From: Rene Rivera (grafik.list_at_[hidden])
Date: 2005-09-26 15:20:51
Robert Ramey wrote:
> Rene Rivera wrote:
>>After having run two cycles of tests for Boost with the
>>mingw-3_4_2-stlport-5_0 configuration and having it take more than 14
>>hours on a 2.2Ghz+1GB machine, most of that in the Boost.Serialization
> BTW - on my machine here I test borland 5.51, borland 5.64, msvc 6.0, msvc
> 6.0-stlport 4.53, VC 7.1, gcc 3.3 with cygwin, and comeau. The whole set of
> tests take about 5 hours on my 2.2 GHz machine for all seven compilers.
> That's just the serialization library - I don't test other libraries. So if
> you're experiencing 11 hours CPU time for one compiler you're definately
> doing something differently than I am.
One likely immediate difference is that I run test with optimizations
> Note in a couple of cases I found test that fail in a particularly
> inconvenient way on some platforms. This is usually that either they loop
> forever or allocate so much memory that they thrash forever. Sinced you're
> running 11 hours CPU time for just one compiler, it would be helpful to
> determine the cause. Its possible that failure in just one test is causing
> the failure.
I'll try and see where the slow downs occur. But just an FYI the long
running times are not just the serialization library. There are some
tests out there that do take a long time to just compile.
> On win32 platforms there is no test cut-off for tests which exceed some
> predetermined resource levels. Obviously that would be helpful, but for now
> we just have to look at them.
Sure, and it's on our TODO list for testing. But if the large resources
needed for one particular test are indicative of using the library it
reflects badly on that library. And users might be frustrated enough to
not use the library if their compile times go from minutes to hours, or
from Megs to Gigs of RAM.
-- -- Grafik - Don't Assume Anything -- Redshift Software, Inc. - http://redshift-software.com -- rrivera/acm.org - grafik/redshift-software.com -- 102708583/icq - grafikrobot/aim - Grafik/jabber.org
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk