[LLVMbugs] [Bug 18591] New: Incorrect optimization of is_simm32 C++ function from HotSpot JVM sources

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=18591

            Bug ID: 18591
           Summary: Incorrect optimization of is_simm32 C++ function from
                    HotSpot JVM sources
           Product: new-bugs
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P
         Component: new bugs
          Assignee: unassignedbugs at nondot.org
          Reporter: mikael.gerdin at oracle.com
                CC: llvmbugs at cs.uiuc.edu
    Classification: Unclassified

Created attachment 11924
  --> http://llvm.org/bugs/attachment.cgi?id=11924&action=edit
test case

I tried to build the OpenJDK JVM with clang++ 3.4 and discovered that the 32
bit version seems to hit an impossible assert.

I did some digging and found out that the function Assembler::is_simm32 is
incorrectly optimized, causing us to always hit the assert.

The JVM sources in question are here:
http://hg.openjdk.java.net/jdk9/hs/hotspot/file/3e2b76368121/src/share/vm/asm/assembler.hpp

The bug appears on -O2 and above, both with clang-3.4 and with llvm+clang trunk
as of 2014-01-23.

I've created a reduced test case (attached)

llvm-build$ Release+Asserts/bin/clang++ -O1 -m32 ~/scratch/is_simm.cc; ./a.out;
echo "Exit code: $?"
is_simm32(1) = 1
min_simm(32) <= 1 && 1 <= max_simm(32) = 1
Exit code: 0
llvm-build$ Release+Asserts/bin/clang++ -O2 -m32 ~/scratch/is_simm.cc; ./a.out;
echo "Exit code: $?"
is_simm32(1) = 0
min_simm(32) <= 1 && 1 <= max_simm(32) = 1
Exit code: 1

-- 
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/20140123/b26aa5ba/attachment.html>