Bitexactness requires to add offsets in different places compared to prores or C, and makes the function approximately 2% slower. For 16 frames of a DNxHD 4:2:2 10bits test sequence: C: 60861 decicycles in idct, 1048205 runs, 371 skips sse2: 27567 decicycles in idct, 1048216 runs, 360 skips avx: 26272 decicycles in idct, 1048171 runs, 405 skips The add version is not implemented, so the corresponding dsp function is set to NULL to make it clear in a code executing it. Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2015-10-12 19:37:47 +02:00
IDCT_FN pw_1, 15, pw_88, 18, pw_4, pw_1019, r3
prores: idct sse2/sse4 optimizations. ~3.0-3.5x as fast as original C version, 1.6x as fast overall.
2011-09-30 14:37:11 +02:00
RET
%endmacro
proresdsp: port x86 assembly to cpuflags.
2012-07-27 05:38:27 +02:00
INIT_XMM sse2
x86: simple_idct(_put): 10bits versions Modeled from the prores version. Clips to [0;1023] and is bitexact. Bitexactness requires to add offsets in different places compared to prores or C, and makes the function approximately 2% slower. For 16 frames of a DNxHD 4:2:2 10bits test sequence: C: 60861 decicycles in idct, 1048205 runs, 371 skips sse2: 27567 decicycles in idct, 1048216 runs, 360 skips avx: 26272 decicycles in idct, 1048171 runs, 405 skips The add version is not implemented, so the corresponding dsp function is set to NULL to make it clear in a code executing it. Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2015-10-12 19:37:47 +02:00
idct_fn
Add some missing _EXTERNAL suffixes to yasm source files.
2012-08-31 15:39:03 +02:00
%if HAVE_AVX_EXTERNAL
proresdsp: port x86 assembly to cpuflags.
2012-07-27 05:38:27 +02:00
INIT_XMM avx
x86: simple_idct(_put): 10bits versions Modeled from the prores version. Clips to [0;1023] and is bitexact. Bitexactness requires to add offsets in different places compared to prores or C, and makes the function approximately 2% slower. For 16 frames of a DNxHD 4:2:2 10bits test sequence: C: 60861 decicycles in idct, 1048205 runs, 371 skips sse2: 27567 decicycles in idct, 1048216 runs, 360 skips avx: 26272 decicycles in idct, 1048171 runs, 405 skips The add version is not implemented, so the corresponding dsp function is set to NULL to make it clear in a code executing it. Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2015-10-12 19:37:47 +02:00
idct_fn
Fix compilation with yasm-0.6.2
2012-01-12 16:35:49 +01:00
%endif
prores: idct sse2/sse4 optimizations. ~3.0-3.5x as fast as original C version, 1.6x as fast overall.
2011-09-30 14:37:11 +02:00
%endif
Reference in New Issue Copy Permalink