Skip to content

Commit 8bf619a

Browse files
committed
[RISCV] Remove fixed vector constraint on masked interleave costing
The last of the fixed vector changes for interleaves with masked loads and stores landed last night, so we no longer need the restriction.
1 parent adb2421 commit 8bf619a

File tree

1 file changed

+1
-3
lines changed

1 file changed

+1
-3
lines changed

llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp

Lines changed: 1 addition & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -983,9 +983,7 @@ InstructionCost RISCVTTIImpl::getInterleavedMemoryOpCost(
983983
// with an adjacent appropriate memory to vlseg/vsseg intrinsics. vlseg/vsseg
984984
// only support masking per-iteration (i.e. condition), not per-segment (i.e.
985985
// gap).
986-
// TODO: Support masked interleaved access for fixed length vector.
987-
if ((isa<ScalableVectorType>(VecTy) || !UseMaskForCond) && !UseMaskForGaps &&
988-
Factor <= TLI->getMaxSupportedInterleaveFactor()) {
986+
if (!UseMaskForGaps && Factor <= TLI->getMaxSupportedInterleaveFactor()) {
989987
auto *VTy = cast<VectorType>(VecTy);
990988
std::pair<InstructionCost, MVT> LT = getTypeLegalizationCost(VTy);
991989
// Need to make sure type has't been scalarized

0 commit comments

Comments
 (0)