qcolombet (Quentin Colombet)
User

Projects

User does not belong to any projects.

User Details

User Since
Dec 17 2012, 10:03 AM (270 w, 2 d)

Recent Activity

Yesterday

qcolombet accepted D43042: [MachineOperand][Target] MachineOperand::isRenamable semantics changes.

Hi Geoff,

Wed, Feb 21, 1:03 PM
qcolombet added inline comments to D43444: [AArch64][GlobalISel] When copying from a gpr32 to an fpr16 reg, convert to fpr32 first.
Wed, Feb 21, 10:50 AM

Fri, Feb 16

qcolombet added a comment to D43042: [MachineOperand][Target] MachineOperand::isRenamable semantics changes.

FYI, I've revert the original commit in r325421.

Fri, Feb 16, 7:15 PM
qcolombet committed rL325421: Revert "[MachineCopyPropagation] Extend pass to do COPY source forwarding".
Revert "[MachineCopyPropagation] Extend pass to do COPY source forwarding"
Fri, Feb 16, 7:08 PM
qcolombet accepted D43270: [GISel][Tablegen]: Make GlobalISelEmitter rule prioritization similar to that of selectionDAG.

Hi Aditya,

Fri, Feb 16, 1:24 PM
qcolombet accepted D43253: bitcode support change for fast flags compatibility.

Thanks all for the great feedbacks!

Fri, Feb 16, 11:04 AM
qcolombet accepted D43310: [AArch64][GlobalISel] Fix an assert fail/miscompile when fp16 types are copied to GPR register banks.

Hi Amara,

Fri, Feb 16, 9:26 AM

Tue, Feb 13

qcolombet accepted D43206: [GISel]: Add pattern matcher for FMUL and FADD.
Tue, Feb 13, 10:02 AM
qcolombet added a comment to D43042: [MachineOperand][Target] MachineOperand::isRenamable semantics changes.

(though it may require the interface to be changed slightly since we'll need the opcode)

Tue, Feb 13, 9:45 AM

Mon, Feb 12

qcolombet accepted D43090: GlobalISel: IRTranslate llvm.fmuladd.* intrinsic.
Mon, Feb 12, 1:58 PM

Fri, Feb 9

qcolombet added inline comments to D43090: GlobalISel: IRTranslate llvm.fmuladd.* intrinsic.
Fri, Feb 9, 5:52 PM
qcolombet added inline comments to D43090: GlobalISel: IRTranslate llvm.fmuladd.* intrinsic.
Fri, Feb 9, 4:28 PM
qcolombet added a comment to D43042: [MachineOperand][Target] MachineOperand::isRenamable semantics changes.

Hi Geoff,

Fri, Feb 9, 2:37 PM
qcolombet added inline comments to D43090: GlobalISel: IRTranslate llvm.fmuladd.* intrinsic.
Fri, Feb 9, 11:27 AM

Thu, Feb 8

qcolombet added a comment to D39304: [IR] redefine 'reassoc' fast-math-flag and add 'trans' fast-math-flag.

BTW, while writing the RFC, I realized that we could potentially generated incorrect code if we were silently downgrade a post-r317488 bitcode with a pre-r317488 compiler. (I.e., running fast math optimizations whereas we only wanted reassoc)

Thu, Feb 8, 5:37 PM
qcolombet added a comment to D39304: [IR] redefine 'reassoc' fast-math-flag and add 'trans' fast-math-flag.

That's http://lists.llvm.org/pipermail/llvm-dev/2018-February/121114.html

Thu, Feb 8, 5:35 PM
qcolombet added a comment to D39304: [IR] redefine 'reassoc' fast-math-flag and add 'trans' fast-math-flag.

I think we have to create a new version of the IR since the bits changed meaning (we can't just flip 'on' new bits).

Thu, Feb 8, 4:56 PM
qcolombet added a comment to D39304: [IR] redefine 'reassoc' fast-math-flag and add 'trans' fast-math-flag.

Hi Sanjay,

Thu, Feb 8, 3:44 PM
qcolombet resigned from D41762: [DWARF] Incorrect prologue end line record..

For the review itself, I leave this to @aprantl. I was just providing insights on the shrink-wrapping question :).

Thu, Feb 8, 9:28 AM

Tue, Feb 6

qcolombet accepted D42984: GlobalISel: Always check operand types when executing match table.

LGTM

Tue, Feb 6, 4:43 PM

Mon, Feb 5

qcolombet added a comment to D42877: [GlobalISel] Print/Parse FailedISel MachineFunction property.

Hi Roman.

Mon, Feb 5, 10:17 AM

Fri, Feb 2

qcolombet accepted D42861: [ReleaseNotes] Add note for enabling GlobalISel for AArch64 -O0.

LGTM

Fri, Feb 2, 1:00 PM
qcolombet accepted D42860: [ReleaseNotes] Add note for the new -fexperimental-isel flag..

LGTM

Fri, Feb 2, 1:00 PM
qcolombet accepted D42832: [AArch64][GlobalISel] Use getRegClassForTypeOnBank() in selectCopy..

One question below.

Fri, Feb 2, 9:31 AM
qcolombet accepted D37775: Add a verifier test to check the access on both sides of COPY are the same.

LGTM, coordinate with the x86 changes and you should be good to go!

Fri, Feb 2, 9:27 AM

Thu, Feb 1

qcolombet added inline comments to D42667: SplitKit: Fix liveness recomputation in some remat cases..
Thu, Feb 1, 11:08 AM
qcolombet accepted D42697: [GlobalISel] Constrain the dest reg of IMPLICT_DEF.

While I was here, I notice that selectCopy was not using getRegClassForTypeOnBank. Could you double check this is intended?

Thu, Feb 1, 11:07 AM
qcolombet accepted D42667: SplitKit: Fix liveness recomputation in some remat cases..

LGTM with the nitpicks mentioned by Krzysztof.

Thu, Feb 1, 10:48 AM

Wed, Jan 31

qcolombet accepted D42103: [LSR] Don't force bases of foldable formulae to the final type..

LGTM

Wed, Jan 31, 10:15 AM

Tue, Jan 30

qcolombet added a comment to D42600: [WIP][Shrink-wrap]split restore point.

Just two remarks:

  • So far shrink-wrapping was solely an analysis pass, i.e., it didn't modify the input code
  • Ideally I'd like we postpone creating new blocks until we decided where we are going to insert the code, otherwise we'll end up with blocks to clean-up later on
Tue, Jan 30, 5:10 PM
qcolombet added a comment to D41585: [Greedy RegAlloc] Take into account the cost of local intervals when selecting split candidate..

Ah makes sense.

Tue, Jan 30, 9:56 AM
qcolombet added a comment to D42103: [LSR] Don't force bases of foldable formulae to the final type..

Let's leave the patch as is.

Tue, Jan 30, 9:53 AM

Mon, Jan 29

qcolombet accepted D42103: [LSR] Don't force bases of foldable formulae to the final type..

Alright, LGTM then.
Please also patches the two other places I mentioned assuming you can have a test case for them.

Mon, Jan 29, 4:16 PM
qcolombet committed rL323710: [RAFast] Don't dereference MBB::end.
[RAFast] Don't dereference MBB::end
Mon, Jan 29, 3:44 PM
qcolombet accepted D42565: [ARM][GISel] PR35965 Constrain RegClasses of nested instructions built from Dst Pattern.

Got a chat with Roman off-line and turns out the problem I mentioned is worked around by everything single target, so indeed no test cases!
LGTM then

Mon, Jan 29, 12:03 PM
qcolombet accepted D42607: [LoopStrengthReduce, x86] don't add cost for a cmp that will be macro-fused (PR35681).

Couple of nits on the test case.

Mon, Jan 29, 11:02 AM

Fri, Jan 26

qcolombet added inline comments to D42565: [ARM][GISel] PR35965 Constrain RegClasses of nested instructions built from Dst Pattern.
Fri, Jan 26, 5:17 PM
qcolombet added a comment to D42103: [LSR] Don't force bases of foldable formulae to the final type..

Basically, I am fine either way (with or without the cast), but we would need to be consistent.

Fri, Jan 26, 3:43 PM
qcolombet added a comment to D42103: [LSR] Don't force bases of foldable formulae to the final type..

Out-of-curiosity do you actually see codegen differences with that patch?

Fri, Jan 26, 3:40 PM
qcolombet added inline comments to D42103: [LSR] Don't force bases of foldable formulae to the final type..
Fri, Jan 26, 3:28 PM
qcolombet accepted D41835: [MachineCopyPropagation] Extend pass to do COPY source forwarding.

LGTM

Fri, Jan 26, 12:18 PM
qcolombet added inline comments to D42449: [MachineVerifier] Add check that renamable operands aren't reserved registers..
Fri, Jan 26, 12:11 PM

Thu, Jan 25

qcolombet accepted D42287: [GlobalISel][X86] Fixing failures after https://reviews.llvm.org/D37775.

LGTM with couple of nits below.

Thu, Jan 25, 6:10 PM
qcolombet accepted D42276: [Driver] Add an -fexperimental-isel driver option to enable/disable GlobalISel.

LGTM

Thu, Jan 25, 1:29 PM
qcolombet added a comment to D41762: [DWARF] Incorrect prologue end line record..

Hi Carlos,

Thu, Jan 25, 1:19 PM

Tue, Jan 23

qcolombet accepted D42402: [GISel]: Eliminate redundant copies b/w VRegs of same regclass at the end of InstructionSelection.

LGTM too. The code is probably simpler to understand like this and if compile-time turns out to be a problem we can revisit.
I wonder though if removing this copies upfront wouldn't have helped tablegen current matching table (because it does not look through COPYs right now).

Tue, Jan 23, 4:51 PM
qcolombet added a comment to D42402: [GISel]: Eliminate redundant copies b/w VRegs of same regclass at the end of InstructionSelection.

Sounds like a good thing to do.

Tue, Jan 23, 2:02 PM
qcolombet added inline comments to D42276: [Driver] Add an -fexperimental-isel driver option to enable/disable GlobalISel.
Tue, Jan 23, 10:50 AM

Jan 22 2018

qcolombet added a comment to D41876: Add a utility to reduce GlobalISel tests.

Could this be extended to reduce any isel related bug?

Yes, it can be extended to reduce the other ISel problems by adding a regex and the specific arguments for llc.

I'm guessing the answer is no since you'd have to update the regex as well, but would it make sense to allow llc arguments to be passed through from the command line or something like that?

I thought we could have a regex, argument list pair for each error type.

Jan 22 2018, 2:25 PM
qcolombet added a comment to D41762: [DWARF] Incorrect prologue end line record..
But that does not imply that there is no correspondence to user code before the frame setup is complete.
Jan 22 2018, 10:13 AM

Jan 19 2018

qcolombet added inline comments to D41835: [MachineCopyPropagation] Extend pass to do COPY source forwarding.
Jan 19 2018, 2:46 PM
qcolombet added a comment to D42287: [GlobalISel][X86] Fixing failures after https://reviews.llvm.org/D37775.

Should I submit autogenerated tests that are not FP/G_COPY/G_TRUNC related, but that still need to be updated for D37775 as a separate patch? Should I have a review for it or can I just submit?

Jan 19 2018, 2:40 PM
qcolombet added a comment to D42287: [GlobalISel][X86] Fixing failures after https://reviews.llvm.org/D37775.

Hi Alexander,

Jan 19 2018, 1:32 PM
qcolombet requested changes to D37775: Add a verifier test to check the access on both sides of COPY are the same.
Jan 19 2018, 12:16 PM
qcolombet added inline comments to D42276: [Driver] Add an -fexperimental-isel driver option to enable/disable GlobalISel.
Jan 19 2018, 10:17 AM

Jan 18 2018

qcolombet added a comment to D41373: [GISel][RFC]: GlobalISel Combiner prototype.

Mostly looks good to me.
We probably want to make two separate patches:

  • One with the pattern matching capabilities
  • One with the new pass
Jan 18 2018, 5:57 PM
qcolombet added a comment to D40308: [RegisterCoalescer] More fixes for subreg join failure in RegCoalescer.

Hi David,

Jan 18 2018, 1:40 PM
qcolombet added a comment to D42103: [LSR] Don't force bases of foldable formulae to the final type..

Could you upload the patch with the full context?

Jan 18 2018, 1:31 PM

Jan 17 2018

qcolombet added a comment to D41835: [MachineCopyPropagation] Extend pass to do COPY source forwarding.

Hi Geoff,

Jan 17 2018, 3:03 PM
qcolombet accepted D42137: Add a TargetOption to enable/disable GlobalISel.

LGTM

Jan 17 2018, 1:49 PM

Jan 12 2018

qcolombet added a comment to D42012: [GlobalISel][TableGen] Add support for SDNodeXForm.

Looks mostly good to me.

Jan 12 2018, 4:48 PM
qcolombet added inline comments to D42012: [GlobalISel][TableGen] Add support for SDNodeXForm.
Jan 12 2018, 4:11 PM
qcolombet accepted D40511: [AArch64] Fix scheduling resources for post indexed loads and stores.

I don't expect you have a test case that exposed the problem, right?
(In particular for Cyclone at least WriteAdr is the same as WriteI).

Jan 12 2018, 9:46 AM

Jan 10 2018

qcolombet added a comment to D41886: [CodeGen][ShrinkWrapping] NFC: Replace isIrreducibleCFG by generic implementation..

Hi Diego,

Jan 10 2018, 9:47 AM
qcolombet added inline comments to D41765: [CodeGen] Provide an advanced shrink-wrapping interface.
Jan 10 2018, 9:38 AM
qcolombet accepted D41585: [Greedy RegAlloc] Take into account the cost of local intervals when selecting split candidate..

Hi Marina,

Jan 10 2018, 9:32 AM

Jan 9 2018

qcolombet added a comment to D41876: Add a utility to reduce GlobalISel tests.

Hi Volkan,

Jan 9 2018, 1:56 PM

Jan 5 2018

qcolombet added a reviewer for D41638: [llvm-extract] Support extracting basic blocks: bogner.
Jan 5 2018, 11:38 AM
qcolombet requested changes to D41098: [InlineSpiller] Fix a crash due to lack of forward progress from remat.
Jan 5 2018, 10:07 AM

Jan 4 2018

qcolombet accepted D41738: [GISel]: Don't create G_MUL with 1 during translation of GetElementPtr.
Jan 4 2018, 4:42 PM

Dec 20 2017

qcolombet added a comment to D41362: [AArch64][GlobalISel] Enable GlobalISel at -O0 by default.

How about checking for EnableGlobalISelAbort.getNumOccurences() > 0 to check for an command line override, otherwise doing what the target wants via the isGlobalISelAbortEnabled() override?

Dec 20 2017, 11:45 PM
qcolombet added a comment to D41362: [AArch64][GlobalISel] Enable GlobalISel at -O0 by default.

Sounds like this option doesn’t use the fallback path.
I thought it did.
Something to double check.

Dec 20 2017, 3:23 PM

Dec 19 2017

qcolombet added a comment to D41373: [GISel][RFC]: GlobalISel Combiner prototype.

One more comment: Do you have an idea how we could make sure that the combiners are making progress?
Basically, I would like we think about two cases that bit us in the past:

  1. Two combines undo each others
  2. Two combines keep piling up (like expand op1 into op2 & op3 and expand op2 into op1 & op3 would keep having the IR growing)
Dec 19 2017, 3:44 PM
qcolombet added a comment to D41373: [GISel][RFC]: GlobalISel Combiner prototype.

Hi Aditya,

Dec 19 2017, 3:41 PM

Dec 18 2017

qcolombet committed rL321053: [TableGen][GlobalISel] Reset the internal map of RuleMatchers just before the….
[TableGen][GlobalISel] Reset the internal map of RuleMatchers just before the…
Dec 18 2017, 6:58 PM
qcolombet added a comment to D41362: [AArch64][GlobalISel] Enable GlobalISel at -O0 by default.

Sounds like this option doesn’t use the fallback path.
I thought it did.
Something to double check.

Dec 18 2017, 3:40 PM
qcolombet accepted D41362: [AArch64][GlobalISel] Enable GlobalISel at -O0 by default.

Hi Amara,

Dec 18 2017, 2:59 PM
qcolombet committed rL321031: [TableGen][GlobalISel] Make the arguments of the Instruction and Operand….
[TableGen][GlobalISel] Make the arguments of the Instruction and Operand…
Dec 18 2017, 2:13 PM
qcolombet committed rL321025: [TableGen][GlobalISel] Refactor optimizeRules related bit to allow code reuse.
[TableGen][GlobalISel] Refactor optimizeRules related bit to allow code reuse
Dec 18 2017, 1:26 PM
qcolombet committed rL321017: [TableGen][GlobalISel] Optimize MatchTable for faster instruction selection.
[TableGen][GlobalISel] Optimize MatchTable for faster instruction selection
Dec 18 2017, 11:48 AM
qcolombet closed D39034: [WIP][GlobalISel][TableGen] Optimize MatchTable for faster instruction selection.
Dec 18 2017, 11:48 AM

Dec 15 2017

qcolombet committed rL320890: [TableGen][GlobalISel] Make the different Matcher comparable.
[TableGen][GlobalISel] Make the different Matcher comparable
Dec 15 2017, 3:25 PM
qcolombet committed rL320889: [TableGen][GlobalISel] Fix unused variable warning in release mode.
[TableGen][GlobalISel] Fix unused variable warning in release mode
Dec 15 2017, 3:25 PM
qcolombet committed rL320887: [TableGen][GlobalISel] Have the predicate directly know which data they are….
[TableGen][GlobalISel] Have the predicate directly know which data they are…
Dec 15 2017, 3:08 PM

Dec 14 2017

qcolombet committed rL320767: [TableGen][GlobalISel] Add a common class for all PredicateMatcher.
[TableGen][GlobalISel] Add a common class for all PredicateMatcher
Dec 14 2017, 3:45 PM
qcolombet added a comment to D41098: [InlineSpiller] Fix a crash due to lack of forward progress from remat.

Hi Philip,

Dec 14 2017, 1:27 PM

Dec 13 2017

qcolombet added inline comments to D39034: [WIP][GlobalISel][TableGen] Optimize MatchTable for faster instruction selection.
Dec 13 2017, 4:33 PM
qcolombet updated the diff for D39034: [WIP][GlobalISel][TableGen] Optimize MatchTable for faster instruction selection.
Dec 13 2017, 4:33 PM
qcolombet accepted D41187: Ignore metainstructions during the shrink wrap analysis.

LGTM

Dec 13 2017, 10:41 AM

Dec 12 2017

qcolombet accepted D40049: [PATCH] Global reassociation for improved CSE.

LGTM

Dec 12 2017, 10:26 AM

Dec 11 2017

qcolombet added inline comments to D39034: [WIP][GlobalISel][TableGen] Optimize MatchTable for faster instruction selection.
Dec 11 2017, 2:47 PM
qcolombet accepted D29294: LSR: Check more intrinsic pointer operands.

LGTM for the generic part.

Dec 11 2017, 10:54 AM

Dec 8 2017

qcolombet added inline comments to D39034: [WIP][GlobalISel][TableGen] Optimize MatchTable for faster instruction selection.
Dec 8 2017, 1:56 PM
qcolombet updated the diff for D39034: [WIP][GlobalISel][TableGen] Optimize MatchTable for faster instruction selection.
  • Address nitpicks
  • Push InsnID and OpIdx in addPredicate
  • Bail on sharing when InsnID != 0 or OpIdx is specified
  • Make NumOperands a member variable instead of computed from the size of the list
Dec 8 2017, 1:55 PM

Dec 4 2017

qcolombet added inline comments to D39034: [WIP][GlobalISel][TableGen] Optimize MatchTable for faster instruction selection.
Dec 4 2017, 4:00 PM
qcolombet accepted D40092: [globalisel][irtranslator] Add support for atomicrmw and (strong) cmpxchg.

LGTM nitpicks below

Dec 4 2017, 1:06 PM
qcolombet accepted D40096: Allow similar TargetOpcodes to use inheritance to factor out commonality. NFC..

LGTM

Dec 4 2017, 12:49 PM
qcolombet added inline comments to D40049: [PATCH] Global reassociation for improved CSE.
Dec 4 2017, 11:43 AM
qcolombet added inline comments to D40049: [PATCH] Global reassociation for improved CSE.
Dec 4 2017, 11:32 AM
qcolombet accepted D40763: PeepholeOptimizer: Handle undef operands correctly.
Dec 4 2017, 10:21 AM