[prev in list] [next in list] [prev in thread] [next in thread] 

List:       gcc-fortran
Subject:    Re: [gfortran, testsuite, ping] Run tests with -fbounds-check, was:  Re:  RFA: matmul/transpose opti
From:       Tobias_Schlüter <tobias.schlueter () physik ! uni-muenchen ! de>
Date:       2005-12-28 12:22:40
Message-ID: 43B28390.1080405 () physik ! uni-muenchen ! de
[Download RAW message or body]

Janne Blomqvist wrote:
> On Tue, Dec 27, 2005 at 01:11:26PM +0100, Tobias Schlüter wrote:
>>>Patch to the testsuite below.  Tested with a 'make check-fortran'
>>>and visual inspection of the testsuite log.  This removes the run
>>>with -funroll-all-loops, as this is a synonym for -funroll-loops
>>>according to the documentation.  I also
> 
> 
> I'm not so sure of that, the docs say:
> 
> -funroll-loops
>     Unroll loops whose number of iterations can be determined at
> compile time or upon entry to the loop. -funroll-loops implies both
> -fstrength-reduce and -frerun-cse-after-loop. This option makes code
> larger, and may or may not make it run faster.
> 
> -funroll-all-loops
>     Unroll all loops, even if their number of iterations is uncertain
> when the loop is entered. This usually makes programs run more
> slowly. -funroll-all-loops implies the same options as -funroll-loops,
> 
> It clearly seems that -funroll-all-loops is different. I think the
> statement that it implies the same options as -funroll-loops refers to
> the -fstrength-reduce and -frerun-cse-after-loop, not that the entire
> option is equivalent.

Yes, that's true.  I had a look at the code, in order to figure out what
exactly the differences are, but I couldn't find any handling of
-funroll-all-loops at all.  Where is the code?

grep -n -e unroll.all *[hc] *opt /dev/null
loop-init.c:302:  return (flag_peel_loops || flag_unroll_loops ||
flag_unroll_all_loops);
loop-init.c:316:      if (flag_unroll_all_loops)
toplev.c:1549:  if (flag_unroll_all_loops)
toplev.h:131:extern int flag_unroll_all_loops;
common.opt:978:funroll-all-loops
common.opt:979:Common Report Var(flag_unroll_all_loops)

grep finished (matches found) at Wed Dec 28 13:16:54

> That being said, perhaps we should test with -funroll-all-loops
> instead of -funroll-loops, since it seems that it applies to all
> situations where the normal unroller is used and then some.

Perhaps.  I don't think it's worth changing again, but if the consensus is
different, I will prepare a patch.

- Tobi

[prev in list] [next in list] [prev in thread] [next in thread]