Update test case.

[oota-llvm.git] / test / Transforms / LoopVectorize / gcc-examples.ll

diff --git a/test/Transforms/LoopVectorize/gcc-examples.ll b/test/Transforms/LoopVectorize/gcc-examples.ll
index c34fd72a46789adff74f76526b6c02dc2c904383..6a2c2c6dd63ec9980b235fe89c5bfa732c37c50d 100644 (file)
--- a/test/Transforms/LoopVectorize/gcc-examples.ll
+++ b/test/Transforms/LoopVectorize/gcc-examples.ll
@@ -1,4 +1,5 @@
-; RUN: opt < %s  -loop-vectorize -force-vector-width=4 -dce -instcombine -licm -S | FileCheck %s
+; RUN: opt < %s  -basicaa -loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -dce -instcombine -S | FileCheck %s
+; RUN: opt < %s  -basicaa -loop-vectorize -force-vector-width=4 -force-vector-interleave=4 -dce -instcombine -S | FileCheck %s -check-prefix=UNROLL
 
 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"
@@ -19,11 +20,25 @@ target triple = "x86_64-apple-macosx10.8.0"
 @dd = common global [1024 x float] zeroinitializer, align 16
 @dj = common global [1024 x i32] zeroinitializer, align 16
 
-;CHECK: @example1
+;CHECK-LABEL: @example1(
 ;CHECK: load <4 x i32>
 ;CHECK: add nsw <4 x i32>
 ;CHECK: store <4 x i32>
 ;CHECK: ret void
+;UNROLL-LABEL: @example1(
+;UNROLL: load <4 x i32>
+;UNROLL: load <4 x i32>
+;UNROLL: load <4 x i32>
+;UNROLL: load <4 x i32>
+;UNROLL: add nsw <4 x i32>
+;UNROLL: add nsw <4 x i32>
+;UNROLL: add nsw <4 x i32>
+;UNROLL: add nsw <4 x i32>
+;UNROLL: store <4 x i32>
+;UNROLL: store <4 x i32>
+;UNROLL: store <4 x i32>
+;UNROLL: store <4 x i32>
+;UNROLL: ret void
 define void @example1() nounwind uwtable ssp {
   br label %1
 
@@ -45,9 +60,15 @@ define void @example1() nounwind uwtable ssp {
   ret void
 }
 
-;CHECK: @example2
+;CHECK-LABEL: @example2(
 ;CHECK: store <4 x i32>
 ;CHECK: ret void
+;UNROLL-LABEL: @example2(
+;UNROLL: store <4 x i32>
+;UNROLL: store <4 x i32>
+;UNROLL: store <4 x i32>
+;UNROLL: store <4 x i32>
+;UNROLL: ret void
 define void @example2(i32 %n, i32 %x) nounwind uwtable ssp {
   %1 = icmp sgt i32 %n, 0
   br i1 %1, label %.lr.ph5, label %.preheader
@@ -89,10 +110,15 @@ define void @example2(i32 %n, i32 %x) nounwind uwtable ssp {
   ret void
 }
 
-; We can't vectorize this loop because it has non constant loop bounds.
-;CHECK: @example3
-;CHECK-NOT: <4 x i32>
+;CHECK-LABEL: @example3(
+;CHECK: <4 x i32>
 ;CHECK: ret void
+;UNROLL-LABEL: @example3(
+;UNROLL: <4 x i32>
+;UNROLL: <4 x i32>
+;UNROLL: <4 x i32>
+;UNROLL: <4 x i32>
+;UNROLL: ret void
 define void @example3(i32 %n, i32* noalias nocapture %p, i32* noalias nocapture %q) nounwind uwtable ssp {
   %1 = icmp eq i32 %n, 0
   br i1 %1, label %._crit_edge, label %.lr.ph
@@ -113,9 +139,15 @@ define void @example3(i32 %n, i32* noalias nocapture %p, i32* noalias nocapture
   ret void
 }
 
-;CHECK: @example4
+;CHECK-LABEL: @example4(
 ;CHECK: load <4 x i32>
 ;CHECK: ret void
+;UNROLL-LABEL: @example4(
+;UNROLL: load <4 x i32>
+;UNROLL: load <4 x i32>
+;UNROLL: load <4 x i32>
+;UNROLL: load <4 x i32>
+;UNROLL: ret void
 define void @example4(i32 %n, i32* noalias nocapture %p, i32* noalias nocapture %q) nounwind uwtable ssp {
   %1 = add nsw i32 %n, -1
   %2 = icmp eq i32 %n, 0
@@ -173,9 +205,15 @@ define void @example4(i32 %n, i32* noalias nocapture %p, i32* noalias nocapture
   ret void
 }
 
-;CHECK: @example8
+;CHECK-LABEL: @example8(
 ;CHECK: store <4 x i32>
 ;CHECK: ret void
+;UNROLL-LABEL: @example8(
+;UNROLL: store <4 x i32>
+;UNROLL: store <4 x i32>
+;UNROLL: store <4 x i32>
+;UNROLL: store <4 x i32>
+;UNROLL: ret void
 define void @example8(i32 %x) nounwind uwtable ssp {
   br label %.preheader
 
@@ -202,7 +240,7 @@ define void @example8(i32 %x) nounwind uwtable ssp {
   ret void
 }
 
-;CHECK: @example9
+;CHECK-LABEL: @example9(
 ;CHECK: phi <4 x i32>
 ;CHECK: ret i32
 define i32 @example9() nounwind uwtable readonly ssp {
@@ -226,7 +264,7 @@ define i32 @example9() nounwind uwtable readonly ssp {
   ret i32 %7
 }
 
-;CHECK: @example10a
+;CHECK-LABEL: @example10a(
 ;CHECK: load <4 x i32>
 ;CHECK: add nsw <4 x i32>
 ;CHECK: load <4 x i16>
@@ -261,7 +299,7 @@ define void @example10a(i16* noalias nocapture %sa, i16* noalias nocapture %sb,
   ret void
 }
 
-;CHECK: @example10b
+;CHECK-LABEL: @example10b(
 ;CHECK: load <4 x i16>
 ;CHECK: sext <4 x i16>
 ;CHECK: store <4 x i32>
@@ -285,7 +323,7 @@ define void @example10b(i16* noalias nocapture %sa, i16* noalias nocapture %sb,
   ret void
 }
 
-;CHECK: @example11
+;CHECK-LABEL: @example11(
 ;CHECK: load i32
 ;CHECK: load i32
 ;CHECK: load i32
@@ -329,8 +367,8 @@ define void @example11() nounwind uwtable ssp {
   ret void
 }
 
-;CHECK: @example12
-;CHECK: trunc <4 x i64>
+;CHECK-LABEL: @example12(
+;CHECK: trunc i64
 ;CHECK: store <4 x i32>
 ;CHECK: ret void
 define void @example12() nounwind uwtable ssp {
@@ -350,9 +388,8 @@ define void @example12() nounwind uwtable ssp {
   ret void
 }
 
-; Can't vectorize because of reductions.
-;CHECK: @example13
-;CHECK-NOT: <4 x i32>
+;CHECK-LABEL: @example13(
+;CHECK: <4 x i32>
 ;CHECK: ret void
 define void @example13(i32** nocapture %A, i32** nocapture %B, i32* nocapture %out) nounwind uwtable ssp {
   br label %.preheader
@@ -392,7 +429,7 @@ define void @example13(i32** nocapture %A, i32** nocapture %B, i32* nocapture %o
 }
 
 ; Can vectorize.
-;CHECK: @example14
+;CHECK-LABEL: @example14(
 ;CHECK: <4 x i32>
 ;CHECK: ret void
 define void @example14(i32** nocapture %in, i32** nocapture %coeff, i32* nocapture %out) nounwind uwtable ssp {
@@ -537,9 +574,9 @@ define void @example14(i32** nocapture %in, i32** nocapture %coeff, i32* nocaptu
   ret void
 }
 
-; Can't vectorize because the src and dst pointers are not disjoint.
-;CHECK: @example21
-;CHECK-NOT: <4 x i32>
+;CHECK-LABEL: @example21(
+;CHECK: load <4 x i32>
+;CHECK: shufflevector {{.*}} <i32 3, i32 2, i32 1, i32 0>
 ;CHECK: ret i32
 define i32 @example21(i32* nocapture %b, i32 %n) nounwind uwtable readonly ssp {
   %1 = icmp sgt i32 %n, 0
@@ -565,7 +602,7 @@ define i32 @example21(i32* nocapture %b, i32 %n) nounwind uwtable readonly ssp {
   ret i32 %a.0.lcssa
 }
 
-;CHECK: @example23
+;CHECK-LABEL: @example23(
 ;CHECK: <4 x i32>
 ;CHECK: ret void
 define void @example23(i16* nocapture %src, i32* nocapture %dst) nounwind uwtable ssp {
@@ -589,7 +626,7 @@ define void @example23(i16* nocapture %src, i32* nocapture %dst) nounwind uwtabl
   ret void
 }
 
-;CHECK: @example24
+;CHECK-LABEL: @example24(
 ;CHECK: shufflevector <4 x i16>
 ;CHECK: ret void
 define void @example24(i16 signext %x, i16 signext %y) nounwind uwtable ssp {
@@ -615,7 +652,7 @@ define void @example24(i16 signext %x, i16 signext %y) nounwind uwtable ssp {
   ret void
 }
 
-;CHECK: @example25
+;CHECK-LABEL: @example25(
 ;CHECK: and <4 x i1>
 ;CHECK: zext <4 x i1>
 ;CHECK: ret void