Page MenuHomePhabricator

gilr (Gil Rapaport)
User

Projects

User does not belong to any projects.

User Details

User Since
Nov 12 2014, 11:00 PM (257 w, 3 d)

Recent Activity

Wed, Oct 16

gilr created D69067: [LV] Reduce ingredient DU usage by ILV (NFC).
Wed, Oct 16, 12:59 PM · Restricted Project

Mon, Oct 14

gilr added a comment to D68577: [LV] Apply sink-after & interleave-groups as VPlan transformations (NFC).

ping

Mon, Oct 14, 12:28 PM · Restricted Project

Sun, Oct 13

gilr updated the diff for D68577: [LV] Apply sink-after & interleave-groups as VPlan transformations (NFC).

Applied review comments.
Reused VPWidenMemoryInstructionRecipe's new getMask() in its execute().

Sun, Oct 13, 11:23 PM · Restricted Project
gilr updated the diff for D68577: [LV] Apply sink-after & interleave-groups as VPlan transformations (NFC).

Applied review comments.

Sun, Oct 13, 1:12 PM · Restricted Project
gilr updated the diff for D68577: [LV] Apply sink-after & interleave-groups as VPlan transformations (NFC).

Applied review comments.

Sun, Oct 13, 4:44 AM · Restricted Project
gilr updated the diff for D68577: [LV] Apply sink-after & interleave-groups as VPlan transformations (NFC).

Applied review comments.

Sun, Oct 13, 3:51 AM · Restricted Project
gilr added inline comments to D68577: [LV] Apply sink-after & interleave-groups as VPlan transformations (NFC).
Sun, Oct 13, 3:50 AM · Restricted Project

Fri, Oct 11

gilr updated the diff for D68577: [LV] Apply sink-after & interleave-groups as VPlan transformations (NFC).

Applied review comments.

Fri, Oct 11, 5:26 AM · Restricted Project

Wed, Oct 9

gilr added a comment to D68577: [LV] Apply sink-after & interleave-groups as VPlan transformations (NFC).

So, IIUC, this is changing tryCreateRecipe to move the interleave recipe creation to the caller, buildVPlanWithVPRecipes. The dependencies with the sink values is recorded initially, then the plans are created, then the sinks are applied and, if any, the interleave groups.

Wed, Oct 9, 11:08 AM · Restricted Project
gilr updated the diff for D68577: [LV] Apply sink-after & interleave-groups as VPlan transformations (NFC).
  • Applied review comment
  • Simplified predicate lambda function
Wed, Oct 9, 5:59 AM · Restricted Project
gilr added inline comments to D68577: [LV] Apply sink-after & interleave-groups as VPlan transformations (NFC).
Wed, Oct 9, 5:49 AM · Restricted Project

Mon, Oct 7

gilr created D68577: [LV] Apply sink-after & interleave-groups as VPlan transformations (NFC).
Mon, Oct 7, 7:39 AM · Restricted Project

Sep 15 2019

gilr committed rL371956: Request commit access for gilr.
Request commit access for gilr
Sep 15 2019, 11:00 PM

Mar 14 2019

gilr added a comment to D58389: [NFC][LSR] Cleanup Cost API.

No problem :)
LGTM

Mar 14 2019, 3:45 AM · Restricted Project
gilr accepted D58389: [NFC][LSR] Cleanup Cost API.
Mar 14 2019, 3:45 AM · Restricted Project

Feb 21 2019

gilr added inline comments to D58389: [NFC][LSR] Cleanup Cost API.
Feb 21 2019, 4:31 AM · Restricted Project
gilr added inline comments to D58389: [NFC][LSR] Cleanup Cost API.
Feb 21 2019, 1:09 AM · Restricted Project

Feb 7 2019

gilr accepted D55373: [LSR] Generate formulae to enable more indexed accesses.

LGTM!
(One last nitpick: I'd consider letting EnableBackedgeIndexing default to true as TTI's default already disables it for Hexagon and all other targets)

Feb 7 2019, 4:28 AM · Restricted Project
gilr added inline comments to D55373: [LSR] Generate formulae to enable more indexed accesses.
Feb 7 2019, 2:27 AM · Restricted Project

Feb 6 2019

gilr added inline comments to D55373: [LSR] Generate formulae to enable more indexed accesses.
Feb 6 2019, 12:40 PM · Restricted Project

Feb 3 2019

gilr added inline comments to D55373: [LSR] Generate formulae to enable more indexed accesses.
Feb 3 2019, 2:29 PM · Restricted Project

Jan 28 2019

gilr added inline comments to D55373: [LSR] Generate formulae to enable more indexed accesses.
Jan 28 2019, 5:52 AM · Restricted Project

Dec 16 2018

gilr added inline comments to D55373: [LSR] Generate formulae to enable more indexed accesses.
Dec 16 2018, 5:57 AM · Restricted Project

Dec 12 2018

gilr added inline comments to D55373: [LSR] Generate formulae to enable more indexed accesses.
Dec 12 2018, 12:13 AM · Restricted Project

Dec 9 2018

gilr added inline comments to D55373: [LSR] Generate formulae to enable more indexed accesses.
Dec 9 2018, 8:49 AM · Restricted Project

Nov 28 2018

gilr updated subscribers of D54899: [LoopStrengthReduce] ComplexityLimit as an option.

Thanks for taking a look and for the value renamer tip! The test case has now been renamed and reduced.

Nov 28 2018, 2:21 PM

Nov 27 2018

gilr added inline comments to D54899: [LoopStrengthReduce] ComplexityLimit as an option.
Nov 27 2018, 8:26 AM
gilr added inline comments to D54899: [LoopStrengthReduce] ComplexityLimit as an option.
Nov 27 2018, 5:37 AM

Nov 8 2018

gilr committed rL346390: [LSR] Combine unfolded offset into invariant register.
[LSR] Combine unfolded offset into invariant register
Nov 8 2018, 1:05 AM
gilr closed D51861: [LSR] Combine unfolded offset into invariant register.
Nov 8 2018, 1:05 AM

Nov 7 2018

gilr updated the diff for D51861: [LSR] Combine unfolded offset into invariant register.

Test case for the two-combinations bug greatly reduced by bugpoint (thanks, Quentin!) and checks the generated IR instead of LSR's debug prints.

Nov 7 2018, 8:28 AM

Nov 5 2018

gilr updated the diff for D51861: [LSR] Combine unfolded offset into invariant register.

Added a test case to cover the bug introduced by the first patch and fixed in the second.

Nov 5 2018, 6:08 AM

Nov 2 2018

gilr added a comment to D51861: [LSR] Combine unfolded offset into invariant register.

I am confused. If they don't do (c) how can they fail?

Nov 2 2018, 6:58 AM

Nov 1 2018

gilr added a comment to D51861: [LSR] Combine unfolded offset into invariant register.

Is this covered by an existing test case?
If not, could you add one?

Nov 1 2018, 9:03 AM

Oct 28 2018

gilr requested review of D51861: [LSR] Combine unfolded offset into invariant register.
Oct 28 2018, 1:01 PM
gilr added a reviewer for D51861: [LSR] Combine unfolded offset into invariant register: delena.
Oct 28 2018, 6:04 AM
gilr updated the diff for D51861: [LSR] Combine unfolded offset into invariant register.

Bug fix: ScalarEvolution::getAddExpr() may modify the vector of SCEVs it is given as argument, causing 2nd formula to be created with only part of the SCEVs in Ops.
This patch fixes this by using a temporary copy of Ops for the 1st formula.

Oct 28 2018, 6:03 AM
gilr reopened D51861: [LSR] Combine unfolded offset into invariant register.
Oct 28 2018, 5:55 AM

Oct 24 2018

gilr committed rL345123: Revert r345114.
Revert r345114
Oct 24 2018, 1:43 AM
gilr committed rL345114: [LSR] Combine unfolded offset into invariant register.
[LSR] Combine unfolded offset into invariant register
Oct 24 2018, 12:10 AM
gilr closed D51861: [LSR] Combine unfolded offset into invariant register.
Oct 24 2018, 12:10 AM

Oct 23 2018

gilr added a comment to D51861: [LSR] Combine unfolded offset into invariant register.

Thanks Quentin!
Regarding effect on compile time please see attached CTMark results on x86.

Oct 23 2018, 7:45 AM

Sep 23 2018

gilr added a comment to D51861: [LSR] Combine unfolded offset into invariant register.

Ping

Sep 23 2018, 12:30 AM

Sep 10 2018

gilr created D51861: [LSR] Combine unfolded offset into invariant register.
Sep 10 2018, 8:57 AM
gilr committed rL341835: [LSR] Add tests for small constants; NFC.
[LSR] Add tests for small constants; NFC
Sep 10 2018, 7:57 AM

Jun 25 2018

gilr committed rL335579: [InstCombine] (A + 1) + (B ^ -1) --> A - B.
[InstCombine] (A + 1) + (B ^ -1) --> A - B
Jun 25 2018, 10:36 PM
gilr closed D48535: [InstCombine] (A + 1) + (B ^ -1) --> A - B.
Jun 25 2018, 10:36 PM
gilr updated the diff for D48535: [InstCombine] (A + 1) + (B ^ -1) --> A - B.

Impl. Sanjay comments.
Thanks for taking the time guys!

Jun 25 2018, 7:10 AM
gilr updated the diff for D48535: [InstCombine] (A + 1) + (B ^ -1) --> A - B.

Impl. Roman's comment - m_Add instead of m_c_Add.

Jun 25 2018, 4:02 AM
gilr created D48535: [InstCombine] (A + 1) + (B ^ -1) --> A - B.
Jun 25 2018, 1:41 AM

Nov 20 2017

gilr committed rL318645: [LV] Model masking in VPlan, introducing VPInstructions.
[LV] Model masking in VPlan, introducing VPInstructions
Nov 20 2017, 4:02 AM
gilr closed D38676: [LV] Model masking in VPlan, introducing VPInstructions by committing rL318645: [LV] Model masking in VPlan, introducing VPInstructions.
Nov 20 2017, 4:01 AM

Nov 19 2017

gilr updated the diff for D38676: [LV] Model masking in VPlan, introducing VPInstructions.

Addressed review comments.

Nov 19 2017, 8:18 AM
gilr added a comment to D38676: [LV] Model masking in VPlan, introducing VPInstructions.

I finished my review, and apart from my two final comments, everything looks fine.

Thanks for the hard work! :)

Nov 19 2017, 8:15 AM

Nov 14 2017

gilr committed rL318149: [LV] Introduce VPBlendRecipe, VPWidenMemoryInstructionRecipe.
[LV] Introduce VPBlendRecipe, VPWidenMemoryInstructionRecipe
Nov 14 2017, 4:10 AM
gilr closed D39068: [LV] Introduce VPBlendRecipe, VPWidenMemoryInstructionRecipe by committing rL318149: [LV] Introduce VPBlendRecipe, VPWidenMemoryInstructionRecipe.
Nov 14 2017, 4:10 AM

Nov 7 2017

gilr updated the diff for D38676: [LV] Model masking in VPlan, introducing VPInstructions.

Rebasing this patch on its preparatory patch D39068.

Nov 7 2017, 7:44 AM
gilr updated the diff for D39068: [LV] Introduce VPBlendRecipe, VPWidenMemoryInstructionRecipe.
  • Added LIT test case to make sure the new tryToWidenMemory() method takes scalarization decisions into account.
  • NFC: move createBlockInMask(), createEdgeMask() to their would-be location in D38676 as part of the Planner's methods to facilitate their review.
Nov 7 2017, 7:26 AM
gilr added a comment to D39068: [LV] Introduce VPBlendRecipe, VPWidenMemoryInstructionRecipe.

Excellent, ok, this patch is looking good to me now. How does this fit into the whole story?

The other review (D38676) still has this patch inside, plus tests. Are you going to clean that up and move the tests here?

It'd be good to have only the refactoring part there, and only the new plans here, with tests.

Nov 7 2017, 6:55 AM

Nov 3 2017

gilr updated the diff for D39068: [LV] Introduce VPBlendRecipe, VPWidenMemoryInstructionRecipe.

Addressed review comment - remove the load/store specialized llvm_unreachable from ILV::widenInstruction, load/store now also handled by the switch's default case.

Nov 3 2017, 10:03 AM

Oct 30 2017

gilr added inline comments to D39068: [LV] Introduce VPBlendRecipe, VPWidenMemoryInstructionRecipe.
Oct 30 2017, 8:33 AM

Oct 28 2017

gilr added a comment to D39068: [LV] Introduce VPBlendRecipe, VPWidenMemoryInstructionRecipe.

Ping

Oct 28 2017, 7:53 AM

Oct 18 2017

gilr added a comment to D38676: [LV] Model masking in VPlan, introducing VPInstructions.
In D38676#900051, @gilr wrote:

One change we can probably peel off before, though, is to first introduce VPBlendRecipe and VPWidenMemoryInstruction w/o the new masking code. This should later narrow the diff in LoopVectorize.cpp and reflect the masking changes on those Recipes as well.

Oct 18 2017, 1:06 PM
gilr created D39068: [LV] Introduce VPBlendRecipe, VPWidenMemoryInstructionRecipe.
Oct 18 2017, 1:01 PM

Oct 17 2017

gilr added a comment to D38676: [LV] Model masking in VPlan, introducing VPInstructions.

A minor nit but LGTM. I don't have an aversion to "dead code" if it's going to be used in the near future, so perhaps to make the patch smaller split it into an initial patch to add VPlanValue.h and then the VPlanBuilder.h as well as the changes to the documentation. Those pieces seem uncontroversial to me, before moving onto the vectorizer. @mkuper does this make sense?

My $0.02:
Basically, I agree with @aemerson and @rengolin, at least to the extent that even if either way could work, it would be best to do it the way that ends up with simpler patches and will take less time (overall) to review well.

I don't think there's a problem of dead code, or "it's unclear how the infrastructure will be used", since you already have the code that actually uses it ready for review as well. So, it would probably best to split it into two dependent patches, and post them for review separately. Then, the second one can be rebased on top of the first one, if it has any significant changes, when it goes in. Admittedly, it's more work, but I think it's worth it.

WDYT?

Oct 17 2017, 1:06 PM

Oct 15 2017

gilr updated the diff for D38676: [LV] Model masking in VPlan, introducing VPInstructions.

Addressed comment: Added a call to Value2VPValue.clear() on ~VPlan().

Oct 15 2017, 6:40 AM

Oct 14 2017

gilr updated the summary of D38676: [LV] Model masking in VPlan, introducing VPInstructions.
Oct 14 2017, 12:26 AM

Oct 13 2017

gilr added a comment to D38676: [LV] Model masking in VPlan, introducing VPInstructions.

Would it be possible to split the patch, so that we can review them in a more concise way?

Oct 13 2017, 2:34 PM

Oct 10 2017

gilr updated the diff for D38676: [LV] Model masking in VPlan, introducing VPInstructions.
  • Adressed comments.
  • Fixed bug: tryToWidenMemory() wasn't considering Cost Model's sink-scalar-operands scalarization decisions.
Oct 10 2017, 7:44 AM
gilr added a comment to D38676: [LV] Model masking in VPlan, introducing VPInstructions.

Thanks, Amara!

Oct 10 2017, 6:31 AM

Oct 8 2017

gilr created D38676: [LV] Model masking in VPlan, introducing VPInstructions.
Oct 8 2017, 12:28 PM

Jul 16 2017

gilr added inline comments to D35227: [LV] Don't allow outside uses of IVs if the SCEV is predicated on loop conditions.
Jul 16 2017, 4:19 AM

Apr 25 2017

gilr committed rL301345: [LV] Make LIT test insensitive to basic block numbering.
[LV] Make LIT test insensitive to basic block numbering
Apr 25 2017, 11:27 AM
gilr closed D32404: [LV] Make LIT test insensitive to basic block numbering by committing rL301345: [LV] Make LIT test insensitive to basic block numbering.
Apr 25 2017, 11:27 AM
gilr added a comment to D32404: [LV] Make LIT test insensitive to basic block numbering.

Hi Matt,

Apr 25 2017, 7:16 AM

Apr 24 2017

gilr committed rL301293: [LV] Remove redundant basic block split.
[LV] Remove redundant basic block split
Apr 24 2017, 11:10 PM
gilr closed D32224: [LV] Remove redundant basic block split by committing rL301293: [LV] Remove redundant basic block split.
Apr 24 2017, 11:10 PM

Apr 23 2017

gilr created D32404: [LV] Make LIT test insensitive to basic block numbering.
Apr 23 2017, 1:04 PM

Apr 19 2017

gilr updated the summary of D32224: [LV] Remove redundant basic block split.
Apr 19 2017, 7:33 AM
gilr updated the summary of D32224: [LV] Remove redundant basic block split.
Apr 19 2017, 7:32 AM
gilr created D32224: [LV] Remove redundant basic block split.
Apr 19 2017, 7:31 AM

Apr 18 2017

gilr committed rL300557: [LV] Cache block mask values.
[LV] Cache block mask values
Apr 18 2017, 8:00 AM
gilr closed D32054: [LV] Cache block mask values by committing rL300557: [LV] Cache block mask values.
Apr 18 2017, 7:59 AM

Apr 14 2017

gilr committed rL300310: [LV] Remove implicit single basic block assumption.
[LV] Remove implicit single basic block assumption
Apr 14 2017, 12:43 AM
gilr closed D32040: [LV] Remove implicit single basic block assumption by committing rL300310: [LV] Remove implicit single basic block assumption.
Apr 14 2017, 12:43 AM
gilr updated the diff for D32040: [LV] Remove implicit single basic block assumption.

Applied Michael's comment and removed the comment explaining how using the body as header/latch is safe at this point.

Apr 14 2017, 12:15 AM
gilr updated the summary of D32040: [LV] Remove implicit single basic block assumption.
Apr 14 2017, 12:05 AM

Apr 13 2017

gilr updated the summary of D32054: [LV] Cache block mask values.
Apr 13 2017, 2:54 PM
gilr created D32054: [LV] Cache block mask values.
Apr 13 2017, 2:34 PM
gilr updated subscribers of D32040: [LV] Remove implicit single basic block assumption.
Apr 13 2017, 12:57 PM
gilr created D32040: [LV] Remove implicit single basic block assumption.
Apr 13 2017, 12:55 PM

Mar 23 2017

gilr committed rL298641: [LV] Add regression test for r297610.
[LV] Add regression test for r297610
Mar 23 2017, 1:14 PM
gilr closed D30972: [LV] Add regression test for r297610 by committing rL298641: [LV] Add regression test for r297610.
Mar 23 2017, 1:14 PM

Mar 15 2017

gilr created D30972: [LV] Add regression test for r297610.
Mar 15 2017, 2:48 AM

Mar 14 2017

gilr committed rL297740: [LV] Refactor cross-iteration phi's back-patching; NFC.
[LV] Refactor cross-iteration phi's back-patching; NFC
Mar 14 2017, 7:02 AM
gilr closed D30755: [LV] Refactor cross-iteration phi's back-patching; NFC by committing rL297740: [LV] Refactor cross-iteration phi's back-patching; NFC.
Mar 14 2017, 7:02 AM

Mar 13 2017

gilr updated the diff for D30755: [LV] Refactor cross-iteration phi's back-patching; NFC.

Merged with rL297302 & ironed the affected phi-handling comments.

Mar 13 2017, 7:30 AM
gilr added a comment to D30755: [LV] Refactor cross-iteration phi's back-patching; NFC.

I did have one minor suggestion to avoid some duplicated text in the comments.

Mar 13 2017, 7:28 AM
gilr committed rL297610: [LV] Set memcheck metadata also for VF==1.
[LV] Set memcheck metadata also for VF==1
Mar 13 2017, 3:36 AM

Mar 12 2017

gilr committed rL297580: [LV] A unified scalarizeInstruction() for Vectorizer and Unroller; NFC.
[LV] A unified scalarizeInstruction() for Vectorizer and Unroller; NFC
Mar 12 2017, 5:43 AM