This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: Results for g++ 3.1 application testing on i686-pc-linux-gnu
Nathan Myers wrote:
> On Thu, Mar 07, 2002 at 08:47:42PM +0100, Peter Schmid wrote:
> > current boost cvs:
> > All tests pass. There are seven "failures" for empty_* types in
> > object_type_traits_test.cpp. I do not know why.|
>
> Can you please post the relevant code from one of the tests?
> Maybe the test is broken.
The problem is not so critical, anyway: there are indeed 7 failures but 7 are
also the *expected* failures, therefore no regression wrt 3.0.x neither vs
2.95.x.
checking value of boost::has_trivial_constructor<empty_UDT>::value...failed
found: 0 expected 1
checking value of boost::has_trivial_copy<empty_UDT>::value...failed
found: 0 expected 1
checking value of boost::has_trivial_assign<empty_UDT>::value...failed
found: 0 expected 1
checking value of boost::has_nothrow_constructor<empty_UDT>::value...failed
found: 0 expected 1
checking value of boost::has_nothrow_copy<empty_UDT>::value...failed
found: 0 expected 1
checking value of boost::has_nothrow_assign<empty_UDT>::value...failed
found: 0 expected 1
checking value of boost::is_empty<empty_union_UDT>::value...failed
boost::is_empty<empty_union_UDT>::value does not compile on this compiler
201 tests completed, 7 failures found, 7 failures expected from this compiler.
no errors detected
> > stepanov_v1p2.C:
> > -O
> > Abstraction Penalty: 1.13
> > -O2
> > Abstraction Penalty: 1.32
>
> This is very odd, and suggests an optimization bug.
> Also, it would be interesting to see what happens with -O0.
on i686 at least stepanov started behaving badly at -O2(3) /many/ months ago.
Currently the best results are always obtained at -O1. On my PII:
-O0 "uncomputable", that is >> 100 secs running time
-O1 1.04
-O2 1.28
-O3 1.26
Ciao, Paolo.