https://gcc.gnu.org/bugzilla/show_bug.cgi?id=30314
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Ever confirmed|0 |1
Keywords| |missed-optimization
Status|UNCONFIRMED |ASSIGNED
Last reconfirmed| |2021-08-11
Assignee|unassigned at gcc dot gnu.org |pinskia at gcc dot
gnu.org
--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
starting in GCC 11, we produce:
movl $140, %edx
movq %rdi, %rax
mulq %rdx
jo .L11
Which corresponds to this gimple:
_7 = .MUL_OVERFLOW (num_3(D), 140);
_8 = IMAGPART_EXPR <_7>;
if (_8 != 0)
So it is definitely better than before.
BUT this is not the best code, clang produces (for 64bit):
movabsq $131762457669353941, %rax # imm = 0x1D41D41D41D41D5
cmpq %rax, %rdi
jb .LBB0_2
So basically ULONGMAX/140 as described.
So something like:
(simplify
(imagpart (IFN_MUL_OVERFLOW:s @0 INTEGER_CST@1))
(if (TREE_CODE (TYPE_MAX_VALUE (type)) == INTEGER_CST)
(convert (le @0 (trunc_div { TYPE_MAX_VALUE (type); } @1))))