llvm-project

History

[LV] Fix cost misaligned when gather/scatter w/ addr is uniform. (#157387 )

This patch fix the assertion when the `isUniform` (from legacy model)
and `isSingleScalar`(from Vplan-based model) mismatch.

The simplify test that cause assertion
```
loop:
  loadA = load %a  => %a is loop invariant.
  loadB = load %LoadA
  ...
```
In the legacy cost model, it cannot analysis that addr of `%loadB` is
uniform but in the Vplan-based cost model both addr in `%loadA` and
`loadB` is single scalar.

Full test caused crash: https://llvm.godbolt.org/z/zEG8YKjqh.

---------

Co-authored-by: Luke Lau <luke@igalia.com>

2025-09-11 07:49:54 +08:00

bf16.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

blend-any-of-reduction-cost.ll

[RISCV] Enable tail folding by default (#151681 )

2025-08-08 14:26:23 +08:00

blocks-with-dead-instructions.ll

[VPlan] Reassociate (x & y) & z -> x & (y & z) (#155383 )

2025-09-03 01:09:19 +00:00

dead-ops-cost.ll

[LV][EVL] Support interleaved access with tail folding by EVL (#152070 )

2025-09-01 21:20:06 +08:00

defaults.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

divrem.ll

[VPlan] Fold common edges away in convertPhisToBlends (#150368 )

2025-09-01 07:03:33 +00:00

evl-compatible-loops.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

evl-iv-simplify.ll

[LV][EVL] Introduce the EVLIndVarSimplify Pass for EVL-vectorized loops (#131005 )

2025-05-14 13:49:50 -07:00

f16.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

first-order-recurrence-scalable-vf1.ll

[VPlan] Add dedicated user for resume phi with epilogue vectorization.

2025-08-10 21:21:16 +01:00

fminimumnum.ll

[VPlan] Add VPlan-based addMinIterCheck, replace ILV for non-epilogue. (#153643 )

2025-08-26 15:52:31 +01:00

force-vect-msg.ll

[LV][RISCV] Try fixing Windows buildbot failure in force-vect-msg.ll. NFC

2025-08-09 21:09:36 +08:00

gather-scatter-cost.ll

[LV] Fix cost misaligned when gather/scatter w/ addr is uniform. (#157387 )

2025-09-11 07:49:54 +08:00

illegal-type.ll

[VPlan] Don't access canonical IV in VPWidenPointerInduction::execute.

2025-03-15 21:32:48 +00:00

induction-costs.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

inloop-reduction.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

interleaved-accesses-zve32x.ll

…

interleaved-accesses.ll

[LV][EVL] Support interleaved access with tail folding by EVL (#152070 )

2025-09-01 21:20:06 +08:00

interleaved-cost.ll

[RISCV] Enable tail folding by default (#151681 )

2025-08-08 14:26:23 +08:00

interleaved-masked-access.ll

[LV][EVL] Support interleaved access with tail folding by EVL (#152070 )

2025-09-01 21:20:06 +08:00

interleaved-store-with-gap.ll

[LV] Fix gap mask requirement for interleaved access (#151105 )

2025-08-01 14:24:30 +08:00

lit.local.cfg

…

lmul.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

low-trip-count.ll

Reapply "[VPlan] Remove trivial dead VPPhi cycles."

2025-08-12 22:09:30 +01:00

mask-index-type.ll

[VPlan] Fold common edges away in convertPhisToBlends (#150368 )

2025-09-01 07:03:33 +00:00

masked_gather_scatter.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

only-compute-cost-for-vplan-vfs.ll

[VPlan] Support VPWidenPointerInductionRecipes with EVL tail folding (#152110 )

2025-08-07 10:54:24 +08:00

ordered-reduction.ll

[VPlan] Remove VPVectorPointer for part 0 after unrolling. (#149735 )

2025-07-27 13:53:26 +01:00

partial-reduce-dot-product.ll

[LV] Use shl for ((VF * Step) * vscale) in createStepForVF. (#153495 )

2025-08-14 19:27:51 +01:00

pr87378-vpinstruction-or-drop-poison-generating-flags.ll

[VPlan] Move findCommonEdgeMask optimization to simplifyBlends (#156304 )

2025-09-05 01:29:22 +00:00

pr88802.ll

[VPlan] Fold common edges away in convertPhisToBlends (#150368 )

2025-09-01 07:03:33 +00:00

pr154103.ll

[Reland] "[RISCV][TTI] Implement getAddressComputationCost() in RISCV TTI. #149955 " (#156386 )

2025-09-02 11:43:27 +08:00

preserve-dbg-loc.ll

[RISCV][LV] Use predicate-else-scalar-epilogue flag in tail folding tests. NFC

2025-07-28 17:04:00 +08:00

reductions.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

reg-usage-bf16.ll

[RISCV] Enable tail folding by default (#151681 )

2025-08-08 14:26:23 +08:00

reg-usage-f16.ll

[RISCV] Enable tail folding by default (#151681 )

2025-08-08 14:26:23 +08:00

reg-usage-maxbandwidth.ll

[LV][VPlan] Reduce register usage of VPEVLBasedIVPHIRecipe. (#154482 )

2025-08-21 07:39:01 +08:00

reg-usage.ll

[RISCV] Enable tail folding by default (#151681 )

2025-08-08 14:26:23 +08:00

remark-reductions.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

riscv-interleaved.ll

[RISCV] Remove -riscv-v-vector-bits-min from LoopVectorize tests. NFC (#148565 )

2025-07-14 21:59:35 +08:00

riscv-unroll.ll

[VPlan] Remove VPVectorPointer for part 0 after unrolling. (#149735 )

2025-07-27 13:53:26 +01:00

riscv-vector-reverse.ll

Reapply "[SCEVExp] Add helper to clean up dead instructions after expansion. (#157308 )"

2025-09-09 09:47:41 +01:00

safe-dep-distance.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

scalable-basics.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

scalable-tailfold.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

scalable-vf-hint.ll

…

select-cmp-reduction.ll

[VPlan] Fold common edges away in convertPhisToBlends (#150368 )

2025-09-01 07:03:33 +00:00

select-invariant-cond-cost.ll

…

short-trip-count.ll

[RISCV] Enable tail folding by default (#151681 )

2025-08-08 14:26:23 +08:00

strided-accesses.ll

[LV] Improve the test coverage for strided access. nfc (#155981 )

2025-09-03 10:19:36 +00:00

tail-folding-bin-unary-ops-args.ll

[VPlan] Add VPlan-based addMinIterCheck, replace ILV for non-epilogue. (#153643 )

2025-08-26 15:52:31 +01:00

tail-folding-call-intrinsics.ll

[VPlan] Add VPlan-based addMinIterCheck, replace ILV for non-epilogue. (#153643 )

2025-08-26 15:52:31 +01:00

tail-folding-cast-intrinsics.ll

[VPlan] Add VPlan-based addMinIterCheck, replace ILV for non-epilogue. (#153643 )

2025-08-26 15:52:31 +01:00

tail-folding-cond-reduction.ll

[VPlan] Fold common edges away in convertPhisToBlends (#150368 )

2025-09-01 07:03:33 +00:00

tail-folding-cost.ll

[RISCV][LV] Remove redundant -force-tail-folding-style from tests. NFC

2025-07-28 16:11:01 +08:00

tail-folding-div.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

tail-folding-fixed-order-recurrence.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

tail-folding-gather-scatter.ll

[Reland] "[RISCV][TTI] Implement getAddressComputationCost() in RISCV TTI. #149955 " (#156386 )

2025-09-02 11:43:27 +08:00

tail-folding-inloop-reduction.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

tail-folding-interleave.ll

[Reland] "[RISCV][TTI] Implement getAddressComputationCost() in RISCV TTI. #149955 " (#156386 )

2025-09-02 11:43:27 +08:00

tail-folding-intermediate-store.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

tail-folding-iv32.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

tail-folding-known-no-overflow.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

tail-folding-masked-loadstore.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

tail-folding-no-masking.ll

[RISCV][LV] Use predicate-else-scalar-epilogue flag in tail folding tests. NFC

2025-07-28 17:04:00 +08:00

tail-folding-ordered-reduction.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

tail-folding-reduction-cost.ll

[RISCV][LV] Use predicate-else-scalar-epilogue flag in tail folding tests. NFC

2025-07-28 17:04:00 +08:00

tail-folding-reduction.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

tail-folding-reverse-load-store.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

tail-folding-safe-dep-distance.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

tail-folding-uniform-store.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

truncate-to-minimal-bitwidth-cost.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

truncate-to-minimal-bitwidth-evl-crash.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

type-info-cache-evl-crash.ll

[VPlan] Simplify x && false -> false, x | 0 -> x (#156345 )

2025-09-04 10:29:59 +01:00

uniform-load-store.ll

[VPlan] Fold common edges away in convertPhisToBlends (#150368 )

2025-09-01 07:03:33 +00:00

unroll-in-loop-vectorizer.ll

[RISCV] Remove -riscv-v-vector-bits-min from LoopVectorize tests. NFC (#148565 )

2025-07-14 21:59:35 +08:00

veclib-function-calls.ll

…

vector-loop-backedge-elimination-with-evl.ll

Reapply "[VPlan] Remove trivial dead VPPhi cycles."

2025-08-12 22:09:30 +01:00

vectorize-vp-intrinsics.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

vf-will-not-generate-any-vector-insts.ll

[VPlan] Replace EVL branch condition with (branch-on-count AVLNext, 0) (#152167 )

2025-08-26 11:19:19 +00:00

vplan-riscv-vector-reverse.ll

[RISCV] Enable tail folding by default (#151681 )

2025-08-08 14:26:23 +08:00

vplan-vp-call-intrinsics.ll

[VPlan] Create AVL as a phi from TC -> 0 with EVL tail folding (#151481 )

2025-08-01 11:00:05 +08:00

vplan-vp-cast-intrinsics.ll

[VPlan] Create AVL as a phi from TC -> 0 with EVL tail folding (#151481 )

2025-08-01 11:00:05 +08:00

vplan-vp-intrinsics-fixed-order-recurrence.ll

[VPlan] Create AVL as a phi from TC -> 0 with EVL tail folding (#151481 )

2025-08-01 11:00:05 +08:00

vplan-vp-intrinsics-reduction.ll

[VPlan] Create AVL as a phi from TC -> 0 with EVL tail folding (#151481 )

2025-08-01 11:00:05 +08:00

vplan-vp-intrinsics.ll

[VPlan] Create AVL as a phi from TC -> 0 with EVL tail folding (#151481 )

2025-08-01 11:00:05 +08:00

zvl32b.ll

…