[cfe-commits] r65259 - /cfe/trunk/lib/CodeGen/CGDecl.cpp

[Daniel Dunbar]

On Sat, Feb 21, 2009 at 7:23 PM, Eli Friedman <eli.friedman at gmail.com> wrote:
> Last part of PR3254: use the same alignment computation in Sema and
> CodeGen.  I'm not sure whether this actually makes any visible
> difference, but it's better to be consistent anyway.

It totally does:
--
ddunbar at milton:single-args-64$ echo "void f0(void) { double x; }" |
clang -emit-llvm | grep alloca
	%x = alloca double, align 8		; <double*> [#uses=0]
--
used to be align 4. It looks like we are still emitting double globals
with an alignment of 4 instead of 8 though (and treating
__alignof__(double) as 4).

Thanks!

 - Daniel