From 859366f93f2a1e443bb0a71da91d42382087b573 Mon Sep 17 00:00:00 2001
From: Nadav Rotem <nrotem@apple.com>
Date: Sat, 27 Oct 2012 04:11:32 +0000
Subject: [PATCH] 1. Fix a bug in getTypeConversion. When a *simple* type is
 split, we need to return the type of the split result. 2. Change the maximum
 vectorization width from 4 to 8. 3. A test for both.

llvm-svn: 166864
---
 llvm/include/llvm/Target/TargetLowering.h          |  3 ++
 llvm/lib/Transforms/Vectorize/LoopVectorize.cpp    |  2 +-
 .../Transforms/LoopVectorize/X86/gcc-examples.ll   | 62 ++++++++++++++++++++++
 3 files changed, 66 insertions(+), 1 deletion(-)
 create mode 100644 llvm/test/Transforms/LoopVectorize/X86/gcc-examples.ll

diff --git a/llvm/include/llvm/Target/TargetLowering.h b/llvm/include/llvm/Target/TargetLowering.h
index 9d0aeaa..13f80fd 100644
--- a/llvm/include/llvm/Target/TargetLowering.h
+++ b/llvm/include/llvm/Target/TargetLowering.h
@@ -1980,6 +1980,9 @@ public:
          ValueTypeActions.getTypeAction(NVT.getSimpleVT()) != TypePromoteInteger)
          && "Promote may not follow Expand or Promote");
 
+      if (LA == TypeSplitVector)
+        NVT = EVT::getVectorVT(Context, VT.getVectorElementType(),
+                               VT.getVectorNumElements() / 2);
       return LegalizeKind(LA, NVT);
     }
 
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 1773812..be197db 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -317,7 +317,7 @@ public:
   /// Returns the most profitable vectorization factor for the loop that is
   /// smaller or equal to the VF argument. This method checks every power
   /// of two up to VF.
-  unsigned findBestVectorizationFactor(unsigned VF = 4);
+  unsigned findBestVectorizationFactor(unsigned VF = 8);
 
 private:
   /// Returns the expected execution cost. The unit of the cost does
diff --git a/llvm/test/Transforms/LoopVectorize/X86/gcc-examples.ll b/llvm/test/Transforms/LoopVectorize/X86/gcc-examples.ll
new file mode 100644
index 0000000..e7a63c9
--- /dev/null
+++ b/llvm/test/Transforms/LoopVectorize/X86/gcc-examples.ll
@@ -0,0 +1,62 @@
+; RUN: opt < %s  -loop-vectorize -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 -dce -instcombine -licm -S | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+@b = common global [2048 x i32] zeroinitializer, align 16
+@c = common global [2048 x i32] zeroinitializer, align 16
+@a = common global [2048 x i32] zeroinitializer, align 16
+
+; Select VF = 8;
+;CHECK: @example1
+;CHECK: load <8 x i32>
+;CHECK: add <8 x i32>
+;CHECK: store <8 x i32>
+;CHECK: ret void
+define void @example1() nounwind uwtable ssp {
+  br label %1
+
+; <label>:1                                       ; preds = %1, %0
+  %indvars.iv = phi i64 [ 0, %0 ], [ %indvars.iv.next, %1 ]
+  %2 = getelementptr inbounds [2048 x i32]* @b, i64 0, i64 %indvars.iv
+  %3 = load i32* %2, align 4
+  %4 = getelementptr inbounds [2048 x i32]* @c, i64 0, i64 %indvars.iv
+  %5 = load i32* %4, align 4
+  %6 = add nsw i32 %5, %3
+  %7 = getelementptr inbounds [2048 x i32]* @a, i64 0, i64 %indvars.iv
+  store i32 %6, i32* %7, align 4
+  %indvars.iv.next = add i64 %indvars.iv, 1
+  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
+  %exitcond = icmp eq i32 %lftr.wideiv, 256
+  br i1 %exitcond, label %8, label %1
+
+; <label>:8                                       ; preds = %1
+  ret void
+}
+
+
+; Select VF=4 because sext <8 x i1> to <8 x i32> is expensive. 
+;CHECK: @example10b
+;CHECK: load <4 x i16>
+;CHECK: sext <4 x i16>
+;CHECK: store <4 x i32>
+;CHECK: ret void
+define void @example10b(i16* noalias nocapture %sa, i16* noalias nocapture %sb, i16* noalias nocapture %sc, i32* noalias nocapture %ia, i32* noalias nocapture %ib, i32* noalias nocapture %ic) nounwind uwtable ssp {
+  br label %1
+
+; <label>:1                                       ; preds = %1, %0
+  %indvars.iv = phi i64 [ 0, %0 ], [ %indvars.iv.next, %1 ]
+  %2 = getelementptr inbounds i16* %sb, i64 %indvars.iv
+  %3 = load i16* %2, align 2
+  %4 = sext i16 %3 to i32
+  %5 = getelementptr inbounds i32* %ia, i64 %indvars.iv
+  store i32 %4, i32* %5, align 4
+  %indvars.iv.next = add i64 %indvars.iv, 1
+  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
+  %exitcond = icmp eq i32 %lftr.wideiv, 1024
+  br i1 %exitcond, label %6, label %1
+
+; <label>:6                                       ; preds = %1
+  ret void
+}
+
-- 
2.7.4