This is the mail archive of the libstdc++@gcc.gnu.org mailing list for the libstdc++ project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [RFA:] Make 23_containers/set/modifiers/16728.cc not time out nonnatively


> Date: Tue, 6 Feb 2007 00:46:22 +0000 (UTC)
> From: "Joseph S. Myers" <joseph@codesourcery.com>

> On Tue, 6 Feb 2007, Hans-Peter Nilsson wrote:
> > libstdc++-v3:
> >         * testsuite/23_containers/set/modifiers/16728.cc: Use only 10
> >         iterations for non-native targets.
> 
> It's wrong for what's tested for a given target to depend on the host.  A 
> target that can be tested both native and cross (say *-linux*) should have 
> the same tests run in both cases.

I disagree: we have keep the tests reasonable for all
environments.  Do you really generally want to test less in the
native environment than can be reasonable in the cross
environment?

> You need some other property specified in some way (in the board file?) to 
> indicate that you have a slow simulator.

It's _not slow_ it's one of the fastest.  It's just not as fast
as native.  (The existence of a slow_simulator board variable is
a red herring.)

Perhaps the number of iterations could be changed to 10 all
over, unless manually compiled with -DITERATIONS=N, something
like this:

+// To test performance compile with -DITERATIONS=10000 and use "time".
+#ifndef ITERATIONS
+#define ITERATIONS 10
+#endif
+
 // The number of iterations to be performed.
-int iterations = 10000;
+int iterations = ITERATIONS;

Does that suit libstdc++ers?

brgds, H-P


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]