LoopVectorizer: Add a basic cost model which uses the VTTI interface.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@166620 91177308-0d34-0410-b5e6-96231b3b80d8

10 files changed, 49 insertions, 9 deletions

diff --git a/test/Transforms/LoopVectorize/2012-10-20-infloop.ll b/test/Transforms/LoopVectorize/2012-10-20-infloop.ll
index 5caaffc8dd..0176c9a189 100644
--- a/test/Transforms/LoopVectorize/2012-10-20-infloop.ll
+++ b/test/Transforms/LoopVectorize/2012-10-20-infloop.ll
@@ -1,4 +1,4 @@
-; RUN: opt < %s  -loop-vectorize -dce
+; RUN: opt < %s  -loop-vectorize -force-vector-width=4 -dce
 
 ; Check that we don't fall into an infinite loop.
 define void @test() nounwind {
diff --git a/test/Transforms/LoopVectorize/cost-model.ll b/test/Transforms/LoopVectorize/cost-model.ll
new file mode 100644
index 0000000000..20c710db46
--- /dev/null
+++ b/test/Transforms/LoopVectorize/cost-model.ll
@@ -0,0 +1,40 @@
+; RUN: opt < %s  -loop-vectorize -mtriple=x86_64-apple-macosx10.8.0 -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"
+
+@c = common global [2048 x i32] zeroinitializer, align 16
+@b = common global [2048 x i32] zeroinitializer, align 16
+@d = common global [2048 x i32] zeroinitializer, align 16
+@a = common global [2048 x i32] zeroinitializer, align 16
+
+; At this point the cost model is pretty bad and we are vectorizing the code below.
+; TODO: This code should not be vectorized on x86.
+;CHECK: cost_model_1
+;CHECK: <4 x i32>
+;CHECK: ret void
+define void @cost_model_1() nounwind uwtable noinline ssp {
+entry:
+  br label %for.body
+
+for.body:                                         ; preds = %for.body, %entry
+  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
+  %0 = shl nsw i64 %indvars.iv, 1
+  %arrayidx = getelementptr inbounds [2048 x i32]* @c, i64 0, i64 %0
+  %1 = load i32* %arrayidx, align 8
+  %idxprom1 = sext i32 %1 to i64
+  %arrayidx2 = getelementptr inbounds [2048 x i32]* @b, i64 0, i64 %idxprom1
+  %2 = load i32* %arrayidx2, align 4
+  %arrayidx4 = getelementptr inbounds [2048 x i32]* @d, i64 0, i64 %indvars.iv
+  %3 = load i32* %arrayidx4, align 4
+  %idxprom5 = sext i32 %3 to i64
+  %arrayidx6 = getelementptr inbounds [2048 x i32]* @a, i64 0, i64 %idxprom5
+  store i32 %2, i32* %arrayidx6, 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 %for.end, label %for.body
+
+for.end:                                          ; preds = %for.body
+  ret void
+}
diff --git a/test/Transforms/LoopVectorize/cpp-new-array.ll b/test/Transforms/LoopVectorize/cpp-new-array.ll
index 12a91b0791..26902eba9e 100644
--- a/test/Transforms/LoopVectorize/cpp-new-array.ll
+++ b/test/Transforms/LoopVectorize/cpp-new-array.ll
@@ -1,4 +1,4 @@
-; RUN: opt < %s  -loop-vectorize -dce -instcombine -licm -S | FileCheck %s
+; RUN: opt < %s  -loop-vectorize -force-vector-width=4 -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"
diff --git a/test/Transforms/LoopVectorize/gcc-examples.ll b/test/Transforms/LoopVectorize/gcc-examples.ll
index 6fb1792b2c..d8942ac861 100644
--- a/test/Transforms/LoopVectorize/gcc-examples.ll
+++ b/test/Transforms/LoopVectorize/gcc-examples.ll
@@ -1,4 +1,4 @@
-; RUN: opt < %s  -loop-vectorize -dce -instcombine -licm -S | FileCheck %s
+; RUN: opt < %s  -loop-vectorize -force-vector-width=4 -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"
diff --git a/test/Transforms/LoopVectorize/increment.ll b/test/Transforms/LoopVectorize/increment.ll
index e818d68562..069b7ea031 100644
--- a/test/Transforms/LoopVectorize/increment.ll
+++ b/test/Transforms/LoopVectorize/increment.ll
@@ -1,4 +1,4 @@
-; RUN: opt < %s  -loop-vectorize -dce -instcombine -licm -S | FileCheck %s
+; RUN: opt < %s  -loop-vectorize -force-vector-width=4 -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"
diff --git a/test/Transforms/LoopVectorize/induction_plus.ll b/test/Transforms/LoopVectorize/induction_plus.ll
index bd90113e52..b31bceb50d 100644
--- a/test/Transforms/LoopVectorize/induction_plus.ll
+++ b/test/Transforms/LoopVectorize/induction_plus.ll
@@ -1,4 +1,4 @@
-; RUN: opt < %s -loop-vectorize -instcombine -S | FileCheck %s
+; RUN: opt < %s -loop-vectorize -force-vector-width=4 -instcombine -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"
diff --git a/test/Transforms/LoopVectorize/non-const-n.ll b/test/Transforms/LoopVectorize/non-const-n.ll
index 04c5c84a4f..7727b0a2dc 100644
--- a/test/Transforms/LoopVectorize/non-const-n.ll
+++ b/test/Transforms/LoopVectorize/non-const-n.ll
@@ -1,4 +1,4 @@
-; RUN: opt < %s  -loop-vectorize -dce -instcombine -licm -S | FileCheck %s
+; RUN: opt < %s  -loop-vectorize -force-vector-width=4 -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"
diff --git a/test/Transforms/LoopVectorize/read-only.ll b/test/Transforms/LoopVectorize/read-only.ll
index 4095ea68ef..b4d1bac132 100644
--- a/test/Transforms/LoopVectorize/read-only.ll
+++ b/test/Transforms/LoopVectorize/read-only.ll
@@ -1,4 +1,4 @@
-; RUN: opt < %s  -loop-vectorize -dce -instcombine -licm -S | FileCheck %s
+; RUN: opt < %s  -loop-vectorize -force-vector-width=4 -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"
diff --git a/test/Transforms/LoopVectorize/reduction.ll b/test/Transforms/LoopVectorize/reduction.ll
index 3e871b229b..ca51c7b37d 100644
--- a/test/Transforms/LoopVectorize/reduction.ll
+++ b/test/Transforms/LoopVectorize/reduction.ll
@@ -1,4 +1,4 @@
-; RUN: opt < %s  -loop-vectorize -dce -instcombine -licm -S | FileCheck %s
+; RUN: opt < %s  -loop-vectorize -force-vector-width=4 -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"
diff --git a/test/Transforms/LoopVectorize/scalar-select.ll b/test/Transforms/LoopVectorize/scalar-select.ll
index 8d5b6fd8af..e537bde31b 100644
--- a/test/Transforms/LoopVectorize/scalar-select.ll
+++ b/test/Transforms/LoopVectorize/scalar-select.ll
@@ -1,4 +1,4 @@
-; RUN: opt < %s  -loop-vectorize -dce -instcombine -licm -S | FileCheck %s
+; RUN: opt < %s  -loop-vectorize -force-vector-width=4 -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"