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

Thomas Koenig <> wrote on Fri, 08 Dec 2006:

There are quite a few tests that only test the library (like the
I/O tests).  Cycling through all the options is fairly unlikely
to uncover any bugs.

Could we put these tests into a separate directory, so they are
run only once?  That could save a lot of time on testing.

Yes, that's a great argument for finally separating the library tests from the rest. I'll look into it, as I currently know my way around the testsuite.

WRT the rest of the discussion, I'd like to point out that -O3 -ftree-vectorize currently ICEs on a few testcases (PR30087), which is why I haven't yet applied a patch. Nevertheless, this makes it look like an important test.

I don't think we should grow the list of options we cycle through, testsuite runtime is an issue, even if the gfortran testsuite is the fastest testsuite in gcc. Anyway, things may look quite different with a separate testsuite for the library.

The option list I have in my tree looks like this:
{ -O0 } { -O1 } { -O2 } \
{ -O2 -ftree-vectorize } \
{ -O2 -fomit-frame-pointer -finline-functions - fbounds-check } \
{ -O2 -fomit-frame-pointer -finline-functions -funroll-loops } \
{ -O3 -g } \
{ -Os }

- Tobi

---------------------------------------------------------------- This message was sent using IMP, the Internet Messaging Program.

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