Enable -widen-vmovs by default.

This will widen 32-bit register vmov instructions to 64-bit when
possible.  The 64-bit vmovd instructions can then be translated to NEON
vorr instructions by the execution dependency fix pass.

The copies are only widened if they are marked as clobbering the whole
D-register.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@144734 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/ARM/ARMBaseInstrInfo.cpp b/lib/Target/ARM/ARMBaseInstrInfo.cpp
index 30f38b8686279d5693bc091d617f9a5a371c5274..93153484dec1a0b8985c13ed225db09ffe46fd83 100644 (file)
--- a/lib/Target/ARM/ARMBaseInstrInfo.cpp
+++ b/lib/Target/ARM/ARMBaseInstrInfo.cpp
@@ -46,7 +46,7 @@ EnableARM3Addr("enable-arm-3-addr-conv", cl::Hidden,
                cl::desc("Enable ARM 2-addr to 3-addr conv"));
 
 static cl::opt<bool>
-WidenVMOVS("widen-vmovs", cl::Hidden,
+WidenVMOVS("widen-vmovs", cl::Hidden, cl::init(true),
            cl::desc("Widen ARM vmovs to vmovd when possible"));
 
 /// ARM_MLxEntry - Record information about MLA / MLS instructions.