1
0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2025-01-19 05:49:09 +02:00

518 Commits

Author SHA1 Message Date
Zhao Zhili
b1240c983f tests/checkasm: Fix build error when enable linux perf on Android
B0 is defined by system header, see f0f596dbc6b for ref.

Reviewed-by: Martin Storsjö <martin@martin.st>
Signed-off-by: Zhao Zhili <zhilizhao@tencent.com>
2024-06-11 01:11:46 +08:00
James Almer
287d139b77 checkasm/sw_rgb: fix alignment of buffers for rgb_to_yuv tests
src is apparently not guaranteed to be >8 byte aligned, but align to 16
nonetheless as the x86 asm will do unaligned loads anyway.
dst is guaranteed to be 32 byte aligned for the Y plane, but 16 byte for UV.

Signed-off-by: James Almer <jamrial@gmail.com>
2024-06-09 14:12:51 -03:00
James Almer
6743c2fc6a checkasm/sw_rgb: test rgb32/rgb32_1 to yuv
Test all four pixel formats, but only bench the two native endian ones for a
given target.

Signed-off-by: James Almer <jamrial@gmail.com>
2024-06-09 12:29:49 -03:00
James Almer
91b9af0058 x86/aacencdsp: add AVX version of quantize_bands
quant_bands_signed_c: 1928.0
quant_bands_signed_sse2: 406.0
quant_bands_signed_avx: 207.0
quant_bands_unsigned_c: 1702.0
quant_bands_unsigned_sse2: 404.0
quant_bands_unsigned_avx: 209.0

Signed-off-by: James Almer <jamrial@gmail.com>
2024-06-09 12:29:49 -03:00
Andreas Rheinhardt
fca796ac3b tests/checkasm/sw_rgb: Be more strict about clobbering MMX state
The MMXEXT versions of the rgb2rgb functions tested here
always emit emms on their own. Therefore one can use
a stricter test to ensure that it stays that way.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-06-09 12:03:47 +02:00
Rémi Denis-Courmont
8d117024fe checkasm: disable unaligned access emulation
The OS may silently fix (emulate) unaligned hardware access exceptions.
This is extremely slow and code should be fixed not to rely on unaligned
access on affected hardware. Accordingly this requests that the OS
disable emulation and instead throw Bus error, which will be caught by
checkasm's signal handler.

This has no effects if the hardware supports unaligned access in
hardware, since no exceptions are generated. prctl() will fail safe in
that case.
2024-06-07 17:53:05 +03:00
Zhao Zhili
47ba87551c checkasm/sw_rgb: test rgb24/bgr24 to yuv
The line width 8 is supposed to test corner case, while the
performance doesn't matter. Width 1080 is also a case of
unaligned to 16.

Width 1920 meant for benchmark (together with --runs options).

Signed-off-by: James Almer <jamrial@gmail.com>
2024-06-05 15:22:49 -03:00
Anton Khirnov
e4601cc339 lavc/hevc*: move to hevc/ subdir 2024-06-04 11:46:27 +02:00
Rémi Denis-Courmont
be6f8c439a checkasm: add aacencdsp.quant_bands test 2024-06-03 22:43:37 +03:00
Rémi Denis-Courmont
fc85aff72f checkasm: add linear least square tests 2024-06-01 18:05:58 +03:00
James Almer
0a949aacae checkasm/lpc: use fixed length to bench apply_welch_window
Signed-off-by: James Almer <jamrial@gmail.com>
2024-05-31 17:06:08 -03:00
Rémi Denis-Courmont
16132a810d checkasm/lpc: test compute_autocorr
Signed-off-by: James Almer <jamrial@gmail.com>
2024-05-31 16:36:43 -03:00
Rémi Denis-Courmont
98405d28fa checkasm/float_dsp: add double-precision scalar product 2024-05-31 22:22:43 +03:00
James Almer
4008a80c1b tests/checkasm/vvc_mc: don't zero the SAD buffers
They will be filled immediately after.

Signed-off-by: James Almer <jamrial@gmail.com>
2024-05-31 20:05:21 +08:00
James Almer
b70289f354 tests/checkasm/vvc_mc: fix indentation
Signed-off-by: James Almer <jamrial@gmail.com>
2024-05-31 20:05:21 +08:00
Wu Jianhua
442e94e5e4 tests/checkasm/vvc_alf: change alf step size to 8
From Benjamin Bross:
> for ALF where functions are in increments of 4 while 8 should be sufficient according to the spec.

Signed-off-by: Wu Jianhua <toqsxw@outlook.com>
2024-05-31 19:57:31 +08:00
Rémi Denis-Courmont
8a96495fef checkasm/vp8dsp: add VP7 tests 2024-05-30 18:30:52 +03:00
Rémi Denis-Courmont
a3d4c73b4e checkasm/vp8dsp: share DSP context across tests
This will simplify later changes.
2024-05-30 18:30:52 +03:00
Stone Chen
d82c503555 tests/checkasm/vvc_mc: for SAD, only test valid subblock sizes
According to the VVC specification (section 8.5.1), the maximum width/height of a subblock passed for DMVR SAD is 16. This along with previous constraint requiring width * height >= 128 means that  8x16, 16x8, and 16x16 are the only allowed sizes.

This changes check_vvc_sad() to only test and benchmark those sizes.
2024-05-29 21:35:34 +08:00
Rémi Denis-Courmont
44f7f6e010 checkasm: add h263dsp.{h,v}_loop_filter 2024-05-27 22:42:07 +03:00
Andreas Rheinhardt
e9197db4f7 tests/checkasm/vvc_alf: Don't use declare_func_emms
VVC does not have MMX code at all, so one can use the stricter
declare_func to also check that the MMX state has not been clobbered
with (which would be an ABI violation).

Reviewed-by: Martin Storsjö <martin@martin.st>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-05-25 14:21:54 +02:00
Rémi Denis-Courmont
d03cdfa2b6 checkasm/riscv: test misaligned before V
Otherwise V functions mask scalar misaligned ones.
2024-05-24 17:53:43 +03:00
James Almer
0920f506a7 checkasm/flacdsp: add a test for lpc33
Signed-off-by: James Almer <jamrial@gmail.com>
2024-05-24 09:23:00 -03:00
Stone Chen
2e877090f9 tests/checkasm: Add check_vvc_sad to vvc_mc.c
Adds checkasm for DMVR SAD AVX2 implementation.

Benchmarks ( AMD 7940HS )
vvc_sad_8x8_c: 50.3
vvc_sad_8x8_avx2: 0.3
vvc_sad_16x16_c: 250.3
vvc_sad_16x16_avx2: 10.3
vvc_sad_32x32_c: 1020.3
vvc_sad_32x32_avx2: 60.3
vvc_sad_64x64_c: 3850.3
vvc_sad_64x64_avx2: 220.3
vvc_sad_128x128_c: 14100.3
vvc_sad_128x128_avx2: 840.3

Reviewed-by: Ronald S. Bultje <rsbultje@gmail.com>
Signed-off-by: James Almer <jamrial@gmail.com>
2024-05-22 20:36:46 -03:00
Martin Storsjö
6093367147 checkasm: h264dsp: Avoid out of buffer writes when benchmarking
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 spurious
failures.

Signed-off-by: Martin Storsjö <martin@martin.st>
2024-05-21 19:20:06 +03:00
Lynne
d43e123837
checkasm: print bench runs when benchmarking
Helps make sense of the possible noise in the results.
2024-05-21 17:48:48 +02:00
J. Dekker
b1adf6d1d0 checkasm: add runs argument to adjust during bench
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 adjust the number of samples by an exponent (--runs) has
been added, allowing developers to increase the sample count for more
reliable results.

Signed-off-by: J. Dekker <jdek@itanimul.li>
2024-05-21 16:47:45 +02:00
Martin Storsjö
a9dc7dd7fd checkasm: vvc_alf: Limit benchmarking to a reasonable subset of functions
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 of 4. Limit benchmarking to
the cases which are powers of two.

This reduces the number of benchmarked cases from 3072 down to 18.
2024-05-21 20:20:50 +08:00
Nuo Mi
aa8d5c6e7e avcodec/vvcdec: add vvc inter filters for RPR 2024-05-21 20:20:25 +08:00
James Almer
727a603158 checkasm/h264dsp: use int64_t scale values
Fixes "signed integer overflow: [varies] * 104858 cannot be represented in type 'int'" errors
under ubsan.

Signed-off-by: James Almer <jamrial@gmail.com>
2024-05-17 17:04:08 -03:00
Andreas Rheinhardt
784672b833 tests/checkasm/sw_gbrp: Use correct function types for calls
E.g. f2de911818fbd7e73343803626b697fd0c968121 forgot to
add the unused void* here.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-05-17 13:36:03 +02:00
Andreas Rheinhardt
9126705e6e tests/checkasm/vf_blend: Update function type
Forgotten in 5b8faaad6c71bbb90951ca1642391e11cf6f5f91,
a69a0b689c3c4e1e0bbf4c17716a25e14332a250.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-05-17 13:35:33 +02:00
Andreas Rheinhardt
caec57eb3a tests/checkasm/vf_bwdif: Use correct function pointer type
Forgotten in fa06f48371ba2b73c139810db5bb893dc27eda7a.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-05-17 13:31:37 +02:00
Andreas Rheinhardt
4c57cbda67 tests/checkasm/vf_colorspace: Use correct function pointer type
Forgotten in 9b26a8077f1c0139fdcc236d3de08cd2bdc4ec0f.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-05-17 13:31:23 +02:00
Andreas Rheinhardt
e3de22e307 tests/checkasm/vvc_mc: Use correct function pointer type
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-05-17 13:31:09 +02:00
Andreas Rheinhardt
6c0994864e tests/checkasm/vp8dsp: Use correct function pointer type
Forgotten in a54e53a1c428299b19c7b4e2b66d01c0482c41dd.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-05-17 13:30:17 +02:00
Andreas Rheinhardt
dcbdcc3bf5 tests/checkasm/motion: Use correct function pointer type
Forgotten in abb85429f3424375f21bdd135656c2d88357b3d5.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-05-17 13:30:06 +02:00
Andreas Rheinhardt
99135a2f8b tests/checkasm/llviddsp: Use correct function pointer type
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-05-17 13:29:51 +02:00
Andreas Rheinhardt
631636e582 tests/checkasm/huffyuvdsp: Use correct function pointer type
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-05-17 13:29:34 +02:00
Andreas Rheinhardt
895dd370a2 tests/checkasm/hevc_*: Fix funtion pointer types
Forgotten in b3bbbb14d0685c8c1fbcf8455e59c7f444290c7c.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-05-17 13:28:20 +02:00
Andreas Rheinhardt
dff0a1557e tests/checkasm/llauddsp: Avoid UB integer overflow
The only multiplicators used in scalarproduct_and_madd_*
are -1, 0 and +1. Yet it is of type int and the checkasm
test uses the complete range of int for it, leading to overflows
that don't happen for actual users.

Fix this by using a more reasonable range for mul: Given
that it is used in v1[i] += v3[i] * mul with v1 being
a 16bit integer, it makes no sense to use values for mul
that don't fit into 16bit.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-05-17 13:16:58 +02:00
James Almer
8670615743 checkasm/h264dsp: add missing pixel_mask values
Fixes "runtime error: index 4 out of bounds for type 'uint32_t [3]'" errors
after commit 4ced36744ee0bea5fb7a20d1b2926c588f89ea0b.

Signed-off-by: James Almer <jamrial@gmail.com>
2024-05-14 23:17:30 -03:00
Rémi Denis-Courmont
b410439263 lavu/riscv: CPU flag for fast misaligned accesses 2024-05-14 19:50:00 +03:00
J. Dekker
4ced36744e checkasm/h264dsp: support checking more idct depths
Signed-off-by: J. Dekker <jdek@itanimul.li>
2024-05-14 17:46:55 +02:00
Wu Jianhua
3fa98f274a tests/checkasm/vvc_alf: add check_alf_classify
Perforamnce Test (fps):
clip                                      before  after delta
Tango2_3840x2160_60_10_420_27_LD.266      56      115   105.36%
RitualDance_1920x1080_60_10_420_32_LD.266 272     481   76.83%
RitualDance_1920x1080_60_10_420_37_RA.266 303     426   40.59%

Signed-off-by: Wu Jianhua <toqsxw@outlook.com>
2024-05-14 19:21:35 +08:00
Wu Jianhua
9ef6e15b04 tests/checkasm: add checkasm_check_vvc_alf and check_alf_filter
Signed-off-by: Wu Jianhua <toqsxw@outlook.com>
2024-05-14 19:21:35 +08:00
James Almer
5ba6f4e63e checkasm/flacdsp: add a test for wasted33
Signed-off-by: James Almer <jamrial@gmail.com>
2024-05-13 12:18:10 -03:00
James Almer
5acec189af checkasm/flacdsp: add a test for wasted32
Signed-off-by: James Almer <jamrial@gmail.com>
2024-05-12 17:24:08 -03:00
James Almer
479d26cea2 checkasm/flacdsp: sanitize lpc arguments
Fixes signed integer overflows as reported by ubsan.

Signed-off-by: James Almer <jamrial@gmail.com>
2024-05-12 12:36:46 -03:00
James Almer
467d84a06d checkasm/flacdsp: run lpc benchmarks with an unmodified buffer
Signed-off-by: James Almer <jamrial@gmail.com>
2024-05-12 12:36:46 -03:00