https://llvm.org/bugs/show_bug.cgi?id=24073
Bug ID: 24073
Summary: fix for #22563 discards the upper half of (non-array)
__m256 objects
Product: clang
Version: 3.6
Hardware: PC
OS: Linux
Status: NEW
Severity: normal
Priority: P
Component: LLVM Codegen
Assignee: [email protected]
Reporter: [email protected]
CC: [email protected]
Classification: Unclassified
The following testcase, compiled with -mavx at any optimization level fails:
#include <x86intrin.h>
class V {
__m256 d;
public:
[[gnu::noinline]] V(__m256 x) : d(x) {}
bool verify() const {
return 0xff == _mm256_movemask_ps(_mm256_cmp_ps(d, _mm256_set1_ps(1.f),
_CMP_EQ_OQ));
}
};
V foo() { return _mm256_set1_ps(1.f); }
int main() { return foo().verify() ? 0 : 1; }
After the V::V constructor, clang emits "vmovaps (%rsp),%xmm0" instead of
"vmovaps (%rsp),%ymm0". Reverting bce1e8184fcb311e1ec3cbf03247bdf24e91d5e2
(r229408 in SVN) solves the issue.
--
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
LLVMbugs mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/llvmbugs
