Page MenuHomePhabricator

RKSimon (Simon Pilgrim)
User

Projects

User does not belong to any projects.

User Details

User Since
May 5 2014, 7:26 AM (250 w, 5 d)

Recent Activity

Today

RKSimon added a comment to D58574: [X86][SSE] Generalize X86ISD::BLENDI support to more value types with fix for revert from r354713.

FWICT from the discussion on D57888, doesn't that mean the problem is really in the commutation code and we should be fixing that?

Sat, Feb 23, 2:26 AM

Yesterday

RKSimon committed rG5d049ce5ef7f: Fix "not all control paths return" warning. NFCI. (authored by RKSimon).
Fix "not all control paths return" warning. NFCI.
Fri, Feb 22, 9:38 AM
RKSimon committed rL354681: Fix "not all control paths return" warning. NFCI..
Fix "not all control paths return" warning. NFCI.
Fri, Feb 22, 9:37 AM
RKSimon committed rC354681: Fix "not all control paths return" warning. NFCI..
Fix "not all control paths return" warning. NFCI.
Fri, Feb 22, 9:37 AM
RKSimon added a comment to D51589: DAG: Fold extract_vector_elt (scalar_to_vector), K to undef.

The patch has rotted and needs rewriting, but putting in an assert causes hits in these tests:

Failing Tests (16):
    LLVM :: CodeGen/AMDGPU/copy-illegal-type.ll
    LLVM :: CodeGen/AMDGPU/cvt_f32_ubyte.ll
    LLVM :: CodeGen/AMDGPU/flat_atomics_i64.ll
    LLVM :: CodeGen/AMDGPU/global_atomics_i64.ll
    LLVM :: CodeGen/AMDGPU/load-constant-i1.ll
    LLVM :: CodeGen/AMDGPU/load-constant-i16.ll
    LLVM :: CodeGen/AMDGPU/load-global-i1.ll
    LLVM :: CodeGen/AMDGPU/load-global-i16.ll
    LLVM :: CodeGen/AMDGPU/load-local-i1.ll
    LLVM :: CodeGen/AMDGPU/load-local-i16.ll
    LLVM :: CodeGen/AMDGPU/max.i16.ll
    LLVM :: CodeGen/AMDGPU/min.ll
    LLVM :: CodeGen/AMDGPU/select.f16.ll
    LLVM :: CodeGen/AMDGPU/selectcc.ll
    LLVM :: CodeGen/AMDGPU/v_madak_f16.ll
    LLVM :: CodeGen/X86/vector-intrinsics.ll

If possible please can you use one of these to show a codegen change? I tried the x86 test but didn't find anything.

Fri, Feb 22, 4:07 AM
RKSimon accepted D57389: [X86] Improve use of SHLD/SHRD.

LGTM

Fri, Feb 22, 1:39 AM · Restricted Project

Thu, Feb 21

RKSimon created D58503: [SelectionDAG] Add demanded elts variants to isConstSplat helpers. NFCI..
Thu, Feb 21, 5:45 AM · Restricted Project
RKSimon committed rGe6b338cbef18: [X86][SSE] combineX86ShufflesRecursively - moved to generic op input index… (authored by RKSimon).
[X86][SSE] combineX86ShufflesRecursively - moved to generic op input index…
Thu, Feb 21, 4:25 AM
RKSimon committed rL354575: [X86][SSE] combineX86ShufflesRecursively - moved to generic op input index….
[X86][SSE] combineX86ShufflesRecursively - moved to generic op input index…
Thu, Feb 21, 4:24 AM
RKSimon added a comment to D58470: [X86] Reduce some patterns by using FP instructions for integer types even when AVX2 is available and execution domain fixing will do the right thing.

Here you're making more use of the domain pass to reduce isel patterns, but in rL354511 you added isel patterns to avoid the load folding pass - what's the difference?

Thu, Feb 21, 1:35 AM

Wed, Feb 20

RKSimon added inline comments to D58282: [x86] scalarize extract element 0 of FP math.
Wed, Feb 20, 12:42 PM · Restricted Project
RKSimon accepted D58464: [SelectionDAG] Teach GetDemandedBits to look at the known zeros of the LHS when handling ISD::AND.

LGTM

Wed, Feb 20, 12:22 PM · Restricted Project
RKSimon accepted D58197: [x86] vectorize more cast ops in lowering to avoid register file transfers.

LGTM, cheers

Wed, Feb 20, 12:18 PM · Restricted Project
RKSimon committed rGdca47c659ce2: [X86][SSE] combineX86ShufflesRecursively - begin generalizing the number of… (authored by RKSimon).
[X86][SSE] combineX86ShufflesRecursively - begin generalizing the number of…
Wed, Feb 20, 10:02 AM
RKSimon committed rL354489: [X86][SSE] combineX86ShufflesRecursively - begin generalizing the number of….
[X86][SSE] combineX86ShufflesRecursively - begin generalizing the number of…
Wed, Feb 20, 10:01 AM
RKSimon added a comment to D57389: [X86] Improve use of SHLD/SHRD.

@deadalnix Please can you rebase this?

Wed, Feb 20, 4:14 AM · Restricted Project
RKSimon committed rG9921e73d9557: [SLPVectorizer][X86] Add add/sub/mul overflow tests (authored by RKSimon).
[SLPVectorizer][X86] Add add/sub/mul overflow tests
Wed, Feb 20, 4:07 AM
RKSimon committed rL354454: [SLPVectorizer][X86] Add add/sub/mul overflow tests.
[SLPVectorizer][X86] Add add/sub/mul overflow tests
Wed, Feb 20, 4:06 AM
RKSimon accepted D57997: [SDAG] Support vector UMULO/SMULO.

LGTM

Wed, Feb 20, 2:47 AM · Restricted Project

Tue, Feb 19

Herald added a project to D47735: [DAGCombiner] Create rotates more aggressively: Restricted Project.

@kparzysz Do we still need this? Does the IR funnel shift work that @spatel did last year make this redundant?

Tue, Feb 19, 10:34 AM · Restricted Project
RKSimon resigned from D40602: [X86] Add MC level selection support for SHLD (64-bit only).
Tue, Feb 19, 10:31 AM
RKSimon resigned from D26855: New unsafe-fp-math implementation for X86 target.
Tue, Feb 19, 10:30 AM
RKSimon added inline comments to D57997: [SDAG] Support vector UMULO/SMULO.
Tue, Feb 19, 10:25 AM · Restricted Project
RKSimon added a comment to D57888: [X86][SSE] Generalize X86ISD::BLENDI support to more value types (WIP).

@dlj @sammccall I think I've fixed the root issue now (rL354358) - please reply on this ticket if there are any more issues. Cheers!

Tue, Feb 19, 10:10 AM · Restricted Project
RKSimon committed rG0b3b9424ca8a: [X86][SSE] Generalize X86ISD::BLENDI support to more value types (authored by RKSimon).
[X86][SSE] Generalize X86ISD::BLENDI support to more value types
Tue, Feb 19, 10:05 AM
RKSimon committed rL354363: [X86][SSE] Generalize X86ISD::BLENDI support to more value types.
[X86][SSE] Generalize X86ISD::BLENDI support to more value types
Tue, Feb 19, 10:05 AM
RKSimon closed D57888: [X86][SSE] Generalize X86ISD::BLENDI support to more value types (WIP).
Tue, Feb 19, 10:05 AM · Restricted Project
RKSimon committed rGd58cc6f9ee35: Fix stupid assembly comment typo (authored by RKSimon).
Fix stupid assembly comment typo
Tue, Feb 19, 9:57 AM
RKSimon committed rL354361: Fix stupid assembly comment typo.
Fix stupid assembly comment typo
Tue, Feb 19, 9:57 AM
RKSimon committed rGe31838f88b60: [X86][SSE] Add pblendw commuted load test case (authored by RKSimon).
[X86][SSE] Add pblendw commuted load test case
Tue, Feb 19, 9:56 AM
RKSimon committed rL354360: [X86][SSE] Add pblendw commuted load test case.
[X86][SSE] Add pblendw commuted load test case
Tue, Feb 19, 9:55 AM
RKSimon committed rGdce9c2a8119c: [X86][AVX2] Hide VPBLENDD instructions behind AVX2 predicate (authored by RKSimon).
[X86][AVX2] Hide VPBLENDD instructions behind AVX2 predicate
Tue, Feb 19, 9:24 AM
RKSimon committed rL354358: [X86][AVX2] Hide VPBLENDD instructions behind AVX2 predicate.
[X86][AVX2] Hide VPBLENDD instructions behind AVX2 predicate
Tue, Feb 19, 9:23 AM
RKSimon committed rG9d575db85ed5: [X86][AVX] Update VBROADCAST folds to always use v2i64 X86vzload (authored by RKSimon).
[X86][AVX] Update VBROADCAST folds to always use v2i64 X86vzload
Tue, Feb 19, 8:33 AM
RKSimon committed rL354346: [X86][AVX] Update VBROADCAST folds to always use v2i64 X86vzload.
[X86][AVX] Update VBROADCAST folds to always use v2i64 X86vzload
Tue, Feb 19, 8:32 AM
RKSimon committed rGd6add7491564: Cast from SDValue directly instead of superfluous getNode(). NFCI. (authored by RKSimon).
Cast from SDValue directly instead of superfluous getNode(). NFCI.
Tue, Feb 19, 8:21 AM
RKSimon committed rL354343: Cast from SDValue directly instead of superfluous getNode(). NFCI..
Cast from SDValue directly instead of superfluous getNode(). NFCI.
Tue, Feb 19, 8:19 AM
RKSimon committed rG952abcefe428: [X86][AVX] EltsFromConsecutiveLoads - Add BROADCAST lowering support (authored by RKSimon).
[X86][AVX] EltsFromConsecutiveLoads - Add BROADCAST lowering support
Tue, Feb 19, 8:00 AM
RKSimon committed rL354340: [X86][AVX] EltsFromConsecutiveLoads - Add BROADCAST lowering support.
[X86][AVX] EltsFromConsecutiveLoads - Add BROADCAST lowering support
Tue, Feb 19, 8:00 AM
RKSimon closed D58053: [X86][AVX] EltsFromConsecutiveLoads - Add BROADCAST lowering support.
Tue, Feb 19, 8:00 AM · Restricted Project
RKSimon added a comment to D57375: X86AsmParser AVX-512: Return error instead of hitting assert.

@craig.topper Any comments?

Tue, Feb 19, 2:20 AM · Restricted Project

Mon, Feb 18

RKSimon added inline comments to D57997: [SDAG] Support vector UMULO/SMULO.
Mon, Feb 18, 12:43 PM · Restricted Project
RKSimon accepted D58302: [X86] Add ImmArg markings to intrinsics..

LGTM once D57825 lands

Mon, Feb 18, 12:22 PM
RKSimon added a comment to D58053: [X86][AVX] EltsFromConsecutiveLoads - Add BROADCAST lowering support.

ping?

Mon, Feb 18, 9:48 AM · Restricted Project
RKSimon added a comment to rL354243: [MC] Make SubtargetFeatureKV only store one FeatureBitset and use an 'unsigned'….

cheers - I tried that but it had stayed red - I don't know what it is about that bot but it acts up a lot......

Mon, Feb 18, 9:45 AM
RKSimon accepted D57789: [CGP] form usub with overflow from sub+icmp.

LGTM - cheers

Mon, Feb 18, 9:26 AM · Restricted Project
RKSimon updated subscribers of rL354243: [MC] Make SubtargetFeatureKV only store one FeatureBitset and use an 'unsigned'….

@craig.topper This appears to breaking some buildbots:

Mon, Feb 18, 8:07 AM
RKSimon added a comment to D58302: [X86] Add ImmArg markings to intrinsics..

Missing context.

Mon, Feb 18, 4:00 AM
RKSimon added inline comments to D57997: [SDAG] Support vector UMULO/SMULO.
Mon, Feb 18, 3:44 AM · Restricted Project
RKSimon accepted D58181: [x86] split more v8f32/v8i32 shuffles in lowering.

LGTM - cheers.

Mon, Feb 18, 3:37 AM · Restricted Project

Fri, Feb 15

RKSimon committed rGa22814a399e2: Fix 80-column limit in SimplifyDemandedBits/SimplifyDemandedVectorElts. NFCI. (authored by RKSimon).
Fix 80-column limit in SimplifyDemandedBits/SimplifyDemandedVectorElts. NFCI.
Fri, Feb 15, 10:17 AM
RKSimon committed rL354152: Fix 80-column limit in SimplifyDemandedBits/SimplifyDemandedVectorElts. NFCI..
Fix 80-column limit in SimplifyDemandedBits/SimplifyDemandedVectorElts. NFCI.
Fri, Feb 15, 10:17 AM
RKSimon added a comment to D58181: [x86] split more v8f32/v8i32 shuffles in lowering.

Better to stage this in 2 parts?
I can add a check for fast-variable-shuffle, so we get the clear improvements. Then, a follow-up can remove that check and see if that results in any real-world fallout.

Fri, Feb 15, 10:08 AM · Restricted Project
RKSimon added inline comments to D58282: [x86] scalarize extract element 0 of FP math.
Fri, Feb 15, 8:10 AM · Restricted Project
RKSimon added a comment to D56772: [MIR] Add simple PRE pass to MachineCSE.

@hfinkel @MatzeB Any comments?

Fri, Feb 15, 4:30 AM
RKSimon accepted D57788: [X86] Don't set exception mask bits when modifying FPCW to change rounding mode for fp->int conversion.

LGTM with one minor

Fri, Feb 15, 4:27 AM · Restricted Project
RKSimon updated subscribers of rL354075: [clang][FileManager] fillRealPathName even if we aren't opening the file.

@jkorous This is failing on windows buildbots - do you have a fix or shall I revert?

Fri, Feb 15, 4:23 AM
RKSimon committed rG623c38d6cd31: Fix "field 'DFS' will be initialized after field 'DTU'" warning. NFCI. (authored by RKSimon).
Fix "field 'DFS' will be initialized after field 'DTU'" warning. NFCI.
Fri, Feb 15, 4:13 AM
RKSimon committed rL354123: Fix "field 'DFS' will be initialized after field 'DTU'" warning. NFCI..
Fix "field 'DFS' will be initialized after field 'DTU'" warning. NFCI.
Fri, Feb 15, 4:13 AM
RKSimon committed rG6ce08672fb4d: [X86][AVX] lowerShuffleAsLanePermuteAndPermute - fully populate the lane… (authored by RKSimon).
[X86][AVX] lowerShuffleAsLanePermuteAndPermute - fully populate the lane…
Fri, Feb 15, 3:39 AM
RKSimon committed rL354117: [X86][AVX] lowerShuffleAsLanePermuteAndPermute - fully populate the lane….
[X86][AVX] lowerShuffleAsLanePermuteAndPermute - fully populate the lane…
Fri, Feb 15, 3:39 AM
RKSimon closed D58237: [X86][AVX] lowerShuffleAsLanePermuteAndPermute - fully populate the lane shuffle mask (PR40730).
Fri, Feb 15, 3:39 AM · Restricted Project

Thu, Feb 14

RKSimon added a reviewer for D58015: [SelectionDAG][AArch64] Legalize VECREDUCE: gnzlbg.
Thu, Feb 14, 12:21 PM · Restricted Project
RKSimon created D58237: [X86][AVX] lowerShuffleAsLanePermuteAndPermute - fully populate the lane shuffle mask (PR40730).
Thu, Feb 14, 7:51 AM · Restricted Project
RKSimon committed rG362fe5603453: [X86][AVX] Add PR40730 test case (authored by RKSimon).
[X86][AVX] Add PR40730 test case
Thu, Feb 14, 6:45 AM
RKSimon committed rL354034: [X86][AVX] Add PR40730 test case.
[X86][AVX] Add PR40730 test case
Thu, Feb 14, 6:45 AM
RKSimon added a reviewer for D58225: [MIPS MSA] Avoid some DAG combines for vector shifts: craig.topper.

Intel targets tend to only accept vector shifts on Port0, while vector logic can use Port0/1/5 - not sure how much of an issue that would be @craig.topper ? Some AMD targets are almost as bad while others (Jaguar) can issue vector immediate shifts to any vector integer pipe.

Thu, Feb 14, 5:25 AM · Restricted Project
RKSimon added inline comments to D58197: [x86] vectorize more cast ops in lowering to avoid register file transfers.
Thu, Feb 14, 4:17 AM · Restricted Project

Wed, Feb 13

RKSimon added inline comments to D58197: [x86] vectorize more cast ops in lowering to avoid register file transfers.
Wed, Feb 13, 12:49 PM · Restricted Project
RKSimon committed rG48d27e839356: [X86][AVX] Add shuffle_v8i32_0dcd3f14 shuffle test case (authored by RKSimon).
[X86][AVX] Add shuffle_v8i32_0dcd3f14 shuffle test case
Wed, Feb 13, 8:12 AM
RKSimon committed rL353953: [X86][AVX] Add shuffle_v8i32_0dcd3f14 shuffle test case.
[X86][AVX] Add shuffle_v8i32_0dcd3f14 shuffle test case
Wed, Feb 13, 8:12 AM

Tue, Feb 12

RKSimon accepted D58066: [MCA][Scheduler] Use latency information to further classify busy instructions..

LGTM with a couple of minors

Tue, Feb 12, 9:53 AM · Restricted Project
RKSimon committed rG5338f41ced1f: [X86][AVX] Enable shuffle combining support for zero_extend (authored by RKSimon).
[X86][AVX] Enable shuffle combining support for zero_extend
Tue, Feb 12, 9:24 AM
RKSimon committed rL353860: [X86][AVX] Enable shuffle combining support for zero_extend .
[X86][AVX] Enable shuffle combining support for zero_extend
Tue, Feb 12, 9:23 AM
RKSimon added a comment to D57888: [X86][SSE] Generalize X86ISD::BLENDI support to more value types (WIP).

Thanks @sammccall I've been able to repro it now.

Tue, Feb 12, 9:18 AM · Restricted Project
RKSimon committed rG015cc0f0fabc: [PowerPC] Regenerate test (authored by RKSimon).
[PowerPC] Regenerate test
Tue, Feb 12, 8:12 AM
RKSimon committed rL353851: [PowerPC] Regenerate test.
[PowerPC] Regenerate test
Tue, Feb 12, 8:12 AM

Mon, Feb 11

RKSimon added inline comments to D58066: [MCA][Scheduler] Use latency information to further classify busy instructions..
Mon, Feb 11, 11:19 AM · Restricted Project
RKSimon added reviewers for D45355: [SelectionDAG] Fix return calling convention in expansion of ?MULO: efriedma, nikic, RKSimon.

Adding @nikic as there's crossover with D58006

Mon, Feb 11, 9:20 AM · Restricted Project
RKSimon committed rG9ea8f49a83ef: [X86] Regenerate insertelement tests (authored by RKSimon).
[X86] Regenerate insertelement tests
Mon, Feb 11, 8:18 AM
RKSimon committed rL353716: [X86] Regenerate insertelement tests.
[X86] Regenerate insertelement tests
Mon, Feb 11, 8:15 AM
RKSimon added inline comments to D57779: [SLP] Add support for throttling..
Mon, Feb 11, 7:36 AM
RKSimon created D58053: [X86][AVX] EltsFromConsecutiveLoads - Add BROADCAST lowering support.
Mon, Feb 11, 7:29 AM · Restricted Project
RKSimon reopened D57888: [X86][SSE] Generalize X86ISD::BLENDI support to more value types (WIP).

Let's try this again....

Mon, Feb 11, 6:45 AM · Restricted Project
RKSimon closed D42042: [X86][SSE] Add custom execution domain fixing for BLENDPD/BLENDPS/PBLENDD/PBLENDW (PR34873).

I hate Mondays - closing this again as its the wrong patch........

Mon, Feb 11, 6:44 AM · Restricted Project
RKSimon reopened D42042: [X86][SSE] Add custom execution domain fixing for BLENDPD/BLENDPS/PBLENDD/PBLENDW (PR34873).

Reopening as it was reverted at rL353699 due to a rather weird regression.....

Mon, Feb 11, 6:43 AM · Restricted Project

Sun, Feb 10

RKSimon committed rGf6e6c369c0ea: [X86] EltsFromConsecutiveLoads - replace SmallBitVector with APInt (NFC). (authored by RKSimon).
[X86] EltsFromConsecutiveLoads - replace SmallBitVector with APInt (NFC).
Sun, Feb 10, 2:46 PM
RKSimon committed rL353655: [X86] EltsFromConsecutiveLoads - replace SmallBitVector with APInt (NFC)..
[X86] EltsFromConsecutiveLoads - replace SmallBitVector with APInt (NFC).
Sun, Feb 10, 2:45 PM
RKSimon created D58017: [DAG] Add SimplifyDemandedBits support for BSWAP/BITREVERSE.
Sun, Feb 10, 11:34 AM · Restricted Project
RKSimon accepted D57348: [CodeGen][X86] Don't scalarize vector saturating add/sub.

LGTM

Sun, Feb 10, 10:49 AM · Restricted Project
RKSimon committed rGa303186ef3c6: [AArch64] Regenerate bswap tests (authored by RKSimon).
[AArch64] Regenerate bswap tests
Sun, Feb 10, 10:30 AM
RKSimon committed rL353648: [AArch64] Regenerate bswap tests.
[AArch64] Regenerate bswap tests
Sun, Feb 10, 10:29 AM
RKSimon added a comment to D57317: [DAGCombine] Deduplicate addcarry node using commutativity..

@RKSimon This can go now, but has no effect in isolation in any of the test cases I have, mainly because DAGCombine is fairly powerless when faced with addcarry as most of the work involve 'deep' patterns. Do you think I should land anyways ?

Sun, Feb 10, 10:19 AM · Restricted Project
RKSimon committed rGce1031298669: [X86] Add basic bitreverse/bswap combine tests (authored by RKSimon).
[X86] Add basic bitreverse/bswap combine tests
Sun, Feb 10, 10:07 AM
RKSimon committed rL353647: [X86] Add basic bitreverse/bswap combine tests.
[X86] Add basic bitreverse/bswap combine tests
Sun, Feb 10, 10:06 AM
RKSimon committed rGc5744d4d69da: [DAG] Add optional AllowUndefs to isNullOrNullSplat (authored by RKSimon).
[DAG] Add optional AllowUndefs to isNullOrNullSplat
Sun, Feb 10, 9:42 AM
RKSimon committed rL353646: [DAG] Add optional AllowUndefs to isNullOrNullSplat.
[DAG] Add optional AllowUndefs to isNullOrNullSplat
Sun, Feb 10, 9:42 AM
RKSimon committed rG5a82a788a28e: [DAGCombine] Simplify funnel shifts with undef/zero args to bitshifts (authored by RKSimon).
[DAGCombine] Simplify funnel shifts with undef/zero args to bitshifts
Sun, Feb 10, 9:04 AM
RKSimon committed rL353645: [DAGCombine] Simplify funnel shifts with undef/zero args to bitshifts.
[DAGCombine] Simplify funnel shifts with undef/zero args to bitshifts
Sun, Feb 10, 9:03 AM
RKSimon closed D58009: [DAGCombine] Simplify funnel shifts with undef/zero args to bitshifts.
Sun, Feb 10, 9:03 AM · Restricted Project