Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
 help / color / mirror / Atom feed
From: flow gg <hlefthleft@gmail.com>
To: FFmpeg development discussions and patches <ffmpeg-devel@ffmpeg.org>
Subject: Re: [FFmpeg-devel] [PATCH] af_afir: RISC-V V fcmul_add
Date: Thu, 28 Sep 2023 13:45:44 +0800
Message-ID: <CAEa-L+t-FxWG7=M-WuL5Tzk6w+xmhVrAnSvWUwjadesxeaXncw@mail.gmail.com> (raw)
In-Reply-To: <2238622.iZASKD2KPV@basile.remlab.net>

[-- Attachment #1: Type: text/plain, Size: 1584 bytes --]

Okay, I revert the volatile in ff_read_time

How about this version?

use vls instead vlseg, and use vfmacc

The benchmark is sometimes better, sometimes the same

fcmul_add_c: 3.5
fcmul_add_rvv_f32: 3.5
 - af_afir.fcmul_add [OK]
fcmul_add_c: 4.5
fcmul_add_rvv_f32: 4.2
 - af_afir.fcmul_add [OK]
fcmul_add_c: 4.2
fcmul_add_rvv_f32: 4.2
 - af_afir.fcmul_add [OK]
fcmul_add_c: 4.5
fcmul_add_rvv_f32: 4.2
 - af_afir.fcmul_add [OK]
fcmul_add_c: 4.7
fcmul_add_rvv_f32: 3.5


Rémi Denis-Courmont <remi@remlab.net> 于2023年9月28日周四 00:41写道:

> Le tiistaina 26. syyskuuta 2023, 12.24.58 EEST flow gg a écrit :
> > benchmark:
> > fcmul_add_c: 19.7
> > fcmul_add_rvv_f32: 6.7
>
> With optimisations enabled and the benchmarking fix, I get this (on the
> same
> hardware, I believe):
>
> fcmul_add_c: 3.5
> fcmul_add_rvv_f32: 6.7
>
> For sure unfortunate design limitations of T-Head C910 are to blame to no
> small extent. It is not the first occurrence of an RVV optimisation that
> turns
> out worse than scalar due to those, and I still have honest hopes that
> newer
> (and conformant) IP would give saner results, but... I also believe that
> the
> code could be improved regardless.
>
> --
> Rémi Denis-Courmont
> 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".
>

[-- Attachment #2: 0001-af_afir-RISC-V-V-fcmul_add.patch --]
[-- Type: text/x-patch, Size: 5159 bytes --]

From 3e956958b01e780b7360dace59b6248f61a6f12c Mon Sep 17 00:00:00 2001
From: h <hlefthleft@gmail.com>
Date: Tue, 26 Sep 2023 15:03:12 +0800
Subject: [PATCH] af_afir: RISC-V V fcmul_add

---
 libavfilter/af_afirdsp.h         |  3 ++
 libavfilter/riscv/Makefile       |  2 ++
 libavfilter/riscv/af_afir_init.c | 39 ++++++++++++++++++++
 libavfilter/riscv/af_afir_rvv.S  | 61 ++++++++++++++++++++++++++++++++
 4 files changed, 105 insertions(+)
 create mode 100644 libavfilter/riscv/Makefile
 create mode 100644 libavfilter/riscv/af_afir_init.c
 create mode 100644 libavfilter/riscv/af_afir_rvv.S

diff --git a/libavfilter/af_afirdsp.h b/libavfilter/af_afirdsp.h
index 4208501393..d2d1e909c1 100644
--- a/libavfilter/af_afirdsp.h
+++ b/libavfilter/af_afirdsp.h
@@ -34,6 +34,7 @@ typedef struct AudioFIRDSPContext {
 } AudioFIRDSPContext;
 
 void ff_afir_init_x86(AudioFIRDSPContext *s);
+void ff_afir_init_riscv(AudioFIRDSPContext *s);
 
 static void fcmul_add_c(float *sum, const float *t, const float *c, ptrdiff_t len)
 {
@@ -76,6 +77,8 @@ static av_unused void ff_afir_init(AudioFIRDSPContext *dsp)
 
 #if ARCH_X86
     ff_afir_init_x86(dsp);
+#elif ARCH_RISCV
+    ff_afir_init_riscv(dsp);
 #endif
 }
 
diff --git a/libavfilter/riscv/Makefile b/libavfilter/riscv/Makefile
new file mode 100644
index 0000000000..0b968a9c0d
--- /dev/null
+++ b/libavfilter/riscv/Makefile
@@ -0,0 +1,2 @@
+OBJS += riscv/af_afir_init.o
+RVV-OBJS += riscv/af_afir_rvv.o
diff --git a/libavfilter/riscv/af_afir_init.c b/libavfilter/riscv/af_afir_init.c
new file mode 100644
index 0000000000..ffa176abd2
--- /dev/null
+++ b/libavfilter/riscv/af_afir_init.c
@@ -0,0 +1,39 @@
+/*
+ * Copyright © 2023 hleft
+ *
+ * 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 <stdint.h>
+
+#include "config.h"
+#include "libavutil/attributes.h"
+#include "libavutil/cpu.h"
+#include "libavfilter/af_afirdsp.h"
+
+void ff_fcmul_add_rvv(float *sum, const float *t, const float *c,
+                       ptrdiff_t len);
+
+av_cold void ff_afir_init_riscv(AudioFIRDSPContext *s)
+{
+#if HAVE_RVV
+    int flags = av_get_cpu_flags();
+
+    if (flags & AV_CPU_FLAG_RVV_F32)
+        s->fcmul_add = ff_fcmul_add_rvv;
+#endif
+}
diff --git a/libavfilter/riscv/af_afir_rvv.S b/libavfilter/riscv/af_afir_rvv.S
new file mode 100644
index 0000000000..6c15586007
--- /dev/null
+++ b/libavfilter/riscv/af_afir_rvv.S
@@ -0,0 +1,61 @@
+/*
+ * Copyright © 2023 hleft
+ *
+ * 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"
+
+//  void ff_fcmul_add(float *sum, const float *t, const float *c, int len)
+func ff_fcmul_add_rvv, zve32f
+1:
+        vsetvli     t0, a3, e32, m1, ta, ma
+        li          t1, 8
+        li          t2, 4
+        vlse32.v    v0, (a1), t1
+        add         a1, a1, t2
+        vlse32.v    v1, (a1), t1
+        sub         a1, a1, t2
+        vlse32.v    v2, (a2), t1
+        add         a2, a2, t2
+        vlse32.v    v3, (a2), t1
+        sub         a2, a2, t2
+        vlse32.v    v4, (a0), t1
+        add         a0, a0, t2
+        vlse32.v    v5, (a0), t1
+        sub         a0, a0, t2
+        vfmacc.vv   v4, v0, v2
+        vfnmsac.vv  v4, v1, v3
+        vfmacc.vv   v5, v0, v3
+        vfmacc.vv   v5, v1, v2
+        vsseg2e32.v v4, (a0)
+        mul         t3, t1, t0
+        add         a0, a0, t3
+        add         a1, a1, t3
+        add         a2, a2, t3
+        sub         a3, a3, t0
+        bgtz        a3, 1b
+
+        flw         fa0, 0(a1)
+        flw         fa1, 0(a2)
+        flw         fa2, 0(a0)
+        fmul.s      fa0, fa0, fa1
+        fadd.s      fa2, fa2, fa0
+        fsw         fa2, 0(a0)
+
+        ret
+endfunc
-- 
2.42.0


[-- Attachment #3: Type: text/plain, Size: 251 bytes --]

_______________________________________________
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".

  reply	other threads:[~2023-09-28  5:46 UTC|newest]

Thread overview: 18+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-09-26  9:24 flow gg
2023-09-26 18:34 ` Rémi Denis-Courmont
2023-09-26 18:40   ` Paul B Mahol
2023-09-26 18:44     ` Rémi Denis-Courmont
2023-09-27  1:47       ` flow gg
2023-09-27 16:01         ` Rémi Denis-Courmont
2023-09-27 16:27         ` Rémi Denis-Courmont
2023-09-26 18:50 ` Rémi Denis-Courmont
2023-09-27 16:41 ` Rémi Denis-Courmont
2023-09-28  5:45   ` flow gg [this message]
2023-09-28 13:33     ` Rémi Denis-Courmont
2023-11-13  9:43       ` flow gg
2023-11-13 15:35         ` Rémi Denis-Courmont
2023-11-13 16:01           ` Paul B Mahol
2023-11-15  8:57           ` flow gg
2023-11-15  8:59           ` flow gg
2023-11-15 15:05             ` Rémi Denis-Courmont
2023-11-15 23:04               ` flow gg

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='CAEa-L+t-FxWG7=M-WuL5Tzk6w+xmhVrAnSvWUwjadesxeaXncw@mail.gmail.com' \
    --to=hlefthleft@gmail.com \
    --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