llvm-project/clang/test/CodeGenCXX/blocks-cxx11.cpp

115 lines
3.9 KiB
C++

// RUN: %clang_cc1 %s -fblocks -triple x86_64-apple-darwin -std=c++11 -emit-llvm -o - | FileCheck %s
template <class T> void takeItByValue(T);
void takeABlock(void (^)());
// rdar://problem/11022704
namespace test_int {
void test() {
const int x = 100;
takeABlock(^{ takeItByValue(x); });
// CHECK: call void @_Z13takeItByValueIiEvT_(i32 100)
}
}
namespace test_int_ref {
void test() {
const int y = 200;
const int &x = y;
takeABlock(^{ takeItByValue(x); });
// TODO: there's no good reason that this isn't foldable.
// CHECK: call void @_Z13takeItByValueIiEvT_(i32 {{%.*}})
}
}
namespace test_float {
void test() {
const float x = 1;
takeABlock(^{ takeItByValue(x); });
// CHECK: call void @_Z13takeItByValueIfEvT_(float 1.0
}
}
namespace test_float_ref {
void test() {
const float y = 100;
const float &x = y;
takeABlock(^{ takeItByValue(x); });
// TODO: there's no good reason that this isn't foldable.
// CHECK: call void @_Z13takeItByValueIfEvT_(float {{%.*}})
}
}
namespace test_complex_int {
void test() {
constexpr _Complex int x = 500;
takeABlock(^{ takeItByValue(x); });
// CHECK: store { i32, i32 } { i32 500, i32 0 },
// CHECK: store i32 500,
// CHECK-NEXT: store i32 0,
// CHECK-NEXT: [[COERCE:%.*]] = bitcast
// CHECK-NEXT: [[CVAL:%.*]] = load i64, i64* [[COERCE]]
// CHECK-NEXT: call void @_Z13takeItByValueICiEvT_(i64 [[CVAL]])
}
}
namespace test_complex_int_ref {
void test() {
const _Complex int y = 100;
const _Complex int &x = y;
takeABlock(^{ takeItByValue(x); });
// CHECK: call void @_Z13takeItByValueICiEvT_(i64
}
}
namespace test_complex_int_ref_mutable {
_Complex int y = 100;
void test() {
const _Complex int &x = y;
takeABlock(^{ takeItByValue(x); });
// CHECK: [[R:%.*]] = load i32, i32* getelementptr inbounds ({ i32, i32 }* @_ZN28test_complex_int_ref_mutable1yE, i32 0, i32 0)
// CHECK-NEXT: [[I:%.*]] = load i32, i32* getelementptr inbounds ({ i32, i32 }* @_ZN28test_complex_int_ref_mutable1yE, i32 0, i32 1)
// CHECK-NEXT: [[RSLOT:%.*]] = getelementptr inbounds { i32, i32 }, { i32, i32 }* [[CSLOT:%.*]], i32 0, i32 0
// CHECK-NEXT: [[ISLOT:%.*]] = getelementptr inbounds { i32, i32 }, { i32, i32 }* [[CSLOT]], i32 0, i32 1
// CHECK-NEXT: store i32 [[R]], i32* [[RSLOT]]
// CHECK-NEXT: store i32 [[I]], i32* [[ISLOT]]
// CHECK-NEXT: [[COERCE:%.*]] = bitcast { i32, i32 }* [[CSLOT]] to i64*
// CHECK-NEXT: [[CVAL:%.*]] = load i64, i64* [[COERCE]],
// CHECK-NEXT: call void @_Z13takeItByValueICiEvT_(i64 [[CVAL]])
}
}
// rdar://13295759
namespace test_block_in_lambda {
void takeBlock(void (^block)());
// The captured variable has to be non-POD so that we have a copy expression.
struct A {
void *p;
A(const A &);
~A();
void use() const;
};
void test(A a) {
auto lambda = [a]() {
takeBlock(^{ a.use(); });
};
lambda(); // make sure we emit the invocation function
}
// CHECK-LABEL: define internal void @"_ZZN20test_block_in_lambda4testENS_1AEENK3$_0clEv"(
// CHECK: [[BLOCK:%.*]] = alloca [[BLOCK_T:<{.*}>]], align 8
// CHECK: [[THIS:%.*]] = load [[LAMBDA_T:%.*]]*, [[LAMBDA_T:%.*]]**
// CHECK: [[TO_DESTROY:%.*]] = getelementptr inbounds [[BLOCK_T]], [[BLOCK_T]]* [[BLOCK]], i32 0, i32 5
// CHECK: [[T0:%.*]] = getelementptr inbounds [[BLOCK_T]], [[BLOCK_T]]* [[BLOCK]], i32 0, i32 5
// CHECK-NEXT: [[T1:%.*]] = getelementptr inbounds [[LAMBDA_T]], [[LAMBDA_T]]* [[THIS]], i32 0, i32 0
// CHECK-NEXT: call void @_ZN20test_block_in_lambda1AC1ERKS0_({{.*}}* [[T0]], {{.*}}* dereferenceable({{[0-9]+}}) [[T1]])
// CHECK-NEXT: [[T0:%.*]] = bitcast [[BLOCK_T]]* [[BLOCK]] to void ()*
// CHECK-NEXT: call void @_ZN20test_block_in_lambda9takeBlockEU13block_pointerFvvE(void ()* [[T0]])
// CHECK-NEXT: call void @_ZN20test_block_in_lambda1AD1Ev({{.*}}* [[TO_DESTROY]])
// CHECK-NEXT: ret void
}