https://bugs.llvm.org/show_bug.cgi?id=50094
Bug ID: 50094
Summary: (popcount(x) ^ popcount(y)) & 1 - > (popcount(x ^ y))
& 1
Product: libraries
Version: trunk
Hardware: PC
OS: Linux
Status: NEW
Severity: enhancement
Priority: P
Component: Scalar Optimizations
Assignee: [email protected]
Reporter: [email protected]
CC: [email protected]
Original pattern: (__builtin_parity(x) ^ __builtin_parity(y))
LLVM rewrites it as: (__builtin_popcount(x) ^ __builtin_popcount(y)) & 1
----------------------------------------
define i32 @src(i32 %0, i32 %1) noread nowrite nofree {
%2:
%3 = ctpop i32 %0
%4 = ctpop i32 %1
%5 = xor i32 %4, %3
%6 = and i32 %5, 1
ret i32 %6
}
=>
define i32 @tgt(i32 %0, i32 %1) noread nowrite nofree {
%2:
%3 = xor i32 %1, %0
%4 = ctpop i32 %3
%5 = and i32 %4, 1
ret i32 %5
}
Transformation seems to be correct!
parity_xor(unsigned int, unsigned int): #
@parity_xor(unsigned int, unsigned int)
popcnt ecx, edi
popcnt eax, esi
xor eax, ecx
and eax, 1
ret
tgt(unsigned int, unsigned int): # @tgt(unsigned
int, unsigned int)
xor edi, esi
popcnt eax, edi
and eax, 1
ret
https://godbolt.org/z/v3h361fz7
--
You are receiving this mail because:
You are on the CC list for the bug._______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
