From: Craig Topper <craig.topper@gmail.com>
Date: Mon, 14 Oct 2013 00:24:33 +0000 (+0000)
Subject: Don't use 64-bit versions of MOVMSKPD in CodeGen. The instructions only produce a... 
X-Git-Url: http://plrg.eecs.uci.edu/git/?a=commitdiff_plain;h=eefca967925d2b8cbb3ed3a24930717ea315b56e;p=oota-llvm.git

Don't use 64-bit versions of MOVMSKPD in CodeGen. The instructions only produce a 1-bit result so we can just use SUBREG_TO_REG to extend the 32-bit versions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@192562 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index 6d00c543388..6f7e9131580 100644
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -2731,11 +2731,13 @@ let Predicates = [HasAVX] in {
   def : Pat<(i32 (X86fgetsign FR32:$src)),
             (VMOVMSKPSrr32 (COPY_TO_REGCLASS FR32:$src, VR128))>;
   def : Pat<(i64 (X86fgetsign FR32:$src)),
-            (VMOVMSKPSrr64 (COPY_TO_REGCLASS FR32:$src, VR128))>;
+            (SUBREG_TO_REG (i64 0),
+             (VMOVMSKPSrr32 (COPY_TO_REGCLASS FR32:$src, VR128)), sub_32bit)>;
   def : Pat<(i32 (X86fgetsign FR64:$src)),
             (VMOVMSKPDrr32 (COPY_TO_REGCLASS FR64:$src, VR128))>;
   def : Pat<(i64 (X86fgetsign FR64:$src)),
-            (VMOVMSKPDrr64 (COPY_TO_REGCLASS FR64:$src, VR128))>;
+            (SUBREG_TO_REG (i64 0),
+             (VMOVMSKPDrr32 (COPY_TO_REGCLASS FR64:$src, VR128)), sub_32bit)>;
 }
 
 defm MOVMSKPS : sse12_extr_sign_mask<VR128, int_x86_sse_movmsk_ps, "movmskps",
@@ -2747,13 +2749,15 @@ def : Pat<(i32 (X86fgetsign FR32:$src)),
           (MOVMSKPSrr32 (COPY_TO_REGCLASS FR32:$src, VR128))>,
       Requires<[UseSSE1]>;
 def : Pat<(i64 (X86fgetsign FR32:$src)),
-          (MOVMSKPSrr64 (COPY_TO_REGCLASS FR32:$src, VR128))>,
+          (SUBREG_TO_REG (i64 0),
+           (MOVMSKPSrr32 (COPY_TO_REGCLASS FR32:$src, VR128)), sub_32bit)>,
       Requires<[UseSSE1]>;
 def : Pat<(i32 (X86fgetsign FR64:$src)),
           (MOVMSKPDrr32 (COPY_TO_REGCLASS FR64:$src, VR128))>,
       Requires<[UseSSE2]>;
 def : Pat<(i64 (X86fgetsign FR64:$src)),
-          (MOVMSKPDrr64 (COPY_TO_REGCLASS FR64:$src, VR128))>,
+          (SUBREG_TO_REG (i64 0),
+           (MOVMSKPDrr32 (COPY_TO_REGCLASS FR64:$src, VR128)), sub_32bit)>,
       Requires<[UseSSE2]>;
 
 //===---------------------------------------------------------------------===//