vendor/llvm/llvm-r142614

11 files changed, 301 insertions, 13 deletions

diff --git a/test/Analysis/BasicAA/2009-10-13-AtomicModRef.ll b/test/Analysis/BasicAA/2009-10-13-AtomicModRef.ll
index 5078dd53a79f..4b6a12e821e1 100644
--- a/test/Analysis/BasicAA/2009-10-13-AtomicModRef.ll
+++ b/test/Analysis/BasicAA/2009-10-13-AtomicModRef.ll
@@ -1,14 +1,12 @@
 ; RUN: opt -basicaa -gvn -instcombine -S < %s | FileCheck %s
 target datalayout = "E-p:64:64:64-a0:0:8-f32:32:32-f64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-v64:64:64-v128:128:128"
 
-declare i8 @llvm.atomic.load.add.i8.p0i8(i8*, i8)
-
 define i8 @foo(i8* %ptr) {
   %P = getelementptr i8* %ptr, i32 0
   %Q = getelementptr i8* %ptr, i32 1
 ; CHECK: getelementptr
   %X = load i8* %P
-  %Y = call i8 @llvm.atomic.load.add.i8.p0i8(i8* %Q, i8 1)
+  %Y = atomicrmw add i8* %Q, i8 1 monotonic
   %Z = load i8* %P
 ; CHECK-NOT: = load
   %A = sub i8 %X, %Z
diff --git a/test/Analysis/BasicAA/cas.ll b/test/Analysis/BasicAA/cas.ll
index 8dd3695d6d57..754309cd81bb 100644
--- a/test/Analysis/BasicAA/cas.ll
+++ b/test/Analysis/BasicAA/cas.ll
@@ -6,10 +6,8 @@
 
 define i32 @main() {
   %a = load i32* @flag0
-  %b = tail call i32 @llvm.atomic.swap.i32.p0i32(i32* @turn, i32 1)
+  %b = atomicrmw xchg i32* @turn, i32 1 monotonic
   %c = load i32* @flag0
   %d = sub i32 %a, %c
   ret i32 %d
 }
-
-declare i32 @llvm.atomic.swap.i32.p0i32(i32*, i32) nounwind
diff --git a/test/Analysis/BasicAA/gep-alias.ll b/test/Analysis/BasicAA/gep-alias.ll
index 69f7fafaca07..4bb28326612c 100644
--- a/test/Analysis/BasicAA/gep-alias.ll
+++ b/test/Analysis/BasicAA/gep-alias.ll
@@ -169,3 +169,35 @@ define i8 @test10([4 x i8] *%P, i32 %i) {
 ; CHECK: @test10
 ; CHECK: ret i8 0
 }
+
+; (This was a miscompilation.)
+define float @test11(i32 %indvar, [4 x [2 x float]]* %q) nounwind ssp {
+  %tmp = mul i32 %indvar, -1
+  %dec = add i32 %tmp, 3
+  %scevgep = getelementptr [4 x [2 x float]]* %q, i32 0, i32 %dec
+  %scevgep35 = bitcast [2 x float]* %scevgep to i64*
+  %arrayidx28 = getelementptr inbounds [4 x [2 x float]]* %q, i32 0, i32 0
+  %y29 = getelementptr inbounds [2 x float]* %arrayidx28, i32 0, i32 1
+  store float 1.0, float* %y29, align 4
+  store i64 0, i64* %scevgep35, align 4
+  %tmp30 = load float* %y29, align 4
+  ret float %tmp30
+  ; CHECK: @test11
+  ; CHECK: ret float %tmp30
+}
+
+; (This was a miscompilation.)
+define i32 @test12(i32 %x, i32 %y, i8* %p) nounwind {
+  %a = bitcast i8* %p to [13 x i8]*
+  %b = getelementptr [13 x i8]* %a, i32 %x
+  %c = bitcast [13 x i8]* %b to [15 x i8]*
+  %d = getelementptr [15 x i8]* %c, i32 %y, i32 8
+  %castd = bitcast i8* %d to i32*
+  %castp = bitcast i8* %p to i32*
+  store i32 1, i32* %castp
+  store i32 0, i32* %castd
+  %r = load i32* %castp
+  ret i32 %r
+  ; CHECK: @test12
+  ; CHECK: ret i32 %r
+}
diff --git a/test/Analysis/BasicAA/memset_pattern.ll b/test/Analysis/BasicAA/memset_pattern.ll
new file mode 100644
index 000000000000..590664c5084e
--- /dev/null
+++ b/test/Analysis/BasicAA/memset_pattern.ll
@@ -0,0 +1,21 @@
+; RUN: opt -S -basicaa -gvn < %s | FileCheck %s
+; PR10872
+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"
+target triple = "x86_64-apple-macosx10.7"
+
+@z = internal global i32 0, align 4
+@y = internal global i32 0, align 4
+@x = internal constant i32 0, align 4
+
+; CHECK: @test
+define i32 @test() nounwind uwtable ssp {
+entry:
+  store i32 1, i32* @z
+  tail call void @memset_pattern16(i8* bitcast (i32* @y to i8*), i8* bitcast (i32* @x to i8*), i64 4) nounwind
+; CHECK-NOT: load
+  %l = load i32* @z
+; CHECK: ret i32 1
+  ret i32 %l
+}
+
+declare void @memset_pattern16(i8*, i8*, i64)
diff --git a/test/Analysis/BasicAA/modref.ll b/test/Analysis/BasicAA/modref.ll
index 233396b7e5ca..8421faf9c2c4 100644
--- a/test/Analysis/BasicAA/modref.ll
+++ b/test/Analysis/BasicAA/modref.ll
@@ -129,8 +129,23 @@ define i8 @test6(i8* %p, i8* noalias %a) {
 ; CHECK: ret
 }
 
+; PR10628
+declare void @test7decl(i32* nocapture %x)
+define i32 @test7() nounwind uwtable ssp {
+entry:
+  %x = alloca i32, align 4
+  store i32 0, i32* %x, align 4
+  %add.ptr = getelementptr inbounds i32* %x, i64 1
+  call void @test7decl(i32* %add.ptr)
+  %tmp = load i32* %x, align 4
+  ret i32 %tmp
+; CHECK: @test7(
+; CHECK: store i32 0
+; CHECK: call void @test7decl
+; CHECK: load i32*
+}
+
 declare void @llvm.memset.p0i8.i32(i8* nocapture, i8, i32, i32, i1) nounwind
 declare void @llvm.memset.p0i8.i8(i8* nocapture, i8, i8, i32, i1) nounwind
 declare void @llvm.memcpy.p0i8.p0i8.i8(i8* nocapture, i8* nocapture, i8, i32, i1) nounwind
 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* nocapture, i8* nocapture, i32, i32, i1) nounwind
-
diff --git a/test/Analysis/BlockFrequencyInfo/basic.ll b/test/Analysis/BlockFrequencyInfo/basic.ll
new file mode 100644
index 000000000000..540d06b1f562
--- /dev/null
+++ b/test/Analysis/BlockFrequencyInfo/basic.ll
@@ -0,0 +1,92 @@
+; RUN: opt < %s -analyze -block-freq | FileCheck %s
+
+define i32 @test1(i32 %i, i32* %a) {
+; CHECK: Printing analysis {{.*}} for function 'test1'
+; CHECK: entry = 1024
+entry:
+  br label %body
+
+; Loop backedges are weighted and thus their bodies have a greater frequency.
+; CHECK: body = 31744
+body:
+  %iv = phi i32 [ 0, %entry ], [ %next, %body ]
+  %base = phi i32 [ 0, %entry ], [ %sum, %body ]
+  %arrayidx = getelementptr inbounds i32* %a, i32 %iv
+  %0 = load i32* %arrayidx
+  %sum = add nsw i32 %0, %base
+  %next = add i32 %iv, 1
+  %exitcond = icmp eq i32 %next, %i
+  br i1 %exitcond, label %exit, label %body
+
+; CHECK: exit = 1024
+exit:
+  ret i32 %sum
+}
+
+define i32 @test2(i32 %i, i32 %a, i32 %b) {
+; CHECK: Printing analysis {{.*}} for function 'test2'
+; CHECK: entry = 1024
+entry:
+  %cond = icmp ult i32 %i, 42
+  br i1 %cond, label %then, label %else, !prof !0
+
+; The 'then' branch is predicted more likely via branch weight metadata.
+; CHECK: then = 963
+then:
+  br label %exit
+
+; CHECK: else = 60
+else:
+  br label %exit
+
+; FIXME: It may be a bug that we don't sum back to 1024.
+; CHECK: exit = 1023
+exit:
+  %result = phi i32 [ %a, %then ], [ %b, %else ]
+  ret i32 %result
+}
+
+!0 = metadata !{metadata !"branch_weights", i32 64, i32 4}
+
+define i32 @test3(i32 %i, i32 %a, i32 %b, i32 %c, i32 %d, i32 %e) {
+; CHECK: Printing analysis {{.*}} for function 'test3'
+; CHECK: entry = 1024
+entry:
+  switch i32 %i, label %case_a [ i32 1, label %case_b
+                                 i32 2, label %case_c
+                                 i32 3, label %case_d
+                                 i32 4, label %case_e ], !prof !1
+
+; CHECK: case_a = 51
+case_a:
+  br label %exit
+
+; CHECK: case_b = 51
+case_b:
+  br label %exit
+
+; The 'case_c' branch is predicted more likely via branch weight metadata.
+; CHECK: case_c = 819
+case_c:
+  br label %exit
+
+; CHECK: case_d = 51
+case_d:
+  br label %exit
+
+; CHECK: case_e = 51
+case_e:
+  br label %exit
+
+; FIXME: It may be a bug that we don't sum back to 1024.
+; CHECK: exit = 1023
+exit:
+  %result = phi i32 [ %a, %case_a ],
+                    [ %b, %case_b ],
+                    [ %c, %case_c ],
+                    [ %d, %case_d ],
+                    [ %e, %case_e ]
+  ret i32 %result
+}
+
+!1 = metadata !{metadata !"branch_weights", i32 4, i32 4, i32 64, i32 4, i32 4}
diff --git a/test/Analysis/BlockFrequencyInfo/dg.exp b/test/Analysis/BlockFrequencyInfo/dg.exp
new file mode 100644
index 000000000000..f2005891a59a
--- /dev/null
+++ b/test/Analysis/BlockFrequencyInfo/dg.exp
@@ -0,0 +1,3 @@
+load_lib llvm.exp
+
+RunLLVMTests [lsort [glob -nocomplain $srcdir/$subdir/*.{ll,c,cpp}]]
diff --git a/test/Analysis/ScalarEvolution/2011-10-04-ConstEvolve.ll b/test/Analysis/ScalarEvolution/2011-10-04-ConstEvolve.ll
new file mode 100644
index 000000000000..29bb64ad642b
--- /dev/null
+++ b/test/Analysis/ScalarEvolution/2011-10-04-ConstEvolve.ll
@@ -0,0 +1,50 @@
+; RUN: opt < %s -analyze -scalar-evolution | FileCheck %s
+
+; Exercise getConstantEvolvingPHIOperands on an interesting loop.
+; This should complete in milliseconds, not minutes.
+
+; Just check that it actually ran trip count analysis.
+; CHECK: Determining loop execution counts for: @test
+define void @test() nounwind {
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i32 [ %30, %loop ], [ 0, %entry ]
+  %0 = add i32 %iv, 1
+  %1 = add i32 %0, 2
+  %2 = add i32 %1, %0
+  %3 = add i32 %2, %1
+  %4 = add i32 %3, %2
+  %5 = add i32 %4, %3
+  %6 = add i32 %5, %4
+  %7 = add i32 %6, %5
+  %8 = add i32 %7, %6
+  %9 = add i32 %8, %7
+  %10 = add i32 %9, %8
+  %11 = add i32 %10, %9
+  %12 = add i32 %11, %10
+  %13 = add i32 %12, %11
+  %14 = add i32 %13, %12
+  %15 = add i32 %14, %13
+  %16 = add i32 %15, %14
+  %17 = add i32 %16, %15
+  %18 = add i32 %17, %16
+  %19 = add i32 %18, %17
+  %20 = add i32 %19, %18
+  %21 = add i32 %20, %19
+  %22 = add i32 %21, %20
+  %23 = add i32 %22, %21
+  %24 = add i32 %23, %22
+  %25 = add i32 %24, %23
+  %26 = add i32 %25, %24
+  %27 = add i32 %26, %25
+  %28 = add i32 %27, %26
+  %29 = add i32 %28, %27
+  %30 = add i32 %29, %28
+  %cmp = icmp eq i32 %30, -108
+  br i1 %cmp, label %exit, label %loop
+
+exit:
+  unreachable
+}
diff --git a/test/Analysis/ScalarEvolution/SolveQuadraticEquation.ll b/test/Analysis/ScalarEvolution/SolveQuadraticEquation.ll
index ec95141fbe92..06f1b6fe33a6 100644
--- a/test/Analysis/ScalarEvolution/SolveQuadraticEquation.ll
+++ b/test/Analysis/ScalarEvolution/SolveQuadraticEquation.ll
@@ -1,11 +1,11 @@
 ; RUN: opt < %s -analyze -scalar-evolution \
-; RUN:   -scalar-evolution-max-iterations=0 | grep {backedge-taken count is 100}
+; RUN:   -scalar-evolution-max-iterations=0 | FileCheck %s
+
 ; PR1101
 
 @A = weak global [1000 x i32] zeroinitializer, align 32         
 
-
-define void @test(i32 %N) {
+define void @test1(i32 %N) {
 entry:
         %"alloca point" = bitcast i32 0 to i32           ; <i32> [#uses=0]
         br label %bb3
@@ -30,3 +30,53 @@ bb5:            ; preds = %bb3
 return:         ; preds = %bb5
         ret void
 }
+; CHECK: Determining loop execution counts for: @test1
+; CHECK-NEXT: backedge-taken count is 100
+
+
+; PR10383
+; These next two used to crash.
+
+define void @test2(i1 %cmp, i64 %n) {
+entry:
+  br label %for.body1
+
+for.body1:
+  %a0.08 = phi i64 [ 0, %entry ], [ %inc512, %for.body1 ]
+  %inc512 = add i64 %a0.08, 1
+  br i1 %cmp, label %preheader, label %for.body1
+
+preheader:
+  br label %for.body2
+
+for.body2:
+  %indvar = phi i64 [ 0, %preheader ], [ %indvar.next, %for.body2 ]
+  %tmp111 = add i64 %n, %indvar
+  %tmp114 = mul i64 %a0.08, %indvar
+  %mul542 = mul i64 %tmp114, %tmp111
+  %indvar.next = add i64 %indvar, 1
+  br i1 undef, label %end, label %for.body2
+
+end:
+  ret void
+}
+; CHECK: Determining loop execution counts for: @test2
+
+define i32 @test3() {
+if.then466:
+  br i1 undef, label %for.cond539.preheader, label %for.inc479
+
+for.inc479:
+  %a2.07 = phi i32 [ %add495, %for.inc479 ], [ 0, %if.then466 ]
+  %j.36 = phi i32 [ %inc497, %for.inc479 ], [ undef, %if.then466 ]
+  %mul484 = mul nsw i32 %j.36, %j.36
+  %mul491 = mul i32 %j.36, %j.36
+  %mul493 = mul i32 %mul491, %mul484
+  %add495 = add nsw i32 %mul493, %a2.07
+  %inc497 = add nsw i32 %j.36, 1
+  br i1 undef, label %for.cond539.preheader, label %for.inc479
+
+for.cond539.preheader:
+  unreachable
+}
+; CHECK: Determining loop execution counts for: @test3
diff --git a/test/Analysis/ScalarEvolution/avoid-smax-1.ll b/test/Analysis/ScalarEvolution/avoid-smax-1.ll
index 474d564c70c6..e90a55559d7f 100644
--- a/test/Analysis/ScalarEvolution/avoid-smax-1.ll
+++ b/test/Analysis/ScalarEvolution/avoid-smax-1.ll
@@ -1,6 +1,4 @@
-; RUN: opt < %s -indvars -S > %t
-; RUN: grep select %t | count 2
-; RUN: grep {icmp ne i32.\* } %t
+; RUN: opt < %s -indvars -S -enable-iv-rewrite | FileCheck %s
 
 ; Indvars should be able to insert a canonical induction variable
 ; for the bb6 loop without using a maximum calculation (icmp, select)
@@ -8,6 +6,9 @@
 ; by an appropriate conditional branch. Unfortunately, indvars is
 ; not yet able to find the comparison for the other two loops in
 ; this testcase.
+; CHECK: entry:
+; CHECK-NOT: select
+; CHECK: bb6:
 
 target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
 target triple = "i386-apple-darwin9"
diff --git a/test/Analysis/ScalarEvolution/max-trip-count.ll b/test/Analysis/ScalarEvolution/max-trip-count.ll
index 843fb073087c..0cdbdf57a64c 100644
--- a/test/Analysis/ScalarEvolution/max-trip-count.ll
+++ b/test/Analysis/ScalarEvolution/max-trip-count.ll
@@ -70,3 +70,31 @@ for.end:                                          ; preds = %for.body, %for.cond
 }
 
 declare i32 @printf(i8*, ...)
+
+define void @test(i8* %a, i32 %n) nounwind {
+entry:
+  %cmp1 = icmp sgt i32 %n, 0
+  br i1 %cmp1, label %for.body.lr.ph, label %for.end
+
+for.body.lr.ph:                                   ; preds = %entry
+  %tmp = zext i32 %n to i64
+  br label %for.body
+
+for.body:                                         ; preds = %for.body, %for.body.lr.ph
+  %indvar = phi i64 [ %indvar.next, %for.body ], [ 0, %for.body.lr.ph ]
+  %arrayidx = getelementptr i8* %a, i64 %indvar
+  store i8 0, i8* %arrayidx, align 1
+  %indvar.next = add i64 %indvar, 1
+  %exitcond = icmp ne i64 %indvar.next, %tmp
+  br i1 %exitcond, label %for.body, label %for.cond.for.end_crit_edge
+
+for.cond.for.end_crit_edge:                       ; preds = %for.body
+  br label %for.end
+
+for.end:                                          ; preds = %for.cond.for.end_crit_edge, %entry
+  ret void
+}
+
+; CHECK: Determining loop execution counts for: @test
+; CHECK-NEXT: backedge-taken count is
+; CHECK-NEXT: max backedge-taken count is -1