Le 16 mai 2024 12:33:36 GMT+03:00, Andrew Sayers
a écrit :
>For example, WSAStartup()'s documentation says:
>
>"A call to the WSAGetLastError function is not needed and should not be
> used"
>---
> libavformat/network.c | 5 -
> libavformat/network.h | 12
> 2 files changed
Le torstaina 16. toukokuuta 2024, 16.09.49 EEST J. Dekker a écrit :
> Some timers on certain device and test combinations can produce noisy
> results, affecting the reliability of performance measurements. One
> notable example of this is the Canaan K230 RISC-V development board.
>
> An option to
vtype_vli computes the VTYPE value with the optimal LMUL for a given
element width, tail and mask policies and a run-time vector length.
vtype_ivli does the same, but with the compile-time constant vector
length.
vwtypei and vntypei can be used to widen or narrow a VTYPE value for
use in mixed-wi
This calculates the optimal vector type value at run-time based on the
hardware vector length and the FLAC LPC prediction order. In this
particular case, the additional computation is easily amortised over
the loop iterations:
T-Head C908:
C V before V after
1 48.0214.7 95.2
2
This is pretty much the same as for lpc16, though it only improves half
as large prediction orders. With 128-bit vectors, this gives:
C V old V new
1 69.2 181.5 95.5
2 107.7 180.7 95.2
3 145.5 180.0 103.5
4 183.0 179.2 102.7
5 220.7 178.5 128.0
6 257.7 194.0 127.5
7
T-Head C908:
add_hfyu_left_pred_bgr32_c: 237.5
add_hfyu_left_pred_bgr32_rvv_i32: 173.5 (before)
add_hfyu_left_pred_bgr32_rvv_i32: 110.0 (after)
---
libavcodec/riscv/huffyuvdsp_rvv.S | 6 --
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/libavcodec/riscv/huffyuvdsp_rvv.S
b
Le torstaina 16. toukokuuta 2024, 19.48.40 EEST Rémi Denis-Courmont a écrit :
> T-Head C908:
> add_hfyu_left_pred_bgr32_c: 237.5
> add_hfyu_left_pred_bgr32_rvv_i32: 173.5 (before)
> add_hfyu_left_pred_bgr32_rvv_i32: 110.0 (after)
> ---
> libavcodec/riscv/huffyuvdsp_rvv
Le 17 mai 2024 11:34:35 GMT+03:00, Sfan5 a écrit :
>Signed-off-by: sfan5
>---
> libavformat/tls_mbedtls.c | 14 ++
> 1 file changed, 14 insertions(+)
>
>diff --git a/libavformat/tls_mbedtls.c b/libavformat/tls_mbedtls.c
>index 24c3afd94c..9508fe3436 100644
>--- a/libavformat/tls_mbed
Le 17 mai 2024 16:49:58 GMT+03:00, Michael Niedermayer
a écrit :
>Hi all
>
>Before this is forgotten again, better start some dicsussion too early than
>too late
>
>I propose that if we have the oppertunity again next year to receive a grant
>from STF. That we use it to fund:
>
>* Paul to work
Le maanantaina 13. toukokuuta 2024, 19.59.22 EEST u...@foxmail.com a écrit :
> From: sunyuechi
>
> C908:
> vp9_avg4_8bpp_c: 1.2
> vp9_avg4_8bpp_rvv_i64: 1.0
> vp9_avg8_8bpp_c: 3.7
> vp9_avg8_8bpp_rvv_i64: 1.5
> vp9_avg16_8bpp_c: 14.7
> vp9_avg16_8bpp_rvv_i64: 3.5
> vp9_avg32_8bpp_c: 57.7
> vp9_av
vcodec/riscv/vc1dsp_rvv.S
+++ b/libavcodec/riscv/vc1dsp_rvv.S
@@ -1,5 +1,6 @@
/*
* Copyright (c) 2023 Institue of Software Chinese Academy of Sciences (ISCAS).
+ * Copyright (c) 2024 Rémi Denis-Courmont.
*
* This file is part of FFmpeg.
*
@@ -159,3 +160,55 @@ func ff_avg_pixels8x8_rvv,
Le 17 mai 2024 21:50:20 GMT+03:00, Michael Niedermayer
a écrit :
>Let me reword this for 2024:
>The people on the booth are predominantly male. Similarly ffmpeg-devel
>is predominantly male. More gender diversity would be good.
As you've noted, this is an obvious problem but not an FFmpeg-spec
tcode_find_candidate_rvv;
+# endif
#endif
}
diff --git a/libavcodec/riscv/startcode_rvv.S b/libavcodec/riscv/startcode_rvv.S
new file mode 100644
index 00..7c43b1d7f3
--- /dev/null
+++ b/libavcodec/riscv/startcode_rvv.S
@@ -0,0 +1,44 @@
+/*
+ * Copyright © 2024 Rémi Denis-Courmont.
+ *
+ * R
---
libavcodec/riscv/vp8dsp_rvv.S | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/libavcodec/riscv/vp8dsp_rvv.S b/libavcodec/riscv/vp8dsp_rvv.S
index 4d7a9f6a2d..0ba9fa443d 100644
--- a/libavcodec/riscv/vp8dsp_rvv.S
+++ b/libavcodec/riscv/vp8dsp_rvv.S
@@ -171,7 +171,7 @@ en
Wrong patch sent, disregard
--
Rémi Denis-Courmont
http://www.remlab.net/
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-requ
---
libavcodec/riscv/vp9_intra_rvv.S | 28 ++--
1 file changed, 14 insertions(+), 14 deletions(-)
diff --git a/libavcodec/riscv/vp9_intra_rvv.S b/libavcodec/riscv/vp9_intra_rvv.S
index 280c497687..7b3c22b99a 100644
--- a/libavcodec/riscv/vp9_intra_rvv.S
+++ b/libavcodec/ri
Le maanantaina 13. toukokuuta 2024, 19.59.23 EEST u...@foxmail.com a écrit :
> From: sunyuechi
>
> C908:
> vp9_avg_bilin_4h_8bpp_c: 5.2
> vp9_avg_bilin_4h_8bpp_rvv_i64: 2.2
> vp9_avg_bilin_4v_8bpp_c: 5.5
> vp9_avg_bilin_4v_8bpp_rvv_i64: 2.2
> vp9_avg_bilin_8h_8bpp_c: 20.0
> vp9_avg_bilin_8h_8bpp_
_check_h264chroma(void);
void checkasm_check_h264dsp(void);
void checkasm_check_h264pred(void);
diff --git a/tests/checkasm/h263dsp.c b/tests/checkasm/h263dsp.c
new file mode 100644
index 00..2d0957a90b
--- /dev/null
+++ b/tests/checkasm/h263dsp.c
@@ -0,0 +1,62 @@
+/*
+ * Copyright (c) 2024 Ré
Hi,
Le 18 mai 2024 21:55:04 GMT+03:00, Cosmin Stejerean via ffmpeg-devel
a écrit :
>
>
>> On May 18, 2024, at 7:04 AM, Nuo Mi wrote:
>>
>> This happened many years ago. See the discussion here:
>> https://patchwork.ffmpeg.org/project/ffmpeg/patch/20201221060710.12230-6-nuomi2...@gmail.com/#605
Le sunnuntaina 19. toukokuuta 2024, 11.49.52 EEST Andreas Rheinhardt a écrit :
> Fixes fate-source.
>
> Signed-off-by: Andreas Rheinhardt
> ---
> An alternative would be to change the license header, but this would
> have to be done by courmisch himself.
A better alternative would be to fix the
git a/libavcodec/riscv/h263dsp_init.c b/libavcodec/riscv/h263dsp_init.c
new file mode 100644
index 00..21b536366c
--- /dev/null
+++ b/libavcodec/riscv/h263dsp_init.c
@@ -0,0 +1,41 @@
+/*
+ * Copyright © 2022 Rémi Denis-Courmont.
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free soft
Le 20 mai 2024 12:33:41 GMT+03:00, Andreas Rheinhardt
a écrit :
>Michael Niedermayer:
>> Fixes: CID1583742 Unchecked return value
>>
>> Sponsored-by: Sovereign Tech Fund
>> Signed-off-by: Michael Niedermayer
>> ---
>> libavutil/hwcontext_drm.c | 5 -
>> 1 file changed, 4 insertions(+), 1
Le maanantaina 20. toukokuuta 2024, 19.33.43 EEST Cosmin Stejerean via ffmpeg-
devel a écrit :
> > Adding a disabled-by-default decoder is adding bloat, if there is not a
> > specific known reason why that is needed.
> > This is especially true for video decoders, where the external decoders
> >
Le sunnuntaina 19. toukokuuta 2024, 14.29.43 EEST Thilo Borgmann via ffmpeg-
devel a écrit :
> [...]
>
> >> * Fund administrative / maintainance work (one example is the mailman
> >> upgrade that is needed>>
> >> with the next OS upgrade on one of our servers (this is not as trivial
> >> as o
Le maanantaina 20. toukokuuta 2024, 21.39.18 EEST Cosmin Stejerean via ffmpeg-
devel a écrit :
> The same way using FDK-AAC as a decoder works, if you want to use it as the
> *AAC decoder you have to specify the decoder with -c:a libfdk_aac before
> the input.-
AFAIK, we don't have hwaccel for aud
Le maanantaina 20. toukokuuta 2024, 22.33.45 EEST Cosmin Stejerean via ffmpeg-
devel a écrit :
> > And again, you can't expect users to select decoders manually. If vvdec is
> > the
> > default, hwaccel won't work. If vvdec is not the default, then it's
> > dead code.
>
> Not sure why you keep re
Hi,
Le 20 mai 2024 03:42:03 GMT+03:00, Stone Chen a
écrit :
>Adds checkasm for DMVR SAD AVX2 implementation.
>
>Benchmarks ( AMD 7940HS )
>vvc_sad_8x8_c: 70.0
>vvc_sad_8x8_avx2: 10.0
>vvc_sad_16x16_c: 280.0
>vvc_sad_16x16_avx2: 20.0
>vvc_sad_32x32_c: 1020.0
>vvc_sad_32x32_avx2: 70.0
>vvc_sad_64x
Le 21 mai 2024 09:37:18 GMT+03:00, "Martin Storsjö" a écrit :
>On Tue, 21 May 2024, Rémi Denis-Courmont wrote:
>
>> Hi,
>>
>> Le 20 mai 2024 03:42:03 GMT+03:00, Stone Chen a
>> écrit :
>>> Adds checkasm for DMVR SAD AVX2 implementation.
>
Le 21 mai 2024 13:04:29 GMT+03:00, "Martin Storsjö" a écrit :
>Don't benchmark every single combination of widths and heights;
>only benchmark cases which are squares (like in vvc_mc.c).
>
>Contrary to vvc_mc, which increases sizes by doubling dimensions,
>vvc_alf tests all sizes in increments o
Le 21 mai 2024 14:04:53 GMT+03:00, "Martin Storsjö" a écrit :
>The loop filters can write before the pointer given to them;
>the actual test invocations correctly used an offset, while
>the benchmark calls were lacking an offset. Therefore, when
>running with benchmarking, these tests could have
Le 21 mai 2024 13:14:59 GMT+03:00, sfan5 a écrit :
>Am 17.05.24 um 11:51 schrieb Rémi Denis-Courmont:
>> Le 17 mai 2024 11:34:35 GMT+03:00, Sfan5 a écrit :
>>> Signed-off-by: sfan5
>>> ---
>>> libavformat/tls_mbedtls.c | 14 ++
>>> 1
\
> +dsp->mc[idx1][FILTER_8TAP_SMOOTH ][1][0][0] = ff_avg##sz##_rvv; \
> +dsp->mc[idx1][FILTER_8TAP_REGULAR][1][0][0] = ff_avg##sz##_rvv; \
> +dsp->mc[idx1][FILTER_8TAP_SHARP ][1][0][0] = ff_avg##sz##_rvv; \
> +dsp->mc[idx1][FILTER_B
his. All non-static symbols in
libavcodec are in the same namespace.
> Many macros in the C file are copied from MIPS, where
> it is called ff_avg4_msa. Here, it has been simply changed to ff_avg4_rvv.
AArch64 uses ff_vp9_avg which seems a lot saner.
--
Rémi Den
t_riscv(VVCDSPContext *const c, const int bd)
> +{
> +#if HAVE_RVV
> +const int flags = av_get_cpu_flags();
> +
> +if ((flags & AV_CPU_FLAG_RVV_I32) && (flags & AV_CPU_FLAG_RVB_ADDR) &&
> +ff_rv_vlen_least(256)) {
> +switch (bd) {
>
The code is already there, we just need to use it.
get_pixels_unaligned_c: 2.2
get_pixels_unaligned_misaligned: 1.7
---
libavcodec/riscv/pixblockdsp_init.c | 7 +++
1 file changed, 7 insertions(+)
diff --git a/libavcodec/riscv/pixblockdsp_init.c
b/libavcodec/riscv/pixblockdsp_init.c
index b
Otherwise V functions mask scalar misaligned ones.
---
tests/checkasm/checkasm.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tests/checkasm/checkasm.c b/tests/checkasm/checkasm.c
index 31ca9f6e2b..76835ab267 100644
--- a/tests/checkasm/checkasm.c
+++ b/tests/checkasm/checka
vv.S
> @@ -53,6 +53,49 @@ func ff_vp9_avg\len\()_rvv, zve32x
> endfunc
> .endm
>
> +.macro bilin_load dst len op type mn
Commas, please.
--
Rémi Denis-Courmont
http://www.remlab.net/
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.
Le tiistaina 21. toukokuuta 2024, 21.43.44 EEST Thilo Borgmann via ffmpeg-devel
a écrit :
> >> Same as above about that we should and STF would.
> >> Especially since no corporate interest usually pays anyone for these
> >> tasks
> >
> > Sadly true, but...
> >
> >> (in case of reviews it might o
Le tiistaina 21. toukokuuta 2024, 22.42.00 EEST Rémi Denis-Courmont a écrit :
> And "I hope you realise that you are arguing for" Intel, Loongson, etc.
> employees to stop reviewing patches.
P.S.: And FFlabs too, since it is a for-profit company.
--
レミ・デニ-クールモン
http:
Le 22 mai 2024 00:34:03 GMT+03:00, Thilo Borgmann via ffmpeg-devel
a écrit :
>>> I hope you realize what you argue in favor of.
>>
>> Yes. It's quoted above.
>>
>> Are you claiming that *no* review is better than *some* review done in
>> *public* for all to see by a paid professional just bec
Le 22 mai 2024 14:04:50 GMT+03:00, Andreas Rheinhardt
a écrit :
>Several files already had standard license header (namely
>2-clause BSD files), yet due to the 80 char line length limit,
>they were not treated as such by source-check.sh (which
>fate-source uses). Therefore relax the BSD check.
This removes one stray LI and reworks the vector arithmetic to avoid
changing the vector configuration. On K230, this takes the 46.5 cycle
count down from 46.5 to 43.5.
---
libavcodec/riscv/rv34dsp_rvv.S | 13 ++---
1 file changed, 6 insertions(+), 7 deletions(-)
diff --git a/libavcodec/r
Le 22 mai 2024 23:28:54 GMT+03:00, "Rémi Denis-Courmont" a
écrit :
>This removes one stray LI and reworks the vector arithmetic to avoid
>changing the vector configuration. On K230, this takes the 46.5 cycle
>count down from 46.5 to 43.5.
>---
> libavcodec
there :-(
FWIW, beware that people will tend to assume that you are a professional if you
assign copyright to a legal entity, even if there is another nonobvious reason
why you do that.
>Rémi Denis-Courmont 于2024年5月23日周四 04:35写道：
>
>>
>>
>> Le 22 mai 2024 23:28:54 GMT+0
---
libavcodec/riscv/rv34dsp_rvv.S | 1 -
1 file changed, 1 deletion(-)
diff --git a/libavcodec/riscv/rv34dsp_rvv.S b/libavcodec/riscv/rv34dsp_rvv.S
index f1f6345012..8eda01665d 100644
--- a/libavcodec/riscv/rv34dsp_rvv.S
+++ b/libavcodec/riscv/rv34dsp_rvv.S
@@ -37,7 +37,6 @@ func ff_rv34_idct_dc
---
libavcodec/riscv/vp8dsp_rvv.S | 8
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/libavcodec/riscv/vp8dsp_rvv.S b/libavcodec/riscv/vp8dsp_rvv.S
index 0ba9fa443d..9c84b1503e 100644
--- a/libavcodec/riscv/vp8dsp_rvv.S
+++ b/libavcodec/riscv/vp8dsp_rvv.S
@@ -94,7 +94,7 @@
Le 24 mai 2024 10:11:49 GMT+03:00, Thilo Borgmann via ffmpeg-devel
a écrit :
>Unfortunately true, yet you argue to pay more companies to do reviews instead
>having reviews funded by unbiased means.
That is a blatant and libelous lie and will be reported to CC promptly.
___
Le 24 mai 2024 09:44:50 GMT+03:00, Andreas Rheinhardt
a écrit :
>Rémi Denis-Courmont:
>> ---
>> libavcodec/riscv/rv34dsp_rvv.S | 1 -
>> 1 file changed, 1 deletion(-)
>>
>> diff --git a/libavcodec/riscv/rv34dsp_rvv.S b/libavcodec/riscv/rv34dsp_rvv.S
>&
This will simplify later changes.
---
tests/checkasm/vp8dsp.c | 82 +
1 file changed, 34 insertions(+), 48 deletions(-)
diff --git a/tests/checkasm/vp8dsp.c b/tests/checkasm/vp8dsp.c
index 6e989819ef..12d2f62f2d 100644
--- a/tests/checkasm/vp8dsp.c
+++ b/te
---
tests/checkasm/vp8dsp.c | 57 ++---
1 file changed, 36 insertions(+), 21 deletions(-)
diff --git a/tests/checkasm/vp8dsp.c b/tests/checkasm/vp8dsp.c
index 12d2f62f2d..e448fe89fc 100644
--- a/tests/checkasm/vp8dsp.c
+++ b/tests/checkasm/vp8dsp.c
@@ -18,8 +18
This loop correctly assumes that VLMAX=16 (4x128-bit vectors
with 32-bit elements) and 32 >= pred_order > 16. We need to alternate
between VL=16 and VL=t2=pred_order-16 elements to add up to pred_order.
The current code requests AVL=a2=pred_order elements. In QEMU and on
thte K230 hardware, this s
flac_lpc_33_13_c: 499.7
flac_lpc_33_13_rvv_i64: 197.7
flac_lpc_33_16_c: 601.5
flac_lpc_33_16_rvv_i64: 195.2
flac_lpc_33_29_c: 1011.5
flac_lpc_33_29_rvv_i64: 300.7
flac_lpc_33_32_c: 1099.0
flac_lpc_33_32_rvv_i64: 296.7
---
libavcodec/riscv/flacdsp_init.c | 7 +--
libavcodec/riscv/flacdsp_rvv.S
Le perjantaina 24. toukokuuta 2024, 23.04.40 EEST Rémi Denis-Courmont a écrit
:
> flac_lpc_33_13_c: 499.7
> flac_lpc_33_13_rvv_i64: 197.7
> flac_lpc_33_16_c: 601.5
> flac_lpc_33_16_rvv_i64: 195.2
> flac_lpc_33_29_c: 1011.5
> flac_lpc_33_29_rvv_i64: 300.7
> fla
Hi,
Sorry for the delay, this is not my day job.
Le tiistaina 21. toukokuuta 2024, 22.24.21 EEST u...@foxmail.com a écrit :
> From: sunyuechi
> ---
> libavcodec/riscv/Makefile | 2 +
> libavcodec/riscv/vvc_mc_rvv.S | 313 +
> libavcodec/riscv/vvcd
nit_fpel
> +
> +#define init_subpel1(idx1, idx2, idxh, idxv, sz, dir, type, vlen) \
> +dsp->mc[idx1][FILTER_8TAP_SMOOTH ][idx2][idxh][idxv] = \
> +ff_##type##_vp9_8tap_smooth_##sz##dir##_rvv##vlen; \
> +dsp->mc[idx1][FILTER_8TAP_REGULAR][idx2][idxh][idxv] = \
> +
Le lauantaina 25. toukokuuta 2024, 14.18.12 EEST Michael Niedermayer a écrit :
> Fixes: CID1559855 Unchecked return value from library
This is again very misleading. Logging an error (that most people won't even
see) is not fixing a bug. Either this is not a bug, and this patch can't be
fixing a
--
Change since previous version:
- Disable test on x86, since the x86 optimisation fails
with a suspiciously large and regular error with 10-20% probability.
---
tests/checkasm/lpc.c | 54 ++--
1 file changed, 52 insertions(+), 2 deletions(-)
diff --git
Le lauantaina 25. toukokuuta 2024, 15.19.51 EEST Michael Niedermayer a écrit :
> On Sat, May 25, 2024 at 02:38:07PM +0300, Rémi Denis-Courmont wrote:
> > Le lauantaina 25. toukokuuta 2024, 14.18.12 EEST Michael Niedermayer a
écrit :
> > > Fixes: CID1559855 Unchecked return
---
libavcodec/riscv/vp8dsp_rvv.S | 24 ++--
1 file changed, 6 insertions(+), 18 deletions(-)
diff --git a/libavcodec/riscv/vp8dsp_rvv.S b/libavcodec/riscv/vp8dsp_rvv.S
index cb9b0b8b5f..bb0c7bf02a 100644
--- a/libavcodec/riscv/vp8dsp_rvv.S
+++ b/libavcodec/riscv/vp8dsp_rvv.S
Use shifts rather than multiply, and save one instruction.
---
libavcodec/riscv/vp8dsp_init.c | 26 ++
libavcodec/riscv/vp8dsp_rvv.S | 7 +++
2 files changed, 17 insertions(+), 16 deletions(-)
diff --git a/libavcodec/riscv/vp8dsp_init.c b/libavcodec/riscv/vp8dsp_init
This saves one instruction and frees up A5, which will be repurposed in
later changes. Unfortunately, we need to add quite a lot of alternative
code for this.
---
libavcodec/riscv/vp8dsp_rvv.S | 24
1 file changed, 16 insertions(+), 8 deletions(-)
diff --git a/libavcodec/
For a given type, only the first VSETVLI instruction varies depending
on the size.
---
libavcodec/riscv/vp8dsp_rvv.S | 37 +--
1 file changed, 27 insertions(+), 10 deletions(-)
diff --git a/libavcodec/riscv/vp8dsp_rvv.S b/libavcodec/riscv/vp8dsp_rvv.S
index bb0c7bf
---
libavcodec/riscv/vp8dsp_rvv.S | 56 ---
1 file changed, 32 insertions(+), 24 deletions(-)
diff --git a/libavcodec/riscv/vp8dsp_rvv.S b/libavcodec/riscv/vp8dsp_rvv.S
index a4fcd158a5..002e7f3174 100644
--- a/libavcodec/riscv/vp8dsp_rvv.S
+++ b/libavcodec/riscv/v
Le lauantaina 18. toukokuuta 2024, 21.44.10 EEST Rémi Denis-Courmont a écrit :
> ---
> tests/checkasm/Makefile | 1 +
> tests/checkasm/checkasm.c | 3 ++
> tests/checkasm/checkasm.h | 1 +
> tests/checkasm/h263dsp.c | 62 +++
> tests
C uses 16-bit
values for no apparent reasons. We should probably change the C code to use 8-
bit on all platforms - except AArch64, because the NEON code probably relies
on 16-bit format.
--
Rémi Denis-Courmont
http://www.remlab.net/
___
ffmpeg
This results in AUIPC; ADDI instead of AUIPC; ADDI; ... ADDI.
---
libavcodec/riscv/sbrdsp_rvv.S | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/libavcodec/riscv/sbrdsp_rvv.S b/libavcodec/riscv/sbrdsp_rvv.S
index aba9a28108..331b88022c 100644
--- a/libavcodec/riscv/sbrdsp_rvv.
128-it is the maximum, not the minimum here. Larger vector sizes can
result in reads past the end of the noise value table.
This partially reverts commit cdcb4b98b7f74d87a6274899ff70724795d551cb.
---
libavcodec/riscv/sbrdsp_init.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git
Le lauantaina 25. toukokuuta 2024, 20.52.44 EEST Rémi Denis-Courmont a écrit :
> 128-it is the maximum, not the minimum here. Larger vector sizes can
> result in reads past the end of the noise value table.
Nit: bit
>
> This partially reverts commit cdcb4b98b7f74d87a6274899ff70
Le lauantaina 25. toukokuuta 2024, 21.16.22 EEST flow gg a écrit :
> Would it be better to replace the two vsetvlstatic8 and vsetvlstatic16 with
> two vsetvl?
The other option is to hard-code the most pessimistic multiplier. That would
be easier to read and save two instructions in the head, it w
Since the SBR noise application optimisations are currently restricted
to hardware with 128-bit vectors, and use a quadruple multipler, they
can load up to 16 32-bit elements. But the "loads" are of 2 segments,
or 16 pairs of single precision float.
Thus we need to expand the dupiclated section of
hf_apply_noise_0_c: 35.7
hf_apply_noise_0_rvv_f32: 9.5
hf_apply_noise_1_c: 38.5
hf_apply_noise_1_rvv_f32: 10.0
hf_apply_noise_2_c: 35.5
hf_apply_noise_2_rvv_f32: 9.7
hf_apply_noise_3_c: 38.5
hf_apply_noise_3_rvv_f32: 10.0
Maybe extending the noise table manually is not such great idea, but I
not q
Le lauantaina 25. toukokuuta 2024, 21.56.58 EEST flow gg a écrit :
> The reduction in code size seems to be due to switching to using j labels,
> doesn't seem to be about vset, but another issue. j labels are indeed
> better. I will make similar modifications.
If we don't use vsetvl, then we have
On SpacemiT X60 (with len == 4000):
autocorr_10_c: 2303.7
autocorr_10_rvv_f64: 1411.5 (before)
autocorr_10_rvv_f64: 842.2 (after)
---
libavcodec/riscv/lpc_init.c | 3 ++-
libavcodec/riscv/lpc_rvv.S | 5 +++--
2 files changed, 5 insertions(+), 3 deletions(-)
diff --git a/libavcodec/riscv/l
Le sunnuntaina 26. toukokuuta 2024, 1.31.18 EEST James Almer a écrit :
> On 5/25/2024 5:57 PM, Lynne via ffmpeg-devel wrote:
> > The inline asm function had issues running under checkasm.
> > So I came to finish what I started, and wrote the last part
> > of LPC computation in assembly.
> >
> > au
This is not great as transposition is poorly supported, but it works:
vp8_luma_dc_wht_c: 2.5
vp8_luma_dc_wht_rvv_i32: 1.7
---
libavcodec/riscv/vp8dsp_init.c | 2 ++
libavcodec/riscv/vp8dsp_rvv.S | 55 ++
2 files changed, 57 insertions(+)
diff --git a/libavc
Le sunnuntaina 26. toukokuuta 2024, 10.29.02 EEST Rémi Denis-Courmont a écrit
:
> diff --git a/libavcodec/riscv/vp8dsp_init.c b/libavcodec/riscv/vp8dsp_init.c
> index 2413fbf449..d48fe08560 100644
> --- a/libavcodec/riscv/vp8dsp_init.c
> +++ b/libavcodec/riscv/vp8dsp_init.c
>
-OBJS-$(CONFIG_VP8DSP) += riscv/vp8dsp_rvv.o
diff --git a/libavcodec/riscv/vp7dsp_init.c b/libavcodec/riscv/vp7dsp_init.c
new file mode 100644
index 00..6d9aba43d9
--- /dev/null
+++ b/libavcodec/riscv/vp7dsp_init.c
@@ -0,0 +1,41 @@
+/*
+ * Copyright (c) 2024 Rémi Denis-Courmont.
+ *
+ * This
This prepares for some code reuse.
---
libavcodec/riscv/vp7dsp_init.c | 3 ++-
libavcodec/riscv/vp7dsp_rvv.S | 23 ++-
2 files changed, 16 insertions(+), 10 deletions(-)
diff --git a/libavcodec/riscv/vp7dsp_init.c b/libavcodec/riscv/vp7dsp_init.c
index 5a57a39d0b..1406eb1c6a
Most of the code is shared with DC, thanks to minor earlier changes.
vp7_idct_add_c: 5.2
vp7_idct_add_rvv_i32: 2.5
---
libavcodec/riscv/vp7dsp_init.c | 2 ++
libavcodec/riscv/vp7dsp_rvv.S | 29 +
2 files changed, 31 insertions(+)
diff --git a/libavcodec/riscv/
RV128 is mostly scifi at this point, so we can just disable it here
(the EPEL HV prologue/epilogue do not save 128-bit registers).
---
libavcodec/riscv/vp8dsp_init.c | 3 ++-
libavcodec/riscv/vp8dsp_rvv.S | 2 ++
2 files changed, 4 insertions(+), 1 deletion(-)
diff --git a/libavcodec/riscv/vp8ds
Le lauantaina 25. toukokuuta 2024, 19.30.44 EEST James Almer a écrit :
> On 5/18/2024 3:44 PM, Rémi Denis-Courmont wrote:
> > ---
> >
> > tests/checkasm/Makefile | 1 +
> > tests/checkasm/checkasm.c | 3 ++
> > tests/checkasm/checkasm.h | 1 +
>
atches.
You can revert the patch and then provide a better version in a patch series.
However it may be prefereable to squash the revert and the improved version
(and describe the improvements in the commit log).
--
Rémi Denis-Courmont
http://ww
Le tiistaina 21. toukokuuta 2024, 21.36.33 EEST flow gg a écrit :
> Do macros definition also need a comma?
Indeed not. I'd rather be safe than sorry though.
--
レミ・デニ-クールモン
http://www.remlab.net/
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
ht
Le lauantaina 25. toukokuuta 2024, 14.58.53 EEST Rémi Denis-Courmont a écrit :
> --
> Change since previous version:
> - Disable test on x86, since the x86 optimisation fails
> with a suspiciously large and regular error with 10-20% probability.
> ---
> tests/c
The generally expected behaviour is for software to append from the
current position, not to keep resetting. See also how shell works:
# (foo && bar) > file
...would normally concatenate the output of both commands to file.
If foo seeks like lavf, bar would instead overwrite the output of
foo (po
---
libavcodec/riscv/lpc_init.c | 2 +-
libavcodec/riscv/lpc_rvv.S | 1 +
2 files changed, 2 insertions(+), 1 deletion(-)
diff --git a/libavcodec/riscv/lpc_init.c b/libavcodec/riscv/lpc_init.c
index f21eca4caa..d9bcbbe604 100644
--- a/libavcodec/riscv/lpc_init.c
+++ b/libavcodec/riscv/lpc_init.c
A constant (-1) is added to the length value, so we can have an added
for free, and optimise the addition away if the addend is exactly 1.
---
libavcodec/riscv/lpc_rvv.S | 2 +-
libavutil/riscv/asm.S | 9 ++---
2 files changed, 7 insertions(+), 4 deletions(-)
diff --git a/libavcodec/risc
---
Changes since v2:
- Scale the error factor to length since this computes sums.
- Check the last element from results.
- Use fixed vector size for benchmarks.
---
tests/checkasm/lpc.c | 51 +---
1 file changed, 48 insertions(+), 3 deletions(-)
diff --gi
Le maanantaina 27. toukokuuta 2024, 19.01.59 EEST Rémi Denis-Courmont a écrit
:
> ---
> Changes since v2:
> - Scale the error factor to length since this computes sums.
> - Check the last element from results.
> - Use fixed vector size for benchmarks.
- Test real use lag values (f
If there are no zero bytes, t2 equals -1. The code cannot simply fall
through to the match case.
---
libavcodec/riscv/startcode_rvv.S | 2 ++
1 file changed, 2 insertions(+)
diff --git a/libavcodec/riscv/startcode_rvv.S b/libavcodec/riscv/startcode_rvv.S
index 7c43b1d7f3..36a3369431 100644
--- a/
Le maanantaina 27. toukokuuta 2024, 22.15.40 EEST James Almer a écrit :
> On 5/27/2024 4:10 PM, James Almer wrote:
> > On 5/27/2024 1:01 PM, Rémi Denis-Courmont wrote:
> >> ---
> >> Changes since v2:
> >> - Scale the error factor to length since this computes
Le 28 mai 2024 06:07:41 GMT+03:00, James Almer a écrit :
>On 5/27/2024 11:49 PM, Andreas Rheinhardt wrote:
>> These strings are so short that they can be put directly
>> into the containing structure, avoiding the pointer
>> and putting it into .rodata.
>> Also use chars for interlaced and level
---
tests/checkasm/Makefile | 1 +
tests/checkasm/checkasm.c | 1 +
tests/checkasm/checkasm.h | 1 +
tests/checkasm/lls.c | 110 ++
tests/fate/checkasm.mak | 1 +
5 files changed, 114 insertions(+)
create mode 100644 tests/checkasm/lls.c
diff
The function pointer is appended to the structure for backward binary
compatibility. Fortunately, this is allocated by libavutil, not by the
user, so increasing the structure size is safe.
---
doc/APIchanges| 3 +++
libavutil/float_dsp.c | 12
libavutil/float_dsp.h | 14 +
---
libavfilter/aap_template.c | 14 +-
libavfilter/anlms_template.c | 16 ++--
libavfilter/arls_template.c | 14 +-
3 files changed, 4 insertions(+), 40 deletions(-)
diff --git a/libavfilter/aap_template.c b/libavfilter/aap_template.c
index ea9c815a89..0e05
---
tests/checkasm/float_dsp.c | 19 +++
1 file changed, 19 insertions(+)
diff --git a/tests/checkasm/float_dsp.c b/tests/checkasm/float_dsp.c
index cadfa65e2a..296db1cff9 100644
--- a/tests/checkasm/float_dsp.c
+++ b/tests/checkasm/float_dsp.c
@@ -278,6 +278,22 @@ static void tes
C908:
scalarproduct_double_c: 39.2
scalarproduct_double_rvv_f64: 10.5
X60:
scalarproduct_double_c: 35.0
scalarproduct_double_rvv_f64: 5.2
---
libavutil/riscv/float_dsp_init.c | 3 +++
libavutil/riscv/float_dsp_rvv.S | 21 +
2 files changed, 24 insertions(+)
dif
til/riscv/lls_init.c b/libavutil/riscv/lls_init.c
new file mode 100644
index 00..b1317dc925
--- /dev/null
+++ b/libavutil/riscv/lls_init.c
@@ -0,0 +1,57 @@
+/*
+ * Copyright © 2024 Rémi Denis-Courmont.
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute
---
tests/checkasm/Makefile | 1 +
tests/checkasm/checkasm.c | 1 +
tests/checkasm/checkasm.h | 1 +
tests/checkasm/lls.c | 101 ++
tests/fate/checkasm.mak | 1 +
5 files changed, 105 insertions(+)
create mode 100644 tests/checkasm/lls.c
diff
alignment beyond that imposed by
> double?
The C and RISC-V implementations require the natural alignment of double, 64-
bit. If somebody wants to increase the required alignment(s), they are free to
audit the call sites and update the comments accordingly, since this won't be
an ABI br
1 - 100 of 1371 matches
Mail list logo
