Page MenuHomePhabricator

RKSimon (Simon Pilgrim)
User

Projects

User does not belong to any projects.

User Details

User Since
May 5 2014, 7:26 AM (259 w, 2 d)

Recent Activity

Today

RKSimon accepted D60546: [X86] Use MOVQ for i64 atomic_stores when SSE2 is enabled.

LGTM

Wed, Apr 24, 12:17 PM · Restricted Project
RKSimon accepted D61028: [DAGCombiner] scale repeated FP divisor by splat factor.

LGTM

Wed, Apr 24, 12:04 PM · Restricted Project
RKSimon added inline comments to D61068: [X86][SSE] Disable shouldFoldConstantShiftPairToMask for btver1/btver2 targets (PR40758).
Wed, Apr 24, 12:01 PM · Restricted Project
RKSimon committed rG10daecba1de8: [X86][SSE] Add tests for bitcasting vXi1 bool vectors to non-simple types. (authored by RKSimon).
[X86][SSE] Add tests for bitcasting vXi1 bool vectors to non-simple types.
Wed, Apr 24, 10:24 AM
RKSimon committed rL359116: [X86][SSE] Add tests for bitcasting vXi1 bool vectors to non-simple types..
[X86][SSE] Add tests for bitcasting vXi1 bool vectors to non-simple types.
Wed, Apr 24, 10:23 AM
RKSimon added inline comments to D61028: [DAGCombiner] scale repeated FP divisor by splat factor.
Wed, Apr 24, 9:56 AM · Restricted Project
RKSimon committed rG55f14dac7455: [InstCombine][X86] Use generic expansion of PACKSS/PACKUS for constant folding. (authored by RKSimon).
[InstCombine][X86] Use generic expansion of PACKSS/PACKUS for constant folding.
Wed, Apr 24, 9:54 AM
RKSimon committed rL359111: [InstCombine][X86] Use generic expansion of PACKSS/PACKUS for constant folding..
[InstCombine][X86] Use generic expansion of PACKSS/PACKUS for constant folding.
Wed, Apr 24, 9:51 AM
RKSimon added inline comments to D61028: [DAGCombiner] scale repeated FP divisor by splat factor.
Wed, Apr 24, 7:54 AM · Restricted Project
RKSimon created D61068: [X86][SSE] Disable shouldFoldConstantShiftPairToMask for btver1/btver2 targets (PR40758).
Wed, Apr 24, 7:18 AM · Restricted Project
RKSimon committed rGd30745b2a04e: [X86] Add shouldFoldConstantShiftPairToMask override placeholder. NFCI. (authored by RKSimon).
[X86] Add shouldFoldConstantShiftPairToMask override placeholder. NFCI.
Wed, Apr 24, 5:33 AM
RKSimon committed rL359088: [X86] Add shouldFoldConstantShiftPairToMask override placeholder. NFCI..
[X86] Add shouldFoldConstantShiftPairToMask override placeholder. NFCI.
Wed, Apr 24, 5:32 AM
RKSimon committed rG039a563e6a29: [X86][SSE] Add masked bit test cases for PR26697 (authored by RKSimon).
[X86][SSE] Add masked bit test cases for PR26697
Wed, Apr 24, 3:33 AM
RKSimon committed rL359082: [X86][SSE] Add masked bit test cases for PR26697.
[X86][SSE] Add masked bit test cases for PR26697
Wed, Apr 24, 3:33 AM

Yesterday

RKSimon added inline comments to D60993: [X86] Initial cleanups on the FixupLEAs pass. Separate Atom LEA creation from other LEA optimizations..
Tue, Apr 23, 1:18 PM · Restricted Project
RKSimon added inline comments to D60975: Convert a masked.gather of at most one element to a masked.load.
Tue, Apr 23, 1:04 PM · Restricted Project
RKSimon added a comment to D60512: [X86][AVX] Combine shuffles extracted from a common vector.

ping?

Tue, Apr 23, 12:34 PM · Restricted Project
RKSimon added inline comments to D61024: [SLP] Fix crash after r358519.
Tue, Apr 23, 9:53 AM · Restricted Project
RKSimon committed rG0e4992ce27b9: [X86] Pull out collectConcatOps helper. NFCI. (authored by RKSimon).
[X86] Pull out collectConcatOps helper. NFCI.
Tue, Apr 23, 7:09 AM
RKSimon committed rL358989: [X86] Pull out collectConcatOps helper. NFCI..
[X86] Pull out collectConcatOps helper. NFCI.
Tue, Apr 23, 7:08 AM
RKSimon added a reviewer for D60993: [X86] Initial cleanups on the FixupLEAs pass. Separate Atom LEA creation from other LEA optimizations.: andreadb.
Tue, Apr 23, 6:04 AM · Restricted Project
RKSimon committed rG6b1825081dcf: Fix "-Wimplicit-fallthrough" warning. NFCI. (authored by RKSimon).
Fix "-Wimplicit-fallthrough" warning. NFCI.
Tue, Apr 23, 4:44 AM
RKSimon committed rC358973: Fix "-Wimplicit-fallthrough" warning. NFCI..
Fix "-Wimplicit-fallthrough" warning. NFCI.
Tue, Apr 23, 4:43 AM
RKSimon committed rL358973: Fix "-Wimplicit-fallthrough" warning. NFCI..
Fix "-Wimplicit-fallthrough" warning. NFCI.
Tue, Apr 23, 4:43 AM
RKSimon accepted D60838: [x86] use psubus for more vsetcc lowering (PR39859).

LGTM

Tue, Apr 23, 4:29 AM · Restricted Project
RKSimon committed rGddd225d1a92e: Fix MSVC "32-bit shift implicitly converted to 64 bits" warning. NFCI. (authored by RKSimon).
Fix MSVC "32-bit shift implicitly converted to 64 bits" warning. NFCI.
Tue, Apr 23, 4:15 AM
RKSimon committed rL358970: Fix MSVC "32-bit shift implicitly converted to 64 bits" warning. NFCI..
Fix MSVC "32-bit shift implicitly converted to 64 bits" warning. NFCI.
Tue, Apr 23, 4:14 AM
RKSimon committed rGe7a68fd93e82: Fix MSVC "32-bit shift implicitly converted to 64 bits" warning. NFCI. (authored by RKSimon).
Fix MSVC "32-bit shift implicitly converted to 64 bits" warning. NFCI.
Tue, Apr 23, 4:10 AM
RKSimon committed rL358969: Fix MSVC "32-bit shift implicitly converted to 64 bits" warning. NFCI..
Fix MSVC "32-bit shift implicitly converted to 64 bits" warning. NFCI.
Tue, Apr 23, 4:09 AM

Mon, Apr 22

RKSimon updated the diff for D56387: [DAGCombiner] Enable SimplifyDemandedBits vector support for TRUNCATE (WIP).

rebase - still showing a number of regressions that are proving tricky to fix

Mon, Apr 22, 7:31 AM · Restricted Project
RKSimon abandoned D59422: [SelectionDAG] Use SimplifyDemandedBits on truncated SCALAR_TO_VECTORs.

Abandoning - the x86 improvements were handled by rL358019

Mon, Apr 22, 7:18 AM · Restricted Project
RKSimon committed rG6276ce014283: [TargetLowering][AMDGPU][X86] Improve SimplifyDemandedBits bitcast handling (authored by RKSimon).
[TargetLowering][AMDGPU][X86] Improve SimplifyDemandedBits bitcast handling
Mon, Apr 22, 7:05 AM
RKSimon committed rL358887: [TargetLowering][AMDGPU][X86] Improve SimplifyDemandedBits bitcast handling.
[TargetLowering][AMDGPU][X86] Improve SimplifyDemandedBits bitcast handling
Mon, Apr 22, 7:03 AM
RKSimon closed D60462: [TargetLowering][AMDGPU][X86] Improve SimplifyDemandedBits bitcast handling.
Mon, Apr 22, 7:02 AM · Restricted Project
RKSimon added a reviewer for D59422: [SelectionDAG] Use SimplifyDemandedBits on truncated SCALAR_TO_VECTORs: spatel.
Mon, Apr 22, 6:54 AM · Restricted Project
RKSimon added inline comments to D60462: [TargetLowering][AMDGPU][X86] Improve SimplifyDemandedBits bitcast handling.
Mon, Apr 22, 6:54 AM · Restricted Project
RKSimon updated the diff for D60462: [TargetLowering][AMDGPU][X86] Improve SimplifyDemandedBits bitcast handling.

Add AMDGPU srl(and(x,m),c) -> and(srl(x,c),srl(m,c)) canonicalization to improve BFE recognition

Mon, Apr 22, 6:47 AM · Restricted Project
RKSimon committed rGffd67233d461: [AMDGPU] Regenerate uitofp i8 to float conversion tests. (authored by RKSimon).
[AMDGPU] Regenerate uitofp i8 to float conversion tests.
Mon, Apr 22, 3:19 AM
RKSimon committed rL358879: [AMDGPU] Regenerate uitofp i8 to float conversion tests. .
[AMDGPU] Regenerate uitofp i8 to float conversion tests.
Mon, Apr 22, 3:19 AM

Fri, Apr 19

RKSimon committed rG4c09b7d921b4: [AMDGPU] Regenerate extractelt->truncate test. (authored by RKSimon).
[AMDGPU] Regenerate extractelt->truncate test.
Fri, Apr 19, 2:49 AM
RKSimon committed rL358746: [AMDGPU] Regenerate extractelt->truncate test. .
[AMDGPU] Regenerate extractelt->truncate test.
Fri, Apr 19, 2:48 AM

Thu, Apr 18

RKSimon added reviewers for D60851: [llvm-exegesis] Insert IACA markers: RKSimon, lebedev.ri, andreadb.

Please add context to the diff

Thu, Apr 18, 10:27 AM · Restricted Project
RKSimon committed rG4171a91e9288: [X86] combineVectorTruncationWithPACKUS - remove split/concatenation of mask (authored by RKSimon).
[X86] combineVectorTruncationWithPACKUS - remove split/concatenation of mask
Thu, Apr 18, 10:22 AM
RKSimon committed rL358692: [X86] combineVectorTruncationWithPACKUS - remove split/concatenation of mask.
[X86] combineVectorTruncationWithPACKUS - remove split/concatenation of mask
Thu, Apr 18, 10:21 AM
RKSimon closed D60375: [X86] combineVectorTruncationWithPACKUS - remove split/concatenation of mask.
Thu, Apr 18, 10:21 AM · Restricted Project
RKSimon added a comment to D60375: [X86] combineVectorTruncationWithPACKUS - remove split/concatenation of mask.

ping?

Thu, Apr 18, 3:14 AM · Restricted Project
RKSimon committed rG8f87e53462a6: [X86][SSE] Lower ICMP EQ(AND(X,C),C) -> SRA(SHL(X,LOG2(C)),BW-1) iff C is power… (authored by RKSimon).
[X86][SSE] Lower ICMP EQ(AND(X,C),C) -> SRA(SHL(X,LOG2(C)),BW-1) iff C is power…
Thu, Apr 18, 2:58 AM
RKSimon committed rL358651: [X86][SSE] Lower ICMP EQ(AND(X,C),C) -> SRA(SHL(X,LOG2(C)),BW-1) iff C is power….
[X86][SSE] Lower ICMP EQ(AND(X,C),C) -> SRA(SHL(X,LOG2(C)),BW-1) iff C is power…
Thu, Apr 18, 2:57 AM
RKSimon closed D60625: [X86][SSE] Lower ICMP EQ(AND(X,C),C) -> SRA(SHL(X,LOG2(C)),BW-1) iff C is power-of-2..
Thu, Apr 18, 2:57 AM · Restricted Project
RKSimon added a comment to D60852: Fix for bug 41512: lower INSERT_VECTOR_ELT(ZeroVec, 0, Elt) to SCALAR_TO_VECTOR(Elt) for all SSE flavors.

Your test cases need to be a lot simpler - I'd recommend looking at buildvec-insertvec.ll and possibly adding your tests to that file instead of adding this new file.

Thu, Apr 18, 2:22 AM · Restricted Project
RKSimon added a reviewer for D60852: Fix for bug 41512: lower INSERT_VECTOR_ELT(ZeroVec, 0, Elt) to SCALAR_TO_VECTOR(Elt) for all SSE flavors: RKSimon.
Thu, Apr 18, 2:17 AM · Restricted Project

Wed, Apr 17

RKSimon requested changes to D60214: [DAGCombiner] move splat-shuffle after binop with splat constant.

Please add support for arm/aarch64 splat-and-multiply instructions

Wed, Apr 17, 1:49 PM · Restricted Project
RKSimon added inline comments to D60552: [X86] Enable intrinsics of AVX512_BF16, which are supported for BFLOAT16 in Cooper Lake.
Wed, Apr 17, 1:32 PM · Restricted Project
RKSimon added a comment to D60550: [X86] Enable AVX512_BF16 instructions, which are supported for BFLOAT16 in Cooper Lake.

Some initial thoughts - I don't know a lot about the bfloat16 instructions so need to read up when I get the chance.

Wed, Apr 17, 1:31 PM · Restricted Project
RKSimon accepted D60214: [DAGCombiner] move splat-shuffle after binop with splat constant.

Cheers, LGTM

Wed, Apr 17, 9:53 AM · Restricted Project
RKSimon committed rGe7fe6dd5edb8: [DAGCombine] Add SimplifyDemandedBits helper that handles demanded elts mask as… (authored by RKSimon).
[DAGCombine] Add SimplifyDemandedBits helper that handles demanded elts mask as…
Wed, Apr 17, 8:46 AM
RKSimon committed rL358585: [DAGCombine] Add SimplifyDemandedBits helper that handles demanded elts mask as….
[DAGCombine] Add SimplifyDemandedBits helper that handles demanded elts mask as…
Wed, Apr 17, 8:45 AM
RKSimon added a comment to D59744: Fix i386 ABI "__m64" type bug.

One last style comment from me but we need somebody better with the different ABIs to finally approve this.

Wed, Apr 17, 8:41 AM · Restricted Project
RKSimon added a comment to D60214: [DAGCombiner] move splat-shuffle after binop with splat constant.

A couple of minors but this looks almost ready to me, the avx512 broadcast folds are a known issue

Wed, Apr 17, 8:33 AM · Restricted Project
RKSimon added a reviewer for D60462: [TargetLowering][AMDGPU][X86] Improve SimplifyDemandedBits bitcast handling: nhaehnle.
Wed, Apr 17, 8:24 AM · Restricted Project
RKSimon updated the diff for D60462: [TargetLowering][AMDGPU][X86] Improve SimplifyDemandedBits bitcast handling.

rebase

Wed, Apr 17, 8:24 AM · Restricted Project
RKSimon updated the diff for D60625: [X86][SSE] Lower ICMP EQ(AND(X,C),C) -> SRA(SHL(X,LOG2(C)),BW-1) iff C is power-of-2..

Use isConstOrConstSplat

Wed, Apr 17, 4:59 AM · Restricted Project
RKSimon added inline comments to D60625: [X86][SSE] Lower ICMP EQ(AND(X,C),C) -> SRA(SHL(X,LOG2(C)),BW-1) iff C is power-of-2..
Wed, Apr 17, 4:59 AM · Restricted Project
RKSimon committed rG9daacec816da: [CostModel][X86] Add bool anyof/allof reduction costs (authored by RKSimon).
[CostModel][X86] Add bool anyof/allof reduction costs
Wed, Apr 17, 4:02 AM
RKSimon committed rL358574: [CostModel][X86] Add bool anyof/allof reduction costs.
[CostModel][X86] Add bool anyof/allof reduction costs
Wed, Apr 17, 4:02 AM
RKSimon closed D60403: [CostModel][X86] Add bool anyof/allof reduction costs.
Wed, Apr 17, 4:02 AM · Restricted Project

Tue, Apr 16

RKSimon updated the diff for D60403: [CostModel][X86] Add bool anyof/allof reduction costs.

Use Type::getVectorElementType()->isIntegerTy(1) - reduction types should always be vectors

Tue, Apr 16, 2:50 PM · Restricted Project
RKSimon updated the diff for D60403: [CostModel][X86] Add bool anyof/allof reduction costs.

Add codegen comments and use Type::isIntegerTy(1)

Tue, Apr 16, 2:41 PM · Restricted Project
RKSimon added a comment to D60403: [CostModel][X86] Add bool anyof/allof reduction costs.

ping

Tue, Apr 16, 2:10 PM · Restricted Project
RKSimon updated the diff for D60512: [X86][AVX] Combine shuffles extracted from a common vector.

rebase

Tue, Apr 16, 2:05 PM · Restricted Project
RKSimon committed rGe5573f4f4eba: [TargetLowering] Rename preferShiftsToClearExtremeBits and… (authored by RKSimon).
[TargetLowering] Rename preferShiftsToClearExtremeBits and…
Tue, Apr 16, 1:56 PM
RKSimon committed rL358526: [TargetLowering] Rename preferShiftsToClearExtremeBits and….
[TargetLowering] Rename preferShiftsToClearExtremeBits and…
Tue, Apr 16, 1:56 PM
RKSimon committed rG82ffa88a04a7: [SLP] Refactoring of the operand reordering code. (authored by RKSimon).
[SLP] Refactoring of the operand reordering code.
Tue, Apr 16, 12:26 PM
RKSimon committed rL358519: [SLP] Refactoring of the operand reordering code..
[SLP] Refactoring of the operand reordering code.
Tue, Apr 16, 12:25 PM
RKSimon closed D59973: [SLP] Refactoring of the operand reordering code..
Tue, Apr 16, 12:25 PM · Restricted Project
RKSimon committed rL358516: [X86][AVX] X86ISD::PERMV/PERMV3 node types can never fold index ops.
[X86][AVX] X86ISD::PERMV/PERMV3 node types can never fold index ops
Tue, Apr 16, 12:18 PM
RKSimon committed rGd769bb1e5836: [X86][AVX] X86ISD::PERMV/PERMV3 node types can never fold index ops (authored by RKSimon).
[X86][AVX] X86ISD::PERMV/PERMV3 node types can never fold index ops
Tue, Apr 16, 12:18 PM
RKSimon closed D60562: [X86][AVX] X86ISD::PERMV/PERMV3 node types can never fold index ops.
Tue, Apr 16, 12:18 PM · Restricted Project
RKSimon accepted D59973: [SLP] Refactoring of the operand reordering code..

Thank you for the reviews. @RKSimon any more comments ?

Tue, Apr 16, 1:59 AM · Restricted Project

Mon, Apr 15

RKSimon added inline comments to D60160: [PowerPC] Update P9 vector costs for insert/extract element.
Mon, Apr 15, 2:30 PM · Restricted Project
RKSimon added inline comments to D59744: Fix i386 ABI "__m64" type bug.
Mon, Apr 15, 10:20 AM · Restricted Project
RKSimon accepted D60674: [X86] Restore the pavg intrinsics..

LGTM - there's too many different optimizations and canonicalizations that can occur on such a pattern to be able to match all of the permutations.

Mon, Apr 15, 4:39 AM · Restricted Project, Restricted Project

Sat, Apr 13

RKSimon updated subscribers of rL356398: [AMDGPU] Asm/disasm v_cndmask_b32_e64 with abs/neg source modifiers.

@tpr Please can you look at the copy+paste error in https://bugs.llvm.org/show_bug.cgi?id=41488

Sat, Apr 13, 7:08 AM

Fri, Apr 12

RKSimon created D60625: [X86][SSE] Lower ICMP EQ(AND(X,C),C) -> SRA(SHL(X,LOG2(C)),BW-1) iff C is power-of-2..
Fri, Apr 12, 10:22 AM · Restricted Project
RKSimon added inline comments to D60462: [TargetLowering][AMDGPU][X86] Improve SimplifyDemandedBits bitcast handling.
Fri, Apr 12, 7:50 AM · Restricted Project
RKSimon committed rG6c8f4ada360d: [X86][SSE] Recognise vXi1 boolean anyof/allof reduction patterns (authored by RKSimon).
[X86][SSE] Recognise vXi1 boolean anyof/allof reduction patterns
Fri, Apr 12, 7:25 AM
RKSimon committed rL358286: [X86][SSE] Recognise vXi1 boolean anyof/allof reduction patterns.
[X86][SSE] Recognise vXi1 boolean anyof/allof reduction patterns
Fri, Apr 12, 7:21 AM
RKSimon closed D60610: [X86][SSE] Recognise vXi1 boolean anyof/allof reduction patterns.
Fri, Apr 12, 7:21 AM · Restricted Project
RKSimon commandeered D59912: [x86] flatten packss+movmsk into 256-bit movmsk.

After chatting with @spatel I'm going to deal with this as a followup to D60610

Fri, Apr 12, 7:06 AM · Restricted Project
RKSimon accepted D60522: [KnownBits] Add computeForAddCarry().

LGTM - let's leave the std::move issue for now

Fri, Apr 12, 7:06 AM · Restricted Project
RKSimon added a comment to D60610: [X86][SSE] Recognise vXi1 boolean anyof/allof reduction patterns.

LGTM - see inline for a minor potential follow-up.

If we can assume vXi1 IR, then I can probably abandon my recent movmsk efforts:
D59669
D59912

Fri, Apr 12, 6:55 AM · Restricted Project
RKSimon created D60610: [X86][SSE] Recognise vXi1 boolean anyof/allof reduction patterns.
Fri, Apr 12, 5:28 AM · Restricted Project
RKSimon updated the diff for D60562: [X86][AVX] X86ISD::PERMV/PERMV3 node types can never fold index ops.

Removed if-else chain

Fri, Apr 12, 4:13 AM · Restricted Project

Thu, Apr 11

RKSimon committed rG8d083c5e0bd3: [ConstantFold] ExtractConstantBytes - handle shifts on large integer types (authored by RKSimon).
[ConstantFold] ExtractConstantBytes - handle shifts on large integer types
Thu, Apr 11, 9:39 AM
RKSimon committed rL358192: [ConstantFold] ExtractConstantBytes - handle shifts on large integer types.
[ConstantFold] ExtractConstantBytes - handle shifts on large integer types
Thu, Apr 11, 9:39 AM
RKSimon updated the diff for D60562: [X86][AVX] X86ISD::PERMV/PERMV3 node types can never fold index ops.

rebase after rL358186 et al

Thu, Apr 11, 8:38 AM · Restricted Project
RKSimon committed rG40b647ae8e5f: [X86] SimplifyDemandedVectorElts - add X86ISD::VPERMV3 mask support (authored by RKSimon).
[X86] SimplifyDemandedVectorElts - add X86ISD::VPERMV3 mask support
Thu, Apr 11, 8:29 AM
RKSimon committed rL358186: [X86] SimplifyDemandedVectorElts - add X86ISD::VPERMV3 mask support.
[X86] SimplifyDemandedVectorElts - add X86ISD::VPERMV3 mask support
Thu, Apr 11, 8:29 AM
RKSimon committed rGa41275a39839: [X86][AVX] Tweak X86ISD::VPERMV3 demandedelts test (authored by RKSimon).
[X86][AVX] Tweak X86ISD::VPERMV3 demandedelts test
Thu, Apr 11, 8:10 AM
RKSimon committed rL358182: [X86][AVX] Tweak X86ISD::VPERMV3 demandedelts test.
[X86][AVX] Tweak X86ISD::VPERMV3 demandedelts test
Thu, Apr 11, 8:10 AM