SIMD support

[leonbotros]

• SIMD enabled by simd feature, make sure to compile using nightly (cargo +nightly) for a supported target (using RUSTFLAGS='-C target-cpu=native' or RUSTFLAGS='-C target_feature=+avx2' for example).
• Checks at compile time if the target supports AVX512/AVX2/WASM SIMD and chooses x16/x8/x4 accordingly.
• Keeps bigger in and out buffers in the SIMD implementation, so only useful if atleast 48 * 16/8/4 byte input/outputs are absorbed/squeezed.
• SIMD xoodoo permutation is fully generic over the amount of SIMD lanes.
• Xoofff implementations are not fully generic due to unrolling on a generic constant not being possible using crunchy.
• Operations converting bytes to N parallel state and vice versa are not optimized. They might be faster using gather/scatter methods.

TODO:

• document how to enable the feature.
• doc test compilation currently fails.
• test simd in CI.
• test or not include the x16 code (support is low, I don't have it either)

[itzmeanjan]

Amazing, thanks for the PR @leonbotros . I'll take a look.

[leonbotros]

Thanks, feedback is appreciated. Since WASM does not have runtime feature detection I decided to not use any of the #[target_feature(enable = "...")] stuff.

[itzmeanjan]

Thanks, feedback is appreciated. Since WASM does not have runtime feature detection I decided to not use any of the #[target_feature(enable = "...")] stuff.

Totally makes sense. I'll need some time before I can review the PR. Though it's a relief to see that tests are passing.

[itzmeanjan]

Hi @leonbotros , I merged back latest from master into your PR branch and when I benchmark on my machine 12th Gen Intel(R) Core(TM) i7-1260P, running GNU/Linux kernel 6.5.0-13-generic, I see ⬇️

It's pretty good 💯

I'd like to merge this and publish on crates.io. Would you doing the remaining or you want me to take over for the rest of the part ?

[leonbotros]

Hi, I currently have no time to clean this up a bit so feel free ;)

…

On Sun, Nov 26, 2023, 06:16 Anjan Roy ***@***.***> wrote: Hi @leonbotros <https://github.com/leonbotros> , I merged back latest from master into your PR branch and when I benchmark on my machine *12th Gen Intel(R) Core(TM) i7-1260P*, running GNU/Linux kernel *6.5.0-13-generic*, I see ⬇️ Screenshot.from.2023-11-26.09-11-40.png (view on web) <https://github.com/itzmeanjan/xoofff/assets/45074836/e4651462-47e6-4a84-95f8-2fdcc6d66f96> It's pretty good 💯 I'd like to merge this and publish on crates.io. Would you doing the remaining or you want me to take over for the rest of the part ? — Reply to this email directly, view it on GitHub <#5 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AANY4MOE3JESN7JMHXNJBY3YGLGETAVCNFSM6AAAAAAZYJZQOOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQMRWGU3DEMRXGU> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[itzmeanjan]

Hi, I currently have no time to clean this up a bit so feel free ;)
…
On Sun, Nov 26, 2023, 06:16 Anjan Roy @.> wrote: Hi @leonbotros https://github.com/leonbotros , I merged back latest from master into your PR branch and when I benchmark on my machine 12th Gen Intel(R) Core(TM) i7-1260P, running GNU/Linux kernel 6.5.0-13-generic, I see ⬇️ Screenshot.from.2023-11-26.09-11-40.png (view on web) https://github.com/itzmeanjan/xoofff/assets/45074836/e4651462-47e6-4a84-95f8-2fdcc6d66f96 It's pretty good 💯 I'd like to merge this and publish on crates.io. Would you doing the remaining or you want me to take over for the rest of the part ? — Reply to this email directly, view it on GitHub <#5 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AANY4MOE3JESN7JMHXNJBY3YGLGETAVCNFSM6AAAAAAZYJZQOOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQMRWGU3DEMRXGU . You are receiving this because you were mentioned.Message ID: @.>

Thanks for your contribution. I'll proceed then ;)