[llvm-bugs] [Bug 39353] New: Exorbitant compile times
via llvm-bugs
llvm-bugs at lists.llvm.org
Fri Oct 19 04:14:41 PDT 2018
https://bugs.llvm.org/show_bug.cgi?id=39353
Bug ID: 39353
Summary: Exorbitant compile times
Product: clang
Version: trunk
Hardware: PC
OS: Linux
Status: NEW
Severity: enhancement
Priority: P
Component: C++
Assignee: unassignedclangbugs at nondot.org
Reporter: lumosimann at gmail.com
CC: dgregor at apple.com, llvm-bugs at lists.llvm.org
I was just experimenting a bit and realized that cache might miss some cache
opportunities:
```
#include <iostream>
#include <tuple>
#include <type_traits>
template <size_t I0, size_t... I>
struct Recursion {
using type = std::tuple<
typename Recursion<I0 - 1, I - 1, I - 1, I - 1, I - 1>::type...>;
};
template <size_t... I>
struct Recursion<0, I...> {
using type = std::integral_constant<int, 0>;
};
template <std::size_t I, typename T>
struct SomeWrapper {
using type = std::tuple_element_t<0, T>;
};
template <size_t I>
using ComplexType = typename Recursion<I, I, I, I, I>::type;
struct SimpleType {
using type = int;
};
template <typename T>
struct Concept : std::bool_constant<std::tuple_size_v<typename T::type> != 4> {
};
template <typename T>
std::enable_if_t<Concept<T>::value, int> g(T) {
return 3;
}
template <typename T>
std::enable_if_t<!Concept<T>::value, int> g(T) {
return 3;
}
template <std::size_t... Is>
int f(std::index_sequence<Is...>) {
return (... + g(SomeWrapper<Is, ComplexType<9>>{}));
}
int main() {
std::cout << f(std::make_index_sequence<200>()) << std::endl;
return 0;
}
```
Compilation time for this code scales linearly with Clang (linear in terms
ofthe number in `make_index_sequence`), but remains constant for gcc.
The whole thing is just setting up a huge type that is used in SFiNAE for
evaluating g. It looks like Concept<T> is not cached during consecutive calls
to g.
--
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20181019/74807aa5/attachment-0001.html>
More information about the llvm-bugs
mailing list