https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90424
Bug ID: 90424
Summary: memcpy into vector builtin not optimized
Product: gcc
Version: 9.1.0
Status: UNCONFIRMED
Keywords: missed-optimization
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: kretz at kde dot org
Target Milestone: ---
Target: x86_64-*-*, i?86-*-*
Testcase (cf. https://godbolt.org/z/LsKcii):
template <class T>
using V [[gnu::vector_size(16)]] = T;
template <class T, unsigned M = sizeof(V<T>)>
V<T> load(const void *p) {
using W = V<T>;
W r;
__builtin_memcpy(&r, p, M);
return r;
}
// movq or movsd
template V<char> load<char, 8>(const void *); // bad
template V<short> load<short, 8>(const void *); // bad
template V<int> load<int, 8>(const void *); // bad
template V<long> load<long, 8>(const void *); // good
template V<float> load<float, 8>(const void *); // bad
template V<double> load<double, 8>(const void *); // good (movsd?)
// movd or movss
template V<char> load<char, 4>(const void *); // bad
template V<short> load<short, 4>(const void *); // bad
template V<int> load<int, 4>(const void *); // good
template V<float> load<float, 4>(const void *); // good
All of these partial loads should be translated to a single mov[qd] or movs[sd]
instruction. But most of them are not.
