https://bugs.llvm.org/show_bug.cgi?id=49654
Bug ID: 49654
Summary: [InstCombine] Incorrect propagation of nsz from fneg
to fdiv
Product: libraries
Version: trunk
Hardware: All
OS: All
Status: NEW
Keywords: miscompilation
Severity: normal
Priority: P
Component: Scalar Optimizations
Assignee: unassignedb...@nondot.org
Reporter: nunoplo...@sapo.pt
CC: cameron.mcina...@nyu.edu, juneyoung....@sf.snu.ac.kr,
llvm-bugs@lists.llvm.org, spatel+l...@rotateright.com
This is from Transforms/InstCombine/fneg.ll, reduced from fast to just nsz.
The issue is that the original fdiv has no nsz but the optimized one has. In
this case, nsz can be propagated, but not added (even if fneg is nsz).
define float @fdiv_op0_constant_fneg_fmf(float %x) {
%d = fdiv float 42.000000, %x
%r = fneg nsz float %d
ret float %r
}
=>
define float @fdiv_op0_constant_fneg_fmf(float %x) {
%r = fdiv nsz float -42.000000, %x
ret float %r
}
Transformation doesn't verify!
ERROR: Value mismatch
Example:
float %x = #x80000000 (-0.0)
Source:
float %d = #xff800000 (-oo)
float %r = #x7f800000 (+oo)
Target:
float %r = #xff800000 (-oo)
Source value: #x7f800000 (+oo)
Target value: #xff800000 (-oo)
--
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
