msearles (Mark Searles)
User

Projects

User does not belong to any projects.

User Details

User Since
Mar 7 2017, 11:02 AM (66 w, 5 d)

Recent Activity

Fri, Jun 15

msearles accepted D48246: [AMDGPU] setcc (select cc, CT, CF), CF, eq | ne -> xor cc, -1 | cc.

LGTM

Fri, Jun 15, 5:10 PM
msearles accepted D48223: DAG combine "and|or (select c, -1, 0), x" -> "select c, x, 0|-1".

LGTM

Fri, Jun 15, 12:37 PM

Wed, Jun 13

msearles accepted D48133: [AMDGPU] Corrected computeKnownBits for V_PERM_B32.

LGTM

Wed, Jun 13, 9:52 AM

Tue, Jun 12

msearles added a comment to D48099: [AMDGPU] DAG combine to produce V_PERM_B32.

LGTM

Tue, Jun 12, 4:43 PM
msearles accepted D48099: [AMDGPU] DAG combine to produce V_PERM_B32.

LGTM; couple of minor nits on couple of comments

Tue, Jun 12, 3:48 PM

Mon, Jun 11

msearles committed rL334459: [AMDGPU] prevent hitting Assertion `isReg() && "Wrong MachineOperand accessor"'.
[AMDGPU] prevent hitting Assertion `isReg() && "Wrong MachineOperand accessor"'
Mon, Jun 11, 5:45 PM
msearles closed D48047: [AMDGPU] findMaskOperands() - prevent hitting Assertion `isReg() && "Wrong MachineOperand accessor"' .
Mon, Jun 11, 5:45 PM · Restricted Project
msearles added inline comments to D48047: [AMDGPU] findMaskOperands() - prevent hitting Assertion `isReg() && "Wrong MachineOperand accessor"' .
Mon, Jun 11, 1:37 PM · Restricted Project
msearles updated the diff for D48047: [AMDGPU] findMaskOperands() - prevent hitting Assertion `isReg() && "Wrong MachineOperand accessor"' .

Removed # REQUIRES: asserts in test

Mon, Jun 11, 1:37 PM · Restricted Project
msearles updated the diff for D48047: [AMDGPU] findMaskOperands() - prevent hitting Assertion `isReg() && "Wrong MachineOperand accessor"' .

Added a test

Mon, Jun 11, 1:22 PM · Restricted Project
msearles created D48047: [AMDGPU] findMaskOperands() - prevent hitting Assertion `isReg() && "Wrong MachineOperand accessor"' .
Mon, Jun 11, 1:11 PM · Restricted Project

Mon, Jun 4

msearles committed rL333926: [AMDGPU][Waitcnt] Fix handling of flat instrs.
[AMDGPU][Waitcnt] Fix handling of flat instrs
Mon, Jun 4, 9:56 AM
msearles closed D46616: [AMDGPU][Waitcnt] Fix handling of flat instrs.
Mon, Jun 4, 9:56 AM · Restricted Project

Wed, May 30

msearles committed rL333559: [AMDGPU][Waitcnt] Fix build error: unused variable 'SWaitInst'.
[AMDGPU][Waitcnt] Fix build error: unused variable 'SWaitInst'
Wed, May 30, 9:32 AM
msearles committed rL333556: [AMDGPU][Waitcnt] Fix handling of loops with many bottom blocks.
[AMDGPU][Waitcnt] Fix handling of loops with many bottom blocks
Wed, May 30, 8:51 AM
msearles closed D47488: [AMDGPU][Waitcnt] Fix handling of loops with many bottom blocks.
Wed, May 30, 8:51 AM · Restricted Project

Tue, May 29

msearles created D47488: [AMDGPU][Waitcnt] Fix handling of loops with many bottom blocks.
Tue, May 29, 10:39 AM · Restricted Project

Fri, May 25

msearles committed rL333303: [AMDGPU][Waitcnt] Remove obsolete waitcnt option.
[AMDGPU][Waitcnt] Remove obsolete waitcnt option
Fri, May 25, 1:28 PM
msearles closed D47378: [AMDGPU][Waitcnt] Remove obsolete waitcnt option.
Fri, May 25, 1:28 PM · Restricted Project
msearles created D47378: [AMDGPU][Waitcnt] Remove obsolete waitcnt option.
Fri, May 25, 10:40 AM · Restricted Project

Thu, May 24

msearles added a comment to D46616: [AMDGPU][Waitcnt] Fix handling of flat instrs.

ping

Thu, May 24, 3:38 PM · Restricted Project

Mon, May 21

msearles committed rC332854: [Clang Tablegen] Add llvm_unreachable() to getModifierName().
[Clang Tablegen] Add llvm_unreachable() to getModifierName()
Mon, May 21, 10:33 AM
msearles committed rL332854: [Clang Tablegen] Add llvm_unreachable() to getModifierName().
[Clang Tablegen] Add llvm_unreachable() to getModifierName()
Mon, May 21, 10:33 AM
msearles closed D47150: [Clang Tablegen] Add llvm_unreachable() to getModifierName().
Mon, May 21, 10:33 AM
msearles updated the diff for D47150: [Clang Tablegen] Add llvm_unreachable() to getModifierName().

Add comment per reviewer suggestion.

Mon, May 21, 10:25 AM
msearles updated the diff for D47150: [Clang Tablegen] Add llvm_unreachable() to getModifierName().

Ah, yes. Adding llvm_unreachable() is better. Done.

Mon, May 21, 10:05 AM
msearles created D47150: [Clang Tablegen] Add llvm_unreachable() to getModifierName().
Mon, May 21, 9:47 AM

May 17 2018

msearles added a comment to D46616: [AMDGPU][Waitcnt] Fix handling of flat instrs.

ping

May 17 2018, 9:28 PM · Restricted Project

May 10 2018

msearles updated the diff for D46616: [AMDGPU][Waitcnt] Fix handling of flat instrs.

Add subtarget-specific checks

May 10 2018, 8:25 AM · Restricted Project

May 8 2018

msearles updated the diff for D46616: [AMDGPU][Waitcnt] Fix handling of flat instrs.

Update comment as suggested by reviewer.

May 8 2018, 6:49 PM · Restricted Project
msearles created D46616: [AMDGPU][Waitcnt] Fix handling of flat instrs.
May 8 2018, 5:43 PM · Restricted Project

May 7 2018

msearles committed rL331641: [AMDGPU][Waitcnt] Remove the old waitcnt pass.
[AMDGPU][Waitcnt] Remove the old waitcnt pass
May 7 2018, 7:47 AM
msearles closed D46448: [AMDGPU][Waitcnt] Remove the old waitcnt pass, which is no longer maintained and getting crufty.
May 7 2018, 7:47 AM · Restricted Project

May 4 2018

msearles added a comment to D46448: [AMDGPU][Waitcnt] Remove the old waitcnt pass, which is no longer maintained and getting crufty.

Please also remove it from AMDGPU.h. Otherwise LGTM.

May 4 2018, 3:26 PM · Restricted Project
msearles updated the diff for D46448: [AMDGPU][Waitcnt] Remove the old waitcnt pass, which is no longer maintained and getting crufty.

Remove old waitcnt pass from lib/Target/AMDGPU/AMDGPU.h

May 4 2018, 3:26 PM · Restricted Project
msearles added reviewers for D46448: [AMDGPU][Waitcnt] Remove the old waitcnt pass, which is no longer maintained and getting crufty: arsenm, rampitec.
May 4 2018, 3:13 PM · Restricted Project
msearles updated the diff for D46448: [AMDGPU][Waitcnt] Remove the old waitcnt pass, which is no longer maintained and getting crufty.

Add waitcnt.mir

May 4 2018, 1:10 PM · Restricted Project
msearles created D46448: [AMDGPU][Waitcnt] Remove the old waitcnt pass, which is no longer maintained and getting crufty.
May 4 2018, 12:52 PM · Restricted Project

May 1 2018

msearles added inline comments to D46298: AMDGPU: Move isSDNodeSourceOfDivergence() implementation to SITargetLowering.
May 1 2018, 8:57 AM

Apr 27 2018

msearles committed rL331062: [AMDGPU][Waitcnt] Update a few tests to use default waitcnt pass (si-insert….
[AMDGPU][Waitcnt] Update a few tests to use default waitcnt pass (si-insert…
Apr 27 2018, 11:02 AM
msearles closed D46154: [AMDGPU][Waitcnt] Update a few lit tests to use the default waitcnt pass.
Apr 27 2018, 11:02 AM · Restricted Project

Apr 26 2018

msearles added reviewers for D46154: [AMDGPU][Waitcnt] Update a few lit tests to use the default waitcnt pass: cfang, FarhanaAleen.
Apr 26 2018, 4:42 PM · Restricted Project
msearles created D46154: [AMDGPU][Waitcnt] Update a few lit tests to use the default waitcnt pass.
Apr 26 2018, 4:41 PM · Restricted Project
msearles committed rL330954: [AMDGPU][Waitcnt] As of gfx7, VMEM operations do not increment the export….
[AMDGPU][Waitcnt] As of gfx7, VMEM operations do not increment the export…
Apr 26 2018, 9:14 AM
msearles closed D46067: [AMDGPU][Waitcnt] Take ISA target into account for s_waitcnt expcnt instr generation.
Apr 26 2018, 9:14 AM · Restricted Project

Apr 25 2018

msearles committed rL330862: [AMDGPU] Waitcnt pass: add debug options.
[AMDGPU] Waitcnt pass: add debug options
Apr 25 2018, 12:26 PM
msearles closed D45888: [AMDGPU][Waitcnt] Add debug options.
Apr 25 2018, 12:26 PM · Restricted Project
msearles created D46067: [AMDGPU][Waitcnt] Take ISA target into account for s_waitcnt expcnt instr generation.
Apr 25 2018, 8:53 AM · Restricted Project

Apr 24 2018

msearles closed D45973: [AMDGPU][Waitcnt] NFC. Cleanup some code.

https://reviews.llvm.org/rL330730

Apr 24 2018, 9:07 AM · Restricted Project
msearles committed rL330730: [AMDGPU][Waitcnt] NFC. Cleanup some code/naming consistency:.
[AMDGPU][Waitcnt] NFC. Cleanup some code/naming consistency:
Apr 24 2018, 9:03 AM

Apr 23 2018

msearles created D45973: [AMDGPU][Waitcnt] NFC. Cleanup some code.
Apr 23 2018, 10:05 AM · Restricted Project
msearles updated the diff for D45888: [AMDGPU][Waitcnt] Add debug options.

To simplify review, remove unrelated code cleanup: naming consistency s/SWaitcnt/Waitcnt s/WaitCnt/Waitcnt

Apr 23 2018, 9:19 AM · Restricted Project

Apr 20 2018

msearles created D45888: [AMDGPU][Waitcnt] Add debug options.
Apr 20 2018, 9:55 AM · Restricted Project

Apr 19 2018

msearles committed rL330337: [AMDGPU] Do not only rely on BB number when finding bottom loop.
[AMDGPU] Do not only rely on BB number when finding bottom loop
Apr 19 2018, 8:45 AM
msearles closed D43831: [AMDGPU] Do not only rely on BB number when finding bottom loop.
Apr 19 2018, 8:45 AM

Apr 17 2018

msearles added a comment to D43831: [AMDGPU] Do not only rely on BB number when finding bottom loop.

Ping; the patch looks good to the original author and to me, the follow-on author. However, it would be nice if someone other than one of us accepts the revision.

Apr 17 2018, 12:02 PM

Apr 10 2018

msearles updated the diff for D43831: [AMDGPU] Do not only rely on BB number when finding bottom loop.
  • Change isLoopBottom() to return true in the case of a single basic block loop
  • Clear LoopWaitcntDataMap in between functions
Apr 10 2018, 9:15 AM
msearles commandeered D43831: [AMDGPU] Do not only rely on BB number when finding bottom loop.

Picking up where @hakzsam left off

Apr 10 2018, 9:13 AM

Apr 7 2018

msearles added inline comments to D43831: [AMDGPU] Do not only rely on BB number when finding bottom loop.
Apr 7 2018, 7:14 AM

Mar 27 2018

msearles accepted D44940: AMDGPU: Fix crash when MachinePointerInfo invalid.

LGTM

Mar 27 2018, 9:01 AM

Mar 15 2018

msearles added inline comments to D43831: [AMDGPU] Do not only rely on BB number when finding bottom loop.
Mar 15 2018, 8:50 AM

Mar 14 2018

msearles committed rL327583: [AMDGPU] Waitcnt pass: Modify the waitcnt pass to propagate info in the case of….
[AMDGPU] Waitcnt pass: Modify the waitcnt pass to propagate info in the case of…
Mar 14 2018, 3:07 PM
msearles closed D44434: [AMDGPU] Waitcnt pass: propagate info in the case of a single basic block loop.
Mar 14 2018, 3:07 PM · Restricted Project
msearles accepted D44319: [AMDGPU]Supported ds_write_b128 generation..

LGTM

Mar 14 2018, 2:55 PM
msearles updated the diff for D44434: [AMDGPU] Waitcnt pass: propagate info in the case of a single basic block loop.

Fix formatting per reviewer comments

Mar 14 2018, 11:50 AM · Restricted Project
msearles added inline comments to D43831: [AMDGPU] Do not only rely on BB number when finding bottom loop.
Mar 14 2018, 9:35 AM
msearles updated the diff for D44434: [AMDGPU] Waitcnt pass: propagate info in the case of a single basic block loop.

Adjust per reviewer comments.

Mar 14 2018, 9:32 AM · Restricted Project

Mar 13 2018

msearles added reviewers for D44434: [AMDGPU] Waitcnt pass: propagate info in the case of a single basic block loop: arsenm, rampitec.
Mar 13 2018, 8:53 AM · Restricted Project
msearles created D44434: [AMDGPU] Waitcnt pass: propagate info in the case of a single basic block loop.
Mar 13 2018, 8:53 AM · Restricted Project

Mar 8 2018

msearles added inline comments to D43831: [AMDGPU] Do not only rely on BB number when finding bottom loop.
Mar 8 2018, 6:42 PM

Mar 5 2018

msearles added a comment to D43831: [AMDGPU] Do not only rely on BB number when finding bottom loop.

For the testcase, would you also run -instnamer to rename instances of %<number>? You may want to combine it with other clean-up options, something like this: opt -S -deadarghaX0r -strip -strip-debug -strip-dead-prototypes -instnamer

Mar 5 2018, 10:14 AM

Feb 23 2018

msearles accepted D43699: [AMDGPU] Shrinking V_SUBBREV_U32.

LGTM

Feb 23 2018, 3:12 PM

Feb 19 2018

msearles closed D43426: [AMDGPU] Make note of exisitng waitcnt instrs.

https://reviews.llvm.org/rL325524

Feb 19 2018, 11:25 AM · Restricted Project
msearles committed rL325524: [AMDGPU] Make note of existing waitcnt instrs; this is add-on work related to….
[AMDGPU] Make note of existing waitcnt instrs; this is add-on work related to…
Feb 19 2018, 11:24 AM
msearles committed rL325518: [AMDGPU] Increased vector length for global/constant loads..
[AMDGPU] Increased vector length for global/constant loads.
Feb 19 2018, 8:47 AM
msearles closed D43275: [AMDGPU]Increased vector length for global/constant loads. .
Feb 19 2018, 8:47 AM

Feb 16 2018

msearles created D43426: [AMDGPU] Make note of exisitng waitcnt instrs.
Feb 16 2018, 5:59 PM · Restricted Project

Feb 15 2018

msearles accepted D43350: [AMDGPU] Combine adjacent waitcounts in a single strongest wait.

LGTM

Feb 15 2018, 1:23 PM

Feb 8 2018

msearles accepted D43094: AMDGPU: Enable PIC by default for amdgcn.

LGTM

Feb 8 2018, 3:42 PM

Feb 6 2018

msearles accepted D42997: [AMDGPU] Fixed wait count reuse.

LGTM; what about a test that exercises the code that was effectively turned off (~line 1129)?

Feb 6 2018, 6:27 PM
msearles committed rL324440: [AMDGPU] Suppress redundant waitcnt instrs..
[AMDGPU] Suppress redundant waitcnt instrs.
Feb 6 2018, 6:23 PM
msearles closed D42854: [AMDGPU] Suppress redundant waitcnt instrs.
Feb 6 2018, 6:23 PM · Restricted Project
msearles added inline comments to D42854: [AMDGPU] Suppress redundant waitcnt instrs.
Feb 6 2018, 5:57 PM · Restricted Project
msearles updated the diff for D42854: [AMDGPU] Suppress redundant waitcnt instrs.

Adjust per reviewer comment

Feb 6 2018, 5:57 PM · Restricted Project
msearles updated the diff for D42854: [AMDGPU] Suppress redundant waitcnt instrs.
  • Don't remove existing waitcnt instrs; if a redundant is to be inserted, keep the existing waitcnt and don't insert the duplicate.
  • Fix mir test
Feb 6 2018, 4:08 PM · Restricted Project

Feb 4 2018

msearles planned changes to D42854: [AMDGPU] Suppress redundant waitcnt instrs.
Feb 4 2018, 12:04 PM · Restricted Project
msearles added inline comments to D42854: [AMDGPU] Suppress redundant waitcnt instrs.
Feb 4 2018, 12:03 PM · Restricted Project
msearles added a comment to D42854: [AMDGPU] Suppress redundant waitcnt instrs.

Can the pass update its internal state while walking the control flow to factor in the consequences of the original waitcnts? That way a decision as to whether a waitcnt is required will take into account these original waitcnts. This means the benefit is obtained regardless of whether the waitcnts are adjacent or separated (even in different BBs).

It seems that a separate pass could be done after the final waitcnts have been decided to collapse adjacent waitncts into a single one if possible. Or perhaps it would be better to postpone inserting the waitcnts until after the dataflow iteration has found a fixed point, at which time any original/deduced waitcnts can be merger if adjacent.

Feb 4 2018, 12:02 PM · Restricted Project

Feb 3 2018

msearles added a comment to D42854: [AMDGPU] Suppress redundant waitcnt instrs.

A concern is that you do not want to remove an original waitcnt when inserting a new one, as the pass may iterate and subsequently decide not to add a waitcnt there, but will have eliminated a waitcnt needed to implement the memory model. Is that an issue?

Feb 3 2018, 9:44 AM · Restricted Project

Feb 2 2018

msearles added reviewers for D42854: [AMDGPU] Suppress redundant waitcnt instrs: kzhuravl, rampitec.
Feb 2 2018, 10:18 AM · Restricted Project
msearles created D42854: [AMDGPU] Suppress redundant waitcnt instrs.
Feb 2 2018, 10:18 AM · Restricted Project

Jan 30 2018

msearles committed rL323791: [AMDGPU] Revert "[AMDGPU] Add options for waitcnt pass debugging; add instr….
[AMDGPU] Revert "[AMDGPU] Add options for waitcnt pass debugging; add instr…
Jan 30 2018, 9:20 AM
msearles committed rL323788: [AMDGPU] Add options for waitcnt pass debugging; add instr count in debug….
[AMDGPU] Add options for waitcnt pass debugging; add instr count in debug…
Jan 30 2018, 8:52 AM
msearles closed D40091: [AMDGPU] Add options for waitcnt pass debugging; add instr count in debug output..
Jan 30 2018, 8:52 AM

Jan 22 2018

msearles committed rL323153: [AMDGPU] SI Load Store Optimizer: When merging with offset, use….
[AMDGPU] SI Load Store Optimizer: When merging with offset, use…
Jan 22 2018, 1:48 PM
msearles closed D42124: SI Load Store Optimizer: When merging with offset, use V_ADD_{I|U}32_e64.
Jan 22 2018, 1:48 PM · Restricted Project
msearles added inline comments to D42124: SI Load Store Optimizer: When merging with offset, use V_ADD_{I|U}32_e64.
Jan 22 2018, 12:34 PM · Restricted Project
msearles updated the diff for D42124: SI Load Store Optimizer: When merging with offset, use V_ADD_{I|U}32_e64.
  • Use scalar mov, not vector mov
  • Add test to merge-load-store-vreg.mir for nocarryadd opcodes
  • Update ds-combine-large-stride.ll . Note that the immediate is not folded if not using nocarryadd opcodes; this is a limitation in foldImmediates() and will be addressed in a follow-on patch
Jan 22 2018, 12:32 PM · Restricted Project

Jan 16 2018

msearles planned changes to D42124: SI Load Store Optimizer: When merging with offset, use V_ADD_{I|U}32_e64.

test/CodeGen/AMDGPU/ds-combine-large-stride.ll is failing with this patch; likely a test update; no review needed until I resolve this.

Jan 16 2018, 2:16 PM · Restricted Project
msearles created D42124: SI Load Store Optimizer: When merging with offset, use V_ADD_{I|U}32_e64.
Jan 16 2018, 12:16 PM · Restricted Project