This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
More on partial overlaps...
- To: egcs at egcs dot cygnus dot com
- Subject: More on partial overlaps...
- From: craig at jcb-sc dot com
- Date: 27 Mar 1999 18:41:50 -0000
- Cc: craig at jcb-sc dot com
- References: <199903271826.KAA28024@kankakee.wrs.com>
[Was Re: Results for egcs-2.93.13 19990327 (gcc2 ss-980929 experimental) testsuite on sparc-sun-solaris2.5.1]
>FAIL: g77.f-torture/execute/19990325-0.f execution, -O0
>FAIL: g77.f-torture/execute/19990325-0.f execution, -O1
>FAIL: g77.f-torture/execute/19990325-0.f execution, -O2
>FAIL: g77.f-torture/execute/19990325-0.f execution, -O2 -fomit-frame-pointer -finline-functions
>FAIL: g77.f-torture/execute/19990325-0.f execution, -Os
>FAIL: g77.f-torture/execute/19990325-1.f execution, -O0
>FAIL: g77.f-torture/execute/19990325-1.f execution, -O1
>FAIL: g77.f-torture/execute/19990325-1.f execution, -O2
>FAIL: g77.f-torture/execute/19990325-1.f execution, -O2 -fomit-frame-pointer -finline-functions
>FAIL: g77.f-torture/execute/19990325-1.f execution, -Os
The fact that 19990325-0.f fails as well as -1 suggests (though I haven't
debugged it myself) that this is a target where even two *32-bit* floats
are copied directly in an assignment. It illustrates my concern that
the g77 front end won't know, in any clean way, whether temporaries
are really needed to avoid the problem, but the back end might, so it
should (ideally) be the back end that is told whether partial overlap
must be accommodated, so it can take the appropriate action.
tq vm, (burley)