aarch64_64bit_simd - Branch - jvdd/argminmax

feat: add 64-bit aarch64 SIMD implementation

#52Merged

Comparing

aarch64_64bit_simd

(

5b51457

) with

main

(

5602f92

)

Untouched: 168

Benchmarks

168 total

impl_f16_argmax_rn

benches/bench_f16_return_nan.rs::benches::argminmax_rn_f16_random_array_long

119.5 µs119.4 µs

impl_f16_argminmax_rn

benches/bench_f16_return_nan.rs::benches::argminmax_rn_f16_random_array_long

119.5 µs119.4 µs

avx2_u8_argmin

benches/bench_u8.rs::benches::argminmax_u8_random_array_long

61.9 µs61.9 µs

avx2_i8_argmax

benches/bench_i8.rs::benches::argminmax_i8_random_array_long

62.5 µs62.5 µs

avx2_f16_argminmax_rn

benches/bench_f16_return_nan.rs::benches::argminmax_rn_f16_random_array_long

119.3 µs119.2 µs

avx2_u8_argminmax

benches/bench_u8.rs::benches::argminmax_u8_random_array_long

73.7 µs73.7 µs

sse_f16_argmax_in

benches/bench_f16_ignore_nan.rs::benches::argminmax_in_f16_random_array_long

162 µs162 µs

avx2_f16_argmin_in

benches/bench_f16_ignore_nan.rs::benches::argminmax_in_f16_random_array_long

121.2 µs121.2 µs

sse_f16_argmax_rn

benches/bench_f16_return_nan.rs::benches::argminmax_rn_f16_random_array_long

159.2 µs159.2 µs

avx2_u16_argminmax

benches/bench_u16.rs::benches::argminmax_u16_random_array_long

113.1 µs113.1 µs

sse_i16_argminmax

benches/bench_i16.rs::benches::argminmax_i16_random_array_long

144.2 µs144.2 µs

avx2_f16_argmin_rn

benches/bench_f16_return_nan.rs::benches::argminmax_rn_f16_random_array_long

119.3 µs119.3 µs

impl_f16_argmin_rn

benches/bench_f16_return_nan.rs::benches::argminmax_rn_f16_random_array_long

119.4 µs119.4 µs

sse_f16_argminmax_in

benches/bench_f16_ignore_nan.rs::benches::argminmax_in_f16_random_array_long

180.2 µs180.1 µs

avx_f32_argmin_in

benches/bench_f32_ignore_nan.rs::benches::argminmax_in_f32_random_array_long

211.5 µs211.4 µs

scalar_i8_argmin

benches/bench_i8.rs::benches::argminmax_i8_random_array_long

253 µs253 µs

scalar_i8_argmax

benches/bench_i8.rs::benches::argminmax_i8_random_array_long

253 µs253 µs

sse_u32_argmax

benches/bench_u32.rs::benches::argminmax_u32_random_array_long

265.5 µs265.5 µs

scalar_u16_argmin

benches/bench_u16.rs::benches::argminmax_u16_random_array_long

292.6 µs292.6 µs

sse_i32_argminmax

benches/bench_i32.rs::benches::argminmax_i32_random_array_long

290.5 µs290.4 µs

sse_u32_argminmax

benches/bench_u32.rs::benches::argminmax_u32_random_array_long

301.2 µs301.2 µs

scalar_i16_argminmax

benches/bench_i16.rs::benches::argminmax_i16_random_array_long

349.5 µs349.5 µs

scalar_f32_argminmax_rn

benches/bench_f32_return_nan.rs::benches::argminmax_rn_f32_random_array_long

528.2 µs528.1 µs

avx_f64_argmax_in

benches/bench_f64_ignore_nan.rs::benches::argminmax_in_f64_random_array_long

422.1 µs422.1 µs

avx_f64_argminmax_in

benches/bench_f64_ignore_nan.rs::benches::argminmax_in_f64_random_array_long

448.8 µs448.8 µs

Commits

Click on a commit to change the comparison range

Base

main

5602f92

+0.01%

fix: fix typo in nightly_simd feature

383c791

3 years ago

by jvdd

fix: use 64-bit lanesize for aarch64 u64

4cbdd0c

3 years ago

by jvdd

feat: add stable f64 ignore nan SIMD for aarch64

fec3b69

3 years ago

by jvdd

feat: add stable f64 return nan SIMD for aarch64

9227628

3 years ago

by jvdd

fix: import std-arch for aarch64

cf6e522

3 years ago

by jvdd

fix: use 64-bit intrinsincs instead of 32-bit ones :see_no_evil:

d0b4d62

3 years ago

by jvdd

fix: MAX_INDEX available on aarch64 for f64 ignore nan

88a0e46

3 years ago

by jvdd

fix: clippy import warnings

0097135

3 years ago

by jvdd

bench: add 64-bit aarch64 benchmarks

8c23a21

3 years ago

by jvdd

perf: use aarch64 where faster than scalar

8d91856

3 years ago

by jvdd

chore: formatting

3b9708c

3 years ago

by jvdd

fix: use correct arg function

b66e8e5

3 years ago

by jvdd

-0.01%

:pen: review code

5b51457

3 years ago

by jvdd

Home Terms Privacy Docs