[LLVMbugs] [Bug 10949] New: [AVX] incorrect code generated for <8 x i16> vector

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=10949

           Summary: [AVX] incorrect code generated for <8 x i16> vector
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: All
            Status: NEW
          Severity: normal
          Priority: P
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: matt at pharr.org
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=7286)
 --> (http://llvm.org/bugs/attachment.cgi?id=7286)
test case

(I'm seeing some recent breakage for 8x vectors of i8, i16, and i64 sizes with
AVX, as of the last day or two.  I assume they're likely to have the same root
cause, so am reporting just this one.)

With the attached test case, the code generated for foo() computes different
results if I use the -mattr=+avx flag to llc or not. The results computed for
the SSE path [0,2,0,4,0,0,0,0] are correct, but the AVX path ends up returning
[0,2,0,0,0,0,0,0].

-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.