Cleanup PALIGNR handling and remove the old palign pattern fragment.

Also make PALIGNR masks to don't match 256-bits, which isn't supported
It's also a step to solve PR10489

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@136448 91177308-0d34-0410-b5e6-96231b3b80d8

4 files changed, 23 insertions, 43 deletions

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index e969553..5dae745 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -3076,6 +3076,8 @@ bool X86::isPSHUFLWMask(ShuffleVectorSDNode *N) {
 static bool isPALIGNRMask(const SmallVectorImpl<int> &Mask, EVT VT,
                           bool hasSSSE3) {
   int i, e = VT.getVectorNumElements();
+  if (VT.getSizeInBits() != 128 && VT.getSizeInBits() != 64)
+    return false;
 
   // Do not handle v2i64 / v2f64 shuffles with palignr.
   if (e < 4 || !hasSSSE3)
@@ -3104,12 +3106,6 @@ static bool isPALIGNRMask(const SmallVectorImpl<int> &Mask, EVT VT,
   return true;
 }
 
-bool X86::isPALIGNRMask(ShuffleVectorSDNode *N) {
-  SmallVector<int, 8> M;
-  N->getMask(M);
-  return ::isPALIGNRMask(M, N->getValueType(0), true);
-}
-
 /// isSHUFPMask - Return true if the specified VECTOR_SHUFFLE operand
 /// specifies a shuffle of elements that is suitable for input to SHUFP*.
 static bool isSHUFPMask(const SmallVectorImpl<int> &Mask, EVT VT) {
diff --git a/lib/Target/X86/X86ISelLowering.h b/lib/Target/X86/X86ISelLowering.h
index 845efcd..76267c1 100644
--- a/lib/Target/X86/X86ISelLowering.h
+++ b/lib/Target/X86/X86ISelLowering.h
@@ -421,10 +421,6 @@ namespace llvm {
     /// specifies a shuffle of elements that is suitable for input to MOVDDUP.
     bool isMOVDDUPMask(ShuffleVectorSDNode *N);
 
-    /// isPALIGNRMask - Return true if the specified VECTOR_SHUFFLE operand
-    /// specifies a shuffle of elements that is suitable for input to PALIGNR.
-    bool isPALIGNRMask(ShuffleVectorSDNode *N);
-
     /// isVEXTRACTF128Index - Return true if the specified
     /// EXTRACT_SUBVECTOR operand specifies a vector extract that is
     /// suitable for input to VEXTRACTF128.
diff --git a/lib/Target/X86/X86InstrFragmentsSIMD.td b/lib/Target/X86/X86InstrFragmentsSIMD.td
index 4e8d6ef..670bbc5 100644
--- a/lib/Target/X86/X86InstrFragmentsSIMD.td
+++ b/lib/Target/X86/X86InstrFragmentsSIMD.td
@@ -436,11 +436,6 @@ def pshuflw : PatFrag<(ops node:$lhs, node:$rhs),
   return X86::isPSHUFLWMask(cast<ShuffleVectorSDNode>(N));
 }], SHUFFLE_get_pshuflw_imm>;
 
-def palign : PatFrag<(ops node:$lhs, node:$rhs),
-                     (vector_shuffle node:$lhs, node:$rhs), [{
-  return X86::isPALIGNRMask(cast<ShuffleVectorSDNode>(N));
-}], SHUFFLE_get_palign_imm>;
-
 def vextractf128_extract : PatFrag<(ops node:$bigvec, node:$index),
                                    (extract_subvector node:$bigvec,
                                                       node:$index), [{
diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index db1ab52..441c146 100644
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -3641,26 +3641,29 @@ multiclass ssse3_palign<string asm, bit Is2Addr = 1> {
 
 let Predicates = [HasAVX] in
   defm VPALIGN : ssse3_palign<"vpalignr", 0>, VEX_4V;
-let Constraints = "$src1 = $dst" in
+let Constraints = "$src1 = $dst", Predicates = [HasSSSE3] in
   defm PALIGN : ssse3_palign<"palignr">;
 
-let AddedComplexity = 5 in {
-def : Pat<(v4i32 (palign:$src3 VR128:$src1, VR128:$src2)),
-          (PALIGNR128rr VR128:$src2, VR128:$src1,
-                        (SHUFFLE_get_palign_imm VR128:$src3))>,
-      Requires<[HasSSSE3]>;
-def : Pat<(v4f32 (palign:$src3 VR128:$src1, VR128:$src2)),
-          (PALIGNR128rr VR128:$src2, VR128:$src1,
-                        (SHUFFLE_get_palign_imm VR128:$src3))>,
-      Requires<[HasSSSE3]>;
-def : Pat<(v8i16 (palign:$src3 VR128:$src1, VR128:$src2)),
-          (PALIGNR128rr VR128:$src2, VR128:$src1,
-                        (SHUFFLE_get_palign_imm VR128:$src3))>,
-      Requires<[HasSSSE3]>;
-def : Pat<(v16i8 (palign:$src3 VR128:$src1, VR128:$src2)),
-          (PALIGNR128rr VR128:$src2, VR128:$src1,
-                        (SHUFFLE_get_palign_imm VR128:$src3))>,
-      Requires<[HasSSSE3]>;
+let Predicates = [HasSSSE3] in {
+def : Pat<(v4i32 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
+          (PALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
+def : Pat<(v4f32 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
+          (PALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
+def : Pat<(v8i16 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
+          (PALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
+def : Pat<(v16i8 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
+          (PALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
+}
+
+let Predicates = [HasAVX] in {
+def : Pat<(v4i32 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
+          (VPALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
+def : Pat<(v4f32 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
+          (VPALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
+def : Pat<(v8i16 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
+          (VPALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
+def : Pat<(v16i8 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
+          (VPALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
 }
 
 //===---------------------------------------------------------------------===//
@@ -5845,16 +5848,6 @@ def : Pat<(v8i16 (X86PShuflw VR128:$src, (i8 imm:$imm))),
 def : Pat<(v8i16 (X86PShuflw (bc_v8i16 (memopv2i64 addr:$src)), (i8 imm:$imm))),
           (PSHUFLWmi addr:$src, imm:$imm)>;
 
-// Shuffle with PALIGN
-def : Pat<(v4i32 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
-          (PALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
-def : Pat<(v4f32 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
-          (PALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
-def : Pat<(v8i16 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
-          (PALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
-def : Pat<(v16i8 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
-          (PALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
-
 // Shuffle with MOVLPS
 def : Pat<(v4f32 (X86Movlps VR128:$src1, (load addr:$src2))),
           (MOVLPSrm VR128:$src1, addr:$src2)>;