[LLVMbugs] [Bug 21140] New: New vector shuffle lowering misses to use movhps

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=21140

            Bug ID: 21140
           Summary: New vector shuffle lowering misses to use movhps
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: All
            Status: NEW
          Severity: normal
          Priority: P
         Component: Backend: X86
          Assignee: unassignedbugs at nondot.org
          Reporter: qcolombet at apple.com
                CC: llvmbugs at cs.uiuc.edu
    Classification: Unclassified

Created attachment 13124
  --> http://llvm.org/bugs/attachment.cgi?id=13124&action=edit
IR to reproduce the problem.

With the attach IR, the new vector shuffle lowering generates a movq followed
by a shufps instead of the direct movhps.

To reproduce:
llc -x86-experimental-vector-shuffle-lowering=true none_shufps_movhps.ll -o
new.s
llc -x86-experimental-vector-shuffle-lowering=false none_shufps_movhps.ll -o
old.s

diff -U 10 old.s new.s
 _foo:                                   ## @foo
     .cfi_startproc
 ## BB#0:
     movq    (%rsi), %xmm0
-    movhps    (%rdi), %xmm0
+    movq    (%rdi), %xmm1
+    shufps    $68, %xmm1, %xmm0       ## xmm0 = xmm0[0,1],xmm1[0,1]
     retq

-- 
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/20141002/5cf6362c/attachment.html>