On 8/16/2020 12:01 PM, Paul B Mahol wrote: > On 8/16/20, James Almer <jamr...@gmail.com> wrote: >> On 8/16/2020 11:09 AM, Paul B Mahol wrote: >>> On 8/16/20, Paul B Mahol <one...@gmail.com> wrote: >>>> Hi, >>>> >>>> patch attached. >>>> >>>> Please help porting this to linux and 64bit calling convention. >>>> >>> >>> New patch attached, could build on x64, please report any build failure. >> >> [...] >> >>> diff --git a/libavcodec/x86/cfhddsp.asm b/libavcodec/x86/cfhddsp.asm >>> new file mode 100644 >>> index 0000000000..80371e65c9 >>> --- /dev/null >>> +++ b/libavcodec/x86/cfhddsp.asm >>> @@ -0,0 +1,626 @@ >>> +;****************************************************************************** >>> +;* x86-optimized functions for the CFHD decoder >>> +;* Copyright (c) 2020 Paul B Mahol >>> +;* >>> +;* 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/x86/x86util.asm" >>> + >>> +SECTION_RODATA >>> + >>> +factor_p1_p1: dw 1, 1, 1, 1, 1, 1, 1, 1, >>> +factor_p1_n1: dw 1, -1, 1, -1, 1, -1, 1, -1, >>> +factor_n1_p1: dw -1, 1, -1, 1, -1, 1, -1, 1, >>> +factor_p11_n4: dw 11, -4, 11, -4, 11, -4, 11, -4, >>> +factor_p5_p4: dw 5, 4, 5, 4, 5, 4, 5, 4, >>> +pd_4: times 4 dd 4 >>> +pw_0: times 8 dw 0 >>> +pw_1023: times 8 dw 1023 >>> +pw_4095: times 8 dw 4095 >>> + >>> +SECTION .text >>> + >>> +%macro CFHD_HORIZ_FILTER 1 >>> +%if %1 == 1023 >>> +cglobal cfhd_horiz_filter_clip10, 5, 6, 8, output, low, high, width, bpc >>> + DEFINE_ARGS output, low, high, width, x, temp >>> + shl widthd, 1 >>> +%define ostrideq widthq >>> +%define lwidthq widthq >>> +%define hwidthq widthq >>> +%elif %1 == 4095 >>> +cglobal cfhd_horiz_filter_clip12, 5, 6, 8, output, low, high, width, bpc >>> + DEFINE_ARGS output, low, high, width, x, temp >>> + shl widthd, 1 >>> +%define ostrideq widthq >>> +%define lwidthq widthq >>> +%define hwidthq widthq >>> +%else >>> +%if ARCH_X86_64 >>> +cglobal cfhd_horiz_filter, 11, 11, 8, output, ostride, low, lwidth, high, >>> hwidth, width, height >>> +DEFINE_ARGS output, ostride, low, lwidth, high, hwidth, width, height, >>> x, y, temp >>> + shl ostrided, 1 >>> + shl lwidthd, 1 >>> + shl hwidthd, 1 >>> + shl widthd, 1 >>> + >>> + mov yq, heightq >>> + neg yq >>> +%else >>> +cglobal cfhd_horiz_filter, 6, 6, 8, 64, output, x, low, y, high, temp, >>> width, height >>> + shl xd, 1 >>> + shl yd, 1 >>> + shl tempd, 1 >>> + >>> + mov dword [rsp + 0], xq >>> + mov dword [rsp + 8], yq >>> + mov dword [rsp + 16], tempq >> >> These are four bytes on x86_32, not eight. Also, since all arguments >> come from stack, you can simply move them back doing >> >> mov xmp, xq >> mov ymp, yq >> mov tempmp, tempq >> %define ostrideq xm >> %define lwidthq ym >> %define hwidthq tempm >> >> Saving you the need to reserve space. >> >>> + >>> + mov yd, r6m >> >> Just load r6/width normally in cglobal, you can use up to seven regs on >> x86_32. > > But I need 11 regs, thus I use stack.
You can use the stack space already used for the input arguments without the need to reserve more, as i explained above. > >> >>> + shl yd, 1 >>> + mov dword [rsp + 24], yq >>> + >>> + mov yd, r7m >>> + neg yq >>> + >>> +%define ostrideq [rsp + 0] >>> +%define lwidthq [rsp + 8] >>> +%define hwidthq [rsp + 16] >>> +%define widthq [rsp + 24] >> >> If you're going to define widthq here like this, then you shouldn't >> define width in cglobal. But as i said above, you have a reg free to >> store it. >> _______________________________________________ >> ffmpeg-devel mailing list >> ffmpeg-devel@ffmpeg.org >> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel >> >> To unsubscribe, visit link above, or email >> ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe". > _______________________________________________ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-devel > > To unsubscribe, visit link above, or email > ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe". > _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
