RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-12-04 09:45:00 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

[ConstantFold][SVE] Fix constant folding for scalable vector compare instruction.

Browse Source 
Summary:
Do not iterate on scalable vector. Also do not return constant scalable vector
from ConstantInt::get().
Fix result type by using getElementCount() instead of getNumElements().

Reviewers: sdesmalen, efriedma, apazos, huntergr, willlovett

Reviewed By: efriedma

Subscribers: tschuett, hiraditya, rkruppe, psnobl, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D73753
This commit is contained in:
Huihui Zhang 2020-03-12 15:59:14 -07:00
parent 997e59690c
commit 2fab04a4e4
3 changed files with 56 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
lib/IR/ConstantFold.cpp
View File

@ -1840,7 +1840,7 @@ Constant *llvm::ConstantFoldCompareInstruction(unsigned short pred,
Type *ResultTy;
if (VectorType *VT = dyn_cast<VectorType>(C1->getType()))
ResultTy = VectorType::get(Type::getInt1Ty(C1->getContext()),
VT->getNumElements());
VT->getElementCount());
else
ResultTy = Type::getInt1Ty(C1->getContext());
@ -1971,6 +1971,11 @@ Constant *llvm::ConstantFoldCompareInstruction(unsigned short pred,
R==APFloat::cmpEqual);
}
} else if (C1->getType()->isVectorTy()) {
// Do not iterate on scalable vector. The number of elements is unknown at
// compile-time.
if (C1->getType()->getVectorIsScalable())
return nullptr;
// If we can constant fold the comparison of each element, constant fold
// the whole vector comparison.
SmallVector<Constant*, 4> ResElts;

4
lib/IR/Constants.cpp
View File

@ -2175,7 +2175,7 @@ Constant *ConstantExpr::getICmp(unsigned short pred, Constant *LHS,
Type *ResultTy = Type::getInt1Ty(LHS->getContext());
if (VectorType *VT = dyn_cast<VectorType>(LHS->getType()))
ResultTy = VectorType::get(ResultTy, VT->getNumElements());
ResultTy = VectorType::get(ResultTy, VT->getElementCount());
LLVMContextImpl *pImpl = LHS->getType()->getContext().pImpl;
return pImpl->ExprConstants.getOrCreate(ResultTy, Key);
@ -2200,7 +2200,7 @@ Constant *ConstantExpr::getFCmp(unsigned short pred, Constant *LHS,
Type *ResultTy = Type::getInt1Ty(LHS->getContext());
if (VectorType *VT = dyn_cast<VectorType>(LHS->getType()))
ResultTy = VectorType::get(ResultTy, VT->getNumElements());
ResultTy = VectorType::get(ResultTy, VT->getElementCount());
LLVMContextImpl *pImpl = LHS->getType()->getContext().pImpl;
return pImpl->ExprConstants.getOrCreate(ResultTy, Key);

48
test/Analysis/ConstantFolding/vscale.ll
View File

@ -235,3 +235,51 @@ define <vscale x 16 x i8> @call() {
%r = call <vscale x 16 x i8> @llvm.sadd.sat.nxv16i8(<vscale x 16 x i8> undef, <vscale x 16 x i8> undef)
ret <vscale x 16 x i8> %r
}
define <vscale x 4 x i1> @icmp_undef() {
; CHECK-LABEL: @icmp_undef(
; CHECK-NEXT: ret <vscale x 4 x i1> undef
;
%r = icmp eq <vscale x 4 x i32> undef, undef
ret <vscale x 4 x i1> %r
}
define <vscale x 4 x i1> @icmp_zero() {
; CHECK-LABEL: @icmp_zero(
; CHECK-NEXT: ret <vscale x 4 x i1> icmp eq (<vscale x 4 x i32> zeroinitializer, <vscale x 4 x i32> zeroinitializer)
;
%r = icmp eq <vscale x 4 x i32> zeroinitializer, zeroinitializer
ret <vscale x 4 x i1> %r
}
define <vscale x 4 x i1> @fcmp_true() {
; CHECK-LABEL: @fcmp_true(
; CHECK-NEXT: ret <vscale x 4 x i1> shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> undef, i1 true, i32 0), <vscale x 4 x i1> undef, <vscale x 4 x i32> zeroinitializer)
;
%r = fcmp true <vscale x 4 x float> undef, undef
ret <vscale x 4 x i1> %r
}
define <vscale x 4 x i1> @fcmp_false() {
; CHECK-LABEL: @fcmp_false(
; CHECK-NEXT: ret <vscale x 4 x i1> zeroinitializer
;
%r = fcmp false <vscale x 4 x float> undef, undef
ret <vscale x 4 x i1> %r
}
define <vscale x 4 x i1> @fcmp_undef() {
; CHECK-LABEL: @fcmp_undef(
; CHECK-NEXT: ret <vscale x 4 x i1> undef
;
%r = icmp ne <vscale x 4 x i32> undef, undef
ret <vscale x 4 x i1> %r
}
define <vscale x 4 x i1> @fcmp_not_equality() {
; CHECK-LABEL: @fcmp_not_equality(
; CHECK-NEXT: ret <vscale x 4 x i1> shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> undef, i1 true, i32 0), <vscale x 4 x i1> undef, <vscale x 4 x i32> zeroinitializer)
;
%r = icmp ule <vscale x 4 x i32> undef, zeroinitializer
ret <vscale x 4 x i1> %r
}