Files
llvm-project/clang/test/CodeGenCXX/implicit-copy-constructor.cpp
Lang Hames 5824a4f1b0 When generating IR for default copy-constructors, copy-assignment operators,
move-constructors and move-assignment operators, use memcpy to copy adjacent
POD members.

Previously, classes with one or more Non-POD members would fall back on
element-wise copies for all members, including POD members. This often
generated a lot of IR. Without padding metadata, it wasn't often possible
for the LLVM optimizers to turn the element-wise copies into a memcpy.

This code hasn't yet received any serious tuning. I didn't see any serious
regressions on a self-hosted clang build, or any of the nightly tests, but
I think it's important to get this out in the wild to get more testing.
Insights, feedback and comments welcome.

Many thanks to David Blaikie, Richard Smith, and especially John McCall for
their help and feedback on this work.

llvm-svn: 174919
2013-02-11 23:44:11 +00:00

108 lines
2.1 KiB
C++

// RUN: %clang_cc1 -triple x86_64-apple-darwin10 -emit-llvm -o - %s -std=c++11 | FileCheck %s
struct A {
A();
A(const A&);
A(A&);
~A();
};
struct B {
B();
B(B&);
};
struct C {
C() {}
C(C& other, A a = A());
int i, j;
};
struct POD {
int array[3][4];
};
struct D : A, B, virtual C {
D();
int scalar;
int scalar_array[2][3];
B class_member;
C class_member_array[2][3];
POD pod_array[2][3];
union {
int x;
float f[3];
};
};
void f(D d) {
D d2(d);
}
// CHECK: define linkonce_odr void @_ZN1DC1ERS_(%struct.D* %this, %struct.D*) unnamed_addr
// CHECK: call void @_ZN1AC1Ev
// CHECK: call void @_ZN1CC2ERS_1A
// CHECK: call void @_ZN1AD1Ev
// CHECK: call void @_ZN1AC2ERS_
// CHECK: call void @_ZN1BC2ERS_
// CHECK: {{call void @llvm.memcpy.p0i8.p0i8.i64.*i64 28}}
// CHECK: call void @_ZN1BC1ERS_
// CHECK: br
// CHECK: {{icmp ult.*, 2}}
// CHECK: {{icmp ult.*, 3}}
// CHECK: call void @_ZN1AC1Ev
// CHECK: call void @_ZN1CC1ERS_1A
// CHECK: call void @_ZN1AD1Ev
// CHECK: {{call void @llvm.memcpy.p0i8.p0i8.i64.*i64 300}}
// CHECK: ret void
template<class T> struct X0 { void f0(T * ) { } };
template <class > struct X1 { X1( X1& , int = 0 ) { } };
struct X2 { X1<int> result; };
void test_X2()
{
typedef X2 impl;
typedef X0<impl> pimpl;
impl* i;
pimpl pdata;
pdata.f0( new impl(*i));
}
// rdar://problem/9598341
namespace test3 {
struct A { A(const A&); A&operator=(const A&); };
struct B { A a; unsigned : 0; };
void test(const B &x) {
B y = x;
y = x;
}
}
namespace test4 {
// When determining whether to implement an array copy as a memcpy, look at
// whether the *selected* constructor is trivial.
struct S {
int arr[5][5];
S(S &);
S(const S &) = default;
};
// CHECK: @_ZN5test42f1
void f1(S a) {
// CHECK-NOT: memcpy
// CHECK: call void @_ZN5test41SC1ERS0_
// CHECK-NOT: memcpy
S b(a);
// CHECK: }
}
// CHECK: @_ZN5test42f2
void f2(const S a) {
// CHECK-NOT: call void @_ZN5test41SC1ERS0_
// CHECK: memcpy
// CHECK-NOT: call void @_ZN5test41SC1ERS0_
S b(a);
// CHECK: }
}
}