This is the mail archive of the gcc-patches@gcc.gnu.org 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]

[reload-v2] Fix testsuite failures after double-word IRA work



While the branch was bootstrapping it was showing some testsuite regressions relative to the trunk due to two issues.


The testsuite failures show up because we were allocating the last hard register in GENERAL_REGS to an XFmode allocno/pseudo. Needless to say this caused problems later.

The first problem was we were creating a single underlying object for the new allocnos because the new allocnos didn't have a cover class assigned yet. This resulted in always assuming the allocno had one underlying object.

Second, we failed to set conflicts for hard regs not in the new allocno's cover class. I'm not sure how we go away with this for so long, but it's clearly wrong.


Attachment: PP
Description: Text document


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