Page MenuHomePhabricator

pgavlin (Pat Gavlin)
User

Projects

User does not belong to any projects.

User Details

User Since
Apr 27 2015, 10:07 AM (374 w, 12 h)

Recent Activity

Jan 22 2016

pgavlin added a comment to D16439: [PlaceSafepoints] Introduce a -spp-no-statepoints flag.

The mechanics look good to me as well. Thanks for the edifying discussion, too :)

Jan 22 2016, 1:07 PM

Jan 20 2016

pgavlin added a comment to D16342: Add a "gc-transition" operand bundle.

LGTM aside from a few nits. Some suggested docs are above.

Jan 20 2016, 10:34 AM

Jan 6 2016

pgavlin accepted D15632: [Statepoints] Relocating vectors of pointers.
Jan 6 2016, 6:35 PM
pgavlin added a comment to D15632: [Statepoints] Relocating vectors of pointers.

LGTM.

Jan 6 2016, 6:35 PM

Nov 24 2015

pgavlin added inline comments to D14665: Support literal structs in mangled intrinsics. .
Nov 24 2015, 2:58 PM

Nov 20 2015

pgavlin added a comment to D14665: Support literal structs in mangled intrinsics. .

@reames/@sanjoy, do you think you'll have a chance to take a look at this soon? This is hampering LLILC's ability to use statepoints with the SysV AMD64 ABI.

Nov 20 2015, 1:05 PM

Nov 18 2015

pgavlin added a comment to D14665: Support literal structs in mangled intrinsics. .

Ping: I think I've addressed all feedback.

Nov 18 2015, 10:26 AM

Nov 17 2015

pgavlin updated the diff for D14665: Support literal structs in mangled intrinsics. .

Addressed review feedback.

Nov 17 2015, 11:57 AM
pgavlin added a comment to D14665: Support literal structs in mangled intrinsics. .

Expand overloaded-intrinsic-name.ll test to include an overload using a literal struct type and execute the new codepath?

Nov 17 2015, 11:00 AM
pgavlin updated the diff for D14665: Support literal structs in mangled intrinsics. .
Nov 17 2015, 10:59 AM
pgavlin closed D14666: Lower statepoints with multi-def targets..

Committed in r253339.

Nov 17 2015, 8:07 AM
pgavlin committed rL253339: Lower statepoints with multi-def targets..
Lower statepoints with multi-def targets.
Nov 17 2015, 8:07 AM

Nov 13 2015

pgavlin retitled D14666: Lower statepoints with multi-def targets. from to Lower statepoints with multi-def targets..
Nov 13 2015, 1:40 PM
pgavlin retitled D14665: Support literal structs in mangled intrinsics. from to Support literal structs in mangled intrinsics. .
Nov 13 2015, 1:39 PM

Aug 14 2015

pgavlin committed rL245115: Add a target environment for CoreCLR..
Add a target environment for CoreCLR.
Aug 14 2015, 3:42 PM
pgavlin closed D11012: Add a target environment for CoreCLR. by committing rL245115: Add a target environment for CoreCLR..
Aug 14 2015, 3:42 PM
pgavlin updated the diff for D11012: Add a target environment for CoreCLR..

Rebased against latest changes.

Aug 14 2015, 3:36 PM

Jul 9 2015

pgavlin committed rL241827: Allow {e,r}bp as the target of {read,write}_register..
Allow {e,r}bp as the target of {read,write}_register.
Jul 9 2015, 10:41 AM
pgavlin closed D10977: Allow {e,r}bp as the target of {read,write}_register. by committing rL241827: Allow {e,r}bp as the target of {read,write}_register..
Jul 9 2015, 10:40 AM

Jul 7 2015

pgavlin retitled D11012: Add a target environment for CoreCLR. from to Add a target environment for CoreCLR..
Jul 7 2015, 2:45 PM
pgavlin updated the diff for D10977: Allow {e,r}bp as the target of {read,write}_register..

Moved frame register assertions into an NDEBUG guard.

Jul 7 2015, 12:55 PM
pgavlin updated the diff for D10977: Allow {e,r}bp as the target of {read,write}_register..

Respond to Hal's feedback re: error conditions.

Jul 7 2015, 11:10 AM

Jul 6 2015

pgavlin updated the diff for D10977: Allow {e,r}bp as the target of {read,write}_register..

Address CR feedback.

Jul 6 2015, 7:40 PM
pgavlin added a comment to D10977: Allow {e,r}bp as the target of {read,write}_register..
In D10977#200036, @rnk wrote:

Is it really that important to prevent people from hurting themselves by accessing ebp/rbp in a function without a frame pointer? I'd rather let them get hurt than suffer the extra code in LLVM.

Jul 6 2015, 5:16 PM
pgavlin retitled D10977: Allow {e,r}bp as the target of {read,write}_register. from to Allow {e,r}bp as the target of {read,write}_register..
Jul 6 2015, 3:56 PM

Jul 2 2015

pgavlin added a comment to D10758: [Inliner] Teach LLVM to inline through statepoints..

Looks good aside from the pending StatepointOffset refactoring and one small comment.

Jul 2 2015, 1:44 PM

Jun 29 2015

pgavlin closed D9299: Run StatepointLowering.{cpp,h} through clang-format..
Jun 29 2015, 10:15 AM

Jun 26 2015

pgavlin accepted D10756: [NFC] Make the Statepoint class more like CallSite.

LGTM.

Jun 26 2015, 11:11 AM
pgavlin added a comment to D10758: [Inliner] Teach LLVM to inline through statepoints..

This is looking quite a bit better. I'm going to hold off on accepting this patch until Chandler has had a chance to follow up on http://reviews.llvm.org/D10631, however.

Jun 26 2015, 11:08 AM

Jun 24 2015

pgavlin added a comment to D10377: Add parser for the stackmap section format.

I prefer Lang's approach as well--I certainly wouldn't mind having this checked in as an intermediate step, however.

Jun 24 2015, 3:27 PM
pgavlin added inline comments to D10633: [Inlining] Teach the inliner pass to inline through statepoints..
Jun 24 2015, 3:20 PM
pgavlin accepted D10680: Enable StackMap Serialization for COFF.

LGTM as well.

Jun 24 2015, 3:06 PM

Jun 18 2015

pgavlin accepted D10461: [RewriteStatepointsForGC] Generalized vector phi/select handling for base pointers.

LGTM modulo a few very minor comments.

Jun 18 2015, 10:27 AM
pgavlin accepted D10526: [CallGraph] Teach the CallGraph about non-leaf intrinsics..

This looks good to me as well. I'm also a bit leery of returning the call target as an argument index. Perhaps it will eventually make more sense to add an overload that operates on the call itself and returns the argument itself, but as Chandler mentioned, it's hard to tell without an example use.

Jun 18 2015, 10:13 AM

Jun 12 2015

pgavlin added a comment to D10377: Add parser for the stackmap section format.

+1 to Swaroop's suggestion to use an ostream rather than printf.

Jun 12 2015, 9:59 AM

Jun 11 2015

pgavlin accepted D10197: [CodeGen] Introduce a FAULTING_LOAD_OP pseudo-op..

I don't know enough about COFF to confidently do that, and I don't have a stackmap example to copy from -- do you mind adding that bit of code once this change has landed?

Jun 11 2015, 10:44 AM

Jun 9 2015

pgavlin added inline comments to D10197: [CodeGen] Introduce a FAULTING_LOAD_OP pseudo-op..
Jun 9 2015, 9:55 AM

May 28 2015

pgavlin added a comment to D10104: [NFCI] Change RewriteStatepointsForGC to a ModulePass..

Okay, I see the dependency in there now. Missed it when I glanced at that review earlier.

May 28 2015, 3:39 PM
pgavlin accepted D10104: [NFCI] Change RewriteStatepointsForGC to a ModulePass..

LGTM as well. Which later change depends on this?

May 28 2015, 3:37 PM

May 18 2015

pgavlin added inline comments to D9767: Add a GCStrategy for CoreCLR .
May 18 2015, 4:50 PM
pgavlin added a comment to D9767: Add a GCStrategy for CoreCLR .

Code changes LGTM. @reames, is there a list of available GC strategies that we should update?

May 18 2015, 10:59 AM
pgavlin added inline comments to D9761: [PlaceSafepoints] Use the analysis infrastructure rather than an inner pass manager.
May 18 2015, 10:24 AM

May 15 2015

pgavlin accepted D9695: [Verifier] Assert gc_relocate always return a pointer type .

LGTM. Thanks for doing this.

May 15 2015, 2:59 PM

May 14 2015

pgavlin added a comment to D9695: [Verifier] Assert gc_relocate always return a pointer type .

That sounds fine to me. Thanks for digging in a bit more!

May 14 2015, 11:16 AM

May 13 2015

pgavlin accepted D9735: [PlaceSafepoints] New attributes for patchable statepoints..

LGTM. Thanks for the updates.

May 13 2015, 12:53 PM
pgavlin requested changes to D9695: [Verifier] Assert gc_relocate always return a pointer type .

Please add a test case for the first assert as well (i.e. a case where relocate is returning a non-pointer type).

May 13 2015, 11:00 AM
pgavlin requested changes to D9735: [PlaceSafepoints] New attributes for patchable statepoints..

It would be nice to have documentation for these attributes somewhere.

May 13 2015, 10:46 AM

May 12 2015

pgavlin committed rL237191: [Statepoints] Clean up statepoint argument accessors..
[Statepoints] Clean up statepoint argument accessors.
May 12 2015, 2:37 PM
pgavlin closed D9622: Clean up statepoint argument accessors. by committing rL237191: [Statepoints] Clean up statepoint argument accessors..
May 12 2015, 2:37 PM
pgavlin committed rL237166: [Statepoints] Split the calling convention and statepoint flags operand to….
[Statepoints] Split the calling convention and statepoint flags operand to…
May 12 2015, 12:54 PM
pgavlin closed D9623: Split the calling convention and statepoint flags operand to STATEPOINT into two separate operands. by committing rL237166: [Statepoints] Split the calling convention and statepoint flags operand to….
May 12 2015, 12:54 PM
pgavlin added inline comments to D9676: [RewriteStatepointsForGC] Extend base pointer to handle more cases w/vectors.
May 12 2015, 10:18 AM
pgavlin updated the diff for D9623: Split the calling convention and statepoint flags operand to STATEPOINT into two separate operands..
May 12 2015, 9:13 AM
pgavlin added a comment to D9623: Split the calling convention and statepoint flags operand to STATEPOINT into two separate operands..

I did, yes.Nearly all of the code that would be sensitive to operand indices is in StackMap.{c,h}, and it is very good about using accessors and/or symbolic offsets.

May 12 2015, 7:59 AM
pgavlin added a comment to D9695: [Verifier] Assert gc_relocate always return a pointer type .

+1 to Sanjoy's comments. Thanks for doing this.

May 12 2015, 7:49 AM

May 11 2015

pgavlin accepted D9546: [Statepoints] Support for "patchable" statepoints..

LGTM modulo the comments above.

May 11 2015, 5:29 PM
pgavlin added inline comments to D9546: [Statepoints] Support for "patchable" statepoints..
May 11 2015, 4:40 PM
pgavlin accepted D9676: [RewriteStatepointsForGC] Extend base pointer to handle more cases w/vectors.
May 11 2015, 4:22 PM
pgavlin added a comment to D9676: [RewriteStatepointsForGC] Extend base pointer to handle more cases w/vectors.

LGTM with one nit.

May 11 2015, 4:21 PM

May 8 2015

pgavlin retitled D9623: Split the calling convention and statepoint flags operand to STATEPOINT into two separate operands. from to Split the calling convention and statepoint flags operand to STATEPOINT into two separate operands..
May 8 2015, 2:33 PM
pgavlin added a comment to D9622: Clean up statepoint argument accessors..

Sanjoy, this might conflict with D9546. I'm happy to wait until that patch has landed before checking this in.

May 8 2015, 1:47 PM
pgavlin added reviewers for D9622: Clean up statepoint argument accessors.: reames, sanjoy.
May 8 2015, 1:46 PM
pgavlin retitled D9622: Clean up statepoint argument accessors. from to Clean up statepoint argument accessors..
May 8 2015, 1:45 PM
pgavlin committed rL236891: Fix a docs build break introduced by rL236888..
Fix a docs build break introduced by rL236888.
May 8 2015, 11:41 AM
pgavlin committed rL236888: Extend the statepoint intrinsic to allow statepoints to be marked as….
Extend the statepoint intrinsic to allow statepoints to be marked as…
May 8 2015, 11:12 AM
pgavlin closed D9501: Extend the statepoint intrinsic to allow statepoints to be marked as transitions from GC-aware code to code that is not GC-aware. by committing rL236888: Extend the statepoint intrinsic to allow statepoints to be marked as….
May 8 2015, 11:11 AM

May 6 2015

pgavlin added a comment to D9501: Extend the statepoint intrinsic to allow statepoints to be marked as transitions from GC-aware code to code that is not GC-aware..

I do have commit access now, so I can commit this once you've signed off.

May 6 2015, 8:41 PM
pgavlin updated the diff for D9501: Extend the statepoint intrinsic to allow statepoints to be marked as transitions from GC-aware code to code that is not GC-aware..

Added a test for GC transition lowering.

May 6 2015, 8:40 PM
pgavlin added inline comments to D9501: Extend the statepoint intrinsic to allow statepoints to be marked as transitions from GC-aware code to code that is not GC-aware..
May 6 2015, 6:05 PM
pgavlin updated the diff for D9501: Extend the statepoint intrinsic to allow statepoints to be marked as transitions from GC-aware code to code that is not GC-aware..

Expand documentation and address further CR feedback.

May 6 2015, 6:02 PM
pgavlin updated the diff for D9501: Extend the statepoint intrinsic to allow statepoints to be marked as transitions from GC-aware code to code that is not GC-aware..

Expand the documentation for GC transitions and address CR feedback.

May 6 2015, 2:48 PM
pgavlin added a comment to D9501: Extend the statepoint intrinsic to allow statepoints to be marked as transitions from GC-aware code to code that is not GC-aware..

Thanks for the feedback! I'll post a new version of the code with the requested changes later today.

May 6 2015, 11:24 AM
pgavlin added inline comments to D9480: [StatepointLowering] Don't create temporary instructions. NFCI..
May 6 2015, 10:12 AM

May 5 2015

pgavlin retitled D9501: Extend the statepoint intrinsic to allow statepoints to be marked as transitions from GC-aware code to code that is not GC-aware. from to Extend the statepoint intrinsic to allow statepoints to be marked as transitions from GC-aware code to code that is not GC-aware..
May 5 2015, 9:20 AM

Apr 29 2015

pgavlin committed rL236166: Run StatepointLowering.{cpp,h} through clang-format..
Run StatepointLowering.{cpp,h} through clang-format.
Apr 29 2015, 2:56 PM

Apr 27 2015

pgavlin retitled D9299: Run StatepointLowering.{cpp,h} through clang-format. from to Run StatepointLowering.{cpp,h} through clang-format..
Apr 27 2015, 10:16 AM