https://bugs.llvm.org/show_bug.cgi?id=45361
Bug ID: 45361
Summary: fold away float to double (to float) conversion in
copysign
Product: libraries
Version: trunk
Hardware: PC
OS: Windows NT
Status: NEW
Severity: enhancement
Priority: P
Component: Scalar Optimizations
Assignee: unassignedb...@nondot.org
Reporter: nok.ra...@gmail.com
CC: llvm-bugs@lists.llvm.org
#include <cmath>
float foo(float x, float y)
{
return std::copysign(x, y);
}
// bug 45305-like foldable float to double promotion
float bar(float x, float y)
{
return std::copysign(x, double(y));
}
// previous + fext-ftrunc pair across copysign
float qaz(float x, float y)
{
return std::copysign(double(x), y);
}
define float @foo(float %0, float %1) {
%3 = tail call float @llvm.copysign.f32(float %0, float %1) #2
ret float %3
}
define float @bar(float %0, float %1) {
%3 = fpext float %1 to double
%4 = fpext float %0 to double
%5 = tail call double @llvm.copysign.f64(double %4, double %3) #2
%6 = fptrunc double %5 to float
ret float %6
}
define float @qaz(float %0, float %1) {
%3 = fpext float %0 to double
%4 = fpext float %1 to double
%5 = tail call double @llvm.copysign.f64(double %3, double %4) #2
%6 = fptrunc double %5 to float
ret float %6
}
https://godbolt.org/z/sKIjDg
--
You are receiving this mail because:
You are on the CC list for the bug._______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
