[llvm-commits] [compiler-rt] r159419 - /compiler-rt/trunk/lib/asan/tests/CMakeLists.txt

[Kostya Serebryany]

On Fri, Jun 29, 2012 at 2:52 PM, Chandler Carruth <chandlerc at google.com>wrote:

> Well, at the most generic level, I'd prefer to have more binaries rather
> than fewer as we can run them in parallel.


That's a good point. But my point is that if we run this as one process we
are more likely to hit scalability issues.
That might be a bad use of a *unit* test, but that helped me find a
scalability bug at least once (one test allocated too much memory; as the
result another test became too slow).

If we want to run a gtest binary in parallel, we can always use shards.


>
> Of course, that's within reason -- when they execute in a few us, that's
> not useful, but we're not there.
>
> That said, the big difference is that it feels like the link step of the
> instrumented side of this is a bit higher risk than the other? Maybe I'm
> just paranoid.
>

Risk? Hm. Dunno. Hardly so.

--kcc
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20120629/3bbe4f3f/attachment.html>