This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC 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: Effect of -fno-reorder-blocks?

> On Mon, 18 Nov 2002, Jan Hubicka wrote:
> 
> > On the other hand, the spec2000 tester is P4 and knowing results there
> > might be interesting because of completely different machine.
> > Also in case the prboelm is really instruction cache trashing (I will
> > try to oprofile it), we probably won't see it on spec95 because it is
> > small.
> > 
> OK.  Lets try it.  It will be enabled for tomorrow's run.  b-i-b
> does not run tonight.
Thanks!  The results are tehre, so you can return it back to original
state.
It looks like -fno-reorder-blocks is win, but majority comes from equake
benchmark.  There is regression on Athlon too, so it looks like the
algorithm does particulary bad there.  I will check.

Honza
> 
> 
> Diego.
Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]