Disable the `SimplifyArmIdentity` pass by wesleywiser · Pull Request #73262 · rust-lang/rust

wesleywiser · 2020-06-12T01:04:08Z

This pass is buggy so I'm disabling it to fix a stable-to-beta
regression.

Related to #73223

rust-highfive · 2020-06-12T01:04:12Z

r? @eddyb

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

oli-obk · 2020-06-12T09:03:27Z

If we aren't in a rush about this, please open a PR against master with this change that we can backport to beta.

This pass is buggy so I'm disabling it to fix a stable-to-beta regression. Related to rust-lang#73223

bors · 2020-06-12T12:43:02Z

☔ The latest upstream changes made this pull request unmergeable. Please resolve the merge conflicts.

wesleywiser · 2020-06-12T12:44:02Z

Rebased

oli-obk · 2020-06-12T12:50:57Z

@bors r+

bors · 2020-06-12T12:50:58Z

📌 Commit f160583 has been approved by oli-obk

wesleywiser · 2020-06-12T13:43:02Z

Since the linked PR is P-critical, I'm raising the priority of this.

@bors p=1

Dylan-DPC-zz · 2020-06-12T18:32:29Z

bumping this higher to run after the current rollup

@bors p=3

bors · 2020-06-12T18:46:55Z

⌛ Testing commit f160583 with merge 664f6df82aa3cc4f3ad1eec7dd277b061938f7da...

bors · 2020-06-12T18:52:30Z

💔 Test failed - checks-azure

Dylan-DPC-zz · 2020-06-12T19:15:27Z

@bors retry

Dylan-DPC-zz · 2020-06-12T19:15:35Z

@bors retry

Dylan-DPC-zz · 2020-06-12T19:37:53Z

@bors retru p=1000

Dylan-DPC-zz · 2020-06-12T19:38:10Z

@bors retry p=1000

bors · 2020-06-12T19:38:22Z

⌛ Testing commit f160583 with merge f4fbb93...

Dylan-DPC-zz · 2020-06-12T23:04:39Z

@bors treeclosed-

Dylan-DPC-zz · 2020-06-12T23:04:48Z

@bors treeclosed--

bors · 2020-06-12T23:11:27Z

☀️ Test successful - checks-azure
Approved by: oli-obk
Pushing f4fbb93 to master...

nnethercote · 2020-06-15T23:58:32Z

This caused a small perf regression in a couple of opt benchmarks.

@wesleywiser: will this pass be re-enabled in the future?

wesleywiser · 2020-06-16T00:00:02Z

@nnethercote Yes! I'm working on finishing up some changes that will re-enable this on nightly.

pnkfelix · 2020-06-17T17:40:44Z

The manner in which this was disabled seems relevant to a recent wg-mir-opt conversation on zulip.

I guess if opt-level >= 2 is generally known to denote potentially buggy optimizations, then what's done here is okay.

But I don't know if that's the consensus based on the above linked conversation...

pnkfelix · 2020-06-17T17:42:53Z

(I was tempted to unilaterally beta approve this, but I'm hesitating due to not knowing if it matches the semantics everyone else expects from the mir-opt levels. We'll just discuss as a group tomorrow instead.)

wesleywiser · 2020-06-17T18:54:26Z

@pnkfelix The current state is that mir-opt-level >= 2 has buggy optimizations. There isn't too much of a semantic difference between 2 or 3 currently although I am in favor of moving to a scheme where mir-opt-level=0 is no optimizations, mir-opt-level=1 is stable optimizations, mir-opt-level=2 is unstable but nearly correct optimizations (or perhaps just optimizations that hurt compile-time), mir-opt-level=3 is code we want to keep from bit-rotting but has issues or something similar.

nikomatsakis · 2020-06-18T14:37:58Z

@wesleywiser I presume you'll add some tests that demonstrate the bug, too?

wesleywiser · 2020-06-18T14:38:59Z

@nikomatsakis Yes, I'm adding some regression tests in the next PR.

pnkfelix · 2020-06-18T14:41:19Z

Discussed at T-compiler meeting; approved for beta backport

RalfJung · 2020-06-22T09:26:44Z

 impl<'tcx> MirPass<'tcx> for SimplifyArmIdentity {
- fn run_pass(&self, _: TyCtxt<'tcx>, source: MirSource<'tcx>, body: &mut Body<'tcx>) {
+ fn run_pass(&self, tcx: TyCtxt<'tcx>, source: MirSource<'tcx>, body: &mut Body<'tcx>) {
+ if tcx.sess.opts.debugging_opts.mir_opt_level < 2 {


Uh, could we maybe not keep buggy passes running at all? I think a -Zmir-opt-level=3 miscompilation is a soundness bug just like it would be at a lower optimization level.
C/C++ programmers seem to associate "high optimization levels" with "creates buggy code", but really that's awful and we shouldn't do it.

@RalfJung I opened a related MCP: rust-lang/compiler-team#319
If you have thoughts about it, please leave a comment there. Thanks!

@ghost

…ulacrum [beta] backports This PR backports the following: * Make novel structural match violations not a `bug` rust-lang#73446 * linker: Never pass `-no-pie` to non-gnu linkers rust-lang#73384 * Disable the `SimplifyArmIdentity` pass rust-lang#73262 * Allow inference regions when relating consts rust-lang#73225 * Fix link error with #[thread_local] introduced by rust-lang#71192 rust-lang#73065 * Ensure stack when building MIR for matches rust-lang#72941 * Don't create impl candidates when obligation contains errors rust-lang#73005 r? @ghost

rust-highfive assigned eddyb Jun 12, 2020

This comment has been minimized.
Sign in to view

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Jun 12, 2020

Disable the SimplifyArmIdentity pass on beta
f160583
This pass is buggy so I'm disabling it to fix a stable-to-beta regression. Related to rust-lang#73223

wesleywiser changed the base branch from beta to masterJune 12, 2020 12:42

wesleywiser force-pushed the simplifyarmidentity_beta_regression branch from 5f63f6d to f160583Compare June 12, 2020 12:43

oli-obk added the beta-nominated Nominated for backporting to the compiler in the beta channel. label Jun 12, 2020

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 Jun 12, 2020

Mark-Simulacrum changed the title ~~Disable the SimplifyArmIdentity pass on beta~~Disable the SimplifyArmIdentity passJun 12, 2020

bors added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels Jun 12, 2020

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 Jun 12, 2020

bors added the merged-by-bors This PR was explicitly merged by bors. label Jun 12, 2020

bors merged commit f4fbb93 into rust-lang:masterJun 12, 2020

spastorino added the T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. label Jun 17, 2020

pnkfelix added the beta-accepted Accepted for backporting to the compiler in the beta channel. label Jun 18, 2020

RalfJung reviewed Jun 22, 2020
View reviewed changes

RalfJung mentioned this pull request Jun 25, 2020
Slice index becomes wrong (beta regression) #73223
Closed

Mark-Simulacrum removed the beta-nominated Nominated for backporting to the compiler in the beta channel. label Jun 26, 2020

Mark-Simulacrum mentioned this pull request Jun 26, 2020
[beta] backports #73766
Merged

wesleywiser mentioned this pull request Oct 9, 2020
Remove SimplifyBranchSame MIR optimization #77706
Closed

cuviper added this to the 1.46 milestone May 2, 2024

Uh oh!

Uh oh!

Conversation

wesleywiser commented Jun 12, 2020

Uh oh!

rust-highfive commented Jun 12, 2020

Uh oh!

This comment has been minimized.

oli-obk commented Jun 12, 2020

Uh oh!

bors commented Jun 12, 2020

Uh oh!

wesleywiser commented Jun 12, 2020

Uh oh!

oli-obk commented Jun 12, 2020

Uh oh!

bors commented Jun 12, 2020

Uh oh!

wesleywiser commented Jun 12, 2020

Uh oh!

Dylan-DPC-zz commented Jun 12, 2020

Uh oh!

bors commented Jun 12, 2020

Uh oh!

bors commented Jun 12, 2020

Uh oh!

Dylan-DPC-zz commented Jun 12, 2020

Uh oh!

Dylan-DPC-zz commented Jun 12, 2020

Uh oh!

Dylan-DPC-zz commented Jun 12, 2020

Uh oh!

Dylan-DPC-zz commented Jun 12, 2020

Uh oh!

bors commented Jun 12, 2020

Uh oh!

Dylan-DPC-zz commented Jun 12, 2020

Uh oh!

Dylan-DPC-zz commented Jun 12, 2020

Uh oh!

bors commented Jun 12, 2020

Uh oh!

nnethercote commented Jun 15, 2020

Uh oh!

wesleywiser commented Jun 16, 2020

Uh oh!

pnkfelix commented Jun 17, 2020

Uh oh!

pnkfelix commented Jun 17, 2020

Uh oh!

wesleywiser commented Jun 17, 2020

Uh oh!

nikomatsakis commented Jun 18, 2020

Uh oh!

wesleywiser commented Jun 18, 2020

Uh oh!

pnkfelix commented Jun 18, 2020

Uh oh!

RalfJungJun 22, 2020

Choose a reason for hiding this comment

Uh oh!

wesleywiserJun 23, 2020

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

13 participants