Page MenuHomePhabricator

RKSimon (Simon Pilgrim)
User

Projects

User does not belong to any projects.

User Details

User Since
May 5 2014, 7:26 AM (240 w, 3 d)

Recent Activity

Today

RKSimon committed rL349057: [X86][SSE] Add SSE vector imm/var shift support to….
[X86][SSE] Add SSE vector imm/var shift support to…
Thu, Dec 13, 8:43 AM
RKSimon committed rL349052: [X86][SSE] Fix all remaining modulo vector rotation amounts (PR38243).
[X86][SSE] Fix all remaining modulo vector rotation amounts (PR38243)
Thu, Dec 13, 7:53 AM
RKSimon committed rL349047: [X86][SSE] Fix modulo rotation amounts for v8i16/v16i16/v4i32 (PR38243).
[X86][SSE] Fix modulo rotation amounts for v8i16/v16i16/v4i32 (PR38243)
Thu, Dec 13, 7:26 AM
RKSimon committed rL349040: [X86][SSE] Merge the vXi16/vXi32 vector rotation expansion cases. NFCI..
[X86][SSE] Merge the vXi16/vXi32 vector rotation expansion cases. NFCI.
Thu, Dec 13, 6:54 AM
RKSimon added inline comments to D55600: [TargetLowering] Add ISD::OR + ISD::XOR handling to SimplifyDemandedVectorElts.
Thu, Dec 13, 6:30 AM
RKSimon committed rL349034: [X86][BWI] Don't custom lower vXi8 rotations..
[X86][BWI] Don't custom lower vXi8 rotations.
Thu, Dec 13, 5:47 AM
RKSimon committed rL349028: [DAGCombine] Moved X86 rotate_amount % bitwidth == 0 early out to DAGCombiner.
[DAGCombine] Moved X86 rotate_amount % bitwidth == 0 early out to DAGCombiner
Thu, Dec 13, 4:26 AM
RKSimon accepted D55603: test-suite: divide avx512 tests to instruction set subdirectories.

LGTM

Thu, Dec 13, 3:44 AM
RKSimon committed rL349025: [TargetLowering] Add ISD::ROTL/ROTR vector expansion.
[TargetLowering] Add ISD::ROTL/ROTR vector expansion
Thu, Dec 13, 3:24 AM
RKSimon added inline comments to D55600: [TargetLowering] Add ISD::OR + ISD::XOR handling to SimplifyDemandedVectorElts.
Thu, Dec 13, 2:21 AM
RKSimon accepted D55603: test-suite: divide avx512 tests to instruction set subdirectories.

LGTM

Thu, Dec 13, 12:49 AM

Yesterday

RKSimon updated the diff for D55600: [TargetLowering] Add ISD::OR + ISD::XOR handling to SimplifyDemandedVectorElts.

rebase

Wed, Dec 12, 1:37 PM
RKSimon accepted D55565: [X86] Don't emit MULX by default with BMI2.

LGTM

Wed, Dec 12, 1:11 PM
RKSimon committed rL348966: [X86] Added missing constant pool checks. NFCI..
[X86] Added missing constant pool checks. NFCI.
Wed, Dec 12, 11:59 AM
RKSimon committed rL348958: Fix Wdocumentation warning. NFCI..
Fix Wdocumentation warning. NFCI.
Wed, Dec 12, 11:05 AM
RKSimon committed rL348956: [X86] Added missing constant pool checks. NFCI..
[X86] Added missing constant pool checks. NFCI.
Wed, Dec 12, 10:56 AM
RKSimon committed rL348953: [SelectionDAG] Add a generic isSplatValue function.
[SelectionDAG] Add a generic isSplatValue function
Wed, Dec 12, 10:36 AM
RKSimon closed D55426: [SelectionDAG] Add a generic isSplatValue function..
Wed, Dec 12, 10:36 AM
RKSimon added a comment to D55603: test-suite: divide avx512 tests to instruction set subdirectories.

Thanks - this approach makes more sense to me. @craig.topper @MatzeB Any other comments?

Wed, Dec 12, 9:56 AM
RKSimon updated the diff for D55426: [SelectionDAG] Add a generic isSplatValue function..

Addressed @andreadb's comments

Wed, Dec 12, 9:47 AM
RKSimon created D55600: [TargetLowering] Add ISD::OR + ISD::XOR handling to SimplifyDemandedVectorElts.
Wed, Dec 12, 7:08 AM
RKSimon added a comment to D55565: [X86] Don't emit MULX by default with BMI2.

I've generalized https://bugs.llvm.org/show_bug.cgi?id=34232 to cover all possible cases where we should/could use MULX

Wed, Dec 12, 5:56 AM
RKSimon committed rL348926: [TargetLowering] Add ISD::AND handling to SimplifyDemandedVectorElts.
[TargetLowering] Add ISD::AND handling to SimplifyDemandedVectorElts
Wed, Dec 12, 5:46 AM
RKSimon closed D55558: [TargetLowering] Add ISD::AND handling to SimplifyDemandedVectorElts.
Wed, Dec 12, 5:46 AM
RKSimon committed rL348925: Regenerate knownbits test. NFCI..
Regenerate knownbits test. NFCI.
Wed, Dec 12, 5:24 AM
RKSimon accepted D55414: [X86] Emit SBB instead of SETCC_CARRY from LowerSELECT. Break false dependency on the SBB input..
Wed, Dec 12, 3:54 AM
RKSimon added a comment to D55414: [X86] Emit SBB instead of SETCC_CARRY from LowerSELECT. Break false dependency on the SBB input..

How should we proceed here? Can we commit this patch and file a bug for AMD improvements?

Wed, Dec 12, 3:54 AM

Tue, Dec 11

RKSimon created D55558: [TargetLowering] Add ISD::AND handling to SimplifyDemandedVectorElts.
Tue, Dec 11, 7:40 AM
RKSimon added a comment to D52109: [TwoAddressInstructionPass] Don't update SrcRegMap for copies inserted for tied register constraint when the src isn't killed.

What do you want to do about the tests (I think just psubus.ll and avx512-regcall-NoMask.ll) that show an increase in moves?

Tue, Dec 11, 7:13 AM
RKSimon added a reviewer for D55557: [llvm-mca] Move llvm-mca library to llvm/lib/MCA.: mattd.
Tue, Dec 11, 6:57 AM
RKSimon added a comment to D55426: [SelectionDAG] Add a generic isSplatValue function..

gentle ping

Tue, Dec 11, 6:23 AM
RKSimon committed rL348839: [TargetLowering] Add ISD::EXTRACT_VECTOR_ELT support to SimplifyDemandedBits.
[TargetLowering] Add ISD::EXTRACT_VECTOR_ELT support to SimplifyDemandedBits
Tue, Dec 11, 3:11 AM
RKSimon committed rL348838: Fix "not all control paths return a value" MSVC warnings. NFCI..
Fix "not all control paths return a value" MSVC warnings. NFCI.
Tue, Dec 11, 2:48 AM

Mon, Dec 10

RKSimon added a comment to D55494: [x86] allow 8-bit adds to be promoted by convertToThreeAddress() to form LEA.

Cheers - @andreadb @craig.topper do you have any comments?

Mon, Dec 10, 11:35 AM
RKSimon committed rL348784: [TargetLowering] Add UNDEF folding to SimplifyDemandedVectorElts.
[TargetLowering] Add UNDEF folding to SimplifyDemandedVectorElts
Mon, Dec 10, 10:33 AM
RKSimon closed D55511: [TargetLowering] Add UNDEF folding to SimplifyDemandedVectorElts.
Mon, Dec 10, 10:32 AM
RKSimon committed rL348781: [DAGCombiner] Remove unnecessary recursive DAGCombiner::visitINSERT_SUBVECTOR….
[DAGCombiner] Remove unnecessary recursive DAGCombiner::visitINSERT_SUBVECTOR…
Mon, Dec 10, 10:24 AM
RKSimon added inline comments to D55515: [x86] increment/decrement constant vector with min/max in vsetcc lowering (PR39859).
Mon, Dec 10, 9:10 AM
RKSimon updated subscribers of rL348759: [DAGCombiner] Use the result value type in visitCONCAT_VECTORS.
Mon, Dec 10, 7:23 AM
RKSimon created D55511: [TargetLowering] Add UNDEF folding to SimplifyDemandedVectorElts.
Mon, Dec 10, 6:56 AM
RKSimon added inline comments to D55414: [X86] Emit SBB instead of SETCC_CARRY from LowerSELECT. Break false dependency on the SBB input..
Mon, Dec 10, 4:13 AM
RKSimon added a comment to D54278: [SelectionDAG] Teach getNode to constant fold SIGN/ZERO/ANY_EXTEND_VECTOR_INREG.

I'm happy to abandon this.

Mon, Dec 10, 1:10 AM

Sun, Dec 9

RKSimon accepted D55480: [CostModel][X86][AArch64] Adjust cost of the scalarization part of min/max reduction..

LGTM

Sun, Dec 9, 9:29 AM
RKSimon added inline comments to D55494: [x86] allow 8-bit adds to be promoted by convertToThreeAddress() to form LEA.
Sun, Dec 9, 8:48 AM
RKSimon committed rL348721: [X86] Extend pfm counter coverage for llvm-exegesis.
[X86] Extend pfm counter coverage for llvm-exegesis
Sun, Dec 9, 5:48 AM
RKSimon added a comment to D54278: [SelectionDAG] Teach getNode to constant fold SIGN/ZERO/ANY_EXTEND_VECTOR_INREG.

Removing any of the constant folding code in X86 requires committing D54276 first so that we can still see constant MULH inputs as build_vectors and DAG combine them that way. With MULH lowering happening during DAG legalization we don't get a chance to constant fold until they have been turned into constant pool entries and that requires target specific constant folding.

Sun, Dec 9, 4:08 AM
RKSimon added inline comments to D55459: [DAGCombiner][X86] Prevent visitSIGN_EXTEND from returning N when (sext (setcc)) already has the target desired type for the setcc.
Sun, Dec 9, 4:07 AM
RKSimon added inline comments to D55480: [CostModel][X86][AArch64] Adjust cost of the scalarization part of min/max reduction..
Sun, Dec 9, 3:57 AM
RKSimon added inline comments to D55485: [X86] Fix avoid store forwarding blocks pass for negative displacements.
Sun, Dec 9, 3:53 AM
RKSimon added a comment to D38028: [X86][TableGen] Recommitting the X86 memory folding tables TableGen backend while disabling it by default..

You're correct, nothing sets it. I'm nervous about ever using the autogenerated table by default. I've used it a while back to copy entries into the manual tables. But I had to add more waivers and adjustments to fix bugs in the autogenerated table in the process. I don't think they are still a perfect match. I worry that if we switch to the autogenerated table and as we add new instructions in the future we can silently introduce bugs because the new table entries weren't reviewed.

Should we remove the autogenerated table again then?

Sun, Dec 9, 3:40 AM
RKSimon added inline comments to D55481: [CostModel][X86] Have getArithmeticReductionCost call back to the target code after the loop reaches a legal type. Only handle legal/simple types from the X86 code.
Sun, Dec 9, 2:48 AM

Sat, Dec 8

RKSimon added a reviewer for D55452: [CostModel][X86] Teach getArithmeticReductionCost to properly cost the shuffles needed for a PairWise reduction before we reach a legal type. And remove one shuffle from the end of the reduction.: ABataev.
Sat, Dec 8, 5:53 AM

Fri, Dec 7

RKSimon added a reviewer for D55449: [InstCombine] Fix negative GEP offset evaluation for 32-bit pointers: majnemer.
Fri, Dec 7, 1:51 PM
RKSimon committed rL348649: Fix unused variable warning. NFCI..
Fix unused variable warning. NFCI.
Fri, Dec 7, 1:47 PM
RKSimon committed rL348626: [X86] Replace instregex with instrs list. NFCI. .
[X86] Replace instregex with instrs list. NFCI.
Fri, Dec 7, 10:50 AM
RKSimon committed rL348624: [llvm-mca][x86] Add RDSEED instruction resource tests for GLM.
[llvm-mca][x86] Add RDSEED instruction resource tests for GLM
Fri, Dec 7, 10:40 AM
RKSimon committed rL348623: [llvm-mca][x86] Add missing AES instruction resource tests.
[llvm-mca][x86] Add missing AES instruction resource tests
Fri, Dec 7, 10:39 AM
RKSimon committed rL348622: [llvm-mca][x86] Add RDRAND/RDSEED instruction resource tests.
[llvm-mca][x86] Add RDRAND/RDSEED instruction resource tests
Fri, Dec 7, 10:33 AM
RKSimon committed rL348617: [X86] Improve pfm counter coverage for llvm-exegesis.
[X86] Improve pfm counter coverage for llvm-exegesis
Fri, Dec 7, 9:51 AM
RKSimon closed D55432: [X86] Improve pfm counter cpu coverage for llvm-exegesis.
Fri, Dec 7, 9:51 AM
RKSimon added a comment to D55432: [X86] Improve pfm counter cpu coverage for llvm-exegesis.

The plan is to only add these mappings, and encourage people to come up with new sched profiles,
or is the auto-sched-profile-generation upcoming?

Fri, Dec 7, 7:15 AM
RKSimon created D55432: [X86] Improve pfm counter cpu coverage for llvm-exegesis.
Fri, Dec 7, 6:44 AM
RKSimon created D55426: [SelectionDAG] Add a generic isSplatValue function..
Fri, Dec 7, 4:43 AM
RKSimon added a comment to D54719: [Intrinsic] Signed Fixed Point Multiplication Intrinsic.

Use getConstantOperandVal where possible

Fri, Dec 7, 4:22 AM
RKSimon accepted D55397: [CostModel][X86] Fix overcounting arithmetic cost in illegal types getArithmeticReductionCost.

Nice catch! LGTM

Fri, Dec 7, 4:02 AM
RKSimon updated subscribers of rL339967: [WebAssembly] CFG stackify support for exception handling.
Fri, Dec 7, 3:26 AM
RKSimon committed rL348581: Fix gcc7.3 -Wparentheses warning. NFCI..
Fix gcc7.3 -Wparentheses warning. NFCI.
Fri, Dec 7, 3:13 AM
RKSimon updated subscribers of rL348578: [yaml2obj] format some codes NFC..

@Higuoxing I'm seeing build failures because ELFYAML::VersionNeedSection is not defined - please revert and recommit with the missing definition.

Fri, Dec 7, 2:52 AM
RKSimon committed rL348575: [X86] Add ivybridge to llvm-exegesis PFM counter mappings.
[X86] Add ivybridge to llvm-exegesis PFM counter mappings
Fri, Dec 7, 1:30 AM
RKSimon committed rL348574: [SelectionDAG] Don't pass on DemandedElts when handling SCALAR_TO_VECTOR.
[SelectionDAG] Don't pass on DemandedElts when handling SCALAR_TO_VECTOR
Fri, Dec 7, 1:22 AM
RKSimon closed D55223: [SelectionDAG] Don't pass on DemandedElts when handling SCALAR_TO_VECTOR.
Fri, Dec 7, 1:21 AM
RKSimon added a reviewer for D55414: [X86] Emit SBB instead of SETCC_CARRY from LowerSELECT. Break false dependency on the SBB input.: andreadb.
Fri, Dec 7, 1:06 AM

Thu, Dec 6

RKSimon accepted D55355: [X86] Directly create ADC/SBB nodes instead of using ADD/SUB with (and SETCC_CARRY, 1).

LGTM - thanks. If you have concerns that the isel pattern is still used anywhere, I'm fine with it staying with a suitable FIXME comment.

Thu, Dec 6, 1:32 PM
RKSimon committed rL348517: Fix Wdocumentation warning. NFCI..
Fix Wdocumentation warning. NFCI.
Thu, Dec 6, 11:20 AM
RKSimon accepted D55223: [SelectionDAG] Don't pass on DemandedElts when handling SCALAR_TO_VECTOR.

LGTM - I'll commit shortly but please make sure you've requested commit access

Thu, Dec 6, 9:52 AM
RKSimon committed rL348498: [X86] Refactored IsSplatVector to use switch. NFCI..
[X86] Refactored IsSplatVector to use switch. NFCI.
Thu, Dec 6, 8:32 AM
RKSimon accepted D55274: [DagCombiner][X86] Simplify a ConcatVectors of a scalar_to_vector with undef..

LGTM - thanks

Thu, Dec 6, 8:24 AM
RKSimon committed rL348494: DAGCombiner::visitINSERT_VECTOR_ELT - pull out repeated VT.getVectorNumElements….
DAGCombiner::visitINSERT_VECTOR_ELT - pull out repeated VT.getVectorNumElements…
Thu, Dec 6, 7:43 AM
RKSimon added a reviewer for D54839: [CodeGen] Enhance machine PHIs optimization: MatzeB.

This looks ok, but I know very little about this pass - @MatzeB @craig.topper any comments?

Thu, Dec 6, 6:24 AM
RKSimon committed rL348486: Fix -Wcovered-switch-default warning. NFCI..
Fix -Wcovered-switch-default warning. NFCI.
Thu, Dec 6, 6:05 AM
RKSimon added inline comments to D55297: [DemandedBits][BDCE] Support vectors of integers.
Thu, Dec 6, 4:27 AM
RKSimon added a comment to D55355: [X86] Directly create ADC/SBB nodes instead of using ADD/SUB with (and SETCC_CARRY, 1).

This addresses a FIXME and avoids depending on an isel pattern match I think.

Thu, Dec 6, 2:36 AM

Wed, Dec 5

RKSimon committed rL348403: [X86][SSE] Fix a copy+paste typo that was folding the sext/zext of partial….
[X86][SSE] Fix a copy+paste typo that was folding the sext/zext of partial…
Wed, Dec 5, 11:35 AM
RKSimon updated subscribers of rL348365: Move detection of libc++ include dirs to Driver on MacOS.

@ilya-biryukov This is causing windows buildbot failures - revert? http://lab.llvm.org:8011/builders/llvm-clang-lld-x86_64-scei-ps4-windows10pro-fast/builds/21985

Wed, Dec 5, 10:55 AM
RKSimon added a comment to rL348365: Move detection of libc++ include dirs to Driver on MacOS.

This is causing windows buildbot failures - revert? http://lab.llvm.org:8011/builders/llvm-clang-lld-x86_64-scei-ps4-windows10pro-fast/builds/21985

Wed, Dec 5, 10:54 AM
RKSimon added a comment to D55274: [DagCombiner][X86] Simplify a ConcatVectors of a scalar_to_vector with undef..

Also, its probably worth doing the initial simplifyConcatVectors refactor as a NFC

Wed, Dec 5, 7:50 AM
RKSimon added inline comments to D55274: [DagCombiner][X86] Simplify a ConcatVectors of a scalar_to_vector with undef..
Wed, Dec 5, 7:46 AM
RKSimon committed rL348366: [X86][SSE] Begun adding modulo rotate support to LowerRotate.
[X86][SSE] Begun adding modulo rotate support to LowerRotate
Wed, Dec 5, 6:49 AM
RKSimon committed rL348361: [TargetLowering] Remove ISD::ANY_EXTEND/ANY_EXTEND_VECTOR_INREG opcodes from….
[TargetLowering] Remove ISD::ANY_EXTEND/ANY_EXTEND_VECTOR_INREG opcodes from…
Wed, Dec 5, 4:23 AM
RKSimon committed rL348358: [DAG] Add fshl/fshr tblgen opcodes.
[DAG] Add fshl/fshr tblgen opcodes
Wed, Dec 5, 3:58 AM
RKSimon committed rL348353: [SelectionDAG] Initial support for FSHL/FSHR funnel shift opcodes (PR39467).
[SelectionDAG] Initial support for FSHL/FSHR funnel shift opcodes (PR39467)
Wed, Dec 5, 3:16 AM
RKSimon closed D54698: [SelectionDAG] Initial support for FSHL/FSHR funnel shift opcodes (PR39467).
Wed, Dec 5, 3:16 AM
RKSimon committed rL348350: Remove superfluous comments. NFCI..
Remove superfluous comments. NFCI.
Wed, Dec 5, 2:48 AM
RKSimon committed rL348348: [TargetLowering] SimplifyDemandedVectorElts - don't alter DemandedElts mask.
[TargetLowering] SimplifyDemandedVectorElts - don't alter DemandedElts mask
Wed, Dec 5, 2:40 AM
RKSimon edited reviewers for D55310: [utils] Use operator "in" instead of bound function "has_key", added: gbedwell, MaskRay; removed: RKSimon.
Wed, Dec 5, 2:02 AM

Tue, Dec 4

RKSimon committed rL348282: [X86][SSE] Add SimplifyDemandedBitsForTargetNode handling for MOVMSK.
[X86][SSE] Add SimplifyDemandedBitsForTargetNode handling for MOVMSK
Tue, Dec 4, 8:55 AM
RKSimon added a comment to D55274: [DagCombiner][X86] Simplify a ConcatVectors of a scalar_to_vector with undef..

Maybe call the test file combine-concatvectors.ll ? That matches the other filenames we have.

Tue, Dec 4, 8:47 AM
RKSimon committed rL348277: [X86][SSE] Add MOVMSK demandedbits/elts tests.
[X86][SSE] Add MOVMSK demandedbits/elts tests
Tue, Dec 4, 8:04 AM
RKSimon committed rL348265: Add common check prefix. NFCI..
Add common check prefix. NFCI.
Tue, Dec 4, 6:36 AM
RKSimon added a comment to D28907: [SLP] Fix for PR30787: Failure to beneficially vectorize 'copyable' elements in integer binary ops..

@dtemirbulatov Any movement on this? It'd be great to get this in for the 8.0 release!

Tue, Dec 4, 6:18 AM