https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115036
Andrew Macleod <amacleod at redhat dot com> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |amacleod at redhat dot com
--- Comment #2 from Andrew Macleod <amacleod at redhat dot com> ---
Where is something suppose to happen?
After VRP2, we have:
Non-varying global ranges:
=========================:
a_3(D) : [irange] long int [0, 65535] MASK 0xffff VALUE 0x0
b_4(D) : [irange] long int [0, 65535] MASK 0xffff VALUE 0x0
_5 : [irange] long int [0, 65535] MASK 0xffff VALUE 0x0
Recording return range of test/1:[irange] long int [0, 65535] MASK 0xffff VALUE
0x0
long int test (long int a, long int b)
{
long int _5;
<bb 2> [local count: 1073741824]:
_5 = a_3(D) / b_4(D);
return _5;
}
And if I try int in GCC11, which is pre-ranger, I get:
Value ranges after VRP:
a_3(D): long int [0, 65535]
b_4(D): long int [0, 65535]
_5: long int [0, 65535]
Substituting values and folding statements
Folding statement: _5 = a_3(D) / b_4(D);
Not folded
Folding statement: return _5;
Not folded
long int test (long int a, long int b)
{
long int _5;
<bb 2> [local count: 1073741824]:
_5 = a_3(D) / b_4(D);
return _5;
}
And it still generates:
movq %rdi, %rax
cqto
idivq %rsi
ret
So I don't think we've lost any op narrowing code.
What are we suppose to do here? I presume something is suppose to recognize
that with 2 32 bit inputs, the divide is suppose to be changed to something
else?
