http://llvm.org/bugs/show_bug.cgi?id=19451
James Molloy <[email protected]> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |RESOLVED CC| |[email protected] Resolution|--- |INVALID --- Comment #7 from James Molloy <[email protected]> --- Hi, I have reproduced the failure locally, but this is not a compiler bug. The testcase creates a structure "foo" and uses it in a std::vector. "foo" has an alignment constraint due to it containing a float32x4_t member of 16 bytes. The compiler honours this in all code it generates. The underlying allocator used by std::vector just calls operator new (_Znwj), and uses the memory returned. operator new calls malloc, and malloc returns memory aligned to 8 bytes. Memory corruption follows due to NEON instructions with :128 alignment hints being used on unaligned addresses. A similar question is asked at StackOverflow here: http://stackoverflow.com/questions/12942548/making-stdvector-allocate-aligned-memory . I'll resolve this as "not a bug". Cheers, James -- 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
