Page MenuHomePhabricator

tpr (Tim Renouf)
User

Projects

User does not belong to any projects.

User Details

User Since
Jan 26 2017, 7:07 AM (108 w, 1 d)

Recent Activity

Mon, Feb 11

tpr accepted D57737: [AMDGPU] Fix DPP sequence in atomic optimizer..

Ah, right, I see about the need for the exclusive and inclusive scan results.

Mon, Feb 11, 5:04 AM · Restricted Project, Restricted Project
tpr added a comment to D57737: [AMDGPU] Fix DPP sequence in atomic optimizer..

But I still don't understand it:

  1. Why do you want an exclusive scan? Surely what you're trying to do is just "sum" up all lanes into lane 63, which is an inclusive scan.
  2. Can't you do an exclusive scan with powers of 2 shifts like an inclusive scan, but just with the wf_sr1 on the front? (Although I think that gives the wrong answer due to (1)).
  3. Isn't the only thing wrong with this code before this fix that you forgot to put the bank masks on steps 2, 3 and 4? (Although you're correct to remove the unnecessary intermediate wwm intrinsic calls.)
Mon, Feb 11, 3:43 AM · Restricted Project, Restricted Project
tpr added a comment to D57737: [AMDGPU] Fix DPP sequence in atomic optimizer..

Oh, it's because you've switched to an exclusive scan.

Mon, Feb 11, 3:20 AM · Restricted Project, Restricted Project
tpr added a comment to D57737: [AMDGPU] Fix DPP sequence in atomic optimizer..

I don't understand this fix. Surely a reduction is done with just power of two shifts. Why do we need the shift by 3 as well? What is the extra wf_sr1 dpp at the start for?

Mon, Feb 11, 3:15 AM · Restricted Project, Restricted Project
tpr added inline comments to D57028: [AMDGPU] Added MsgPack format PAL metadata.
Mon, Feb 11, 1:57 AM · Restricted Project

Sun, Feb 10

tpr added inline comments to D57028: [AMDGPU] Added MsgPack format PAL metadata.
Sun, Feb 10, 11:04 AM · Restricted Project

Mon, Feb 4

tpr updated the diff for D57024: [AMDGPU] Switched HSA metadata to use MsgPackDocument.

V2: Addressed review comment.

Mon, Feb 4, 10:12 AM · Restricted Project
tpr added inline comments to D57027: [AMDGPU] Factored PAL metadata handling out into its own class.
Mon, Feb 4, 9:54 AM · Restricted Project
tpr added a comment to D57027: [AMDGPU] Factored PAL metadata handling out into its own class.

The code looks good to me, but I'm wondering why the move from doing this in AsmPrinter::EmitEndOfAsmFile to TargetStreamer::finish? It seems like it is fine to emit directives in either, but I don't understand why the change here.

Mon, Feb 4, 9:51 AM · Restricted Project

Tue, Jan 29

tpr accepted D56002: [AMDGPU] Fix a weird WWM intrinsic issue..

I was thinking it is a bit odd that we're having to handle WQM and WWM in this way even though they are target instructions. However I think it is ok because they act like the non-target COPY instruction in that they do not specify a register class for operands.

Tue, Jan 29, 3:21 AM · Restricted Project

Jan 23 2019

tpr updated the diff for D57023: [MsgPack] New MsgPackDocument class.

V2: Addressed review comments.

Jan 23 2019, 12:07 PM · Restricted Project
tpr committed rL351950: [AMDGPU] With XNACK, cannot clause a load with result coalesced with operand.
[AMDGPU] With XNACK, cannot clause a load with result coalesced with operand
Jan 23 2019, 5:38 AM
tpr closed D57008: [AMDGPU] With XNACK, cannot clause a load with result coalesced with operand.
Jan 23 2019, 5:38 AM

Jan 22 2019

tpr updated the diff for D57008: [AMDGPU] With XNACK, cannot clause a load with result coalesced with operand.

V2: Addressed review comments.

Jan 22 2019, 12:40 AM
tpr added inline comments to D57008: [AMDGPU] With XNACK, cannot clause a load with result coalesced with operand.
Jan 22 2019, 12:40 AM

Jan 21 2019

tpr added parent revisions for D57028: [AMDGPU] Added MsgPack format PAL metadata: D57027: [AMDGPU] Factored PAL metadata handling out into its own class, D57023: [MsgPack] New MsgPackDocument class.
Jan 21 2019, 9:49 AM · Restricted Project
tpr added a child revision for D57023: [MsgPack] New MsgPackDocument class: D57028: [AMDGPU] Added MsgPack format PAL metadata.
Jan 21 2019, 9:49 AM · Restricted Project
tpr added a child revision for D57027: [AMDGPU] Factored PAL metadata handling out into its own class: D57028: [AMDGPU] Added MsgPack format PAL metadata.
Jan 21 2019, 9:49 AM · Restricted Project
tpr added reviewers for D57028: [AMDGPU] Added MsgPack format PAL metadata: scott.linder, kzhuravl.
Jan 21 2019, 9:49 AM · Restricted Project
tpr created D57028: [AMDGPU] Added MsgPack format PAL metadata.
Jan 21 2019, 9:48 AM · Restricted Project
tpr added reviewers for D57027: [AMDGPU] Factored PAL metadata handling out into its own class: scott.linder, kzhuravl.
Jan 21 2019, 9:47 AM · Restricted Project
tpr created D57027: [AMDGPU] Factored PAL metadata handling out into its own class.
Jan 21 2019, 9:47 AM · Restricted Project
tpr added a parent revision for D57024: [AMDGPU] Switched HSA metadata to use MsgPackDocument: D57023: [MsgPack] New MsgPackDocument class.
Jan 21 2019, 9:44 AM · Restricted Project
tpr added a child revision for D57023: [MsgPack] New MsgPackDocument class: D57024: [AMDGPU] Switched HSA metadata to use MsgPackDocument.
Jan 21 2019, 9:44 AM · Restricted Project
tpr added a child revision for D57024: [AMDGPU] Switched HSA metadata to use MsgPackDocument: D57025: [MsgPack] Removed MsgPackTypes.
Jan 21 2019, 9:44 AM · Restricted Project
tpr added a parent revision for D57025: [MsgPack] Removed MsgPackTypes: D57024: [AMDGPU] Switched HSA metadata to use MsgPackDocument.
Jan 21 2019, 9:44 AM · Restricted Project
tpr added reviewers for D57025: [MsgPack] Removed MsgPackTypes: scott.linder, arsenm, t-tye.
Jan 21 2019, 9:42 AM · Restricted Project
tpr created D57025: [MsgPack] Removed MsgPackTypes.
Jan 21 2019, 9:42 AM · Restricted Project
tpr added reviewers for D57024: [AMDGPU] Switched HSA metadata to use MsgPackDocument: scott.linder, arsenm, t-tye.
Jan 21 2019, 9:38 AM · Restricted Project
tpr created D57024: [AMDGPU] Switched HSA metadata to use MsgPackDocument.
Jan 21 2019, 9:35 AM · Restricted Project
tpr added reviewers for D57023: [MsgPack] New MsgPackDocument class: scott.linder, dblaikie.
Jan 21 2019, 9:31 AM · Restricted Project
tpr created D57023: [MsgPack] New MsgPackDocument class.
Jan 21 2019, 9:29 AM · Restricted Project
tpr added reviewers for D57008: [AMDGPU] With XNACK, cannot clause a load with result coalesced with operand: rampitec, vpykhtin.
Jan 21 2019, 2:12 AM
tpr created D57008: [AMDGPU] With XNACK, cannot clause a load with result coalesced with operand.
Jan 21 2019, 2:09 AM

Nov 28 2018

tpr abandoned D54855: [AMDGPU] An exp must be branched over if exec=0.

Yes, sorry, I pushed this fix to phab without even checking that it applies to trunk. I had completely forgotten that you have already fixed this in trunk (D48431) and I reverted your fix in our local branch.

Nov 28 2018, 7:03 AM

Nov 23 2018

tpr added reviewers for D54855: [AMDGPU] An exp must be branched over if exec=0: nhaehnle, arsenm.
Nov 23 2018, 4:15 AM
tpr created D54855: [AMDGPU] An exp must be branched over if exec=0.
Nov 23 2018, 4:11 AM

Nov 16 2018

tpr abandoned D54516: [AMDGPU] Do not mark llvm.amdgcn.set.inactive as IntrNoMem.

For a quick fix for my specific problem, I have gone for the frontend hack, so I am abandoning this review.

Nov 16 2018, 10:05 AM
tpr added a comment to D54516: [AMDGPU] Do not mark llvm.amdgcn.set.inactive as IntrNoMem.

Oh, I've found it already in use in a different case in LLPC:

Nov 16 2018, 6:26 AM
tpr added a comment to D54516: [AMDGPU] Do not mark llvm.amdgcn.set.inactive as IntrNoMem.

EarlyCSE does seem to common up in this situation. And, if I disable that, I get GVN commoning it up.

Nov 16 2018, 6:17 AM
tpr added a reviewer for D54516: [AMDGPU] Do not mark llvm.amdgcn.set.inactive as IntrNoMem: sameerds.
Nov 16 2018, 3:59 AM
tpr added a comment to D54516: [AMDGPU] Do not mark llvm.amdgcn.set.inactive as IntrNoMem.

It's EarlyCSE, which seems to completely ignore Convergent. That's bad, right?

Nov 16 2018, 3:29 AM

Nov 14 2018

tpr added reviewers for D54516: [AMDGPU] Do not mark llvm.amdgcn.set.inactive as IntrNoMem: nhaehnle, cwabbott, rampitec.
Nov 14 2018, 12:52 AM
tpr created D54516: [AMDGPU] Do not mark llvm.amdgcn.set.inactive as IntrNoMem.
Nov 14 2018, 12:50 AM

Nov 5 2018

tpr added a reviewer for D54100: [LegacyPassManager] Add NL to -print-before/-print-after banner: chandlerc.
Nov 5 2018, 5:16 AM
tpr added reviewers for D54100: [LegacyPassManager] Add NL to -print-before/-print-after banner: fedor.sergeev, atrick.
Nov 5 2018, 5:15 AM
tpr created D54100: [LegacyPassManager] Add NL to -print-before/-print-after banner.
Nov 5 2018, 5:13 AM

Oct 31 2018

tpr added a comment to D52821: [Disassembler][llvm-readobj] ELF note dumper abstraction.

It would be good to have comments from other reviewers re the whole idea of having a TargetRegistry interface to allow readobj-style dumping of a target-specific note type to be inside the target.

Oct 31 2018, 6:51 AM

Oct 24 2018

tpr committed rC345198: Add gfx909 to GPU Arch.
Add gfx909 to GPU Arch
Oct 24 2018, 2:22 PM
tpr committed rL345198: Add gfx909 to GPU Arch.
Add gfx909 to GPU Arch
Oct 24 2018, 2:22 PM
tpr closed D53558: Add gfx909 to GPU Arch.
Oct 24 2018, 2:21 PM
tpr committed rL345120: [AMDGPU] Defined gfx909 Raven Ridge 2.
[AMDGPU] Defined gfx909 Raven Ridge 2
Oct 24 2018, 1:16 AM
tpr closed D53418: [AMDGPU] Defined gfx909 Raven Ridge 2.
Oct 24 2018, 1:16 AM

Oct 23 2018

tpr abandoned D53584: [Transforms/Utils] New -instanonymizer pass.

-strip actually does more, but will be fine for the application I need it for.

Oct 23 2018, 1:20 PM
tpr added reviewers for D53584: [Transforms/Utils] New -instanonymizer pass: mgorny, chandlerc.

Hi Michał and Chandler. I wasn't sure who to add as reviewers for this. If you can think of anyone more appropriate, please add them. Thanks.

Oct 23 2018, 9:32 AM
tpr created D53584: [Transforms/Utils] New -instanonymizer pass.
Oct 23 2018, 9:26 AM
tpr created D53558: Add gfx909 to GPU Arch.
Oct 23 2018, 1:26 AM
tpr added reviewers for D53558: Add gfx909 to GPU Arch: yaxunl, kzhuravl.
Oct 23 2018, 1:26 AM
tpr updated the diff for D53418: [AMDGPU] Defined gfx909 Raven Ridge 2.

V2: Addressed review comments.

Oct 23 2018, 12:54 AM

Oct 19 2018

tpr added reviewers for D53418: [AMDGPU] Defined gfx909 Raven Ridge 2: t-tye, kzhuravl.
Oct 19 2018, 12:31 AM
tpr created D53418: [AMDGPU] Defined gfx909 Raven Ridge 2.
Oct 19 2018, 12:29 AM

Oct 17 2018

tpr added a comment to D52821: [Disassembler][llvm-readobj] ELF note dumper abstraction.

This particular family of notes is target specific because it contains information on how the driver should program the hardware in order to be able to run the shaders in this graphics pipeline. So I think that justifies treating the notes as target specific.

Oct 17 2018, 10:28 PM

Oct 10 2018

tpr added a comment to D52823: [Disassembler] MCELFNoteDisassembler abstraction.

Ping.

Oct 10 2018, 10:52 PM
tpr added inline comments to D52821: [Disassembler][llvm-readobj] ELF note dumper abstraction.
Oct 10 2018, 10:51 PM

Oct 5 2018

tpr added inline comments to D52821: [Disassembler][llvm-readobj] ELF note dumper abstraction.
Oct 5 2018, 12:06 PM
tpr added inline comments to D52821: [Disassembler][llvm-readobj] ELF note dumper abstraction.
Oct 5 2018, 1:08 AM

Oct 3 2018

tpr committed rL343672: [AMDGPU] Fix for negative offsets in buffer/tbuffer intrinsics.
[AMDGPU] Fix for negative offsets in buffer/tbuffer intrinsics
Oct 3 2018, 3:31 AM
tpr closed D52683: [AMDGPU] Fix for negative offsets in buffer/tbuffer intrinsics.
Oct 3 2018, 3:31 AM
tpr added a comment to D52683: [AMDGPU] Fix for negative offsets in buffer/tbuffer intrinsics.

For the old intrinsics, there used to be code called from dag pattern matching to convert a single offset into some combination of voffset, soffset and instoffset. Is that what you're referring to?

Oct 3 2018, 2:58 AM
tpr updated subscribers of D52824: [AMDGPU] Implemented MCELFNoteDisassembler for PAL metadata note.
Oct 3 2018, 2:01 AM
tpr updated subscribers of D52823: [Disassembler] MCELFNoteDisassembler abstraction.
Oct 3 2018, 2:01 AM
tpr updated subscribers of D52822: [llvm-readobj][AMDGPU] Moved AMDGPU-specific note record dumping into target.
Oct 3 2018, 2:01 AM
tpr added a reviewer for D52823: [Disassembler] MCELFNoteDisassembler abstraction: kzhuravl.
Oct 3 2018, 2:01 AM
tpr added a reviewer for D52824: [AMDGPU] Implemented MCELFNoteDisassembler for PAL metadata note: kzhuravl.
Oct 3 2018, 2:01 AM
tpr updated subscribers of D52821: [Disassembler][llvm-readobj] ELF note dumper abstraction.
Oct 3 2018, 2:01 AM
tpr added a reviewer for D52822: [llvm-readobj][AMDGPU] Moved AMDGPU-specific note record dumping into target: kzhuravl.
Oct 3 2018, 2:01 AM
tpr added a reviewer for D52821: [Disassembler][llvm-readobj] ELF note dumper abstraction: kzhuravl.
Oct 3 2018, 1:58 AM
tpr added a reviewer for D52824: [AMDGPU] Implemented MCELFNoteDisassembler for PAL metadata note: scott.linder.

Requires D52823. Currently no in-tree user of this code.

Oct 3 2018, 1:51 AM
tpr added reviewers for D52823: [Disassembler] MCELFNoteDisassembler abstraction: pcc, asb, scott.linder.

See D52824 for the start of AMDGPU implementing this.

Oct 3 2018, 1:51 AM
tpr added reviewers for D52821: [Disassembler][llvm-readobj] ELF note dumper abstraction: pcc, asb, scott.linder, abdulras, compnerd.

See D52822 for AMDGPU implementing ELFNoteDumper.

Oct 3 2018, 1:48 AM
tpr added reviewers for D52822: [llvm-readobj][AMDGPU] Moved AMDGPU-specific note record dumping into target: scott.linder, abdulras, compnerd.

Requires D52821.

Oct 3 2018, 1:48 AM
tpr created D52824: [AMDGPU] Implemented MCELFNoteDisassembler for PAL metadata note.
Oct 3 2018, 1:32 AM
tpr created D52823: [Disassembler] MCELFNoteDisassembler abstraction.
Oct 3 2018, 1:30 AM
tpr created D52822: [llvm-readobj][AMDGPU] Moved AMDGPU-specific note record dumping into target.
Oct 3 2018, 1:29 AM
tpr created D52821: [Disassembler][llvm-readobj] ELF note dumper abstraction.
Oct 3 2018, 1:29 AM

Sep 29 2018

tpr added reviewers for D52683: [AMDGPU] Fix for negative offsets in buffer/tbuffer intrinsics: nhaehnle, mareko.
Sep 29 2018, 1:07 AM
tpr created D52683: [AMDGPU] Fix for negative offsets in buffer/tbuffer intrinsics.
Sep 29 2018, 1:04 AM

Sep 20 2018

tpr accepted D52291: AMDGPU: Future-proof {raw,struct}.buffer.atomic intrinsics.

LGTM

Sep 20 2018, 6:09 PM

Sep 15 2018

tpr added a comment to D52061: [RegisterCoalescer] Only look at main ranges in valuesIdentical/followCopyChain.

Hi Quentin

Sep 15 2018, 10:01 AM

Sep 14 2018

tpr added a comment to D51849: [RegisterCoalescer] Avoid "Use not jointly dominated by defs" in removePartialRedundancy.
  1. That test only fails without this fix if you have the fix in D51257.
Sep 14 2018, 3:41 PM
tpr updated the diff for D51849: [RegisterCoalescer] Avoid "Use not jointly dominated by defs" in removePartialRedundancy.

V2: mir test

Sep 14 2018, 3:35 PM
tpr added a comment to D51849: [RegisterCoalescer] Avoid "Use not jointly dominated by defs" in removePartialRedundancy.

Hmm, me too for trunk. It only fails on our fork that is 340950 plus local changes, so the code generated by instruction selection is probably different.

Sep 14 2018, 5:04 AM

Sep 13 2018

tpr committed rL342189: [AMDGPU] Removed unused method.
[AMDGPU] Removed unused method
Sep 13 2018, 2:59 PM
tpr closed D52022: [AMDGPU] Removed unused method.
Sep 13 2018, 2:59 PM
tpr added a comment to D51848: [RegisterCoalescer] Fixup "Fixed inconsistent followCopyChain with subreg".

This is now superseded by D52061. If that one is approved, I will abandon this one.

Sep 13 2018, 2:46 PM
tpr added reviewers for D52061: [RegisterCoalescer] Only look at main ranges in valuesIdentical/followCopyChain: qcolombet, MatzeB, kparzysz.
Sep 13 2018, 2:44 PM
tpr created D52061: [RegisterCoalescer] Only look at main ranges in valuesIdentical/followCopyChain.
Sep 13 2018, 2:41 PM
tpr added reviewers for D52022: [AMDGPU] Removed unused method: nhaehnle, mareko.
Sep 13 2018, 3:38 AM
tpr created D52022: [AMDGPU] Removed unused method.
Sep 13 2018, 3:37 AM

Sep 10 2018

tpr added reviewers for D51849: [RegisterCoalescer] Avoid "Use not jointly dominated by defs" in removePartialRedundancy: MatzeB, kparzysz, qcolombet.
Sep 10 2018, 12:44 AM