llvm-project

History

[SLP] Initial support for ordered reductions

Patch models ordered reductions as a series of extractelements for the
cases which cannot be modeled as unordered reductions.

Fixes #50590

Reviewers: RKSimon, hiraditya

Pull Request: https://github.com/llvm/llvm-project/pull/182644

2026-03-20 13:45:14 -04:00

32-bit.ll

[llvm] Regenerate test checks including TBAA semantics (NFC)

2025-09-12 20:01:17 +02:00

64-bit-vector.ll

…

aarch64-vector-functions.ll

[SLP] Use the correct calling convention for vector math routines (#180759 )

2026-02-11 10:52:58 +00:00

accelerate-vector-functions-inseltpoison.ll

…

alternate-vectorization-split-node.ll

[SLP]Support LShr as base for copyable elements

2025-08-14 19:12:27 -07:00

buildvector-reduce.ll

…

buildvector-vectorize.ll

…

commute.ll

[SLP]Initial FMAD support (#149102 )

2025-08-11 05:53:55 -07:00

div-like-mixed-with-undefs.ll

[SLP][NFC]Fix UB and constant folded ops in test, NFC

2025-11-01 11:50:39 -07:00

div.ll

[SLP]Exclude non-profitable subtrees.

2025-12-30 09:32:05 -08:00

exp.ll

Reland [VectorUtils] Trivially vectorize ldexp, [l]lround (#152476 )

2025-08-07 12:07:29 +01:00

ext-trunc.ll

[AArch64,TTI] Remove RealUse check for vector insert/extract costs. (#146526 )

2025-07-15 15:19:27 +01:00

external-non-inst-use.ll

…

external-use-icmp.ll

[AArch64,TTI] Remove RealUse check for vector insert/extract costs. (#146526 )

2025-07-15 15:19:27 +01:00

externally-used-copyables.ll

[SLP]Prefer copyable over alternate

2026-03-20 11:59:59 -04:00

extract-subvector-long-input.ll

[AArch64,TTI] Remove RealUse check for vector insert/extract costs. (#146526 )

2025-07-15 15:19:27 +01:00

extractelements-to-shuffle.ll

[AArch64] use isTRNMask to calculate shuffle costs (#171524 )

2025-12-15 05:34:30 +00:00

extracts-from-scalarizable-vector.ll

…

fminimumnum.ll

…

fmulladd.ll

…

fpsat.ll

…

fround.ll

Reland [VectorUtils] Trivially vectorize ldexp, [l]lround (#152476 )

2025-08-07 12:07:29 +01:00

fshl.ll

…

gather-buildvector-with-minbitwidth-user.ll

[SLP]Exclude non-profitable subtrees.

2025-12-30 09:32:05 -08:00

gather-cost.ll

[SLP] Drop unnecessary '' from around -passes=... arg lists to appease update_test_checks.py when run on DOS. NFC.

2025-07-10 12:18:41 +01:00

gather-load-128.ll

…

gather-load-min-required-vf-2.ll

…

gather-reduce.ll

…

gather-root.ll

…

gather-with-minbith-user.ll

…

getelementptr2.ll

[AArch64,TTI] Remove RealUse check for vector insert/extract costs. (#146526 )

2025-07-15 15:19:27 +01:00

getelementptr.ll

[SLP] Loop aware cost model/tree building

2026-03-18 17:54:01 -04:00

horizontal.ll

[SLP] Loop aware cost model/tree building

2026-03-18 17:54:01 -04:00

insertelement-inseltpoison.ll

…

insertelement.ll

…

InstructionsState-is-invalid-0.ll

[SLP]Remove emission of vector_insert/vector_extract intrinsics

2025-07-11 15:26:45 -04:00

invalid_type.ll

…

landing_pad.ll

[SLP] Loop aware cost model/tree building

2026-03-18 17:54:01 -04:00

lit.local.cfg

…

load-store-q.ll

…

loadi8.ll

…

loadorder.ll

[SLP]Remove emission of vector_insert/vector_extract intrinsics

2025-07-11 15:26:45 -04:00

many-uses-fma-candidate.ll

[SLP]Improved/fixed FMAD support in reductions

2025-09-02 13:09:57 -07:00

masked-loads-side-effects-after-vec.ll

…

matmul.ll

…

memory-runtime-checks-in-loops.ll

…

memory-runtime-checks.ll

[AArch64,TTI] Remove RealUse check for vector insert/extract costs. (#146526 )

2025-07-15 15:19:27 +01:00

minbitwidth-cast-node-length.ll

…

minimum-sizes.ll

…

mismatched-intrinsics.ll

…

mul.ll

…

multiple_reduction.ll

[AArch64,TTI] Remove RealUse check for vector insert/extract costs. (#146526 )

2025-07-15 15:19:27 +01:00

non-inst-abs-sub-copyable-value.ll

[SLP] Loop aware cost model/tree building

2026-03-18 17:54:01 -04:00

nontemporal.ll

…

phi-node-bitwidt-op-not.ll

[AArch64,TTI] Remove RealUse check for vector insert/extract costs. (#146526 )

2025-07-15 15:19:27 +01:00

phi-use-order-scalable.ll

…

PR38339.ll

…

reduce-add-i64.ll

…

reduce-fadd.ll

[SLP]Remove emission of vector_insert/vector_extract intrinsics

2025-07-11 15:26:45 -04:00

reduce-maximumnum-minimumnum.ll

…

remarks.ll

…

reorder-fmuladd-crash.ll

…

reordered-loads.ll

…

reused-scalar-repeated-in-node.ll

[SLP] Loop aware cost model/tree building

2026-03-18 17:54:01 -04:00

revec-non-pow2.ll

[SLP]Fix an ArrayRef out-of-bounds access in slice

2026-02-17 10:00:13 -08:00

revec-reductions.ll

[SLP]Fix types for reductions in revec

2026-03-01 07:54:13 -08:00

revec-select-scalar-vector-cond.ll

[SLP]Do not vectorize select nodes with scalar and vector conditions

2026-03-01 07:01:04 -08:00

scalable-external-user.ll

…

scalable-vector.ll

[IR] Replace alignment argument with attribute on masked intrinsics (#163802 )

2025-10-20 08:50:09 +00:00

scalarization-overhead.ll

[SLP]Initial FMAD support (#149102 )

2025-08-11 05:53:55 -07:00

sdiv-pow2.ll

…

select-zext-analysis.ll

[SLP]Skip operands comparing on non-matching (but compatible) instructions

2026-02-11 04:55:29 -08:00

shuffle-vectors-mask-size.ll

[SLP]Enable float point math ops as copyables elements.

2025-12-26 11:55:58 -08:00

slp-abs.ll

…

slp-and-reduction.ll

…

slp-fma-loss.ll

[SLP] Loop aware cost model/tree building

2026-03-18 17:54:01 -04:00

slp-frem.ll

…

slp-or-reduction.ll

…

slp-xor-reduction.ll

…

spillcost-di.ll

[llvm] Regenerate test checks including TBAA semantics (NFC)

2025-09-12 20:01:17 +02:00

spillcost-order.ll

…

splat-loads.ll

[AArch64,TTI] Remove RealUse check for vector insert/extract costs. (#146526 )

2025-07-15 15:19:27 +01:00

store-ptr.ll

[llvm] Regenerate test checks including TBAA semantics (NFC)

2025-09-12 20:01:17 +02:00

transpose-inseltpoison.ll

…

transpose-with-constants.ll

[AArch64] use isTRNMask to calculate shuffle costs (#171524 )

2025-12-15 05:34:30 +00:00

transpose.ll

…

trunc-insertion.ll

[SLP] Loop aware cost model/tree building

2026-03-18 17:54:01 -04:00

tsc-s116.ll

…

tsc-s352.ll

[SLP] Initial support for ordered reductions

2026-03-20 13:45:14 -04:00

unique_values.ll

[SLP] Report the correct operand to getArithmeticInstrCost() when duplicated scalars (#174442 )

2026-01-05 22:25:25 +00:00

unreachable-blocks-with-phis.ll

[AArch64,TTI] Remove RealUse check for vector insert/extract costs. (#146526 )

2025-07-15 15:19:27 +01:00

unsigned-after-sext-node.ll

…

uselistorder.ll

…

user-node-not-in-bitwidths.ll

…

vec3-base.ll

…

vec3-calls.ll

[SLP]Remove emission of vector_insert/vector_extract intrinsics

2025-07-11 15:26:45 -04:00

vec3-reorder-reshuffle.ll

[AArch64] Change the cost of fma and fmuladd to match fmul. (#152963 )

2025-08-14 21:53:45 +01:00

vec15-base.ll

…

vec-node-dropped-with-gather-user.ll

[SLP][NFC]Add a test that reveals the issue with non-profitable subtrees deletion

2025-12-30 08:50:51 -08:00

vecreduceadd.ll

[AArch64] Improve the cost model for extending mull (#125651 )

2025-11-04 07:50:51 +00:00

vector-getelementptr.ll

[AArch64,TTI] Remove RealUse check for vector insert/extract costs. (#146526 )

2025-07-15 15:19:27 +01:00

vectorizable-selects-min-max.ll

[SLP] Drop unnecessary '' from around -passes=... arg lists to appease update_test_checks.py when run on DOS. NFC.

2025-07-10 12:18:41 +01:00

vectorizable-selects-uniform-cmps.ll

[AArch64,TTI] Remove RealUse check for vector insert/extract costs. (#146526 )

2025-07-15 15:19:27 +01:00

vectorize-free-extracts-inserts.ll

…

widen.ll

…