Thank you for your guidance, I finally understand.. How about choosing
manual zero-extension for rv64? I modified the patch.
#if (__riscv_xlen == 64)
slli a2, a2, 32
srli a2, a2, 32
#endif
Rémi Denis-Courmont <[email protected]> 于2023年11月22日周三 22:51写道：
>
>
> Le 22 novembre 2023 16:30:44 GMT+02:00, flow gg <[email protected]> a
> écrit :
> >> How did you test it?
> >
> >I wrote a test, but it was a bit rough, so I want to modify it before
> >submitting. I've added it to this reply.
> >
> >> This does not seem according to the C ABI. AFAIK `unsigned` is
> >sign-extended.
> >
> >I'm a bit confused... because this passed in the tests I wrote in qemu.
> >Maybe there's a problem with my test？
>
> You probably didn't test sizes between 2^31 and 2^32-1. This might not
> even be feasible in QEMU.
>
> Ideally the prototype would use size_t, then the problem wouldn't exist.
>
> >
> >> ALU right before dependent conditional branch should be avoided.
> >
> >Should the sub be moved forward? I've modified it.
> >
> >> SHxADD can be used advantageously.
> >
> >Okay, I've made the modification
> >
> >Rémi Denis-Courmont <[email protected]> 于2023年11月22日周三 21:41写道：
> >
> >> Hi,
> >>
> >> How did you test it? As per
> >> http://ffmpeg.org/pipermail/ffmpeg-devel/2023-June/310720.html we still
> >> don't have a FATE instance set up with the RISC-V Vector extension. The
> >> only testing consists of my manual runs of checkasm on a K230 board. (We
> >> *do* have Zba and Zbb now though, hence the existing
> extract_exponents()).
> >>
> >> Also:
> >> - This does not seem according to the C ABI. AFAIK `unsigned` is
> >> sign-extended.
> >> - ALU right before dependent conditional branch should be avoided.
> >> - SHxADD can be used advantageously.
> >>
> >>
> >> Le 22 novembre 2023 14:00:07 GMT+02:00, flow gg <[email protected]>
> a
> >> écrit :
> >> >c910
> >> > float_to_fixed24_c: 208.2
> >> > float_to_fixed24_rvv_f32: 71.5
> >> _______________________________________________
> >> ffmpeg-devel mailing list
> >> [email protected]
> >> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> >>
> >> To unsubscribe, visit link above, or email
> >> [email protected] with subject "unsubscribe".
> >>
> _______________________________________________
> ffmpeg-devel mailing list
> [email protected]
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
> To unsubscribe, visit link above, or email
> [email protected] with subject "unsubscribe".
>
From d709519219138b746ff622b15bb004b27eed7333 Mon Sep 17 00:00:00 2001
From: sunyuechi <[email protected]>
Date: Wed, 22 Nov 2023 14:57:29 +0800
Subject: [PATCH] lavc/ac3dsp: R-V V float_to_fixed24
c910
float_to_fixed24_c: 208.2
float_to_fixed24_rvv_f32: 71.5
---
libavcodec/riscv/Makefile | 3 ++-
libavcodec/riscv/ac3dsp_init.c | 5 ++++
libavcodec/riscv/ac3dsp_rvv.S | 45 ++++++++++++++++++++++++++++++++++
3 files changed, 52 insertions(+), 1 deletion(-)
create mode 100644 libavcodec/riscv/ac3dsp_rvv.S
diff --git a/libavcodec/riscv/Makefile b/libavcodec/riscv/Makefile
index 8f2a519827..ac7b7c2929 100644
--- a/libavcodec/riscv/Makefile
+++ b/libavcodec/riscv/Makefile
@@ -1,7 +1,8 @@
OBJS-$(CONFIG_AAC_DECODER) += riscv/aacpsdsp_init.o riscv/sbrdsp_init.o
RVV-OBJS-$(CONFIG_AAC_DECODER) += riscv/aacpsdsp_rvv.o riscv/sbrdsp_rvv.o
OBJS-$(CONFIG_AC3DSP) += riscv/ac3dsp_init.o \
- riscv/ac3dsp_rvb.o
+ riscv/ac3dsp_rvb.o \
+ riscv/ac3dsp_rvv.o
OBJS-$(CONFIG_ALAC_DECODER) += riscv/alacdsp_init.o
RVV-OBJS-$(CONFIG_ALAC_DECODER) += riscv/alacdsp_rvv.o
OBJS-$(CONFIG_AUDIODSP) += riscv/audiodsp_init.o \
diff --git a/libavcodec/riscv/ac3dsp_init.c b/libavcodec/riscv/ac3dsp_init.c
index 20f294f1de..de82d1c7a7 100644
--- a/libavcodec/riscv/ac3dsp_init.c
+++ b/libavcodec/riscv/ac3dsp_init.c
@@ -26,6 +26,7 @@
#include "libavcodec/ac3dsp.h"
void ff_extract_exponents_rvb(uint8_t *exp, int32_t *coef, int nb_coefs);
+void ff_float_to_fixed24_rvv(int32_t *dst, const float *src, unsigned int len);
av_cold void ff_ac3dsp_init_riscv(AC3DSPContext *c)
{
@@ -35,4 +36,8 @@ av_cold void ff_ac3dsp_init_riscv(AC3DSPContext *c)
if (flags & AV_CPU_FLAG_RVB_BASIC)
c->extract_exponents = ff_extract_exponents_rvb;
}
+
+ if (flags & AV_CPU_FLAG_RVV_F32) {
+ c->float_to_fixed24 = ff_float_to_fixed24_rvv;
+ }
}
diff --git a/libavcodec/riscv/ac3dsp_rvv.S b/libavcodec/riscv/ac3dsp_rvv.S
new file mode 100644
index 0000000000..4d4a566659
--- /dev/null
+++ b/libavcodec/riscv/ac3dsp_rvv.S
@@ -0,0 +1,45 @@
+/*
+ * Copyright (c) 2023 Institue of Software Chinese Academy of Sciences (ISCAS).
+ *
+ * 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 "config.h"
+#include "libavutil/riscv/asm.S"
+
+func ff_float_to_fixed24_rvv, zve32f
+ li t1, 1 << 24
+ fcvt.s.w f0, t1
+
+#if (__riscv_xlen == 64)
+ slli a2, a2, 32
+ srli a2, a2, 32
+#endif
+
+1:
+ vsetvli t0, a2, e32, m4, ta, ma
+ sub a2, a2, t0
+ vle32.v v0, (a1)
+ vfmul.vf v0, v0, f0
+ vfcvt.x.f.v v0, v0
+ sh2add a1, t0, a1
+ vse32.v v0, (a0)
+ sh2add a0, t0, a0
+ bgtz a2, 1b
+
+ ret
+endfunc
--
2.43.0
_______________________________________________
ffmpeg-devel mailing list
[email protected]
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
[email protected] with subject "unsubscribe".
