[experimental] Unify VBMI untranspose onto one vpermi2b kernel for all widths
#146Comparing
claude/dreamy-goodall-vbmi-uniform
(ec0597e
) withdevelop
(6c10ea7
)0%
- Untouched
- 158
- Skipped
- 123
Benchmarks
123 total
The benchmarks below were skipped, so their baseline results are used instead. If they were deleted in your codebase, archive them to remove them from the performance reports.
Learn more about archiving benchmarks
Learn more about archiving benchmarks
dispatch_untranspose
benches/bit_transpose.rsSkipped
404.6 µs*
bitpacking_cmp_fused[u32, 26]
benches/bitpacking_cmp.rs::benchSkipped
3.7 µs*
bitpacking_cmp_fused[u16, 11]
benches/bitpacking_cmp.rs::benchSkipped
1.8 µs*
bitpacking_cmp_fused[u32, 18]
benches/bitpacking_cmp.rs::benchSkipped
3.1 µs*
scalar_untranspose
benches/bit_transpose.rsSkipped
606.2 µs*
bitpacking_cmp_fused[u32, 11]
benches/bitpacking_cmp.rs::benchSkipped
2.6 µs*
bitpacking_cmp_fused[u32, 27]
benches/bitpacking_cmp.rs::benchSkipped
3.7 µs*
bitpacking_cmp_fused[u16, 4]
benches/bitpacking_cmp.rs::benchSkipped
1.6 µs*
bitpacking_cmp_fused[u32, 24]
benches/bitpacking_cmp.rs::benchSkipped
2.7 µs*
bitpacking_cmp_fused[u16, 9]
benches/bitpacking_cmp.rs::benchSkipped
1.7 µs*
bitpacking_cmp_fused[u16, 8]
benches/bitpacking_cmp.rs::benchSkipped
1.8 µs*
bitpacking_cmp_fused[u32, 10]
benches/bitpacking_cmp.rs::benchSkipped
2.5 µs*
bitpacking_cmp_fused[u32, 25]
benches/bitpacking_cmp.rs::benchSkipped
3.6 µs*
bitpacking_cmp_fused[u16, 10]
benches/bitpacking_cmp.rs::benchSkipped
1.8 µs*
bitpacking_cmp_fused[u32, 12]
benches/bitpacking_cmp.rs::benchSkipped
2.3 µs*
bitpacking_cmp_fused[u32, 28]
benches/bitpacking_cmp.rs::benchSkipped
3 µs*
bitpacking_cmp_fused[u16, 12]
benches/bitpacking_cmp.rs::benchSkipped
1.8 µs*
bitpacking_cmp_fused[u32, 13]
benches/bitpacking_cmp.rs::benchSkipped
2.7 µs*
bitpacking_cmp_fused[u16, 13]
benches/bitpacking_cmp.rs::benchSkipped
2.3 µs*
bitpacking_cmp_fused[u32, 29]
benches/bitpacking_cmp.rs::benchSkipped
3.8 µs*
bitpacking_cmp_fused[u32, 6]
benches/bitpacking_cmp.rs::benchSkipped
2.2 µs*
bitpacking_cmp_fused[u32, 14]
benches/bitpacking_cmp.rs::benchSkipped
2.8 µs*
bitpacking_cmp_fused[u32, 2]
benches/bitpacking_cmp.rs::benchSkipped
1.9 µs*
bitpacking_cmp_fused[u32, 30]
benches/bitpacking_cmp.rs::benchSkipped
3.9 µs*
bitpacking_cmp_fused[u32, 15]
benches/bitpacking_cmp.rs::benchSkipped
2.9 µs*
Commits
Click on a commit to change the comparison rangeBase
claude/dreamy-goodall-3vpHT
6c10ea7
0%
Unify VBMI untranspose onto a single vpermi2b kernel for all widths
ec0597e
25 days ago
by joseph-isaacs