Index: lib/Analysis/ScopDetection.cpp
===================================================================
--- lib/Analysis/ScopDetection.cpp
+++ lib/Analysis/ScopDetection.cpp
@@ -59,6 +59,7 @@
 #include "llvm/Analysis/ScalarEvolution.h"
 #include "llvm/Analysis/ScalarEvolutionExpressions.h"
 #include "llvm/IR/DebugInfo.h"
+#include "llvm/IR/IntrinsicInst.h"
 #include "llvm/IR/DiagnosticInfo.h"
 #include "llvm/IR/DiagnosticPrinter.h"
 #include "llvm/IR/LLVMContext.h"
@@ -299,7 +300,7 @@
 }
 
 bool ScopDetection::isValidCallInst(CallInst &CI) {
-  if (CI.mayHaveSideEffects() || CI.doesNotReturn())
+  if (CI.doesNotReturn())
     return false;
 
   if (CI.doesNotAccessMemory())
@@ -311,7 +312,29 @@
   if (CalledFunction == 0)
     return false;
 
-  // TODO: Intrinsics.
+  // Check if we can handle the intrinsic call.
+  if (auto *IT = dyn_cast<IntrinsicInst>(&CI)) {
+    switch (IT->getIntrinsicID()) {
+    // Lifetime markers are supported/ignored.
+    case llvm::Intrinsic::lifetime_start:
+    case llvm::Intrinsic::lifetime_end:
+    // Invariant markers are supported/ignored.
+    case llvm::Intrinsic::invariant_start:
+    case llvm::Intrinsic::invariant_end:
+    // Some misc annotations are supported/ignored.
+    case llvm::Intrinsic::var_annotation:
+    case llvm::Intrinsic::ptr_annotation:
+    case llvm::Intrinsic::annotation:
+    case llvm::Intrinsic::donothing:
+    case llvm::Intrinsic::assume:
+    case llvm::Intrinsic::expect:
+      return true;
+    default:
+      // Other intrinsics which may access the memory are not yet supported.
+      break;
+    }
+  }
+
   return false;
 }
 
Index: lib/CodeGen/BlockGenerators.cpp
===================================================================
--- lib/CodeGen/BlockGenerators.cpp
+++ lib/CodeGen/BlockGenerators.cpp
@@ -267,6 +267,30 @@
     return;
   }
 
+  // Skip some special intrinsics for which we do not adjust the semantics to
+  // the new schedule. All others are handled like every other instruction.
+  if (auto *IT = dyn_cast<IntrinsicInst>(Inst)) {
+    switch (IT->getIntrinsicID()) {
+    // Lifetime markers are ignored.
+    case llvm::Intrinsic::lifetime_start:
+    case llvm::Intrinsic::lifetime_end:
+    // Invariant markers are ignored.
+    case llvm::Intrinsic::invariant_start:
+    case llvm::Intrinsic::invariant_end:
+    // Some misc annotations are ignored.
+    case llvm::Intrinsic::var_annotation:
+    case llvm::Intrinsic::ptr_annotation:
+    case llvm::Intrinsic::annotation:
+    case llvm::Intrinsic::donothing:
+    case llvm::Intrinsic::assume:
+    case llvm::Intrinsic::expect:
+      return;
+    default:
+      // Other intrinsics are copied.
+      break;
+    }
+  }
+
   copyInstScalar(Inst, BBMap, GlobalMap, LTS);
 }
 
Index: test/Isl/CodeGen/intrinsics_lifetime.ll
===================================================================
--- /dev/null
+++ test/Isl/CodeGen/intrinsics_lifetime.ll
@@ -0,0 +1,86 @@
+; RUN: opt %loadPolly -basicaa -polly-codegen-isl -polly-codegen-scev -S < %s | FileCheck %s
+;
+; Verify that we remove the lifetime markers from the optimized SCoP.
+;
+; CHECK: for.body:
+; CHECK:   call void @llvm.lifetime.start
+; CHECK: for.end:
+; CHECK:   call void @llvm.lifetime.end
+; CHECK-NOT: call void @llvm.lifetime.start
+; CHECK-NOT: call void @llvm.lifetime.end
+;
+;    int A[1024];
+;    void jd() {
+;      for (int i = 0; i < 1024; i++) {
+;        int tmp[1024];
+;        for (int j = i; j < 1024; j++)
+;          tmp[i] += A[j];
+;        A[i] = tmp[i];
+;      }
+;    }
+;
+; ModuleID = 'test/Isl/CodeGen/lifetime_intrinsics.ll'
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+@A = common global [1024 x i32] zeroinitializer, align 16
+
+; Function Attrs: nounwind uwtable
+define void @jd() #0 {
+entry:
+  %tmp = alloca [1024 x i32], align 16
+  %tmp3 = bitcast [1024 x i32]* %tmp to i8*
+  br label %for.cond
+
+for.cond:                                         ; preds = %for.inc11, %entry
+  %indvars.iv3 = phi i64 [ %indvars.iv.next4, %for.inc11 ], [ 0, %entry ]
+  %exitcond5 = icmp ne i64 %indvars.iv3, 1024
+  br i1 %exitcond5, label %for.body, label %for.end13
+
+for.body:                                         ; preds = %for.cond
+  call void @llvm.lifetime.start(i64 4096, i8* %tmp3) #1
+  br label %for.cond2
+
+for.cond2:                                        ; preds = %for.inc, %for.body
+  %indvars.iv1 = phi i64 [ %indvars.iv.next2, %for.inc ], [ %indvars.iv3, %for.body ]
+  %lftr.wideiv = trunc i64 %indvars.iv1 to i32
+  %exitcond = icmp ne i32 %lftr.wideiv, 1024
+  br i1 %exitcond, label %for.body4, label %for.end
+
+for.body4:                                        ; preds = %for.cond2
+  %arrayidx = getelementptr inbounds [1024 x i32]* @A, i64 0, i64 %indvars.iv1
+  %tmp6 = load i32* %arrayidx, align 4
+  %arrayidx6 = getelementptr inbounds [1024 x i32]* %tmp, i64 0, i64 %indvars.iv3
+  %tmp7 = load i32* %arrayidx6, align 4
+  %add = add nsw i32 %tmp7, %tmp6
+  store i32 %add, i32* %arrayidx6, align 4
+  br label %for.inc
+
+for.inc:                                          ; preds = %for.body4
+  %indvars.iv.next2 = add nuw nsw i64 %indvars.iv1, 1
+  br label %for.cond2
+
+for.end:                                          ; preds = %for.cond2
+  %arrayidx8 = getelementptr inbounds [1024 x i32]* %tmp, i64 0, i64 %indvars.iv3
+  %tmp8 = load i32* %arrayidx8, align 4
+  %arrayidx10 = getelementptr inbounds [1024 x i32]* @A, i64 0, i64 %indvars.iv3
+  store i32 %tmp8, i32* %arrayidx10, align 4
+  call void @llvm.lifetime.end(i64 4096, i8* %tmp3) #1
+  br label %for.inc11
+
+for.inc11:                                        ; preds = %for.end
+  %indvars.iv.next4 = add nuw nsw i64 %indvars.iv3, 1
+  br label %for.cond
+
+for.end13:                                        ; preds = %for.cond
+  ret void
+}
+
+; Function Attrs: nounwind
+declare void @llvm.lifetime.start(i64, i8* nocapture) #1
+
+; Function Attrs: nounwind
+declare void @llvm.lifetime.end(i64, i8* nocapture) #1
+
+attributes #0 = { nounwind uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="true" "no-nans-fp-math"="true" "stack-protector-buffer-size"="8" "unsafe-fp-math"="true" "use-soft-float"="false" }
+attributes #1 = { nounwind }
Index: test/Isl/CodeGen/intrinsics_misc.ll
===================================================================
--- /dev/null
+++ test/Isl/CodeGen/intrinsics_misc.ll
@@ -0,0 +1,101 @@
+; RUN: opt %loadPolly -basicaa -polly-codegen-isl -polly-codegen-scev -S < %s | FileCheck %s
+;
+; Verify that we remove the misc intrinsics  from the optimized SCoP.
+;
+; CHECK: for.body:
+; CHECK:   call {}* @llvm.invariant.start
+; CHECK: for.body4:
+; CHECK:   call void @llvm.assume
+; CHECK:   call i1 @llvm.expect.i1
+; CHECK:   call void @llvm.donothing
+; CHECK: for.end:
+; CHECK:   call void @llvm.invariant.end
+; CHECK-NOT: call void @llvm.{{.*}}
+;
+;    int A[1024];
+;    void jd() {
+;      for (int i = 0; i < 1024; i++) {
+;        int tmp[1024];
+;        for (int j = i; j < 1024; j++)
+;          tmp[i] += A[j];
+;        A[i] = tmp[i];
+;      }
+;    }
+;
+; ModuleID = 'test/Isl/CodeGen/lifetime_intrinsics.ll'
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+@A = common global [1024 x i32] zeroinitializer, align 16
+
+; Function Attrs: nounwind uwtable
+define void @jd() #0 {
+entry:
+  %tmp = alloca [1024 x i32], align 16
+  %tmp3 = bitcast [1024 x i32]* @A to i8*
+  br label %for.cond
+
+for.cond:                                         ; preds = %for.inc11, %entry
+  %indvars.iv3 = phi i64 [ %indvars.iv.next4, %for.inc11 ], [ 0, %entry ]
+  %exitcond5 = icmp ne i64 %indvars.iv3, 1024
+  br i1 %exitcond5, label %for.body, label %for.end13
+
+for.body:                                         ; preds = %for.cond
+  %lis = call {}* @llvm.invariant.start(i64 4096, i8* %tmp3) #1
+  br label %for.cond2
+
+for.cond2:                                        ; preds = %for.inc, %for.body
+  %indvars.iv1 = phi i64 [ %indvars.iv.next2, %for.inc ], [ %indvars.iv3, %for.body ]
+  %lftr.wideiv = trunc i64 %indvars.iv1 to i32
+  %exitcond = icmp ne i32 %lftr.wideiv, 1024
+  br i1 %exitcond, label %for.body4, label %for.end
+
+for.body4:                                        ; preds = %for.cond2
+  call void @llvm.assume(i1 %exitcond)
+  call i1 @llvm.expect.i1(i1 %exitcond, i1 1)
+  %arrayidx = getelementptr inbounds [1024 x i32]* @A, i64 0, i64 %indvars.iv1
+  %tmp6 = load i32* %arrayidx, align 4
+  %arrayidx6 = getelementptr inbounds [1024 x i32]* %tmp, i64 0, i64 %indvars.iv3
+  call void @llvm.donothing()
+  %tmp7 = load i32* %arrayidx6, align 4
+  %add = add nsw i32 %tmp7, %tmp6
+  store i32 %add, i32* %arrayidx6, align 4
+  br label %for.inc
+
+for.inc:                                          ; preds = %for.body4
+  %indvars.iv.next2 = add nuw nsw i64 %indvars.iv1, 1
+  br label %for.cond2
+
+for.end:                                          ; preds = %for.cond2
+  %arrayidx8 = getelementptr inbounds [1024 x i32]* %tmp, i64 0, i64 %indvars.iv3
+  %tmp8 = load i32* %arrayidx8, align 4
+  %arrayidx10 = getelementptr inbounds [1024 x i32]* @A, i64 0, i64 %indvars.iv3
+  call void @llvm.invariant.end({}* %lis, i64 4096, i8* %tmp3) #1
+  store i32 %tmp8, i32* %arrayidx10, align 4
+  br label %for.inc11
+
+for.inc11:                                        ; preds = %for.end
+  %indvars.iv.next4 = add nuw nsw i64 %indvars.iv3, 1
+  br label %for.cond
+
+for.end13:                                        ; preds = %for.cond
+  ret void
+}
+
+; Function Attrs: nounwind
+declare void @llvm.donothing() #1
+
+; Function Attrs: nounwind
+declare void @llvm.assume(i1) #1
+
+; Function Attrs: nounwind
+declare i1 @llvm.expect.i1(i1, i1) #1
+
+; Function Attrs: nounwind
+declare {}* @llvm.invariant.start(i64, i8* nocapture) #1
+
+; Function Attrs: nounwind
+declare void @llvm.invariant.end({}*, i64, i8* nocapture) #1
+
+attributes #0 = { nounwind uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="true" "no-nans-fp-math"="true" "stack-protector-buffer-size"="8" "unsafe-fp-math"="true" "use-soft-float"="false" }
+attributes #1 = { nounwind }
Index: test/ScopDetect/intrinsics_1.ll
===================================================================
--- /dev/null
+++ test/ScopDetect/intrinsics_1.ll
@@ -0,0 +1,106 @@
+; RUN: opt %loadPolly -basicaa -polly-detect -analyze < %s | FileCheck %s
+;
+; CHECK: Valid Region for Scop: for.cond => for.end
+;
+;    #include "math.h"
+;
+;    void jd(int *restrict A, float *restrict B) {
+;      for (int i = 0; i < 1024; i++) {
+;        A[i] = pow(ceil(log10(sqrt(i))), floor(log2(i)));
+;        B[i] = fabs(log(sin(i)) + exp2(cos(i))) + exp(i);
+;      }
+;    }
+;
+; ModuleID = '/home/johannes/repos/polly/test/ScopDetect/intrinsics.ll'
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+
+; Function Attrs: nounwind uwtable
+define void @jd(i32* noalias %A, float* noalias %B) #0 {
+entry:
+  br label %for.cond
+
+for.cond:                                         ; preds = %for.inc, %entry
+  %indvars.iv = phi i64 [ %indvars.iv.next, %for.inc ], [ 0, %entry ]
+  %exitcond = icmp ne i64 %indvars.iv, 1024
+  br i1 %exitcond, label %for.body, label %for.end
+
+for.body:                                         ; preds = %for.cond
+  %tmp = trunc i64 %indvars.iv to i32
+  %conv = sitofp i32 %tmp to double
+  %tmp1 = call double @llvm.sqrt.f64(double %conv)
+  %call = call double @__log10_finite(double %tmp1) #2
+  %call1 = call double @ceil(double %call) #2
+  %tmp2 = trunc i64 %indvars.iv to i32
+  %conv2 = sitofp i32 %tmp2 to double
+  %call3 = call double @__log2_finite(double %conv2) #2
+  %call4 = call double @floor(double %call3) #2
+  %tmp3 = call double @llvm.pow.f64(double %call1, double %call4)
+  %conv5 = fptosi double %tmp3 to i32
+  %arrayidx = getelementptr inbounds i32* %A, i64 %indvars.iv
+  store i32 %conv5, i32* %arrayidx, align 4
+  %tmp4 = trunc i64 %indvars.iv to i32
+  %conv6 = sitofp i32 %tmp4 to double
+  %call7 = call double @sin(double %conv6) #2
+  %call8 = call double @__log_finite(double %call7) #2
+  %tmp5 = trunc i64 %indvars.iv to i32
+  %conv9 = sitofp i32 %tmp5 to double
+  %call10 = call double @cos(double %conv9) #2
+  %call11 = call double @__exp2_finite(double %call10) #2
+  %add = fadd fast double %call8, %call11
+  %call12 = call double @fabs(double %add) #2
+  %tmp6 = trunc i64 %indvars.iv to i32
+  %conv13 = sitofp i32 %tmp6 to double
+  %call14 = call double @__exp_finite(double %conv13) #2
+  %add15 = fadd fast double %call12, %call14
+  %conv16 = fptrunc double %add15 to float
+  %arrayidx18 = getelementptr inbounds float* %B, i64 %indvars.iv
+  store float %conv16, float* %arrayidx18, align 4
+  br label %for.inc
+
+for.inc:                                          ; preds = %for.body
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  br label %for.cond
+
+for.end:                                          ; preds = %for.cond
+  ret void
+}
+
+; Function Attrs: nounwind readnone
+declare double @ceil(double) #1
+
+; Function Attrs: nounwind readnone
+declare double @__log10_finite(double) #1
+
+; Function Attrs: nounwind readnone
+declare double @llvm.sqrt.f64(double) #2
+
+; Function Attrs: nounwind readnone
+declare double @floor(double) #1
+
+; Function Attrs: nounwind readnone
+declare double @__log2_finite(double) #1
+
+; Function Attrs: nounwind readnone
+declare double @llvm.pow.f64(double, double) #2
+
+; Function Attrs: nounwind readnone
+declare double @fabs(double) #1
+
+; Function Attrs: nounwind readnone
+declare double @__log_finite(double) #1
+
+; Function Attrs: nounwind readnone
+declare double @sin(double) #1
+
+; Function Attrs: nounwind readnone
+declare double @__exp2_finite(double) #1
+
+; Function Attrs: nounwind readnone
+declare double @cos(double) #1
+
+; Function Attrs: nounwind readnone
+declare double @__exp_finite(double) #1
+
+attributes #0 = { nounwind uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="true" "no-nans-fp-math"="true" "stack-protector-buffer-size"="8" "unsafe-fp-math"="true" "use-soft-float"="false" }
+attributes #1 = { nounwind readnone "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="true" "no-nans-fp-math"="true" "stack-protector-buffer-size"="8" "unsafe-fp-math"="true" "use-soft-float"="false" }
+attributes #2 = { nounwind readnone }
Index: test/ScopDetect/intrinsics_2.ll
===================================================================
--- /dev/null
+++ test/ScopDetect/intrinsics_2.ll
@@ -0,0 +1,80 @@
+; RUN: opt %loadPolly -basicaa -polly-detect -polly-codegen-scev -analyze < %s | FileCheck %s
+;
+; Verify that we allow the lifetime markers for the tmp array.
+;
+; CHECK: Valid Region for Scop: for.cond => for.end13
+;
+;    int A[1024];
+;    void jd() {
+;      for (int i = 0; i < 1024; i++) {
+;        int tmp[1024];
+;        for (int j = i; j < 1024; j++)
+;          tmp[i] += A[j];
+;        A[i] = tmp[i];
+;      }
+;    }
+;
+; ModuleID = 'test/Isl/CodeGen/lifetime_intrinsics.ll'
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+
+@A = common global [1024 x i32] zeroinitializer, align 16
+
+; Function Attrs: nounwind uwtable
+define void @jd() #0 {
+entry:
+  %tmp = alloca [1024 x i32], align 16
+  %tmp3 = bitcast [1024 x i32]* %tmp to i8*
+  br label %for.cond
+
+for.cond:                                         ; preds = %for.inc11, %entry
+  %indvars.iv3 = phi i64 [ %indvars.iv.next4, %for.inc11 ], [ 0, %entry ]
+  %exitcond5 = icmp ne i64 %indvars.iv3, 1024
+  br i1 %exitcond5, label %for.body, label %for.end13
+
+for.body:                                         ; preds = %for.cond
+  call void @llvm.lifetime.start(i64 4096, i8* %tmp3) #1
+  br label %for.cond2
+
+for.cond2:                                        ; preds = %for.inc, %for.body
+  %indvars.iv1 = phi i64 [ %indvars.iv.next2, %for.inc ], [ %indvars.iv3, %for.body ]
+  %lftr.wideiv = trunc i64 %indvars.iv1 to i32
+  %exitcond = icmp ne i32 %lftr.wideiv, 1024
+  br i1 %exitcond, label %for.body4, label %for.end
+
+for.body4:                                        ; preds = %for.cond2
+  %arrayidx = getelementptr inbounds [1024 x i32]* @A, i64 0, i64 %indvars.iv1
+  %tmp6 = load i32* %arrayidx, align 4
+  %arrayidx6 = getelementptr inbounds [1024 x i32]* %tmp, i64 0, i64 %indvars.iv3
+  %tmp7 = load i32* %arrayidx6, align 4
+  %add = add nsw i32 %tmp7, %tmp6
+  store i32 %add, i32* %arrayidx6, align 4
+  br label %for.inc
+
+for.inc:                                          ; preds = %for.body4
+  %indvars.iv.next2 = add nuw nsw i64 %indvars.iv1, 1
+  br label %for.cond2
+
+for.end:                                          ; preds = %for.cond2
+  %arrayidx8 = getelementptr inbounds [1024 x i32]* %tmp, i64 0, i64 %indvars.iv3
+  %tmp8 = load i32* %arrayidx8, align 4
+  %arrayidx10 = getelementptr inbounds [1024 x i32]* @A, i64 0, i64 %indvars.iv3
+  store i32 %tmp8, i32* %arrayidx10, align 4
+  br label %for.inc11
+
+for.inc11:                                        ; preds = %for.end
+  %indvars.iv.next4 = add nuw nsw i64 %indvars.iv3, 1
+  call void @llvm.lifetime.end(i64 4096, i8* %tmp3) #1
+  br label %for.cond
+
+for.end13:                                        ; preds = %for.cond
+  ret void
+}
+
+; Function Attrs: nounwind
+declare void @llvm.lifetime.start(i64, i8* nocapture) #1
+
+; Function Attrs: nounwind
+declare void @llvm.lifetime.end(i64, i8* nocapture) #1
+
+attributes #0 = { nounwind uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="true" "no-nans-fp-math"="true" "stack-protector-buffer-size"="8" "unsafe-fp-math"="true" "use-soft-float"="false" }
+attributes #1 = { nounwind }
Index: test/ScopDetect/intrinsics_3.ll
===================================================================
--- /dev/null
+++ test/ScopDetect/intrinsics_3.ll
@@ -0,0 +1,93 @@
+; RUN: opt %loadPolly -basicaa -polly-detect -polly-codegen-scev -analyze < %s | FileCheck %s
+;
+; Verify that we allow the misc intrinsics.
+;
+; CHECK: Valid Region for Scop: for.cond => for.end13
+;
+;    int A[1024];
+;    void jd() {
+;      for (int i = 0; i < 1024; i++) {
+;        int tmp[1024];
+;        for (int j = i; j < 1024; j++)
+;          tmp[i] += A[j];
+;        A[i] = tmp[i];
+;      }
+;    }
+;
+; ModuleID = 'test/Isl/CodeGen/lifetime_intrinsics.ll'
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+@A = common global [1024 x i32] zeroinitializer, align 16
+
+; Function Attrs: nounwind uwtable
+define void @jd() #0 {
+entry:
+  %tmp = alloca [1024 x i32], align 16
+  %tmp3 = bitcast [1024 x i32]* @A to i8*
+  br label %for.cond
+
+for.cond:                                         ; preds = %for.inc11, %entry
+  %indvars.iv3 = phi i64 [ %indvars.iv.next4, %for.inc11 ], [ 0, %entry ]
+  %exitcond5 = icmp ne i64 %indvars.iv3, 1024
+  br i1 %exitcond5, label %for.body, label %for.end13
+
+for.body:                                         ; preds = %for.cond
+  %lis = call {}* @llvm.invariant.start(i64 4096, i8* %tmp3) #1
+  br label %for.cond2
+
+for.cond2:                                        ; preds = %for.inc, %for.body
+  %indvars.iv1 = phi i64 [ %indvars.iv.next2, %for.inc ], [ %indvars.iv3, %for.body ]
+  %lftr.wideiv = trunc i64 %indvars.iv1 to i32
+  %exitcond = icmp ne i32 %lftr.wideiv, 1024
+  br i1 %exitcond, label %for.body4, label %for.end
+
+for.body4:                                        ; preds = %for.cond2
+  call void @llvm.assume(i1 %exitcond)
+  call i1 @llvm.expect.i1(i1 %exitcond, i1 1)
+  %arrayidx = getelementptr inbounds [1024 x i32]* @A, i64 0, i64 %indvars.iv1
+  %tmp6 = load i32* %arrayidx, align 4
+  %arrayidx6 = getelementptr inbounds [1024 x i32]* %tmp, i64 0, i64 %indvars.iv3
+  call void @llvm.donothing()
+  %tmp7 = load i32* %arrayidx6, align 4
+  %add = add nsw i32 %tmp7, %tmp6
+  store i32 %add, i32* %arrayidx6, align 4
+  br label %for.inc
+
+for.inc:                                          ; preds = %for.body4
+  %indvars.iv.next2 = add nuw nsw i64 %indvars.iv1, 1
+  br label %for.cond2
+
+for.end:                                          ; preds = %for.cond2
+  %arrayidx8 = getelementptr inbounds [1024 x i32]* %tmp, i64 0, i64 %indvars.iv3
+  %tmp8 = load i32* %arrayidx8, align 4
+  %arrayidx10 = getelementptr inbounds [1024 x i32]* @A, i64 0, i64 %indvars.iv3
+  call void @llvm.invariant.end({}* %lis, i64 4096, i8* %tmp3) #1
+  store i32 %tmp8, i32* %arrayidx10, align 4
+  br label %for.inc11
+
+for.inc11:                                        ; preds = %for.end
+  %indvars.iv.next4 = add nuw nsw i64 %indvars.iv3, 1
+  br label %for.cond
+
+for.end13:                                        ; preds = %for.cond
+  ret void
+}
+
+; Function Attrs: nounwind
+declare void @llvm.donothing() #1
+
+; Function Attrs: nounwind
+declare void @llvm.assume(i1) #1
+
+; Function Attrs: nounwind
+declare i1 @llvm.expect.i1(i1, i1) #1
+
+; Function Attrs: nounwind
+declare {}* @llvm.invariant.start(i64, i8* nocapture) #1
+
+; Function Attrs: nounwind
+declare void @llvm.invariant.end({}*, i64, i8* nocapture) #1
+
+attributes #0 = { nounwind uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="true" "no-nans-fp-math"="true" "stack-protector-buffer-size"="8" "unsafe-fp-math"="true" "use-soft-float"="false" }
+attributes #1 = { nounwind }