CodeGen peephole: fold redundant phys reg copies

[oota-llvm.git] / test / CodeGen / X86 / atomic-dagsched.ll

diff --git a/test/CodeGen/X86/atomic-dagsched.ll b/test/CodeGen/X86/atomic-dagsched.ll
index 00891d6434c196ea180e2f35b7243bad3a113442..97bb1afa47a7b298a5808566e0964add52363531 100644 (file)
--- a/test/CodeGen/X86/atomic-dagsched.ll
+++ b/test/CodeGen/X86/atomic-dagsched.ll
@@ -1,13 +1,13 @@
-; RUN: llc < %s -march=x86-64 -mcpu=corei7 | FileCheck %s
+; RUN: llc < %s -march=x86-64 -mcpu=corei7 -verify-machineinstrs | FileCheck %s
 
 define void @test(i8** %a, i64* %b, i64 %c, i64 %d) nounwind {
 entry:
-  %ptrtoarg4 = load i8** %a, align 8
-  %brglist1 = getelementptr i8** %a, i64 1
-  %ptrtoarg25 = load i8** %brglist1, align 8
-  %0 = load i64* %b, align 8
+  %ptrtoarg4 = load i8*, i8** %a, align 8
+  %brglist1 = getelementptr i8*, i8** %a, i64 1
+  %ptrtoarg25 = load i8*, i8** %brglist1, align 8
+  %0 = load i64, i64* %b, align 8
   %1 = mul i64 %0, 4
-  %scevgep = getelementptr i8* %ptrtoarg25, i64 %1
+  %scevgep = getelementptr i8, i8* %ptrtoarg25, i64 %1
   %2 = mul i64 %d, 4
   br label %loop.cond
 
@@ -18,8 +18,8 @@ loop.cond:                                        ; preds = %test.exit, %entry
   br i1 %3, label %return, label %loop
 
 loop:                                             ; preds = %loop.cond
-  %4 = load i64* addrspace(256)* inttoptr (i64 264 to i64* addrspace(256)*), align 8, !tbaa !0
-  %5 = load i64* %4, align 8, !tbaa !3
+  %4 = load i64*, i64* addrspace(256)* inttoptr (i64 264 to i64* addrspace(256)*), align 8
+  %5 = load i64, i64* %4, align 8
   %vector.size.i = ashr i64 %5, 3
   %num.vector.wi.i = shl i64 %vector.size.i, 3
   %6 = icmp eq i64 %vector.size.i, 0
@@ -34,9 +34,9 @@ dim_0_vector_pre_head.i:                          ; preds = %loop
 vector_kernel_entry.i:                            ; preds = %vector_kernel_entry.i, %dim_0_vector_pre_head.i
   %asr.iv9 = phi i8* [ %scevgep10, %vector_kernel_entry.i ], [ %asr.iv6, %dim_0_vector_pre_head.i ]
   %asr.iv = phi i64 [ %asr.iv.next, %vector_kernel_entry.i ], [ %vector.size.i, %dim_0_vector_pre_head.i ]
-  %8 = bitcast i8* %ptrtoarg4 to i32 addrspace(1)*
-  %asr.iv911 = bitcast i8* %asr.iv9 to <8 x i32> addrspace(1)*
-  %9 = load <8 x i32> addrspace(1)* %asr.iv911, align 4
+  %8 = addrspacecast i8* %ptrtoarg4 to i32 addrspace(1)*
+  %asr.iv911 = addrspacecast i8* %asr.iv9 to <8 x i32> addrspace(1)*
+  %9 = load <8 x i32>, <8 x i32> addrspace(1)* %asr.iv911, align 4
   %extract8vector_func.i = extractelement <8 x i32> %9, i32 0
   %extract9vector_func.i = extractelement <8 x i32> %9, i32 1
   %extract10vector_func.i = extractelement <8 x i32> %9, i32 2
@@ -55,7 +55,7 @@ vector_kernel_entry.i:                            ; preds = %vector_kernel_entry
   %17 = atomicrmw min i32 addrspace(1)* %8, i32 %extract15vector_func.i seq_cst
   store <8 x i32> %vectorvector_func.i, <8 x i32> addrspace(1)* %asr.iv911, align 4
   %asr.iv.next = add i64 %asr.iv, -1
-  %scevgep10 = getelementptr i8* %asr.iv9, i64 32
+  %scevgep10 = getelementptr i8, i8* %asr.iv9, i64 32
   %dim_0_vector_cmp.to.max.i = icmp eq i64 %asr.iv.next, 0
   br i1 %dim_0_vector_cmp.to.max.i, label %scalarIf.i, label %vector_kernel_entry.i
 
@@ -65,21 +65,21 @@ scalarIf.i:                                       ; preds = %vector_kernel_entry
   br i1 %18, label %test.exit, label %dim_0_pre_head.i
 
 dim_0_pre_head.i:                                 ; preds = %scalarIf.i
-  %19 = load i64* addrspace(256)* inttoptr (i64 264 to i64* addrspace(256)*), align 8, !tbaa !0
-  %20 = load i64* %19, align 8, !tbaa !3
+  %19 = load i64*, i64* addrspace(256)* inttoptr (i64 264 to i64* addrspace(256)*), align 8
+  %20 = load i64, i64* %19, align 8
   %21 = trunc i64 %20 to i32
   %22 = mul i64 %vector.size.i, 8
   br label %scalar_kernel_entry.i
 
 scalar_kernel_entry.i:                            ; preds = %scalar_kernel_entry.i, %dim_0_pre_head.i
   %asr.iv12 = phi i64 [ %asr.iv.next13, %scalar_kernel_entry.i ], [ %22, %dim_0_pre_head.i ]
-  %23 = bitcast i8* %asr.iv6 to i32 addrspace(1)*
-  %24 = bitcast i8* %ptrtoarg4 to i32 addrspace(1)*
-  %scevgep16 = getelementptr i32 addrspace(1)* %23, i64 %asr.iv12
-  %25 = load i32 addrspace(1)* %scevgep16, align 4, !tbaa !4
+  %23 = addrspacecast i8* %asr.iv6 to i32 addrspace(1)*
+  %24 = addrspacecast i8* %ptrtoarg4 to i32 addrspace(1)*
+  %scevgep16 = getelementptr i32, i32 addrspace(1)* %23, i64 %asr.iv12
+  %25 = load i32, i32 addrspace(1)* %scevgep16, align 4
   %26 = atomicrmw min i32 addrspace(1)* %24, i32 %25 seq_cst
-  %scevgep15 = getelementptr i32 addrspace(1)* %23, i64 %asr.iv12
-  store i32 %21, i32 addrspace(1)* %scevgep15, align 4, !tbaa !4
+  %scevgep15 = getelementptr i32, i32 addrspace(1)* %23, i64 %asr.iv12
+  store i32 %21, i32 addrspace(1)* %scevgep15, align 4
   %asr.iv.next13 = add i64 %asr.iv12, 1
   %dim_0_cmp.to.max.i = icmp eq i64 %5, %asr.iv.next13
   br i1 %dim_0_cmp.to.max.i, label %test.exit, label %scalar_kernel_entry.i
@@ -88,7 +88,7 @@ test.exit:                     ; preds = %scalar_kernel_entry.i, %scalarIf.i
   %27 = bitcast i8* %asr.iv6 to i1*
   %28 = add i64 %iv, %d
   store i64 %28, i64* %b, align 8
-  %scevgep8 = getelementptr i1* %27, i64 %2
+  %scevgep8 = getelementptr i1, i1* %27, i64 %2
   %29 = bitcast i1* %scevgep8 to i8*
   br label %loop.cond
 
@@ -97,12 +97,6 @@ return:                                           ; preds = %loop.cond
   ret void
 }
 
-!0 = metadata !{metadata !"any pointer", metadata !1}
-!1 = metadata !{metadata !"omnipotent char", metadata !2}
-!2 = metadata !{metadata !"Simple C/C++ TBAA"}
-!3 = metadata !{metadata !"long", metadata !1}
-!4 = metadata !{metadata !"int", metadata !1}
-
 ; CHECK: test
 ; CHECK: decq
 ; CHECK-NOT: cmpxchgl