1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-18 18:42:46 +02:00

Reapply [ConstantFold] Handle vectors in ConstantFoldLoadThroughBitcast()

There seems to be an impedance mismatch between what the type
system considers an aggregate (structs and arrays) and what
constants consider an aggregate (structs, arrays and vectors).

Adjust the type check to consider vectors as well. The previous
version of the patch dropped the type check entirely, but it
turns out that getAggregateElement() does require the constant
to be an aggregate in some edge cases: For Poison/Undef the
getNumElements() API is called, without checking in advance that
we're dealing with an aggregate. Possibly the implementation should
avoid doing that, but for now I'm adding an assert so the next
person doesn't fall into this trap.
This commit is contained in:
Nikita Popov 2021-03-06 12:14:48 +01:00
parent b3cc8585de
commit 4766db4e38
4 changed files with 6 additions and 7 deletions

View File

@ -389,7 +389,7 @@ Constant *llvm::ConstantFoldLoadThroughBitcast(Constant *C, Type *DestTy,
// If this isn't an aggregate type, there is nothing we can do to drill down
// and find a bitcastable constant.
if (!SrcTy->isAggregateType())
if (!SrcTy->isAggregateType() && !SrcTy->isVectorTy())
return nullptr;
// We're simulating a load through a pointer that was bitcast to point to

View File

@ -419,6 +419,9 @@ Constant *Constant::getAllOnesValue(Type *Ty) {
}
Constant *Constant::getAggregateElement(unsigned Elt) const {
assert((getType()->isAggregateType() || getType()->isVectorTy()) &&
"Must be an aggregate/vector constant");
if (const auto *CC = dyn_cast<ConstantAggregate>(this))
return Elt < CC->getNumOperands() ? CC->getOperand(Elt) : nullptr;

View File

@ -213,14 +213,12 @@ entry:
ret i64 addrspace(4)* %ref
}
; TODO: missed optimization
define i8 addrspace(4)* @forward_memcopy(i8 addrspace(4)* addrspace(4)* %loc) {
; CHECK-LABEL: @forward_memcopy(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[LOC_BC:%.*]] = bitcast i8 addrspace(4)* addrspace(4)* [[LOC:%.*]] to i8 addrspace(4)*
; CHECK-NEXT: call void @llvm.memcpy.p4i8.p0i8.i64(i8 addrspace(4)* align 4 [[LOC_BC]], i8* bitcast (<4 x i64 addrspace(4)*>* @NonZeroConstant2 to i8*), i64 8, i1 false)
; CHECK-NEXT: [[REF:%.*]] = load i8 addrspace(4)*, i8 addrspace(4)* addrspace(4)* [[LOC]], align 8
; CHECK-NEXT: ret i8 addrspace(4)* [[REF]]
; CHECK-NEXT: ret i8 addrspace(4)* bitcast (i64 addrspace(4)* getelementptr (i64, i64 addrspace(4)* null, i32 3) to i8 addrspace(4)*)
;
entry:
%loc.bc = bitcast i8 addrspace(4)* addrspace(4)* %loc to i8 addrspace(4)*

View File

@ -213,14 +213,12 @@ entry:
ret i64 addrspace(4)* %ref
}
; TODO: missed optimization
define i8 addrspace(4)* @forward_memcopy(i8 addrspace(4)* addrspace(4)* %loc) {
; CHECK-LABEL: @forward_memcopy(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[LOC_BC:%.*]] = bitcast i8 addrspace(4)* addrspace(4)* [[LOC:%.*]] to i8 addrspace(4)*
; CHECK-NEXT: call void @llvm.memcpy.p4i8.p0i8.i64(i8 addrspace(4)* align 4 [[LOC_BC]], i8* bitcast (<4 x i64 addrspace(4)*>* @NonZeroConstant2 to i8*), i64 8, i1 false)
; CHECK-NEXT: [[REF:%.*]] = load i8 addrspace(4)*, i8 addrspace(4)* addrspace(4)* [[LOC]], align 8
; CHECK-NEXT: ret i8 addrspace(4)* [[REF]]
; CHECK-NEXT: ret i8 addrspace(4)* bitcast (i64 addrspace(4)* getelementptr (i64, i64 addrspace(4)* null, i32 3) to i8 addrspace(4)*)
;
entry:
%loc.bc = bitcast i8 addrspace(4)* addrspace(4)* %loc to i8 addrspace(4)*