[PowerPC] VSX loads and stores support unaligned access

I've not yet updated PPCTTI because I'm not sure what the actual relative cost
is compared to the aligned uses.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@204848 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/PowerPC/PPCISelLowering.cpp b/lib/Target/PowerPC/PPCISelLowering.cpp
index e09725d59d2a00ace4d08083cbe866369cd157f9..87639247e21a5984a860c695168d9877211425d0 100644 (file)
--- a/lib/Target/PowerPC/PPCISelLowering.cpp
+++ b/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -7960,7 +7960,6 @@ SDValue PPCTargetLowering::PerformDAGCombine(SDNode *N,
     unsigned ABIAlignment = getDataLayout()->getABITypeAlignment(Ty);
     if (ISD::isNON_EXTLoad(N) && VT.isVector() &&
         TM.getSubtarget<PPCSubtarget>().hasAltivec() &&
-        // FIXME: Update this for VSX!
         (VT == MVT::v16i8 || VT == MVT::v8i16 ||
          VT == MVT::v4i32 || VT == MVT::v4f32) &&
         LD->getAlignment() < ABIAlignment) {
@@ -8716,8 +8715,14 @@ bool PPCTargetLowering::allowsUnalignedMemoryAccesses(EVT VT,
   if (!VT.isSimple())
     return false;
 
-  if (VT.getSimpleVT().isVector())
-    return false;
+  if (VT.getSimpleVT().isVector()) {
+    if (PPCSubTarget.hasVSX()) {
+      if (VT != MVT::v2f64 && VT != MVT::v2i64)
+        return false;
+    } else {
+      return false;
+    }
+  }
 
   if (VT == MVT::ppcf128)
     return false;

diff --git a/lib/Target/PowerPC/PPCTargetTransformInfo.cpp b/lib/Target/PowerPC/PPCTargetTransformInfo.cpp
index 2491a9e53920a2ac29a82a57206986c65a2a05b6..a915b04fa500838542d7bd0c730cdc2faac2fd49 100644 (file)
--- a/lib/Target/PowerPC/PPCTargetTransformInfo.cpp
+++ b/lib/Target/PowerPC/PPCTargetTransformInfo.cpp
@@ -244,6 +244,8 @@ unsigned PPCTTI::getMemoryOpCost(unsigned Opcode, Type *Src, unsigned Alignment,
   // Each load/store unit costs 1.
   unsigned Cost = LT.first * 1;
 
+  // FIXME: Update this for VSX loads/stores that support unaligned access.
+
   // PPC in general does not support unaligned loads and stores. They'll need
   // to be decomposed based on the alignment factor.
   unsigned SrcBytes = LT.second.getStoreSize();

diff --git a/test/CodeGen/PowerPC/vsx.ll b/test/CodeGen/PowerPC/vsx.ll
index d2f8bb8ee55ffec810671c102f1660f7a908518f..36848427a552344f55dcd01763e673a9e4aea83d 100644 (file)
--- a/test/CodeGen/PowerPC/vsx.ll
+++ b/test/CodeGen/PowerPC/vsx.ll
@@ -314,6 +314,24 @@ define void @test29(<2 x double>* %a, <2 x double> %b) {
 ; CHECK: blr
 }
 
+define <2 x double> @test28u(<2 x double>* %a) {
+  %v = load <2 x double>* %a, align 8
+  ret <2 x double> %v
+
+; CHECK-LABEL: @test28u
+; CHECK: lxvd2x 34, 0, 3
+; CHECK: blr
+}
+
+define void @test29u(<2 x double>* %a, <2 x double> %b) {
+  store <2 x double> %b, <2 x double>* %a, align 8
+  ret void
+
+; CHECK-LABEL: @test29u
+; CHECK: stxvd2x 34, 0, 3
+; CHECK: blr
+}
+
 define <2 x i64> @test30(<2 x i64>* %a) {
   %v = load <2 x i64>* %a, align 16
   ret <2 x i64> %v