perf: loadu_si512 instead of slower loadu_epi
#49Comparing
optimize_loadu_avx512
(6804059
) withmain
(55c8c5a
)0%
- Untouched
- 168
Benchmarks
168 total
avx2_f16_argmin_in
benches/bench_f16_ignore_nan.rs::benches::argminmax_in_f16_random_array_long0%
121.2 µs121.2 µs
impl_i16_argmax
benches/bench_i16.rs::benches::argminmax_i16_random_array_long0%
106.8 µs106.8 µs
impl_i16_argmin
benches/bench_i16.rs::benches::argminmax_i16_random_array_long0%
106.8 µs106.8 µs
impl_u16_argmin
benches/bench_u16.rs::benches::argminmax_u16_random_array_long0%
106.9 µs106.9 µs
impl_u16_argmax
benches/bench_u16.rs::benches::argminmax_u16_random_array_long0%
106.9 µs106.9 µs
avx2_f16_argmax_in
benches/bench_f16_ignore_nan.rs::benches::argminmax_in_f16_random_array_long0%
121.3 µs121.3 µs
sse_f16_argminmax_rn
benches/bench_f16_return_nan.rs::benches::argminmax_rn_f16_random_array_long0%
159.1 µs159.1 µs
sse_f16_argmin_rn
benches/bench_f16_return_nan.rs::benches::argminmax_rn_f16_random_array_long0%
159.1 µs159.2 µs
avx2_f16_argmax_rn
benches/bench_f16_return_nan.rs::benches::argminmax_rn_f16_random_array_long0%
119.3 µs119.3 µs
avx2_f16_argminmax_in
benches/bench_f16_ignore_nan.rs::benches::argminmax_in_f16_random_array_long0%
128.8 µs128.8 µs
sse_f16_argminmax_in
benches/bench_f16_ignore_nan.rs::benches::argminmax_in_f16_random_array_long0%
180.2 µs180.1 µs
sse_f16_argmax_rn
benches/bench_f16_return_nan.rs::benches::argminmax_rn_f16_random_array_long0%
159.2 µs159.2 µs
impl_i32_argmin
benches/bench_i32.rs::benches::argminmax_i32_random_array_long0%
211.5 µs211.5 µs
impl_i32_argmax
benches/bench_i32.rs::benches::argminmax_i32_random_array_long0%
211.5 µs211.5 µs
impl_u32_argmin
benches/bench_u32.rs::benches::argminmax_u32_random_array_long0%
216 µs215.9 µs
impl_u32_argmax
benches/bench_u32.rs::benches::argminmax_u32_random_array_long0%
215.9 µs215.9 µs
impl_f32_argminmax_in
benches/bench_f32_ignore_nan.rs::benches::argminmax_in_f32_random_array_long0%
225.2 µs225.2 µs
avx2_f32_argminmax_rn
benches/bench_f32_return_nan.rs::benches::argminmax_rn_f32_random_array_long0%
235.5 µs235.6 µs
avx2_f32_argmin_rn
benches/bench_f32_return_nan.rs::benches::argminmax_rn_f32_random_array_long0%
235.6 µs235.6 µs
sse_i32_argminmax
benches/bench_i32.rs::benches::argminmax_i32_random_array_long0%
290.5 µs290.4 µs
impl_f16_argmax_in
benches/bench_f16_ignore_nan.rs::benches::argminmax_in_f16_random_array_long0%
121.5 µs121.5 µs
sse_f32_argminmax_rn
benches/bench_f32_return_nan.rs::benches::argminmax_rn_f32_random_array_long0%
319 µs319 µs
sse_f32_argmin_rn
benches/bench_f32_return_nan.rs::benches::argminmax_rn_f32_random_array_long0%
319.1 µs319.1 µs
impl_f16_argmin_rn
benches/bench_f16_return_nan.rs::benches::argminmax_rn_f16_random_array_long0%
119.4 µs119.4 µs
impl_f16_argmax_rn
benches/bench_f16_return_nan.rs::benches::argminmax_rn_f16_random_array_long0%
119.5 µs119.5 µs
Commits
Click on a commit to change the comparison rangeBase
main
55c8c5a
0%
perf: loadu_si512 instead of slower loadu_epi
6804059
3 years ago
by jvdd