Page MenuHomePhabricator

dfukalov (Daniil Fukalov)
Compiler Engineer at AMD

Projects

User does not belong to any projects.

User Details

User Since
Mar 27 2014, 8:40 AM (327 w, 5 d)

Recent Activity

Yesterday

dfukalov added a comment to D82761: SpeculativeExecution: Fix for logic change introduced in D81730..

Ping.

Tue, Jul 7, 10:25 AM · Restricted Project

Thu, Jul 2

dfukalov accepted D82453: [CostModel] Fix for cast crash.
Thu, Jul 2, 6:57 AM · Restricted Project
dfukalov added inline comments to D82453: [CostModel] Fix for cast crash.
Thu, Jul 2, 3:10 AM · Restricted Project

Tue, Jun 30

dfukalov added inline comments to D82761: SpeculativeExecution: Fix for logic change introduced in D81730..
Tue, Jun 30, 7:00 AM · Restricted Project
dfukalov updated the diff for D82761: SpeculativeExecution: Fix for logic change introduced in D81730..

Test updated

Tue, Jun 30, 7:00 AM · Restricted Project
dfukalov committed rG1a6cebb4d12c: [PM] Fix new PM to perform SpeculativeExecution as in old PM (authored by dfukalov).
[PM] Fix new PM to perform SpeculativeExecution as in old PM
Tue, Jun 30, 5:23 AM
dfukalov closed D82735: [PM] Fix new PM to perform SpeculativeExecution as in old PM.
Tue, Jun 30, 5:23 AM · Restricted Project

Mon, Jun 29

dfukalov committed rG8cc722ffc749: [NFC] Fixed ignored .hip test. (authored by dfukalov).
[NFC] Fixed ignored .hip test.
Mon, Jun 29, 9:11 AM
dfukalov closed D82764: [NFC] Fixed ignored .hip test..
Mon, Jun 29, 9:10 AM · Restricted Project
dfukalov added a reviewer for D82735: [PM] Fix new PM to perform SpeculativeExecution as in old PM: arsenm.
Mon, Jun 29, 6:58 AM · Restricted Project
dfukalov created D82764: [NFC] Fixed ignored .hip test..
Mon, Jun 29, 6:58 AM · Restricted Project
dfukalov created D82761: SpeculativeExecution: Fix for logic change introduced in D81730..
Mon, Jun 29, 5:54 AM · Restricted Project

Sun, Jun 28

dfukalov created D82735: [PM] Fix new PM to perform SpeculativeExecution as in old PM.
Sun, Jun 28, 6:13 PM · Restricted Project
dfukalov committed rGc7bcd431d9c4: SpeculativeExecution: fix incorrect debug info move (authored by dfukalov).
SpeculativeExecution: fix incorrect debug info move
Sun, Jun 28, 4:57 AM
dfukalov closed D81730: SpeculativeExecution: fix incorrect debug info move.
Sun, Jun 28, 4:57 AM · debug-info, Restricted Project
dfukalov abandoned D82650: [HIP] Set default FP_CONTRACT to ON.

I was completely wrong

Sun, Jun 28, 3:21 AM · Restricted Project

Fri, Jun 26

dfukalov planned changes to D82650: [HIP] Set default FP_CONTRACT to ON.

I think the point of -ffp-contract=fast is that it allows emitting fma across statements if there are such opportunities whereas -ffp-contract=on only allows fma in one statement

https://llvm.org/docs/CompileCudaWithLLVM.html#flags-that-control-numerical-code

Yes, you're right. I was confused by current FE behavior and a number of different commits, reverts and re-commits.

Fri, Jun 26, 7:37 AM · Restricted Project
dfukalov created D82650: [HIP] Set default FP_CONTRACT to ON.
Fri, Jun 26, 5:56 AM · Restricted Project

Wed, Jun 24

dfukalov added inline comments to D82453: [CostModel] Fix for cast crash.
Wed, Jun 24, 2:40 PM · Restricted Project
dfukalov committed rG7ddee0922fc2: [NFCI][CostModel] Add const to Value*. (authored by dfukalov).
[NFCI][CostModel] Add const to Value*.
Wed, Jun 24, 1:35 PM
dfukalov closed D82383: [NFCI][CostModel] Add const to Value*..
Wed, Jun 24, 1:35 PM · Restricted Project

Tue, Jun 23

dfukalov created D82383: [NFCI][CostModel] Add const to Value*..
Tue, Jun 23, 7:26 AM · Restricted Project

Fri, Jun 19

dfukalov added a comment to D81730: SpeculativeExecution: fix incorrect debug info move.

ping...

Fri, Jun 19, 4:52 PM · debug-info, Restricted Project
dfukalov committed rG129388ddc4df: [AMDGPU][CostModel] Add fneg cost estimation (authored by dfukalov).
[AMDGPU][CostModel] Add fneg cost estimation
Fri, Jun 19, 7:33 AM
dfukalov closed D82065: [AMDGPU][CostModel] Add fneg cost estimation.
Fri, Jun 19, 7:33 AM · Restricted Project

Thu, Jun 18

dfukalov added inline comments to D82065: [AMDGPU][CostModel] Add fneg cost estimation.
Thu, Jun 18, 5:57 AM · Restricted Project
dfukalov updated the diff for D82065: [AMDGPU][CostModel] Add fneg cost estimation.

Added comment, fixed typo

Thu, Jun 18, 5:57 AM · Restricted Project

Wed, Jun 17

dfukalov created D82065: [AMDGPU][CostModel] Add fneg cost estimation.
Wed, Jun 17, 5:51 PM · Restricted Project

Sun, Jun 14

dfukalov accepted D81635: [NFCI][CostModel] Unify FNeg cost.
Sun, Jun 14, 6:39 PM · Restricted Project

Sat, Jun 13

dfukalov accepted D81643: [NFCI][CostModel] Unify ExtractElement cost..
Sat, Jun 13, 4:18 AM · Restricted Project

Fri, Jun 12

dfukalov updated the diff for D81730: SpeculativeExecution: fix incorrect debug info move.

updated comments

Fri, Jun 12, 4:34 PM · debug-info, Restricted Project
dfukalov added a comment to D79164: [CostModel] getCFInstrCost.

Actually it seems all of these dozens of "CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32" may be removed from most of tests not related to getCFInstrCost

Fri, Jun 12, 8:37 AM · Restricted Project
dfukalov accepted D79164: [CostModel] getCFInstrCost.

LGTM, although minor change in test' comment

Fri, Jun 12, 8:37 AM · Restricted Project
dfukalov created D81730: SpeculativeExecution: fix incorrect debug info move.
Fri, Jun 12, 4:49 AM · debug-info, Restricted Project

Wed, Jun 10

dfukalov added a comment to D74825: SpeculativeExecution: fixed ingoring free execution.

I've just got the bug, will take a look

Wed, Jun 10, 8:12 AM · Restricted Project
dfukalov added inline comments to D81448: [CostModel] Unify Shuffle and InsertElement Costs.
Wed, Jun 10, 4:20 AM · Restricted Project

Tue, Jun 9

dfukalov accepted D81448: [CostModel] Unify Shuffle and InsertElement Costs.
Tue, Jun 9, 6:15 PM · Restricted Project
dfukalov added inline comments to D79164: [CostModel] getCFInstrCost.
Tue, Jun 9, 5:42 PM · Restricted Project
dfukalov accepted D80992: [CostModel] Unify getArithmeticInstrCost .
Tue, Jun 9, 5:09 PM · Restricted Project

Mon, Jun 8

dfukalov committed rG6a31a9a5431f: [AMDGPU][NFC] Skip processing intrinsics that do not become real instructions (authored by dfukalov).
[AMDGPU][NFC] Skip processing intrinsics that do not become real instructions
Mon, Jun 8, 6:17 PM
dfukalov closed D81260: [AMDGPU][NFC] Skip processing intrinsics that do not become real instructions.
Mon, Jun 8, 6:17 PM · Restricted Project

Jun 6 2020

dfukalov committed rGc94d32a6b302: [AMDGPU] Increase max iterations count to analyze complete unroll (authored by dfukalov).
[AMDGPU] Increase max iterations count to analyze complete unroll
Jun 6 2020, 6:55 AM
dfukalov closed D81204: [AMDGPU] Increase max iterations count to analyze complete unroll.
Jun 6 2020, 6:54 AM · Restricted Project
dfukalov updated the diff for D81204: [AMDGPU] Increase max iterations count to analyze complete unroll.

fixed spelling

Jun 6 2020, 4:14 AM · Restricted Project
dfukalov added a comment to D81260: [AMDGPU][NFC] Skip processing intrinsics that do not become real instructions.

There are the tests in llvm/test/CodeGen/AMDGPU/simplify-libcalls.ll

Jun 6 2020, 1:02 AM · Restricted Project

Jun 5 2020

dfukalov updated the diff for D81260: [AMDGPU][NFC] Skip processing intrinsics that do not become real instructions.

diff reuploaded

Jun 5 2020, 6:27 PM · Restricted Project
dfukalov added a comment to D81260: [AMDGPU][NFC] Skip processing intrinsics that do not become real instructions.
Jun 5 2020, 5:55 PM · Restricted Project
dfukalov updated the diff for D81260: [AMDGPU][NFC] Skip processing intrinsics that do not become real instructions.

added test

Jun 5 2020, 5:54 PM · Restricted Project
dfukalov updated the diff for D81260: [AMDGPU][NFC] Skip processing intrinsics that do not become real instructions.

test added

Jun 5 2020, 5:54 PM · Restricted Project
dfukalov updated the diff for D81204: [AMDGPU] Increase max iterations count to analyze complete unroll.

test updated as requested

Jun 5 2020, 3:10 PM · Restricted Project
dfukalov added a comment to D81260: [AMDGPU][NFC] Skip processing intrinsics that do not become real instructions.

But how did it work?! It should have skipped any call which is not a lifetime intrinsic.

Jun 5 2020, 3:10 PM · Restricted Project
dfukalov added a comment to D81260: [AMDGPU][NFC] Skip processing intrinsics that do not become real instructions.

Phabricator somehow shows wrong diff after patch update. Correct diff is

-       if (!CI) continue;
+       if (!CI || isa<DbgInfoIntrinsic>(CI) || CI->isLifetimeStartOrEnd())
+         continue;
Jun 5 2020, 12:54 PM · Restricted Project
dfukalov retitled D81260: [AMDGPU][NFC] Skip processing intrinsics that do not become real instructions from [AMDGPU] Skip processing intrinsics that do not become real instructions to [AMDGPU][NFC] Skip processing intrinsics that do not become real instructions.
Jun 5 2020, 12:54 PM · Restricted Project
dfukalov added a comment to D80992: [CostModel] Unify getArithmeticInstrCost .

AMDGPU part LGTM,
let's wait a couple of days to allow others to take a look

Jun 5 2020, 10:35 AM · Restricted Project
dfukalov updated the diff for D81260: [AMDGPU][NFC] Skip processing intrinsics that do not become real instructions.

fixed typo

Jun 5 2020, 10:01 AM · Restricted Project
dfukalov created D81260: [AMDGPU][NFC] Skip processing intrinsics that do not become real instructions.
Jun 5 2020, 7:11 AM · Restricted Project

Jun 4 2020

dfukalov updated the diff for D81204: [AMDGPU] Increase max iterations count to analyze complete unroll.

fixing lint message

Jun 4 2020, 5:09 PM · Restricted Project
dfukalov created D81204: [AMDGPU] Increase max iterations count to analyze complete unroll.
Jun 4 2020, 5:09 PM · Restricted Project
dfukalov added inline comments to D80992: [CostModel] Unify getArithmeticInstrCost .
Jun 4 2020, 3:30 PM · Restricted Project

Jun 3 2020

dfukalov accepted D80994: [NFCI][CostModel][AMDGPU] Simplify getUserCost.

Thanks!

Jun 3 2020, 6:45 PM · Restricted Project

May 22 2020

dfukalov added a comment to D79164: [CostModel] getCFInstrCost.

AMDGPU part LGTM

May 22 2020, 6:56 AM · Restricted Project

May 21 2020

dfukalov accepted D79849: [NFCI][CostModel] Unify getCFInstrCost.
May 21 2020, 2:40 AM · Restricted Project
dfukalov accepted D79848: [CostModel] Unify getCastInstrCost.
May 21 2020, 2:40 AM · Restricted Project

May 20 2020

dfukalov accepted D79941: [NFCI][CostModel] Refactor getIntrinsicInstrCost.
May 20 2020, 3:45 AM · Restricted Project

May 19 2020

dfukalov added a comment to D79941: [NFCI][CostModel] Refactor getIntrinsicInstrCost.

I like the idea to simplify cost model implementation since it was not trivial to make even a small changes.

May 19 2020, 8:41 AM · Restricted Project

May 18 2020

dfukalov accepted D80012: [CostModel] Unify Intrinsic Costs..
May 18 2020, 11:54 AM · Restricted Project

Feb 27 2020

Gerrit Code Review <gerritcr@atlswgit01.amd.com> committed rG93a40cd800be: Merge "fix for codeobject disasm dump" into amd-common (authored by dfukalov).
Merge "fix for codeobject disasm dump" into amd-common
Feb 27 2020, 4:46 AM
dfukalov committed rG594f5e471df5: fix for codeobject disasm dump (authored by dfukalov).
fix for codeobject disasm dump
Feb 27 2020, 4:46 AM
Gerrit Code Review <gerritcr@atlswgit01.amd.com> committed rG1a57162d23d3: Merge "remove warning on unised canHandled() in case of release build" into amd… (authored by dfukalov).
Merge "remove warning on unised canHandled() in case of release build" into amd…
Feb 27 2020, 3:53 AM
dfukalov committed rGe37edcf452ef: remove warning on unised canHandled() in case of release build (authored by dfukalov).
remove warning on unised canHandled() in case of release build
Feb 27 2020, 3:53 AM
Gerrit Code Review <gerritcr@atlswgit01.amd.com> committed rGb0084c83d31e: Merge "fix include path to LLVM_SOURCE_DIR since in case if llvm tree is a… (authored by dfukalov).
Merge "fix include path to LLVM_SOURCE_DIR since in case if llvm tree is a…
Feb 27 2020, 3:48 AM
dfukalov committed rG16dd4e7fd468: fix include path to LLVM_SOURCE_DIR since in case if llvm tree is a… (authored by dfukalov).
fix include path to LLVM_SOURCE_DIR since in case if llvm tree is a…
Feb 27 2020, 3:48 AM
dfukalov committed rG1a4c91b32cb5: Fix and refine AMDGPUConvertAtomicLibCalls pass (authored by dfukalov).
Fix and refine AMDGPUConvertAtomicLibCalls pass
Feb 27 2020, 3:47 AM
GitHub <noreply@github.com> committed rG8bc3626c2906: LowerKernelCalls pass ported from HSAIL compiler (#43) (authored by dfukalov).
LowerKernelCalls pass ported from HSAIL compiler (#43)
Feb 27 2020, 3:38 AM
GitHub <noreply@github.com> committed rG6e06de958418: fixed clp pass when processing library functions with half type parameters (#46) (authored by dfukalov).
fixed clp pass when processing library functions with half type parameters (#46)
Feb 27 2020, 3:38 AM
GitHub <noreply@github.com> committed rG9a219c0af06c: Merge pull request #34 from RadeonOpenCompute/clpVectorExpansion (authored by dfukalov).
Merge pull request #34 from RadeonOpenCompute/clpVectorExpansion
Feb 27 2020, 3:37 AM
dfukalov committed rGaeb93640fd33: Merge branch 'amd-common' of https://github.com/RadeonOpenCompute/llvm into… (authored by dfukalov).
Merge branch 'amd-common' of https://github.com/RadeonOpenCompute/llvm into…
Feb 27 2020, 3:37 AM
dfukalov committed rGd05c16645fd6: fixed processing of functions with mixed vector and scalar parameters, test… (authored by dfukalov).
fixed processing of functions with mixed vector and scalar parameters, test…
Feb 27 2020, 3:37 AM
dfukalov committed rGa9fdac06d4ab: Merge branch 'amd-common' of https://github.com/RadeonOpenCompute/llvm into… (authored by dfukalov).
Merge branch 'amd-common' of https://github.com/RadeonOpenCompute/llvm into…
Feb 27 2020, 3:37 AM
Tom Stellard <thomas.stellard@and.com> committed rGeb4b13dd5a63: fixed SWDEV-106711 "Assert in atomic pre link pass" (authored by dfukalov).
fixed SWDEV-106711 "Assert in atomic pre link pass"
Feb 27 2020, 3:36 AM
dfukalov committed rG3b6fd3e1c2b3: Merge branch 'amd-common' of https://github.com/RadeonOpenCompute/llvm into… (authored by dfukalov).
Merge branch 'amd-common' of https://github.com/RadeonOpenCompute/llvm into…
Feb 27 2020, 3:36 AM
dfukalov committed rGef145ee3e742: Merge branch 'amd-common' of https://github.com/RadeonOpenCompute/llvm into… (authored by dfukalov).
Merge branch 'amd-common' of https://github.com/RadeonOpenCompute/llvm into…
Feb 27 2020, 3:35 AM
GitHub <noreply@github.com> committed rG3efb661d7642: Merge pull request #36 from RadeonOpenCompute/revertingLowerKernelCalls (authored by dfukalov).
Merge pull request #36 from RadeonOpenCompute/revertingLowerKernelCalls
Feb 27 2020, 3:35 AM
dfukalov committed rGb8d205c05c7e: Revert "Ported lower kernels calls pre-link pass from hsail compiler" (authored by dfukalov).
Revert "Ported lower kernels calls pre-link pass from hsail compiler"
Feb 27 2020, 3:35 AM
dfukalov committed rGcd6b588486c8: Revert "removed declarations processing since the pass should clone kernel… (authored by dfukalov).
Revert "removed declarations processing since the pass should clone kernel…
Feb 27 2020, 3:35 AM
dfukalov committed rG093d528df4f6: removed declarations processing since the pass should clone kernel functions (authored by dfukalov).
removed declarations processing since the pass should clone kernel functions
Feb 27 2020, 3:35 AM
dfukalov committed rGbf1858ed94ae: Ported lower kernels calls pre-link pass from hsail compiler (authored by dfukalov).
Ported lower kernels calls pre-link pass from hsail compiler
Feb 27 2020, 3:35 AM
dfukalov committed rGf59a9f350a63: 1. revised as per Valery comment 2. extended regression test (authored by dfukalov).
1. revised as per Valery comment 2. extended regression test
Feb 27 2020, 3:35 AM
dfukalov committed rG929691646418: fixed almost all of coding style issues (authored by dfukalov).
fixed almost all of coding style issues
Feb 27 2020, 3:34 AM
dfukalov committed rG572f7b1ca0e9: fixed a number of Tom's suggestions (authored by dfukalov).
fixed a number of Tom's suggestions
Feb 27 2020, 3:34 AM
dfukalov committed rG9dd2cef9e055: SWDEV-103116 "[OCL-LC-ROCm] Add OpenCL specific passes - clp" (authored by dfukalov).
SWDEV-103116 "[OCL-LC-ROCm] Add OpenCL specific passes - clp"
Feb 27 2020, 3:34 AM

Feb 20 2020

dfukalov committed rGdbfc682e2ba2: SpeculativeExecution: fixed ingoring free execution (authored by dfukalov).
SpeculativeExecution: fixed ingoring free execution
Feb 20 2020, 3:48 AM
dfukalov closed D74825: SpeculativeExecution: fixed ingoring free execution.
Feb 20 2020, 3:48 AM · Restricted Project
dfukalov updated the diff for D74825: SpeculativeExecution: fixed ingoring free execution.

Test updated as requested

Feb 20 2020, 3:43 AM · Restricted Project

Feb 19 2020

dfukalov created D74825: SpeculativeExecution: fixed ingoring free execution.
Feb 19 2020, 3:37 AM · Restricted Project

Jan 20 2020

dfukalov committed rGde34b54edce4: [SCEV] Swap guards estimation sequence. NFC (authored by dfukalov).
[SCEV] Swap guards estimation sequence. NFC
Jan 20 2020, 5:49 AM
dfukalov closed D72929: [SCEV] Swap guards estimation sequence. NFC.
Jan 20 2020, 5:49 AM · Restricted Project

Jan 17 2020

dfukalov created D72929: [SCEV] Swap guards estimation sequence. NFC.
Jan 17 2020, 8:34 AM · Restricted Project

Dec 17 2019

dfukalov accepted D71622: [AMDGPU] Fixed cost model for packed 16 bit ops.

Thanks!

Dec 17 2019, 2:37 PM · Restricted Project

Dec 10 2019

dfukalov abandoned D71123: [NFC] Fix compilation with gcc 5.4.
Dec 10 2019, 4:03 AM · Restricted Project