| Message ID | CAEa-L+uKuW31MEJE=o-DJ58kxPjoWENpJjKdEVfyuPTr-t8=bw@mail.gmail.com |
|---|---|
| State | New |
| Headers | show |
| Series | [FFmpeg-devel] ac3dsp: RISC-V V float_to_fixed24 | expand |
| Context | Check | Description |
|---|---|---|
| yinshiyou/make_loongarch64 | success | Make finished |
| yinshiyou/make_fate_loongarch64 | success | Make fate finished |
| andriy/make_x86 | success | Make finished |
| andriy/make_fate_x86 | success | Make fate finished |
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
> 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". >
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". >> >
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". >>
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".
>
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".
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 >
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 ...
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.
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 >
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.
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". >
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.
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.
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". >
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". >> >
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
c910 float_to_fixed24_c: 208.2 float_to_fixed24_rvv_f32: 71.5