https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77993
--- Comment #7 from Michael Meissner <meissner at gcc dot gnu.org> --- Created attachment 39951 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=39951&action=edit Patch to disable IFmode if -msoft-float I believe this patch fixes the problem. It disables IFmode if -msoft-float was used. The problem was some of the insns that support IFmode were enabled if -msoft-float was used. Since the 64-bit Linux distributions stopped supporting -msoft-float some time ago, -msoft-float was no longer built on those systems, and the bug was not noticed. However, 32-bit Linux systems and other PowerPC systems do build software floating point. IFmode is the __ibm128 type added as part of the __float128/_Float128 support to allow access to the current IBM extended double support if/when long double becomes _Float128. I'm attempting to do a normal bootstrap on 64-bit Linux right now, and it looks like the build is broken due to other reasons. So, I'm going back to a previous version that did build to make sure the patch is ok. Eric, if you could try it in your environment to see if it solves the problem, it would be helpful.
