Page MenuHomePhabricator

piotr (Piotr Sobczak)
User

Projects

User does not belong to any projects.

User Details

User Since
Dec 4 2018, 6:02 AM (58 w, 6 d)

Recent Activity

Fri, Jan 17

piotr added a comment to D71785: [InstCombine][AMDGPU] Trim components of s_buffer_load.

I intend to merge the patch soon, based on the fact that the patch was reviewed in D70315. Compared to that version, 3 problematic cases were removed and tests updated.

Fri, Jan 17, 7:46 AM · Restricted Project

Fri, Jan 10

piotr added a comment to D71785: [InstCombine][AMDGPU] Trim components of s_buffer_load.

Ping.

Fri, Jan 10, 2:11 AM · Restricted Project

Dec 20 2019

piotr added reviewers for D71785: [InstCombine][AMDGPU] Trim components of s_buffer_load: mareko, nhaehnle.
Dec 20 2019, 1:58 PM · Restricted Project
piotr added a comment to D71785: [InstCombine][AMDGPU] Trim components of s_buffer_load.

This is another attempt to trim unused components of buffer loads. Comparing to D70315, this commit does not handle formatted buffer loads, as it is not safe.

Dec 20 2019, 1:58 PM · Restricted Project
piotr created D71785: [InstCombine][AMDGPU] Trim components of s_buffer_load.
Dec 20 2019, 1:47 PM · Restricted Project

Dec 18 2019

piotr committed rG40b5a0f7c89c: Revert "[InstCombine][AMDGPU] Trim more components of *buffer_load" (authored by piotr).
Revert "[InstCombine][AMDGPU] Trim more components of *buffer_load"
Dec 18 2019, 1:13 PM
piotr added a reverting change for rG65f94b33808d: [InstCombine][AMDGPU] Trim more components of *buffer_load: rG40b5a0f7c89c: Revert "[InstCombine][AMDGPU] Trim more components of *buffer_load".
Dec 18 2019, 1:13 PM
piotr added a reverting change for D70315: [InstCombine][AMDGPU] Trim more components of *buffer_load: rG40b5a0f7c89c: Revert "[InstCombine][AMDGPU] Trim more components of *buffer_load".
Dec 18 2019, 1:13 PM · Restricted Project

Dec 17 2019

piotr committed rG65f94b33808d: [InstCombine][AMDGPU] Trim more components of *buffer_load (authored by piotr).
[InstCombine][AMDGPU] Trim more components of *buffer_load
Dec 17 2019, 8:53 AM
piotr closed D70315: [InstCombine][AMDGPU] Trim more components of *buffer_load.
Dec 17 2019, 8:53 AM · Restricted Project

Dec 12 2019

piotr added a comment to D70315: [InstCombine][AMDGPU] Trim more components of *buffer_load.

Ping.

Dec 12 2019, 12:34 AM · Restricted Project

Nov 20 2019

piotr committed rG4a801170f36a: [AMDGPU][SILoadStoreOptimizer] Merge TBUFFER loads/stores (authored by piotr).
[AMDGPU][SILoadStoreOptimizer] Merge TBUFFER loads/stores
Nov 20 2019, 2:00 PM
piotr closed D69794: [AMDGPU][SILoadStoreOptimizer] Merge TBUFFER loads/stores.
Nov 20 2019, 2:00 PM · Restricted Project

Nov 19 2019

piotr updated the diff for D70315: [InstCombine][AMDGPU] Trim more components of *buffer_load.

Addressed review comments, updated var name and fixed language in a comment. Regarding the new test: the code is in the function that can only be called by SimplifyDemandedVectorElts, so only for intrinsics with vector types, not pointer type.

Nov 19 2019, 2:06 PM · Restricted Project

Nov 18 2019

piotr added a comment to D69794: [AMDGPU][SILoadStoreOptimizer] Merge TBUFFER loads/stores.

Ping

Nov 18 2019, 11:53 AM · Restricted Project

Nov 15 2019

piotr added a reviewer for D70315: [InstCombine][AMDGPU] Trim more components of *buffer_load: arsenm.
Nov 15 2019, 8:00 AM · Restricted Project
piotr created D70315: [InstCombine][AMDGPU] Trim more components of *buffer_load.
Nov 15 2019, 8:00 AM · Restricted Project
piotr committed rG02419ab5c739: [AMDGPU] Lower llvm.amdgcn.s.buffer.load.v3[i|f]32 (authored by piotr).
[AMDGPU] Lower llvm.amdgcn.s.buffer.load.v3[i|f]32
Nov 15 2019, 6:04 AM
piotr closed D70118: [AMDGPU] Lower llvm.amdgcn.s.buffer.load.v3[i|f]32.
Nov 15 2019, 6:04 AM · Restricted Project
piotr updated the diff for D70118: [AMDGPU] Lower llvm.amdgcn.s.buffer.load.v3[i|f]32.

Widening instead of splitting.

Nov 15 2019, 4:30 AM · Restricted Project
piotr added inline comments to D70118: [AMDGPU] Lower llvm.amdgcn.s.buffer.load.v3[i|f]32.
Nov 15 2019, 12:46 AM · Restricted Project

Nov 13 2019

piotr added inline comments to D70118: [AMDGPU] Lower llvm.amdgcn.s.buffer.load.v3[i|f]32.
Nov 13 2019, 3:29 AM · Restricted Project
piotr updated the diff for D70118: [AMDGPU] Lower llvm.amdgcn.s.buffer.load.v3[i|f]32.

Using the alignment from the ABI type alignment.

Nov 13 2019, 1:46 AM · Restricted Project

Nov 12 2019

piotr added reviewers for D70118: [AMDGPU] Lower llvm.amdgcn.s.buffer.load.v3[i|f]32: arsenm, nhaehnle.
Nov 12 2019, 4:04 AM · Restricted Project
piotr created D70118: [AMDGPU] Lower llvm.amdgcn.s.buffer.load.v3[i|f]32.
Nov 12 2019, 4:04 AM · Restricted Project

Nov 7 2019

piotr added a comment to D69794: [AMDGPU][SILoadStoreOptimizer] Merge TBUFFER loads/stores.

I investigated briefly the possibility of extending LoadStoreVectorizer, but using that pass looked tricky because, among other things, the intrinsics which I want to handle here are not instances of LoadInst for which that pass is well-suited. In fact, extending the si-load-store-opt pass seemed simpler. This patch adds support for MTBUF instructions (analogous to the existing MUBUF) and handles querying of buffer formats.

Nov 7 2019, 5:50 AM · Restricted Project

Nov 4 2019

piotr added a reviewer for D69794: [AMDGPU][SILoadStoreOptimizer] Merge TBUFFER loads/stores: nhaehnle.
Nov 4 2019, 2:49 AM · Restricted Project
piotr created D69794: [AMDGPU][SILoadStoreOptimizer] Merge TBUFFER loads/stores.
Nov 4 2019, 2:43 AM · Restricted Project

Oct 21 2019

piotr committed rGa861c9aef926: [InstCombine] Allow values with multiple users in SimplifyDemandedVectorElts (authored by piotr).
[InstCombine] Allow values with multiple users in SimplifyDemandedVectorElts
Oct 21 2019, 1:19 AM
piotr closed D67345: [InstCombine] Allow values with multiple users in SimplifyDemandedVectorElts.
Oct 21 2019, 1:19 AM · Restricted Project
piotr committed rL375395: [InstCombine] Allow values with multiple users in SimplifyDemandedVectorElts.
[InstCombine] Allow values with multiple users in SimplifyDemandedVectorElts
Oct 21 2019, 1:10 AM

Oct 16 2019

piotr committed rG79769a4475b3: [InstCombine][AMDGPU] Fix crash with v3i16/v3f16 buffer intrinsics (authored by piotr).
[InstCombine][AMDGPU] Fix crash with v3i16/v3f16 buffer intrinsics
Oct 16 2019, 4:18 AM
piotr committed rL374993: [InstCombine][AMDGPU] Fix crash with v3i16/v3f16 buffer intrinsics.
[InstCombine][AMDGPU] Fix crash with v3i16/v3f16 buffer intrinsics
Oct 16 2019, 4:18 AM
piotr closed D68865: [InstCombine][AMDGPU] Fix crash with v3i16/v3f16 buffer intrinsics.
Oct 16 2019, 4:18 AM · Restricted Project
piotr committed rG02baaca742f7: [AMDGPU] Extend the SI Load/Store optimizer (authored by piotr).
[AMDGPU] Extend the SI Load/Store optimizer
Oct 16 2019, 3:20 AM
piotr committed rL374984: [AMDGPU] Extend the SI Load/Store optimizer.
[AMDGPU] Extend the SI Load/Store optimizer
Oct 16 2019, 3:20 AM
piotr closed D64911: [AMDGPU] Extend the SI Load/Store optimizer.
Oct 16 2019, 3:19 AM · Restricted Project
piotr updated the diff for D67345: [InstCombine] Allow values with multiple users in SimplifyDemandedVectorElts.

As suggested by Nicolai, I extended the patch to also handle shufflevector users in addition to extractelement. The new generalized version also simplifies the case where three extract elements are used (the previous patch only handled two). The new code also makes it easy to extend the case list even more in the future.

Oct 16 2019, 12:11 AM · Restricted Project

Oct 14 2019

piotr updated the diff for D64911: [AMDGPU] Extend the SI Load/Store optimizer.

Added TODO.

Oct 14 2019, 11:20 PM · Restricted Project
piotr added inline comments to D64911: [AMDGPU] Extend the SI Load/Store optimizer.
Oct 14 2019, 11:09 PM · Restricted Project
piotr updated the diff for D68865: [InstCombine][AMDGPU] Fix crash with v3i16/v3f16 buffer intrinsics.

Changed "== 16" to "!= 32".

Oct 14 2019, 5:01 AM · Restricted Project

Oct 11 2019

piotr added inline comments to D64911: [AMDGPU] Extend the SI Load/Store optimizer.
Oct 11 2019, 7:08 AM · Restricted Project
piotr updated the diff for D64911: [AMDGPU] Extend the SI Load/Store optimizer.

Rebased and addressed review comments.

Oct 11 2019, 7:01 AM · Restricted Project
piotr added a reviewer for D68865: [InstCombine][AMDGPU] Fix crash with v3i16/v3f16 buffer intrinsics: arsenm.
Oct 11 2019, 5:26 AM · Restricted Project
piotr added reviewers for D68865: [InstCombine][AMDGPU] Fix crash with v3i16/v3f16 buffer intrinsics: tpr, nhaehnle.
Oct 11 2019, 5:26 AM · Restricted Project
piotr updated the diff for D68865: [InstCombine][AMDGPU] Fix crash with v3i16/v3f16 buffer intrinsics.

Fixed wrong spacing.

Oct 11 2019, 5:18 AM · Restricted Project
piotr created D68865: [InstCombine][AMDGPU] Fix crash with v3i16/v3f16 buffer intrinsics.
Oct 11 2019, 5:11 AM · Restricted Project

Oct 4 2019

piotr updated the diff for D64911: [AMDGPU] Extend the SI Load/Store optimizer.

Rebased after D68384 has been merged.

Oct 4 2019, 4:54 AM · Restricted Project
piotr committed rG165e469145ec: [AMDGPU][SILoadStoreOptimizer] NFC: Refactor code (authored by piotr).
[AMDGPU][SILoadStoreOptimizer] NFC: Refactor code
Oct 4 2019, 12:09 AM
piotr committed rL373699: [AMDGPU][SILoadStoreOptimizer] NFC: Refactor code.
[AMDGPU][SILoadStoreOptimizer] NFC: Refactor code
Oct 4 2019, 12:08 AM
piotr closed D68384: [AMDGPU][SILoadStoreOptimizer] NFC: Refactor code.
Oct 4 2019, 12:08 AM · Restricted Project

Oct 3 2019

piotr added reviewers for D68384: [AMDGPU][SILoadStoreOptimizer] NFC: Refactor code: nhaehnle, arsenm, tstellar.
Oct 3 2019, 5:41 AM · Restricted Project
piotr updated the diff for D64911: [AMDGPU] Extend the SI Load/Store optimizer.

A recent patch in SILoadOptimizer (D65496) conflicted with this patch. I have rebased it to the latest master, but for clarity I will split the review into two parts: a separate review for the NFC refactoring (D68384) and the current review only for merging MIMG instruction.

Oct 3 2019, 5:29 AM · Restricted Project
piotr created D68384: [AMDGPU][SILoadStoreOptimizer] NFC: Refactor code.
Oct 3 2019, 5:26 AM · Restricted Project

Oct 2 2019

piotr committed rL373491: [AMDGPU] Extend buffer intrinsics with swizzling.
[AMDGPU] Extend buffer intrinsics with swizzling
Oct 2 2019, 10:25 AM
piotr committed rG265e94e65798: [AMDGPU] Extend buffer intrinsics with swizzling (authored by piotr).
[AMDGPU] Extend buffer intrinsics with swizzling
Oct 2 2019, 10:25 AM
piotr closed D68200: [AMDGPU] Extend buffer intrinsics with swizzling.
Oct 2 2019, 10:25 AM · Restricted Project
piotr added inline comments to D68200: [AMDGPU] Extend buffer intrinsics with swizzling.
Oct 2 2019, 2:40 AM · Restricted Project
piotr updated the diff for D68200: [AMDGPU] Extend buffer intrinsics with swizzling.

Addressed review comments and rebased.

Oct 2 2019, 2:38 AM · Restricted Project

Oct 1 2019

piotr added a comment to D67345: [InstCombine] Allow values with multiple users in SimplifyDemandedVectorElts.

This is a useful change, but there is an unfortunate asymmetry here in how the code is structured: in addition to extractelement, we could also have shufflevector users (or masked stores etc.). Presumably we'd be able to handle all of those together without duplicating the code. Is there a way to take this into account?

Also, two comments inline.

Oct 1 2019, 5:05 AM · Restricted Project

Sep 30 2019

piotr added a comment to D68200: [AMDGPU] Extend buffer intrinsics with swizzling.

I thought this was a property of the resource descriptor? Why do you need to add it to the intrinsic?

Sep 30 2019, 7:53 AM · Restricted Project
piotr added reviewers for D68200: [AMDGPU] Extend buffer intrinsics with swizzling: arsenm, nhaehnle, tpr.
Sep 30 2019, 1:42 AM · Restricted Project
piotr created D68200: [AMDGPU] Extend buffer intrinsics with swizzling.
Sep 30 2019, 1:42 AM · Restricted Project

Sep 27 2019

piotr added a comment to D67345: [InstCombine] Allow values with multiple users in SimplifyDemandedVectorElts.

Ping

Sep 27 2019, 12:29 AM · Restricted Project

Sep 23 2019

piotr committed rL372571: Request commit access for piotr.
Request commit access for piotr
Sep 23 2019, 2:57 AM

Sep 19 2019

piotr added a comment to D67345: [InstCombine] Allow values with multiple users in SimplifyDemandedVectorElts.

Ping

Sep 19 2019, 7:31 AM · Restricted Project

Sep 9 2019

piotr added a reviewer for D67345: [InstCombine] Allow values with multiple users in SimplifyDemandedVectorElts: majnemer.
Sep 9 2019, 3:49 AM · Restricted Project
piotr added reviewers for D67345: [InstCombine] Allow values with multiple users in SimplifyDemandedVectorElts: reames, arsenm.
Sep 9 2019, 3:49 AM · Restricted Project
piotr created D67345: [InstCombine] Allow values with multiple users in SimplifyDemandedVectorElts.
Sep 9 2019, 3:46 AM · Restricted Project

Sep 2 2019

piotr updated the diff for D64911: [AMDGPU] Extend the SI Load/Store optimizer.

Rebased and added missing 'const' in SILoadStoreOptimizer::dmasksCanBeCombined.

Sep 2 2019, 6:59 AM · Restricted Project
piotr committed rG252a584cbd01: [AMDGPU] Add test (authored by piotr).
[AMDGPU] Add test
Sep 2 2019, 3:03 AM
piotr committed rL370634: [AMDGPU] Add test.
[AMDGPU] Add test
Sep 2 2019, 3:02 AM
piotr closed D63860: [AMDGPU] Add test.
Sep 2 2019, 3:02 AM · Restricted Project

Aug 30 2019

piotr retitled D63860: [AMDGPU] Add test from [MachineCSE] Extend CSE heuristic to [AMDGPU] Add test.
Aug 30 2019, 10:01 AM · Restricted Project
piotr updated the diff for D63860: [AMDGPU] Add test.

Reverting the patch and leaving the test only as it passes without the change in code.

Aug 30 2019, 9:58 AM · Restricted Project
piotr committed rG67b979466a61: [InstCombine][AMDGPU] Simplify tbuffer loads (authored by piotr).
[InstCombine][AMDGPU] Simplify tbuffer loads
Aug 30 2019, 7:21 AM
piotr committed rL370475: [InstCombine][AMDGPU] Simplify tbuffer loads.
[InstCombine][AMDGPU] Simplify tbuffer loads
Aug 30 2019, 7:21 AM
piotr closed D66926: [InstCombine][AMDGPU] Simplify tbuffer loads.
Aug 30 2019, 7:21 AM · Restricted Project
piotr added a comment to D63860: [AMDGPU] Add test.

Thank you for bearing with me, I have been away.

Aug 30 2019, 6:02 AM · Restricted Project

Aug 29 2019

piotr added reviewers for D66926: [InstCombine][AMDGPU] Simplify tbuffer loads: arsenm, nhaehnle.
Aug 29 2019, 12:06 AM · Restricted Project
piotr created D66926: [InstCombine][AMDGPU] Simplify tbuffer loads.
Aug 29 2019, 12:06 AM · Restricted Project

Jul 29 2019

piotr added a comment to D64911: [AMDGPU] Extend the SI Load/Store optimizer.

Ping

Jul 29 2019, 7:36 AM · Restricted Project
piotr added a comment to D63860: [AMDGPU] Add test.

Ping

Jul 29 2019, 7:36 AM · Restricted Project

Jul 19 2019

piotr added a comment to D64911: [AMDGPU] Extend the SI Load/Store optimizer.

I think a generic pass would not be suited for our image instructions, due to the dmask special treatment. I had decided to extend the si-load-store-opt pass, because similar transformations were already handled there.

Jul 19 2019, 8:15 AM · Restricted Project
piotr updated the diff for D64911: [AMDGPU] Extend the SI Load/Store optimizer.

V2: Addressed review comments.

Jul 19 2019, 7:09 AM · Restricted Project
piotr updated the diff for D63860: [AMDGPU] Add test.

V5: Refactored code.

Jul 19 2019, 5:05 AM · Restricted Project

Jul 18 2019

piotr added a reviewer for D64911: [AMDGPU] Extend the SI Load/Store optimizer: nhaehnle.
Jul 18 2019, 4:01 AM · Restricted Project
piotr created D64911: [AMDGPU] Extend the SI Load/Store optimizer.
Jul 18 2019, 4:01 AM · Restricted Project
piotr retitled D63860: [AMDGPU] Add test from [MachineCSE] Improve CSE on phi node incoming value to [MachineCSE] Extend CSE heuristic.
Jul 18 2019, 3:22 AM · Restricted Project
piotr updated the diff for D63860: [AMDGPU] Add test.

V4: Simplified the condition and test.

Jul 18 2019, 3:20 AM · Restricted Project

Jul 10 2019

piotr added inline comments to D63860: [AMDGPU] Add test.
Jul 10 2019, 5:04 AM · Restricted Project
piotr added a comment to D63860: [AMDGPU] Add test.

Hi @piotr, have you tried regression benchmarking with this patch? For instance, test suite.

Hi Anton, Yes, I have just run regression testing on the test-suite and the tests pass ("Expected Passes : 917").

Hmm, I'm talking about performance benchmarking, from llvm Test Suite: https://llvm.org/docs/TestingGuide.html#test-suite.

Jul 10 2019, 5:01 AM · Restricted Project

Jul 8 2019

piotr added a comment to D63860: [AMDGPU] Add test.

Hi @piotr, have you tried regression benchmarking with this patch? For instance, test suite.

Jul 8 2019, 4:27 PM · Restricted Project

Jul 2 2019

piotr updated the summary of D63860: [AMDGPU] Add test.
Jul 2 2019, 3:29 AM · Restricted Project
piotr updated the diff for D63860: [AMDGPU] Add test.

V3: Updating commit message.

Jul 2 2019, 3:29 AM · Restricted Project
piotr added inline comments to D63860: [AMDGPU] Add test.
Jul 2 2019, 3:24 AM · Restricted Project
piotr updated the diff for D63860: [AMDGPU] Add test.

V2: Removing condition that BB has to contain phis only. Added a MIR test.

Jul 2 2019, 3:20 AM · Restricted Project

Jun 27 2019

piotr added a comment to D63860: [AMDGPU] Add test.

Related to D63709.

Jun 27 2019, 1:35 AM · Restricted Project
piotr added a reviewer for D63860: [AMDGPU] Add test: arsenm.
Jun 27 2019, 1:34 AM · Restricted Project
piotr created D63860: [AMDGPU] Add test.
Jun 27 2019, 1:31 AM · Restricted Project