[llvm-bugs] [Bug 49132] New: double-double constant expression evaluation can produce spurious NaN

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=49132

            Bug ID: 49132
           Summary: double-double constant expression evaluation can
                    produce spurious NaN
           Product: clang
           Version: trunk
          Hardware: Other
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: -New Bugs
          Assignee: unassignedclangbugs at nondot.org
          Reporter: bbaren at google.com
                CC: htmldeveloper at gmail.com, llvm-bugs at lists.llvm.org,
                    neeilans at live.com, richard-llvm at metafoo.co.uk

Created attachment 24512
  --> https://bugs.llvm.org/attachment.cgi?id=24512&action=edit
minimal example

When compiling for POWER (-target powerpc64le-linux-gnu), Clang can incorrectly
simplify long double (double-double) expressions to NaN:

        $ uname -m
        ppc64le
        $ cat ldbl.c
        extern int printf(const char* restrict, ...);
        int main() {
                long double a = -0x1.1e282e024debdb875f47f6c85p+1021L;
                printf("%Le\n", a + 0x1.fffffffffffff7ffffffffffff8p+1023L);
        }
        $ clang -Wall -Wextra -O1 -o ldbl ldbl.c
        $ ./ldbl
        nan

Inspecting the generated bitcode shows that Clang has statically evaluated the
double-double addition to a NaN, but by IEEE 754, adding two non-NaN inputs
should never produce a NaN output.

GCC produces the correct answer for this computation, though only when
optimizations are enabled (see
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99048).

-- 
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/20210210/0888e233/attachment.html>