https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97459
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Attachment #49633|0 |1
is obsolete| |
--- Comment #19 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 49634
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=49634&action=edit
gcc11-pr97459-wip.patch
The earlier patch was buggy, we can't really consider the bit 63 for 128-bit
umod or 31 for 64-bit umod, because the sum of the 3 numbers like:
0x7fffffffffffffffULL + 0x7fffffffffffffffULL + 0x3ULL can overflow and unlike
the bit 64 (or 32) case it would need more code to handle those 2 values where
it overflowed.
I'll see if I figure out how to do signed modulo too.