adamg/actually-compare-packed-bools - Branch - spiraldb/fastlanes

Unpack compare into an appropriately sized bit array

#125

Comparing

adamg/actually-compare-packed-bools

(

56236df

) with

develop

(

7c3e7e5

)

-10%

Improvement: 18

Regression: 1

Untouched: 17

Skipped: 16

Benchmarks

52 total

bitpacking_cmp_seq[u32]

benches/bitpacking_cmp_cod.rs

-10%

5.2 µs5.8 µs

bitpacking_cmp_unpack[u64]

benches/bitpacking_cmp_cod.rs

×2.8

9.6 µs3.4 µs

bitpacking_cmp_unpack[u32]

benches/bitpacking_cmp_cod.rs

×2.8

5.2 µs1.9 µs

bitpacking_cmp_unpack[u16]

benches/bitpacking_cmp_cod.rs

×2.4

3.1 µs1.3 µs

unfor_pack_16_from_3_stack

benches/ffor.rs

+80%

3 µs1.7 µs

unchecked_unfor_pack_16_from_3_stack

benches/ffor.rs

+69%

3.2 µs1.9 µs

unchecked_unpack_then_add_reference_16_from_3_stack

benches/ffor.rs

+68%

3.9 µs2.3 µs

unpack_then_add_reference_16_from_3_stack

benches/ffor.rs

+67%

3.7 µs2.2 µs

delta_u16_fused

benches/delta.rs

+57%

3.6 µs2.3 µs

unpack_16_from_3_stack

benches/bitpacking.rs

+54%

2.4 µs1.6 µs

unchecked_unpack_16_from_3_stack

benches/bitpacking.rs

+47%

2.6 µs1.8 µs

bitpacking_cmp_fused[u64]

benches/bitpacking_cmp_cod.rs

+35%

3.9 µs2.9 µs

bitpacking_cmp_fused[u32]

benches/bitpacking_cmp_cod.rs

+34%

2.9 µs2.1 µs

delta_u16_unfused

benches/delta.rs

+33%

5.1 µs3.9 µs

pack_16_to_3_stack

benches/bitpacking.rs

+32%

2.2 µs1.7 µs

for_pack_16_to_3_stack

benches/ffor.rs

+27%

2.3 µs1.8 µs

transpose_u16

benches/transpose.rs

+14%

10.8 µs9.5 µs

bitpacking_cmp_fused[u16]

benches/bitpacking_cmp_cod.rs

+12%

2.2 µs2 µs

bitpacking_cmp_seq[u16]

benches/bitpacking_cmp_cod.rs

+10%

5.8 µs5.2 µs

bitpacking_cmp_seq[u64]

benches/bitpacking_cmp_cod.rs

+5%

7.6 µs7.3 µs

throughput_decompress_unchecked_separate_reference

benches/ffor.rs

1.4 ms1.4 ms

delta_throughput_compress

benches/delta.rs

804.7 µs804.7 µs

throughput_decompress

benches/ffor.rs

1.2 ms1.2 ms

throughput_decompress_unchecked

benches/ffor.rs

1.3 ms1.3 ms

throughput_decompress_separate_reference

benches/ffor.rs

1.4 ms1.4 ms

Commits

Click on a commit to change the comparison range

Base

develop

7c3e7e5

-16.48%

Actually compare into a packed mask

2649497

9 days ago

by AdamGS

+1.91%

Improve benchmark accuracy

aff6e2d

8 days ago

by AdamGS

-0.06%

maybe like this?

3a5de82

8 days ago

by AdamGS

what about this

ef48533

8 days ago

by AdamGS

maybe not

623311a

8 days ago

by AdamGS

+0.84%

MOAR

7c28298

8 days ago

by AdamGS

+3.36%

Some benchmark cleanup

56236df

8 days ago

by AdamGS

Home Terms Privacy Docs