This is the mail archive of the 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: [gfortran, testsuite] Fix bounds checking for substrings, enable bounds chekcing in dg testsuite, assorted fixes

On Tue, 2006-12-05 at 19:35 +0100, Paul Thomas wrote:
> Tobias,
> I'll do the commit later on tonight.
> Is the -Os test necessary for gfortran?  What does it test?

I think it is a good thing to test more than necessary.  Also I know of
one target which is going to support gfortran where size is very very
important because there is only 256k of (non DMA'able) memory.
Also -Os just changes the back-end so testing the back-end more is a
good thing because one pass which is disabled at -Os could expose in a
different pass that is enabled with -O2 or -O3.

> What is the value of "-O3 -g"?  For the testsuite, I think we should
> drop the "-g".

We should have at least one level of optimization that uses -g as it
also checks the debugging generators to make sure they don't have issues
(like ICEs).

Remember the gfortran testsuite is a testsuite for both the front-end
and the back-end and all parts inbetween. 

Here is what I would use (and no this is not too long and gfortran
testsuite is shorter to run than even the libstdc++ testsuite):
-O3 -fomit-frame-pointer
-O3 -fomit-frame-pointer -funroll-loops
-O3 -fomit-frame-pointer -funroll-all-loops -finline-functions
-O3 -g
-O2 -fbounds-check
-O2 -fno-repack-arrays

Yes there are 10 of them but remember the testsuite is for both the
front-end usage and the back-end usage.  I might remove the
unroll-all-loops one but I know gfortran usages are more likely to use
that than normal GCC users.

Andrew Pinski

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