[llvm] 58d79aa - [AArch64] Guard against non-simple types in udiv sve costs. (#148580)
via llvm-commits
llvm-commits at lists.llvm.org
Tue Jul 15 02:25:11 PDT 2025
Author: David Green
Date: 2025-07-15T10:25:08+01:00
New Revision: 58d79aaba6765606f78fc6ec6b9ae1db3b44d834
URL: https://github.com/llvm/llvm-project/commit/58d79aaba6765606f78fc6ec6b9ae1db3b44d834
DIFF: https://github.com/llvm/llvm-project/commit/58d79aaba6765606f78fc6ec6b9ae1db3b44d834.diff
LOG: [AArch64] Guard against non-simple types in udiv sve costs. (#148580)
The code here probably needs to change to handle types more uniformly,
but this patch prevents it from trying to use a simple type where it does
not exist.
Fixes #148438.
Added:
Modified:
llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
llvm/test/Analysis/CostModel/AArch64/sve-div.ll
Removed:
################################################################################
diff --git a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
index 8904402065696..07baf29ce7016 100644
--- a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
@@ -4126,10 +4126,8 @@ InstructionCost AArch64TTIImpl::getArithmeticInstrCost(
if (TLI->isOperationLegalOrCustom(ISD, LT.second) && ST->hasSVE()) {
// SDIV/UDIV operations are lowered using SVE, then we can have less
// costs.
- if (isa<FixedVectorType>(Ty) && cast<FixedVectorType>(Ty)
- ->getPrimitiveSizeInBits()
- .getFixedValue() < 128) {
- EVT VT = TLI->getValueType(DL, Ty);
+ if (VT.isSimple() && isa<FixedVectorType>(Ty) &&
+ Ty->getPrimitiveSizeInBits().getFixedValue() < 128) {
static const CostTblEntry DivTbl[]{
{ISD::SDIV, MVT::v2i8, 5}, {ISD::SDIV, MVT::v4i8, 8},
{ISD::SDIV, MVT::v8i8, 8}, {ISD::SDIV, MVT::v2i16, 5},
diff --git a/llvm/test/Analysis/CostModel/AArch64/sve-div.ll b/llvm/test/Analysis/CostModel/AArch64/sve-div.ll
index c055d3218f65b..84a9ca0fab6d8 100644
--- a/llvm/test/Analysis/CostModel/AArch64/sve-div.ll
+++ b/llvm/test/Analysis/CostModel/AArch64/sve-div.ll
@@ -10,6 +10,7 @@ define void @sdiv() {
; CHECK-NEXT: Cost Model: Found costs of RThru:8 CodeSize:4 Lat:4 SizeLat:4 for: %V8i64 = sdiv <8 x i64> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:1 CodeSize:4 Lat:4 SizeLat:4 for: %V2i32 = sdiv <2 x i32> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:2 CodeSize:4 Lat:4 SizeLat:4 for: %V4i32 = sdiv <4 x i32> undef, undef
+; CHECK-NEXT: Cost Model: Found costs of 4 for: %V6i32 = sdiv <6 x i32> undef, undef
; CHECK-NEXT: Cost Model: Found costs of 4 for: %V8i32 = sdiv <8 x i32> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:8 CodeSize:4 Lat:4 SizeLat:4 for: %V16i32 = sdiv <16 x i32> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:5 CodeSize:4 Lat:4 SizeLat:4 for: %V2i16 = sdiv <2 x i16> undef, undef
@@ -19,6 +20,7 @@ define void @sdiv() {
; CHECK-NEXT: Cost Model: Found costs of RThru:32 CodeSize:4 Lat:4 SizeLat:4 for: %V32i16 = sdiv <32 x i16> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:5 CodeSize:4 Lat:4 SizeLat:4 for: %V2i8 = sdiv <2 x i8> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:8 CodeSize:4 Lat:4 SizeLat:4 for: %V4i8 = sdiv <4 x i8> undef, undef
+; CHECK-NEXT: Cost Model: Found costs of RThru:16 CodeSize:4 Lat:4 SizeLat:4 for: %V6i8 = sdiv <6 x i8> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:8 CodeSize:4 Lat:4 SizeLat:4 for: %V8i8 = sdiv <8 x i8> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:16 CodeSize:4 Lat:4 SizeLat:4 for: %V16i8 = sdiv <16 x i8> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:32 CodeSize:4 Lat:4 SizeLat:4 for: %V32i8 = sdiv <32 x i8> undef, undef
@@ -48,6 +50,7 @@ define void @sdiv() {
%V8i64 = sdiv <8 x i64> undef, undef
%V2i32 = sdiv <2 x i32> undef, undef
%V4i32 = sdiv <4 x i32> undef, undef
+ %V6i32 = sdiv <6 x i32> undef, undef
%V8i32 = sdiv <8 x i32> undef, undef
%V16i32 = sdiv <16 x i32> undef, undef
%V2i16 = sdiv <2 x i16> undef, undef
@@ -57,6 +60,7 @@ define void @sdiv() {
%V32i16 = sdiv <32 x i16> undef, undef
%V2i8 = sdiv <2 x i8> undef, undef
%V4i8 = sdiv <4 x i8> undef, undef
+ %V6i8 = sdiv <6 x i8> undef, undef
%V8i8 = sdiv <8 x i8> undef, undef
%V16i8 = sdiv <16 x i8> undef, undef
%V32i8 = sdiv <32 x i8> undef, undef
@@ -89,6 +93,7 @@ define void @udiv() {
; CHECK-NEXT: Cost Model: Found costs of RThru:8 CodeSize:4 Lat:4 SizeLat:4 for: %V8i64 = udiv <8 x i64> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:1 CodeSize:4 Lat:4 SizeLat:4 for: %V2i32 = udiv <2 x i32> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:2 CodeSize:4 Lat:4 SizeLat:4 for: %V4i32 = udiv <4 x i32> undef, undef
+; CHECK-NEXT: Cost Model: Found costs of 4 for: %V6i32 = udiv <6 x i32> undef, undef
; CHECK-NEXT: Cost Model: Found costs of 4 for: %V8i32 = udiv <8 x i32> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:8 CodeSize:4 Lat:4 SizeLat:4 for: %V16i32 = udiv <16 x i32> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:5 CodeSize:4 Lat:4 SizeLat:4 for: %V2i16 = udiv <2 x i16> undef, undef
@@ -98,6 +103,7 @@ define void @udiv() {
; CHECK-NEXT: Cost Model: Found costs of RThru:32 CodeSize:4 Lat:4 SizeLat:4 for: %V32i16 = udiv <32 x i16> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:5 CodeSize:4 Lat:4 SizeLat:4 for: %V2i8 = udiv <2 x i8> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:8 CodeSize:4 Lat:4 SizeLat:4 for: %V4i8 = udiv <4 x i8> undef, undef
+; CHECK-NEXT: Cost Model: Found costs of RThru:16 CodeSize:4 Lat:4 SizeLat:4 for: %V6i8 = udiv <6 x i8> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:8 CodeSize:4 Lat:4 SizeLat:4 for: %V8i8 = udiv <8 x i8> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:16 CodeSize:4 Lat:4 SizeLat:4 for: %V16i8 = udiv <16 x i8> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:32 CodeSize:4 Lat:4 SizeLat:4 for: %V32i8 = udiv <32 x i8> undef, undef
@@ -127,6 +133,7 @@ define void @udiv() {
%V8i64 = udiv <8 x i64> undef, undef
%V2i32 = udiv <2 x i32> undef, undef
%V4i32 = udiv <4 x i32> undef, undef
+ %V6i32 = udiv <6 x i32> undef, undef
%V8i32 = udiv <8 x i32> undef, undef
%V16i32 = udiv <16 x i32> undef, undef
%V2i16 = udiv <2 x i16> undef, undef
@@ -136,6 +143,7 @@ define void @udiv() {
%V32i16 = udiv <32 x i16> undef, undef
%V2i8 = udiv <2 x i8> undef, undef
%V4i8 = udiv <4 x i8> undef, undef
+ %V6i8 = udiv <6 x i8> undef, undef
%V8i8 = udiv <8 x i8> undef, undef
%V16i8 = udiv <16 x i8> undef, undef
%V32i8 = udiv <32 x i8> undef, undef
More information about the llvm-commits
mailing list