Page MenuHomePhabricator

piotr (Piotr Sobczak)
User

Projects

User does not belong to any projects.

User Details

User Since
Dec 4 2018, 6:02 AM (116 w, 1 d)

Recent Activity

Today

piotr added inline comments to D96258: [AMDGPU] Introduce Strict WQM mode.
Wed, Feb 24, 5:48 AM · Restricted Project
piotr updated the summary of D96258: [AMDGPU] Introduce Strict WQM mode.
Wed, Feb 24, 5:48 AM · Restricted Project
piotr updated the diff for D96258: [AMDGPU] Introduce Strict WQM mode.

Rebased, changed strictwqm to strict_wqm and updated tests.

Wed, Feb 24, 5:44 AM · Restricted Project

Yesterday

piotr updated the diff for D96257: [AMDGPU] Rename amdgcn_wwm to amdgcn_strict_wwm.

Added missing dot in the intrinsic name.

Tue, Feb 23, 1:37 AM · Restricted Project

Mon, Feb 22

piotr updated the diff for D96257: [AMDGPU] Rename amdgcn_wwm to amdgcn_strict_wwm.

Rebased and updated tests.

Mon, Feb 22, 6:10 AM · Restricted Project

Fri, Feb 19

piotr accepted D95503: [AMDGPU] WQM/WWM: Fix marking of partial definitions.
Fri, Feb 19, 2:40 AM · Restricted Project

Wed, Feb 17

piotr added a comment to D95503: [AMDGPU] WQM/WWM: Fix marking of partial definitions.

Looks go to me (with two nits inline).

Wed, Feb 17, 9:39 AM · Restricted Project
piotr committed rGc72a63b4b06a: [AMDGPU] Add implicit vcc_lo on S_CBRANCH_VCCNZ in wave32 (authored by piotr).
[AMDGPU] Add implicit vcc_lo on S_CBRANCH_VCCNZ in wave32
Wed, Feb 17, 6:17 AM
piotr closed D96793: [AMDGPU] Add implicit vcc_lo on S_CBRANCH_VCCNZ in wave32.
Wed, Feb 17, 6:17 AM · Restricted Project
piotr committed rG08131c743933: [AMDGPU] Fix a miscompile with S_ADD/S_SUB (authored by piotr).
[AMDGPU] Fix a miscompile with S_ADD/S_SUB
Wed, Feb 17, 3:27 AM
piotr closed D86878: [AMDGPU] Fix a miscompile with S_ADD/S_SUB.
Wed, Feb 17, 3:26 AM · Restricted Project
piotr updated the diff for D96793: [AMDGPU] Add implicit vcc_lo on S_CBRANCH_VCCNZ in wave32.

Commoned up the SI and GFX10-WAVE64 checks.

Wed, Feb 17, 2:19 AM · Restricted Project

Tue, Feb 16

piotr added a comment to D96793: [AMDGPU] Add implicit vcc_lo on S_CBRANCH_VCCNZ in wave32.

Updated commit message with more details and pasted the original error message.

Tue, Feb 16, 9:14 AM · Restricted Project
piotr retitled D96793: [AMDGPU] Add implicit vcc_lo on S_CBRANCH_VCCNZ in wave32 from [AMDGPU] Fix implicit operand to [AMDGPU] Add implicit vcc_lo on S_CBRANCH_VCCNZ in wave32.
Tue, Feb 16, 9:12 AM · Restricted Project
piotr added reviewers for D96793: [AMDGPU] Add implicit vcc_lo on S_CBRANCH_VCCNZ in wave32: arsenm, foad, critson.
Tue, Feb 16, 8:58 AM · Restricted Project
piotr requested review of D96793: [AMDGPU] Add implicit vcc_lo on S_CBRANCH_VCCNZ in wave32.
Tue, Feb 16, 8:57 AM · Restricted Project
piotr updated the diff for D86878: [AMDGPU] Fix a miscompile with S_ADD/S_SUB.

Tests rewritten.

Tue, Feb 16, 1:10 AM · Restricted Project

Mon, Feb 15

piotr retitled D96257: [AMDGPU] Rename amdgcn_wwm to amdgcn_strict_wwm from [AMDGPU] Rename amdgcn_wwm to amdgcn_strictwwm to [AMDGPU] Rename amdgcn_wwm to amdgcn_strict_wwm.
Mon, Feb 15, 8:13 AM · Restricted Project
piotr updated the diff for D96257: [AMDGPU] Rename amdgcn_wwm to amdgcn_strict_wwm.

Add the underscore between strict and wwm. Restored the old tests for amdgcn_wwm.

Mon, Feb 15, 8:09 AM · Restricted Project
piotr abandoned D96714: [AMDGPU] Rename amdgcn_wwm to amdgcn_strict_wwm.
Mon, Feb 15, 7:59 AM · Restricted Project
piotr requested review of D96714: [AMDGPU] Rename amdgcn_wwm to amdgcn_strict_wwm.
Mon, Feb 15, 7:58 AM · Restricted Project

Thu, Feb 11

piotr added a comment to D86878: [AMDGPU] Fix a miscompile with S_ADD/S_SUB.

This is the intended patch, no need to change AMDGPUISelDAGToDAG.cpp in the most recent version. Here, due to the updated check in isBoolSGPR the problematic combines that would strip zext will not happen (see SITargetLowering::performAddCombine/performSubCombine). This is similar to the very first approach I put up for a review.

Thu, Feb 11, 6:44 AM · Restricted Project
piotr updated the summary of D86878: [AMDGPU] Fix a miscompile with S_ADD/S_SUB.
Thu, Feb 11, 1:12 AM · Restricted Project
piotr updated the diff for D86878: [AMDGPU] Fix a miscompile with S_ADD/S_SUB.

Following the suggesstions of updating isBoolSGPR.

Thu, Feb 11, 1:11 AM · Restricted Project

Wed, Feb 10

piotr added inline comments to D96257: [AMDGPU] Rename amdgcn_wwm to amdgcn_strict_wwm.
Wed, Feb 10, 12:56 AM · Restricted Project

Tue, Feb 9

piotr added inline comments to D96336: [AMDGPU] Save VGPR of whole wave when spilling.
Tue, Feb 9, 8:52 AM · Restricted Project
piotr added inline comments to D96258: [AMDGPU] Introduce Strict WQM mode.
Tue, Feb 9, 5:47 AM · Restricted Project
piotr added a comment to D96257: [AMDGPU] Rename amdgcn_wwm to amdgcn_strict_wwm.

Is it worth adding a test to ensure old intrinsic still works?

Tue, Feb 9, 5:28 AM · Restricted Project

Mon, Feb 8

piotr accepted D94746: [AMDGPU] Move kill lowering to WQM pass and add live mask tracking.
Mon, Feb 8, 5:44 AM · Restricted Project
piotr added reviewers for D96258: [AMDGPU] Introduce Strict WQM mode: arsenm, critson.
Mon, Feb 8, 4:31 AM · Restricted Project
piotr added reviewers for D96257: [AMDGPU] Rename amdgcn_wwm to amdgcn_strict_wwm: arsenm, critson.
Mon, Feb 8, 4:31 AM · Restricted Project
piotr added a comment to D96257: [AMDGPU] Rename amdgcn_wwm to amdgcn_strict_wwm.

See also D96258.

Mon, Feb 8, 4:30 AM · Restricted Project
piotr requested review of D96258: [AMDGPU] Introduce Strict WQM mode.
Mon, Feb 8, 4:29 AM · Restricted Project
piotr requested review of D96257: [AMDGPU] Rename amdgcn_wwm to amdgcn_strict_wwm.
Mon, Feb 8, 4:25 AM · Restricted Project

Wed, Jan 27

piotr committed rGfc8e7411218c: [AMDGPU] Avoid an illegal operand in si-shrink-instructions (authored by piotr).
[AMDGPU] Avoid an illegal operand in si-shrink-instructions
Wed, Jan 27, 11:50 PM
piotr closed D95527: [AMDGPU] Avoid an illegal operand in si-shrink-instructions.
Wed, Jan 27, 11:50 PM · Restricted Project
piotr accepted D95509: [AMDGPU] Mark V_SET_INACTIVE as defining SCC.
Wed, Jan 27, 11:27 PM · Restricted Project
piotr accepted D94747: [AMDGPU] Add llvm.amdgcn.wqm.demote intrinsic.
Wed, Jan 27, 11:16 PM · Restricted Project
piotr added reviewers for D95527: [AMDGPU] Avoid an illegal operand in si-shrink-instructions: arsenm, rampitec.
Wed, Jan 27, 7:24 AM · Restricted Project
piotr requested review of D95527: [AMDGPU] Avoid an illegal operand in si-shrink-instructions.
Wed, Jan 27, 7:18 AM · Restricted Project

Jan 25 2021

piotr added a comment to D94747: [AMDGPU] Add llvm.amdgcn.wqm.demote intrinsic.

LGTM with a few nits - feel free to ignore them.

Jan 25 2021, 2:31 AM · Restricted Project

Jan 21 2021

piotr added a comment to D95132: [AMDGPU][GlobalISel] Run SIAddImgInit.

Thanks!

Jan 21 2021, 7:18 AM · Restricted Project
piotr added inline comments to D95132: [AMDGPU][GlobalISel] Run SIAddImgInit.
Jan 21 2021, 7:11 AM · Restricted Project
piotr added a comment to D86878: [AMDGPU] Fix a miscompile with S_ADD/S_SUB.

Sorry for the lack of activity lately - I plan to get back to this patch in the near future.

Jan 21 2021, 1:01 AM · Restricted Project

Jan 20 2021

piotr added inline comments to D94746: [AMDGPU] Move kill lowering to WQM pass and add live mask tracking.
Jan 20 2021, 2:18 AM · Restricted Project

Jan 18 2021

piotr added a comment to D94746: [AMDGPU] Move kill lowering to WQM pass and add live mask tracking.

LGTM with a few more nits (inline).

Jan 18 2021, 5:58 AM · Restricted Project

Jan 15 2021

piotr added inline comments to D94746: [AMDGPU] Move kill lowering to WQM pass and add live mask tracking.
Jan 15 2021, 3:28 AM · Restricted Project
piotr added inline comments to D94746: [AMDGPU] Move kill lowering to WQM pass and add live mask tracking.
Jan 15 2021, 2:21 AM · Restricted Project
piotr added inline comments to D94746: [AMDGPU] Move kill lowering to WQM pass and add live mask tracking.
Jan 15 2021, 2:12 AM · Restricted Project

Dec 16 2020

piotr committed rGc7afb698caf0: [AMDGPU] Avoid calling copyFastMathFlags in wrong context (authored by piotr).
[AMDGPU] Avoid calling copyFastMathFlags in wrong context
Dec 16 2020, 1:24 AM

Nov 18 2020

piotr committed rGb3b9be4ae77e: SpeculativeExecution: Allow speculating more instruction types (authored by piotr).
SpeculativeExecution: Allow speculating more instruction types
Nov 18 2020, 8:01 AM
piotr closed D91688: SpeculativeExecution: Allow speculating more instruction types.
Nov 18 2020, 8:00 AM · Restricted Project
piotr added a comment to D91688: SpeculativeExecution: Allow speculating more instruction types.

I asked myself the same question - this was tried before - see https://reviews.llvm.org/D24544. I do not know any more details.

Nov 18 2020, 6:27 AM · Restricted Project
piotr updated the summary of D91688: SpeculativeExecution: Allow speculating more instruction types.
Nov 18 2020, 6:20 AM · Restricted Project
piotr updated the diff for D91688: SpeculativeExecution: Allow speculating more instruction types.

Added two more instructions: trunc and freeze (thanks @foad).

Nov 18 2020, 6:10 AM · Restricted Project
piotr added inline comments to D91636: [AMDGPU] Fix and extend vccz workarounds.
Nov 18 2020, 1:48 AM · Restricted Project
piotr added a comment to D91688: SpeculativeExecution: Allow speculating more instruction types.

As Matt pointed out in D91633, extractvalue and insertvalue were also missing, so I am adding them here (the new tests are taken from D24543). Also the tests added earlier in D91633 are being moved to spec-other.ll.

Nov 18 2020, 1:37 AM · Restricted Project
piotr added a reviewer for D91688: SpeculativeExecution: Allow speculating more instruction types: arsenm.
Nov 18 2020, 1:32 AM · Restricted Project
piotr requested review of D91688: SpeculativeExecution: Allow speculating more instruction types.
Nov 18 2020, 1:31 AM · Restricted Project
piotr committed rGc173f1b8ebba: SpeculativeExecution: Allow speculating more instruction types (authored by piotr).
SpeculativeExecution: Allow speculating more instruction types
Nov 18 2020, 12:47 AM
piotr closed D91633: SpeculativeExecution: Allow speculating more instruction types.
Nov 18 2020, 12:47 AM · Restricted Project

Nov 17 2020

piotr retitled D91633: SpeculativeExecution: Allow speculating more instruction types from SpeculativeExecution: Allow speculating more instructions types to SpeculativeExecution: Allow speculating more instruction types.
Nov 17 2020, 8:06 AM · Restricted Project
piotr added reviewers for D91633: SpeculativeExecution: Allow speculating more instruction types: arsenm, dfukalov.
Nov 17 2020, 7:46 AM · Restricted Project
piotr requested review of D91633: SpeculativeExecution: Allow speculating more instruction types.
Nov 17 2020, 7:45 AM · Restricted Project

Nov 13 2020

piotr committed rG47dec5aa603b: [DivergenceAnalysis] Use addRequiredTransitive (authored by piotr).
[DivergenceAnalysis] Use addRequiredTransitive
Nov 13 2020, 5:40 AM
piotr closed D91335: [DivergenceAnalysis] Use addRequiredTransitive.
Nov 13 2020, 5:40 AM · Restricted Project
piotr updated the diff for D91335: [DivergenceAnalysis] Use addRequiredTransitive.

Rebased to retrigger the buildbot checks.

Nov 13 2020, 4:47 AM · Restricted Project

Nov 12 2020

piotr added reviewers for D91335: [DivergenceAnalysis] Use addRequiredTransitive: rampitec, nhaehnle, arsenm.
Nov 12 2020, 4:36 AM · Restricted Project
piotr requested review of D91335: [DivergenceAnalysis] Use addRequiredTransitive.
Nov 12 2020, 4:35 AM · Restricted Project

Oct 22 2020

piotr added inline comments to D86878: [AMDGPU] Fix a miscompile with S_ADD/S_SUB.
Oct 22 2020, 11:43 PM · Restricted Project
piotr committed rG7ae0033ca881: [AMDGPU] Fix expansion of i16 MULH (authored by piotr).
[AMDGPU] Fix expansion of i16 MULH
Oct 22 2020, 8:06 AM
piotr closed D89965: [AMDGPU] Fix expansion of i16 MULH.
Oct 22 2020, 8:06 AM · Restricted Project
piotr added a comment to D86878: [AMDGPU] Fix a miscompile with S_ADD/S_SUB.

Ping.

Oct 22 2020, 7:59 AM · Restricted Project
piotr added a reviewer for D89965: [AMDGPU] Fix expansion of i16 MULH: arsenm.
Oct 22 2020, 7:57 AM · Restricted Project
piotr requested review of D89965: [AMDGPU] Fix expansion of i16 MULH.
Oct 22 2020, 7:56 AM · Restricted Project

Oct 21 2020

piotr added a comment to D80485: [DAGCombiner][PowerPC] Remove isMulhCheaperThanMulShift TLI hook. Use isOperationLegalOrCustom directly instead..

Hi! This commit causes problems for AMDGPU backend - see attached file

. Any ideas before I start investigating this in detail?

LLVM ERROR: Cannot select: t56: i16 = mulhs t42, Constant:i16<-32509>

t42: i16 = truncate t67
  t67: i32 = add t66, t28
    t66: i32 = add t37, t34
      t37: i32 = shl nuw nsw t12, Constant:i32<13>
        t12: i32,ch = CopyFromReg t0, Register:i32 %5
          t11: i32 = Register %5
        t36: i32 = Constant<13>
      t34: i32 = shl nuw nsw t10, Constant:i32<7>
        t10: i32,ch = CopyFromReg t0, Register:i32 %4
          t9: i32 = Register %4
        t26: i32 = Constant<7>
    t28: i32 = add t16, t27
      t16: i32,ch = CopyFromReg t0, Register:i32 %7
        t15: i32 = Register %7
      t27: i32 = shl t8, Constant:i32<7>
        t8: i32,ch = CopyFromReg t0, Register:i32 %3
          t7: i32 = Register %3
        t26: i32 = Constant<7>
t52: i16 = Constant<-32509>

Hi, I agree with the suggestion by @dmgreen. Could you see if that works?

Oct 21 2020, 2:53 AM · Restricted Project

Oct 20 2020

piotr added a comment to D80485: [DAGCombiner][PowerPC] Remove isMulhCheaperThanMulShift TLI hook. Use isOperationLegalOrCustom directly instead..

Hi! This commit causes problems for AMDGPU backend - see attached file

. Any ideas before I start investigating this in detail?

Oct 20 2020, 8:08 AM · Restricted Project

Oct 19 2020

piotr committed rGc872faf6e091: [AMDGPU] Do not generate S_CMP_LG_U64 on gfx7 (authored by piotr).
[AMDGPU] Do not generate S_CMP_LG_U64 on gfx7
Oct 19 2020, 5:45 AM
piotr closed D89536: [AMDGPU] Do not generate S_CMP_LG_U64 on gfx7.
Oct 19 2020, 5:45 AM · Restricted Project

Oct 16 2020

piotr added a reviewer for D89536: [AMDGPU] Do not generate S_CMP_LG_U64 on gfx7: rampitec.
Oct 16 2020, 3:47 AM · Restricted Project
piotr added a comment to D89536: [AMDGPU] Do not generate S_CMP_LG_U64 on gfx7.

Related to D78091.

Oct 16 2020, 3:46 AM · Restricted Project
piotr added reviewers for D89536: [AMDGPU] Do not generate S_CMP_LG_U64 on gfx7: arsenm, alex-t.
Oct 16 2020, 3:42 AM · Restricted Project
piotr requested review of D89536: [AMDGPU] Do not generate S_CMP_LG_U64 on gfx7.
Oct 16 2020, 3:40 AM · Restricted Project

Oct 15 2020

piotr updated the diff for D86878: [AMDGPU] Fix a miscompile with S_ADD/S_SUB.

Renamed variable to avoid a multiline assignment.

Oct 15 2020, 7:13 AM · Restricted Project
piotr updated the diff for D86878: [AMDGPU] Fix a miscompile with S_ADD/S_SUB.

Good spot - I looked at the shader database and indeed the pattern with two connected setcc is frequent enough to special case it so I updated the patch.

Oct 15 2020, 7:07 AM · Restricted Project

Oct 9 2020

piotr updated the diff for D86878: [AMDGPU] Fix a miscompile with S_ADD/S_SUB.

Invert the condition.

Oct 9 2020, 3:54 PM · Restricted Project

Oct 6 2020

piotr added a comment to D86878: [AMDGPU] Fix a miscompile with S_ADD/S_SUB.

Ah now I see what you meant, sorry.

Oct 6 2020, 7:47 AM · Restricted Project
piotr added a comment to D86878: [AMDGPU] Fix a miscompile with S_ADD/S_SUB.

Ping.

Oct 6 2020, 5:19 AM · Restricted Project

Sep 23 2020

piotr committed rG8d7fd73c3a8c: [AMDGPU] Fix merging m0 inits (authored by piotr).
[AMDGPU] Fix merging m0 inits
Sep 23 2020, 12:15 AM
piotr closed D87882: [AMDGPU] Fix merging m0 inits.
Sep 23 2020, 12:15 AM · Restricted Project

Sep 22 2020

piotr added a comment to D87882: [AMDGPU] Fix merging m0 inits.
bb.1:
  ..
  TO_inst   (inits m0)
  ...       (uses m0)
  FROM_inst (clobbers m0)
  ...
  S_CBRANCH_VCCZ %bb.1, undef
  S_BRANCH %bb.2
Sep 22 2020, 6:47 AM · Restricted Project
piotr added inline comments to D86878: [AMDGPU] Fix a miscompile with S_ADD/S_SUB.
Sep 22 2020, 6:18 AM · Restricted Project
piotr updated the diff for D86878: [AMDGPU] Fix a miscompile with S_ADD/S_SUB.

Simplified test and added 5 more.

Sep 22 2020, 6:14 AM · Restricted Project

Sep 20 2020

piotr added a comment to D87882: [AMDGPU] Fix merging m0 inits.

Doesn't loop block self dominate?

You mean, why the condition "MDT.dominates(From, To)" returns false if From and To are in the same BB? Inside that function there is a bb dominance check if the instructions are in different blocks, but if the instructions are in the same block the code checks whether From comes before To.

The bug I am trying to fix is simplified in the test m0-in-loop-0 where before my patch SI_INIT from line 333 was incorrectly removed. Although "m0 = COPY" comes after SI_INIT in that bb, this is a loop and in the second iteration DS_WRITE would see a clobbered m0, so SI_INIT has to be preserved.

That sounds like a bug in the dominate() to me.

Sep 20 2020, 11:43 PM · Restricted Project

Sep 18 2020

piotr added a comment to D87882: [AMDGPU] Fix merging m0 inits.

Doesn't loop block self dominate?

Sep 18 2020, 1:43 PM · Restricted Project

Sep 17 2020

piotr added reviewers for D87882: [AMDGPU] Fix merging m0 inits: rampitec, kerbowa, arsenm.
Sep 17 2020, 10:14 PM · Restricted Project
piotr requested review of D87882: [AMDGPU] Fix merging m0 inits.
Sep 17 2020, 10:13 PM · Restricted Project
piotr added inline comments to D86878: [AMDGPU] Fix a miscompile with S_ADD/S_SUB.
Sep 17 2020, 7:46 AM · Restricted Project

Sep 11 2020

piotr retitled D86878: [AMDGPU] Fix a miscompile with S_ADD/S_SUB from [AMDGPU] Fix a miscompile in add combine to [AMDGPU] Fix a miscompile with S_ADD/S_SUB.
Sep 11 2020, 2:41 AM · Restricted Project
piotr updated the diff for D86878: [AMDGPU] Fix a miscompile with S_ADD/S_SUB.

I think you are right - I moved the fix to the better place.

Sep 11 2020, 2:39 AM · Restricted Project