[PowerPC] Emit VMX loads/stores for aligned ops to avoid adding swaps on LE

This patch corresponds to review:
https://reviews.llvm.org/D26861

It also fixes PR30730.

Committing on behalf of Lei Huang.

llvm-svn: 287679

2 files changed, 23 insertions, 6 deletions

diff --git a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
index 2c2fa62d551..8232132948f 100644
--- a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
+++ b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -10814,6 +10814,14 @@ SDValue PPCTargetLowering::expandVSXLoadForLE(SDNode *N,
   }
 
   MVT VecTy = N->getValueType(0).getSimpleVT();
+
+  // Do not expand to PPCISD::LXVD2X + PPCISD::XXSWAPD when the load is
+  // aligned and the type is a vector with elements up to 4 bytes
+  if (Subtarget.needsSwapsForVSXMemOps() && !(MMO->getAlignment()%16)
+      && VecTy.getScalarSizeInBits() <= 32 ) {
+    return SDValue();
+  }
+
   SDValue LoadOps[] = { Chain, Base };
   SDValue Load = DAG.getMemIntrinsicNode(PPCISD::LXVD2X, dl,
                                          DAG.getVTList(MVT::v2f64, MVT::Other),
@@ -10878,6 +10886,13 @@ SDValue PPCTargetLowering::expandVSXStoreForLE(SDNode *N,
   SDValue Src = N->getOperand(SrcOpnd);
   MVT VecTy = Src.getValueType().getSimpleVT();
 
+  // Do not expand to PPCISD::XXSWAPD and PPCISD::STXVD2X when the load is
+  // aligned and the type is a vector with elements up to 4 bytes
+  if (Subtarget.needsSwapsForVSXMemOps() && !(MMO->getAlignment()%16)
+      && VecTy.getScalarSizeInBits() <= 32 ) {
+    return SDValue();
+  }
+
   // All stores are done as v2f64 and possible bit cast.
   if (VecTy != MVT::v2f64) {
     Src = DAG.getNode(ISD::BITCAST, dl, MVT::v2f64, Src);
diff --git a/llvm/lib/Target/PowerPC/PPCInstrVSX.td b/llvm/lib/Target/PowerPC/PPCInstrVSX.td
index a90989fefe8..c9bd16977ec 100644
--- a/llvm/lib/Target/PowerPC/PPCInstrVSX.td
+++ b/llvm/lib/Target/PowerPC/PPCInstrVSX.td
@@ -138,7 +138,7 @@ let Uses = [RM] in {
     def LXVW4X : XX1Form<31, 780,
                          (outs vsrc:$XT), (ins memrr:$src),
                          "lxvw4x $XT, $src", IIC_LdStLFD,
-                         [(set v4i32:$XT, (int_ppc_vsx_lxvw4x xoaddr:$src))]>;
+                         []>;
   } // mayLoad
 
   // Store indexed instructions
@@ -160,7 +160,7 @@ let Uses = [RM] in {
     def STXVW4X : XX1Form<31, 908,
                          (outs), (ins vsrc:$XT, memrr:$dst),
                          "stxvw4x $XT, $dst", IIC_LdStSTFD,
-                         [(store v4i32:$XT, xoaddr:$dst)]>;
+                         []>;
     }
   } // mayStore
 
@@ -1018,8 +1018,6 @@ let Predicates = [HasVSX, HasOnlySwappingMemOps] in {
   // Stores.
   def : Pat<(int_ppc_vsx_stxvd2x v2f64:$rS, xoaddr:$dst),
             (STXVD2X $rS, xoaddr:$dst)>;
-  def : Pat<(int_ppc_vsx_stxvw4x v4i32:$rS, xoaddr:$dst),
-            (STXVW4X $rS, xoaddr:$dst)>;
   def : Pat<(int_ppc_vsx_stxvd2x_be v2f64:$rS, xoaddr:$dst),
             (STXVD2X $rS, xoaddr:$dst)>;
   def : Pat<(int_ppc_vsx_stxvw4x_be v4i32:$rS, xoaddr:$dst),
@@ -1030,8 +1028,12 @@ let Predicates = [IsBigEndian, HasVSX, HasOnlySwappingMemOps] in {
   def : Pat<(v2f64 (load xoaddr:$src)), (LXVD2X xoaddr:$src)>;
   def : Pat<(v2i64 (load xoaddr:$src)), (LXVD2X xoaddr:$src)>;
   def : Pat<(v4i32 (load xoaddr:$src)), (LXVW4X xoaddr:$src)>;
+  def : Pat<(v4i32 (int_ppc_vsx_lxvw4x xoaddr:$src)), (LXVW4X xoaddr:$src)>;
   def : Pat<(store v2f64:$rS, xoaddr:$dst), (STXVD2X $rS, xoaddr:$dst)>;
   def : Pat<(store v2i64:$rS, xoaddr:$dst), (STXVD2X $rS, xoaddr:$dst)>;
+  def : Pat<(store v4i32:$XT, xoaddr:$dst), (STXVW4X $XT, xoaddr:$dst)>;
+  def : Pat<(int_ppc_vsx_stxvw4x v4i32:$rS, xoaddr:$dst),
+            (STXVW4X $rS, xoaddr:$dst)>;
 }
 
 // Permutes.
@@ -1852,8 +1854,8 @@ let Predicates = [IsLittleEndian, HasVSX] in
   def : Pat<(f64 (vector_extract v2f64:$S, i64:$Idx)),
             (f64 VectorExtractions.LE_VARIABLE_DOUBLE)>;
 
-  def : Pat<(v4i32 (int_ppc_vsx_lxvw4x_be xoaddr:$src)), (LXVW4X xoaddr:$src)>;
-  def : Pat<(v2f64 (int_ppc_vsx_lxvd2x_be xoaddr:$src)), (LXVD2X xoaddr:$src)>;
+def : Pat<(v4i32 (int_ppc_vsx_lxvw4x_be xoaddr:$src)), (LXVW4X xoaddr:$src)>;
+def : Pat<(v2f64 (int_ppc_vsx_lxvd2x_be xoaddr:$src)), (LXVD2X xoaddr:$src)>;
 
 let Predicates = [IsLittleEndian, HasDirectMove] in {
   // v16i8 scalar <-> vector conversions (LE)