Fix TurboQuant centroid initialization wasting most codes at high dimensions
#8076Comparing
mp/tq-centroid-init
(8db4b68
) withdevelop
(2699169
)-16%
- Regression
- 1
- Untouched
- 1250
Unknown Walltime execution environment detected
Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data. For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.Benchmarks
1251 total
chunked_varbinview_opt_canonical_into[(1000, 10)]
vortex-array/benches/chunk_array_builder.rs-16%
187.4 µs224.5 µs
iter_arrow_buffer[128]
vortex-buffer/benches/vortex_bitbuffer.rs+7%
890.3 ns831.9 ns
set_indices_arrow_buffer[128]
vortex-buffer/benches/vortex_bitbuffer.rs+6%
1,010.3 ns951.9 ns
take_filter_primitive_small_random_mask_random_indices[4096, 10]
vortex-array/benches/take_filter.rs+5%
30 µs28.5 µs
slice_vortex_buffer[16384]
vortex-buffer/benches/vortex_bitbuffer.rs+5%
1.2 µs1.2 µs
slice_vortex_buffer[128]
vortex-buffer/benches/vortex_bitbuffer.rs+5%
1.2 µs1.2 µs
slice_vortex_buffer[65536]
vortex-buffer/benches/vortex_bitbuffer.rs+5%
1.2 µs1.2 µs
slice_vortex_buffer[1024]
vortex-buffer/benches/vortex_bitbuffer.rs+5%
1.2 µs1.2 µs
slice_vortex_buffer[2048]
vortex-buffer/benches/vortex_bitbuffer.rs+5%
1.2 µs1.2 µs
rank_single[(1024, 0.1)]
vortex-mask/benches/rank.rs+4%
1.5 µs1.5 µs
rank_single[(1024, 0.9)]
vortex-mask/benches/rank.rs+4%
1.5 µs1.5 µs
rank_single[(16384, 0.9)]
vortex-mask/benches/rank.rs+3%
1.9 µs1.8 µs
decompress[u8, (1000, 16)]
encodings/runend/benches/run_end_compress.rs+3%
18.9 µs18.3 µs
rank_single[(16384, 0.1)]
vortex-mask/benches/rank.rs+3%
2.1 µs2 µs
patched_take_10_contiguous
encodings/fastlanes/benches/bitpacking_take.rs+3%
30.8 µs30 µs
push_n_vortex_buffer[u32, 128]
vortex-buffer/benches/vortex_buffer.rs+2%
1.7 µs1.7 µs
transpose_scalar
encodings/fastlanes/benches/bit_transpose.rs+2%
1.8 µs1.8 µs
rank_single[(65536, 0.1)]
vortex-mask/benches/rank.rs+2%
3.9 µs3.8 µs
rank_single[(65536, 0.9)]
vortex-mask/benches/rank.rs+2%
3.9 µs3.9 µs
bitwise_not_arrow_buffer[128]
vortex-buffer/benches/vortex_bitbuffer.rs+1%
2.1 µs2.1 µs
append_vortex_buffer[128]
vortex-buffer/benches/vortex_bitbuffer.rs+1%
2.1 µs2.1 µs
take_map[(0.1, 0.5)]
vortex-array/benches/take_patches.rs+1%
993.5 µs980.2 µs
cuda/alp_f32/1%[100M]
vortex-cuda/benches/alp_cuda.rs::benches::benchmark_alp_decode::cuda+1%
4.8 ms4.8 ms
set_indices_arrow_buffer[1024]
vortex-buffer/benches/vortex_bitbuffer.rs+1%
2.2 µs2.2 µs
iter_arrow_buffer[1024]
vortex-buffer/benches/vortex_bitbuffer.rs+1%
4.4 µs4.3 µs
Commits
Click on a commit to change the comparison rangeBase
develop
2699169
-16.5%
Fix TurboQuant centroid initialization wasting most codes at high dimensions
8db4b68
2 hours ago
by mprammer