Hi,
On 2018-09-05 10:05:26 -0400, Tom Lane wrote:
> Peter Eisentraut <peter.eisentraut@2ndquadrant.com> writes:
> > On 05/09/2018 02:51, Andres Freund wrote:
> >> My current proposal is thus to do add a check that does
> >> #if defined(__clang__) && defined(__i386__) && !defined(__SSE2_MATH__)
> >> something-that-fails
> >> #endif
> >> in an autoconf test, and have configure complain if that fails.
>
> > Would this not be invalidated if the bug you have filed gets fixed?
Realistically we're going to be running into old versions of clang for a
long time. And the importance of running i386 without SSE2 surely isn't
increasing. So I don't really see an urgent need to do anything about
it. And if it gets fixed, and we care, we can just add a clang version
check to the test.
> Perhaps, but how would autoconf tell that? I wouldn't have any confidence
> in a runtime test, as it might or might not trigger the bug depending on
> all sorts of factors like -O level.
>
> In any case, it seems like "use -msse2 on x86" is good advice from a
> performance standpoint even if it doesn't prevent a compiler bug.
Indeed.
> One thought is that maybe we should provide a way to override this,
> in case somebody really can't or doesn't want to use -msse2, and
> is willing to put up with platform-dependent float behavior.
IDK, people that are capable of making that decision can just hack
configure.
Greetings,
Andres Freund