Update stdsimd to undo an accidental stabilization #52535

alexcrichton · 2018-07-19T14:02:44Z

Closes #52403

rust-highfive · 2018-07-19T14:02:48Z

r? @aturon

(rust_highfive has picked a reviewer for you, use r? to override)

rust-highfive · 2018-07-19T14:02:49Z

Warning

These commits modify submodules.

alexcrichton · 2018-07-19T14:02:59Z

@bors: r=Mark-Simulacrum

carrying over approval from #52429

bors · 2018-07-19T14:03:00Z

📌 Commit ec4aaf1 has been approved by Mark-Simulacrum

alexcrichton · 2018-07-19T14:03:59Z

This error was suspected to be caused by this update, so I'm running a test locally

…-Simulacrum Update stdsimd to undo an accidental stabilization Closes rust-lang#52403

kennytm · 2018-07-20T16:15:36Z

The same powerpc error in #52570 (comment), but again it is a rollup 🤷.

alexcrichton · 2018-07-20T16:31:22Z

@bors: r-

I'll take a look

We're running into issues updating with rust-lang/rust#52535, so we need to get this working without `RUSTFLAGS` enabling the `altivec` feature

alexcrichton · 2018-07-20T16:58:17Z

@gnzlbg as a heads up I had to update this to stdsimd master to pull in rust-lang/stdarch#531 which also pulls in the deletion of std::simd

@bors: r=Mark-Simulacrum

bors · 2018-07-20T16:58:18Z

📌 Commit a25dfa4dacb39d892ab7ff6ab2dda4d30f260ad9 has been approved by Mark-Simulacrum

gnzlbg · 2018-07-20T21:49:18Z

@alexcrichton the std::simd replacement is already fully compatible with what we had before so people can just use that. Let me know if you want to pull it into the nursery.

…

On Fri 20. Jul 2018 at 18:59, Alex Crichton ***@***.***> wrote: @gnzlbg <https://github.com/gnzlbg> as a heads up I had to update this to stdsimd master to pull in rust-lang/stdarch#531 <rust-lang/stdarch#531> which also pulls in the deletion of std::simd @bors <https://github.com/bors>: r=Mark-Simulacrum — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#52535 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AA3Npv8MZJS5jH8hZ0PC5UB8Go57ZTKzks5uIgxagaJpZM4VWdK0> .

bors · 2018-07-20T21:49:19Z

@gnzlbg: 🔑 Insufficient privileges: Not in reviewers

alexcrichton · 2018-07-20T22:25:31Z

@gnzlbg oh I thought I recalled you saying you didn't want to delete std::simd just yet while you waited to publish your first version, but I may be misremembering! Other than that it sounds great! If you'd like I can move it to the nursery as well

bors · 2018-07-21T03:01:32Z

⌛ Testing commit a25dfa4dacb39d892ab7ff6ab2dda4d30f260ad9 with merge ba65701428f1c1ae53aee962ac32bd0771d4e922...

bors · 2018-07-21T04:19:12Z

💔 Test failed - status-appveyor

bors · 2018-07-21T06:26:24Z

⌛ Testing commit d77defc with merge 17eb392...

Update stdsimd to undo an accidental stabilization Closes #52403

gnzlbg · 2018-07-21T08:02:20Z

@alexcrichton

oh I thought I recalled you saying you didn't want to delete std::simd just yet while you waited to publish your first version, but I may be misremembering! Other than that it sounds great!

You are remembering it correctly, but I changed my mind about doing releases. Those who want to use this will need to add a git dependency on it, which ties them to nightly, and forbids them from releasing stuff.

That's unfortunate, but it will prevent the issue that we had with faster where the compiler moved on, and faster and all their dependencies stopped working because the released versions of coresimd became unsupported.

If you'd like I can move it to the nursery as well

I think that would be best. I guess one just needs to fork it right?

bors · 2018-07-21T08:31:28Z

☀️ Test successful - status-appveyor, status-travis
Approved by: Mark-Simulacrum
Pushing 17eb392 to master...

alexcrichton · 2018-07-21T15:39:25Z

@gnzlbg ok, if you want to catch me on IRC I can help you transfer the repo to the nursery

gnzlbg · 2018-07-21T16:14:55Z

I’m travelling this weekend and connectivity is low, but I think this can wait till Monday.

…

On Sat 21. Jul 2018 at 17:39, Alex Crichton ***@***.***> wrote: @gnzlbg <https://github.com/gnzlbg> ok, if you want to catch me on IRC I can help you transfer the repo to the nursery — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#52535 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AA3Nprw1fMu3DMhdRvAQMr6Qux4boGH2ks5uI0tGgaJpZM4VWdK0> .

pitdicker · 2018-07-22T05:15:31Z

src/libcore/lib.rs

@@ -244,9 +244,6 @@ macro_rules! vector_impl { ($([$f:ident, $($args:tt)*]),*) => { $($f!($($args)*)
 #[cfg(not(stage0))] // allow changes to how stdsimd works in stage0
 mod coresimd;

-#[unstable(feature = "stdsimd", issue = "48556")]
-#[cfg(not(stage0))]
-pub use coresimd::simd;


This removal broke the CI of rand, but I have trouble finding the reason. Is there an alternative to use simd with no_std, or is this no longer supported?

Indeed! This has moved over to https://github.com/gnzlbg/packed_simd

nnethercote · 2018-07-25T00:32:10Z

perf.rust-lang.org indicates that this caused a significant compile-time win, which is (a) nice, and (b) surprising to me.

@alexcrichton, do you know why this might have improved rustc speed?

gnzlbg · 2018-07-25T07:51:25Z

@nnethercote we removed a very big module from stdsimd: rust-lang/stdarch#528 but I don't know how could that make rustc compile things faster. AFAICT none of the benchmarks being compiled there were using that module (otherwise they would have stopped compiling when we removed it).

This might mean that adding stuff to the std library incurs a compile-time cost for users, even if they don't use the stuff, which would be very bad.

alexcrichton · 2018-07-25T09:00:49Z

Yes this removed a large unstable module from libcore. I believe it had a ton of trait impls, so maybe metadata loading for libstd is faster now? It could mean perhaps that metadata loading isn't as efficient as it could otherwise be perhaps?

gnzlbg · 2018-07-25T09:14:07Z

I believe it had a ton of trait impls,

This is correct (it really had a ton of trait impls). It also had a ton of types with tons of methods.

FYI what was removed now lives in the nursery: https://github.com/rust-lang-nursery/packed_simd

MoSal · 2018-07-25T15:58:55Z

FYI, while it's obviously removed from the nightly documentation, std::simd is still in the official documentation here.

This may confuse some people (e.g. game developers) who are googling "Rust SIMD".

@steveklabnik

gnzlbg · 2018-07-25T22:03:12Z

That's weird, the APIs are unstable so they should only appear on the nightly docs, but those are not in nightly/... :/

MoSal · 2018-07-25T22:53:01Z

@gnzlbg

Unstable APIs appear in stable docs. Maybe they shouldn't, but they do. And maybe it's too late to change that behavior.

Edit: A possible improvement would be to not show unstable modules, but continue showing unstable functions/methods. But that's a topic that may deserve its own issue.

nnethercote · 2018-07-26T05:24:03Z

so maybe metadata loading for libstd is faster now?

Seems plausible... especially given that the speedups were largest for short-running benchmarks (esp. "check" builds) which suggests a constant-ish time reduction.

I intended to do a Cachegrind diff but when I ran locally I saw a negligible difference caused by the change. I don't know why.

std::simd was removed in rust-lang/rust#52535

`std::simd` was removed in rust-lang/rust#52535. https://crates.io/crates/packed_simd is in the rust-lang-nursery org.

Use the packed_simd crate instead of std::simd `std::simd` was removed in rust-lang/rust#52535. https://crates.io/crates/packed_simd is in the rust-lang-nursery org.  --- This change is [<img src="https://reviewable.io/review_button.svg" height="34" align="absmiddle" alt="Reviewable"/>](https://reviewable.io/reviews/servo/servo/21272)

rust-highfive assigned aturon Jul 19, 2018

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Jul 19, 2018

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Jul 19, 2018

pietroalbini added a commit to pietroalbini/rust that referenced this pull request Jul 20, 2018

Rollup merge of rust-lang#52535 - alexcrichton:update-stdsimd, r=Mark…

7a0ffda

…-Simulacrum Update stdsimd to undo an accidental stabilization Closes rust-lang#52403

pietroalbini mentioned this pull request Jul 20, 2018

Rollup of 8 pull requests #52570

Closed

bors added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels Jul 20, 2018

alexcrichton mentioned this pull request Jul 20, 2018

Fix compile of stdsimd on powerpc with no flags rust-lang/stdarch#531

Merged

alexcrichton force-pushed the update-stdsimd branch from ec4aaf1 to a25dfa4 Compare July 20, 2018 16:57

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Jul 20, 2018

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Jul 21, 2018

bors added a commit that referenced this pull request Jul 21, 2018

Auto merge of #52535 - alexcrichton:update-stdsimd, r=Mark-Simulacrum

17eb392

Update stdsimd to undo an accidental stabilization Closes #52403

bors merged commit d77defc into rust-lang:master Jul 21, 2018

alexcrichton deleted the update-stdsimd branch July 21, 2018 15:05

pitdicker reviewed Jul 22, 2018

View reviewed changes

pitdicker mentioned this pull request Jul 22, 2018

Depend on packed_simd rust-random/rand#565

Closed

mati865 mentioned this pull request Jul 27, 2018

Can't build for nightly Rust AdamNiederer/faster#38

Open

SimonSapin added a commit to servo/servo that referenced this pull request Jul 27, 2018

Use the packed_simd crate instead of std::simd

789a61c

std::simd was removed in rust-lang/rust#52535

SimonSapin added a commit to servo/servo that referenced this pull request Jul 27, 2018

Use the packed_simd crate instead of std::simd

6473c28

`std::simd` was removed in rust-lang/rust#52535. https://crates.io/crates/packed_simd is in the rust-lang-nursery org.

SimonSapin mentioned this pull request Jul 27, 2018

Use the packed_simd crate instead of std::simd servo/servo#21272

Merged

SimonSapin added a commit to servo/servo that referenced this pull request Jul 27, 2018

Use the packed_simd crate instead of std::simd

1c0941f

`std::simd` was removed in rust-lang/rust#52535. https://crates.io/crates/packed_simd is in the rust-lang-nursery org.

termoshtt mentioned this pull request Aug 9, 2018

Cannot build latest nightly rust-math/sfmt#20

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Update stdsimd to undo an accidental stabilization #52535

Update stdsimd to undo an accidental stabilization #52535

alexcrichton commented Jul 19, 2018

rust-highfive commented Jul 19, 2018

rust-highfive commented Jul 19, 2018

alexcrichton commented Jul 19, 2018

bors commented Jul 19, 2018

alexcrichton commented Jul 19, 2018

kennytm commented Jul 20, 2018

alexcrichton commented Jul 20, 2018

alexcrichton commented Jul 20, 2018

bors commented Jul 20, 2018

gnzlbg commented Jul 20, 2018 via email

bors commented Jul 20, 2018

alexcrichton commented Jul 20, 2018

bors commented Jul 21, 2018

bors commented Jul 21, 2018

bors commented Jul 21, 2018

gnzlbg commented Jul 21, 2018 •

edited

Loading

bors commented Jul 21, 2018

alexcrichton commented Jul 21, 2018

gnzlbg commented Jul 21, 2018 via email

pitdicker Jul 22, 2018

alexcrichton Jul 22, 2018

pitdicker Jul 22, 2018

nnethercote commented Jul 25, 2018

gnzlbg commented Jul 25, 2018 •

edited

Loading

alexcrichton commented Jul 25, 2018

gnzlbg commented Jul 25, 2018 •

edited

Loading

MoSal commented Jul 25, 2018

gnzlbg commented Jul 25, 2018

MoSal commented Jul 25, 2018 •

edited

Loading

nnethercote commented Jul 26, 2018

Update stdsimd to undo an accidental stabilization #52535

Update stdsimd to undo an accidental stabilization #52535

Conversation

alexcrichton commented Jul 19, 2018

rust-highfive commented Jul 19, 2018

rust-highfive commented Jul 19, 2018

alexcrichton commented Jul 19, 2018

bors commented Jul 19, 2018

alexcrichton commented Jul 19, 2018

kennytm commented Jul 20, 2018

alexcrichton commented Jul 20, 2018

alexcrichton commented Jul 20, 2018

bors commented Jul 20, 2018

gnzlbg commented Jul 20, 2018 via email

bors commented Jul 20, 2018

alexcrichton commented Jul 20, 2018

bors commented Jul 21, 2018

bors commented Jul 21, 2018

bors commented Jul 21, 2018

gnzlbg commented Jul 21, 2018 • edited Loading

bors commented Jul 21, 2018

alexcrichton commented Jul 21, 2018

gnzlbg commented Jul 21, 2018 via email

pitdicker Jul 22, 2018

Choose a reason for hiding this comment

alexcrichton Jul 22, 2018

Choose a reason for hiding this comment

pitdicker Jul 22, 2018

Choose a reason for hiding this comment

nnethercote commented Jul 25, 2018

gnzlbg commented Jul 25, 2018 • edited Loading

alexcrichton commented Jul 25, 2018

gnzlbg commented Jul 25, 2018 • edited Loading

MoSal commented Jul 25, 2018

gnzlbg commented Jul 25, 2018

MoSal commented Jul 25, 2018 • edited Loading

nnethercote commented Jul 26, 2018

gnzlbg commented Jul 21, 2018 •

edited

Loading

gnzlbg commented Jul 25, 2018 •

edited

Loading

gnzlbg commented Jul 25, 2018 •

edited

Loading

MoSal commented Jul 25, 2018 •

edited

Loading