54ed459e3e
The wrapper used to hold the handle for resource type has just the default copy constructor and assignment operator. This causes clang to insert memcpys when it does an assignment of a resource type. The memcpy then cause optimizations to fail when the memcpy is turned into a load and store of an i64. To fix this, we should define copying of a resource type by adding the operator= and copy constructor. Partially fixes #154669
63 lines
3.2 KiB
HLSL
63 lines
3.2 KiB
HLSL
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.6-compute -finclude-default-header \
|
|
// RUN: -emit-llvm -disable-llvm-passes -o - %s | FileCheck %s
|
|
|
|
// This test verifies handling of local arrays of resources when used
|
|
// as a function argument that is modified inside the function.
|
|
|
|
// CHECK: @_ZL1X = internal global %"class.hlsl::RWBuffer" poison, align 4
|
|
// CHECK: @_ZL1Y = internal global %"class.hlsl::RWBuffer" poison, align 4
|
|
|
|
RWBuffer<int> X : register(u0);
|
|
RWBuffer<int> Y : register(u1);
|
|
|
|
// CHECK: define {{.*}} @_Z6SomeFnA2_N4hlsl8RWBufferIiEEji(
|
|
// CHECK-SAME: ptr noundef byval([2 x %"class.hlsl::RWBuffer"]) align 4 %B, i32 noundef %Idx, i32 noundef %Val0)
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: %[[Idx_addr:.*]] = alloca i32, align 4
|
|
// CHECK-NEXT: %[[Val0_addr:.*]] = alloca i32, align 4
|
|
// CHECK-NEXT: store i32 %Idx, ptr %[[Idx_addr]], align 4
|
|
// CHECK-NEXT: store i32 %Val0, ptr %[[Val0_addr]], align 4
|
|
void SomeFn(RWBuffer<int> B[2], uint Idx, int Val0) {
|
|
|
|
// CHECK-NEXT: %[[B_0_Ptr:.*]] = getelementptr inbounds [2 x %"class.hlsl::RWBuffer"], ptr %B, i32 0, i32 0
|
|
// CHECK-NEXT: call {{.*}} @_ZN4hlsl8RWBufferIiEaSERKS1_(ptr {{.*}} %[[B_0_Ptr]], ptr {{.*}} @_ZL1Y)
|
|
B[0] = Y;
|
|
|
|
// NOTE: _ZN4hlsl8RWBufferIiEixEj is the subscript operator for RWBuffer<int>
|
|
|
|
// CHECK-NEXT: %[[Val0:.*]] = load i32, ptr %[[Val0_addr]], align 4
|
|
// CHECK-NEXT: %[[B_0_Ptr:.*]] = getelementptr inbounds [2 x %"class.hlsl::RWBuffer"], ptr %B, i32 0, i32 0
|
|
// CHECK-NEXT: %[[Idx:.*]] = load i32, ptr %[[Idx_addr]], align 4
|
|
// CHECK-NEXT: %[[BufPtr:.*]] = call {{.*}} ptr @_ZN4hlsl8RWBufferIiEixEj(ptr {{.*}} %[[B_0_Ptr]], i32 noundef %[[Idx]])
|
|
// CHECK-NEXT: store i32 %[[Val0]], ptr %[[BufPtr]], align 4
|
|
B[0][Idx] = Val0;
|
|
}
|
|
|
|
// CHECK: define {{.*}} void @_Z4mainj(i32 noundef %GI)
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: %[[GI_addr:.*]] = alloca i32, align 4
|
|
[numthreads(4,1,1)]
|
|
void main(uint GI : SV_GroupIndex) {
|
|
// CHECK-NEXT: %A = alloca [2 x %"class.hlsl::RWBuffer"], align 4
|
|
// CHECK-NEXT: %[[Tmp:.*]] = alloca [2 x %"class.hlsl::RWBuffer"], align 4
|
|
// CHECK-NEXT: store i32 %GI, ptr %GI.addr, align 4
|
|
|
|
// Initialization of array A with resources X and Y
|
|
// CHECK-NEXT: call void @_ZN4hlsl8RWBufferIiEC1ERKS1_(ptr {{.*}} %A, ptr {{.*}} @_ZL1X)
|
|
// CHECK-NEXT: %[[A_1_Ptr:.*]] = getelementptr inbounds %"class.hlsl::RWBuffer", ptr %A, i32 1
|
|
// CHECK-NEXT: call void @_ZN4hlsl8RWBufferIiEC1ERKS1_(ptr {{.*}} %[[A_1_Ptr]], ptr {{.*}} @_ZL1Y)
|
|
RWBuffer<int> A[2] = {X, Y};
|
|
|
|
// Verify that SomeFn is called with a local copy of the array A
|
|
// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 4 %[[Tmp]], ptr align 4 %A, i32 8, i1 false)
|
|
// CHECK-NEXT: %[[GI:.*]] = load i32, ptr %[[GI_addr]], align 4
|
|
// CHECK-NEXT: call void @_Z6SomeFnA2_N4hlsl8RWBufferIiEEji(ptr noundef byval([2 x %"class.hlsl::RWBuffer"]) align 4 %[[Tmp]], i32 noundef %[[GI]], i32 noundef 1)
|
|
SomeFn(A, GI, 1);
|
|
|
|
// CHECK-NEXT: %[[A_0_Ptr:.*]] = getelementptr inbounds [2 x %"class.hlsl::RWBuffer"], ptr %A, i32 0, i32 0
|
|
// CHECK-NEXT: %[[GI:.*]] = load i32, ptr %[[GI_addr]], align 4
|
|
// CHECK-NEXT: %[[BufPtr:.*]] = call {{.*}} ptr @_ZN4hlsl8RWBufferIiEixEj(ptr {{.*}} %[[A_0_Ptr]], i32 noundef %[[GI]])
|
|
// CHECK-NEXT: store i32 2, ptr %[[BufPtr]], align 4
|
|
A[0][GI] = 2;
|
|
}
|