* [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24
@ 2023-11-22 12:00 flow gg
2023-11-22 13:40 ` Rémi Denis-Courmont
2023-11-23 7:11 ` flow gg
0 siblings, 2 replies; 17+ messages in thread
From: flow gg @ 2023-11-22 12:00 UTC (permalink / raw)
To: FFmpeg development discussions and patches
[-- Attachment #1: Type: text/plain, Size: 70 bytes --]
c910
float_to_fixed24_c: 208.2
float_to_fixed24_rvv_f32: 71.5
[-- Attachment #2: lavc-ac3dsp-R-V-V-float_to_fixed24.patch --]
[-- Type: text/x-patch, Size: 3554 bytes --]
From 69da974fd0febaa74db4dd551b05172caeefb846 Mon Sep 17 00:00:00 2001
From: sunyuechi <sunyuechi@iscas.ac.cn>
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 | 40 ++++++++++++++++++++++++++++++++++
3 files changed, 47 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..4d8ab060e7
--- /dev/null
+++ b/libavcodec/riscv/ac3dsp_rvv.S
@@ -0,0 +1,40 @@
+/*
+ * 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
+1:
+ vsetvli t0, a2, e32, m4, ta, ma
+ vle32.v v0, (a1)
+ slli t3, t0, 2
+ vfmul.vf v0, v0, f0
+ vfcvt.x.f.v v0, v0
+ add a1, a1, t3
+ vse32.v v0, (a0)
+ add a0, a0, t3
+ sub a2, a2, t0
+ bgtz a2, 1b
+
+ ret
+endfunc
--
2.43.0
[-- 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".
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24
2023-11-22 12:00 [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24 flow gg
@ 2023-11-22 13:40 ` Rémi Denis-Courmont
2023-11-22 14:30 ` flow gg
2023-11-23 7:11 ` flow gg
1 sibling, 1 reply; 17+ messages in thread
From: Rémi Denis-Courmont @ 2023-11-22 13:40 UTC (permalink / raw)
To: FFmpeg development discussions and patches
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 <hlefthleft@gmail.com> a écrit :
>c910
> float_to_fixed24_c: 208.2
> float_to_fixed24_rvv_f32: 71.5
_______________________________________________
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".
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24
2023-11-22 13:40 ` Rémi Denis-Courmont
@ 2023-11-22 14:30 ` flow gg
2023-11-22 14:35 ` flow gg
` (2 more replies)
0 siblings, 3 replies; 17+ messages in thread
From: flow gg @ 2023-11-22 14:30 UTC (permalink / raw)
To: FFmpeg development discussions and patches
[-- Attachment #1: Type: text/plain, Size: 1628 bytes --]
> 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?
> 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 <remi@remlab.net> 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 <hlefthleft@gmail.com> a
> écrit :
> >c910
> > float_to_fixed24_c: 208.2
> > float_to_fixed24_rvv_f32: 71.5
> _______________________________________________
> 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: lavc-ac3dsp-R-V-V-float_to_fixed24_2.patch --]
[-- Type: text/x-patch, Size: 3521 bytes --]
From 3e790fdccd780257f464aa8f8a56a37321ddd429 Mon Sep 17 00:00:00 2001
From: sunyuechi <sunyuechi@iscas.ac.cn>
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 | 39 ++++++++++++++++++++++++++++++++++
3 files changed, 46 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..c0e2880e28
--- /dev/null
+++ b/libavcodec/riscv/ac3dsp_rvv.S
@@ -0,0 +1,39 @@
+/*
+ * 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
+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
[-- Attachment #3: 0001-add-ac3dsp-test.patch --]
[-- Type: text/x-patch, Size: 4930 bytes --]
From 08a012d86db51275fd2cda8dd7ad47cc1f1481ce Mon Sep 17 00:00:00 2001
From: sunyuechi <sunyuechi@iscas.ac.cn>
Date: Wed, 22 Nov 2023 14:57:29 +0800
Subject: [PATCH] lavc/ac3dsp: R-V V float_to_fixed24
---
tests/checkasm/Makefile | 1 +
tests/checkasm/ac3dsp.c | 88 +++++++++++++++++++++++++++++++++++++++
tests/checkasm/checkasm.c | 3 ++
tests/checkasm/checkasm.h | 1 +
4 files changed, 93 insertions(+)
create mode 100644 tests/checkasm/ac3dsp.c
diff --git a/tests/checkasm/Makefile b/tests/checkasm/Makefile
index 8bc241d29b..8c714c2a07 100644
--- a/tests/checkasm/Makefile
+++ b/tests/checkasm/Makefile
@@ -5,6 +5,7 @@ AVCODECOBJS-$(CONFIG_BLOCKDSP) += blockdsp.o
AVCODECOBJS-$(CONFIG_BSWAPDSP) += bswapdsp.o
AVCODECOBJS-$(CONFIG_FMTCONVERT) += fmtconvert.o
AVCODECOBJS-$(CONFIG_G722DSP) += g722dsp.o
+AVCODECOBJS-$(CONFIG_AC3DSP) += ac3dsp.o
AVCODECOBJS-$(CONFIG_H264CHROMA) += h264chroma.o
AVCODECOBJS-$(CONFIG_H264DSP) += h264dsp.o
AVCODECOBJS-$(CONFIG_H264PRED) += h264pred.o
diff --git a/tests/checkasm/ac3dsp.c b/tests/checkasm/ac3dsp.c
new file mode 100644
index 0000000000..ebebe06990
--- /dev/null
+++ b/tests/checkasm/ac3dsp.c
@@ -0,0 +1,88 @@
+#include "checkasm.h"
+#include <stdio.h>
+
+
+#include <string.h>
+
+#include "libavutil/common.h"
+#include "libavutil/intreadwrite.h"
+#include "libavutil/mem.h"
+#include "libavutil/mem_internal.h"
+
+#include "libavcodec/ac3dsp.h"
+
+/**
+ * Convert an array of float in range [-1.0,1.0] to int32_t with range
+ * [-(1<<24),(1<<24)]
+ *
+ * @param dst destination array of int32_t.
+ * constraints: 16-byte aligned
+ * @param src source array of float.
+ * constraints: 16-byte aligned
+ * @param len number of elements to convert.
+ * constraints: multiple of 32 greater than zero
+ */
+// void (*float_to_fixed24)(int32_t *dst, const float *src, unsigned int len);
+
+
+#define randomize_float(buf, len) \
+ do { \
+ int i; \
+ for (i = 0; i < len; i++) { \
+ float f = (float)rnd() / (UINT_MAX >> 5) - 16.0f; \
+ buf[i] = f; \
+ } \
+ } while (0)
+
+#define randomize_int(buf, len, size, bits) \
+ do { \
+ int i; \
+ for (i = 0; i < len; i++) { \
+ uint ## size ## _t r = rnd() & ((1LL << bits) - 1); \
+ AV_WN ## size ## A(buf + i, -(1LL << (bits - 1)) + r); \
+ } \
+ } while (0)
+
+static void check_float_to_fixed24(AC3DSPContext *c) {
+#define BUF_SIZE 800
+ LOCAL_ALIGNED_32(int32_t, v1, [BUF_SIZE]);
+ LOCAL_ALIGNED_32(float, v2, [BUF_SIZE]);
+
+ declare_func(void, int32_t *, const float *, unsigned int);
+
+ randomize_int(v1, BUF_SIZE, 32, 10);
+ randomize_float(v2, BUF_SIZE);
+
+ if (check_func(c->float_to_fixed24, "float_to_fixed24")) {
+ LOCAL_ALIGNED_32(int32_t, dst, [BUF_SIZE]);
+ LOCAL_ALIGNED_32(int32_t, dst2, [BUF_SIZE]);
+
+ call_ref(dst, v2, 80);
+ call_new(dst2, v2, 80);
+
+ if (memcmp(dst, dst2, sizeof(*dst) * 10) != 0){
+ puts(">>>>>>>>>>>>>> fail --------------------");
+ for(int i = 0 ; i < 10; i++){
+ printf("dst[%d] = %d, dst2[%d] = %d\n", i, dst[i], i, dst2[i]);
+ }
+ puts("");
+
+ fail();
+ } else {
+ puts(">>>>>>>>>>>>>> ok --------------------");
+ }
+
+ bench_new(v1, v2, 80);
+ }
+
+
+ report("float_to_fixed24");
+}
+
+void checkasm_check_ac3dsp(void)
+{
+ AC3DSPContext c;
+ ff_ac3dsp_init(&c);
+
+ check_float_to_fixed24(&c);
+}
diff --git a/tests/checkasm/checkasm.c b/tests/checkasm/checkasm.c
index 708119e7c6..9502e372a1 100644
--- a/tests/checkasm/checkasm.c
+++ b/tests/checkasm/checkasm.c
@@ -105,6 +105,9 @@ static const struct {
#if CONFIG_G722DSP
{ "g722dsp", checkasm_check_g722dsp },
#endif
+ #if CONFIG_AC3DSP
+ { "ac3dsp", checkasm_check_ac3dsp },
+ #endif
#if CONFIG_H264CHROMA
{ "h264chroma", checkasm_check_h264chroma },
#endif
diff --git a/tests/checkasm/checkasm.h b/tests/checkasm/checkasm.h
index cfea868ff1..4c73589606 100644
--- a/tests/checkasm/checkasm.h
+++ b/tests/checkasm/checkasm.h
@@ -96,6 +96,7 @@ void checkasm_check_vp8dsp(void);
void checkasm_check_vp9dsp(void);
void checkasm_check_videodsp(void);
void checkasm_check_vorbisdsp(void);
+void checkasm_check_ac3dsp(void);
struct CheckasmPerf;
--
2.43.0
[-- Attachment #4: 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".
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24
2023-11-22 14:30 ` flow gg
@ 2023-11-22 14:35 ` flow gg
2023-11-22 14:51 ` Rémi Denis-Courmont
2023-11-22 17:18 ` James Almer
2 siblings, 0 replies; 17+ messages in thread
From: flow gg @ 2023-11-22 14:35 UTC (permalink / raw)
To: FFmpeg development discussions and patches
qemu-riscv64 -cpu rv64,v=true,g=true,c=true,zba=true,vlen=128 checkasm
--test=ac3dsp
flow gg <hlefthleft@gmail.com> 于2023年11月22日周三 22:30写道:
> > 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?
>
> > 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 <remi@remlab.net> 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 <hlefthleft@gmail.com> a
>> écrit :
>> >c910
>> > float_to_fixed24_c: 208.2
>> > float_to_fixed24_rvv_f32: 71.5
>> _______________________________________________
>> 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".
>>
>
_______________________________________________
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".
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24
2023-11-22 14:30 ` flow gg
2023-11-22 14:35 ` flow gg
@ 2023-11-22 14:51 ` Rémi Denis-Courmont
2023-11-22 16:37 ` flow gg
2023-11-22 17:18 ` James Almer
2 siblings, 1 reply; 17+ messages in thread
From: Rémi Denis-Courmont @ 2023-11-22 14:51 UTC (permalink / raw)
To: FFmpeg development discussions and patches
Le 22 novembre 2023 16:30:44 GMT+02:00, flow gg <hlefthleft@gmail.com> 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 <remi@remlab.net> 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 <hlefthleft@gmail.com> a
>> écrit :
>> >c910
>> > float_to_fixed24_c: 208.2
>> > float_to_fixed24_rvv_f32: 71.5
>> _______________________________________________
>> 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".
>>
_______________________________________________
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".
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24
2023-11-22 14:51 ` Rémi Denis-Courmont
@ 2023-11-22 16:37 ` flow gg
2023-11-22 16:49 ` James Almer
2023-11-22 23:17 ` flow gg
0 siblings, 2 replies; 17+ messages in thread
From: flow gg @ 2023-11-22 16:37 UTC (permalink / raw)
To: FFmpeg development discussions and patches
[-- Attachment #1: Type: text/plain, Size: 2637 bytes --]
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 <remi@remlab.net> 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 <hlefthleft@gmail.com> 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 <remi@remlab.net> 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 <hlefthleft@gmail.com>
> a
> >> écrit :
> >> >c910
> >> > float_to_fixed24_c: 208.2
> >> > float_to_fixed24_rvv_f32: 71.5
> >> _______________________________________________
> >> 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".
> >>
> _______________________________________________
> 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: lavc-ac3dsp-R-V-V-float_to_fixed24.patch --]
[-- Type: text/x-patch, Size: 3608 bytes --]
From d709519219138b746ff622b15bb004b27eed7333 Mon Sep 17 00:00:00 2001
From: sunyuechi <sunyuechi@iscas.ac.cn>
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
[-- 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".
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24
2023-11-22 16:37 ` flow gg
@ 2023-11-22 16:49 ` James Almer
2023-11-22 23:17 ` flow gg
1 sibling, 0 replies; 17+ messages in thread
From: James Almer @ 2023-11-22 16:49 UTC (permalink / raw)
To: ffmpeg-devel
On 11/22/2023 1:37 PM, flow gg wrote:
> 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
Please, don't top post.
I think it will be better to change the prototype to use ptrdiff_t for
len, as it's done in other dps functions.
>
> Rémi Denis-Courmont <remi@remlab.net> 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 <hlefthleft@gmail.com> 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 <remi@remlab.net> 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 <hlefthleft@gmail.com>
>> a
>>>> écrit :
>>>>> c910
>>>>> float_to_fixed24_c: 208.2
>>>>> float_to_fixed24_rvv_f32: 71.5
>>>> _______________________________________________
>>>> 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".
>>>>
>> _______________________________________________
>> 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".
>>
>>
>> _______________________________________________
>> 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".
_______________________________________________
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".
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24
2023-11-22 14:30 ` flow gg
2023-11-22 14:35 ` flow gg
2023-11-22 14:51 ` Rémi Denis-Courmont
@ 2023-11-22 17:18 ` James Almer
2023-11-22 17:34 ` flow gg
2 siblings, 1 reply; 17+ messages in thread
From: James Almer @ 2023-11-22 17:18 UTC (permalink / raw)
To: ffmpeg-devel
On 11/22/2023 11:30 AM, flow gg wrote:
>> 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.
> From 08a012d86db51275fd2cda8dd7ad47cc1f1481ce Mon Sep 17 00:00:00 2001
> From: sunyuechi <sunyuechi@iscas.ac.cn>
> Date: Wed, 22 Nov 2023 14:57:29 +0800
> Subject: [PATCH] lavc/ac3dsp: R-V V float_to_fixed24
>
> ---
> tests/checkasm/Makefile | 1 +
> tests/checkasm/ac3dsp.c | 88 +++++++++++++++++++++++++++++++++++++++
> tests/checkasm/checkasm.c | 3 ++
> tests/checkasm/checkasm.h | 1 +
> 4 files changed, 93 insertions(+)
> create mode 100644 tests/checkasm/ac3dsp.c
>
> diff --git a/tests/checkasm/Makefile b/tests/checkasm/Makefile
> index 8bc241d29b..8c714c2a07 100644
> --- a/tests/checkasm/Makefile
> +++ b/tests/checkasm/Makefile
> @@ -5,6 +5,7 @@ AVCODECOBJS-$(CONFIG_BLOCKDSP) += blockdsp.o
> AVCODECOBJS-$(CONFIG_BSWAPDSP) += bswapdsp.o
> AVCODECOBJS-$(CONFIG_FMTCONVERT) += fmtconvert.o
> AVCODECOBJS-$(CONFIG_G722DSP) += g722dsp.o
> +AVCODECOBJS-$(CONFIG_AC3DSP) += ac3dsp.o
> AVCODECOBJS-$(CONFIG_H264CHROMA) += h264chroma.o
> AVCODECOBJS-$(CONFIG_H264DSP) += h264dsp.o
> AVCODECOBJS-$(CONFIG_H264PRED) += h264pred.o
> diff --git a/tests/checkasm/ac3dsp.c b/tests/checkasm/ac3dsp.c
> new file mode 100644
> index 0000000000..ebebe06990
> --- /dev/null
> +++ b/tests/checkasm/ac3dsp.c
> @@ -0,0 +1,88 @@
> +#include "checkasm.h"
> +#include <stdio.h>
> +
> +
> +#include <string.h>
> +
> +#include "libavutil/common.h"
> +#include "libavutil/intreadwrite.h"
> +#include "libavutil/mem.h"
> +#include "libavutil/mem_internal.h"
> +
> +#include "libavcodec/ac3dsp.h"
> +
> +/**
> + * Convert an array of float in range [-1.0,1.0] to int32_t with range
> + * [-(1<<24),(1<<24)]
> + *
> + * @param dst destination array of int32_t.
> + * constraints: 16-byte aligned
> + * @param src source array of float.
> + * constraints: 16-byte aligned
> + * @param len number of elements to convert.
> + * constraints: multiple of 32 greater than zero
> + */
> +// void (*float_to_fixed24)(int32_t *dst, const float *src, unsigned int len);
> +
> +
> +#define randomize_float(buf, len) \
> + do { \
> + int i; \
> + for (i = 0; i < len; i++) { \
> + float f = (float)rnd() / (UINT_MAX >> 5) - 16.0f; \
> + buf[i] = f; \
> + } \
> + } while (0)
> +
> +#define randomize_int(buf, len, size, bits) \
> + do { \
> + int i; \
> + for (i = 0; i < len; i++) { \
> + uint ## size ## _t r = rnd() & ((1LL << bits) - 1); \
> + AV_WN ## size ## A(buf + i, -(1LL << (bits - 1)) + r); \
> + } \
> + } while (0)
> +
> +static void check_float_to_fixed24(AC3DSPContext *c) {
> +#define BUF_SIZE 800
800, if this is meant to be used as len, is not a multiple of 32.
> + LOCAL_ALIGNED_32(int32_t, v1, [BUF_SIZE]);
> + LOCAL_ALIGNED_32(float, v2, [BUF_SIZE]);
> +
> + declare_func(void, int32_t *, const float *, unsigned int);
> +
> + randomize_int(v1, BUF_SIZE, 32, 10);
This is not really used at all. The input is floats, and the output is
write only.
> + randomize_float(v2, BUF_SIZE);
> +
> + if (check_func(c->float_to_fixed24, "float_to_fixed24")) {
> + LOCAL_ALIGNED_32(int32_t, dst, [BUF_SIZE]);
> + LOCAL_ALIGNED_32(int32_t, dst2, [BUF_SIZE]);
The requirement is 16 byte alignment.
> +
> + call_ref(dst, v2, 80);
This should be BUF_SIZE. And 80 is also not a multiple of 32.
> + call_new(dst2, v2, 80);
> +
> + if (memcmp(dst, dst2, sizeof(*dst) * 10) != 0){
memcmp(dst, dst2, sizeof(dst))
> + puts(">>>>>>>>>>>>>> fail --------------------");
No puts(), please. This line is also not needed.
> + for(int i = 0 ; i < 10; i++){
> + printf("dst[%d] = %d, dst2[%d] = %d\n", i, dst[i], i, dst2[i]);
fprintf(stderr, ...);
> + }
> + puts("");
> +
> + fail();
> + } else {
> + puts(">>>>>>>>>>>>>> ok --------------------");
Same.
> + }
> +
> + bench_new(v1, v2, 80);
bench_new(dst2, v2...
> + }
> +
> +
> + report("float_to_fixed24");
> +}
> +
> +void checkasm_check_ac3dsp(void)
> +{
> + AC3DSPContext c;
> + ff_ac3dsp_init(&c);
> +
> + check_float_to_fixed24(&c);
> +}
> diff --git a/tests/checkasm/checkasm.c b/tests/checkasm/checkasm.c
> index 708119e7c6..9502e372a1 100644
> --- a/tests/checkasm/checkasm.c
> +++ b/tests/checkasm/checkasm.c
> @@ -105,6 +105,9 @@ static const struct {
> #if CONFIG_G722DSP
> { "g722dsp", checkasm_check_g722dsp },
> #endif
> + #if CONFIG_AC3DSP
> + { "ac3dsp", checkasm_check_ac3dsp },
> + #endif
> #if CONFIG_H264CHROMA
> { "h264chroma", checkasm_check_h264chroma },
> #endif
> diff --git a/tests/checkasm/checkasm.h b/tests/checkasm/checkasm.h
> index cfea868ff1..4c73589606 100644
> --- a/tests/checkasm/checkasm.h
> +++ b/tests/checkasm/checkasm.h
> @@ -96,6 +96,7 @@ void checkasm_check_vp8dsp(void);
> void checkasm_check_vp9dsp(void);
> void checkasm_check_videodsp(void);
> void checkasm_check_vorbisdsp(void);
> +void checkasm_check_ac3dsp(void);
>
> struct CheckasmPerf;
>
> --
> 2.43.0
>
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24
2023-11-22 17:18 ` James Almer
@ 2023-11-22 17:34 ` flow gg
0 siblings, 0 replies; 17+ messages in thread
From: flow gg @ 2023-11-22 17:34 UTC (permalink / raw)
To: FFmpeg development discussions and patches
Wow, thank you for reviewing this. I just wanted to see if the function was
working properly. There are so many bugs in the test code ...
_______________________________________________
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".
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24
2023-11-22 16:37 ` flow gg
2023-11-22 16:49 ` James Almer
@ 2023-11-22 23:17 ` flow gg
2023-11-23 17:08 ` Rémi Denis-Courmont
1 sibling, 1 reply; 17+ messages in thread
From: flow gg @ 2023-11-22 23:17 UTC (permalink / raw)
To: FFmpeg development discussions and patches
[-- Attachment #1: Type: text/plain, Size: 157 bytes --]
Hello, I saw the new commit "avcodec/ac3dsp: make len a size_t in
float_to_fixed24."
So I removed the part #if (__riscv_xlen == 64) and restored the patch.
[-- Attachment #2: lavc-ac3dsp-R-V-V-float_to_fixed24.patch --]
[-- Type: text/x-patch, Size: 3521 bytes --]
From 3e790fdccd780257f464aa8f8a56a37321ddd429 Mon Sep 17 00:00:00 2001
From: sunyuechi <sunyuechi@iscas.ac.cn>
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 | 39 ++++++++++++++++++++++++++++++++++
3 files changed, 46 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..c0e2880e28
--- /dev/null
+++ b/libavcodec/riscv/ac3dsp_rvv.S
@@ -0,0 +1,39 @@
+/*
+ * 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
+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
[-- 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".
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24
2023-11-22 12:00 [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24 flow gg
2023-11-22 13:40 ` Rémi Denis-Courmont
@ 2023-11-23 7:11 ` flow gg
1 sibling, 0 replies; 17+ messages in thread
From: flow gg @ 2023-11-23 7:11 UTC (permalink / raw)
To: FFmpeg development discussions and patches
[-- Attachment #1: Type: text/plain, Size: 424 bytes --]
I modified the temporary test and sent it in "[FFmpeg-devel] [PATCH]
checkasm/ac3dsp: add float_to_fixed24 test".
So the test time results have changed, and I updated them in the patch.
c910
float_to_fixed24_c: 2207.2
float_to_fixed24_rvv_f32: 696.2
flow gg <hlefthleft@gmail.com> 于2023年11月22日周三 20:00写道:
> c910
> float_to_fixed24_c: 208.2
> float_to_fixed24_rvv_f32: 71.5
>
[-- Attachment #2: lavc-ac3dsp-R-V-V-float_to_fixed24.patch --]
[-- Type: text/x-patch, Size: 3519 bytes --]
From 3e790fdccd780257f464aa8f8a56a37321ddd429 Mon Sep 17 00:00:00 2001
From: sunyuechi <sunyuechi@iscas.ac.cn>
Date: Wed, 22 Nov 2023 14:57:29 +0800
Subject: [PATCH] lavc/ac3dsp: R-V V float_to_fixed24
c910
float_to_fixed24_c: 2207.2
float_to_fixed24_rvv_f32: 696.2
---
libavcodec/riscv/Makefile | 3 ++-
libavcodec/riscv/ac3dsp_init.c | 5 +++++
libavcodec/riscv/ac3dsp_rvv.S | 39 ++++++++++++++++++++++++++++++++++
3 files changed, 46 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..c0e2880e28
--- /dev/null
+++ b/libavcodec/riscv/ac3dsp_rvv.S
@@ -0,0 +1,39 @@
+/*
+ * 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
+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
[-- 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".
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24
2023-11-22 23:17 ` flow gg
@ 2023-11-23 17:08 ` Rémi Denis-Courmont
2023-11-23 22:39 ` flow gg
0 siblings, 1 reply; 17+ messages in thread
From: Rémi Denis-Courmont @ 2023-11-23 17:08 UTC (permalink / raw)
To: FFmpeg development discussions and patches
Le torstaina 23. marraskuuta 2023, 1.17.03 EET flow gg a écrit :
> Hello, I saw the new commit "avcodec/ac3dsp: make len a size_t in
> float_to_fixed24."
>
> So I removed the part #if (__riscv_xlen == 64) and restored the patch.
You're not checking for Zba. Also 'bnez' would be more logical than 'bgtz'
for an unsigned counter.
--
レミ・デニ-クールモン
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".
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24
2023-11-23 17:08 ` Rémi Denis-Courmont
@ 2023-11-23 22:39 ` flow gg
2023-12-01 18:35 ` Rémi Denis-Courmont
0 siblings, 1 reply; 17+ messages in thread
From: flow gg @ 2023-11-23 22:39 UTC (permalink / raw)
To: FFmpeg development discussions and patches
[-- Attachment #1: Type: text/plain, Size: 812 bytes --]
Okay, changed
Rémi Denis-Courmont <remi@remlab.net> 于2023年11月24日周五 01:09写道:
> Le torstaina 23. marraskuuta 2023, 1.17.03 EET flow gg a écrit :
> > Hello, I saw the new commit "avcodec/ac3dsp: make len a size_t in
> > float_to_fixed24."
> >
> > So I removed the part #if (__riscv_xlen == 64) and restored the patch.
>
> You're not checking for Zba. Also 'bnez' would be more logical than
> 'bgtz'
> for an unsigned counter.
>
> --
> レミ・デニ-クールモン
> 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: lavc-ac3dsp-R-V-V-float_to_fixed24.patch --]
[-- Type: text/x-patch, Size: 3559 bytes --]
From af221d659ebc1e97b6d274681061fa8331d0b147 Mon Sep 17 00:00:00 2001
From: sunyuechi <sunyuechi@iscas.ac.cn>
Date: Wed, 22 Nov 2023 14:57:29 +0800
Subject: [PATCH] lavc/ac3dsp: R-V V float_to_fixed24
c910
float_to_fixed24_c: 2207.2
float_to_fixed24_rvv_f32: 696.2
---
libavcodec/riscv/Makefile | 3 ++-
libavcodec/riscv/ac3dsp_init.c | 3 +++
libavcodec/riscv/ac3dsp_rvv.S | 39 ++++++++++++++++++++++++++++++++++
3 files changed, 44 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..25244943cb 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)
{
@@ -34,5 +35,7 @@ av_cold void ff_ac3dsp_init_riscv(AC3DSPContext *c)
if (flags & AV_CPU_FLAG_RVB_ADDR) {
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..82c14ea275
--- /dev/null
+++ b/libavcodec/riscv/ac3dsp_rvv.S
@@ -0,0 +1,39 @@
+/*
+ * 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
+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
+ bnez a2, 1b
+
+ ret
+endfunc
--
2.43.0
[-- 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".
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24
2023-11-23 22:39 ` flow gg
@ 2023-12-01 18:35 ` Rémi Denis-Courmont
2023-12-01 18:38 ` Rémi Denis-Courmont
0 siblings, 1 reply; 17+ messages in thread
From: Rémi Denis-Courmont @ 2023-12-01 18:35 UTC (permalink / raw)
To: FFmpeg development discussions and patches
Le perjantaina 24. marraskuuta 2023, 0.39.39 EET flow gg a écrit :
> Okay, changed
src/libavcodec/riscv/ac3dsp_init.c: In function ‘ff_ac3dsp_init_riscv’:
src/libavcodec/riscv/ac3dsp_init.c:39:33: warning: assignment to ‘void (*)
(int32_t *, const float *, size_t)’ {aka ‘void (*)(int *, const float *, long
unsigned int)’} from incompatible pointer type ‘void (*)(int32_t *, const float
*, unsigned int)’ {aka ‘void (*)(int *, const float *, unsigned int)’} [-
Wincompatible-pointer-types]
39 | c->float_to_fixed24 = ff_float_to_fixed24_rvv;
| ^
Also the Makefile precondition is inaccurate.
--
雷米‧德尼-库尔蒙
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".
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24
2023-12-01 18:35 ` Rémi Denis-Courmont
@ 2023-12-01 18:38 ` Rémi Denis-Courmont
2023-12-01 19:50 ` flow gg
0 siblings, 1 reply; 17+ messages in thread
From: Rémi Denis-Courmont @ 2023-12-01 18:38 UTC (permalink / raw)
To: FFmpeg development discussions and patches
Le perjantaina 1. joulukuuta 2023, 20.35.10 EET Rémi Denis-Courmont a écrit :
> Le perjantaina 24. marraskuuta 2023, 0.39.39 EET flow gg a écrit :
> > Okay, changed
>
> src/libavcodec/riscv/ac3dsp_init.c: In function ‘ff_ac3dsp_init_riscv’:
> src/libavcodec/riscv/ac3dsp_init.c:39:33: warning: assignment to ‘void (*)
> (int32_t *, const float *, size_t)’ {aka ‘void (*)(int *, const float *,
> long unsigned int)’} from incompatible pointer type ‘void (*)(int32_t *,
> const float *, unsigned int)’ {aka ‘void (*)(int *, const float *, unsigned
> int)’} [- Wincompatible-pointer-types]
> 39 | c->float_to_fixed24 = ff_float_to_fixed24_rvv;
>
> | ^
>
> Also the Makefile precondition is inaccurate.
Oh, and on C908, LMUL=8 is actually faster than LMUL=4. Generally speaking,
you should maximise the LMUL unless there is a *specific* reason not to.
--
レミ・デニ-クールモン
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".
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24
2023-12-01 18:38 ` Rémi Denis-Courmont
@ 2023-12-01 19:50 ` flow gg
2023-12-01 20:16 ` flow gg
0 siblings, 1 reply; 17+ messages in thread
From: flow gg @ 2023-12-01 19:50 UTC (permalink / raw)
To: FFmpeg development discussions and patches
[-- Attachment #1: Type: text/plain, Size: 1456 bytes --]
Okay, changed and attached
Rémi Denis-Courmont <remi@remlab.net> 于2023年12月2日周六 02:38写道:
> Le perjantaina 1. joulukuuta 2023, 20.35.10 EET Rémi Denis-Courmont a
> écrit :
> > Le perjantaina 24. marraskuuta 2023, 0.39.39 EET flow gg a écrit :
> > > Okay, changed
> >
> > src/libavcodec/riscv/ac3dsp_init.c: In function ‘ff_ac3dsp_init_riscv’:
> > src/libavcodec/riscv/ac3dsp_init.c:39:33: warning: assignment to ‘void
> (*)
> > (int32_t *, const float *, size_t)’ {aka ‘void (*)(int *, const float *,
> > long unsigned int)’} from incompatible pointer type ‘void (*)(int32_t *,
> > const float *, unsigned int)’ {aka ‘void (*)(int *, const float *,
> unsigned
> > int)’} [- Wincompatible-pointer-types]
> > 39 | c->float_to_fixed24 = ff_float_to_fixed24_rvv;
> >
> > | ^
> >
> > Also the Makefile precondition is inaccurate.
>
> Oh, and on C908, LMUL=8 is actually faster than LMUL=4. Generally
> speaking,
> you should maximise the LMUL unless there is a *specific* reason not to.
>
> --
> レミ・デニ-クールモン
> 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: lavc-ac3dsp-R-V-V-float_to_fixed24.patch --]
[-- Type: text/x-patch, Size: 3725 bytes --]
From db03232a9ef1caab333b2fb5a1b684a68c7b0114 Mon Sep 17 00:00:00 2001
From: sunyuechi <sunyuechi@iscas.ac.cn>
Date: Wed, 22 Nov 2023 14:57:29 +0800
Subject: [PATCH] lavc/ac3dsp: R-V V float_to_fixed24
c910
float_to_fixed24_c: 2207.2
float_to_fixed24_rvv_f32: 696.2
---
libavcodec/riscv/Makefile | 3 ++-
libavcodec/riscv/ac3dsp_init.c | 4 ++++
libavcodec/riscv/ac3dsp_rvv.S | 39 ++++++++++++++++++++++++++++++++++
3 files changed, 45 insertions(+), 1 deletion(-)
create mode 100644 libavcodec/riscv/ac3dsp_rvv.S
diff --git a/libavcodec/riscv/Makefile b/libavcodec/riscv/Makefile
index 2d0e6c19c8..b00db279c1 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..118b2955ca 100644
--- a/libavcodec/riscv/ac3dsp_init.c
+++ b/libavcodec/riscv/ac3dsp_init.c
@@ -18,6 +18,7 @@
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
+#include <stddef.h>
#include <stdint.h>
#include "config.h"
@@ -26,6 +27,7 @@
#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, size_t len);
av_cold void ff_ac3dsp_init_riscv(AC3DSPContext *c)
{
@@ -34,5 +36,7 @@ av_cold void ff_ac3dsp_init_riscv(AC3DSPContext *c)
if (flags & AV_CPU_FLAG_RVB_ADDR) {
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..b8d32c4677
--- /dev/null
+++ b/libavcodec/riscv/ac3dsp_rvv.S
@@ -0,0 +1,39 @@
+/*
+ * 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
+1:
+ vsetvli t0, a2, e32, m8, 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
+ bnez a2, 1b
+
+ ret
+endfunc
--
2.43.0
[-- 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".
^ permalink raw reply [flat|nested] 17+ messages in thread
* Re: [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24
2023-12-01 19:50 ` flow gg
@ 2023-12-01 20:16 ` flow gg
0 siblings, 0 replies; 17+ messages in thread
From: flow gg @ 2023-12-01 20:16 UTC (permalink / raw)
To: FFmpeg development discussions and patches
[-- Attachment #1: Type: text/plain, Size: 1650 bytes --]
I forgot to modify the Makefile; I've made the changes in this reply.
flow gg <hlefthleft@gmail.com> 于2023年12月2日周六 03:50写道:
> Okay, changed and attached
>
> Rémi Denis-Courmont <remi@remlab.net> 于2023年12月2日周六 02:38写道:
>
>> Le perjantaina 1. joulukuuta 2023, 20.35.10 EET Rémi Denis-Courmont a
>> écrit :
>> > Le perjantaina 24. marraskuuta 2023, 0.39.39 EET flow gg a écrit :
>> > > Okay, changed
>> >
>> > src/libavcodec/riscv/ac3dsp_init.c: In function ‘ff_ac3dsp_init_riscv’:
>> > src/libavcodec/riscv/ac3dsp_init.c:39:33: warning: assignment to ‘void
>> (*)
>> > (int32_t *, const float *, size_t)’ {aka ‘void (*)(int *, const float *,
>> > long unsigned int)’} from incompatible pointer type ‘void (*)(int32_t *,
>> > const float *, unsigned int)’ {aka ‘void (*)(int *, const float *,
>> unsigned
>> > int)’} [- Wincompatible-pointer-types]
>> > 39 | c->float_to_fixed24 = ff_float_to_fixed24_rvv;
>> >
>> > | ^
>> >
>> > Also the Makefile precondition is inaccurate.
>>
>> Oh, and on C908, LMUL=8 is actually faster than LMUL=4. Generally
>> speaking,
>> you should maximise the LMUL unless there is a *specific* reason not to.
>>
>> --
>> レミ・デニ-クールモン
>> 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: lavc-ac3dsp-R-V-V-float_to_fixed24.patch --]
[-- Type: text/x-patch, Size: 3664 bytes --]
From 2aa06d9d8d4853ac089a13ed6a758f9ecb0aa5a9 Mon Sep 17 00:00:00 2001
From: sunyuechi <sunyuechi@iscas.ac.cn>
Date: Wed, 22 Nov 2023 14:57:29 +0800
Subject: [PATCH] lavc/ac3dsp: R-V V float_to_fixed24
c910
float_to_fixed24_c: 2207.2
float_to_fixed24_rvv_f32: 696.2
---
libavcodec/riscv/Makefile | 1 +
libavcodec/riscv/ac3dsp_init.c | 4 ++++
libavcodec/riscv/ac3dsp_rvv.S | 39 ++++++++++++++++++++++++++++++++++
3 files changed, 44 insertions(+)
create mode 100644 libavcodec/riscv/ac3dsp_rvv.S
diff --git a/libavcodec/riscv/Makefile b/libavcodec/riscv/Makefile
index 2d0e6c19c8..29a7fec455 100644
--- a/libavcodec/riscv/Makefile
+++ b/libavcodec/riscv/Makefile
@@ -2,6 +2,7 @@ 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-$(CONFIG_AC3DSP) += riscv/ac3dsp_init.o \
riscv/ac3dsp_rvb.o
+RVV-OBJS-$(CONFIG_AC3DSP) += 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..118b2955ca 100644
--- a/libavcodec/riscv/ac3dsp_init.c
+++ b/libavcodec/riscv/ac3dsp_init.c
@@ -18,6 +18,7 @@
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
+#include <stddef.h>
#include <stdint.h>
#include "config.h"
@@ -26,6 +27,7 @@
#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, size_t len);
av_cold void ff_ac3dsp_init_riscv(AC3DSPContext *c)
{
@@ -34,5 +36,7 @@ av_cold void ff_ac3dsp_init_riscv(AC3DSPContext *c)
if (flags & AV_CPU_FLAG_RVB_ADDR) {
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..b8d32c4677
--- /dev/null
+++ b/libavcodec/riscv/ac3dsp_rvv.S
@@ -0,0 +1,39 @@
+/*
+ * 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
+1:
+ vsetvli t0, a2, e32, m8, 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
+ bnez a2, 1b
+
+ ret
+endfunc
--
2.43.0
[-- 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".
^ permalink raw reply [flat|nested] 17+ messages in thread
end of thread, other threads:[~2023-12-01 20:16 UTC | newest]
Thread overview: 17+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-11-22 12:00 [FFmpeg-devel] [PATCH] ac3dsp: RISC-V V float_to_fixed24 flow gg
2023-11-22 13:40 ` Rémi Denis-Courmont
2023-11-22 14:30 ` flow gg
2023-11-22 14:35 ` flow gg
2023-11-22 14:51 ` Rémi Denis-Courmont
2023-11-22 16:37 ` flow gg
2023-11-22 16:49 ` James Almer
2023-11-22 23:17 ` flow gg
2023-11-23 17:08 ` Rémi Denis-Courmont
2023-11-23 22:39 ` flow gg
2023-12-01 18:35 ` Rémi Denis-Courmont
2023-12-01 18:38 ` Rémi Denis-Courmont
2023-12-01 19:50 ` flow gg
2023-12-01 20:16 ` flow gg
2023-11-22 17:18 ` James Almer
2023-11-22 17:34 ` flow gg
2023-11-23 7:11 ` flow gg
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