Re: [PING] Splay tree speedup

[Daniel Berlin <dberlin at dberlin dot org>]

On Sun, 2005-11-06 at 21:11 -0500, Kazu Hirata wrote:
> Hi Brian,
> 
> > Please pardon my ignorance, but how can I run the
> > compile-time regressions?  Are they run with make
> > check?
> 
> No, our regression testsuite does not check compile-time regressions.  (There is 
> a time-out mechanism in the testsuite, but that's to prevent gcc from running 
> forever, not to detect compile-time regressions.)
> 
> > Also, how are the times normalized for different hosts?
> 
> I am not aware of any normalization.  People pick their favorite programs and 
> compile them on their own machines.  There is a nice graph that CSiBE people 
> maintains though:
> 
>    http://www.inf.u-szeged.hu/csibe/s-i686-linux.php
> 


For trivial patches that are supposed to be compile time speedup, people
generally run the following:

time make bootstrap
time Compile the testcase from PR 8361

For both an unpatched compiler, and a patched compiler.