[LLVMbugs] [Bug 21130] gvn badly replaces a call to fptosi

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=21130

Sanjay Patel <spatel+llvm at rotateright.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
                 CC|                            |spatel+llvm at rotateright.com
         Resolution|---                         |INVALID

--- Comment #3 from Sanjay Patel <spatel+llvm at rotateright.com> ---
Both the LLVM Language Ref and the C standard (annex F.4) say that the code is
undefined, so anything goes:

fptosi:
"If the value won’t fit in the integer type, the results are undefined."

C - floating to integer conversion:
"...if the integral part of the floating value exceeds the range of the integer
type, then the "invalid" floating-point exception is raised and the resulting
value is unspecified."

Try using lrint()?

Clang does issue warnings when it's obvious that there's a conversion problem:
$ cat 21130.c
char foo() { return 156.0; }

$ ./clang  21130.c -c
21130.c:3:12: warning: implicit conversion from 'double' to 'char' changes
value from 156 to 127 [-Wliteral-conversion]
    return 156.0;
    ~~~~~~ ^~~~~
1 warning generated.

...but your C test case is probably too complicated for that check.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20141002/9ba8556a/attachment.html>