virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-12-28 20:53:54 +02:00
Code Issues Releases Activity
d61386a690
FFmpeg/libavcodec/x86/flac_dsp_gpl.asm

102 lines
2.8 KiB
NASM
Raw Normal View History

lavc/flacenc: add sse4 version of the 16-bit lpc encoder From 1.8 to 2.4 times faster. Runtime is reduced by 2 to 39%. The speed-up generally increases with compression_level. This lpc encoder is not used with levels < 3 so it provides no speed-up in these cases. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:02 +03:00
;******************************************************************************
;* FLAC DSP functions
;*
;* Copyright (c) 2014 James Darnley <james.darnley@gmail.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 General Public License as published by
;* the Free Software Foundation; either version 2 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 General Public License for more details.
;*
;* You should have received a copy of the GNU 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/x86/x86util.asm"
SECTION_TEXT
INIT_XMM sse4
%if ARCH_X86_64
lavc/flacenc: partially unroll loop in flac_enc_lpc_16 It now does 12 samples per iteration, up from 4. From 1.8 to 3.2 times faster again. 3.6 to 5.7 times faster overall. Runtime is reduced by a further 2 to 18%. Overall runtime reduced by 4 to 50%. Same conditions as before apply. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:03 +03:00
cglobal flac_enc_lpc_16, 5, 7, 8, 0, res, smp, len, order, coefs
lavc/flacenc: add sse4 version of the 16-bit lpc encoder From 1.8 to 2.4 times faster. Runtime is reduced by 2 to 39%. The speed-up generally increases with compression_level. This lpc encoder is not used with levels < 3 so it provides no speed-up in these cases. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:02 +03:00
DECLARE_REG_TMP 5, 6
%define length r2d
movsxd orderq, orderd
%else
lavc/flacenc: partially unroll loop in flac_enc_lpc_16 It now does 12 samples per iteration, up from 4. From 1.8 to 3.2 times faster again. 3.6 to 5.7 times faster overall. Runtime is reduced by a further 2 to 18%. Overall runtime reduced by 4 to 50%. Same conditions as before apply. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:03 +03:00
cglobal flac_enc_lpc_16, 5, 6, 8, 0, res, smp, len, order, coefs
lavc/flacenc: add sse4 version of the 16-bit lpc encoder From 1.8 to 2.4 times faster. Runtime is reduced by 2 to 39%. The speed-up generally increases with compression_level. This lpc encoder is not used with levels < 3 so it provides no speed-up in these cases. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:02 +03:00
DECLARE_REG_TMP 2, 5
%define length r2mp
%endif
; Here we assume that the maximum order value is 32. This means that we only
; need to copy a maximum of 32 samples. Therefore we let the preprocessor
; unroll this loop and copy all 32.
%assign iter 0
%rep 32/(mmsize/4)
movu m0, [smpq+iter]
movu [resq+iter], m0
%assign iter iter+mmsize
%endrep
lea resq, [resq+orderq*4]
lea smpq, [smpq+orderq*4]
lea coefsq, [coefsq+orderq*4]
sub length, orderd
movd m3, r5m
neg orderq
%define posj t0q
%define negj t1q
.looplen:
pxor m0, m0
lavc/flacenc: partially unroll loop in flac_enc_lpc_16 It now does 12 samples per iteration, up from 4. From 1.8 to 3.2 times faster again. 3.6 to 5.7 times faster overall. Runtime is reduced by a further 2 to 18%. Overall runtime reduced by 4 to 50%. Same conditions as before apply. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:03 +03:00
pxor m4, m4
pxor m6, m6
lavc/flacenc: add sse4 version of the 16-bit lpc encoder From 1.8 to 2.4 times faster. Runtime is reduced by 2 to 39%. The speed-up generally increases with compression_level. This lpc encoder is not used with levels < 3 so it provides no speed-up in these cases. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:02 +03:00
mov posj, orderq
xor negj, negj
.looporder:
movd m2, [coefsq+posj*4] ; c = coefs[j]
SPLATD m2
movu m1, [smpq+negj*4-4] ; s = smp[i-j-1]
lavc/flacenc: partially unroll loop in flac_enc_lpc_16 It now does 12 samples per iteration, up from 4. From 1.8 to 3.2 times faster again. 3.6 to 5.7 times faster overall. Runtime is reduced by a further 2 to 18%. Overall runtime reduced by 4 to 50%. Same conditions as before apply. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:03 +03:00
movu m5, [smpq+negj*4-4+mmsize]
movu m7, [smpq+negj*4-4+mmsize*2]
lavc/flacenc: add sse4 version of the 16-bit lpc encoder From 1.8 to 2.4 times faster. Runtime is reduced by 2 to 39%. The speed-up generally increases with compression_level. This lpc encoder is not used with levels < 3 so it provides no speed-up in these cases. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:02 +03:00
pmulld m1, m2
lavc/flacenc: partially unroll loop in flac_enc_lpc_16 It now does 12 samples per iteration, up from 4. From 1.8 to 3.2 times faster again. 3.6 to 5.7 times faster overall. Runtime is reduced by a further 2 to 18%. Overall runtime reduced by 4 to 50%. Same conditions as before apply. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:03 +03:00
pmulld m5, m2
pmulld m7, m2
lavc/flacenc: add sse4 version of the 16-bit lpc encoder From 1.8 to 2.4 times faster. Runtime is reduced by 2 to 39%. The speed-up generally increases with compression_level. This lpc encoder is not used with levels < 3 so it provides no speed-up in these cases. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:02 +03:00
paddd m0, m1 ; p += c * s
lavc/flacenc: partially unroll loop in flac_enc_lpc_16 It now does 12 samples per iteration, up from 4. From 1.8 to 3.2 times faster again. 3.6 to 5.7 times faster overall. Runtime is reduced by a further 2 to 18%. Overall runtime reduced by 4 to 50%. Same conditions as before apply. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:03 +03:00
paddd m4, m5
paddd m6, m7
lavc/flacenc: add sse4 version of the 16-bit lpc encoder From 1.8 to 2.4 times faster. Runtime is reduced by 2 to 39%. The speed-up generally increases with compression_level. This lpc encoder is not used with levels < 3 so it provides no speed-up in these cases. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:02 +03:00
dec negj
inc posj
jnz .looporder
psrad m0, m3 ; p >>= shift
lavc/flacenc: partially unroll loop in flac_enc_lpc_16 It now does 12 samples per iteration, up from 4. From 1.8 to 3.2 times faster again. 3.6 to 5.7 times faster overall. Runtime is reduced by a further 2 to 18%. Overall runtime reduced by 4 to 50%. Same conditions as before apply. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:03 +03:00
psrad m4, m3
psrad m6, m3
lavc/flacenc: add sse4 version of the 16-bit lpc encoder From 1.8 to 2.4 times faster. Runtime is reduced by 2 to 39%. The speed-up generally increases with compression_level. This lpc encoder is not used with levels < 3 so it provides no speed-up in these cases. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:02 +03:00
movu m1, [smpq]
lavc/flacenc: partially unroll loop in flac_enc_lpc_16 It now does 12 samples per iteration, up from 4. From 1.8 to 3.2 times faster again. 3.6 to 5.7 times faster overall. Runtime is reduced by a further 2 to 18%. Overall runtime reduced by 4 to 50%. Same conditions as before apply. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:03 +03:00
movu m5, [smpq+mmsize]
movu m7, [smpq+mmsize*2]
lavc/flacenc: add sse4 version of the 16-bit lpc encoder From 1.8 to 2.4 times faster. Runtime is reduced by 2 to 39%. The speed-up generally increases with compression_level. This lpc encoder is not used with levels < 3 so it provides no speed-up in these cases. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:02 +03:00
psubd m1, m0 ; smp[i] - p
lavc/flacenc: partially unroll loop in flac_enc_lpc_16 It now does 12 samples per iteration, up from 4. From 1.8 to 3.2 times faster again. 3.6 to 5.7 times faster overall. Runtime is reduced by a further 2 to 18%. Overall runtime reduced by 4 to 50%. Same conditions as before apply. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:03 +03:00
psubd m5, m4
psubd m7, m6
lavc/flacenc: add sse4 version of the 16-bit lpc encoder From 1.8 to 2.4 times faster. Runtime is reduced by 2 to 39%. The speed-up generally increases with compression_level. This lpc encoder is not used with levels < 3 so it provides no speed-up in these cases. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:02 +03:00
movu [resq], m1 ; res[i] = smp[i] - (p >> shift)
lavc/flacenc: partially unroll loop in flac_enc_lpc_16 It now does 12 samples per iteration, up from 4. From 1.8 to 3.2 times faster again. 3.6 to 5.7 times faster overall. Runtime is reduced by a further 2 to 18%. Overall runtime reduced by 4 to 50%. Same conditions as before apply. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:03 +03:00
movu [resq+mmsize], m5
movu [resq+mmsize*2], m7
lavc/flacenc: add sse4 version of the 16-bit lpc encoder From 1.8 to 2.4 times faster. Runtime is reduced by 2 to 39%. The speed-up generally increases with compression_level. This lpc encoder is not used with levels < 3 so it provides no speed-up in these cases. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:02 +03:00
lavc/flacenc: partially unroll loop in flac_enc_lpc_16 It now does 12 samples per iteration, up from 4. From 1.8 to 3.2 times faster again. 3.6 to 5.7 times faster overall. Runtime is reduced by a further 2 to 18%. Overall runtime reduced by 4 to 50%. Same conditions as before apply. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:03 +03:00
add resq, 3*mmsize
add smpq, 3*mmsize
sub length, (3*mmsize)/4
lavc/flacenc: add sse4 version of the 16-bit lpc encoder From 1.8 to 2.4 times faster. Runtime is reduced by 2 to 39%. The speed-up generally increases with compression_level. This lpc encoder is not used with levels < 3 so it provides no speed-up in these cases. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-08-13 00:22:02 +03:00
jg .looplen
RET
Reference in New Issue Copy Permalink