From 1d5a8cca004c29d99526b3c327d948f8ab197c90 Mon Sep 17 00:00:00 2001
From: Evan Cheng <evan.cheng@apple.com>
Date: Wed, 5 Apr 2006 07:20:06 +0000
Subject: [PATCH] Handle canonical form of e.g. vector_shuffle v1, v1, <0, 4,
 1, 5, 2, 6, 3, 7>

This is turned into
vector_shuffle v1, <undef>, <0, 0, 1, 1, 2, 2, 3, 3>
by dag combiner.

It would match a {p}unpckl on x86.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@27437 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/X86/X86ISelLowering.cpp | 33 ++++++++++++++++++++++++++++++
 lib/Target/X86/X86ISelLowering.h   |  5 +++++
 lib/Target/X86/X86InstrSSE.td      | 18 ++++++++++++++++
 3 files changed, 56 insertions(+)

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index a2f4cbe8202..976286678da 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -1664,6 +1664,37 @@ bool X86::isUNPCKHMask(SDNode *N) {
   return true;
 }
 
+/// isUNPCKL_v_undef_Mask - Special case of isUNPCKLMask for canonical form
+/// of vector_shuffle v, v, <0, 4, 1, 5>, i.e. vector_shuffle v, undef,
+/// <0, 0, 1, 1>
+bool X86::isUNPCKL_v_undef_Mask(SDNode *N) {
+  assert(N->getOpcode() == ISD::BUILD_VECTOR);
+
+  unsigned NumElems = N->getNumOperands();
+  if (NumElems != 4 && NumElems != 8 && NumElems != 16)
+    return false;
+
+  for (unsigned i = 0, j = 0; i != NumElems; i += 2, ++j) {
+    SDOperand BitI  = N->getOperand(i);
+    SDOperand BitI1 = N->getOperand(i+1);
+
+    if (BitI.getOpcode() != ISD::UNDEF) {
+      assert(isa<ConstantSDNode>(BitI) && "Invalid VECTOR_SHUFFLE mask!");
+      if (cast<ConstantSDNode>(BitI)->getValue() != j)
+        return false;
+    }
+
+    if (BitI1.getOpcode() != ISD::UNDEF) {
+      assert(isa<ConstantSDNode>(BitI1) && "Invalid VECTOR_SHUFFLE mask!");
+      if (cast<ConstantSDNode>(BitI1)->getValue() != j)
+        return false;
+    }
+  }
+
+  return true;
+}
+
+
 /// isSplatMask - Return true if the specified VECTOR_SHUFFLE operand specifies
 /// a splat of a single element.
 bool X86::isSplatMask(SDNode *N) {
@@ -2604,6 +2635,7 @@ SDOperand X86TargetLowering::LowerOperation(SDOperand Op, SelectionDAG &DAG) {
     }
 
     if (X86::isUNPCKLMask(PermMask.Val) ||
+        X86::isUNPCKL_v_undef_Mask(PermMask.Val) ||
         X86::isUNPCKHMask(PermMask.Val))
       // Leave the VECTOR_SHUFFLE alone. It matches {P}UNPCKL*.
       return Op;
@@ -2929,5 +2961,6 @@ X86TargetLowering::isShuffleMaskLegal(SDOperand Mask, MVT::ValueType VT) const {
           isPSHUFHW_PSHUFLWMask(Mask.Val) ||
           X86::isSHUFPMask(Mask.Val) ||
           X86::isUNPCKLMask(Mask.Val) ||
+          X86::isUNPCKL_v_undef_Mask(Mask.Val) ||
           X86::isUNPCKHMask(Mask.Val));
 }
diff --git a/lib/Target/X86/X86ISelLowering.h b/lib/Target/X86/X86ISelLowering.h
index 9722d6fd4fa..ddb2d307ba9 100644
--- a/lib/Target/X86/X86ISelLowering.h
+++ b/lib/Target/X86/X86ISelLowering.h
@@ -220,6 +220,11 @@ namespace llvm {
    /// specifies a shuffle of elements that is suitable for input to UNPCKH.
    bool isUNPCKHMask(SDNode *N);
 
+   /// isUNPCKL_v_undef_Mask - Special case of isUNPCKLMask for canonical form
+   /// of vector_shuffle v, v, <0, 4, 1, 5>, i.e. vector_shuffle v, undef,
+   /// <0, 0, 1, 1>
+   bool isUNPCKL_v_undef_Mask(SDNode *N);
+
    /// isSplatMask - Return true if the specified VECTOR_SHUFFLE operand
    /// specifies a splat of a single element.
    bool isSplatMask(SDNode *N);
diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index 3fd71c89dfc..bdf5fe2de85 100644
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -100,6 +100,10 @@ def UNPCKH_shuffle_mask : PatLeaf<(build_vector), [{
   return X86::isUNPCKHMask(N);
 }]>;
 
+def UNPCKL_v_undef_shuffle_mask : PatLeaf<(build_vector), [{
+  return X86::isUNPCKL_v_undef_Mask(N);
+}]>;
+
 def PSHUFD_shuffle_mask : PatLeaf<(build_vector), [{
   return X86::isPSHUFDMask(N);
 }], SHUFFLE_get_shuf_imm>;
@@ -1733,6 +1737,20 @@ def : Pat<(vector_shuffle (loadv4f32 addr:$src1), (undef),
           (v4f32 (PSHUFLWmi addr:$src1, PSHUFLW_fp_shuffle_mask:$sm))>,
       Requires<[HasSSE2]>;
 
+// vector_shuffle v1, <undef>, <0, 0, 1, 1, ...>
+def : Pat<(v4f32 (vector_shuffle VR128:$src, (undef),
+                  UNPCKL_v_undef_shuffle_mask)),
+          (UNPCKLPSrr VR128:$src, VR128:$src)>, Requires<[HasSSE2]>;
+def : Pat<(v16i8 (vector_shuffle VR128:$src, (undef),
+                  UNPCKL_v_undef_shuffle_mask)),
+          (PUNPCKLBWrr VR128:$src, VR128:$src)>, Requires<[HasSSE2]>;
+def : Pat<(v8i16 (vector_shuffle VR128:$src, (undef),
+                  UNPCKL_v_undef_shuffle_mask)),
+          (PUNPCKLWDrr VR128:$src, VR128:$src)>, Requires<[HasSSE2]>;
+def : Pat<(v4i32 (vector_shuffle VR128:$src, (undef),
+                  UNPCKL_v_undef_shuffle_mask)),
+          (PUNPCKLDQrr VR128:$src, VR128:$src)>, Requires<[HasSSE1]>;
+
 // 128-bit logical shifts
 def : Pat<(int_x86_sse2_psll_dq VR128:$src1, imm:$src2),
           (v2i64 (PSLLDQri VR128:$src1, (PSxLDQ_imm imm:$src2)))>;