19 Dec
2012
19 Dec
'12
4:51 p.m.
On Wed, Dec 19, 2012 at 6:34 AM, Tim Blechmann <tim@klingt.org> wrote:
hi andrey and peter,
to be more specific: of the test machines at sandia sometimes has test failures for boost.lockfree, which i cannot reproduce on my machine. the algorithm in question (the spsc_queue) only makes use of atomic reads and writes and only requires two memory barriers. the sandia machine is a 4 cpu, 8 core/cpu machine, while my machines are all single-cpu multi-core.
...
imo, the platform_fence_before/after is not necessary ...
tim
If there is a bug, _removing_ barriers isn't going to solve it. Or do I misunderstand the situation? (I agree that the barriers may be unnecessary, but I don't see how it helps with the test failures.) Tony