Re: [FFmpeg-devel] [PATCH] swscale/x86/rgb2rgb: add AVX512ICL versions of shuffle_bytes

[James Almer <jamrial@gmail.com>]

[-- Attachment #1.1.1: Type: text/plain, Size: 2039 bytes --]

On 1/25/2025 12:50 PM, Shreesh Adiga wrote:
>> Try running it several times using the same seed, so
>> "tests/checkasm/checkasm --test=sw_rgb --bench 17575157", and make sure
>> no power saving feature is enabled (so the CPU frequency doesn't change
>> based on load). That may help getting consistent results.
> 
> After running "echo performance | tee
> /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor" and
> recompiling ffmpeg with "--enable-linux-perf", I am seeing the below
> numbers:
> 
> shuffle_bytes_0321_c:                                   56.5 ( 1.00x)
> shuffle_bytes_0321_ssse3:                               18.0 ( 3.14x)
> shuffle_bytes_0321_avx2:                                10.0 ( 5.65x)
> shuffle_bytes_0321_avx512icl:                            9.0 ( 6.28x)
> shuffle_bytes_1230_c:                                   84.5 ( 1.00x)
> shuffle_bytes_1230_ssse3:                               18.2 ( 4.63x)
> shuffle_bytes_1230_avx2:                                22.2 ( 3.80x)
> shuffle_bytes_1230_avx512icl:                           10.0 ( 8.45x)
> shuffle_bytes_2103_c:                                   49.8 ( 1.00x)
> shuffle_bytes_2103_ssse3:                               21.2 ( 2.34x)
> shuffle_bytes_2103_avx2:                                17.5 ( 2.84x)
> shuffle_bytes_2103_avx512icl:                            7.5 ( 6.63x)
> shuffle_bytes_3012_c:                                   84.5 ( 1.00x)
> shuffle_bytes_3012_ssse3:                               17.0 ( 4.97x)
> shuffle_bytes_3012_avx2:                                16.0 ( 5.28x)
> shuffle_bytes_3012_avx512icl:                           16.2 ( 5.20x)
> shuffle_bytes_3210_c:                                   92.8 ( 1.00x)
> shuffle_bytes_3210_ssse3:                               25.8 ( 3.60x)
> shuffle_bytes_3210_avx2:                                14.0 ( 6.62x)
> shuffle_bytes_3210_avx512icl:                            9.0 (10.31x)

Excellent. Will apply the patch later unless somebody objects.


[-- Attachment #1.2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 495 bytes --]

[-- Attachment #2: 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".