Simon Pilgrim 24d9550b27

[VectorCombine] foldShuffleOfBinops - if both operands are the same don't duplicate the total new cost (#172719) ...

If we're shuffling/concatenating the same operands then ensure we don't
duplicate the total cost, ensure we reuse the final shuffle and
recognise that we reduce the total instruction count (so fold even when
NewCost == OldCost, not just NewCost < OldCost).

2025-12-18 07:03:06 +00:00

..

AArch64

[VectorCombine] Fold permute of intrinsics into intrinsic of permutes: shuffle(intrinsic, poison/undef) -> intrinsic(shuffle) (#170052)

2025-12-05 15:54:53 +00:00

AMDGPU

VectorCombine: Improve the insert/extract fold in the narrowing case (#168820)

2025-12-15 11:25:51 -08:00

ARM

[VectorCombine][AMDGPU] Narrow Phi of Shuffles. (#140188)

2025-08-12 18:45:11 +01:00

Hexagon

[VectorCombine][AMDGPU] Narrow Phi of Shuffles. (#140188)

2025-08-12 18:45:11 +01:00

PowerPC

[VectorCombine] Fix scalarizeExtExtract for big-endian (#157962)

2025-09-15 10:08:16 +00:00

RISCV

[VectorCombine][AMDGPU] Narrow Phi of Shuffles. (#140188)

2025-08-12 18:45:11 +01:00

SPIRV

[VectorCombine][TTI] Prevent extract/ins rewrite to GEP (#150216)

2025-07-31 14:14:00 +02:00

X86

[VectorCombine] foldShuffleOfBinops - if both operands are the same don't duplicate the total new cost (#172719)

2025-12-18 07:03:06 +00:00

alias.ll

[VectorCombine] Preserve scoped alias metadata (#153714)

2025-08-18 18:16:32 +00:00

binop-scalarize.ll

[VectorCombine] Apply InstSimplify in scalarizeOpOrCmp to avoid infinite loop (#153069)

2025-08-15 18:38:04 +00:00

fold-binop-of-reductions.ll

…

fold-shuffle-chains-to-reduce.ll

[VectorCombine] New folding pattern for extract/binop/shuffle chains (#145232)

2025-08-24 14:21:48 +05:30

intrinsic-scalarize.ll

[IR] NFC: Remove 'experimental' from partial.reduce.add intrinsic (#158637)

2025-09-17 11:44:47 +01:00

load-insert-store.ll

…

load-shufflevector.ll

[VectorCombine] Shrink loads used in shufflevector rebroadcasts. (#153138)

2025-08-12 14:08:37 +01:00

narrow-phi-of-shuffles.ll

[VectorCombine][AMDGPU] Narrow Phi of Shuffles. (#140188)

2025-08-12 18:45:11 +01:00

pr88796.ll

…

pr132563.ll

…

pr157335.ll

[VectorCombine] foldSelectShuffle - early-out cases where the max vector register width isn't large enough (#157430)

2025-09-08 12:04:23 +00:00

unary-op-scalarize.ll

[VectorCombine] Support nary operands and intrinsics in scalarizeOpOrCmp (#138406)

2025-05-28 09:45:54 +01:00