portable_simd version of Sub (3bpp/4bpp) #642

Sentimentron · 2025-09-09T18:36:50Z

Results from silicon (bpp=3)

CPU	Baseline	Result	Speedup
Arm Cortex A520	910.9 MiB/s	2025.9 MiB/s	122.40%
Arm Cortex X4	6551.8 MiB/s	6313 MiB/s	-3.64%
Apple Silicon M2	5839.2 MiB/s	5751.8 MiB/s	-1.33%
AMD EPYC 7B13	3830.6 MiB/s	5472.9 MiB/s	42.87%

Results from silicon (bpp=4)

CPU	Baseline	Result	Speedup
Arm Cortex A520	607.0 MiB/s	3226.8 MiB/s	431.62%
Arm Cortex X4	6551.8 MiB/s	6313.0 MiB/s	105.71%
Apple Silicon M2	5800.3 MiB/s	10616.0 MiB/s	83.03%
AMD EPYC 7B13	10796.0 MiB/s	15268.0 MiB/s	41.42%

Opened as a draft until #632 is resolved.

Prefix sum improves performance on A520 and Epyc by 122.40% and 42.87% respectively.

Improves performance by around 40% on the Epyc system, 431% on the Cortex-A520.

Sentimentron added 4 commits September 3, 2025 20:54

perf: portable_simd 3bpp paeth filter

f6ba382

fix: restrict portable_simd feature to "unstable"

1b94193

perf: portable_simd version of sub (3bpp)

22d45f7

Prefix sum improves performance on A520 and Epyc by 122.40% and 42.87% respectively.

perf: 4bpp version of the sub filter

a807fe4

Improves performance by around 40% on the Epyc system, 431% on the Cortex-A520.

Provide feedback