This is a revision of the patches I sent on May 18th.
I tested it on 3 platforms:
* Power9 little endian, --with-code=power9;
* Power8 big endian, --with-code=power8, both 32/64-bit tests done;
* Power10 little endian, --with-code=power10.
All systems bootstrapped and there were no new regressions. I believe I have
addressed the issues with the last patch.
The first patch in this set contains the same GCC code and new test as in the
previous patch, since I don't believe there was a problem with those bits.
I moved the changes for the existing test 'float128-minmax.c' to patch number
two. Rather than using '#pragma GCC target' to force power9 code generation on
power10, instead I used conditional scan-assembler statements to deliniate the
power9 and power10 code generation.
The third patch of this set fixes the complicated test that was complained
about in the previous second patch.
Can I check these patches into the master branch. Ideally, I think these
should go into GCC 11.2 after a soak-in period.
--
Michael Meissner, IBM
IBM, M/S 2506R, 550 King Street, Littleton, MA 01460-6245, USA
email: [email protected], phone: +1 (978) 899-4797
