Page MenuHomePhabricator

george.burgess.iv (George Burgess IV)
User

Projects

User does not belong to any projects.

User Details

User Since
Jul 15 2014, 12:28 PM (236 w, 1 d)

Recent Activity

Yesterday

george.burgess.iv added inline comments to D57129: [SimpleLoopUnswitch] Early check exit for trivial unswitch with MemorySSA..
Wed, Jan 23, 5:44 PM

Tue, Jan 15

george.burgess.iv added a comment to D56760: Add a new builtin: __builtin_dynamic_object_size.

Thanks for this!

Tue, Jan 15, 6:35 PM
george.burgess.iv added a comment to D56720: [MemorySSA] Add caching results of reaching LiveOnEntry MemoryDef to ClobberWalker.

Thanks for this!

Tue, Jan 15, 12:01 PM

Tue, Jan 8

george.burgess.iv accepted D56467: [MemorySSA] Remove optimized value when reseting optimized..

Thanks!

Tue, Jan 8, 7:27 PM

Fri, Jan 4

george.burgess.iv accepted D55944: [MemorySSA] Extend the clobber walker with the option to skip the starting access..

LGTM; thanks again.

Fri, Jan 4, 9:41 AM
george.burgess.iv added inline comments to D56285: [MemorySSA] Add SkipSelfWalker..
Fri, Jan 4, 9:32 AM

Thu, Jan 3

george.burgess.iv accepted D56285: [MemorySSA] Add SkipSelfWalker..

LGTM modulo a naming nit. If you're happy with the suggestion, feel free to submit. Thanks!

Thu, Jan 3, 1:41 PM
george.burgess.iv accepted D55957: [MemorySSA] Refactor CachingWalker..

Thanks!

Thu, Jan 3, 11:18 AM
george.burgess.iv added a comment to D55957: [MemorySSA] Refactor CachingWalker..

This LGTM at a high level; just a handful of nits for you, and I think this should be good to submit

Thu, Jan 3, 9:45 AM

Wed, Jan 2

george.burgess.iv added a comment to D55957: [MemorySSA] Refactor CachingWalker..

Thanks for this!

Wed, Jan 2, 9:07 AM
george.burgess.iv added a comment to D55944: [MemorySSA] Extend the clobber walker with the option to skip the starting access..

Thanks for this! I like it when patches are small and split up like this one :)

Wed, Jan 2, 9:07 AM

Dec 15 2018

george.burgess.iv added a comment to D55641: [CallSite removal] Migrate all Alias Analysis APIs to use the newly minted `CallBase` class instead of the `CallSite` wrapper..

MemorySSA changes LGTM, modulo one comment. Thanks!

Dec 15 2018, 11:42 PM

Dec 13 2018

george.burgess.iv added a comment to D55679: Place .text.hot section before .text..

Chrome OS developed a dependency (perf, not functional) on things in .text.hot coming before things in .text. Namely, they have code that maps the first N MB of Chrome's .text section to hugepages, under the assumption that the code near the beginning of the binary will be hotter than the code near the end of the binary.

Dec 13 2018, 3:58 PM

Dec 4 2018

george.burgess.iv accepted D55268: HowToBuildWithPGO.rst: Fix a few details in the manual steps.

(Also thanks for the reminder that I still need to update this with commentary for the purely-cmake-based approach. :) )

Dec 4 2018, 9:30 AM

Dec 3 2018

george.burgess.iv added a comment to D55060: [ThinLTO] Look through aliases in cache key calculations.

Thanks again! :)

Dec 3 2018, 4:05 PM
george.burgess.iv added a comment to D55060: [ThinLTO] Look through aliases in cache key calculations.

Thanks for the feedback!

Dec 3 2018, 10:42 AM
george.burgess.iv updated the diff for D55060: [ThinLTO] Look through aliases in cache key calculations.

Addressed feedback

Dec 3 2018, 10:41 AM

Nov 29 2018

george.burgess.iv created D55060: [ThinLTO] Look through aliases in cache key calculations.
Nov 29 2018, 9:39 AM

Nov 6 2018

george.burgess.iv added a comment to D4276: Added llvm.is.constant intrinsic.

It may make sense to iterate, as you say, but that can/should be a different patch.

Nov 6 2018, 2:07 PM · deleted

Nov 3 2018

george.burgess.iv added a comment to D53598: Add docs+a script for building clang/LLVM with PGO.

I notice that you're using LLVM_BUILD_INSTRUMENTED=IR, which corresponds to -fprofile-generate (IR-level profiling), instead of -fprofile-instr-generate (clang-level profiling). Did you play around with both and observe that IR-level profiling gave you better results?

Nov 3 2018, 4:13 PM

Oct 31 2018

george.burgess.iv added a comment to D53598: Add docs+a script for building clang/LLVM with PGO.

Thanks for pointing that out!

Oct 31 2018, 2:45 PM

Oct 30 2018

george.burgess.iv added a comment to D49103: Lower llvm.objectsize earlier in our optimization pipeline.

And numbers for building ${large_project}, as promised:

Oct 30 2018, 7:02 PM

Oct 29 2018

george.burgess.iv updated the diff for D49103: Lower llvm.objectsize earlier in our optimization pipeline.

Rebase, and only run this in the non-prelink full-LTO pipeline

Oct 29 2018, 6:55 PM
george.burgess.iv added a comment to D49103: Lower llvm.objectsize earlier in our optimization pipeline.

You mentioned a 1.9% degradation with the current placement. Did you happen to try any other locations in the pipeline?

Oct 29 2018, 6:19 PM
george.burgess.iv added a comment to D49103: Lower llvm.objectsize earlier in our optimization pipeline.

Ping :)

Oct 29 2018, 1:39 PM

Oct 27 2018

george.burgess.iv accepted D53795: [utils] Fix _run_benchmark in collect_and_build_with_pgo.py.

LGTM assuming the _run_benchmark changes are undone.

Oct 27 2018, 3:26 PM

Oct 26 2018

george.burgess.iv added a comment to D53598: Add docs+a script for building clang/LLVM with PGO.

Thanks again!

Oct 26 2018, 1:58 PM

Oct 25 2018

george.burgess.iv added a comment to D53598: Add docs+a script for building clang/LLVM with PGO.

Thank you!

Oct 25 2018, 6:19 PM
george.burgess.iv updated the diff for D53598: Add docs+a script for building clang/LLVM with PGO.

Address feedback

Oct 25 2018, 6:18 PM

Oct 23 2018

george.burgess.iv created D53598: Add docs+a script for building clang/LLVM with PGO.
Oct 23 2018, 11:59 AM

Oct 20 2018

george.burgess.iv abandoned D52245: No review requested: precursor changes for D44748.

Landed these a while back; closing.

Oct 20 2018, 1:19 PM

Oct 9 2018

george.burgess.iv added a comment to D44748: Track whether the size of a MemoryLocation is precise.

Thanks again! :)

Oct 9 2018, 11:41 PM
george.burgess.iv accepted D52924: Make __builtin_object_size use the EM_IgnoreSideEffects evaluation mode..

Thanks!

Oct 9 2018, 1:12 PM

Oct 4 2018

george.burgess.iv added a comment to D44748: Track whether the size of a MemoryLocation is precise.

Woo! Thank you very much for the review (and for bearing with me in your repeated requests to minimize this patch :) ).

Oct 4 2018, 8:17 PM

Oct 3 2018

george.burgess.iv added inline comments to D44748: Track whether the size of a MemoryLocation is precise.
Oct 3 2018, 10:38 AM
george.burgess.iv updated the diff for D44748: Track whether the size of a MemoryLocation is precise.
Oct 3 2018, 10:38 AM

Sep 28 2018

george.burgess.iv added inline comments to D44748: Track whether the size of a MemoryLocation is precise.
Sep 28 2018, 3:53 PM
george.burgess.iv updated the diff for D44748: Track whether the size of a MemoryLocation is precise.

Address feedback

Sep 28 2018, 3:52 PM

Sep 24 2018

george.burgess.iv added inline comments to D51664: [IR] Lazily number instructions for local dominance queries.
Sep 24 2018, 11:15 AM

Sep 21 2018

george.burgess.iv accepted D52268: [AST] Squeeze some bits in LinkageComputer.

LGTM too.

Sep 21 2018, 11:22 AM · Restricted Project

Sep 19 2018

george.burgess.iv added a comment to D52268: [AST] Squeeze some bits in LinkageComputer.

Thanks for this! LGTM after erichkeane's comments are resolved.

Sep 19 2018, 3:47 PM · Restricted Project

Sep 18 2018

george.burgess.iv updated the diff for D44748: Track whether the size of a MemoryLocation is precise.

Shrunk the diff as requested; the few pre-commit patches required to make this build are available as one patch in https://reviews.llvm.org/D52245 .

Sep 18 2018, 12:55 PM
george.burgess.iv added inline comments to D44748: Track whether the size of a MemoryLocation is precise.
Sep 18 2018, 12:55 PM
george.burgess.iv created D52245: No review requested: precursor changes for D44748.
Sep 18 2018, 12:49 PM

Sep 12 2018

george.burgess.iv added a comment to D51664: [IR] Lazily number instructions for local dominance queries.

This looks good to me, but like said, I don't have sufficient context to stamp it with great confidence.

Sep 12 2018, 3:32 PM
george.burgess.iv accepted D51960: [EarlyCSEwMemorySSA] Add MSSA verification and tests to make EarlyCSE failures easier to track..

LGTM, modulo one nit. Thanks!

Sep 12 2018, 10:35 AM

Sep 6 2018

george.burgess.iv added inline comments to D51717: [EarlyCSEwMemorySSA] Fix failure (w/ expensive checks). Need to resetOptimizeUses for replaced instructions..
Sep 6 2018, 4:55 PM
george.burgess.iv added a comment to D51717: [EarlyCSEwMemorySSA] Fix failure (w/ expensive checks). Need to resetOptimizeUses for replaced instructions..

Thanks for this!

Sep 6 2018, 4:55 PM

Sep 5 2018

george.burgess.iv accepted D45299: API to update MemorySSA for cloned blocks..

LGTM. Thanks for this!

Sep 5 2018, 5:02 PM
george.burgess.iv accepted D51563: [MemorySSA] Update MemoryPhi wiring for block splitting to consider if identical edges were merged..

Thanks!

Sep 5 2018, 9:25 AM
george.burgess.iv added a comment to D51664: [IR] Lazily number instructions for local dominance queries.

I have zero context on the various ways we can/do add instructions to a BB, so I can't immediately be helpful with the tricky part of this review. :) I've no complaints about the general direction of this patch, though.

Sep 5 2018, 8:48 AM

Aug 29 2018

george.burgess.iv accepted D51327: [MemorySSA] Add expesive check for validating clobber accesses..

LGTM after the requested nit. Thanks again!

Aug 29 2018, 10:59 AM
george.burgess.iv accepted D51423: Fix DebugCounterTest when running all SupportTests.

LGTM. Thanks again!

Aug 29 2018, 9:06 AM
george.burgess.iv added a comment to D51423: Fix DebugCounterTest when running all SupportTests.

Thanks for this!

Aug 29 2018, 8:20 AM

Aug 27 2018

george.burgess.iv added a comment to D51327: [MemorySSA] Add expesive check for validating clobber accesses..

Thanks for this!

Aug 27 2018, 3:31 PM

Aug 22 2018

george.burgess.iv added a comment to D45299: API to update MemorySSA for cloned blocks..

Lots of nits, and a few functional concerns. For the functional concerns that revolve around dubious edge additions, I'm happy with fixes or "it's impossible for that to actually happen; added a comment explaining that."

Aug 22 2018, 3:33 PM

Aug 20 2018

george.burgess.iv added inline comments to D51015: [MemorySSA] Make the Updater movable..
Aug 20 2018, 4:53 PM

Aug 14 2018

george.burgess.iv added a comment to D48790: Update MemorySSA in Local utils removing blocks..

Looks fine to me, though I'd prefer if someone familiar with this code had a look, as well.

Aug 14 2018, 7:40 PM
george.burgess.iv accepted D50749: [MemorySSA] Expose the verify as a debug option..

LGTM with a few comments. Thanks!

Aug 14 2018, 7:30 PM
george.burgess.iv added a comment to D45299: API to update MemorySSA for cloned blocks..

I'm not entirely through this patch (I hope to make more progress in the coming few days), but this is what I have for now. Logic's looking solid so far; I'm just nitpicky. :)

Aug 14 2018, 7:18 PM

Aug 13 2018

george.burgess.iv added a comment to D50650: [MemorySSA] Add API to update an incoming block into a MemoryPhi..

George, let me know if you agree with using the generic one and I'll abandon this revision. Thanks!

Aug 13 2018, 8:14 PM
george.burgess.iv added a comment to D50675: [IDF] Teach Iterated Dominance Frontier to use a snapshot CFG based on a GraphDiff..

Never touched this code, so all I can offer is a style nit. Thanks for the patch!

Aug 13 2018, 8:03 PM

Aug 9 2018

george.burgess.iv added a comment to D43269: [MemorySSA] Be less aggressive with @llvm.lifetime.start.

Workaround committed as r339411

Aug 9 2018, 10:16 PM

Aug 8 2018

george.burgess.iv added a comment to D43269: [MemorySSA] Be less aggressive with @llvm.lifetime.start.

(FWIW, I plan to commit a workaround like this tomorrow with said big scary comment, assuming no complaints appear before then. Like said, I'm more than happy to discuss the ideal path to take here even after that lands. But this is a known miscompile with a trivial workaround that should presumably also find its way into the release branch, so...)

Aug 8 2018, 6:15 PM

Aug 7 2018

george.burgess.iv added a comment to D43269: [MemorySSA] Be less aggressive with @llvm.lifetime.start.

Currently, MSSA treats lifetime intrinsics as clobbers, even though they're technically not. This patch would further solidify that behavior, which might not be a good thing (e.g. there's a good argument that we shouldn't be treating lifetime intrinsics as clobbers in the first place). I'm happy to rip out what few lifetime-related pieces MSSA has, but before doing so:

Aug 7 2018, 2:34 PM

Aug 6 2018

george.burgess.iv added a comment to D44748: Track whether the size of a MemoryLocation is precise.

ping :)

Aug 6 2018, 2:43 PM

Aug 4 2018

george.burgess.iv added a comment to D50295: Let GetUnderlyingObject/Objects use MemorySSA.

Thanks for this! In addition to Hal's comments...

Aug 4 2018, 10:07 AM

Aug 2 2018

george.burgess.iv added a comment to D50150: Require users to opt into DebugCounting.

Thanks!

Aug 2 2018, 12:51 PM

Aug 1 2018

george.burgess.iv added inline comments to D50094: Introduce DebugCounter into ConstProp pass.
Aug 1 2018, 1:31 PM
george.burgess.iv added inline comments to D50094: Introduce DebugCounter into ConstProp pass.
Aug 1 2018, 11:18 AM
george.burgess.iv created D50150: Require users to opt into DebugCounting.
Aug 1 2018, 10:58 AM

Jul 31 2018

george.burgess.iv added inline comments to D50094: Introduce DebugCounter into ConstProp pass.
Jul 31 2018, 1:36 PM
george.burgess.iv added a comment to D49103: Lower llvm.objectsize earlier in our optimization pipeline.

Ping :)

Jul 31 2018, 10:55 AM
george.burgess.iv added a comment to D50080: [Support] Use atomics in DebugCounter to silence TSAN.

Thanks for this!

Jul 31 2018, 10:35 AM

Jul 30 2018

george.burgess.iv added a comment to D50031: Add pass to print out DebugCounter info.

Thanks for this!

Jul 30 2018, 6:29 PM
george.burgess.iv edited reviewers for D50033: Introduce DebugCounter into DevRemPairs pass, added: spatel; removed: george.burgess.iv.

Looks straightforward enough to me, but spatel authored this pass (and I've never touched it), so I'm going to defer to them :)

Jul 30 2018, 6:09 PM

Jul 23 2018

george.burgess.iv added a comment to D49560: Add support to track total counts for DebugCounter.

Thanks again!

Jul 23 2018, 2:50 PM
george.burgess.iv added a comment to D49103: Lower llvm.objectsize earlier in our optimization pipeline.

Ping :)

Jul 23 2018, 1:53 PM
george.burgess.iv added a comment to D49560: Add support to track total counts for DebugCounter.

As a result, I also reverted the changes to make CounterDesc part of CounterInfo.

Jul 23 2018, 10:42 AM

Jul 20 2018

george.burgess.iv added a comment to D49560: Add support to track total counts for DebugCounter.

Running ninja check-llvm after applying this on top of r337631, I see many test failures in CodeGen/AMDGPU. Are you seeing these, as well?

Jul 20 2018, 4:56 PM
george.burgess.iv added a comment to D49560: Add support to track total counts for DebugCounter.

This LGTM; thanks again for the patch. Will land it for you once davide's happy

Jul 20 2018, 2:53 PM
george.burgess.iv accepted D49425: [MemorySSAUpdater] Update Phi operands after trivial Phi elimination.

LGTM after one nit is addressed. Thanks again!

Jul 20 2018, 2:05 PM
george.burgess.iv added inline comments to D49560: Add support to track total counts for DebugCounter.
Jul 20 2018, 2:04 PM

Jul 19 2018

george.burgess.iv added a reviewer for D49560: Add support to track total counts for DebugCounter: davide.

Thanks for this!

Jul 19 2018, 2:33 PM
george.burgess.iv added a comment to D49425: [MemorySSAUpdater] Update Phi operands after trivial Phi elimination.

As all the PhiOps are collected we call tryRemoveTrivialPhi which might have replaced the Phi with another MemoryAccess. Then we try to cache that MemoryAccess with the same BB key (line 110), which should return false

Jul 19 2018, 11:10 AM

Jul 17 2018

george.burgess.iv added a comment to D49425: [MemorySSAUpdater] Update Phi operands after trivial Phi elimination.

Thanks for this!

Jul 17 2018, 5:13 PM
george.burgess.iv added a comment to D49103: Lower llvm.objectsize earlier in our optimization pipeline.

I don't really like adding arbitrary flags to existing passes; better to have a separate pass, I think.

Jul 17 2018, 11:59 AM
george.burgess.iv updated the diff for D49103: Lower llvm.objectsize earlier in our optimization pipeline.

Switch to having a pass to do this, as requested.

Jul 17 2018, 11:58 AM

Jul 13 2018

george.burgess.iv accepted D48372: [MemorySSAUpdater] Remove deleted trivial Phis from active workset.

LGTM. Thanks again!

Jul 13 2018, 11:07 AM

Jul 12 2018

george.burgess.iv accepted D49156: [MemorySSA] Add API to update MemoryPhis, following CFG changes..

Thanks again!

Jul 12 2018, 4:02 PM
george.burgess.iv added inline comments to D49156: [MemorySSA] Add API to update MemoryPhis, following CFG changes..
Jul 12 2018, 3:30 PM
george.burgess.iv added a comment to D48372: [MemorySSAUpdater] Remove deleted trivial Phis from active workset.

It also adds and additional overhead

Jul 12 2018, 2:23 PM

Jul 10 2018

george.burgess.iv added inline comments to D49156: [MemorySSA] Add API to update MemoryPhis, following CFG changes..
Jul 10 2018, 4:47 PM
george.burgess.iv added a comment to D44748: Track whether the size of a MemoryLocation is precise.

Ping :)

Jul 10 2018, 1:01 PM
george.burgess.iv accepted D48897: [MemorySSA] Add APIs to move memory accesses between blocks, following CFG changes..

LGTM with a few comments; thanks again!

Jul 10 2018, 11:58 AM
george.burgess.iv added a comment to D48372: [MemorySSAUpdater] Remove deleted trivial Phis from active workset.

I'm still in favor of just replacing SmallVector<MemoryPhi *, 8> InsertedPHIs with SmallVector<WeakVH, 8> InsertedPHIs, since:

  • I think it meets all of our needs here,
  • it's a very straightforward change (no need to deal with removal, new containers, etc.), and
  • I don't see how we can ever get a duplicate in InsertedPHIs to start with.
Jul 10 2018, 11:47 AM

Jul 9 2018

george.burgess.iv added a comment to D48860: Make llvm.objectsize more conservative with null in non-zero address spaces.

(I'll go ahead and commit assuming my interpretation of arsenm's message was correct. Happy to revert or land follow-ups if not)

Jul 9 2018, 3:26 PM
george.burgess.iv created D49103: Lower llvm.objectsize earlier in our optimization pipeline.
Jul 9 2018, 3:04 PM
george.burgess.iv added a comment to D48897: [MemorySSA] Add APIs to move memory accesses between blocks, following CFG changes..

Thanks for the patch!

Jul 9 2018, 2:37 PM

Jul 7 2018

george.burgess.iv accepted D47895: llvm: Add support for "-fno-delete-null-pointer-checks".

Nit fixes LGTM. Please wait for an official stamp from Eli before committing, since I can’t tell if he got the second look he wanted. (I’m inclined to believe he did, but don’t want to jump the gun)

Jul 7 2018, 7:36 PM