Page MenuHomePhabricator

asbirlea (Alina Sbirlea)
Google

Projects

User does not belong to any projects.

User Details

User Since
Feb 10 2016, 9:51 AM (279 w, 5 d)

Recent Activity

Today

asbirlea added a comment to D104641: [LICM] Strip context sensitive attributes after call hoisting.

It may be too much to overload the existing APIs, since this applies only to calls (e.g. in GVNHoist, metadata is dropped for geps, this would be unnecessary). So either doing the two calls independently as in this patch or add another API that does both.
I don't have strong opinions either way.

Mon, Jun 21, 1:43 PM · Restricted Project

Tue, Jun 15

asbirlea added a comment to D104180: [NFC] [LICM] Create LoopNest Invariant Code Motion (LNICM) pass.

Got it! Can you get some tests where some of these differences appear?
It would be a good motivator to see a concrete example; here are a couple of scenarios I can think of, but perhaps your use case is different.

  • LNICM can hoist additionally where LICM doesn't. If that can happen, we should show case with a test.
  • another loop nest pass, in the same LPM with LNICM, and how optimizations would differ if LICM was separated.
Tue, Jun 15, 11:21 AM · Restricted Project
asbirlea added a comment to D104180: [NFC] [LICM] Create LoopNest Invariant Code Motion (LNICM) pass.

Actually, LNICM is to enable other optimizations in loop pipeline that require perfect loop nest by hoisting innermost invariants out of loop nest at once.
I said that LNICM can efficiently perform LICM, but after tweaking the code and running some tests, I realized that it might not get faster.
However, even if it gets no faster, LNICM is worth adding as a new pass.

Tue, Jun 15, 10:47 AM · Restricted Project

Mon, Jun 14

asbirlea added a comment to D104238: [LoopDeletion] Check for irreducible cycles when deleting loops..

See clang-format suggestion, otherwise LGTM.

Mon, Jun 14, 1:49 PM · Restricted Project
asbirlea added a comment to D104180: [NFC] [LICM] Create LoopNest Invariant Code Motion (LNICM) pass.

I will second @fhahn's comments. Please include follow up patches with the improvements you mention, and tests that motivate those.

Mon, Jun 14, 12:24 PM · Restricted Project

Fri, Jun 4

asbirlea added a comment to D103016: NewGVN: Relax assertion about number of times a value is seen.

The issue in PR31613 is a legitimate bug, since two instructions that are not congruent end up in the same congruence class based on the PredicateInfo data. Increasing the limit will not help there since the exhibited behavior in much larger apps is "out of memory", the testcase I provided is only a reduced version triggering a debuggable assertion.

Fri, Jun 4, 3:36 PM · Restricted Project

Wed, May 26

asbirlea added a comment to D103009: [DSE] Transform memset + malloc --> calloc (PR25892).

This looks good to me, but please wait for the reviewers' comments from the previous patch version.

Wed, May 26, 10:57 PM · Restricted Project

May 20 2021

asbirlea added a comment to D102113: [MemCpyOpt] Remove MemDepAnalysis-based implementation.

I believe the process is to additionally test a release after it's been cut, and potential issues that arise can be fixed and cherry-picked into that release. Flipping back the flag would probably not be a useful fix after 6+month, rather resolving the bug would be.
As an additional data point, we've had several internal releases since the flag was flipped and haven't encountered any issues. I'm not saying bugs are sure not to exist, just that there has been some fairly thorough testing done and I'd opt to fix forward rather than switch back to the MemDepAnalysis implementation.
If you have strong reasons to hold back another couple of month on this cleanup, that's an option too, I'm just outlining my reasoning for moving forward.

My motivation to keep the code until the release was not so we can flip back the flag in the codebase; I agree we should just fix the bugs in the new version. I meant the flag may be helpful to end-users who cannot easily update after the release and could use the flag as workaround. It is all what-ifs at this point though :) If it makes planned upcoming work easier, removing the code now sounds good. If it is mostly a matter of removing unused code a bit earlier, I have a slight preference to keep it until the next release branch (I'm happy with the patch moving ahead, I mainly wanted to share another experience to consider)

May 20 2021, 11:59 AM · Restricted Project

May 18 2021

asbirlea accepted D102196: [NewPM] Add options to PrintPassInstrumentation.
May 18 2021, 4:25 PM · Restricted Project, Restricted Project
asbirlea added a comment to D102113: [MemCpyOpt] Remove MemDepAnalysis-based implementation.

I believe the process is to additionally test a release after it's been cut, and potential issues that arise can be fixed and cherry-picked into that release. Flipping back the flag would probably not be a useful fix after 6+month, rather resolving the bug would be.
As an additional data point, we've had several internal releases since the flag was flipped and haven't encountered any issues. I'm not saying bugs are sure not to exist, just that there has been some fairly thorough testing done and I'd opt to fix forward rather than switch back to the MemDepAnalysis implementation.
If you have strong reasons to hold back another couple of month on this cleanup, that's an option too, I'm just outlining my reasoning for moving forward.

May 18 2021, 3:26 PM · Restricted Project
asbirlea accepted D102032: [NewPM] Don't mark AA analyses as preserved.

LGTM.

May 18 2021, 1:29 PM · Restricted Project

May 17 2021

asbirlea accepted D102113: [MemCpyOpt] Remove MemDepAnalysis-based implementation.

Thank you for cleaning this up!

May 17 2021, 12:47 PM · Restricted Project

May 4 2021

asbirlea committed rGb14c8f5f6eb9: Add cal entry for MemorySSA syncs. (authored by asbirlea).
Add cal entry for MemorySSA syncs.
May 4 2021, 12:56 PM
asbirlea committed rG974ff623aa89: Add monthly MemorySSA sync. (authored by asbirlea).
Add monthly MemorySSA sync.
May 4 2021, 11:25 AM

Apr 27 2021

asbirlea accepted D100912: [docs][NewPM] Add section on analyses.

Thank you writing this, well written and useful doc.
No more big comments from my side, please wait to see if the others have additional comments.

Apr 27 2021, 12:14 PM · Restricted Project

Apr 24 2021

asbirlea added inline comments to D100912: [docs][NewPM] Add section on analyses.
Apr 24 2021, 12:09 PM · Restricted Project

Apr 23 2021

asbirlea accepted D100560: [NewGVN] Use ExprResult to add extra predicate users..

Thanks, looks good.
Could you add the const to resolve the clang-tidy warning?

Apr 23 2021, 12:41 PM · Restricted Project
asbirlea accepted D99990: [NewGVN] Use performSymbolicEvaluation instead of createExpression..

Minor typos in description.
s/no update the/no update of the
Hence I won't/Hence it won't

Apr 23 2021, 11:42 AM · Restricted Project

Apr 22 2021

asbirlea accepted D99987: [NewGVN] Track simplification dependencies for phi-of-ops..

Thank you! Looks good.

Apr 22 2021, 3:28 PM · Restricted Project
asbirlea added a comment to D99987: [NewGVN] Track simplification dependencies for phi-of-ops..

I think it makes sense to add the users at the upper call level where the real instruction is known. Just one comment inline.

Apr 22 2021, 1:58 PM · Restricted Project

Apr 20 2021

asbirlea accepted D100893: [FuncAttrs] Always preserve FunctionAnalysisManagerCGSCCProxy.

Sounds good, thank you!

Apr 20 2021, 3:19 PM · Restricted Project

Apr 14 2021

asbirlea accepted D66924: [NewGVN] Add phi-of-ops operands if no real PHI is created..

From my understanding so far, this is the correct approach.
I am going to need a few more days to go over the entire NewGVN pass functionality for the other patches, and I will only be back at work on Monday. Apologies for the delay on this.

Apr 14 2021, 3:38 PM · Restricted Project
asbirlea added a comment to D100470: [LICM][NFC] Fix typo.

Thank you for the typo fixes.

Apr 14 2021, 11:29 AM · Restricted Project

Apr 13 2021

asbirlea abandoned D74640: [LoopRotate] Add explicit flag to require MSSA..
Apr 13 2021, 11:49 AM · Restricted Project
asbirlea added a comment to D100163: [MSSA] Rename uses in IDF regardless of new def position in basic block..

Test added in 1e0b813fc082.

Apr 13 2021, 10:58 AM · Restricted Project
asbirlea committed rG1e0b813fc082: [MemorySSA] Add test for PR49859. (authored by asbirlea).
[MemorySSA] Add test for PR49859.
Apr 13 2021, 10:57 AM

Apr 12 2021

asbirlea added a comment to D98103: [NPM] Do not run function simplification pipeline unnecessarily.

I agree with @aeubanks on removing the NFC from the title. This is changing the functionality, even if it's under a disabled flag.

Apr 12 2021, 3:37 PM · Restricted Project
asbirlea accepted D98103: [NPM] Do not run function simplification pipeline unnecessarily.
Apr 12 2021, 11:02 AM · Restricted Project

Apr 9 2021

asbirlea committed rGf6bff8d15778: [MSSA] Rename uses in IDF regardless of new def position in basic block. (authored by asbirlea).
[MSSA] Rename uses in IDF regardless of new def position in basic block.
Apr 9 2021, 12:33 PM
asbirlea closed D100163: [MSSA] Rename uses in IDF regardless of new def position in basic block..
Apr 9 2021, 12:33 PM · Restricted Project
asbirlea requested review of D100163: [MSSA] Rename uses in IDF regardless of new def position in basic block..
Apr 9 2021, 12:00 AM · Restricted Project

Apr 8 2021

asbirlea added a comment to D95543: [GVN] Clobber partially aliased loads..

Performance-wise this looks good.
Please wait to see if the other reviewers have comments on this.

Apr 8 2021, 10:39 PM · Restricted Project
asbirlea accepted D98027: [NFC][AA] Prepare to convert AliasResult to class with PartialAlias offset..
Apr 8 2021, 10:33 PM · Restricted Project
asbirlea accepted D98718: [AA][NFC] Convert AliasResult to class containing offset for PartialAlias case..

This is good to go from my side.

Apr 8 2021, 10:33 PM · Restricted Project
asbirlea added a comment to D98103: [NPM] Do not run function simplification pipeline unnecessarily.

Restating summary of offline discussion:
Up until now, the repetition of simplification passes has hidden phase ordering issues or the need to rerun certain passes. Turning this flag on will improve compile time and resolve potential exponential compile-time increases, but will also reveal all the cases where the pass pipeline needs to be adjusted to not miss optimizations that were previously done on the follow-up optimization runs.

Apr 8 2021, 11:49 AM · Restricted Project

Apr 6 2021

asbirlea added a comment to D98288: [DSE] Translate killing locations through phis..

I'm seeing a crash with this. Repro: opt -passes=dse on:

Apr 6 2021, 3:59 PM · Restricted Project
asbirlea added a comment to D98718: [AA][NFC] Convert AliasResult to class containing offset for PartialAlias case..

I'm running some additional tests for this patch.
Could you rebase the GVN dependent patch so I get some testing done in parallel?

Apr 6 2021, 11:26 AM · Restricted Project

Apr 5 2021

asbirlea accepted D99843: [LoopRotate] Don't split loop pass manager.
Apr 5 2021, 8:58 AM · Restricted Project

Mar 31 2021

asbirlea accepted D99249: [PassManager] Run additional LICM before LoopRotate.

Most performance results look ok. There are a few that are mixed (+ & -), one regression that's still investigated that so far seems architecture specific and not related to this patch, a couple of noticeable improvements and a few marginal improvements.
Please note I have only tested the new pass manager.
So, green light from my side assuming the other reviewers are also on board with this.

Mar 31 2021, 10:40 AM · Restricted Project

Mar 24 2021

asbirlea added a comment to D99249: [PassManager] Run additional LICM before LoopRotate.

Depending on what the run-time impact is in practice, the additional compile-time may or may not be worth it.
I'm running some additional testing offline on this, please give me a couple of days to test and analyze.

Mar 24 2021, 10:35 AM · Restricted Project

Mar 18 2021

asbirlea accepted D98820: [NewPM] Verify LoopAnalysisResults after a loop pass.
Mar 18 2021, 6:16 PM · Restricted Project

Mar 17 2021

asbirlea accepted D98805: [NewPM] Verify LoopAnalysisResults after a loop pass.

Seems reasonable to verify MSSA unconditionally at this point but if the debug build becomes too overloaded, it can be put under the VerifyMemorySSA flag, like in other loop passes.

Mar 17 2021, 12:55 PM · Restricted Project
asbirlea added a comment to D98718: [AA][NFC] Convert AliasResult to class containing offset for PartialAlias case..

@nikic, could you check this patch through the compiler-tracker, please?

Mar 17 2021, 11:34 AM · Restricted Project

Mar 12 2021

asbirlea accepted D98484: add missing MemorySSAWrapperPass dependency macro to MemCpyOptLegacyPass.
Mar 12 2021, 6:24 PM · Restricted Project
asbirlea added a comment to D98027: [NFC][AA] Prepare to convert AliasResult to class with PartialAlias offset..

This change looks fine. Could you update the follow-up as well as a dependent patch?

Mar 12 2021, 5:19 PM · Restricted Project
asbirlea accepted D98557: [MemorySSA] Don't bail on phi starting access.

Your point is valid and the patch looks good.

Mar 12 2021, 5:11 PM · Restricted Project
asbirlea added a comment to D98469: [NVPTX] Avoid temp copy of byval kernel parameters..

Updates look good.

Mar 12 2021, 4:22 PM · Restricted Project

Mar 11 2021

asbirlea accepted D94619: [SplitEdge] Update SplitCriticalEdge to return a nullptr only when the edge is not critical.

I'd have liked the thoughts of folks familiar with coroutines here, but overall the change looks like an improvement.

Mar 11 2021, 5:35 PM · Restricted Project
asbirlea added inline comments to D98469: [NVPTX] Avoid temp copy of byval kernel parameters..
Mar 11 2021, 5:05 PM · Restricted Project
asbirlea accepted D98440: [NPM][CGSCC] FunctionAnalysisManagerCGSCCProxy: do not clear immutable function passes.
Mar 11 2021, 4:52 PM · Restricted Project, Restricted Project

Mar 10 2021

asbirlea accepted D89264: [LICM] Make promotion faster.

LGTM.

Mar 10 2021, 5:20 PM · Restricted Project

Mar 9 2021

asbirlea accepted D98284: [DSE] Handle memcpy/memset with equal non-const sizes..

Changes looks good.
+1 on the additional testcase.

Mar 9 2021, 2:03 PM · Restricted Project

Mar 8 2021

asbirlea added a reverting change for rG3d8f842712d4: [LICM] Make promotion faster: rG29482426b58e: Revert "[LICM] Make promotion faster".
Mar 8 2021, 12:56 PM
asbirlea committed rG29482426b58e: Revert "[LICM] Make promotion faster" (authored by asbirlea).
Revert "[LICM] Make promotion faster"
Mar 8 2021, 12:56 PM
asbirlea added a reverting change for D89264: [LICM] Make promotion faster: rG29482426b58e: Revert "[LICM] Make promotion faster".
Mar 8 2021, 12:56 PM · Restricted Project

Mar 2 2021

asbirlea accepted D89264: [LICM] Make promotion faster.
Mar 2 2021, 12:58 PM · Restricted Project

Feb 26 2021

asbirlea accepted D97515: [docs] Add documentation on using the new pass manager.

Thank you for adding this, it was long overdue and it is well written!
LGTM.

Feb 26 2021, 3:19 PM · Restricted Project

Feb 18 2021

asbirlea added a comment to D89264: [LICM] Make promotion faster.

Runtime impact looks reasonable now. What's the compile-time impact for the patch now?

Feb 18 2021, 3:24 PM · Restricted Project
asbirlea accepted D96996: [BasicAA] Add simple depth limit to avoid stack overflow (PR49151).

sgtm.

Feb 18 2021, 2:52 PM · Restricted Project
asbirlea added a comment to D96647: [BasicAA] Add depth limit.

I tested this with 512 threshold and I'm only seeing a notable runtime regression in the testsuite multisource in MallocBench. I would not block on that, the stack overflow issue is a larger concern.

Feb 18 2021, 1:40 PM · Restricted Project

Feb 17 2021

asbirlea accepted D96727: [NPM] Properly reset parent loop after loop passes.

Fix looks good. As a detail, perhaps move the NDEBUG inside the body of setParentLoop to keep the code cleaner?

Feb 17 2021, 5:58 PM · Restricted Project
asbirlea added a comment to D89264: [LICM] Make promotion faster.

Ack, testing in progress.

Feb 17 2021, 2:02 PM · Restricted Project
asbirlea accepted D94376: [MemCpyOpt] Enable MemorySSA by default.

This looks good on my end, including run time testing.

Feb 17 2021, 1:24 PM · Restricted Project, Restricted Project

Feb 16 2021

asbirlea requested changes to D96754: [NewPM] Use stale divergence analysis with SimpleLoopUnswitch.

This is absolutely not the right way resolve this.
First, the restriction to not allow getting a stale analysis is very much intentional and part of the design of the new pass manager. The API being added here must not exist.
Second, it is not safe to use the stale DA. LoopUnswitch gets the LegacyDivergenceAnalysis only when making the final unswitching decision, and it does not reuse a stale instance. The same needs to happen in SimpleLoopUnswitch.
A proper solution is to change the divergence analysis pass so it can be created as an object. An example of a pass used as an object is the OptimizationRemarkEmitter.

Feb 16 2021, 12:13 PM · Restricted Project

Feb 9 2021

asbirlea accepted D96101: [opt][NewPM] Add a --print-passes flag to print all available passes.
Feb 9 2021, 4:32 PM · Restricted Project

Feb 8 2021

asbirlea accepted D96288: [SimpleLoopUnswitch] Don't non-trivially unswitch loops that are unsafe to clone.
Feb 8 2021, 1:04 PM · Restricted Project

Feb 5 2021

asbirlea accepted D96100: Specify that some flags are legacy PM-specific.
Feb 5 2021, 2:11 PM · Restricted Project

Feb 3 2021

asbirlea accepted D95763: [MemorySSA] Don't treat lifetime.end as NoAlias.

This looks good!

Feb 3 2021, 4:47 PM · Restricted Project

Feb 2 2021

asbirlea added a comment to D95690: [LoopVectorize] improve IR fast-math-flags propagation in reductions.

Here's what the test reduced to. To reproduce run opt -passes='loop-vectorize' test.ll with the assert in place.

target datalayout = "e-m:e-p:32:32-i64:64-n32:64-S128"
target triple = "wasm32-unknown--wasm"
Feb 2 2021, 11:09 PM · Restricted Project
asbirlea accepted D95380: Turn on the new pass manager by default.
Feb 2 2021, 2:21 PM · Restricted Project

Feb 1 2021

asbirlea added a comment to D95763: [MemorySSA] Don't treat lifetime.end as NoAlias.

I'm not familiar with the original reasoning, so here are some thoughts.

Feb 1 2021, 2:00 PM · Restricted Project
asbirlea accepted D95796: [NewPM][Unswitch] Add option to disable -O3 non-trivial unswitching.

lgtm.

Feb 1 2021, 10:29 AM · Restricted Project

Jan 23 2021

asbirlea added a comment to D94376: [MemCpyOpt] Enable MemorySSA by default.

I'm seeing crashes with this at stage3 in llvm::MemCpyOptPass::processMemCpyMemCpyDependence when invoking MSSA's Walker. I should have a repro next week.

Jan 23 2021, 12:03 AM · Restricted Project, Restricted Project

Jan 22 2021

asbirlea accepted D93503: [GVN] Propagate llvm.access.group metadata of loads.

LGTM.
Please wait to see if @fhahn has additional feedback.

Jan 22 2021, 11:53 PM · Restricted Project
asbirlea added a comment to D93529: [AA] Cache (optionally) estimated PartialAlias offsets..

Thanks! Diff looks reasonable to me, but I'd like to see the use case changes as well. Could you (re-)upload a concrete use case as dependent patch?
One reason I have in mind is we may want to concretely enable this caching solely for a particular BatchAA instance, i.e. pass true to BatchAA and change the meaning in AAQI from BatchMode to CacheOffsets (and, same as now, only allow BatchAA to set this bool). Then only pass true at the BatchAA callsites where it will be used.

Jan 22 2021, 11:39 PM · Restricted Project
asbirlea accepted D95254: [NewPM][opt] Make -enable-new-pm default to LLVM_ENABLE_NEW_PASS_MANAGER.

Woohoo!

Jan 22 2021, 5:50 PM · Restricted Project
asbirlea added inline comments to D93039: Introduce llvm.noalias.decl intrinsic.
Jan 22 2021, 10:22 AM · Restricted Project

Jan 21 2021

asbirlea added a comment to D92887: [LoopUnroll] Use llvm.experimental.noalias.scope.decl for duplicating noalias metadata as needed.

My concern with this patch is in the iteration on all instructions in the blocks in cloneAndAdaptNoAliasScopes and identifyNoAliasScopesToClone.
Perhaps for LoopUnroll is not an issue, if unroll cost model is taken into account before this (i.e. unroll doesn't happen for large BBs). But when doing loop rotate, this may explode for IRs with large BBs, so I'd say some measurements would be helpful here.
For loop rotate, which instructions are cloned from the header to the preheader is known at the callsite, so as to not iterate over the entire original preheader, but I'm not sure this is sufficient.

Jan 21 2021, 5:40 PM · Restricted Project
asbirlea added a comment to D93040: [InlineFunction] Use llvm.experimental.noalias.scope.decl for noalias arguments..

FWIW, this also lgtm.

Jan 21 2021, 5:07 PM · Restricted Project, Restricted Project
asbirlea added a comment to rGb1c2f1282a23: [BasicAA] Move assumption tracking into AAQI.

Thanks for the fix.

Jan 21 2021, 4:48 PM
asbirlea added a comment to D93039: Introduce llvm.noalias.decl intrinsic.

Forgot to send this nits, feel free to add in a follow up.

Jan 21 2021, 4:45 PM · Restricted Project
asbirlea requested changes to D93529: [AA] Cache (optionally) estimated PartialAlias offsets..

Adding this only seems to make sense when using BatchAA.
Could we restrict populating this cache? e.g. extend AAQueryInfo constructor to take bool BatchMode = false, and set it to true for BatchAA. Then update the get/set methods to first check this (check in the setter and assert in the getter).

Jan 21 2021, 4:22 PM · Restricted Project
asbirlea added inline comments to D94630: [LTO] Add support for existing Config::Freestanding option..
Jan 21 2021, 1:24 PM · Restricted Project
asbirlea accepted D95117: [NewPM][opt] Run the "default" AA pipeline by default.

lgtm.

Jan 21 2021, 1:08 PM · Restricted Project
asbirlea added a comment to D95117: [NewPM][opt] Run the "default" AA pipeline by default.

Could we have at least one test showing how to pass an empty pipeline, solely for testing purposes?
AFAICT, currently all tests use the 'default', 'basic-aa' or 'tbaa, basic-aa'. Potentially add the test to test/Other/pass-pipeline-parsing.ll, which is testing 'bad' AA pipeline, or in one of the AA test folders.

Jan 21 2021, 9:11 AM · Restricted Project

Jan 14 2021

asbirlea added a comment to D94363: [AA] Pass AAResults through AAQueryInfo.

Do you have an idea if there is overhead of adding the AAR object to each query (each query info object)? (vs the setAAResults being done once for all the AAs available).

Jan 14 2021, 4:50 PM · Restricted Project

Jan 13 2021

asbirlea accepted D90094: [BasicAA] Handle recursive queries more efficiently (NFCI).

lgtm.
Runtime impact results look good.

Jan 13 2021, 4:15 PM · Restricted Project
asbirlea accepted D94559: [NewPM] Only non-trivially loop unswitch at -O3 and for non-optsize functions.

Thank you for confirming @SjoerdMeijer.

Jan 13 2021, 1:11 PM · Restricted Project

Jan 12 2021

asbirlea added a comment to D94376: [MemCpyOpt] Enable MemorySSA by default.

Could you test clang as well, and update test?

Jan 12 2021, 3:42 PM · Restricted Project, Restricted Project
asbirlea added a comment to D94448: [NewPM] Run non-trivial loop unswitching under -O2/3/s/z.

I'm curious if the compile time impact is complemented by impact in run time. Looking into that for O3.

Jan 12 2021, 3:00 PM · Restricted Project

Jan 11 2021

asbirlea accepted D94448: [NewPM] Run non-trivial loop unswitching under -O2/3/s/z.

lgtm.

Jan 11 2021, 2:28 PM · Restricted Project

Jan 7 2021

asbirlea added a comment to D93743: [NewPM] Run ObjC ARC passes.

The changes LGTM, but I'd rather defer final review to ahatanak@.
It would be good to have some testing but it could be addressed in a separate patch, potentially by folks more familiar with the ObjCARC passes.

Jan 7 2021, 3:26 PM · Restricted Project, Restricted Project
asbirlea added a reviewer for D93929: [NewPM][Hexagon] Fix HexagonVectorLoopCarriedReusePass position in pipeline: quic_aankit.
Jan 7 2021, 3:17 PM · Restricted Project

Jan 6 2021

asbirlea accepted D92486: Set option default for enabling memory ssa for new pass manager loop sink pass to true..

I believe for the NPM this should work fine.
@nikic @fhahn Any objections?

Jan 6 2021, 6:30 PM · Restricted Project
asbirlea accepted D94176: [SplitEdge] Add new parameter to SplitEdge to name the newly created basic block.

lgtm.

Jan 6 2021, 6:21 PM · Restricted Project
asbirlea added a comment to D93503: [GVN] Propagate llvm.access.group metadata of loads.

Looks reasonable to me as well, that the metadata info should hold at the same loop level.

Jan 6 2021, 5:54 PM · Restricted Project
asbirlea committed rG63aeaf754a78: [DominatorTree] Add support for mixed pre/post CFG views. (authored by asbirlea).
[DominatorTree] Add support for mixed pre/post CFG views.
Jan 6 2021, 2:54 PM
asbirlea closed D93371: [DominatorTree] Add support for mixed pre/post CFG views..
Jan 6 2021, 2:53 PM · Restricted Project

Jan 5 2021

asbirlea accepted D91936: [BasicAA] Fix BatchAA results for phi-phi assumptions.

Apologies for the delay, I spent some more time understanding how the assumptions changed in the new implementation.

Jan 5 2021, 5:50 PM · Restricted Project
asbirlea accepted D94085: [simd_ops] Build tests with -fpic.
Jan 5 2021, 10:13 AM · Restricted Project