Allow generation of vmla.f32 instructions when targeting Cortex-A15. The patch also adds the VFP4 feature to Cortex-A15 and fixes the DontUseFusedMAC predicate so that we can still generate vmla.f32 instructions on non-darwin targets with VFP4.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187349 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/Target/ARM/ARM.td b/lib/Target/ARM/ARM.td
index 46928dc..e5da3a5 100644
--- a/lib/Target/ARM/ARM.td
+++ b/lib/Target/ARM/ARM.td
@@ -179,7 +179,7 @@
// FIXME: It has not been determined if A15 has these features.
def ProcA15 : SubtargetFeature<"a15", "ARMProcFamily", "CortexA15",
"Cortex-A15 ARM processors",
- [FeatureT2XtPk, FeatureFP16,
+ [FeatureT2XtPk, FeatureFP16, FeatureVFP4,
FeatureAvoidPartialCPSR,
FeatureTrustZone]>;
def ProcR5 : SubtargetFeature<"r5", "ARMProcFamily", "CortexR5",
diff --git a/lib/Target/ARM/ARMISelDAGToDAG.cpp b/lib/Target/ARM/ARMISelDAGToDAG.cpp
index 31ce38e..4ca3af6 100644
--- a/lib/Target/ARM/ARMISelDAGToDAG.cpp
+++ b/lib/Target/ARM/ARMISelDAGToDAG.cpp
@@ -422,7 +422,7 @@
if (!CheckVMLxHazard)
return true;
- if (!Subtarget->isCortexA8() && !Subtarget->isLikeA9() &&
+ if (!Subtarget->isCortexA8() && !Subtarget->isCortexA9() &&
!Subtarget->isSwift())
return true;
diff --git a/lib/Target/ARM/ARMInstrInfo.td b/lib/Target/ARM/ARMInstrInfo.td
index f543e5d..c243402 100644
--- a/lib/Target/ARM/ARMInstrInfo.td
+++ b/lib/Target/ARM/ARMInstrInfo.td
@@ -262,7 +262,9 @@
def UseFusedMAC : Predicate<"(TM.Options.AllowFPOpFusion =="
" FPOpFusion::Fast) && "
"!Subtarget->isTargetDarwin()">;
-def DontUseFusedMAC : Predicate<"!Subtarget->hasVFP4() || "
+def DontUseFusedMAC : Predicate<"!(TM.Options.AllowFPOpFusion =="
+ " FPOpFusion::Fast &&"
+ " Subtarget->hasVFP4()) || "
"Subtarget->isTargetDarwin()">;
// VGETLNi32 is microcoded on Swift - prefer VMOV.
diff --git a/lib/Target/ARM/ARMTargetMachine.cpp b/lib/Target/ARM/ARMTargetMachine.cpp
index 354a779..1ba78e4 100644
--- a/lib/Target/ARM/ARMTargetMachine.cpp
+++ b/lib/Target/ARM/ARMTargetMachine.cpp
@@ -169,7 +169,7 @@
// FIXME: temporarily disabling load / store optimization pass for Thumb1.
if (getOptLevel() != CodeGenOpt::None && !getARMSubtarget().isThumb1Only())
addPass(createARMLoadStoreOptimizationPass(true));
- if (getOptLevel() != CodeGenOpt::None && getARMSubtarget().isLikeA9())
+ if (getOptLevel() != CodeGenOpt::None && getARMSubtarget().isCortexA9())
addPass(createMLxExpansionPass());
// Since the A15SDOptimizer pass can insert VDUP instructions, it can only be
// enabled when NEON is available.