>>> please pad mnemonics to at least 8 columns for consistency
okay, changed
>>> It seems that you could just as well use vlseg2 without register
stride, no?
yes, vlseg will better, changed
>>> Note that you could do the double versions with very little extra
efforts.
okay
>>> But really, DO NOT use a fixed vector length here. At best, you're
wasting half
>>> the vector width. Your input has a variable size, use it.
okay, changed
>>> I'm a bit surprised that the performance improves this much,
considering that
>>> the C910 is notoriously bad at both segmented strided loads. It might
be that
>>> the C versions is just very bad due to lack of aliasing optimisations.
thanks, You reminded me.
Sorry I had forgotten that there was a problem..
A few days ago, I wanted to try running some existing benchmarks,
```
tests/checkasm/checkasm --bench --test=aacpsdsp
tests/checkasm/checkasm --bench --test=alacdsp
tests/checkasm/checkasm --bench --test=audiodsp
tests/checkasm/checkasm --bench --test=g722dsp
tests/checkasm/checkasm --bench --test=vorbisdsp
tests/checkasm/checkasm --bench --test=float_dsp
tests/checkasm/checkasm --bench --test=fixed_dsp
tests/checkasm/checkasm --bench --test=af_afir
```
but they all returned 0.0.
For example,
```
butterflies_float_c: 0.0
butterflies_float_rvv_f32: 0.0
scalarproduct_float_c: 0.0
scalarproduct_float_rvv_f32: 0.0
vector_dmac_scalar_c: 0.0
vector_dmac_scalar_rvv_f64: 0.0
...
```
I tried changing the -O3 in configure to -O2 or -O1, but still got 0.0.
Only by changing to -O0 did I receive non-zero results.
So, the benchmark I conducted was based on this, and I obtained the initial
results…
fcmul_add_c: 19.7
fcmul_add_rvv_f32: 6.7
Rémi Denis-Courmont <r...@remlab.net> 于2023年9月27日周三 02:44写道：
> Le tiistaina 26. syyskuuta 2023, 21.40.12 EEST Paul B Mahol a écrit :
> > On Tue, Sep 26, 2023 at 8:35 PM Rémi Denis-Courmont <r...@remlab.net>
> wrote:
> > > 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
> > >
> > > Nit: please pad mnemonics to at least 8 columns for consistency.
> > >
> > > I'm a bit surprised that the performance improves this much,
> considering
> > > that
> > > the C910 is notoriously bad at both segmented strided loads. It might
> be
> > > that
> > > the C versions is just very bad due to lack of aliasing optimisations.
> Oh
> > > well.
> >
> > What you mean exactly that C version is missing?
>
> The C version does not have any restrict qualifier. This potentially
> prevents
> the C compiler from unrolling. Adding the keyword can improve performance
> gains of 20-30% on RISC-V scalar floating point.
>
> That said, sometimes you can't validly use restrict, and you simply can't
> tell
> the C compiler how to optimise properly. In those cases, even scalar
> floating
> point optimisations improve performance.
>
> --
> 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-requ...@ffmpeg.org with subject "unsubscribe".
>
From 5343b40c2e1849daadd79d6d531aeb7a762e0e06 Mon Sep 17 00:00:00 2001
From: h <hlefthl...@gmail.com>
Date: Tue, 26 Sep 2023 15:03:12 +0800
Subject: [PATCH] af_afir: RISC-V V fcmul_add
fcmul_add_c: 19.7
fcmul_add_rvv_f32: 6.7
---
libavfilter/af_afirdsp.h | 3 ++
libavfilter/riscv/Makefile | 2 ++
libavfilter/riscv/af_afir_init.c | 39 ++++++++++++++++++++++
libavfilter/riscv/af_afir_rvv.S | 55 ++++++++++++++++++++++++++++++++
4 files changed, 99 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..e060d03b6e
--- /dev/null
+++ b/libavfilter/riscv/af_afir_rvv.S
@@ -0,0 +1,55 @@
+/*
+ * 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
+ vlseg2e32.v v0, (a1)
+ vlseg2e32.v v2, (a2)
+ vlseg2e32.v v4, (a0)
+ vfmul.vv v6, v0, v2
+ vfmul.vv v7, v1, v3
+ vfmul.vv v8, v0, v3
+ vfmul.vv v9, v1, v2
+ vfadd.vv v4, v4, v6
+ vfsub.vv v4, v4, v7
+ vfadd.vv v5, v5, v8
+ vfadd.vv v5, v5, v9
+ vsseg2e32.v v4, (a0)
+ li t2, 8
+ mul t3, t2, 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
_______________________________________________
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".
