From: "Rémi Denis-Courmont" <remi@remlab.net> To: ffmpeg-devel@ffmpeg.org Subject: Re: [FFmpeg-devel] [PATCH 3/3] lavc/audiodsp: add RISC-V F float vector clip Date: Wed, 14 Sep 2022 20:56:52 +0300 Message-ID: <4740913.GXAFRqVoOG@basile.remlab.net> (raw) In-Reply-To: <20220914175031.162194-3-remi@remlab.net> Le keskiviikkona 14. syyskuuta 2022, 20.50.31 EEST remi@remlab.net a écrit : > From: Rémi Denis-Courmont <remi@remlab.net> > > RV64G supports MIN & MAX instructions natively only on floating point > registers, not general purpose ones. The later would require the Zbb > extension. Due to that, it is actually faster to perform the clipping > "properly" in FPU. > > Benchmarked on SiFive U74-MC: > > audiodsp.vector_clipf_c: 29551.5 > audiodsp.vector_clipf_f: 17871.0 > > Also tried unrolling with 2 or 8 elements but it gets worse either way. > --- > libavcodec/audiodsp.c | 2 ++ > libavcodec/audiodsp.h | 1 + > libavcodec/riscv/Makefile | 2 ++ > libavcodec/riscv/audiodsp_init.c | 31 +++++++++++++++++++++ > libavcodec/riscv/audiodsp_rvf.S | 46 ++++++++++++++++++++++++++++++++ > 5 files changed, 82 insertions(+) > create mode 100644 libavcodec/riscv/Makefile > create mode 100644 libavcodec/riscv/audiodsp_init.c > create mode 100644 libavcodec/riscv/audiodsp_rvf.S > > diff --git a/libavcodec/audiodsp.c b/libavcodec/audiodsp.c > index ff43e87dce..eba6e809fd 100644 > --- a/libavcodec/audiodsp.c > +++ b/libavcodec/audiodsp.c > @@ -113,6 +113,8 @@ av_cold void ff_audiodsp_init(AudioDSPContext *c) > ff_audiodsp_init_arm(c); > #elif ARCH_PPC > ff_audiodsp_init_ppc(c); > +#elif ARCH_RISCV > + ff_audiodsp_init_riscv(c); > #elif ARCH_X86 > ff_audiodsp_init_x86(c); > #endif > diff --git a/libavcodec/audiodsp.h b/libavcodec/audiodsp.h > index aa6fa7898b..485b512839 100644 > --- a/libavcodec/audiodsp.h > +++ b/libavcodec/audiodsp.h > @@ -55,6 +55,7 @@ typedef struct AudioDSPContext { > void ff_audiodsp_init(AudioDSPContext *c); > void ff_audiodsp_init_arm(AudioDSPContext *c); > void ff_audiodsp_init_ppc(AudioDSPContext *c); > +void ff_audiodsp_init_riscv(AudioDSPContext *c); > void ff_audiodsp_init_x86(AudioDSPContext *c); > > #endif /* AVCODEC_AUDIODSP_H */ > diff --git a/libavcodec/riscv/Makefile b/libavcodec/riscv/Makefile > new file mode 100644 > index 0000000000..a1f67ed55b > --- /dev/null > +++ b/libavcodec/riscv/Makefile > @@ -0,0 +1,2 @@ > +OBJS += riscv/audiodsp_init.o \ > + riscv/audiodsp_rvf.o > diff --git a/libavcodec/riscv/audiodsp_init.c > b/libavcodec/riscv/audiodsp_init.c new file mode 100644 > index 0000000000..7ffd7e8162 > --- /dev/null > +++ b/libavcodec/riscv/audiodsp_init.c > @@ -0,0 +1,31 @@ > +/* > + * 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/attributes.h" > +#include "libavutil/cpu.h" > +#include "libavcodec/audiodsp.h" > + > +void ff_vector_clipf_rvf(float *dst, const float *src, int len, float min, > float max); + > +av_cold void ff_audiodsp_init_riscv(AudioDSPContext *c) > +{ > + int flags = av_get_cpu_flags(); > + > + if (flags & AV_CPU_FLAG_F) > + c->vector_clipf = ff_vector_clipf_rvf; > +} > diff --git a/libavcodec/riscv/audiodsp_rvf.S > b/libavcodec/riscv/audiodsp_rvf.S new file mode 100644 > index 0000000000..148af96ea2 > --- /dev/null > +++ b/libavcodec/riscv/audiodsp_rvf.S > @@ -0,0 +1,46 @@ > +/* > + * 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/riscv/asm.S" > + > +func ff_vector_clipf_rvf, f > +NOHWF fmv.w.x fa0, a3 > +NOHWF fmv.w.v fa1, a4 This should be fmv.w.x, though I don't know a distro that would use the soft float ABI. Need more sleep. > +1: > + flw ft0, (a1) > + flw ft1, 4(a1) > + fmax.s ft0, ft0, fa0 > + flw ft2, 8(a1) > + fmax.s ft1, ft1, fa0 > + flw ft3, 12(a1) > + fmax.s ft2, ft2, fa0 > + addi a2, a2, -4 > + fmax.s ft3, ft3, fa0 > + addi a1, a1, 16 > + fmin.s ft0, ft0, fa1 > + fmin.s ft1, ft1, fa1 > + fsw ft0, (a0) > + fmin.s ft2, ft2, fa1 > + fsw ft1, 4(a0) > + fmin.s ft3, ft3, fa1 > + fsw ft2, 8(a0) > + fsw ft3, 12(a0) > + addi a0, a0, 16 > + bnez a2, 1b > + ret > +endfunc -- Реми Дёни-Курмон http://www.remlab.net/ _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
prev parent reply other threads:[~2022-09-14 17:57 UTC|newest] Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top 2022-09-14 17:50 [FFmpeg-devel] [RFC] [PATCH 0/3] RISC-V CPU detection Rémi Denis-Courmont 2022-09-14 17:50 ` [FFmpeg-devel] [PATCH 1/3] lavu: detect RISC-V F extension (i.e. float) remi 2022-09-14 19:28 ` Lynne 2022-09-14 19:43 ` Rémi Denis-Courmont 2022-09-14 17:50 ` [FFmpeg-devel] [PATCH 2/3] lavu/riscv: initial common header for assembler macros remi 2022-09-14 17:50 ` [FFmpeg-devel] [PATCH 3/3] lavc/audiodsp: add RISC-V F float vector clip remi 2022-09-14 17:56 ` Rémi Denis-Courmont [this message]
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=4740913.GXAFRqVoOG@basile.remlab.net \ --to=remi@remlab.net \ --cc=ffmpeg-devel@ffmpeg.org \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: link
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel This inbox may be cloned and mirrored by anyone: git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git # If you have public-inbox 1.1+ installed, you may # initialize and index your mirror using the following commands: public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \ ffmpegdev@gitmailbox.com public-inbox-index ffmpegdev Example config snippet for mirrors. AGPL code for this site: git clone https://public-inbox.org/public-inbox.git