Intel's x86-simd-sort 4.0 Delivers A 2x Boost For AVX-512 Performance, Adds AVX2 Code

[phoronix]

Phoronix: Intel's x86-simd-sort 4.0 Delivers A 2x Boost For AVX-512 Performance, Adds AVX2 Code

Earlier this year Intel published x86-simd-sort as a very speedy sorting library that initially leveraged AVX-512 instructions for 10~17x faster sorts. Numpy was one of the first major projects to adopt x86-simd-sort and OpenJDK more recently adopted it. Since the initial release we've seen more features and performance optimizations added. Today marks the release of x86-simd-sort 4.0 and it's delivering even greater performance while also adding an AVX2 code path to help those without AVX-512...

https://www.phoronix.com/news/Intel-x86-simd-sort-4.0

[JEBjames]

Michael

typo/unclear

"~" means approximately...so "10~17x" i.e. "10 approximately 17x" makes no sense?

I think you meant a 10 to 17x speedup? i.e. "10-17x"?

n.b. you copy-pasted this line from a previous article so same thing applies there as well.

[zexelon]

Fairly deep dive into information on performance: https://github.com/Voultapher/sort-r...avx512/text.md

For those interested in a deeper technical dive into simd-sort.

[NathanG]

Michael

typo/unclear

"~" means approximately...so "10~17x" i.e. "10 approximately 17x" makes no sense?

I think you meant a 10 to 17x speedup? i.e. "10-17x"?

n.b. you copy-pasted this line from a previous article so same thing applies there as well.

~ comes from the project release information.

[NathanG]

Fairly deep dive into information on performance: https://github.com/Voultapher/sort-r...avx512/text.md

For those interested in a deeper technical dive into simd-sort.

Thanks, that very nicely illustrates why you should understand your data and test with representative data, rather than just going with a recommendation.

Also a lot of D20 rolls.

[illwieckz]

Michael

typo/unclear

"~" means approximately...so "10~17x" i.e. "10 approximately 17x" makes no sense?

I think you meant a 10 to 17x speedup? i.e. "10-17x"?

n.b. you copy-pasted this line from a previous article so same thing applies there as well.

When used between two numbers (not as a prefix), it means “approximately between 10 and 17”.

[Ladis]

Would be nice, if Intel re-added AVX-512 to its CPUs. Especially when every AMD CPU has it now.