diff options
author | Shih-wei Liao <sliao@google.com> | 2010-02-10 11:10:31 -0800 |
---|---|---|
committer | Shih-wei Liao <sliao@google.com> | 2010-02-10 11:10:31 -0800 |
commit | e264f62ca09a8f65c87a46d562a4d0f9ec5d457e (patch) | |
tree | 59e3d57ef656cef79afa708ae0a3daf25cd91fcf /test/Analysis | |
download | external_llvm-e264f62ca09a8f65c87a46d562a4d0f9ec5d457e.zip external_llvm-e264f62ca09a8f65c87a46d562a4d0f9ec5d457e.tar.gz external_llvm-e264f62ca09a8f65c87a46d562a4d0f9ec5d457e.tar.bz2 |
Check in LLVM r95781.
Diffstat (limited to 'test/Analysis')
155 files changed, 7000 insertions, 0 deletions
diff --git a/test/Analysis/Andersens/2007-11-19-InlineAsm.ll b/test/Analysis/Andersens/2007-11-19-InlineAsm.ll new file mode 100644 index 0000000..5ba3499 --- /dev/null +++ b/test/Analysis/Andersens/2007-11-19-InlineAsm.ll @@ -0,0 +1,8 @@ +; RUN: opt < %s -anders-aa -disable-output + +define void @x(i16 %Y) { +entry: + %tmp = call i16 asm "bswap $0", "=r,r"(i16 %Y) + ret void +} + diff --git a/test/Analysis/Andersens/2008-03-19-External.ll b/test/Analysis/Andersens/2008-03-19-External.ll new file mode 100644 index 0000000..a973103 --- /dev/null +++ b/test/Analysis/Andersens/2008-03-19-External.ll @@ -0,0 +1,12 @@ +; RUN: opt < %s -anders-aa -gvn -S | not grep undef +; PR2160 + +declare void @f(i32*) + +define i32 @g() { +entry: + %tmp = alloca i32 ; <i32*> [#uses=2] + call void @f( i32* %tmp ) + %tmp2 = load i32* %tmp ; <i32> [#uses=1] + ret i32 %tmp2 +} diff --git a/test/Analysis/Andersens/2008-04-07-Memcpy.ll b/test/Analysis/Andersens/2008-04-07-Memcpy.ll new file mode 100644 index 0000000..5a50dd5 --- /dev/null +++ b/test/Analysis/Andersens/2008-04-07-Memcpy.ll @@ -0,0 +1,14 @@ +; RUN: opt < %s -anders-aa -gvn -S | not grep undef +; PR2169 + +declare void @llvm.memcpy.i32(i8*, i8*, i32, i32) nounwind +declare void @use(i8) + +define void @f(i8* %x) { +entry: + %copy = alloca i8 ; <i8*> [#uses=6] + call void @llvm.memcpy.i32( i8* %copy, i8* %x, i32 1, i32 4 ) + %tmp = load i8* %copy ; <i8> [#uses=1] + call void @use(i8 %tmp) + ret void +} diff --git a/test/Analysis/Andersens/2008-12-27-BuiltinWrongType.ll b/test/Analysis/Andersens/2008-12-27-BuiltinWrongType.ll new file mode 100644 index 0000000..da67511 --- /dev/null +++ b/test/Analysis/Andersens/2008-12-27-BuiltinWrongType.ll @@ -0,0 +1,19 @@ +; RUN: opt < %s -anders-aa +; PR3262 + +@.str15 = external global [3 x i8] ; <[3 x i8]*> [#uses=1] + +declare i8* @strtok(...) +declare i8* @memmove(...) + +define void @test1(i8* %want1) nounwind { +entry: + %0 = call i8* (...)* @strtok(i32 0, i8* getelementptr ([3 x i8]* @.str15, i32 0, i32 0)) nounwind ; <i8*> [#uses=0] + unreachable +} + +define void @test2() nounwind { +entry: + %0 = call i8* (...)* @memmove() + unreachable +} diff --git a/test/Analysis/Andersens/basictest.ll b/test/Analysis/Andersens/basictest.ll new file mode 100644 index 0000000..47226dd --- /dev/null +++ b/test/Analysis/Andersens/basictest.ll @@ -0,0 +1,28 @@ +; RUN: opt < %s -anders-aa -aa-eval 2>/dev/null + +define void @test1() { + %X = malloc i32* + %Y = malloc i32 + %Z = ptrtoint i32* %Y to i32 + %W = inttoptr i32 %Z to i32* + store i32* %W, i32** %X + ret void +} + +define void @test2(i32* %P) { + %X = malloc i32* + %Y = malloc i32 + store i32* %P, i32** %X + ret void +} + +define internal i32 *@test3(i32* %P) { + ret i32* %P +} + +define void @test4() { + %X = malloc i32 + %Y = call i32* @test3(i32* %X) + %ZZ = getelementptr i32* null, i32 17 + ret void +} diff --git a/test/Analysis/Andersens/dg.exp b/test/Analysis/Andersens/dg.exp new file mode 100644 index 0000000..1eb4755 --- /dev/null +++ b/test/Analysis/Andersens/dg.exp @@ -0,0 +1,4 @@ +load_lib llvm.exp + +RunLLVMTests [lsort [glob -nocomplain $srcdir/$subdir/*.{ll,c,cpp}]] + diff --git a/test/Analysis/Andersens/external.ll b/test/Analysis/Andersens/external.ll new file mode 100644 index 0000000..13c12dc --- /dev/null +++ b/test/Analysis/Andersens/external.ll @@ -0,0 +1,20 @@ +; RUN: opt < %s -anders-aa -gvn -deadargelim -S | grep store | not grep null + +; Because the 'internal' function is passed to an external function, we don't +; know what the incoming values will alias. As such, we cannot do the +; optimization checked by the 'arg-must-alias.ll' test. + +declare void @external(i32(i32*)*) +@G = internal constant i32* null + +define internal i32 @internal(i32* %ARG) { + ;;; We *DON'T* know that ARG always points to null! + store i32* %ARG, i32** @G + ret i32 0 +} + +define i32 @foo() { + call void @external(i32(i32*)* @internal) + %V = call i32 @internal(i32* null) + ret i32 %V +} diff --git a/test/Analysis/Andersens/modreftest.ll b/test/Analysis/Andersens/modreftest.ll new file mode 100644 index 0000000..e0c2edc --- /dev/null +++ b/test/Analysis/Andersens/modreftest.ll @@ -0,0 +1,15 @@ +; RUN: opt < %s -anders-aa -gvn -instcombine -S \ +; RUN: | grep {ret i1 true} + +@G = internal global i32* null +declare i32 *@ext() + +define i1 @bar() { + %V1 = load i32** @G + %X2 = call i32 *@ext() + %V2 = load i32** @G + store i32* %X2, i32** @G + + %C = icmp eq i32* %V1, %V2 + ret i1 %C +} diff --git a/test/Analysis/Andersens/modreftest2.ll b/test/Analysis/Andersens/modreftest2.ll new file mode 100644 index 0000000..562c961 --- /dev/null +++ b/test/Analysis/Andersens/modreftest2.ll @@ -0,0 +1,14 @@ +; RUN: opt < %s -anders-aa -gvn -S \ +; RUN: | not grep {ret i32 undef} + +;; From PR 2160 +declare void @f(i32*) + +define i32 @g() { +entry: + %tmp = alloca i32 ; <i32*> [#uses=2] + call void @f( i32* %tmp ) + %tmp2 = load i32* %tmp ; <i32> [#uses=1] + ret i32 %tmp2 +} + diff --git a/test/Analysis/Andersens/trivialtest.ll b/test/Analysis/Andersens/trivialtest.ll new file mode 100644 index 0000000..f9f938f --- /dev/null +++ b/test/Analysis/Andersens/trivialtest.ll @@ -0,0 +1,3 @@ +; RUN: opt < %s -anders-aa -disable-output + +define void @foo() { ret void } diff --git a/test/Analysis/BasicAA/2003-02-26-AccessSizeTest.ll b/test/Analysis/BasicAA/2003-02-26-AccessSizeTest.ll new file mode 100644 index 0000000..6b50a16 --- /dev/null +++ b/test/Analysis/BasicAA/2003-02-26-AccessSizeTest.ll @@ -0,0 +1,18 @@ +; This testcase makes sure that size is taken to account when alias analysis +; is performed. It is not legal to delete the second load instruction because +; the value computed by the first load instruction is changed by the store. + +; RUN: opt < %s -gvn -instcombine -S | grep DONOTREMOVE + +define i32 @test() { + %A = alloca i32 + store i32 0, i32* %A + %X = load i32* %A + %B = bitcast i32* %A to i8* + %C = getelementptr i8* %B, i64 1 + store i8 1, i8* %C ; Aliases %A + %Y.DONOTREMOVE = load i32* %A + %Z = sub i32 %X, %Y.DONOTREMOVE + ret i32 %Z +} + diff --git a/test/Analysis/BasicAA/2003-03-04-GEPCrash.ll b/test/Analysis/BasicAA/2003-03-04-GEPCrash.ll new file mode 100644 index 0000000..4f8eabb --- /dev/null +++ b/test/Analysis/BasicAA/2003-03-04-GEPCrash.ll @@ -0,0 +1,7 @@ +; RUN: opt < %s -basicaa -aa-eval -disable-output 2>/dev/null +; Test for a bug in BasicAA which caused a crash when querying equality of P1&P2 +define void @test({[2 x i32],[2 x i32]}* %A, i64 %X, i64 %Y) { + %P1 = getelementptr {[2 x i32],[2 x i32]}* %A, i64 0, i32 0, i64 %X + %P2 = getelementptr {[2 x i32],[2 x i32]}* %A, i64 0, i32 1, i64 %Y + ret void +} diff --git a/test/Analysis/BasicAA/2003-04-22-GEPProblem.ll b/test/Analysis/BasicAA/2003-04-22-GEPProblem.ll new file mode 100644 index 0000000..f7e8295 --- /dev/null +++ b/test/Analysis/BasicAA/2003-04-22-GEPProblem.ll @@ -0,0 +1,15 @@ +; RUN: opt < %s -gvn -instcombine -S | grep sub + +; BasicAA was incorrectly concluding that P1 and P2 didn't conflict! + +define i32 @test(i32 *%Ptr, i64 %V) { + %P2 = getelementptr i32* %Ptr, i64 1 + %P1 = getelementptr i32* %Ptr, i64 %V + %X = load i32* %P1 + store i32 5, i32* %P2 + + %Y = load i32* %P1 + + %Z = sub i32 %X, %Y + ret i32 %Z +} diff --git a/test/Analysis/BasicAA/2003-04-25-GEPCrash.ll b/test/Analysis/BasicAA/2003-04-25-GEPCrash.ll new file mode 100644 index 0000000..97bc38e --- /dev/null +++ b/test/Analysis/BasicAA/2003-04-25-GEPCrash.ll @@ -0,0 +1,7 @@ +; RUN: opt < %s -basicaa -aa-eval -disable-output 2>/dev/null +; Test for a bug in BasicAA which caused a crash when querying equality of P1&P2 +define void @test([17 x i16]* %mask_bits) { + %P1 = getelementptr [17 x i16]* %mask_bits, i64 0, i64 0 + %P2 = getelementptr [17 x i16]* %mask_bits, i64 252645134, i64 0 + ret void +} diff --git a/test/Analysis/BasicAA/2003-05-21-GEP-Problem.ll b/test/Analysis/BasicAA/2003-05-21-GEP-Problem.ll new file mode 100644 index 0000000..d439dfc --- /dev/null +++ b/test/Analysis/BasicAA/2003-05-21-GEP-Problem.ll @@ -0,0 +1,21 @@ +; RUN: opt < %s -licm -disable-output + %struct..apr_array_header_t = type { i32*, i32, i32, i32, i8* } + %struct..apr_table_t = type { %struct..apr_array_header_t, i32, [32 x i32], [32 x i32] } + +define void @table_reindex(%struct..apr_table_t* %t.1) { ; No predecessors! + br label %loopentry + +loopentry: ; preds = %0, %no_exit + %tmp.101 = getelementptr %struct..apr_table_t* %t.1, i64 0, i32 0, i32 2 + %tmp.11 = load i32* %tmp.101 ; <i32> [#uses=0] + br i1 false, label %no_exit, label %UnifiedExitNode + +no_exit: ; preds = %loopentry + %tmp.25 = sext i32 0 to i64 ; <i64> [#uses=1] + %tmp.261 = getelementptr %struct..apr_table_t* %t.1, i64 0, i32 3, i64 %tmp.25 ; <i32*> [#uses=1] + store i32 0, i32* %tmp.261 + br label %loopentry + +UnifiedExitNode: ; preds = %loopentry + ret void +} diff --git a/test/Analysis/BasicAA/2003-06-01-AliasCrash.ll b/test/Analysis/BasicAA/2003-06-01-AliasCrash.ll new file mode 100644 index 0000000..0abd384 --- /dev/null +++ b/test/Analysis/BasicAA/2003-06-01-AliasCrash.ll @@ -0,0 +1,11 @@ +; RUN: opt < %s -basicaa -aa-eval -disable-output 2>/dev/null + +define i32 @MTConcat([3 x i32]* %a.1) { + %tmp.961 = getelementptr [3 x i32]* %a.1, i64 0, i64 4 + %tmp.97 = load i32* %tmp.961 + %tmp.119 = getelementptr [3 x i32]* %a.1, i64 1, i64 0 + %tmp.120 = load i32* %tmp.119 + %tmp.1541 = getelementptr [3 x i32]* %a.1, i64 0, i64 4 + %tmp.155 = load i32* %tmp.1541 + ret i32 0 +} diff --git a/test/Analysis/BasicAA/2003-07-03-BasicAACrash.ll b/test/Analysis/BasicAA/2003-07-03-BasicAACrash.ll new file mode 100644 index 0000000..3e813fa --- /dev/null +++ b/test/Analysis/BasicAA/2003-07-03-BasicAACrash.ll @@ -0,0 +1,10 @@ +; RUN: opt < %s -basicaa -aa-eval -disable-output 2>/dev/null + +%struct..RefPoint = type { i32, { i32, i8, i8 } } +%struct..RefRect = type { %struct..RefPoint, %struct..RefPoint } + +define i32 @BMT_CommitPartDrawObj() { + %tmp.19111 = getelementptr %struct..RefRect* null, i64 0, i32 0, i32 1, i32 2 + %tmp.20311 = getelementptr %struct..RefRect* null, i64 0, i32 1, i32 1, i32 2 + ret i32 0 +} diff --git a/test/Analysis/BasicAA/2003-09-19-LocalArgument.ll b/test/Analysis/BasicAA/2003-09-19-LocalArgument.ll new file mode 100644 index 0000000..637d8f0 --- /dev/null +++ b/test/Analysis/BasicAA/2003-09-19-LocalArgument.ll @@ -0,0 +1,12 @@ +; In this test, a local alloca cannot alias an incoming argument. + +; RUN: opt < %s -gvn -instcombine -S | not grep sub + +define i32 @test(i32* %P) { + %X = alloca i32 + %V1 = load i32* %P + store i32 0, i32* %X + %V2 = load i32* %P + %Diff = sub i32 %V1, %V2 + ret i32 %Diff +} diff --git a/test/Analysis/BasicAA/2003-11-04-SimpleCases.ll b/test/Analysis/BasicAA/2003-11-04-SimpleCases.ll new file mode 100644 index 0000000..911f78c --- /dev/null +++ b/test/Analysis/BasicAA/2003-11-04-SimpleCases.ll @@ -0,0 +1,16 @@ +; This testcase consists of alias relations which should be completely +; resolvable by basicaa. + +; RUN: opt < %s -aa-eval -print-may-aliases -disable-output \ +; RUN: |& not grep May: + +%T = type { i32, [10 x i8] } + +define void @test(%T* %P) { + %A = getelementptr %T* %P, i64 0 + %B = getelementptr %T* %P, i64 0, i32 0 + %C = getelementptr %T* %P, i64 0, i32 1 + %D = getelementptr %T* %P, i64 0, i32 1, i64 0 + %E = getelementptr %T* %P, i64 0, i32 1, i64 5 + ret void +} diff --git a/test/Analysis/BasicAA/2003-12-11-ConstExprGEP.ll b/test/Analysis/BasicAA/2003-12-11-ConstExprGEP.ll new file mode 100644 index 0000000..8166b97 --- /dev/null +++ b/test/Analysis/BasicAA/2003-12-11-ConstExprGEP.ll @@ -0,0 +1,18 @@ +; This testcase consists of alias relations which should be completely +; resolvable by basicaa, but require analysis of getelementptr constant exprs. + +; RUN: opt < %s -aa-eval -print-may-aliases -disable-output \ +; RUN: |& not grep May: + +%T = type { i32, [10 x i8] } + +@G = external global %T + +define void @test() { + %D = getelementptr %T* @G, i64 0, i32 0 + %E = getelementptr %T* @G, i64 0, i32 1, i64 5 + %F = getelementptr i32* getelementptr (%T* @G, i64 0, i32 0), i64 0 + %X = getelementptr [10 x i8]* getelementptr (%T* @G, i64 0, i32 1), i64 0, i64 5 + + ret void +} diff --git a/test/Analysis/BasicAA/2004-07-28-MustAliasbug.ll b/test/Analysis/BasicAA/2004-07-28-MustAliasbug.ll new file mode 100644 index 0000000..e1cfd03 --- /dev/null +++ b/test/Analysis/BasicAA/2004-07-28-MustAliasbug.ll @@ -0,0 +1,10 @@ +; RUN: opt < %s -dse -S | grep {store i32 0} + +define void @test({i32,i32 }* %P) { + %Q = getelementptr {i32,i32}* %P, i32 1 + %X = getelementptr {i32,i32}* %Q, i32 0, i32 1 + %Y = getelementptr {i32,i32}* %Q, i32 1, i32 1 + store i32 0, i32* %X + store i32 1, i32* %Y + ret void +} diff --git a/test/Analysis/BasicAA/2004-12-08-BasicAACrash.ll b/test/Analysis/BasicAA/2004-12-08-BasicAACrash.ll new file mode 100644 index 0000000..81248db --- /dev/null +++ b/test/Analysis/BasicAA/2004-12-08-BasicAACrash.ll @@ -0,0 +1,22 @@ +; RUN: opt < %s -licm + +%"java/lang/Object" = type { %struct.llvm_java_object_base } +%"java/lang/StringBuffer" = type { "java/lang/Object", i32, { "java/lang/Object", i32, [0 x i8] }*, i1 } +%struct.llvm_java_object_base = type opaque + +define void @"java/lang/StringBuffer/setLength(I)V"(%struct.llvm_java_object_base*) { +bc0: + br i1 false, label %bc40, label %bc65 + +bc65: ; preds = %bc0, %bc40 + ret void + +bc40: ; preds = %bc0, %bc40 + %tmp75 = bitcast %struct.llvm_java_object_base* %0 to %"java/lang/StringBuffer"* ; <"java/lang/StringBuffer"*> [#uses=1] + %tmp76 = getelementptr %"java/lang/StringBuffer"* %tmp75, i32 0, i32 1 ; <i32*> [#uses=1] + store i32 0, i32* %tmp76 + %tmp381 = bitcast %struct.llvm_java_object_base* %0 to %"java/lang/StringBuffer"* ; <"java/lang/StringBuffer"*> [#uses=1] + %tmp392 = getelementptr %"java/lang/StringBuffer"* %tmp381, i32 0, i32 1 ; <i32*> [#uses=1] + %tmp403 = load i32* %tmp392 ; <i32> [#uses=0] + br i1 false, label %bc40, label %bc65 +} diff --git a/test/Analysis/BasicAA/2004-12-08-BasicAACrash2.ll b/test/Analysis/BasicAA/2004-12-08-BasicAACrash2.ll new file mode 100644 index 0000000..0e03db3 --- /dev/null +++ b/test/Analysis/BasicAA/2004-12-08-BasicAACrash2.ll @@ -0,0 +1,20 @@ +; RUN: opt < %s -dse + +%"java/lang/Object" = type { %struct.llvm_java_object_base } +%"java/lang/StringBuffer" = type { "java/lang/Object", i32, { "java/lang/Object", i32, [0 x i8] }*, i1 } +%struct.llvm_java_object_base = type opaque + +define void @"java/lang/StringBuffer/ensureCapacity_unsynchronized(I)V"() { +bc0: + %tmp = getelementptr %"java/lang/StringBuffer"* null, i32 0, i32 3 ; <i1*> [#uses=1] + br i1 false, label %bc16, label %bc7 + +bc16: ; preds = %bc0 + %tmp91 = getelementptr %"java/lang/StringBuffer"* null, i32 0, i32 2 ; <{ "java/lang/Object", i32, [0 x i8] }**> [#uses=1] + store { %"java/lang/Object", i32, [0 x i8] }* null, { %"java/lang/Object", i32, [0 x i8] }** %tmp91 + store i1 false, i1* %tmp + ret void + +bc7: ; preds = %bc0 + ret void +} diff --git a/test/Analysis/BasicAA/2005-03-09-BrokenBasicAA.ll b/test/Analysis/BasicAA/2005-03-09-BrokenBasicAA.ll new file mode 100644 index 0000000..4564263 --- /dev/null +++ b/test/Analysis/BasicAA/2005-03-09-BrokenBasicAA.ll @@ -0,0 +1,15 @@ +; RUN: opt < %s -basicaa -gvn -instcombine |\ +; RUN: llvm-dis | grep {load i32\\* %A} + +declare double* @useit(i32*) + +define i32 @foo(i32 %Amt) { + %A = malloc i32, i32 %Amt + %P = call double* @useit(i32* %A) + + %X = load i32* %A + store double 0.0, double* %P + %Y = load i32* %A + %Z = sub i32 %X, %Y + ret i32 %Z +} diff --git a/test/Analysis/BasicAA/2006-03-03-BadArraySubscript.ll b/test/Analysis/BasicAA/2006-03-03-BadArraySubscript.ll new file mode 100644 index 0000000..49327ac --- /dev/null +++ b/test/Analysis/BasicAA/2006-03-03-BadArraySubscript.ll @@ -0,0 +1,31 @@ +; RUN: opt < %s -aa-eval -disable-output |& grep {2 no alias respon} +; TEST that A[1][0] may alias A[0][i]. +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" + +define void @test(i32 %N) { +entry: + %X = alloca [3 x [3 x i32]] ; <[3 x [3 x i32]]*> [#uses=4] + %tmp.24 = icmp sgt i32 %N, 0 ; <i1> [#uses=1] + br i1 %tmp.24, label %no_exit, label %loopexit + +no_exit: ; preds = %no_exit, %entry + %i.0.0 = phi i32 [ 0, %entry ], [ %inc, %no_exit ] ; <i32> [#uses=2] + %tmp.6 = getelementptr [3 x [3 x i32]]* %X, i32 0, i32 0, i32 %i.0.0 ; <i32*> [#uses=1] + store i32 1, i32* %tmp.6 + %tmp.8 = getelementptr [3 x [3 x i32]]* %X, i32 0, i32 0, i32 0 ; <i32*> [#uses=1] + %tmp.9 = load i32* %tmp.8 ; <i32> [#uses=1] + %tmp.11 = getelementptr [3 x [3 x i32]]* %X, i32 0, i32 1, i32 0 ; <i32*> [#uses=1] + %tmp.12 = load i32* %tmp.11 ; <i32> [#uses=1] + %tmp.13 = add i32 %tmp.12, %tmp.9 ; <i32> [#uses=1] + %inc = add i32 %i.0.0, 1 ; <i32> [#uses=2] + %tmp.2 = icmp slt i32 %inc, %N ; <i1> [#uses=1] + br i1 %tmp.2, label %no_exit, label %loopexit + +loopexit: ; preds = %no_exit, %entry + %Y.0.1 = phi i32 [ 0, %entry ], [ %tmp.13, %no_exit ] ; <i32> [#uses=1] + %tmp.4 = getelementptr [3 x [3 x i32]]* %X, i32 0, i32 0 ; <[3 x i32]*> [#uses=1] + %tmp.15 = call i32 (...)* @foo( [3 x i32]* %tmp.4, i32 %Y.0.1 ) ; <i32> [#uses=0] + ret void +} + +declare i32 @foo(...) diff --git a/test/Analysis/BasicAA/2006-11-03-BasicAAVectorCrash.ll b/test/Analysis/BasicAA/2006-11-03-BasicAAVectorCrash.ll new file mode 100644 index 0000000..85f53a6 --- /dev/null +++ b/test/Analysis/BasicAA/2006-11-03-BasicAAVectorCrash.ll @@ -0,0 +1,48 @@ +; RUN: opt < %s -licm -disable-output +target datalayout = "E-p:32:32" +target triple = "powerpc-apple-darwin8.7.0" + +define void @glgRunProcessor() { +entry: + br i1 false, label %bb2037.i, label %cond_true.i18 + +cond_true.i18: ; preds = %entry + ret void + +bb205.i: ; preds = %bb2037.i + switch i32 0, label %bb1013.i [ + i32 14, label %bb239.i + i32 15, label %bb917.i + ] + +bb239.i: ; preds = %bb205.i + br i1 false, label %cond_false277.i, label %cond_true264.i + +cond_true264.i: ; preds = %bb239.i + ret void + +cond_false277.i: ; preds = %bb239.i + %tmp1062.i = getelementptr [2 x <4 x i32>]* null, i32 0, i32 1 ; <<4 x i32>*> [#uses=1] + store <4 x i32> zeroinitializer, <4 x i32>* %tmp1062.i + br i1 false, label %cond_true1032.i, label %cond_false1063.i85 + +bb917.i: ; preds = %bb205.i + ret void + +bb1013.i: ; preds = %bb205.i + ret void + +cond_true1032.i: ; preds = %cond_false277.i + %tmp1187.i = getelementptr [2 x <4 x i32>]* null, i32 0, i32 0, i32 7 ; <i32*> [#uses=1] + store i32 0, i32* %tmp1187.i + br label %bb2037.i + +cond_false1063.i85: ; preds = %cond_false277.i + ret void + +bb2037.i: ; preds = %cond_true1032.i, %entry + br i1 false, label %bb205.i, label %cond_next2042.i + +cond_next2042.i: ; preds = %bb2037.i + ret void +} diff --git a/test/Analysis/BasicAA/2007-01-13-BasePointerBadNoAlias.ll b/test/Analysis/BasicAA/2007-01-13-BasePointerBadNoAlias.ll new file mode 100644 index 0000000..917bf25 --- /dev/null +++ b/test/Analysis/BasicAA/2007-01-13-BasePointerBadNoAlias.ll @@ -0,0 +1,35 @@ +; PR1109 +; RUN: opt < %s -basicaa -gvn -instcombine -S | \ +; RUN: grep {sub i32} +; RUN: opt < %s -basicaa -gvn -instcombine -S | \ +; RUN: not grep {ret i32 0} +; END. + +target datalayout = "e-p:32:32" +target triple = "i686-apple-darwin8" + %struct.CONSTRAINT = type { i32, i32, i32, i32 } + %struct.FILE_POS = type { i8, i8, i16, i32 } + %struct.FIRST_UNION = type { %struct.FILE_POS } + %struct.FOURTH_UNION = type { %struct.CONSTRAINT } + %struct.GAP = type { i8, i8, i16 } + %struct.LIST = type { %struct.rec*, %struct.rec* } + %struct.SECOND_UNION = type { { i16, i8, i8 } } + %struct.STYLE = type { { %struct.GAP }, { %struct.GAP }, i16, i16, i16, i8, i8 } + %struct.THIRD_UNION = type { { [2 x i32], [2 x i32] } } + %struct.closure_type = type { [2 x %struct.LIST], %struct.FIRST_UNION, %struct.SECOND_UNION, %struct.THIRD_UNION, %struct.FOURTH_UNION, %struct.rec*, { %struct.rec* } } + %struct.head_type = type { [2 x %struct.LIST], %struct.FIRST_UNION, %struct.SECOND_UNION, %struct.THIRD_UNION, %struct.FOURTH_UNION, %struct.rec*, { %struct.rec* }, %struct.rec*, %struct.rec*, %struct.rec*, %struct.rec*, %struct.rec*, %struct.rec*, %struct.rec*, %struct.rec*, i32 } + %struct.rec = type { %struct.head_type } + + +define i32 @test(%struct.closure_type* %tmp18169) { + %tmp18174 = getelementptr %struct.closure_type* %tmp18169, i32 0, i32 4, i32 0, i32 0 ; <i32*> [#uses=2] + %tmp18269 = bitcast i32* %tmp18174 to %struct.STYLE* ; <%struct.STYLE*> [#uses=1] + %A = load i32* %tmp18174 ; <i32> [#uses=1] + + %tmp18272 = getelementptr %struct.STYLE* %tmp18269, i32 0, i32 0, i32 0, i32 2 ; <i16*> [#uses=1] + store i16 123, i16* %tmp18272 + + %Q = load i32* %tmp18174 ; <i32> [#uses=1] + %Z = sub i32 %A, %Q ; <i32> [#uses=1] + ret i32 %Z +} diff --git a/test/Analysis/BasicAA/2007-08-01-NoAliasAndCalls.ll b/test/Analysis/BasicAA/2007-08-01-NoAliasAndCalls.ll new file mode 100644 index 0000000..e6a26e3 --- /dev/null +++ b/test/Analysis/BasicAA/2007-08-01-NoAliasAndCalls.ll @@ -0,0 +1,10 @@ +; RUN: opt < %s -basicaa -aa-eval -print-all-alias-modref-info -disable-output |& grep {MayAlias:.*i32\\* %., i32\\* %.} | grep {%x} | grep {%y} + +declare i32* @unclear(i32* %a) + +define void @foo(i32* noalias %x) { + %y = call i32* @unclear(i32* %x) + store i32 0, i32* %x + store i32 0, i32* %y + ret void +} diff --git a/test/Analysis/BasicAA/2007-08-01-NoAliasAndGEP.ll b/test/Analysis/BasicAA/2007-08-01-NoAliasAndGEP.ll new file mode 100644 index 0000000..7f33fa4 --- /dev/null +++ b/test/Analysis/BasicAA/2007-08-01-NoAliasAndGEP.ll @@ -0,0 +1,17 @@ +; RUN: opt < %s -basicaa -aa-eval -print-all-alias-modref-info -disable-output |& grep {9 no alias} +; RUN: opt < %s -basicaa -aa-eval -print-all-alias-modref-info -disable-output |& grep {6 may alias} +; RUN: opt < %s -basicaa -aa-eval -print-all-alias-modref-info -disable-output |& grep {MayAlias:.*i32\\* %Ipointer, i32\\* %Jpointer} + +define void @foo(i32* noalias %p, i32* noalias %q, i32 %i, i32 %j) { + %Ipointer = getelementptr i32* %p, i32 %i + %qi = getelementptr i32* %q, i32 %i + %Jpointer = getelementptr i32* %p, i32 %j + %qj = getelementptr i32* %q, i32 %j + store i32 0, i32* %p + store i32 0, i32* %Ipointer + store i32 0, i32* %Jpointer + store i32 0, i32* %q + store i32 0, i32* %qi + store i32 0, i32* %qj + ret void +} diff --git a/test/Analysis/BasicAA/2007-08-05-GetOverloadedModRef.ll b/test/Analysis/BasicAA/2007-08-05-GetOverloadedModRef.ll new file mode 100644 index 0000000..035299e --- /dev/null +++ b/test/Analysis/BasicAA/2007-08-05-GetOverloadedModRef.ll @@ -0,0 +1,17 @@ +; PR1600 +; RUN: opt < %s -basicaa -gvn -instcombine -S | \ +; RUN: grep {ret i32 0} +; END. + +declare i16 @llvm.cttz.i16(i16) + +define i32 @test(i32* %P, i16* %Q) { + %A = load i16* %Q ; <i16> [#uses=1] + %x = load i32* %P ; <i32> [#uses=1] + %B = call i16 @llvm.cttz.i16( i16 %A ) ; <i16> [#uses=1] + %y = load i32* %P ; <i32> [#uses=1] + store i16 %B, i16* %Q + %z = sub i32 %x, %y ; <i32> [#uses=1] + ret i32 %z +} + diff --git a/test/Analysis/BasicAA/2007-10-24-ArgumentsGlobals.ll b/test/Analysis/BasicAA/2007-10-24-ArgumentsGlobals.ll new file mode 100644 index 0000000..78f24b5 --- /dev/null +++ b/test/Analysis/BasicAA/2007-10-24-ArgumentsGlobals.ll @@ -0,0 +1,14 @@ +; RUN: opt < %s -basicaa -gvn -dce -S | grep tmp7 + + %struct.A = type { i32 } + %struct.B = type { %struct.A } +@a = global %struct.B zeroinitializer ; <%struct.B*> [#uses=2] + +define i32 @_Z3fooP1A(%struct.A* %b) { +entry: + store i32 1, i32* getelementptr (%struct.B* @a, i32 0, i32 0, i32 0), align 8 + %tmp4 = getelementptr %struct.A* %b, i32 0, i32 0 ;<i32*> [#uses=1] + store i32 0, i32* %tmp4, align 4 + %tmp7 = load i32* getelementptr (%struct.B* @a, i32 0, i32 0, i32 0), align 8 ; <i32> [#uses=1] + ret i32 %tmp7 +} diff --git a/test/Analysis/BasicAA/2007-11-05-SizeCrash.ll b/test/Analysis/BasicAA/2007-11-05-SizeCrash.ll new file mode 100644 index 0000000..f699ba2 --- /dev/null +++ b/test/Analysis/BasicAA/2007-11-05-SizeCrash.ll @@ -0,0 +1,34 @@ +; RUN: opt < %s -gvn -disable-output +; PR1774 + +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" +target triple = "x86_64-unknown-linux-gnu" + %struct.device = type { [20 x i8] } + %struct.pci_device_id = type { i32, i32, i32, i32, i32, i32, i64 } + %struct.usb_bus = type { %struct.device* } + %struct.usb_hcd = type { %struct.usb_bus, i64, [0 x i64] } +@uhci_pci_ids = external constant [1 x %struct.pci_device_id] ; <[1 x %struct.pci_device_id]*> [#uses=1] + +@__mod_pci_device_table = alias [1 x %struct.pci_device_id]* @uhci_pci_ids + ; <[1 x %struct.pci_device_id]*> [#uses=0] + +define i32 @uhci_suspend(%struct.usb_hcd* %hcd) { +entry: + %tmp17 = getelementptr %struct.usb_hcd* %hcd, i32 0, i32 2, i64 1 + ; <i64*> [#uses=1] + %tmp1718 = bitcast i64* %tmp17 to i32* ; <i32*> [#uses=1] + %tmp19 = load i32* %tmp1718, align 4 ; <i32> [#uses=0] + br i1 false, label %cond_true34, label %done_okay + +cond_true34: ; preds = %entry + %tmp631 = getelementptr %struct.usb_hcd* %hcd, i32 0, i32 2, i64 +2305843009213693950 ; <i64*> [#uses=1] + %tmp70 = bitcast i64* %tmp631 to %struct.device** + + %tmp71 = load %struct.device** %tmp70, align 8 + + ret i32 undef + +done_okay: ; preds = %entry + ret i32 undef +} diff --git a/test/Analysis/BasicAA/2007-12-08-OutOfBoundsCrash.ll b/test/Analysis/BasicAA/2007-12-08-OutOfBoundsCrash.ll new file mode 100644 index 0000000..8028afb --- /dev/null +++ b/test/Analysis/BasicAA/2007-12-08-OutOfBoundsCrash.ll @@ -0,0 +1,31 @@ +; RUN: opt < %s -gvn -disable-output +; PR1782 + +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" +target triple = "x86_64-unknown-linux-gnu" + %struct.device = type { [20 x i8] } + %struct.pci_device_id = type { i32, i32, i32, i32, i32, i32, i64 } + %struct.usb_bus = type { %struct.device* } + %struct.usb_hcd = type { %struct.usb_bus, [0 x i64] } +@pci_ids = external constant [1 x %struct.pci_device_id] ; <[1 x %struct.pci_device_id]*> [#uses=1] + +@__mod_pci_device_table = alias [1 x %struct.pci_device_id]* @pci_ids ; <[1 x %struct.pci_device_id]*> [#uses=0] + +define i32 @ehci_pci_setup(%struct.usb_hcd* %hcd) { +entry: + %tmp14 = getelementptr %struct.usb_hcd* %hcd, i32 0, i32 0, i32 0 ; <%struct.device**> [#uses=1] + %tmp15 = load %struct.device** %tmp14, align 8 ; <%struct.device*> [#uses=0] + br i1 false, label %bb25, label %return + +bb25: ; preds = %entry + br i1 false, label %cond_true, label %return + +cond_true: ; preds = %bb25 + %tmp601 = getelementptr %struct.usb_hcd* %hcd, i32 0, i32 1, i64 2305843009213693951 ; <i64*> [#uses=1] + %tmp67 = bitcast i64* %tmp601 to %struct.device** ; <%struct.device**> [#uses=1] + %tmp68 = load %struct.device** %tmp67, align 8 ; <%struct.device*> [#uses=0] + ret i32 undef + +return: ; preds = %bb25, %entry + ret i32 undef +} diff --git a/test/Analysis/BasicAA/2008-04-15-Byval.ll b/test/Analysis/BasicAA/2008-04-15-Byval.ll new file mode 100644 index 0000000..2069401 --- /dev/null +++ b/test/Analysis/BasicAA/2008-04-15-Byval.ll @@ -0,0 +1,18 @@ +; RUN: opt < %s -std-compile-opts -S | grep store +; ModuleID = 'small2.c' +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-darwin8" + %struct.x = type { [4 x i32] } + +define void @foo(%struct.x* byval align 4 %X) nounwind { +entry: + %tmp = getelementptr %struct.x* %X, i32 0, i32 0 ; <[4 x i32]*> [#uses=1] + %tmp1 = getelementptr [4 x i32]* %tmp, i32 0, i32 3 ; <i32*> [#uses=1] + store i32 2, i32* %tmp1, align 4 + %tmp2 = call i32 (...)* @bar( %struct.x* byval align 4 %X ) nounwind ; <i32> [#uses=0] + br label %return +return: ; preds = %entry + ret void +} + +declare i32 @bar(...) diff --git a/test/Analysis/BasicAA/2008-06-02-GEPTailCrash.ll b/test/Analysis/BasicAA/2008-06-02-GEPTailCrash.ll new file mode 100644 index 0000000..ba29f3a --- /dev/null +++ b/test/Analysis/BasicAA/2008-06-02-GEPTailCrash.ll @@ -0,0 +1,15 @@ +; RUN: opt < %s -gvn -disable-output +; PR2395 + +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-f80:32:32" +target triple = "i686-pc-linux-gnu" + %struct.S291 = type <{ %union.anon, i32 }> + %union.anon = type { } +@a291 = external global [5 x %struct.S291] ; <[5 x %struct.S291]*> [#uses=2] + +define void @test291() nounwind { +entry: + store i32 1138410269, i32* getelementptr ([5 x %struct.S291]* @a291, i32 0, i32 2, i32 1) + %tmp54 = load i32* bitcast (%struct.S291* getelementptr ([5 x %struct.S291]* @a291, i32 0, i32 2) to i32*), align 4 ; <i32> [#uses=0] + unreachable +} diff --git a/test/Analysis/BasicAA/2008-11-23-NoaliasRet.ll b/test/Analysis/BasicAA/2008-11-23-NoaliasRet.ll new file mode 100644 index 0000000..06018cc --- /dev/null +++ b/test/Analysis/BasicAA/2008-11-23-NoaliasRet.ll @@ -0,0 +1,12 @@ +; RUN: opt < %s -aa-eval |& grep {1 no alias response} + +declare noalias i32* @_Znwj(i32 %x) nounwind + +define i32 @foo() { + %A = call i32* @_Znwj(i32 4) + %B = call i32* @_Znwj(i32 4) + store i32 1, i32* %A + store i32 2, i32* %B + %C = load i32* %A + ret i32 %C +} diff --git a/test/Analysis/BasicAA/2009-03-04-GEPNoalias.ll b/test/Analysis/BasicAA/2009-03-04-GEPNoalias.ll new file mode 100644 index 0000000..3ab5d03 --- /dev/null +++ b/test/Analysis/BasicAA/2009-03-04-GEPNoalias.ll @@ -0,0 +1,13 @@ +; RUN: opt < %s -basicaa -gvn -S | grep load + +declare noalias i32* @noalias() + +define i32 @test(i32 %x) { + %a = call i32* @noalias() + store i32 1, i32* %a + %b = getelementptr i32* %a, i32 %x + store i32 2, i32* %b + + %c = load i32* %a + ret i32 %c +} diff --git a/test/Analysis/BasicAA/2009-10-13-AtomicModRef.ll b/test/Analysis/BasicAA/2009-10-13-AtomicModRef.ll new file mode 100644 index 0000000..6475471 --- /dev/null +++ b/test/Analysis/BasicAA/2009-10-13-AtomicModRef.ll @@ -0,0 +1,17 @@ +; RUN: opt -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) + %Z = load i8* %P +; CHECK-NOT: = load + %A = sub i8 %X, %Z + ret i8 %A +; CHECK: ret i8 0 +} diff --git a/test/Analysis/BasicAA/2009-10-13-GEP-BaseNoAlias.ll b/test/Analysis/BasicAA/2009-10-13-GEP-BaseNoAlias.ll new file mode 100644 index 0000000..771636f --- /dev/null +++ b/test/Analysis/BasicAA/2009-10-13-GEP-BaseNoAlias.ll @@ -0,0 +1,30 @@ +; RUN: opt < %s -aa-eval -print-all-alias-modref-info -disable-output |& grep {NoAlias:.*%P,.*@Z} +; If GEP base doesn't alias Z, then GEP doesn't alias Z. +; rdar://7282591 + +@Y = common global i32 0 +@Z = common global i32 0 + +define void @foo(i32 %cond) nounwind ssp { +entry: + %a = alloca i32 + %tmp = icmp ne i32 %cond, 0 + br i1 %tmp, label %bb, label %bb1 + +bb: + %b = getelementptr i32* %a, i32 0 + br label %bb2 + +bb1: + br label %bb2 + +bb2: + %P = phi i32* [ %b, %bb ], [ @Y, %bb1 ] + %tmp1 = load i32* @Z, align 4 + store i32 123, i32* %P, align 4 + %tmp2 = load i32* @Z, align 4 + br label %return + +return: + ret void +} diff --git a/test/Analysis/BasicAA/byval.ll b/test/Analysis/BasicAA/byval.ll new file mode 100644 index 0000000..cdcafdf --- /dev/null +++ b/test/Analysis/BasicAA/byval.ll @@ -0,0 +1,18 @@ +; RUN: opt < %s -gvn -S | grep {ret i32 1} +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 = "i686-apple-darwin8" + %struct.x = type { i32, i32, i32, i32 } +@g = weak global i32 0 ; <i32*> [#uses=1] + +define i32 @foo(%struct.x* byval %a) nounwind { +entry: + %tmp1 = tail call i32 (...)* @bar( %struct.x* %a ) nounwind ; <i32> [#uses=0] + %tmp2 = getelementptr %struct.x* %a, i32 0, i32 0 ; <i32*> [#uses=2] + store i32 1, i32* %tmp2, align 4 + store i32 2, i32* @g, align 4 + %tmp4 = load i32* %tmp2, align 4 ; <i32> [#uses=1] + ret i32 %tmp4 +} + +declare i32 @bar(...) + diff --git a/test/Analysis/BasicAA/cas.ll b/test/Analysis/BasicAA/cas.ll new file mode 100644 index 0000000..4ce7811 --- /dev/null +++ b/test/Analysis/BasicAA/cas.ll @@ -0,0 +1,15 @@ +; RUN: opt < %s -basicaa -gvn -instcombine -S | grep {ret i32 0} + +@flag0 = internal global i32 zeroinitializer +@turn = internal global i32 zeroinitializer + + +define i32 @main() { + %a = load i32* @flag0 + %b = tail call i32 @llvm.atomic.swap.i32.p0i32(i32* @turn, i32 1) + %c = load i32* @flag0 + %d = sub i32 %a, %c + ret i32 %d +} + +declare i32 @llvm.atomic.swap.i32.p0i32(i32*, i32) nounwind
\ No newline at end of file diff --git a/test/Analysis/BasicAA/constant-over-index.ll b/test/Analysis/BasicAA/constant-over-index.ll new file mode 100644 index 0000000..95f94d0 --- /dev/null +++ b/test/Analysis/BasicAA/constant-over-index.ll @@ -0,0 +1,27 @@ +; RUN: opt < %s -aa-eval -print-all-alias-modref-info \ +; RUN: |& grep {MayAlias: double\\* \[%\]p.0.i.0, double\\* \[%\]p3\$} +; PR4267 + +; %p3 is equal to %p.0.i.0 on the second iteration of the loop, +; so MayAlias is needed. + +define void @foo([3 x [3 x double]]* noalias %p) { +entry: + %p3 = getelementptr [3 x [3 x double]]* %p, i64 0, i64 0, i64 3 + br label %loop + +loop: + %i = phi i64 [ 0, %entry ], [ %i.next, %loop ] + + %p.0.i.0 = getelementptr [3 x [3 x double]]* %p, i64 0, i64 %i, i64 0 + + volatile store double 0.0, double* %p3 + volatile store double 0.1, double* %p.0.i.0 + + %i.next = add i64 %i, 1 + %cmp = icmp slt i64 %i.next, 3 + br i1 %cmp, label %loop, label %exit + +exit: + ret void +} diff --git a/test/Analysis/BasicAA/dg.exp b/test/Analysis/BasicAA/dg.exp new file mode 100644 index 0000000..f200589 --- /dev/null +++ b/test/Analysis/BasicAA/dg.exp @@ -0,0 +1,3 @@ +load_lib llvm.exp + +RunLLVMTests [lsort [glob -nocomplain $srcdir/$subdir/*.{ll,c,cpp}]] diff --git a/test/Analysis/BasicAA/featuretest.ll b/test/Analysis/BasicAA/featuretest.ll new file mode 100644 index 0000000..50dc886 --- /dev/null +++ b/test/Analysis/BasicAA/featuretest.ll @@ -0,0 +1,83 @@ +; This testcase tests for various features the basicaa test should be able to +; determine, as noted in the comments. + +; RUN: opt < %s -basicaa -gvn -instcombine -dce -S | not grep REMOVE +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" + +@Global = external global { i32 } + +; Array test: Test that operations on one local array do not invalidate +; operations on another array. Important for scientific codes. +; +define i32 @different_array_test(i64 %A, i64 %B) { + %Array1 = alloca i32, i32 100 + %Array2 = alloca i32, i32 200 + + %pointer = getelementptr i32* %Array1, i64 %A + %val = load i32* %pointer + + %pointer2 = getelementptr i32* %Array2, i64 %B + store i32 7, i32* %pointer2 + + %REMOVE = load i32* %pointer ; redundant with above load + %retval = sub i32 %REMOVE, %val + ret i32 %retval +} + +; Constant index test: Constant indexes into the same array should not +; interfere with each other. Again, important for scientific codes. +; +define i32 @constant_array_index_test() { + %Array = alloca i32, i32 100 + %P1 = getelementptr i32* %Array, i64 7 + %P2 = getelementptr i32* %Array, i64 6 + + %A = load i32* %P1 + store i32 1, i32* %P2 ; Should not invalidate load + %BREMOVE = load i32* %P1 + %Val = sub i32 %A, %BREMOVE + ret i32 %Val +} + +; Test that if two pointers are spaced out by a constant getelementptr, that +; they cannot alias. +define i32 @gep_distance_test(i32* %A) { + %REMOVEu = load i32* %A + %B = getelementptr i32* %A, i64 2 ; Cannot alias A + store i32 7, i32* %B + %REMOVEv = load i32* %A + %r = sub i32 %REMOVEu, %REMOVEv + ret i32 %r +} + +; Test that if two pointers are spaced out by a constant offset, that they +; cannot alias, even if there is a variable offset between them... +define i32 @gep_distance_test2({i32,i32}* %A, i64 %distance) { + %A1 = getelementptr {i32,i32}* %A, i64 0, i32 0 + %REMOVEu = load i32* %A1 + %B = getelementptr {i32,i32}* %A, i64 %distance, i32 1 + store i32 7, i32* %B ; B cannot alias A, it's at least 4 bytes away + %REMOVEv = load i32* %A1 + %r = sub i32 %REMOVEu, %REMOVEv + ret i32 %r +} + +; Test that we can do funny pointer things and that distance calc will still +; work. +define i32 @gep_distance_test3(i32 * %A) { + %X = load i32* %A + %B = bitcast i32* %A to i8* + %C = getelementptr i8* %B, i64 4 + %Y = load i8* %C + ret i32 8 +} + +; Test that we can disambiguate globals reached through constantexpr geps +define i32 @constexpr_test() { + %X = alloca i32 + %Y = load i32* %X + store i32 5, i32* getelementptr ({ i32 }* @Global, i64 0, i32 0) + %REMOVE = load i32* %X + %retval = sub i32 %Y, %REMOVE + ret i32 %retval +} diff --git a/test/Analysis/BasicAA/gcsetest.ll b/test/Analysis/BasicAA/gcsetest.ll new file mode 100644 index 0000000..a903362 --- /dev/null +++ b/test/Analysis/BasicAA/gcsetest.ll @@ -0,0 +1,46 @@ +; Test that GCSE uses basicaa to do alias analysis, which is capable of +; disambiguating some obvious cases. All loads should be removable in +; this testcase. + +; RUN: opt < %s -basicaa -gvn -instcombine -dce -S \ +; RUN: | not grep load + +@A = global i32 7 +@B = global i32 8 + +define i32 @test() { + %A1 = load i32* @A + + store i32 123, i32* @B ; Store cannot alias @A + + %A2 = load i32* @A + %X = sub i32 %A1, %A2 + ret i32 %X +} + +define i32 @test2() { + %A1 = load i32* @A + br label %Loop +Loop: + %AP = phi i32 [0, %0], [%X, %Loop] + store i32 %AP, i32* @B ; Store cannot alias @A + + %A2 = load i32* @A + %X = sub i32 %A1, %A2 + %c = icmp eq i32 %X, 0 + br i1 %c, label %out, label %Loop + +out: + ret i32 %X +} + +declare void @external() + +define i32 @test3() { + %X = alloca i32 + store i32 7, i32* %X + call void @external() + %V = load i32* %X + ret i32 %V +} + diff --git a/test/Analysis/BasicAA/gep-alias.ll b/test/Analysis/BasicAA/gep-alias.ll new file mode 100644 index 0000000..1ed0312 --- /dev/null +++ b/test/Analysis/BasicAA/gep-alias.ll @@ -0,0 +1,171 @@ +; RUN: opt < %s -gvn -instcombine -S |& FileCheck %s + +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" + +; Make sure that basicaa thinks R and r are must aliases. +define i32 @test1(i8 * %P) { +entry: + %Q = bitcast i8* %P to {i32, i32}* + %R = getelementptr {i32, i32}* %Q, i32 0, i32 1 + %S = load i32* %R + + %q = bitcast i8* %P to {i32, i32}* + %r = getelementptr {i32, i32}* %q, i32 0, i32 1 + %s = load i32* %r + + %t = sub i32 %S, %s + ret i32 %t +; CHECK: @test1 +; CHECK: ret i32 0 +} + +define i32 @test2(i8 * %P) { +entry: + %Q = bitcast i8* %P to {i32, i32, i32}* + %R = getelementptr {i32, i32, i32}* %Q, i32 0, i32 1 + %S = load i32* %R + + %r = getelementptr {i32, i32, i32}* %Q, i32 0, i32 2 + store i32 42, i32* %r + + %s = load i32* %R + + %t = sub i32 %S, %s + ret i32 %t +; CHECK: @test2 +; CHECK: ret i32 0 +} + + +; This was a miscompilation. +define i32 @test3({float, {i32, i32, i32}}* %P) { +entry: + %P2 = getelementptr {float, {i32, i32, i32}}* %P, i32 0, i32 1 + %R = getelementptr {i32, i32, i32}* %P2, i32 0, i32 1 + %S = load i32* %R + + %r = getelementptr {i32, i32, i32}* %P2, i32 0, i32 2 + store i32 42, i32* %r + + %s = load i32* %R + + %t = sub i32 %S, %s + ret i32 %t +; CHECK: @test3 +; CHECK: ret i32 0 +} + + +;; This is reduced from the SmallPtrSet constructor. +%SmallPtrSetImpl = type { i8**, i32, i32, i32, [1 x i8*] } +%SmallPtrSet64 = type { %SmallPtrSetImpl, [64 x i8*] } + +define i32 @test4(%SmallPtrSet64* %P) { +entry: + %tmp2 = getelementptr inbounds %SmallPtrSet64* %P, i64 0, i32 0, i32 1 + store i32 64, i32* %tmp2, align 8 + %tmp3 = getelementptr inbounds %SmallPtrSet64* %P, i64 0, i32 0, i32 4, i64 64 + store i8* null, i8** %tmp3, align 8 + %tmp4 = load i32* %tmp2, align 8 + ret i32 %tmp4 +; CHECK: @test4 +; CHECK: ret i32 64 +} + +; P[i] != p[i+1] +define i32 @test5(i32* %p, i64 %i) { + %pi = getelementptr i32* %p, i64 %i + %i.next = add i64 %i, 1 + %pi.next = getelementptr i32* %p, i64 %i.next + %x = load i32* %pi + store i32 42, i32* %pi.next + %y = load i32* %pi + %z = sub i32 %x, %y + ret i32 %z +; CHECK: @test5 +; CHECK: ret i32 0 +} + +; P[i] != p[(i*4)|1] +define i32 @test6(i32* %p, i64 %i1) { + %i = shl i64 %i1, 2 + %pi = getelementptr i32* %p, i64 %i + %i.next = or i64 %i, 1 + %pi.next = getelementptr i32* %p, i64 %i.next + %x = load i32* %pi + store i32 42, i32* %pi.next + %y = load i32* %pi + %z = sub i32 %x, %y + ret i32 %z +; CHECK: @test6 +; CHECK: ret i32 0 +} + +; P[1] != P[i*4] +define i32 @test7(i32* %p, i64 %i) { + %pi = getelementptr i32* %p, i64 1 + %i.next = shl i64 %i, 2 + %pi.next = getelementptr i32* %p, i64 %i.next + %x = load i32* %pi + store i32 42, i32* %pi.next + %y = load i32* %pi + %z = sub i32 %x, %y + ret i32 %z +; CHECK: @test7 +; CHECK: ret i32 0 +} + +; P[zext(i)] != p[zext(i+1)] +; PR1143 +define i32 @test8(i32* %p, i32 %i) { + %i1 = zext i32 %i to i64 + %pi = getelementptr i32* %p, i64 %i1 + %i.next = add i32 %i, 1 + %i.next2 = zext i32 %i.next to i64 + %pi.next = getelementptr i32* %p, i64 %i.next2 + %x = load i32* %pi + store i32 42, i32* %pi.next + %y = load i32* %pi + %z = sub i32 %x, %y + ret i32 %z +; CHECK: @test8 +; CHECK: ret i32 0 +} + +define i8 @test9([4 x i8] *%P, i32 %i, i32 %j) { + %i2 = shl i32 %i, 2 + %i3 = add i32 %i2, 1 + ; P2 = P + 1 + 4*i + %P2 = getelementptr [4 x i8] *%P, i32 0, i32 %i3 + + %j2 = shl i32 %j, 2 + + ; P4 = P + 4*j + %P4 = getelementptr [4 x i8]* %P, i32 0, i32 %j2 + + %x = load i8* %P2 + store i8 42, i8* %P4 + %y = load i8* %P2 + %z = sub i8 %x, %y + ret i8 %z +; CHECK: @test9 +; CHECK: ret i8 0 +} + +define i8 @test10([4 x i8] *%P, i32 %i) { + %i2 = shl i32 %i, 2 + %i3 = add i32 %i2, 4 + ; P2 = P + 4 + 4*i + %P2 = getelementptr [4 x i8] *%P, i32 0, i32 %i3 + + ; P4 = P + 4*i + %P4 = getelementptr [4 x i8]* %P, i32 0, i32 %i2 + + %x = load i8* %P2 + store i8 42, i8* %P4 + %y = load i8* %P2 + %z = sub i8 %x, %y + ret i8 %z +; CHECK: @test10 +; CHECK: ret i8 0 +} diff --git a/test/Analysis/BasicAA/global-size.ll b/test/Analysis/BasicAA/global-size.ll new file mode 100644 index 0000000..b9cbbcc --- /dev/null +++ b/test/Analysis/BasicAA/global-size.ll @@ -0,0 +1,16 @@ +; A store or load cannot alias a global if the accessed amount is larger then +; the global. + +; RUN: opt < %s -basicaa -gvn -instcombine -S | not grep load +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" + +@B = global i16 8 ; <i16*> [#uses=2] + +define i16 @test(i32* %P) { + %X = load i16* @B ; <i16> [#uses=1] + store i32 7, i32* %P + %Y = load i16* @B ; <i16> [#uses=1] + %Z = sub i16 %Y, %X ; <i16> [#uses=1] + ret i16 %Z +} + diff --git a/test/Analysis/BasicAA/modref.ll b/test/Analysis/BasicAA/modref.ll new file mode 100644 index 0000000..4a61636 --- /dev/null +++ b/test/Analysis/BasicAA/modref.ll @@ -0,0 +1,125 @@ +; RUN: opt < %s -basicaa -gvn -dse -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 void @llvm.memset.i32(i8*, i8, i32, i32) +declare void @llvm.memset.i8(i8*, i8, i8, i32) +declare void @llvm.memcpy.i8(i8*, i8*, i8, i32) +declare void @llvm.memcpy.i32(i8*, i8*, i32, i32) +declare void @llvm.lifetime.end(i64, i8* nocapture) + +declare void @external(i32*) + +define i32 @test0(i8* %P) { + %A = alloca i32 + call void @external(i32* %A) + + store i32 0, i32* %A + + call void @llvm.memset.i32(i8* %P, i8 0, i32 42, i32 1) + + %B = load i32* %A + ret i32 %B + +; CHECK: @test0 +; CHECK: ret i32 0 +} + +declare void @llvm.memcpy.i8(i8*, i8*, i8, i32) + +define i8 @test1() { +; CHECK: @test1 + %A = alloca i8 + %B = alloca i8 + + store i8 2, i8* %B ;; Not written to by memcpy + + call void @llvm.memcpy.i8(i8* %A, i8* %B, i8 -1, i32 0) + + %C = load i8* %B + ret i8 %C +; CHECK: ret i8 2 +} + +define i8 @test2(i8* %P) { +; CHECK: @test2 + %P2 = getelementptr i8* %P, i32 127 + store i8 1, i8* %P2 ;; Not dead across memset + call void @llvm.memset.i8(i8* %P, i8 2, i8 127, i32 0) + %A = load i8* %P2 + ret i8 %A +; CHECK: ret i8 1 +} + +define i8 @test2a(i8* %P) { +; CHECK: @test2 + %P2 = getelementptr i8* %P, i32 126 + + ;; FIXME: DSE isn't zapping this dead store. + store i8 1, i8* %P2 ;; Dead, clobbered by memset. + + call void @llvm.memset.i8(i8* %P, i8 2, i8 127, i32 0) + %A = load i8* %P2 + ret i8 %A +; CHECK-NOT: load +; CHECK: ret i8 2 +} + +define void @test3(i8* %P, i8 %X) { +; CHECK: @test3 +; CHECK-NOT: store +; CHECK-NOT: %Y + %Y = add i8 %X, 1 ;; Dead, because the only use (the store) is dead. + + %P2 = getelementptr i8* %P, i32 2 + store i8 %Y, i8* %P2 ;; Not read by lifetime.end, should be removed. +; CHECK: store i8 2, i8* %P2 + call void @llvm.lifetime.end(i64 1, i8* %P) + store i8 2, i8* %P2 +; CHECK-NOT: store + ret void +; CHECK: ret void +} + +define void @test3a(i8* %P, i8 %X) { +; CHECK: @test3a + %Y = add i8 %X, 1 ;; Dead, because the only use (the store) is dead. + + %P2 = getelementptr i8* %P, i32 2 + store i8 %Y, i8* %P2 ;; FIXME: Killed by llvm.lifetime.end, should be zapped. +; CHECK: store i8 %Y, i8* %P2 + call void @llvm.lifetime.end(i64 10, i8* %P) + ret void +; CHECK: ret void +} + +@G1 = external global i32 +@G2 = external global [4000 x i32] + +define i32 @test4(i8* %P) { + %tmp = load i32* @G1 + call void @llvm.memset.i32(i8* bitcast ([4000 x i32]* @G2 to i8*), i8 0, i32 4000, i32 1) + %tmp2 = load i32* @G1 + %sub = sub i32 %tmp2, %tmp + ret i32 %sub +; CHECK: @test4 +; CHECK: load i32* @G +; CHECK: memset.i32 +; CHECK-NOT: load +; CHECK: sub i32 %tmp, %tmp +} + +; Verify that basicaa is handling variable length memcpy, knowing it doesn't +; write to G1. +define i32 @test5(i8* %P, i32 %Len) { + %tmp = load i32* @G1 + call void @llvm.memcpy.i32(i8* bitcast ([4000 x i32]* @G2 to i8*), i8* bitcast (i32* @G1 to i8*), i32 %Len, i32 1) + %tmp2 = load i32* @G1 + %sub = sub i32 %tmp2, %tmp + ret i32 %sub +; CHECK: @test5 +; CHECK: load i32* @G +; CHECK: memcpy.i32 +; CHECK-NOT: load +; CHECK: sub i32 %tmp, %tmp +} + diff --git a/test/Analysis/BasicAA/no-escape-call.ll b/test/Analysis/BasicAA/no-escape-call.ll new file mode 100644 index 0000000..ccabce9 --- /dev/null +++ b/test/Analysis/BasicAA/no-escape-call.ll @@ -0,0 +1,23 @@ +; RUN: opt < %s -basicaa -gvn -instcombine -S | grep {ret i1 true} +; PR2436 +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-darwin8" + +define i1 @foo(i32 %i) nounwind { +entry: + %arr = alloca [10 x i8*] ; <[10 x i8*]*> [#uses=1] + %tmp2 = call i8* @getPtr( ) nounwind ; <i8*> [#uses=2] + %tmp4 = getelementptr [10 x i8*]* %arr, i32 0, i32 %i ; <i8**> [#uses=2] + store i8* %tmp2, i8** %tmp4, align 4 + %tmp10 = getelementptr i8* %tmp2, i32 10 ; <i8*> [#uses=1] + store i8 42, i8* %tmp10, align 1 + %tmp14 = load i8** %tmp4, align 4 ; <i8*> [#uses=1] + %tmp16 = getelementptr i8* %tmp14, i32 10 ; <i8*> [#uses=1] + %tmp17 = load i8* %tmp16, align 1 ; <i8> [#uses=1] + %tmp19 = icmp eq i8 %tmp17, 42 ; <i1> [#uses=1] + ret i1 %tmp19 +} + +declare i8* @getPtr() + +declare void @abort() noreturn nounwind diff --git a/test/Analysis/BasicAA/nocapture.ll b/test/Analysis/BasicAA/nocapture.ll new file mode 100644 index 0000000..7970fbb --- /dev/null +++ b/test/Analysis/BasicAA/nocapture.ll @@ -0,0 +1,14 @@ +; RUN: opt < %s -basicaa -gvn -instcombine -S | grep {ret i32 0} + +declare i32* @test(i32* nocapture) + +define i32 @test2() { + %P = alloca i32 + %Q = call i32* @test(i32* %P) + %a = load i32* %P + store i32 4, i32* %Q ;; cannot clobber P since it is nocapture. + %b = load i32* %P + %c = sub i32 %a, %b + ret i32 %c +} + diff --git a/test/Analysis/BasicAA/phi-aa.ll b/test/Analysis/BasicAA/phi-aa.ll new file mode 100644 index 0000000..0288960 --- /dev/null +++ b/test/Analysis/BasicAA/phi-aa.ll @@ -0,0 +1,29 @@ +; RUN: opt < %s -basicaa -aa-eval -print-all-alias-modref-info -disable-output |& grep {NoAlias:.*%P,.*@Z} +; rdar://7282591 + +@X = common global i32 0 +@Y = common global i32 0 +@Z = common global i32 0 + +define void @foo(i32 %cond) nounwind ssp { +entry: + %"alloca point" = bitcast i32 0 to i32 + %tmp = icmp ne i32 %cond, 0 + br i1 %tmp, label %bb, label %bb1 + +bb: + br label %bb2 + +bb1: + br label %bb2 + +bb2: + %P = phi i32* [ @X, %bb ], [ @Y, %bb1 ] + %tmp1 = load i32* @Z, align 4 + store i32 123, i32* %P, align 4 + %tmp2 = load i32* @Z, align 4 + br label %return + +return: + ret void +} diff --git a/test/Analysis/BasicAA/phi-and-select.ll b/test/Analysis/BasicAA/phi-and-select.ll new file mode 100644 index 0000000..c69e824 --- /dev/null +++ b/test/Analysis/BasicAA/phi-and-select.ll @@ -0,0 +1,73 @@ +; RUN: opt < %s -aa-eval -print-all-alias-modref-info -disable-output \ +; RUN: |& grep {NoAlias: double\\* \[%\]a, double\\* \[%\]b\$} | count 4 + +; BasicAA should detect NoAliases in PHIs and Selects. + +; Two PHIs in the same block. +define void @foo(i1 %m, double* noalias %x, double* noalias %y) { +entry: + br i1 %m, label %true, label %false + +true: + br label %exit + +false: + br label %exit + +exit: + %a = phi double* [ %x, %true ], [ %y, %false ] + %b = phi double* [ %x, %false ], [ %y, %true ] + volatile store double 0.0, double* %a + volatile store double 1.0, double* %b + ret void +} + +; Two selects with the same condition. +define void @bar(i1 %m, double* noalias %x, double* noalias %y) { +entry: + %a = select i1 %m, double* %x, double* %y + %b = select i1 %m, double* %y, double* %x + volatile store double 0.000000e+00, double* %a + volatile store double 1.000000e+00, double* %b + ret void +} + +; Two PHIs with disjoint sets of inputs. +define void @qux(i1 %m, double* noalias %x, double* noalias %y, + i1 %n, double* noalias %v, double* noalias %w) { +entry: + br i1 %m, label %true, label %false + +true: + br label %exit + +false: + br label %exit + +exit: + %a = phi double* [ %x, %true ], [ %y, %false ] + br i1 %n, label %ntrue, label %nfalse + +ntrue: + br label %nexit + +nfalse: + br label %nexit + +nexit: + %b = phi double* [ %v, %ntrue ], [ %w, %nfalse ] + volatile store double 0.0, double* %a + volatile store double 1.0, double* %b + ret void +} + +; Two selects with disjoint sets of arms. +define void @fin(i1 %m, double* noalias %x, double* noalias %y, + i1 %n, double* noalias %v, double* noalias %w) { +entry: + %a = select i1 %m, double* %x, double* %y + %b = select i1 %n, double* %v, double* %w + volatile store double 0.000000e+00, double* %a + volatile store double 1.000000e+00, double* %b + ret void +} diff --git a/test/Analysis/BasicAA/pure-const-dce.ll b/test/Analysis/BasicAA/pure-const-dce.ll new file mode 100644 index 0000000..54e6e79 --- /dev/null +++ b/test/Analysis/BasicAA/pure-const-dce.ll @@ -0,0 +1,33 @@ +; RUN: opt < %s -basicaa -gvn -S | grep TestConst | count 2 +; RUN: opt < %s -basicaa -gvn -S | grep TestPure | count 3 +; RUN: opt < %s -basicaa -gvn -S | grep TestNone | count 4 +@g = global i32 0 ; <i32*> [#uses=1] + +define i32 @test() { +entry: + %tmp0 = call i32 @TestConst( i32 5 ) readnone ; <i32> [#uses=1] + %tmp1 = call i32 @TestPure( i32 6 ) readonly ; <i32> [#uses=1] + %tmp2 = call i32 @TestNone( i32 7 ) ; <i32> [#uses=1] + store i32 1, i32* @g + %tmp3 = call i32 @TestConst( i32 5 ) readnone ; <i32> [#uses=1] + %tmp4 = call i32 @TestConst( i32 5 ) readnone ; <i32> [#uses=1] + %tmp5 = call i32 @TestPure( i32 6 ) readonly ; <i32> [#uses=1] + %tmp6 = call i32 @TestPure( i32 6 ) readonly ; <i32> [#uses=1] + %tmp7 = call i32 @TestNone( i32 7 ) ; <i32> [#uses=1] + %tmp8 = call i32 @TestNone( i32 7 ) ; <i32> [#uses=1] + %sum0 = add i32 %tmp0, %tmp1 ; <i32> [#uses=1] + %sum1 = add i32 %sum0, %tmp2 ; <i32> [#uses=1] + %sum2 = add i32 %sum1, %tmp3 ; <i32> [#uses=1] + %sum3 = add i32 %sum2, %tmp4 ; <i32> [#uses=1] + %sum4 = add i32 %sum3, %tmp5 ; <i32> [#uses=1] + %sum5 = add i32 %sum4, %tmp6 ; <i32> [#uses=1] + %sum6 = add i32 %sum5, %tmp7 ; <i32> [#uses=1] + %sum7 = add i32 %sum6, %tmp8 ; <i32> [#uses=1] + ret i32 %sum7 +} + +declare i32 @TestConst(i32) readnone + +declare i32 @TestPure(i32) readonly + +declare i32 @TestNone(i32) diff --git a/test/Analysis/BasicAA/store-promote.ll b/test/Analysis/BasicAA/store-promote.ll new file mode 100644 index 0000000..33d0f3a --- /dev/null +++ b/test/Analysis/BasicAA/store-promote.ll @@ -0,0 +1,54 @@ +; Test that LICM uses basicaa to do alias analysis, which is capable of +; disambiguating some obvious cases. If LICM is able to disambiguate the +; two pointers, then the load should be hoisted, and the store sunk. + +; RUN: opt < %s -basicaa -licm -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" + +@A = global i32 7 ; <i32*> [#uses=3] +@B = global i32 8 ; <i32*> [#uses=2] +@C = global [2 x i32] [ i32 4, i32 8 ] ; <[2 x i32]*> [#uses=2] + +define i32 @test1(i1 %c) { + %Atmp = load i32* @A ; <i32> [#uses=2] + br label %Loop + +Loop: ; preds = %Loop, %0 + %ToRemove = load i32* @A ; <i32> [#uses=1] + store i32 %Atmp, i32* @B + br i1 %c, label %Out, label %Loop + +Out: ; preds = %Loop + %X = sub i32 %ToRemove, %Atmp ; <i32> [#uses=1] + ret i32 %X + +; The Loop block should be empty after the load/store are promoted. +; CHECK: @test1 +; CHECK: load i32* @B +; CHECK: Loop: +; CHECK-NEXT: br i1 %c, label %Out, label %Loop +; CHECK: Out: +; CHECK: store i32 %Atmp, i32* @B +} + +define i32 @test2(i1 %c) { + br label %Loop + +Loop: ; preds = %Loop, %0 + %AVal = load i32* @A ; <i32> [#uses=2] + %C0 = getelementptr [2 x i32]* @C, i64 0, i64 0 ; <i32*> [#uses=1] + store i32 %AVal, i32* %C0 + %BVal = load i32* @B ; <i32> [#uses=2] + %C1 = getelementptr [2 x i32]* @C, i64 0, i64 1 ; <i32*> [#uses=1] + store i32 %BVal, i32* %C1 + br i1 %c, label %Out, label %Loop + +Out: ; preds = %Loop + %X = sub i32 %AVal, %BVal ; <i32> [#uses=1] + ret i32 %X +; The Loop block should be empty after the load/store are promoted. +; CHECK: @test2 +; CHECK: Loop: +; CHECK-NEXT: br i1 %c, label %Out, label %Loop +} + diff --git a/test/Analysis/BasicAA/tailcall-modref.ll b/test/Analysis/BasicAA/tailcall-modref.ll new file mode 100644 index 0000000..f7d6c57 --- /dev/null +++ b/test/Analysis/BasicAA/tailcall-modref.ll @@ -0,0 +1,16 @@ +; RUN: opt < %s -basicaa -gvn -instcombine |\ +; RUN: llvm-dis | grep {ret i32 0} + +declare void @foo(i32*) + +declare void @bar() + +define i32 @test() { + %A = alloca i32 ; <i32*> [#uses=3] + call void @foo( i32* %A ) + %X = load i32* %A ; <i32> [#uses=1] + tail call void @bar( ) + %Y = load i32* %A ; <i32> [#uses=1] + %Z = sub i32 %X, %Y ; <i32> [#uses=1] + ret i32 %Z +} diff --git a/test/Analysis/CallGraph/2008-09-09-DirectCall.ll b/test/Analysis/CallGraph/2008-09-09-DirectCall.ll new file mode 100644 index 0000000..6e34209 --- /dev/null +++ b/test/Analysis/CallGraph/2008-09-09-DirectCall.ll @@ -0,0 +1,13 @@ +; RUN: opt < %s -print-callgraph -disable-output |& \ +; RUN: grep {Calls function 'callee'} | count 2 + +define internal void @callee(...) { +entry: + unreachable +} + +define void @caller() { +entry: + call void (...)* @callee( void (...)* @callee ) + unreachable +} diff --git a/test/Analysis/CallGraph/2008-09-09-UsedByGlobal.ll b/test/Analysis/CallGraph/2008-09-09-UsedByGlobal.ll new file mode 100644 index 0000000..12849b7 --- /dev/null +++ b/test/Analysis/CallGraph/2008-09-09-UsedByGlobal.ll @@ -0,0 +1,8 @@ +; RUN: opt < %s -print-callgraph -disable-output |& \ +; RUN: grep {Calls function} + +@a = global void ()* @f ; <void ()**> [#uses=0] + +define internal void @f() { + unreachable +} diff --git a/test/Analysis/CallGraph/dg.exp b/test/Analysis/CallGraph/dg.exp new file mode 100644 index 0000000..f200589 --- /dev/null +++ b/test/Analysis/CallGraph/dg.exp @@ -0,0 +1,3 @@ +load_lib llvm.exp + +RunLLVMTests [lsort [glob -nocomplain $srcdir/$subdir/*.{ll,c,cpp}]] diff --git a/test/Analysis/Dominators/2006-10-02-BreakCritEdges.ll b/test/Analysis/Dominators/2006-10-02-BreakCritEdges.ll new file mode 100644 index 0000000..e31f416 --- /dev/null +++ b/test/Analysis/Dominators/2006-10-02-BreakCritEdges.ll @@ -0,0 +1,20 @@ +; RUN: opt < %s -domtree -break-crit-edges -analyze \ +; RUN: -domtree | grep {3.*%brtrue } +; PR932 + +declare void @use1(i32) + +define void @f(i32 %i, i1 %c) { +entry: + %A = icmp eq i32 %i, 0 ; <i1> [#uses=1] + br i1 %A, label %brtrue, label %brfalse + +brtrue: ; preds = %brtrue, %entry + %B = phi i1 [ true, %brtrue ], [ false, %entry ] ; <i1> [#uses=1] + call void @use1( i32 %i ) + br i1 %B, label %brtrue, label %brfalse + +brfalse: ; preds = %brtrue, %entry + call void @use1( i32 %i ) + ret void +} diff --git a/test/Analysis/Dominators/2007-01-14-BreakCritEdges.ll b/test/Analysis/Dominators/2007-01-14-BreakCritEdges.ll new file mode 100644 index 0000000..96dc739 --- /dev/null +++ b/test/Analysis/Dominators/2007-01-14-BreakCritEdges.ll @@ -0,0 +1,187 @@ +; RUN: opt < %s -domtree -break-crit-edges -domtree -disable-output +; PR1110 + + %struct.OggVorbis_File = type { i8*, i32, i64, i64, %struct.ogg_sync_state, i32, i64*, i64*, i32*, i64*, %struct.vorbis_info*, %struct.vorbis_comment*, i64, i32, i32, i32, double, double, %struct.ogg_stream_state, %struct.vorbis_dsp_state, %struct.vorbis_block, %struct.ov_callbacks } + %struct.alloc_chain = type { i8*, %struct.alloc_chain* } + %struct.ogg_stream_state = type { i8*, i32, i32, i32, i32*, i64*, i32, i32, i32, i32, [282 x i8], i32, i32, i32, i32, i32, i64, i64 } + %struct.ogg_sync_state = type { i8*, i32, i32, i32, i32, i32, i32 } + %struct.oggpack_buffer = type { i32, i32, i8*, i8*, i32 } + %struct.ov_callbacks = type { i32 (i8*, i32, i32, i8*)*, i32 (i8*, i64, i32)*, i32 (i8*)*, i32 (i8*)* } + %struct.vorbis_block = type { float**, %struct.oggpack_buffer, i32, i32, i32, i32, i32, i32, i64, i64, %struct.vorbis_dsp_state*, i8*, i32, i32, i32, %struct.alloc_chain*, i32, i32, i32, i32, i8* } + %struct.vorbis_comment = type { i8**, i32*, i32, i8* } + %struct.vorbis_dsp_state = type { i32, %struct.vorbis_info*, float**, float**, i32, i32, i32, i32, i32, i32, i32, i32, i32, i64, i64, i64, i64, i64, i64, i8* } + %struct.vorbis_info = type { i32, i32, i32, i32, i32, i32, i32, i8* } + + +define void @ov_read() { +entry: + br i1 false, label %bb, label %return + +bb: ; preds = %cond_next22, %entry + br i1 false, label %cond_true8, label %cond_next15 + +cond_true8: ; preds = %bb + br i1 false, label %cond_next15, label %bb29 + +cond_next15: ; preds = %cond_true8, %bb + br i1 false, label %return, label %cond_next22 + +cond_next22: ; preds = %cond_next15 + br i1 false, label %bb, label %return + +bb29: ; preds = %cond_true8 + br i1 false, label %cond_true32, label %return + +cond_true32: ; preds = %bb29 + br i1 false, label %cond_false37.i, label %cond_true.i11 + +cond_true.i11: ; preds = %cond_true32 + br i1 false, label %cond_true8.i, label %ov_info.exit + +cond_true8.i: ; preds = %cond_true.i11 + br i1 false, label %cond_true44, label %cond_next48 + +cond_false37.i: ; preds = %cond_true32 + br label %ov_info.exit + +ov_info.exit: ; preds = %cond_false37.i, %cond_true.i11 + br i1 false, label %cond_true44, label %cond_next48 + +cond_true44: ; preds = %ov_info.exit, %cond_true8.i + br label %cond_next48 + +cond_next48: ; preds = %cond_true44, %ov_info.exit, %cond_true8.i + br i1 false, label %cond_next53, label %return + +cond_next53: ; preds = %cond_next48 + br i1 false, label %cond_true56, label %cond_false97 + +cond_true56: ; preds = %cond_next53 + br i1 false, label %bb85, label %cond_next304 + +bb63: ; preds = %bb85 + br i1 false, label %cond_next78, label %cond_false73 + +cond_false73: ; preds = %bb63 + br i1 false, label %cond_true76, label %cond_next78 + +cond_true76: ; preds = %cond_false73 + br label %cond_next78 + +cond_next78: ; preds = %cond_true76, %cond_false73, %bb63 + br label %bb85 + +bb85: ; preds = %bb89, %cond_next78, %cond_true56 + br i1 false, label %bb63, label %bb89 + +bb89: ; preds = %bb85 + br i1 false, label %bb85, label %cond_next304 + +cond_false97: ; preds = %cond_next53 + br i1 false, label %cond_true108, label %bb248 + +cond_true108: ; preds = %cond_false97 + br i1 false, label %bb196, label %bb149 + +bb112: ; preds = %bb149, %bb146 + br i1 false, label %bb119, label %bb146 + +bb119: ; preds = %cond_next134, %bb112 + br i1 false, label %cond_next134, label %cond_false129 + +cond_false129: ; preds = %bb119 + br i1 false, label %cond_true132, label %cond_next134 + +cond_true132: ; preds = %cond_false129 + br label %cond_next134 + +cond_next134: ; preds = %cond_true132, %cond_false129, %bb119 + br i1 false, label %bb119, label %bb146 + +bb146: ; preds = %cond_next134, %bb112 + br i1 false, label %bb112, label %cond_next304 + +bb149: ; preds = %cond_true108 + br i1 false, label %bb112, label %cond_next304 + +bb155: ; preds = %bb196, %bb193 + br i1 false, label %bb165, label %bb193 + +bb165: ; preds = %cond_next180, %bb155 + br i1 false, label %cond_next180, label %cond_false175 + +cond_false175: ; preds = %bb165 + br i1 false, label %cond_true178, label %cond_next180 + +cond_true178: ; preds = %cond_false175 + br label %cond_next180 + +cond_next180: ; preds = %cond_true178, %cond_false175, %bb165 + br i1 false, label %bb165, label %bb193 + +bb193: ; preds = %cond_next180, %bb155 + br i1 false, label %bb155, label %cond_next304 + +bb196: ; preds = %cond_true108 + br i1 false, label %bb155, label %cond_next304 + +bb207: ; preds = %bb241 + br i1 false, label %cond_next225, label %cond_false220 + +cond_false220: ; preds = %bb207 + br i1 false, label %cond_true223, label %cond_next225 + +cond_true223: ; preds = %cond_false220 + br label %cond_next225 + +cond_next225: ; preds = %cond_true223, %cond_false220, %bb207 + br label %bb241 + +bb241: ; preds = %bb248, %bb245, %cond_next225 + br i1 false, label %bb207, label %bb245 + +bb245: ; preds = %bb241 + br i1 false, label %bb241, label %cond_next304 + +bb248: ; preds = %cond_false97 + br i1 false, label %bb241, label %cond_next304 + +bb256: ; preds = %bb290 + br i1 false, label %cond_next274, label %cond_false269 + +cond_false269: ; preds = %bb256 + br i1 false, label %cond_true272, label %cond_next274 + +cond_true272: ; preds = %cond_false269 + br label %cond_next274 + +cond_next274: ; preds = %cond_true272, %cond_false269, %bb256 + br label %bb290 + +bb290: ; preds = %bb294, %cond_next274 + br i1 false, label %bb256, label %bb294 + +bb294: ; preds = %bb290 + br i1 false, label %bb290, label %cond_next304 + +cond_next304: ; preds = %bb294, %bb248, %bb245, %bb196, %bb193, %bb149, %bb146, %bb89, %cond_true56 + br i1 false, label %cond_next11.i, label %cond_true.i + +cond_true.i: ; preds = %cond_next304 + br i1 false, label %vorbis_synthesis_read.exit, label %cond_next11.i + +cond_next11.i: ; preds = %cond_true.i, %cond_next304 + br label %vorbis_synthesis_read.exit + +vorbis_synthesis_read.exit: ; preds = %cond_next11.i, %cond_true.i + br i1 false, label %cond_next321, label %cond_true316 + +cond_true316: ; preds = %vorbis_synthesis_read.exit + ret void + +cond_next321: ; preds = %vorbis_synthesis_read.exit + ret void + +return: ; preds = %cond_next48, %bb29, %cond_next22, %cond_next15, %entry + ret void +} diff --git a/test/Analysis/Dominators/2007-07-11-SplitBlock.ll b/test/Analysis/Dominators/2007-07-11-SplitBlock.ll new file mode 100644 index 0000000..52fdd2b --- /dev/null +++ b/test/Analysis/Dominators/2007-07-11-SplitBlock.ll @@ -0,0 +1,21 @@ +; RUN: opt < %s -loop-rotate -loop-unswitch -disable-output + +define i32 @stringSearch_Clib(i32 %count) { +entry: + br i1 false, label %bb36, label %bb44 + +cond_true20: ; preds = %bb36 + %tmp33 = add i32 0, 0 ; <i32> [#uses=1] + br label %bb36 + +bb36: ; preds = %cond_true20, %entry + %c.2 = phi i32 [ %tmp33, %cond_true20 ], [ 0, %entry ] ; <i32> [#uses=1] + br i1 false, label %cond_true20, label %bb41 + +bb41: ; preds = %bb36 + %c.2.lcssa = phi i32 [ %c.2, %bb36 ] ; <i32> [#uses=0] + ret i32 0 + +bb44: ; preds = %entry + ret i32 0 +} diff --git a/test/Analysis/Dominators/2007-07-12-SplitBlock.ll b/test/Analysis/Dominators/2007-07-12-SplitBlock.ll new file mode 100644 index 0000000..b46f0c7 --- /dev/null +++ b/test/Analysis/Dominators/2007-07-12-SplitBlock.ll @@ -0,0 +1,13 @@ +; RUN: opt < %s -loop-rotate -licm -loop-unswitch -disable-output + +define i32 @main(i32 %argc, i8** %argv) { +entry: + br label %bb7 + +bb7: ; preds = %bb7, %entry + %tmp54 = icmp slt i32 0, 2000000 ; <i1> [#uses=1] + br i1 %tmp54, label %bb7, label %bb56 + +bb56: ; preds = %bb7 + ret i32 0 +} diff --git a/test/Analysis/Dominators/dg.exp b/test/Analysis/Dominators/dg.exp new file mode 100644 index 0000000..f200589 --- /dev/null +++ b/test/Analysis/Dominators/dg.exp @@ -0,0 +1,3 @@ +load_lib llvm.exp + +RunLLVMTests [lsort [glob -nocomplain $srcdir/$subdir/*.{ll,c,cpp}]] diff --git a/test/Analysis/GlobalsModRef/2008-09-03-ReadGlobals.ll b/test/Analysis/GlobalsModRef/2008-09-03-ReadGlobals.ll new file mode 100644 index 0000000..17ace8a --- /dev/null +++ b/test/Analysis/GlobalsModRef/2008-09-03-ReadGlobals.ll @@ -0,0 +1,18 @@ +; RUN: opt < %s -globalsmodref-aa -gvn -S | grep call | count 2 + +@g = internal global i32 0 ; <i32*> [#uses=2] + +define i32 @r() { + %tmp = load i32* @g ; <i32> [#uses=1] + ret i32 %tmp +} + +define i32 @f() { +entry: + %tmp = call i32 @e( ) ; <i32> [#uses=1] + store i32 %tmp, i32* @g + %tmp2 = call i32 @e( ) ; <i32> [#uses=1] + ret i32 %tmp2 +} + +declare i32 @e() readonly ; might call @r diff --git a/test/Analysis/GlobalsModRef/aliastest.ll b/test/Analysis/GlobalsModRef/aliastest.ll new file mode 100644 index 0000000..3e5d119 --- /dev/null +++ b/test/Analysis/GlobalsModRef/aliastest.ll @@ -0,0 +1,9 @@ +; RUN: opt < %s -globalsmodref-aa -gvn -S | not grep load +@X = internal global i32 4 ; <i32*> [#uses=1] + +define i32 @test(i32* %P) { + store i32 7, i32* %P + store i32 12, i32* @X + %V = load i32* %P ; <i32> [#uses=1] + ret i32 %V +} diff --git a/test/Analysis/GlobalsModRef/chaining-analysis.ll b/test/Analysis/GlobalsModRef/chaining-analysis.ll new file mode 100644 index 0000000..b1d4593 --- /dev/null +++ b/test/Analysis/GlobalsModRef/chaining-analysis.ll @@ -0,0 +1,20 @@ +; RUN: opt < %s -globalsmodref-aa -gvn -S | not grep load + +; This test requires the use of previous analyses to determine that +; doesnotmodX does not modify X (because 'sin' doesn't). + +@X = internal global i32 4 ; <i32*> [#uses=2] + +declare double @sin(double) readnone + +define i32 @test(i32* %P) { + store i32 12, i32* @X + call double @doesnotmodX( double 1.000000e+00 ) ; <double>:1 [#uses=0] + %V = load i32* @X ; <i32> [#uses=1] + ret i32 %V +} + +define double @doesnotmodX(double %V) { + %V2 = call double @sin( double %V ) readnone ; <double> [#uses=1] + ret double %V2 +} diff --git a/test/Analysis/GlobalsModRef/dg.exp b/test/Analysis/GlobalsModRef/dg.exp new file mode 100644 index 0000000..f200589 --- /dev/null +++ b/test/Analysis/GlobalsModRef/dg.exp @@ -0,0 +1,3 @@ +load_lib llvm.exp + +RunLLVMTests [lsort [glob -nocomplain $srcdir/$subdir/*.{ll,c,cpp}]] diff --git a/test/Analysis/GlobalsModRef/indirect-global.ll b/test/Analysis/GlobalsModRef/indirect-global.ll new file mode 100644 index 0000000..4074909 --- /dev/null +++ b/test/Analysis/GlobalsModRef/indirect-global.ll @@ -0,0 +1,20 @@ +; RUN: opt < %s -globalsmodref-aa -gvn -instcombine -S | \ +; RUN: grep {ret i32 0} + +@G = internal global i32* null ; <i32**> [#uses=3] + +define void @test() { + %A = malloc i32 ; <i32*> [#uses=1] + store i32* %A, i32** @G + ret void +} + +define i32 @test1(i32* %P) { + %g1 = load i32** @G ; <i32*> [#uses=2] + %h1 = load i32* %g1 ; <i32> [#uses=1] + store i32 123, i32* %P + %g2 = load i32** @G ; <i32*> [#uses=0] + %h2 = load i32* %g1 ; <i32> [#uses=1] + %X = sub i32 %h1, %h2 ; <i32> [#uses=1] + ret i32 %X +} diff --git a/test/Analysis/GlobalsModRef/modreftest.ll b/test/Analysis/GlobalsModRef/modreftest.ll new file mode 100644 index 0000000..257c0ee --- /dev/null +++ b/test/Analysis/GlobalsModRef/modreftest.ll @@ -0,0 +1,13 @@ +; RUN: opt < %s -globalsmodref-aa -gvn -S | not grep load +@X = internal global i32 4 ; <i32*> [#uses=2] + +define i32 @test(i32* %P) { + store i32 12, i32* @X + call void @doesnotmodX( ) + %V = load i32* @X ; <i32> [#uses=1] + ret i32 %V +} + +define void @doesnotmodX() { + ret void +} diff --git a/test/Analysis/GlobalsModRef/purecse.ll b/test/Analysis/GlobalsModRef/purecse.ll new file mode 100644 index 0000000..994aff8 --- /dev/null +++ b/test/Analysis/GlobalsModRef/purecse.ll @@ -0,0 +1,23 @@ +; Test that pure functions are cse'd away +; RUN: opt < %s -globalsmodref-aa -gvn -instcombine | \ +; RUN: llvm-dis | not grep sub + +define i32 @pure(i32 %X) { + %Y = add i32 %X, 1 ; <i32> [#uses=1] + ret i32 %Y +} + +define i32 @test1(i32 %X) { + %A = call i32 @pure( i32 %X ) ; <i32> [#uses=1] + %B = call i32 @pure( i32 %X ) ; <i32> [#uses=1] + %C = sub i32 %A, %B ; <i32> [#uses=1] + ret i32 %C +} + +define i32 @test2(i32 %X, i32* %P) { + %A = call i32 @pure( i32 %X ) ; <i32> [#uses=1] + store i32 %X, i32* %P ;; Does not invalidate 'pure' call. + %B = call i32 @pure( i32 %X ) ; <i32> [#uses=1] + %C = sub i32 %A, %B ; <i32> [#uses=1] + ret i32 %C +} diff --git a/test/Analysis/LoopDependenceAnalysis/alias.ll b/test/Analysis/LoopDependenceAnalysis/alias.ll new file mode 100644 index 0000000..97be3fd --- /dev/null +++ b/test/Analysis/LoopDependenceAnalysis/alias.ll @@ -0,0 +1,44 @@ +; RUN: opt < %s -analyze -lda | FileCheck %s + +;; x[5] = x[6] // with x being a pointer passed as argument + +define void @f1(i32* nocapture %xptr) nounwind { +entry: + %x.ld.addr = getelementptr i32* %xptr, i64 6 + %x.st.addr = getelementptr i32* %xptr, i64 5 + br label %for.body + +for.body: + %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ] + %x = load i32* %x.ld.addr + store i32 %x, i32* %x.st.addr +; CHECK: 0,1: dep + %i.next = add i64 %i, 1 + %exitcond = icmp eq i64 %i.next, 256 + br i1 %exitcond, label %for.end, label %for.body + +for.end: + ret void +} + +;; x[5] = x[6] // with x being an array on the stack + +define void @foo(...) nounwind { +entry: + %xptr = alloca [256 x i32], align 4 + %x.ld.addr = getelementptr [256 x i32]* %xptr, i64 0, i64 6 + %x.st.addr = getelementptr [256 x i32]* %xptr, i64 0, i64 5 + br label %for.body + +for.body: + %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ] + %x = load i32* %x.ld.addr + store i32 %x, i32* %x.st.addr +; CHECK: 0,1: ind + %i.next = add i64 %i, 1 + %exitcond = icmp eq i64 %i.next, 256 + br i1 %exitcond, label %for.end, label %for.body + +for.end: + ret void +} diff --git a/test/Analysis/LoopDependenceAnalysis/dg.exp b/test/Analysis/LoopDependenceAnalysis/dg.exp new file mode 100644 index 0000000..f200589 --- /dev/null +++ b/test/Analysis/LoopDependenceAnalysis/dg.exp @@ -0,0 +1,3 @@ +load_lib llvm.exp + +RunLLVMTests [lsort [glob -nocomplain $srcdir/$subdir/*.{ll,c,cpp}]] diff --git a/test/Analysis/LoopDependenceAnalysis/siv-strong.ll b/test/Analysis/LoopDependenceAnalysis/siv-strong.ll new file mode 100644 index 0000000..36ac153 --- /dev/null +++ b/test/Analysis/LoopDependenceAnalysis/siv-strong.ll @@ -0,0 +1,110 @@ +; RUN: opt < %s -analyze -lda | FileCheck %s + +@x = common global [256 x i32] zeroinitializer, align 4 +@y = common global [256 x i32] zeroinitializer, align 4 + +;; for (i = 0; i < 256; i++) +;; x[i] = x[i] + y[i] + +define void @f1(...) nounwind { +entry: + br label %for.body + +for.body: + %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ] + %y.addr = getelementptr [256 x i32]* @y, i64 0, i64 %i + %x.addr = getelementptr [256 x i32]* @x, i64 0, i64 %i + %x = load i32* %x.addr ; 0 + %y = load i32* %y.addr ; 1 + %r = add i32 %y, %x + store i32 %r, i32* %x.addr ; 2 +; CHECK: 0,2: dep +; CHECK: 1,2: ind + %i.next = add i64 %i, 1 + %exitcond = icmp eq i64 %i.next, 256 + br i1 %exitcond, label %for.end, label %for.body + +for.end: + ret void +} + +;; for (i = 0; i < 256; i++) +;; x[i+1] = x[i] + y[i] + +define void @f2(...) nounwind { +entry: + br label %for.body + +for.body: + %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ] + %y.ld.addr = getelementptr [256 x i32]* @y, i64 0, i64 %i + %x.ld.addr = getelementptr [256 x i32]* @x, i64 0, i64 %i + %i.next = add i64 %i, 1 + %x.st.addr = getelementptr [256 x i32]* @x, i64 0, i64 %i.next + %x = load i32* %x.ld.addr ; 0 + %y = load i32* %y.ld.addr ; 1 + %r = add i32 %y, %x + store i32 %r, i32* %x.st.addr ; 2 +; CHECK: 0,2: dep +; CHECK: 1,2: ind + %exitcond = icmp eq i64 %i.next, 256 + br i1 %exitcond, label %for.end, label %for.body + +for.end: + ret void +} + +;; for (i = 0; i < 10; i++) +;; x[i+20] = x[i] + y[i] + +define void @f3(...) nounwind { +entry: + br label %for.body + +for.body: + %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ] + %y.ld.addr = getelementptr [256 x i32]* @y, i64 0, i64 %i + %x.ld.addr = getelementptr [256 x i32]* @x, i64 0, i64 %i + %i.20 = add i64 %i, 20 + %x.st.addr = getelementptr [256 x i32]* @x, i64 0, i64 %i.20 + %x = load i32* %x.ld.addr ; 0 + %y = load i32* %y.ld.addr ; 1 + %r = add i32 %y, %x + store i32 %r, i32* %x.st.addr ; 2 +; CHECK: 0,2: dep +; CHECK: 1,2: ind + %i.next = add i64 %i, 1 + %exitcond = icmp eq i64 %i.next, 10 + br i1 %exitcond, label %for.end, label %for.body + +for.end: + ret void +} + +;; for (i = 0; i < 10; i++) +;; x[10*i+1] = x[10*i] + y[i] + +define void @f4(...) nounwind { +entry: + br label %for.body + +for.body: + %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ] + %i.10 = mul i64 %i, 10 + %y.ld.addr = getelementptr [256 x i32]* @y, i64 0, i64 %i.10 + %x.ld.addr = getelementptr [256 x i32]* @x, i64 0, i64 %i.10 + %i.10.1 = add i64 %i.10, 1 + %x.st.addr = getelementptr [256 x i32]* @x, i64 0, i64 %i.10.1 + %x = load i32* %x.ld.addr ; 0 + %y = load i32* %y.ld.addr ; 1 + %r = add i32 %y, %x + store i32 %r, i32* %x.st.addr ; 2 +; CHECK: 0,2: dep +; CHECK: 1,2: ind + %i.next = add i64 %i, 1 + %exitcond = icmp eq i64 %i.next, 10 + br i1 %exitcond, label %for.end, label %for.body + +for.end: + ret void +} diff --git a/test/Analysis/LoopDependenceAnalysis/siv-weak-crossing.ll b/test/Analysis/LoopDependenceAnalysis/siv-weak-crossing.ll new file mode 100644 index 0000000..a7f9bda --- /dev/null +++ b/test/Analysis/LoopDependenceAnalysis/siv-weak-crossing.ll @@ -0,0 +1,118 @@ +; RUN: opt < %s -analyze -lda | FileCheck %s + +@x = common global [256 x i32] zeroinitializer, align 4 +@y = common global [256 x i32] zeroinitializer, align 4 + +;; for (i = 0; i < 256; i++) +;; x[i] = x[255 - i] + y[i] + +define void @f1(...) nounwind { +entry: + br label %for.body + +for.body: + %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ] + %i.255 = sub i64 255, %i + %y.ld.addr = getelementptr [256 x i32]* @y, i64 0, i64 %i + %x.ld.addr = getelementptr [256 x i32]* @x, i64 0, i64 %i.255 + %x.st.addr = getelementptr [256 x i32]* @x, i64 0, i64 %i + %x = load i32* %x.ld.addr ; 0 + %y = load i32* %y.ld.addr ; 1 + %r = add i32 %y, %x + store i32 %r, i32* %x.st.addr ; 2 +; CHECK: 0,2: dep +; CHECK: 1,2: ind + %i.next = add i64 %i, 1 + %exitcond = icmp eq i64 %i.next, 256 + br i1 %exitcond, label %for.end, label %for.body + +for.end: + ret void +} + +;; for (i = 0; i < 100; i++) +;; x[i] = x[255 - i] + y[i] + +define void @f2(...) nounwind { +entry: + br label %for.body + +for.body: + %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ] + %i.255 = sub i64 255, %i + %y.ld.addr = getelementptr [256 x i32]* @y, i64 0, i64 %i + %x.ld.addr = getelementptr [256 x i32]* @x, i64 0, i64 %i.255 + %x.st.addr = getelementptr [256 x i32]* @x, i64 0, i64 %i + %x = load i32* %x.ld.addr ; 0 + %y = load i32* %y.ld.addr ; 1 + %r = add i32 %y, %x + store i32 %r, i32* %x.st.addr ; 2 +; CHECK: 0,2: dep +; CHECK: 1,2: ind + %i.next = add i64 %i, 1 + %exitcond = icmp eq i64 %i.next, 100 + br i1 %exitcond, label %for.end, label %for.body + +for.end: + ret void +} + +;; // the first iteration (i=0) leads to an out-of-bounds access of x. as the +;; // result of this access is undefined, _any_ dependence result is safe. +;; for (i = 0; i < 256; i++) +;; x[i] = x[256 - i] + y[i] + +define void @f3(...) nounwind { +entry: + br label %for.body + +for.body: + %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ] + %i.256 = sub i64 0, %i + %y.ld.addr = getelementptr [256 x i32]* @y, i64 0, i64 %i + %x.ld.addr = getelementptr [256 x i32]* @x, i64 1, i64 %i.256 + %x.st.addr = getelementptr [256 x i32]* @x, i64 0, i64 %i + %x = load i32* %x.ld.addr ; 0 + %y = load i32* %y.ld.addr ; 1 + %r = add i32 %y, %x + store i32 %r, i32* %x.st.addr ; 2 +; CHECK: 0,2: dep +; CHECK: 1,2: + %i.next = add i64 %i, 1 + %exitcond = icmp eq i64 %i.next, 256 + br i1 %exitcond, label %for.end, label %for.body + +for.end: + ret void +} + +;; // slightly contrived but valid IR for the following loop, where all +;; // accesses in all iterations are within bounds. while this example's first +;; // (ZIV-)subscript is (0, 1), accesses are dependent. +;; for (i = 1; i < 256; i++) +;; x[i] = x[256 - i] + y[i] + +define void @f4(...) nounwind { +entry: + br label %for.body + +for.body: + %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ] + %i.1 = add i64 1, %i + %i.256 = sub i64 -1, %i + %y.ld.addr = getelementptr [256 x i32]* @y, i64 0, i64 %i.1 + %x.ld.addr = getelementptr [256 x i32]* @x, i64 1, i64 %i.256 + %x.st.addr = getelementptr [256 x i32]* @x, i64 0, i64 %i.1 + %x = load i32* %x.ld.addr ; 0 + %y = load i32* %y.ld.addr ; 1 + %r = add i32 %y, %x + store i32 %r, i32* %x.st.addr ; 2 +; CHECK: 0,2: dep +; CHECK: 1,2: ind + %i.next = add i64 %i, 1 + %exitcond = icmp eq i64 %i.next, 256 + br i1 %exitcond, label %for.end, label %for.body + +for.end: + ret void +} diff --git a/test/Analysis/LoopDependenceAnalysis/siv-weak-zero.ll b/test/Analysis/LoopDependenceAnalysis/siv-weak-zero.ll new file mode 100644 index 0000000..e75aefd --- /dev/null +++ b/test/Analysis/LoopDependenceAnalysis/siv-weak-zero.ll @@ -0,0 +1,56 @@ +; RUN: opt < %s -analyze -lda | FileCheck %s + +@x = common global [256 x i32] zeroinitializer, align 4 +@y = common global [256 x i32] zeroinitializer, align 4 + +;; for (i = 0; i < 256; i++) +;; x[i] = x[42] + y[i] + +define void @f1(...) nounwind { +entry: + %x.ld.addr = getelementptr [256 x i32]* @x, i64 0, i64 42 + br label %for.body + +for.body: + %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ] + %x.addr = getelementptr [256 x i32]* @x, i64 0, i64 %i + %y.addr = getelementptr [256 x i32]* @y, i64 0, i64 %i + %x = load i32* %x.ld.addr ; 0 + %y = load i32* %y.addr ; 1 + %r = add i32 %y, %x + store i32 %r, i32* %x.addr ; 2 +; CHECK: 0,2: dep +; CHECK: 1,2: ind + %i.next = add i64 %i, 1 + %exitcond = icmp eq i64 %i.next, 256 + br i1 %exitcond, label %for.end, label %for.body + +for.end: + ret void +} + +;; for (i = 0; i < 250; i++) +;; x[i] = x[255] + y[i] + +define void @f2(...) nounwind { +entry: + %x.ld.addr = getelementptr [256 x i32]* @x, i64 0, i64 255 + br label %for.body + +for.body: + %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ] + %x.addr = getelementptr [256 x i32]* @x, i64 0, i64 %i + %y.addr = getelementptr [256 x i32]* @y, i64 0, i64 %i + %x = load i32* %x.ld.addr ; 0 + %y = load i32* %y.addr ; 1 + %r = add i32 %y, %x + store i32 %r, i32* %x.addr ; 2 +; CHECK: 0,2: dep +; CHECK: 1,2: ind + %i.next = add i64 %i, 1 + %exitcond = icmp eq i64 %i.next, 250 + br i1 %exitcond, label %for.end, label %for.body + +for.end: + ret void +} diff --git a/test/Analysis/LoopDependenceAnalysis/ziv.ll b/test/Analysis/LoopDependenceAnalysis/ziv.ll new file mode 100644 index 0000000..ba45948 --- /dev/null +++ b/test/Analysis/LoopDependenceAnalysis/ziv.ll @@ -0,0 +1,63 @@ +; RUN: opt < %s -analyze -lda | FileCheck %s + +@x = common global [256 x i32] zeroinitializer, align 4 + +;; x[5] = x[6] + +define void @f1(...) nounwind { +entry: + br label %for.body + +for.body: + %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ] + %x = load i32* getelementptr ([256 x i32]* @x, i32 0, i64 6) + store i32 %x, i32* getelementptr ([256 x i32]* @x, i32 0, i64 5) +; CHECK: 0,1: ind + %i.next = add i64 %i, 1 + %exitcond = icmp eq i64 %i.next, 256 + br i1 %exitcond, label %for.end, label %for.body + +for.end: + ret void +} + +;; x[c] = x[c+1] // with c being a loop-invariant constant + +define void @f2(i64 %c0) nounwind { +entry: + %c1 = add i64 %c0, 1 + %x.ld.addr = getelementptr [256 x i32]* @x, i64 0, i64 %c0 + %x.st.addr = getelementptr [256 x i32]* @x, i64 0, i64 %c1 + br label %for.body + +for.body: + %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ] + %x = load i32* %x.ld.addr + store i32 %x, i32* %x.st.addr +; CHECK: 0,1: ind + %i.next = add i64 %i, 1 + %exitcond = icmp eq i64 %i.next, 256 + br i1 %exitcond, label %for.end, label %for.body + +for.end: + ret void +} + +;; x[6] = x[6] + +define void @f3(...) nounwind { +entry: + br label %for.body + +for.body: + %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ] + %x = load i32* getelementptr ([256 x i32]* @x, i32 0, i64 6) + store i32 %x, i32* getelementptr ([256 x i32]* @x, i32 0, i64 6) +; CHECK: 0,1: dep + %i.next = add i64 %i, 1 + %exitcond = icmp eq i64 %i.next, 256 + br i1 %exitcond, label %for.end, label %for.body + +for.end: + ret void +} diff --git a/test/Analysis/LoopInfo/2003-05-15-NestingProblem.ll b/test/Analysis/LoopInfo/2003-05-15-NestingProblem.ll new file mode 100644 index 0000000..9355aee --- /dev/null +++ b/test/Analysis/LoopInfo/2003-05-15-NestingProblem.ll @@ -0,0 +1,30 @@ +; This testcase was incorrectly computing that the loopentry.7 loop was +; not a child of the loopentry.6 loop. +; +; RUN: opt < %s -analyze -loops | \ +; RUN: grep {^ Loop at depth 4 containing: %loopentry.7<header><latch><exiting>} + +define void @getAndMoveToFrontDecode() { + br label %endif.2 + +endif.2: ; preds = %loopexit.5, %0 + br i1 false, label %loopentry.5, label %UnifiedExitNode + +loopentry.5: ; preds = %loopexit.6, %endif.2 + br i1 false, label %loopentry.6, label %UnifiedExitNode + +loopentry.6: ; preds = %loopentry.7, %loopentry.5 + br i1 false, label %loopentry.7, label %loopexit.6 + +loopentry.7: ; preds = %loopentry.7, %loopentry.6 + br i1 false, label %loopentry.7, label %loopentry.6 + +loopexit.6: ; preds = %loopentry.6 + br i1 false, label %loopentry.5, label %loopexit.5 + +loopexit.5: ; preds = %loopexit.6 + br i1 false, label %endif.2, label %UnifiedExitNode + +UnifiedExitNode: ; preds = %loopexit.5, %loopentry.5, %endif.2 + ret void +} diff --git a/test/Analysis/LoopInfo/dg.exp b/test/Analysis/LoopInfo/dg.exp new file mode 100644 index 0000000..f200589 --- /dev/null +++ b/test/Analysis/LoopInfo/dg.exp @@ -0,0 +1,3 @@ +load_lib llvm.exp + +RunLLVMTests [lsort [glob -nocomplain $srcdir/$subdir/*.{ll,c,cpp}]] diff --git a/test/Analysis/PointerTracking/dg.exp b/test/Analysis/PointerTracking/dg.exp new file mode 100644 index 0000000..f200589 --- /dev/null +++ b/test/Analysis/PointerTracking/dg.exp @@ -0,0 +1,3 @@ +load_lib llvm.exp + +RunLLVMTests [lsort [glob -nocomplain $srcdir/$subdir/*.{ll,c,cpp}]] diff --git a/test/Analysis/PointerTracking/sizes.ll b/test/Analysis/PointerTracking/sizes.ll new file mode 100644 index 0000000..c8ca648 --- /dev/null +++ b/test/Analysis/PointerTracking/sizes.ll @@ -0,0 +1,86 @@ +; RUN: opt < %s -pointertracking -analyze | 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" +target triple = "x86_64-unknown-linux-gnu" +@.str = internal constant [5 x i8] c"1234\00" ; <[5 x i8]*> [#uses=1] +@test1p = global i8* getelementptr ([5 x i8]* @.str, i32 0, i32 0), align 8 ; <i8**> [#uses=1] +@test1a = global [5 x i8] c"1234\00", align 1 ; <[5 x i8]*> [#uses=1] +@test2a = global [5 x i32] [i32 1, i32 2, i32 3, i32 4, i32 5], align 4 ; <[5 x i32]*> [#uses=2] +@test2p = global i32* getelementptr ([5 x i32]* @test2a, i32 0, i32 0), align 8 ; <i32**> [#uses=1] +@test0p = common global i32* null, align 8 ; <i32**> [#uses=1] +@test0i = common global i32 0, align 4 ; <i32*> [#uses=1] + +define i32 @foo0() nounwind { +entry: + %tmp = load i32** @test0p ; <i32*> [#uses=1] + %conv = bitcast i32* %tmp to i8* ; <i8*> [#uses=1] + %call = tail call i32 @bar(i8* %conv) nounwind ; <i32> [#uses=1] + %tmp1 = load i8** @test1p ; <i8*> [#uses=1] + %call2 = tail call i32 @bar(i8* %tmp1) nounwind ; <i32> [#uses=1] + %call3 = tail call i32 @bar(i8* getelementptr ([5 x i8]* @test1a, i32 0, i32 0)) nounwind ; <i32> [#uses=1] + %call5 = tail call i32 @bar(i8* bitcast ([5 x i32]* @test2a to i8*)) nounwind ; <i32> [#uses=1] + %tmp7 = load i32** @test2p ; <i32*> [#uses=1] + %conv8 = bitcast i32* %tmp7 to i8* ; <i8*> [#uses=1] + %call9 = tail call i32 @bar(i8* %conv8) nounwind ; <i32> [#uses=1] + %call11 = tail call i32 @bar(i8* bitcast (i32* @test0i to i8*)) nounwind ; <i32> [#uses=1] + %add = add i32 %call2, %call ; <i32> [#uses=1] + %add4 = add i32 %add, %call3 ; <i32> [#uses=1] + %add6 = add i32 %add4, %call5 ; <i32> [#uses=1] + %add10 = add i32 %add6, %call9 ; <i32> [#uses=1] + %add12 = add i32 %add10, %call11 ; <i32> [#uses=1] + ret i32 %add12 +} + +declare i32 @bar(i8*) + +define i32 @foo1(i32 %n) nounwind { +entry: +; CHECK: 'foo1': + %test4a = alloca [10 x i8], align 1 ; <[10 x i8]*> [#uses=1] +; CHECK: %test4a = +; CHECK: ==> 1 elements, 10 bytes allocated + %test6a = alloca [10 x i32], align 4 ; <[10 x i32]*> [#uses=1] +; CHECK: %test6a = +; CHECK: ==> 1 elements, 40 bytes allocated + %vla = alloca i8, i32 %n, align 1 ; <i8*> [#uses=1] +; CHECK: %vla = +; CHECK: ==> %n elements, %n bytes allocated + %0 = shl i32 %n, 2 ; <i32> [#uses=1] + %vla7 = alloca i8, i32 %0, align 1 ; <i8*> [#uses=1] +; CHECK: %vla7 = +; CHECK: ==> (4 * %n) elements, (4 * %n) bytes allocated + %call = call i32 @bar(i8* %vla) nounwind ; <i32> [#uses=1] + %arraydecay = getelementptr [10 x i8]* %test4a, i64 0, i64 0 ; <i8*> [#uses=1] + %call10 = call i32 @bar(i8* %arraydecay) nounwind ; <i32> [#uses=1] + %call11 = call i32 @bar(i8* %vla7) nounwind ; <i32> [#uses=1] + %ptrconv14 = bitcast [10 x i32]* %test6a to i8* ; <i8*> [#uses=1] + %call15 = call i32 @bar(i8* %ptrconv14) nounwind ; <i32> [#uses=1] + %add = add i32 %call10, %call ; <i32> [#uses=1] + %add12 = add i32 %add, %call11 ; <i32> [#uses=1] + %add16 = add i32 %add12, %call15 ; <i32> [#uses=1] + ret i32 %add16 +} + +define i32 @foo2(i64 %n) nounwind { +entry: + %call = tail call i8* @malloc(i64 %n) ; <i8*> [#uses=1] +; CHECK: %call = +; CHECK: ==> %n elements, %n bytes allocated + %call2 = tail call i8* @calloc(i64 2, i64 4) nounwind ; <i8*> [#uses=1] +; CHECK: %call2 = +; CHECK: ==> 8 elements, 8 bytes allocated + %call4 = tail call i8* @realloc(i8* null, i64 16) nounwind ; <i8*> [#uses=1] +; CHECK: %call4 = +; CHECK: ==> 16 elements, 16 bytes allocated + %call6 = tail call i32 @bar(i8* %call) nounwind ; <i32> [#uses=1] + %call8 = tail call i32 @bar(i8* %call2) nounwind ; <i32> [#uses=1] + %call10 = tail call i32 @bar(i8* %call4) nounwind ; <i32> [#uses=1] + %add = add i32 %call8, %call6 ; <i32> [#uses=1] + %add11 = add i32 %add, %call10 ; <i32> [#uses=1] + ret i32 %add11 +} + +declare noalias i8* @malloc(i64) nounwind + +declare noalias i8* @calloc(i64, i64) nounwind + +declare noalias i8* @realloc(i8* nocapture, i64) nounwind diff --git a/test/Analysis/PostDominators/2006-09-26-PostDominanceFrontier.ll b/test/Analysis/PostDominators/2006-09-26-PostDominanceFrontier.ll new file mode 100644 index 0000000..b73b7f0 --- /dev/null +++ b/test/Analysis/PostDominators/2006-09-26-PostDominanceFrontier.ll @@ -0,0 +1,97 @@ +; RUN: opt < %s -analyze -postdomfrontier \ +; RUN: -disable-verify +; ModuleID = '2006-09-26-PostDominanceFrontier.bc' +target datalayout = "e-p:64:64" +target triple = "alphaev67-unknown-linux-gnu" + %struct.FILE = type { i32, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, %struct._IO_marker*, %struct.FILE*, i32, i32, i64, i16, i8, [1 x i8], i8*, i64, i8*, i8*, i32, [44 x i8] } + %struct._IO_marker = type { %struct._IO_marker*, %struct.FILE*, i32 } +@TOP = external global i64* ; <i64**> [#uses=1] +@BOT = external global i64* ; <i64**> [#uses=1] +@str = external global [2 x i8] ; <[2 x i8]*> [#uses=0] + +declare void @fopen() + +define void @main(i8** %argv) { +entry: + %netSelect.i507 = alloca i64, align 8 ; <i64*> [#uses=0] + %topStart.i = alloca i64, align 8 ; <i64*> [#uses=0] + %topEnd.i = alloca i64, align 8 ; <i64*> [#uses=0] + %botStart.i = alloca i64, align 8 ; <i64*> [#uses=0] + %botEnd.i = alloca i64, align 8 ; <i64*> [#uses=0] + %c1.i154 = alloca i32, align 4 ; <i32*> [#uses=0] + %b1.i155 = alloca i32, align 4 ; <i32*> [#uses=0] + %t1.i156 = alloca i32, align 4 ; <i32*> [#uses=0] + %c1.i = alloca i32, align 4 ; <i32*> [#uses=0] + %b1.i = alloca i32, align 4 ; <i32*> [#uses=0] + %t1.i = alloca i32, align 4 ; <i32*> [#uses=0] + %netSelect.i5 = alloca i64, align 8 ; <i64*> [#uses=0] + %netSelect.i = alloca i64, align 8 ; <i64*> [#uses=0] + %tmp2.i = getelementptr i8** %argv, i32 1 ; <i8**> [#uses=1] + %tmp3.i4 = load i8** %tmp2.i ; <i8*> [#uses=0] + call void @fopen( ) + br i1 false, label %DimensionChannel.exit, label %bb.backedge.i + +bb.backedge.i: ; preds = %entry + ret void + +DimensionChannel.exit: ; preds = %entry + %tmp13.i137 = malloc i64, i32 0 ; <i64*> [#uses=1] + %tmp610.i = malloc i64, i32 0 ; <i64*> [#uses=1] + br label %cond_true.i143 + +cond_true.i143: ; preds = %cond_true.i143, %DimensionChannel.exit + %tmp9.i140 = getelementptr i64* %tmp13.i137, i64 0 ; <i64*> [#uses=0] + %tmp12.i = getelementptr i64* %tmp610.i, i64 0 ; <i64*> [#uses=0] + br i1 false, label %bb18.i144, label %cond_true.i143 + +bb18.i144: ; preds = %cond_true.i143 + call void @fopen( ) + %tmp76.i105 = malloc i64, i32 0 ; <i64*> [#uses=3] + %tmp674.i = malloc i64, i32 0 ; <i64*> [#uses=2] + %tmp1072.i = malloc i64, i32 0 ; <i64*> [#uses=2] + %tmp1470.i = malloc i64, i32 0 ; <i64*> [#uses=1] + br label %cond_true.i114 + +cond_true.i114: ; preds = %cond_true.i114, %bb18.i144 + %tmp17.i108 = getelementptr i64* %tmp76.i105, i64 0 ; <i64*> [#uses=0] + %tmp20.i = getelementptr i64* %tmp674.i, i64 0 ; <i64*> [#uses=0] + %tmp23.i111 = getelementptr i64* %tmp1470.i, i64 0 ; <i64*> [#uses=0] + br i1 false, label %cond_true40.i, label %cond_true.i114 + +cond_true40.i: ; preds = %cond_true40.i, %cond_true.i114 + %tmp33.i115 = getelementptr i64* %tmp1072.i, i64 0 ; <i64*> [#uses=0] + br i1 false, label %bb142.i, label %cond_true40.i + +cond_next54.i: ; preds = %cond_true76.i + %tmp57.i = getelementptr i64* %tmp55.i, i64 0 ; <i64*> [#uses=0] + br i1 false, label %bb64.i, label %bb69.i + +bb64.i: ; preds = %cond_true76.i, %cond_next54.i + %tmp67.i117 = getelementptr i64* %tmp76.i105, i64 0 ; <i64*> [#uses=0] + br i1 false, label %bb114.i, label %cond_true111.i + +bb69.i: ; preds = %cond_next54.i + br i1 false, label %bb79.i, label %cond_true76.i + +cond_true76.i: ; preds = %bb142.i, %bb69.i + %tmp48.i = getelementptr i64* %tmp46.i, i64 0 ; <i64*> [#uses=0] + br i1 false, label %bb64.i, label %cond_next54.i + +bb79.i: ; preds = %bb69.i + br i1 false, label %bb114.i, label %cond_true111.i + +cond_true111.i: ; preds = %bb79.i, %bb64.i + %tmp84.i127 = getelementptr i64* %tmp46.i, i64 0 ; <i64*> [#uses=0] + ret void + +bb114.i: ; preds = %bb142.i, %bb79.i, %bb64.i + %tmp117.i = getelementptr i64* %tmp76.i105, i64 0 ; <i64*> [#uses=0] + %tmp132.i131 = getelementptr i64* %tmp674.i, i64 0 ; <i64*> [#uses=0] + %tmp122.i = getelementptr i64* %tmp1072.i, i64 0 ; <i64*> [#uses=0] + ret void + +bb142.i: ; preds = %cond_true40.i + %tmp46.i = load i64** @BOT ; <i64*> [#uses=2] + %tmp55.i = load i64** @TOP ; <i64*> [#uses=1] + br i1 false, label %bb114.i, label %cond_true76.i +} diff --git a/test/Analysis/PostDominators/2007-04-17-PostDominanceFrontier.ll b/test/Analysis/PostDominators/2007-04-17-PostDominanceFrontier.ll new file mode 100644 index 0000000..1ec056b --- /dev/null +++ b/test/Analysis/PostDominators/2007-04-17-PostDominanceFrontier.ll @@ -0,0 +1,692 @@ +; RUN: opt < %s -postdomfrontier -disable-output + +define void @SManager() { +entry: + br label %bb.outer + +bb.outer: ; preds = %bb193, %entry + br label %bb.outer156 + +bb.loopexit: ; preds = %bb442 + br label %bb.outer156 + +bb.outer156: ; preds = %bb.loopexit, %bb.outer + br label %bb + +bb: ; preds = %bb.backedge, %bb.outer156 + br i1 false, label %cond_true, label %bb.cond_next_crit_edge + +bb.cond_next_crit_edge: ; preds = %bb + br label %cond_next + +cond_true: ; preds = %bb + br label %cond_next + +cond_next: ; preds = %cond_true, %bb.cond_next_crit_edge + br i1 false, label %cond_next.bb.backedge_crit_edge, label %cond_next107 + +cond_next.bb.backedge_crit_edge: ; preds = %cond_next + br label %bb.backedge + +bb.backedge: ; preds = %cond_true112.bb.backedge_crit_edge, %cond_next.bb.backedge_crit_edge + br label %bb + +cond_next107: ; preds = %cond_next + br i1 false, label %cond_true112, label %cond_next197 + +cond_true112: ; preds = %cond_next107 + br i1 false, label %cond_true118, label %cond_true112.bb.backedge_crit_edge + +cond_true112.bb.backedge_crit_edge: ; preds = %cond_true112 + br label %bb.backedge + +cond_true118: ; preds = %cond_true112 + br i1 false, label %bb123.preheader, label %cond_true118.bb148_crit_edge + +cond_true118.bb148_crit_edge: ; preds = %cond_true118 + br label %bb148 + +bb123.preheader: ; preds = %cond_true118 + br label %bb123 + +bb123: ; preds = %bb142.bb123_crit_edge, %bb123.preheader + br i1 false, label %bb123.bb142_crit_edge, label %cond_next.i57 + +bb123.bb142_crit_edge: ; preds = %bb123 + br label %bb142 + +cond_next.i57: ; preds = %bb123 + br i1 false, label %cond_true135, label %cond_next.i57.bb142_crit_edge + +cond_next.i57.bb142_crit_edge: ; preds = %cond_next.i57 + br label %bb142 + +cond_true135: ; preds = %cond_next.i57 + br label %bb142 + +bb142: ; preds = %cond_true135, %cond_next.i57.bb142_crit_edge, %bb123.bb142_crit_edge + br i1 false, label %bb148.loopexit, label %bb142.bb123_crit_edge + +bb142.bb123_crit_edge: ; preds = %bb142 + br label %bb123 + +bb148.loopexit: ; preds = %bb142 + br label %bb148 + +bb148: ; preds = %bb148.loopexit, %cond_true118.bb148_crit_edge + br i1 false, label %bb151.preheader, label %bb148.bb177_crit_edge + +bb148.bb177_crit_edge: ; preds = %bb148 + br label %bb177 + +bb151.preheader: ; preds = %bb148 + br label %bb151 + +bb151: ; preds = %bb171.bb151_crit_edge, %bb151.preheader + br i1 false, label %bb151.bb171_crit_edge, label %cond_next.i49 + +bb151.bb171_crit_edge: ; preds = %bb151 + br label %bb171 + +cond_next.i49: ; preds = %bb151 + br i1 false, label %cond_true164, label %cond_next.i49.bb171_crit_edge + +cond_next.i49.bb171_crit_edge: ; preds = %cond_next.i49 + br label %bb171 + +cond_true164: ; preds = %cond_next.i49 + br label %bb171 + +bb171: ; preds = %cond_true164, %cond_next.i49.bb171_crit_edge, %bb151.bb171_crit_edge + br i1 false, label %bb177.loopexit, label %bb171.bb151_crit_edge + +bb171.bb151_crit_edge: ; preds = %bb171 + br label %bb151 + +bb177.loopexit: ; preds = %bb171 + br label %bb177 + +bb177: ; preds = %bb177.loopexit, %bb148.bb177_crit_edge + br i1 false, label %bb180.preheader, label %bb177.bb193_crit_edge + +bb177.bb193_crit_edge: ; preds = %bb177 + br label %bb193 + +bb180.preheader: ; preds = %bb177 + br label %bb180 + +bb180: ; preds = %bb180.bb180_crit_edge, %bb180.preheader + br i1 false, label %bb193.loopexit, label %bb180.bb180_crit_edge + +bb180.bb180_crit_edge: ; preds = %bb180 + br label %bb180 + +bb193.loopexit: ; preds = %bb180 + br label %bb193 + +bb193: ; preds = %bb193.loopexit, %bb177.bb193_crit_edge + br label %bb.outer + +cond_next197: ; preds = %cond_next107 + br i1 false, label %cond_next210, label %cond_true205 + +cond_true205: ; preds = %cond_next197 + br i1 false, label %cond_true205.bb213_crit_edge, label %cond_true205.bb299_crit_edge + +cond_true205.bb299_crit_edge: ; preds = %cond_true205 + br label %bb299 + +cond_true205.bb213_crit_edge: ; preds = %cond_true205 + br label %bb213 + +cond_next210: ; preds = %cond_next197 + br label %bb293 + +bb213: ; preds = %bb293.bb213_crit_edge, %cond_true205.bb213_crit_edge + br i1 false, label %bb213.cond_next290_crit_edge, label %cond_true248 + +bb213.cond_next290_crit_edge: ; preds = %bb213 + br label %cond_next290 + +cond_true248: ; preds = %bb213 + br i1 false, label %cond_true248.cond_next290_crit_edge, label %cond_true255 + +cond_true248.cond_next290_crit_edge: ; preds = %cond_true248 + br label %cond_next290 + +cond_true255: ; preds = %cond_true248 + br i1 false, label %cond_true266, label %cond_true255.cond_next271_crit_edge + +cond_true255.cond_next271_crit_edge: ; preds = %cond_true255 + br label %cond_next271 + +cond_true266: ; preds = %cond_true255 + br label %cond_next271 + +cond_next271: ; preds = %cond_true266, %cond_true255.cond_next271_crit_edge + br label %cond_next290 + +cond_next290: ; preds = %cond_next271, %cond_true248.cond_next290_crit_edge, %bb213.cond_next290_crit_edge + br label %bb293 + +bb293: ; preds = %cond_next290, %cond_next210 + br i1 false, label %bb293.bb213_crit_edge, label %bb293.bb299_crit_edge + +bb293.bb299_crit_edge: ; preds = %bb293 + br label %bb299 + +bb293.bb213_crit_edge: ; preds = %bb293 + br label %bb213 + +bb299: ; preds = %bb293.bb299_crit_edge, %cond_true205.bb299_crit_edge + br i1 false, label %bb302.preheader, label %bb299.bb390_crit_edge + +bb299.bb390_crit_edge: ; preds = %bb299 + br label %bb390 + +bb302.preheader: ; preds = %bb299 + br label %bb302 + +bb302: ; preds = %bb384.bb302_crit_edge, %bb302.preheader + br i1 false, label %bb302.bb384_crit_edge, label %cond_true339 + +bb302.bb384_crit_edge: ; preds = %bb302 + br label %bb384 + +cond_true339: ; preds = %bb302 + br i1 false, label %cond_true339.bb384_crit_edge, label %cond_true346 + +cond_true339.bb384_crit_edge: ; preds = %cond_true339 + br label %bb384 + +cond_true346: ; preds = %cond_true339 + br i1 false, label %cond_true357, label %cond_true346.cond_next361_crit_edge + +cond_true346.cond_next361_crit_edge: ; preds = %cond_true346 + br label %cond_next361 + +cond_true357: ; preds = %cond_true346 + br label %cond_next361 + +cond_next361: ; preds = %cond_true357, %cond_true346.cond_next361_crit_edge + br label %bb384 + +bb384: ; preds = %cond_next361, %cond_true339.bb384_crit_edge, %bb302.bb384_crit_edge + br i1 false, label %bb390.loopexit, label %bb384.bb302_crit_edge + +bb384.bb302_crit_edge: ; preds = %bb384 + br label %bb302 + +bb390.loopexit: ; preds = %bb384 + br label %bb390 + +bb390: ; preds = %bb390.loopexit, %bb299.bb390_crit_edge + br i1 false, label %bb391.preheader, label %bb390.bb442.preheader_crit_edge + +bb390.bb442.preheader_crit_edge: ; preds = %bb390 + br label %bb442.preheader + +bb391.preheader: ; preds = %bb390 + br label %bb391 + +bb391: ; preds = %bb413.bb391_crit_edge, %bb391.preheader + br i1 false, label %bb391.bb413_crit_edge, label %cond_next404 + +bb391.bb413_crit_edge: ; preds = %bb391 + br label %bb413 + +cond_next404: ; preds = %bb391 + br i1 false, label %cond_next404.HWrite.exit_crit_edge, label %cond_next.i13 + +cond_next404.HWrite.exit_crit_edge: ; preds = %cond_next404 + br label %HWrite.exit + +cond_next.i13: ; preds = %cond_next404 + br i1 false, label %cond_next.i13.cond_next13.i_crit_edge, label %cond_true12.i + +cond_next.i13.cond_next13.i_crit_edge: ; preds = %cond_next.i13 + br label %cond_next13.i + +cond_true12.i: ; preds = %cond_next.i13 + br label %cond_next13.i + +cond_next13.i: ; preds = %cond_true12.i, %cond_next.i13.cond_next13.i_crit_edge + br i1 false, label %cond_next13.i.bb.i22_crit_edge, label %cond_next43.i + +cond_next13.i.bb.i22_crit_edge: ; preds = %cond_next13.i + br label %bb.i22 + +cond_next43.i: ; preds = %cond_next13.i + br i1 false, label %cond_next43.i.bb.i22_crit_edge, label %bb60.i + +cond_next43.i.bb.i22_crit_edge: ; preds = %cond_next43.i + br label %bb.i22 + +bb.i22: ; preds = %cond_next43.i.bb.i22_crit_edge, %cond_next13.i.bb.i22_crit_edge + br label %bb413 + +bb60.i: ; preds = %cond_next43.i + br i1 false, label %bb60.i.HWrite.exit_crit_edge, label %cond_true81.i + +bb60.i.HWrite.exit_crit_edge: ; preds = %bb60.i + br label %HWrite.exit + +cond_true81.i: ; preds = %bb60.i + br label %bb413 + +HWrite.exit: ; preds = %bb60.i.HWrite.exit_crit_edge, %cond_next404.HWrite.exit_crit_edge + br label %bb413 + +bb413: ; preds = %HWrite.exit, %cond_true81.i, %bb.i22, %bb391.bb413_crit_edge + br i1 false, label %bb442.preheader.loopexit, label %bb413.bb391_crit_edge + +bb413.bb391_crit_edge: ; preds = %bb413 + br label %bb391 + +bb442.preheader.loopexit: ; preds = %bb413 + br label %bb442.preheader + +bb442.preheader: ; preds = %bb442.preheader.loopexit, %bb390.bb442.preheader_crit_edge + br label %bb442.outer + +bb420: ; preds = %bb442 + br i1 false, label %bb439.loopexit, label %cond_next433 + +cond_next433: ; preds = %bb420 + br i1 false, label %cond_next433.HRead.exit.loopexit_crit_edge, label %cond_next.i + +cond_next433.HRead.exit.loopexit_crit_edge: ; preds = %cond_next433 + br label %HRead.exit.loopexit + +cond_next.i: ; preds = %cond_next433 + br i1 false, label %cond_true9.i, label %cond_false223.i + +cond_true9.i: ; preds = %cond_next.i + switch i32 0, label %cond_false.i [ + i32 1, label %cond_true9.i.cond_true15.i_crit_edge + i32 5, label %cond_true9.i.cond_true15.i_crit_edge9 + ] + +cond_true9.i.cond_true15.i_crit_edge9: ; preds = %cond_true9.i + br label %cond_true15.i + +cond_true9.i.cond_true15.i_crit_edge: ; preds = %cond_true9.i + br label %cond_true15.i + +cond_true15.i: ; preds = %cond_true9.i.cond_true15.i_crit_edge, %cond_true9.i.cond_true15.i_crit_edge9 + br i1 false, label %cond_true15.i.cond_true44.i_crit_edge, label %cond_true15.i.cond_false49.i_crit_edge + +cond_true15.i.cond_false49.i_crit_edge: ; preds = %cond_true15.i + br label %cond_false49.i + +cond_true15.i.cond_true44.i_crit_edge: ; preds = %cond_true15.i + br label %cond_true44.i + +cond_false.i: ; preds = %cond_true9.i + br i1 false, label %cond_false.i.cond_next39.i_crit_edge, label %cond_true30.i + +cond_false.i.cond_next39.i_crit_edge: ; preds = %cond_false.i + br label %cond_next39.i + +cond_true30.i: ; preds = %cond_false.i + br label %cond_next39.i + +cond_next39.i: ; preds = %cond_true30.i, %cond_false.i.cond_next39.i_crit_edge + br i1 false, label %cond_next39.i.cond_true44.i_crit_edge, label %cond_next39.i.cond_false49.i_crit_edge + +cond_next39.i.cond_false49.i_crit_edge: ; preds = %cond_next39.i + br label %cond_false49.i + +cond_next39.i.cond_true44.i_crit_edge: ; preds = %cond_next39.i + br label %cond_true44.i + +cond_true44.i: ; preds = %cond_next39.i.cond_true44.i_crit_edge, %cond_true15.i.cond_true44.i_crit_edge + br i1 false, label %cond_true44.i.cond_next70.i_crit_edge, label %cond_true44.i.cond_true61.i_crit_edge + +cond_true44.i.cond_true61.i_crit_edge: ; preds = %cond_true44.i + br label %cond_true61.i + +cond_true44.i.cond_next70.i_crit_edge: ; preds = %cond_true44.i + br label %cond_next70.i + +cond_false49.i: ; preds = %cond_next39.i.cond_false49.i_crit_edge, %cond_true15.i.cond_false49.i_crit_edge + br i1 false, label %cond_false49.i.cond_next70.i_crit_edge, label %cond_false49.i.cond_true61.i_crit_edge + +cond_false49.i.cond_true61.i_crit_edge: ; preds = %cond_false49.i + br label %cond_true61.i + +cond_false49.i.cond_next70.i_crit_edge: ; preds = %cond_false49.i + br label %cond_next70.i + +cond_true61.i: ; preds = %cond_false49.i.cond_true61.i_crit_edge, %cond_true44.i.cond_true61.i_crit_edge + br i1 false, label %cond_true61.i.cond_next70.i_crit_edge, label %cond_true67.i + +cond_true61.i.cond_next70.i_crit_edge: ; preds = %cond_true61.i + br label %cond_next70.i + +cond_true67.i: ; preds = %cond_true61.i + br label %cond_next70.i + +cond_next70.i: ; preds = %cond_true67.i, %cond_true61.i.cond_next70.i_crit_edge, %cond_false49.i.cond_next70.i_crit_edge, %cond_true44.i.cond_next70.i_crit_edge + br i1 false, label %cond_true77.i, label %cond_next81.i + +cond_true77.i: ; preds = %cond_next70.i + br label %bb442.outer.backedge + +cond_next81.i: ; preds = %cond_next70.i + br i1 false, label %cond_true87.i, label %cond_false94.i + +cond_true87.i: ; preds = %cond_next81.i + br i1 false, label %cond_true87.i.cond_true130.i_crit_edge, label %cond_true87.i.cond_next135.i_crit_edge + +cond_true87.i.cond_next135.i_crit_edge: ; preds = %cond_true87.i + br label %cond_next135.i + +cond_true87.i.cond_true130.i_crit_edge: ; preds = %cond_true87.i + br label %cond_true130.i + +cond_false94.i: ; preds = %cond_next81.i + switch i32 0, label %cond_false94.i.cond_next125.i_crit_edge [ + i32 1, label %cond_false94.i.cond_true100.i_crit_edge + i32 5, label %cond_false94.i.cond_true100.i_crit_edge10 + ] + +cond_false94.i.cond_true100.i_crit_edge10: ; preds = %cond_false94.i + br label %cond_true100.i + +cond_false94.i.cond_true100.i_crit_edge: ; preds = %cond_false94.i + br label %cond_true100.i + +cond_false94.i.cond_next125.i_crit_edge: ; preds = %cond_false94.i + br label %cond_next125.i + +cond_true100.i: ; preds = %cond_false94.i.cond_true100.i_crit_edge, %cond_false94.i.cond_true100.i_crit_edge10 + br i1 false, label %cond_true107.i, label %cond_true100.i.cond_next109.i_crit_edge + +cond_true100.i.cond_next109.i_crit_edge: ; preds = %cond_true100.i + br label %cond_next109.i + +cond_true107.i: ; preds = %cond_true100.i + br label %cond_next109.i + +cond_next109.i: ; preds = %cond_true107.i, %cond_true100.i.cond_next109.i_crit_edge + br i1 false, label %cond_next109.i.cond_next125.i_crit_edge, label %cond_true116.i + +cond_next109.i.cond_next125.i_crit_edge: ; preds = %cond_next109.i + br label %cond_next125.i + +cond_true116.i: ; preds = %cond_next109.i + br label %cond_next125.i + +cond_next125.i: ; preds = %cond_true116.i, %cond_next109.i.cond_next125.i_crit_edge, %cond_false94.i.cond_next125.i_crit_edge + br i1 false, label %cond_next125.i.cond_true130.i_crit_edge, label %cond_next125.i.cond_next135.i_crit_edge + +cond_next125.i.cond_next135.i_crit_edge: ; preds = %cond_next125.i + br label %cond_next135.i + +cond_next125.i.cond_true130.i_crit_edge: ; preds = %cond_next125.i + br label %cond_true130.i + +cond_true130.i: ; preds = %cond_next125.i.cond_true130.i_crit_edge, %cond_true87.i.cond_true130.i_crit_edge + br label %cond_next135.i + +cond_next135.i: ; preds = %cond_true130.i, %cond_next125.i.cond_next135.i_crit_edge, %cond_true87.i.cond_next135.i_crit_edge + br i1 false, label %cond_true142.i, label %cond_next135.i.cond_next149.i_crit_edge + +cond_next135.i.cond_next149.i_crit_edge: ; preds = %cond_next135.i + br label %cond_next149.i + +cond_true142.i: ; preds = %cond_next135.i + br label %cond_next149.i + +cond_next149.i: ; preds = %cond_true142.i, %cond_next135.i.cond_next149.i_crit_edge + br i1 false, label %cond_true156.i, label %cond_next149.i.cond_next163.i_crit_edge + +cond_next149.i.cond_next163.i_crit_edge: ; preds = %cond_next149.i + br label %cond_next163.i + +cond_true156.i: ; preds = %cond_next149.i + br label %cond_next163.i + +cond_next163.i: ; preds = %cond_true156.i, %cond_next149.i.cond_next163.i_crit_edge + br i1 false, label %cond_true182.i, label %cond_next163.i.cond_next380.i_crit_edge + +cond_next163.i.cond_next380.i_crit_edge: ; preds = %cond_next163.i + br label %cond_next380.i + +cond_true182.i: ; preds = %cond_next163.i + br i1 false, label %cond_true182.i.cond_next380.i_crit_edge, label %cond_true196.i + +cond_true182.i.cond_next380.i_crit_edge: ; preds = %cond_true182.i + br label %cond_next380.i + +cond_true196.i: ; preds = %cond_true182.i + br i1 false, label %cond_true210.i, label %cond_true196.i.cond_next380.i_crit_edge + +cond_true196.i.cond_next380.i_crit_edge: ; preds = %cond_true196.i + br label %cond_next380.i + +cond_true210.i: ; preds = %cond_true196.i + br i1 false, label %cond_true216.i, label %cond_true210.i.cond_next380.i_crit_edge + +cond_true210.i.cond_next380.i_crit_edge: ; preds = %cond_true210.i + br label %cond_next380.i + +cond_true216.i: ; preds = %cond_true210.i + br label %cond_next380.i + +cond_false223.i: ; preds = %cond_next.i + br i1 false, label %cond_true229.i, label %cond_false355.i + +cond_true229.i: ; preds = %cond_false223.i + br i1 false, label %cond_true229.i.HRead.exit.loopexit_crit_edge, label %cond_next243.i + +cond_true229.i.HRead.exit.loopexit_crit_edge: ; preds = %cond_true229.i + br label %HRead.exit.loopexit + +cond_next243.i: ; preds = %cond_true229.i + br i1 false, label %cond_true248.i, label %cond_false255.i + +cond_true248.i: ; preds = %cond_next243.i + br label %cond_next260.i + +cond_false255.i: ; preds = %cond_next243.i + br label %cond_next260.i + +cond_next260.i: ; preds = %cond_false255.i, %cond_true248.i + br i1 false, label %cond_true267.i, label %cond_next273.i + +cond_true267.i: ; preds = %cond_next260.i + br label %bb442.backedge + +bb442.backedge: ; preds = %bb.i, %cond_true267.i + br label %bb442 + +cond_next273.i: ; preds = %cond_next260.i + br i1 false, label %cond_true281.i, label %cond_next273.i.cond_next288.i_crit_edge + +cond_next273.i.cond_next288.i_crit_edge: ; preds = %cond_next273.i + br label %cond_next288.i + +cond_true281.i: ; preds = %cond_next273.i + br label %cond_next288.i + +cond_next288.i: ; preds = %cond_true281.i, %cond_next273.i.cond_next288.i_crit_edge + br i1 false, label %cond_true295.i, label %cond_next288.i.cond_next302.i_crit_edge + +cond_next288.i.cond_next302.i_crit_edge: ; preds = %cond_next288.i + br label %cond_next302.i + +cond_true295.i: ; preds = %cond_next288.i + br label %cond_next302.i + +cond_next302.i: ; preds = %cond_true295.i, %cond_next288.i.cond_next302.i_crit_edge + br i1 false, label %cond_next302.i.cond_next380.i_crit_edge, label %cond_true328.i + +cond_next302.i.cond_next380.i_crit_edge: ; preds = %cond_next302.i + br label %cond_next380.i + +cond_true328.i: ; preds = %cond_next302.i + br i1 false, label %cond_true343.i, label %cond_true328.i.cond_next380.i_crit_edge + +cond_true328.i.cond_next380.i_crit_edge: ; preds = %cond_true328.i + br label %cond_next380.i + +cond_true343.i: ; preds = %cond_true328.i + br i1 false, label %cond_true349.i, label %cond_true343.i.cond_next380.i_crit_edge + +cond_true343.i.cond_next380.i_crit_edge: ; preds = %cond_true343.i + br label %cond_next380.i + +cond_true349.i: ; preds = %cond_true343.i + br label %cond_next380.i + +cond_false355.i: ; preds = %cond_false223.i + br i1 false, label %cond_false355.i.bb.i_crit_edge, label %cond_next363.i + +cond_false355.i.bb.i_crit_edge: ; preds = %cond_false355.i + br label %bb.i + +cond_next363.i: ; preds = %cond_false355.i + br i1 false, label %bb377.i, label %cond_next363.i.bb.i_crit_edge + +cond_next363.i.bb.i_crit_edge: ; preds = %cond_next363.i + br label %bb.i + +bb.i: ; preds = %cond_next363.i.bb.i_crit_edge, %cond_false355.i.bb.i_crit_edge + br label %bb442.backedge + +bb377.i: ; preds = %cond_next363.i + br label %cond_next380.i + +cond_next380.i: ; preds = %bb377.i, %cond_true349.i, %cond_true343.i.cond_next380.i_crit_edge, %cond_true328.i.cond_next380.i_crit_edge, %cond_next302.i.cond_next380.i_crit_edge, %cond_true216.i, %cond_true210.i.cond_next380.i_crit_edge, %cond_true196.i.cond_next380.i_crit_edge, %cond_true182.i.cond_next380.i_crit_edge, %cond_next163.i.cond_next380.i_crit_edge + br i1 false, label %cond_next380.i.HRead.exit_crit_edge, label %cond_true391.i + +cond_next380.i.HRead.exit_crit_edge: ; preds = %cond_next380.i + br label %HRead.exit + +cond_true391.i: ; preds = %cond_next380.i + br label %bb442.outer.backedge + +bb442.outer.backedge: ; preds = %bb439, %cond_true391.i, %cond_true77.i + br label %bb442.outer + +HRead.exit.loopexit: ; preds = %cond_true229.i.HRead.exit.loopexit_crit_edge, %cond_next433.HRead.exit.loopexit_crit_edge + br label %HRead.exit + +HRead.exit: ; preds = %HRead.exit.loopexit, %cond_next380.i.HRead.exit_crit_edge + br label %bb439 + +bb439.loopexit: ; preds = %bb420 + br label %bb439 + +bb439: ; preds = %bb439.loopexit, %HRead.exit + br label %bb442.outer.backedge + +bb442.outer: ; preds = %bb442.outer.backedge, %bb442.preheader + br label %bb442 + +bb442: ; preds = %bb442.outer, %bb442.backedge + br i1 false, label %bb420, label %bb.loopexit +} + +define void @Invalidate() { +entry: + br i1 false, label %cond_false, label %cond_true + +cond_true: ; preds = %entry + br i1 false, label %cond_true40, label %cond_true.cond_next_crit_edge + +cond_true.cond_next_crit_edge: ; preds = %cond_true + br label %cond_next + +cond_true40: ; preds = %cond_true + br label %cond_next + +cond_next: ; preds = %cond_true40, %cond_true.cond_next_crit_edge + br i1 false, label %cond_true68, label %cond_next.cond_next73_crit_edge + +cond_next.cond_next73_crit_edge: ; preds = %cond_next + br label %cond_next73 + +cond_true68: ; preds = %cond_next + br label %cond_next73 + +cond_next73: ; preds = %cond_true68, %cond_next.cond_next73_crit_edge + br i1 false, label %cond_true91, label %cond_next73.cond_next96_crit_edge + +cond_next73.cond_next96_crit_edge: ; preds = %cond_next73 + br label %cond_next96 + +cond_true91: ; preds = %cond_next73 + br label %cond_next96 + +cond_next96: ; preds = %cond_true91, %cond_next73.cond_next96_crit_edge + br i1 false, label %cond_next96.cond_next112_crit_edge, label %cond_true105 + +cond_next96.cond_next112_crit_edge: ; preds = %cond_next96 + br label %cond_next112 + +cond_true105: ; preds = %cond_next96 + br label %cond_next112 + +cond_next112: ; preds = %cond_true105, %cond_next96.cond_next112_crit_edge + br i1 false, label %cond_next112.cond_next127_crit_edge, label %cond_true119 + +cond_next112.cond_next127_crit_edge: ; preds = %cond_next112 + br label %cond_next127 + +cond_true119: ; preds = %cond_next112 + br label %cond_next127 + +cond_next127: ; preds = %cond_true119, %cond_next112.cond_next127_crit_edge + br i1 false, label %cond_next141, label %cond_true134 + +cond_true134: ; preds = %cond_next127 + br i1 false, label %cond_true134.bb161_crit_edge, label %cond_true134.bb_crit_edge + +cond_true134.bb_crit_edge: ; preds = %cond_true134 + br label %bb + +cond_true134.bb161_crit_edge: ; preds = %cond_true134 + br label %bb161 + +cond_next141: ; preds = %cond_next127 + br label %bb154 + +bb: ; preds = %bb154.bb_crit_edge, %cond_true134.bb_crit_edge + br label %bb154 + +bb154: ; preds = %bb, %cond_next141 + br i1 false, label %bb154.bb161_crit_edge, label %bb154.bb_crit_edge + +bb154.bb_crit_edge: ; preds = %bb154 + br label %bb + +bb154.bb161_crit_edge: ; preds = %bb154 + br label %bb161 + +bb161: ; preds = %bb154.bb161_crit_edge, %cond_true134.bb161_crit_edge + br i1 false, label %bb161.cond_next201_crit_edge, label %cond_true198 + +bb161.cond_next201_crit_edge: ; preds = %bb161 + br label %cond_next201 + +cond_true198: ; preds = %bb161 + br label %cond_next201 + +cond_next201: ; preds = %cond_true198, %bb161.cond_next201_crit_edge + br i1 false, label %cond_next212, label %cond_true206 + +cond_true206: ; preds = %cond_next201 + br label %UnifiedReturnBlock + +cond_false: ; preds = %entry + br label %UnifiedReturnBlock + +cond_next212: ; preds = %cond_next201 + br label %UnifiedReturnBlock + +UnifiedReturnBlock: ; preds = %cond_next212, %cond_false, %cond_true206 + ret void +} diff --git a/test/Analysis/PostDominators/2007-04-20-PostDom-Reset.ll b/test/Analysis/PostDominators/2007-04-20-PostDom-Reset.ll new file mode 100644 index 0000000..767e5db --- /dev/null +++ b/test/Analysis/PostDominators/2007-04-20-PostDom-Reset.ll @@ -0,0 +1,28 @@ +; RUN: opt < %s -postdomfrontier -disable-output + +define void @args_out_of_range() { +entry: + br label %bb + +bb: ; preds = %bb, %entry + br label %bb +} + +define void @args_out_of_range_3() { +entry: + br label %bb + +bb: ; preds = %bb, %entry + br label %bb +} + +define void @Feq() { +entry: + br i1 false, label %cond_true, label %cond_next + +cond_true: ; preds = %entry + unreachable + +cond_next: ; preds = %entry + unreachable +} diff --git a/test/Analysis/PostDominators/dg.exp b/test/Analysis/PostDominators/dg.exp new file mode 100644 index 0000000..f200589 --- /dev/null +++ b/test/Analysis/PostDominators/dg.exp @@ -0,0 +1,3 @@ +load_lib llvm.exp + +RunLLVMTests [lsort [glob -nocomplain $srcdir/$subdir/*.{ll,c,cpp}]] diff --git a/test/Analysis/PostDominators/pr1098.ll b/test/Analysis/PostDominators/pr1098.ll new file mode 100644 index 0000000..afb4776 --- /dev/null +++ b/test/Analysis/PostDominators/pr1098.ll @@ -0,0 +1,14 @@ +; RUN: opt < %s -postdomtree -analyze | grep entry +; PR932 + +define void @foo(i1 %x) { +entry: + br i1 %x, label %bb1, label %bb0 +bb0: ; preds = %entry, bb0 + br label %bb0 +bb1: ; preds = %entry + br label %bb2 +bb2: ; preds = %bb1 + ret void +} + diff --git a/test/Analysis/PostDominators/pr6047_a.ll b/test/Analysis/PostDominators/pr6047_a.ll new file mode 100644 index 0000000..ec1455b --- /dev/null +++ b/test/Analysis/PostDominators/pr6047_a.ll @@ -0,0 +1,15 @@ +; RUN: opt < %s -postdomtree -analyze | FileCheck %s +define internal void @f() { +entry: + br i1 undef, label %bb35, label %bb3.i + +bb3.i: + br label %bb3.i + +bb35.loopexit3: + br label %bb35 + +bb35: + ret void +} +; CHECK: [3] %entry diff --git a/test/Analysis/PostDominators/pr6047_b.ll b/test/Analysis/PostDominators/pr6047_b.ll new file mode 100644 index 0000000..7bd2c86 --- /dev/null +++ b/test/Analysis/PostDominators/pr6047_b.ll @@ -0,0 +1,19 @@ +; RUN: opt < %s -postdomtree -analyze | FileCheck %s +define internal void @f() { +entry: + br i1 undef, label %a, label %bb3.i + +a: + br i1 undef, label %bb35, label %bb3.i + +bb3.i: + br label %bb3.i + + +bb35.loopexit3: + br label %bb35 + +bb35: + ret void +} +; CHECK: [4] %entry diff --git a/test/Analysis/PostDominators/pr6047_c.ll b/test/Analysis/PostDominators/pr6047_c.ll new file mode 100644 index 0000000..08c9551 --- /dev/null +++ b/test/Analysis/PostDominators/pr6047_c.ll @@ -0,0 +1,147 @@ +; RUN: opt < %s -postdomtree -analyze | FileCheck %s +define internal void @f() { +entry: + br i1 undef, label %bb35, label %bb3.i + +bb3.i: + br label %bb3.i + +bb: + br label %bb35 + +bb.i: + br label %bb35 + +_float32_unpack.exit: + br label %bb35 + +bb.i5: + br label %bb35 + +_float32_unpack.exit8: + br label %bb35 + +bb32.preheader: + br label %bb35 + +bb3: + br label %bb35 + +bb3.split.us: + br label %bb35 + +bb.i4.us: + br label %bb35 + +bb7.i.us: + br label %bb35 + +bb.i4.us.backedge: + br label %bb35 + +bb1.i.us: + br label %bb35 + +bb6.i.us: + br label %bb35 + +bb4.i.us: + br label %bb35 + +bb8.i.us: + br label %bb35 + +bb3.i.loopexit.us: + br label %bb35 + +bb.nph21: + br label %bb35 + +bb4: + br label %bb35 + +bb5: + br label %bb35 + +bb14.preheader: + br label %bb35 + +bb.nph18: + br label %bb35 + +bb8.us.preheader: + br label %bb35 + +bb8.preheader: + br label %bb35 + +bb8.us: + br label %bb35 + +bb8: + br label %bb35 + +bb15.loopexit: + br label %bb35 + +bb15.loopexit2: + br label %bb35 + +bb15: + br label %bb35 + +bb16: + br label %bb35 + +bb17.loopexit.split: + br label %bb35 + +bb.nph14: + br label %bb35 + +bb19: + br label %bb35 + +bb20: + br label %bb35 + +bb29.preheader: + br label %bb35 + +bb.nph: + br label %bb35 + +bb23.us.preheader: + br label %bb35 + +bb23.preheader: + br label %bb35 + +bb23.us: + br label %bb35 + +bb23: + br label %bb35 + +bb30.loopexit: + br label %bb35 + +bb30.loopexit1: + br label %bb35 + +bb30: + br label %bb35 + +bb31: + br label %bb35 + +bb35.loopexit: + br label %bb35 + +bb35.loopexit3: + br label %bb35 + +bb35: + ret void +} +; CHECK: [3] %entry diff --git a/test/Analysis/PostDominators/pr6047_d.ll b/test/Analysis/PostDominators/pr6047_d.ll new file mode 100644 index 0000000..4cfa880 --- /dev/null +++ b/test/Analysis/PostDominators/pr6047_d.ll @@ -0,0 +1,24 @@ +; RUN: opt < %s -postdomtree -analyze | FileCheck %s +define internal void @f() { +entry: + br i1 1, label %a, label %b + +a: +br label %c + +b: +br label %c + +c: + br i1 undef, label %bb35, label %bb3.i + +bb3.i: + br label %bb3.i + +bb35.loopexit3: + br label %bb35 + +bb35: + ret void +} +; CHECK: [4] %entry diff --git a/test/Analysis/Profiling/dg.exp b/test/Analysis/Profiling/dg.exp new file mode 100644 index 0000000..1eb4755 --- /dev/null +++ b/test/Analysis/Profiling/dg.exp @@ -0,0 +1,4 @@ +load_lib llvm.exp + +RunLLVMTests [lsort [glob -nocomplain $srcdir/$subdir/*.{ll,c,cpp}]] + diff --git a/test/Analysis/Profiling/edge-profiling.ll b/test/Analysis/Profiling/edge-profiling.ll new file mode 100644 index 0000000..cbaf476 --- /dev/null +++ b/test/Analysis/Profiling/edge-profiling.ll @@ -0,0 +1,139 @@ +; Test the edge profiling instrumentation. +; RUN: opt < %s -insert-edge-profiling -S | FileCheck %s + +; ModuleID = '<stdin>' + +@.str = private constant [12 x i8] c"hello world\00", align 1 ; <[12 x i8]*> [#uses=1] +@.str1 = private constant [6 x i8] c"franz\00", align 1 ; <[6 x i8]*> [#uses=1] +@.str2 = private constant [9 x i8] c"argc > 2\00", align 1 ; <[9 x i8]*> [#uses=1] +@.str3 = private constant [9 x i8] c"argc = 1\00", align 1 ; <[9 x i8]*> [#uses=1] +@.str4 = private constant [6 x i8] c"fritz\00", align 1 ; <[6 x i8]*> [#uses=1] +@.str5 = private constant [10 x i8] c"argc <= 1\00", align 1 ; <[10 x i8]*> [#uses=1] +; CHECK:@EdgeProfCounters +; CHECK:[19 x i32] +; CHECK:zeroinitializer + +define void @oneblock() nounwind { +entry: +; CHECK:entry: +; CHECK:%OldFuncCounter +; CHECK:load +; CHECK:getelementptr +; CHECK:@EdgeProfCounters +; CHECK:i32 0 +; CHECK:i32 0 +; CHECK:%NewFuncCounter +; CHECK:add +; CHECK:%OldFuncCounter +; CHECK:store +; CHECK:%NewFuncCounter +; CHECK:getelementptr +; CHECK:@EdgeProfCounters + %0 = call i32 @puts(i8* getelementptr inbounds ([12 x i8]* @.str, i64 0, i64 0)) nounwind ; <i32> [#uses=0] + ret void +} + +declare i32 @puts(i8*) + +define i32 @main(i32 %argc, i8** %argv) nounwind { +entry: +; CHECK:entry: + %argc_addr = alloca i32 ; <i32*> [#uses=4] + %argv_addr = alloca i8** ; <i8***> [#uses=1] + %retval = alloca i32 ; <i32*> [#uses=2] + %j = alloca i32 ; <i32*> [#uses=4] + %i = alloca i32 ; <i32*> [#uses=4] + %0 = alloca i32 ; <i32*> [#uses=2] +; CHECK:call +; CHECK:@llvm_start_edge_profiling +; CHECK:@EdgeProfCounters + %"alloca point" = bitcast i32 0 to i32 ; <i32> [#uses=0] + store i32 %argc, i32* %argc_addr + store i8** %argv, i8*** %argv_addr + store i32 0, i32* %i, align 4 + br label %bb10 + +bb: ; preds = %bb10 +; CHECK:bb: + %1 = load i32* %argc_addr, align 4 ; <i32> [#uses=1] + %2 = icmp sgt i32 %1, 1 ; <i1> [#uses=1] + br i1 %2, label %bb1, label %bb8 + +bb1: ; preds = %bb +; CHECK:bb1: + store i32 0, i32* %j, align 4 + br label %bb6 + +bb2: ; preds = %bb6 +; CHECK:bb2: + %3 = call i32 @puts(i8* getelementptr inbounds ([6 x i8]* @.str1, i64 0, i64 0)) nounwind ; <i32> [#uses=0] + %4 = load i32* %argc_addr, align 4 ; <i32> [#uses=1] + %5 = icmp sgt i32 %4, 2 ; <i1> [#uses=1] + br i1 %5, label %bb3, label %bb4 + +bb3: ; preds = %bb2 +; CHECK:bb3: + %6 = call i32 @puts(i8* getelementptr inbounds ([9 x i8]* @.str2, i64 0, i64 0)) nounwind ; <i32> [#uses=0] + br label %bb5 + +bb4: ; preds = %bb2 +; CHECK:bb4: + %7 = call i32 @puts(i8* getelementptr inbounds ([9 x i8]* @.str3, i64 0, i64 0)) nounwind ; <i32> [#uses=0] + br label %bb11 + +bb5: ; preds = %bb3 +; CHECK:bb5: + %8 = call i32 @puts(i8* getelementptr inbounds ([6 x i8]* @.str4, i64 0, i64 0)) nounwind ; <i32> [#uses=0] + %9 = load i32* %j, align 4 ; <i32> [#uses=1] + %10 = add nsw i32 %9, 1 ; <i32> [#uses=1] + store i32 %10, i32* %j, align 4 + br label %bb6 + +bb6: ; preds = %bb5, %bb1 +; CHECK:bb6: + %11 = load i32* %j, align 4 ; <i32> [#uses=1] + %12 = load i32* %argc_addr, align 4 ; <i32> [#uses=1] + %13 = icmp slt i32 %11, %12 ; <i1> [#uses=1] + br i1 %13, label %bb2, label %bb7 + +bb7: ; preds = %bb6 +; CHECK:bb7: + br label %bb9 + +bb8: ; preds = %bb +; CHECK:bb8: + %14 = call i32 @puts(i8* getelementptr inbounds ([10 x i8]* @.str5, i64 0, i64 0)) nounwind ; <i32> [#uses=0] + br label %bb9 + +bb9: ; preds = %bb8, %bb7 +; CHECK:bb9: + %15 = load i32* %i, align 4 ; <i32> [#uses=1] + %16 = add nsw i32 %15, 1 ; <i32> [#uses=1] + store i32 %16, i32* %i, align 4 + br label %bb10 + +bb10: ; preds = %bb9, %entry +; CHECK:bb10: + %17 = load i32* %i, align 4 ; <i32> [#uses=1] + %18 = icmp ne i32 %17, 3 ; <i1> [#uses=1] + br i1 %18, label %bb, label %bb11 +; CHECK:br +; CHECK:label %bb10.bb11_crit_edge + +; CHECK:bb10.bb11_crit_edge: +; CHECK:br +; CHECK:label %bb11 + +bb11: ; preds = %bb10, %bb4 +; CHECK:bb11: + call void @oneblock() nounwind + store i32 0, i32* %0, align 4 + %19 = load i32* %0, align 4 ; <i32> [#uses=1] + store i32 %19, i32* %retval, align 4 + br label %return + +return: ; preds = %bb11 +; CHECK:return: + %retval12 = load i32* %retval ; <i32> [#uses=1] + ret i32 %retval12 +} diff --git a/test/Analysis/Profiling/profiling-tool-chain.ll b/test/Analysis/Profiling/profiling-tool-chain.ll new file mode 100644 index 0000000..5ac31b5 --- /dev/null +++ b/test/Analysis/Profiling/profiling-tool-chain.ll @@ -0,0 +1,212 @@ +; RUN: llvm-as %s -o %t1 + +; FIXME: The RUX parts of the test are disabled for now, they aren't working on +; llvm-gcc-x86_64-darwin10-selfhost. + +; Test the edge optimal profiling instrumentation. +; RUN: opt %t1 -insert-optimal-edge-profiling -o %t2 +; RUX: llvm-dis < %t2 | FileCheck --check-prefix=INST %s + +; Test the creation, reading and displaying of profile +; RUX: rm -f llvmprof.out +; RUX: lli -load %llvmlibsdir/profile_rt%shlibext %t2 +; RUX: lli -load %llvmlibsdir/profile_rt%shlibext %t2 1 2 +; RUX: llvm-prof -print-all-code %t1 | FileCheck --check-prefix=PROF %s + +; Test the loaded profile also with verifier. +; RUX opt %t1 -profile-loader -profile-verifier -o %t3 + +; Test profile estimator. +; RUN: opt %t1 -profile-estimator -profile-verifier -o %t3 + +; PROF: 1. 2/4 oneblock +; PROF: 2. 2/4 main +; PROF: 1. 15.7895% 12/76 main() - bb6 +; PROF: 2. 11.8421% 9/76 main() - bb2 +; PROF: 3. 11.8421% 9/76 main() - bb3 +; PROF: 4. 11.8421% 9/76 main() - bb5 +; PROF: 5. 10.5263% 8/76 main() - bb10 +; PROF: 6. 7.89474% 6/76 main() - bb +; PROF: 7. 7.89474% 6/76 main() - bb9 +; PROF: 8. 3.94737% 3/76 main() - bb1 +; PROF: 9. 3.94737% 3/76 main() - bb7 +; PROF: 10. 3.94737% 3/76 main() - bb8 +; PROF: 11. 2.63158% 2/76 oneblock() - entry +; PROF: 12. 2.63158% 2/76 main() - entry +; PROF: 13. 2.63158% 2/76 main() - bb11 +; PROF: 14. 2.63158% 2/76 main() - return + +; ModuleID = '<stdin>' + +@.str = private constant [12 x i8] c"hello world\00", align 1 ; <[12 x i8]*> [#uses=1] +@.str1 = private constant [6 x i8] c"franz\00", align 1 ; <[6 x i8]*> [#uses=1] +@.str2 = private constant [9 x i8] c"argc > 2\00", align 1 ; <[9 x i8]*> [#uses=1] +@.str3 = private constant [9 x i8] c"argc = 1\00", align 1 ; <[9 x i8]*> [#uses=1] +@.str4 = private constant [6 x i8] c"fritz\00", align 1 ; <[6 x i8]*> [#uses=1] +@.str5 = private constant [10 x i8] c"argc <= 1\00", align 1 ; <[10 x i8]*> [#uses=1] +; INST:@OptEdgeProfCounters +; INST:[21 x i32] +; INST:[i32 0, +; INST:i32 -1, +; INST:i32 -1, +; INST:i32 -1, +; INST:i32 -1, +; INST:i32 -1, +; INST:i32 -1, +; INST:i32 -1, +; INST:i32 -1, +; INST:i32 0, +; INST:i32 0, +; INST:i32 -1, +; INST:i32 -1, +; INST:i32 -1, +; INST:i32 0, +; INST:i32 0, +; INST:i32 -1, +; INST:i32 -1, +; INST:i32 0, +; INST:i32 -1, +; INST:i32 -1] + +; PROF:;;; %oneblock called 2 times. +; PROF:;;; +define void @oneblock() nounwind { +entry: +; PROF:entry: +; PROF: ;;; Basic block executed 2 times. + %0 = call i32 @puts(i8* getelementptr inbounds ([12 x i8]* @.str, i64 0, i64 0)) nounwind ; <i32> [#uses=0] + ret void +} + +declare i32 @puts(i8*) + +; PROF:;;; %main called 2 times. +; PROF:;;; +define i32 @main(i32 %argc, i8** %argv) nounwind { +entry: +; PROF:entry: +; PROF: ;;; Basic block executed 2 times. + %argc_addr = alloca i32 ; <i32*> [#uses=4] + %argv_addr = alloca i8** ; <i8***> [#uses=1] + %retval = alloca i32 ; <i32*> [#uses=2] + %j = alloca i32 ; <i32*> [#uses=4] + %i = alloca i32 ; <i32*> [#uses=4] + %0 = alloca i32 ; <i32*> [#uses=2] +; INST:call +; INST:@llvm_start_opt_edge_profiling +; INST:@OptEdgeProfCounters + %"alloca point" = bitcast i32 0 to i32 ; <i32> [#uses=0] + store i32 %argc, i32* %argc_addr + store i8** %argv, i8*** %argv_addr + store i32 0, i32* %i, align 4 + br label %bb10 +; PROF: ;;; Out-edge counts: [2.000000e+00 -> bb10] + +bb: ; preds = %bb10 +; PROF:bb: +; PROF: ;;; Basic block executed 6 times. + %1 = load i32* %argc_addr, align 4 ; <i32> [#uses=1] + %2 = icmp sgt i32 %1, 1 ; <i1> [#uses=1] + br i1 %2, label %bb1, label %bb8 +; PROF: ;;; Out-edge counts: [3.000000e+00 -> bb1] [3.000000e+00 -> bb8] + +bb1: ; preds = %bb +; PROF:bb1: +; PROF: ;;; Basic block executed 3 times. + store i32 0, i32* %j, align 4 + br label %bb6 +; PROF: ;;; Out-edge counts: [3.000000e+00 -> bb6] + +bb2: ; preds = %bb6 +; PROF:bb2: +; PROF: ;;; Basic block executed 9 times. + %3 = call i32 @puts(i8* getelementptr inbounds ([6 x i8]* @.str1, i64 0, i64 0)) nounwind ; <i32> [#uses=0] + %4 = load i32* %argc_addr, align 4 ; <i32> [#uses=1] + %5 = icmp sgt i32 %4, 2 ; <i1> [#uses=1] + br i1 %5, label %bb3, label %bb4 +; PROF: ;;; Out-edge counts: [9.000000e+00 -> bb3] + +bb3: ; preds = %bb2 +; PROF:bb3: +; PROF: ;;; Basic block executed 9 times. + %6 = call i32 @puts(i8* getelementptr inbounds ([9 x i8]* @.str2, i64 0, i64 0)) nounwind ; <i32> [#uses=0] + br label %bb5 +; PROF: ;;; Out-edge counts: [9.000000e+00 -> bb5] + +bb4: ; preds = %bb2 +; PROF:bb4: +; PROF: ;;; Never executed! + %7 = call i32 @puts(i8* getelementptr inbounds ([9 x i8]* @.str3, i64 0, i64 0)) nounwind ; <i32> [#uses=0] + br label %bb11 + +bb5: ; preds = %bb3 +; PROF:bb5: +; PROF: ;;; Basic block executed 9 times. + %8 = call i32 @puts(i8* getelementptr inbounds ([6 x i8]* @.str4, i64 0, i64 0)) nounwind ; <i32> [#uses=0] + %9 = load i32* %j, align 4 ; <i32> [#uses=1] + %10 = add nsw i32 %9, 1 ; <i32> [#uses=1] + store i32 %10, i32* %j, align 4 + br label %bb6 +; PROF: ;;; Out-edge counts: [9.000000e+00 -> bb6] + +bb6: ; preds = %bb5, %bb1 +; PROF:bb6: +; PROF: ;;; Basic block executed 12 times. + %11 = load i32* %j, align 4 ; <i32> [#uses=1] + %12 = load i32* %argc_addr, align 4 ; <i32> [#uses=1] + %13 = icmp slt i32 %11, %12 ; <i1> [#uses=1] + br i1 %13, label %bb2, label %bb7 +; PROF: ;;; Out-edge counts: [9.000000e+00 -> bb2] [3.000000e+00 -> bb7] + +bb7: ; preds = %bb6 +; PROF:bb7: +; PROF: ;;; Basic block executed 3 times. + br label %bb9 +; PROF: ;;; Out-edge counts: [3.000000e+00 -> bb9] + +bb8: ; preds = %bb +; PROF:bb8: +; PROF: ;;; Basic block executed 3 times. + %14 = call i32 @puts(i8* getelementptr inbounds ([10 x i8]* @.str5, i64 0, i64 0)) nounwind ; <i32> [#uses=0] + br label %bb9 +; PROF: ;;; Out-edge counts: [3.000000e+00 -> bb9] + +bb9: ; preds = %bb8, %bb7 +; PROF:bb9: +; PROF: ;;; Basic block executed 6 times. + %15 = load i32* %i, align 4 ; <i32> [#uses=1] + %16 = add nsw i32 %15, 1 ; <i32> [#uses=1] + store i32 %16, i32* %i, align 4 + br label %bb10 +; PROF: ;;; Out-edge counts: [6.000000e+00 -> bb10] + +bb10: ; preds = %bb9, %entry +; PROF:bb10: +; PROF: ;;; Basic block executed 8 times. + %17 = load i32* %i, align 4 ; <i32> [#uses=1] + %18 = icmp ne i32 %17, 3 ; <i1> [#uses=1] + br i1 %18, label %bb, label %bb11 +; INST:br +; INST:label %bb10.bb11_crit_edge +; PROF: ;;; Out-edge counts: [6.000000e+00 -> bb] [2.000000e+00 -> bb11] + +; INST:bb10.bb11_crit_edge: +; INST:br +; INST:label %bb11 + +bb11: ; preds = %bb10, %bb4 +; PROF:bb11: +; PROF: ;;; Basic block executed 2 times. + call void @oneblock() nounwind + store i32 0, i32* %0, align 4 + %19 = load i32* %0, align 4 ; <i32> [#uses=1] + store i32 %19, i32* %retval, align 4 + br label %return +; PROF: ;;; Out-edge counts: [2.000000e+00 -> return] + +return: ; preds = %bb11 +; PROF:return: +; PROF: ;;; Basic block executed 2 times. + %retval12 = load i32* %retval ; <i32> [#uses=1] + ret i32 %retval12 +} diff --git a/test/Analysis/ScalarEvolution/2007-07-15-NegativeStride.ll b/test/Analysis/ScalarEvolution/2007-07-15-NegativeStride.ll new file mode 100644 index 0000000..7ff130f --- /dev/null +++ b/test/Analysis/ScalarEvolution/2007-07-15-NegativeStride.ll @@ -0,0 +1,21 @@ +; RUN: opt < %s -analyze -scalar-evolution \ +; RUN: -scalar-evolution-max-iterations=0 | grep {Loop %bb: backedge-taken count is 100} +; PR1533 + +@array = weak global [101 x i32] zeroinitializer, align 32 ; <[100 x i32]*> [#uses=1] + +define void @loop(i32 %x) { +entry: + br label %bb + +bb: ; preds = %bb, %entry + %i.01.0 = phi i32 [ 100, %entry ], [ %tmp4, %bb ] ; <i32> [#uses=2] + %tmp1 = getelementptr [101 x i32]* @array, i32 0, i32 %i.01.0 ; <i32*> [#uses=1] + store i32 %x, i32* %tmp1 + %tmp4 = add i32 %i.01.0, -1 ; <i32> [#uses=2] + %tmp7 = icmp sgt i32 %tmp4, -1 ; <i1> [#uses=1] + br i1 %tmp7, label %bb, label %return + +return: ; preds = %bb + ret void +} diff --git a/test/Analysis/ScalarEvolution/2007-08-06-MisinterpretBranch.ll b/test/Analysis/ScalarEvolution/2007-08-06-MisinterpretBranch.ll new file mode 100644 index 0000000..e67e4d0 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2007-08-06-MisinterpretBranch.ll @@ -0,0 +1,18 @@ +; RUN: opt < %s -indvars -adce -simplifycfg -S | grep "icmp s" +; PR1598 + +define i32 @f(i32 %a, i32 %b, i32 %x, i32 %y) { +entry: + %tmp3 = icmp eq i32 %a, %b ; <i1> [#uses=1] + br i1 %tmp3, label %return, label %bb + +bb: ; preds = %bb, %entry + %x_addr.0 = phi i32 [ %tmp6, %bb ], [ %x, %entry ] ; <i32> [#uses=1] + %tmp6 = add i32 %x_addr.0, 1 ; <i32> [#uses=3] + %tmp9 = icmp slt i32 %tmp6, %y ; <i1> [#uses=1] + br i1 %tmp9, label %bb, label %return + +return: ; preds = %bb, %entry + %x_addr.1 = phi i32 [ %x, %entry ], [ %tmp6, %bb ] ; <i32> [#uses=1] + ret i32 %x_addr.1 +} diff --git a/test/Analysis/ScalarEvolution/2007-08-06-Unsigned.ll b/test/Analysis/ScalarEvolution/2007-08-06-Unsigned.ll new file mode 100644 index 0000000..ab96243 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2007-08-06-Unsigned.ll @@ -0,0 +1,30 @@ +; RUN: opt < %s -scalar-evolution -analyze | grep {Loop %bb: backedge-taken count is (-1 + (-1 \\* %x) + %y)} +; PR1597 + +define i32 @f(i32 %x, i32 %y) { +entry: + %tmp63 = icmp ult i32 %x, %y ; <i1> [#uses=1] + br i1 %tmp63, label %bb.preheader, label %bb8 + +bb.preheader: ; preds = %entry + br label %bb + +bb: ; preds = %bb3, %bb.preheader + %x_addr.0 = phi i32 [ %tmp2, %bb3 ], [ %x, %bb.preheader ] ; <i32> [#uses=1] + %tmp2 = add i32 %x_addr.0, 1 ; <i32> [#uses=3] + br label %bb3 + +bb3: ; preds = %bb + %tmp6 = icmp ult i32 %tmp2, %y ; <i1> [#uses=1] + br i1 %tmp6, label %bb, label %bb8.loopexit + +bb8.loopexit: ; preds = %bb3 + br label %bb8 + +bb8: ; preds = %bb8.loopexit, %entry + %x_addr.1 = phi i32 [ %x, %entry ], [ %tmp2, %bb8.loopexit ] ; <i32> [#uses=1] + br label %return + +return: ; preds = %bb8 + ret i32 %x_addr.1 +} diff --git a/test/Analysis/ScalarEvolution/2007-09-27-LargeStepping.ll b/test/Analysis/ScalarEvolution/2007-09-27-LargeStepping.ll new file mode 100644 index 0000000..b678fee --- /dev/null +++ b/test/Analysis/ScalarEvolution/2007-09-27-LargeStepping.ll @@ -0,0 +1,22 @@ +; RUN: opt < %s -analyze -scalar-evolution \ +; RUN: -scalar-evolution-max-iterations=0 | grep {backedge-taken count is 13} +; PR1706 + +define i32 @f() { +entry: + br label %bb5 + +bb: ; preds = %bb5 + %tmp2 = shl i32 %j.0, 1 ; <i32> [#uses=1] + %tmp4 = add i32 %i.0, 268435456 ; <i32> [#uses=1] + br label %bb5 + +bb5: ; preds = %bb, %entry + %j.0 = phi i32 [ 1, %entry ], [ %tmp2, %bb ] ; <i32> [#uses=2] + %i.0 = phi i32 [ -1879048192, %entry ], [ %tmp4, %bb ] ; <i32> [#uses=2] + %tmp7 = icmp slt i32 %i.0, 1610612736 ; <i1> [#uses=1] + br i1 %tmp7, label %bb, label %return + +return: ; preds = %bb5 + ret i32 %j.0 +} diff --git a/test/Analysis/ScalarEvolution/2007-11-14-SignedAddRec.ll b/test/Analysis/ScalarEvolution/2007-11-14-SignedAddRec.ll new file mode 100644 index 0000000..514920f --- /dev/null +++ b/test/Analysis/ScalarEvolution/2007-11-14-SignedAddRec.ll @@ -0,0 +1,24 @@ +; RUN: opt < %s -indvars -S | grep printd | grep 1206807378 +; PR1798 + +declare void @printd(i32) + +define i32 @test() { +entry: + br label %bb6 + +bb: ; preds = %bb6 + %tmp3 = add i32 %x.0, %i.0 ; <i32> [#uses=1] + %tmp5 = add i32 %i.0, 1 ; <i32> [#uses=1] + br label %bb6 + +bb6: ; preds = %bb, %entry + %i.0 = phi i32 [ 0, %entry ], [ %tmp5, %bb ] ; <i32> [#uses=3] + %x.0 = phi i32 [ 0, %entry ], [ %tmp3, %bb ] ; <i32> [#uses=3] + %tmp8 = icmp slt i32 %i.0, 123456789 ; <i1> [#uses=1] + br i1 %tmp8, label %bb, label %bb10 + +bb10: ; preds = %bb6 + call void @printd(i32 %x.0) + ret i32 0 +} diff --git a/test/Analysis/ScalarEvolution/2007-11-18-OrInstruction.ll b/test/Analysis/ScalarEvolution/2007-11-18-OrInstruction.ll new file mode 100644 index 0000000..c12721d --- /dev/null +++ b/test/Analysis/ScalarEvolution/2007-11-18-OrInstruction.ll @@ -0,0 +1,21 @@ +; RUN: opt < %s -analyze -scalar-evolution | FileCheck %s +; PR1810 + +define void @fun() { +entry: + br label %header +header: + %i = phi i32 [ 1, %entry ], [ %i.next, %body ] + %cond = icmp eq i32 %i, 10 + br i1 %cond, label %exit, label %body +body: + %a = mul i32 %i, 5 + %b = or i32 %a, 1 + %i.next = add i32 %i, 1 + br label %header +exit: + ret void +} + +; CHECK: --> %b + diff --git a/test/Analysis/ScalarEvolution/2008-02-11-ReversedCondition.ll b/test/Analysis/ScalarEvolution/2008-02-11-ReversedCondition.ll new file mode 100644 index 0000000..fe3a7f4 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-02-11-ReversedCondition.ll @@ -0,0 +1,15 @@ +; RUN: opt < %s -scalar-evolution -analyze | grep {Loop %header: backedge-taken count is (0 smax %n)} + +define void @foo(i32 %n) { +entry: + br label %header +header: + %i = phi i32 [ 0, %entry ], [ %i.inc, %next ] + %cond = icmp sgt i32 %n, %i + br i1 %cond, label %next, label %return +next: + %i.inc = add i32 %i, 1 + br label %header +return: + ret void +} diff --git a/test/Analysis/ScalarEvolution/2008-02-12-SMAXTripCount.ll b/test/Analysis/ScalarEvolution/2008-02-12-SMAXTripCount.ll new file mode 100644 index 0000000..4f14a0d --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-02-12-SMAXTripCount.ll @@ -0,0 +1,16 @@ +; RUN: opt < %s -scalar-evolution -analyze | grep {Loop %loop: backedge-taken count is (100 + (-100 smax %n))} +; PR2002 + +define void @foo(i8 %n) { +entry: + br label %loop +loop: + %i = phi i8 [ -100, %entry ], [ %i.inc, %next ] + %cond = icmp slt i8 %i, %n + br i1 %cond, label %next, label %return +next: + %i.inc = add i8 %i, 1 + br label %loop +return: + ret void +} diff --git a/test/Analysis/ScalarEvolution/2008-02-15-UMax.ll b/test/Analysis/ScalarEvolution/2008-02-15-UMax.ll new file mode 100644 index 0000000..52c7985 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-02-15-UMax.ll @@ -0,0 +1,17 @@ +; RUN: opt < %s -analyze -scalar-evolution | grep umax +; PR2003 + +define i32 @foo(i32 %n) { +entry: + br label %header +header: + %i = phi i32 [ 100, %entry ], [ %i.inc, %next ] + %cond = icmp ult i32 %i, %n + br i1 %cond, label %next, label %return +next: + %i.inc = add i32 %i, 1 + br label %header +return: + ret i32 %i +} + diff --git a/test/Analysis/ScalarEvolution/2008-05-25-NegativeStepToZero.ll b/test/Analysis/ScalarEvolution/2008-05-25-NegativeStepToZero.ll new file mode 100644 index 0000000..bcc124d --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-05-25-NegativeStepToZero.ll @@ -0,0 +1,22 @@ +; RUN: opt < %s -analyze -scalar-evolution \ +; RUN: -scalar-evolution-max-iterations=0 | grep {backedge-taken count is 61} +; PR2364 + +define i32 @func_6() nounwind { +entry: + br label %bb5 + +bb: ; preds = %bb5 + %tmp2 = add i32 %i.0, 1 ; <i32> [#uses=1] + %tmp4 = add i8 %x.0, -4 ; <i8> [#uses=1] + br label %bb5 + +bb5: ; preds = %bb, %entry + %x.0 = phi i8 [ 0, %entry ], [ %tmp4, %bb ] ; <i8> [#uses=2] + %i.0 = phi i32 [ 0, %entry ], [ %tmp2, %bb ] ; <i32> [#uses=2] + %tmp7 = icmp eq i8 %x.0, 12 ; <i1> [#uses=1] + br i1 %tmp7, label %return, label %bb + +return: ; preds = %bb5 + ret i32 %i.0 +} diff --git a/test/Analysis/ScalarEvolution/2008-06-12-BinomialInt64.ll b/test/Analysis/ScalarEvolution/2008-06-12-BinomialInt64.ll new file mode 100644 index 0000000..d503329 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-06-12-BinomialInt64.ll @@ -0,0 +1,43 @@ +; RUN: opt < %s -analyze -scalar-evolution 2>/dev/null +; PR2433 + +define i32 @main1(i32 %argc, i8** %argv) nounwind { +entry: + br i1 false, label %bb10, label %bb23 + +bb10: ; preds = %bb10, %entry + %accum.03 = phi i64 [ %tmp14, %bb10 ], [ 0, %entry ] ; <i64> [#uses=1] + %i.02 = phi i32 [ %tmp16, %bb10 ], [ 0, %entry ] ; <i32> [#uses=1] + %d.1.01 = phi i64 [ %tmp5.i, %bb10 ], [ 0, %entry ] ; <i64> [#uses=1] + %tmp5.i = add i64 %d.1.01, 1 ; <i64> [#uses=2] + %tmp14 = add i64 %accum.03, %tmp5.i ; <i64> [#uses=2] + %tmp16 = add i32 %i.02, 1 ; <i32> [#uses=2] + %tmp20 = icmp slt i32 %tmp16, 0 ; <i1> [#uses=1] + br i1 %tmp20, label %bb10, label %bb23 + +bb23: ; preds = %bb10, %entry + %accum.0.lcssa = phi i64 [ 0, %entry ], [ %tmp14, %bb10 ] ; <i64> [#uses=0] + ret i32 0 +} + +define i32 @main2(i32 %argc, i8** %argv) { +entry: + %tmp8 = tail call i32 @atoi( i8* null ) nounwind readonly ; <i32> [#uses=1] + br i1 false, label %bb9, label %bb21 + +bb9: ; preds = %bb9, %entry + %accum.03 = phi i64 [ %tmp12, %bb9 ], [ 0, %entry ] ; <i64> [#uses=1] + %i.02 = phi i32 [ %tmp14, %bb9 ], [ 0, %entry ] ; <i32> [#uses=1] + %d.1.01 = phi i64 [ %tmp4.i, %bb9 ], [ 0, %entry ] ; <i64> [#uses=1] + %tmp4.i = add i64 %d.1.01, 1 ; <i64> [#uses=2] + %tmp12 = add i64 %accum.03, %tmp4.i ; <i64> [#uses=2] + %tmp14 = add i32 %i.02, 1 ; <i32> [#uses=2] + %tmp18 = icmp slt i32 %tmp14, %tmp8 ; <i1> [#uses=1] + br i1 %tmp18, label %bb9, label %bb21 + +bb21: ; preds = %bb9, %entry + %accum.0.lcssa = phi i64 [ 0, %entry ], [ %tmp12, %bb9 ] ; <i64> [#uses=0] + ret i32 0 +} + +declare i32 @atoi(i8*) nounwind readonly diff --git a/test/Analysis/ScalarEvolution/2008-07-12-UnneededSelect1.ll b/test/Analysis/ScalarEvolution/2008-07-12-UnneededSelect1.ll new file mode 100644 index 0000000..9db9b71 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-07-12-UnneededSelect1.ll @@ -0,0 +1,36 @@ +; RUN: opt < %s -analyze -scalar-evolution |& not grep smax +; PR2261 + +@lut = common global [256 x i8] zeroinitializer, align 32 ; <[256 x i8]*> [#uses=1] + +define void @foo(i32 %count, i32* %srcptr, i32* %dstptr) nounwind { +entry: + icmp sgt i32 %count, 0 ; <i1>:0 [#uses=1] + br i1 %0, label %bb.nph, label %return + +bb.nph: ; preds = %entry + br label %bb + +bb: ; preds = %bb1, %bb.nph + %j.01 = phi i32 [ %8, %bb1 ], [ 0, %bb.nph ] ; <i32> [#uses=1] + load i32* %srcptr, align 4 ; <i32>:1 [#uses=2] + and i32 %1, 255 ; <i32>:2 [#uses=1] + and i32 %1, -256 ; <i32>:3 [#uses=1] + getelementptr [256 x i8]* @lut, i32 0, i32 %2 ; <i8*>:4 [#uses=1] + load i8* %4, align 1 ; <i8>:5 [#uses=1] + zext i8 %5 to i32 ; <i32>:6 [#uses=1] + or i32 %6, %3 ; <i32>:7 [#uses=1] + store i32 %7, i32* %dstptr, align 4 + add i32 %j.01, 1 ; <i32>:8 [#uses=2] + br label %bb1 + +bb1: ; preds = %bb + icmp slt i32 %8, %count ; <i1>:9 [#uses=1] + br i1 %9, label %bb, label %bb1.return_crit_edge + +bb1.return_crit_edge: ; preds = %bb1 + br label %return + +return: ; preds = %bb1.return_crit_edge, %entry + ret void +} diff --git a/test/Analysis/ScalarEvolution/2008-07-12-UnneededSelect2.ll b/test/Analysis/ScalarEvolution/2008-07-12-UnneededSelect2.ll new file mode 100644 index 0000000..1847665 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-07-12-UnneededSelect2.ll @@ -0,0 +1,30 @@ +; RUN: opt < %s -analyze -scalar-evolution |& not grep smax +; PR2070 + +define i32 @a(i32 %x) nounwind { +entry: + icmp sgt i32 %x, 1 ; <i1>:0 [#uses=1] + br i1 %0, label %bb.nph, label %bb2 + +bb.nph: ; preds = %entry + br label %bb + +bb: ; preds = %bb1, %bb.nph + %z.02 = phi i32 [ %1, %bb1 ], [ 1, %bb.nph ] ; <i32> [#uses=1] + %i.01 = phi i32 [ %2, %bb1 ], [ 1, %bb.nph ] ; <i32> [#uses=2] + mul i32 %z.02, %i.01 ; <i32>:1 [#uses=2] + add i32 %i.01, 1 ; <i32>:2 [#uses=2] + br label %bb1 + +bb1: ; preds = %bb + icmp slt i32 %2, %x ; <i1>:3 [#uses=1] + br i1 %3, label %bb, label %bb1.bb2_crit_edge + +bb1.bb2_crit_edge: ; preds = %bb1 + %.lcssa = phi i32 [ %1, %bb1 ] ; <i32> [#uses=1] + br label %bb2 + +bb2: ; preds = %bb1.bb2_crit_edge, %entry + %z.0.lcssa = phi i32 [ %.lcssa, %bb1.bb2_crit_edge ], [ 1, %entry ] ; <i32> [#uses=1] + ret i32 %z.0.lcssa +} diff --git a/test/Analysis/ScalarEvolution/2008-07-19-InfiniteLoop.ll b/test/Analysis/ScalarEvolution/2008-07-19-InfiniteLoop.ll new file mode 100644 index 0000000..1865c05 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-07-19-InfiniteLoop.ll @@ -0,0 +1,15 @@ +; RUN: opt < %s -analyze -scalar-evolution \ +; RUN: -scalar-evolution-max-iterations=0 | grep Unpredictable +; PR2088 + +define void @fun() { +entry: + br label %loop +loop: + %i = phi i8 [ 0, %entry ], [ %i.next, %loop ] + %i.next = add i8 %i, 4 + %cond = icmp ne i8 %i.next, 6 + br i1 %cond, label %loop, label %exit +exit: + ret void +} diff --git a/test/Analysis/ScalarEvolution/2008-07-19-WrappingIV.ll b/test/Analysis/ScalarEvolution/2008-07-19-WrappingIV.ll new file mode 100644 index 0000000..86e07ec4 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-07-19-WrappingIV.ll @@ -0,0 +1,15 @@ +; RUN: opt < %s -analyze -scalar-evolution \ +; RUN: -scalar-evolution-max-iterations=0 | grep {backedge-taken count is 113} +; PR2088 + +define void @fun() { +entry: + br label %loop +loop: + %i = phi i8 [ 0, %entry ], [ %i.next, %loop ] + %i.next = add i8 %i, 18 + %cond = icmp ne i8 %i.next, 4 + br i1 %cond, label %loop, label %exit +exit: + ret void +} diff --git a/test/Analysis/ScalarEvolution/2008-07-29-SGTTripCount.ll b/test/Analysis/ScalarEvolution/2008-07-29-SGTTripCount.ll new file mode 100644 index 0000000..75bd634 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-07-29-SGTTripCount.ll @@ -0,0 +1,28 @@ +; RUN: opt < %s -analyze -scalar-evolution \ +; RUN: -scalar-evolution-max-iterations=0 | FileCheck %s +; PR2607 + +define i32 @_Z1aj(i32 %j) nounwind { +entry: + icmp sgt i32 0, %j ; <i1>:0 [#uses=1] + br i1 %0, label %bb.preheader, label %return + +bb.preheader: ; preds = %entry + br label %bb + +bb: ; preds = %bb, %bb.preheader + %i.01 = phi i32 [ %1, %bb ], [ 0, %bb.preheader ] ; <i32> [#uses=1] + add i32 %i.01, -1 ; <i32>:1 [#uses=3] + icmp sgt i32 %1, %j ; <i1>:2 [#uses=1] + br i1 %2, label %bb, label %return.loopexit + +return.loopexit: ; preds = %bb + br label %return + +return: ; preds = %return.loopexit, %entry + %i.0.lcssa = phi i32 [ 0, %entry ], [ %1, %return.loopexit ] ; <i32> [#uses=1] + ret i32 %i.0.lcssa +} + +; CHECK: backedge-taken count is (-1 + (-1 * %j)) + diff --git a/test/Analysis/ScalarEvolution/2008-07-29-SMinExpr.ll b/test/Analysis/ScalarEvolution/2008-07-29-SMinExpr.ll new file mode 100644 index 0000000..1626c1f --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-07-29-SMinExpr.ll @@ -0,0 +1,26 @@ +; RUN: opt < %s -analyze -scalar-evolution \ +; RUN: -scalar-evolution-max-iterations=0 | FileCheck %s +; PR2607 + +define i32 @b(i32 %x, i32 %y) nounwind { +entry: + %cmp2 = icmp slt i32 %y, %x + %cond3 = select i1 %cmp2, i32 %y, i32 %x + %cmp54 = icmp slt i32 %cond3, -2147483632 + br i1 %cmp54, label %forinc, label %afterfor + +forinc: ; preds = %forinc, %entry + %j.01 = phi i32 [ %dec, %forinc ], [ -2147483632, %entry ] + %dec = add i32 %j.01, -1 + %cmp = icmp slt i32 %y, %x + %cond = select i1 %cmp, i32 %y, i32 %x + %cmp5 = icmp sgt i32 %dec, %cond + br i1 %cmp5, label %forinc, label %afterfor + +afterfor: ; preds = %forinc, %entry + %j.0.lcssa = phi i32 [ -2147483632, %entry ], [ %dec, %forinc ] + ret i32 %j.0.lcssa +} + +; CHECK: backedge-taken count is (-2147483632 + ((-1 + (-1 * %x)) smax (-1 + (-1 * %y)))) + diff --git a/test/Analysis/ScalarEvolution/2008-08-04-IVOverflow.ll b/test/Analysis/ScalarEvolution/2008-08-04-IVOverflow.ll new file mode 100644 index 0000000..3b31d79 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-08-04-IVOverflow.ll @@ -0,0 +1,27 @@ +; RUN: opt < %s -analyze -scalar-evolution \ +; RUN: -scalar-evolution-max-iterations=0 | FileCheck %s +; PR2621 + +define i32 @a() nounwind { +entry: + br label %bb1 + +bb: + trunc i32 %i.0 to i16 + add i16 %0, %x16.0 + add i32 %i.0, 1 + br label %bb1 + +bb1: + %i.0 = phi i32 [ 0, %entry ], [ %2, %bb ] + %x16.0 = phi i16 [ 0, %entry ], [ %1, %bb ] + icmp ult i32 %i.0, 888888 + br i1 %3, label %bb, label %bb2 + +bb2: + zext i16 %x16.0 to i32 + ret i32 %4 +} + +; CHECK: Exits: 20028 + diff --git a/test/Analysis/ScalarEvolution/2008-08-04-LongAddRec.ll b/test/Analysis/ScalarEvolution/2008-08-04-LongAddRec.ll new file mode 100644 index 0000000..b296a19 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-08-04-LongAddRec.ll @@ -0,0 +1,58 @@ +; RUN: opt < %s -analyze -scalar-evolution \ +; RUN: -scalar-evolution-max-iterations=0 | FileCheck %s +; PR2621 + +define i32 @a() nounwind { +entry: + br label %bb1 + +bb: ; preds = %bb1 + add i16 %x17.0, 1 ; <i16>:0 [#uses=2] + add i16 %0, %x16.0 ; <i16>:1 [#uses=2] + add i16 %1, %x15.0 ; <i16>:2 [#uses=2] + add i16 %2, %x14.0 ; <i16>:3 [#uses=2] + add i16 %3, %x13.0 ; <i16>:4 [#uses=2] + add i16 %4, %x12.0 ; <i16>:5 [#uses=2] + add i16 %5, %x11.0 ; <i16>:6 [#uses=2] + add i16 %6, %x10.0 ; <i16>:7 [#uses=2] + add i16 %7, %x9.0 ; <i16>:8 [#uses=2] + add i16 %8, %x8.0 ; <i16>:9 [#uses=2] + add i16 %9, %x7.0 ; <i16>:10 [#uses=2] + add i16 %10, %x6.0 ; <i16>:11 [#uses=2] + add i16 %11, %x5.0 ; <i16>:12 [#uses=2] + add i16 %12, %x4.0 ; <i16>:13 [#uses=2] + add i16 %13, %x3.0 ; <i16>:14 [#uses=2] + add i16 %14, %x2.0 ; <i16>:15 [#uses=2] + add i16 %15, %x1.0 ; <i16>:16 [#uses=1] + add i32 %i.0, 1 ; <i32>:17 [#uses=1] + br label %bb1 + +bb1: ; preds = %bb, %entry + %x2.0 = phi i16 [ 0, %entry ], [ %15, %bb ] ; <i16> [#uses=1] + %x3.0 = phi i16 [ 0, %entry ], [ %14, %bb ] ; <i16> [#uses=1] + %x4.0 = phi i16 [ 0, %entry ], [ %13, %bb ] ; <i16> [#uses=1] + %x5.0 = phi i16 [ 0, %entry ], [ %12, %bb ] ; <i16> [#uses=1] + %x6.0 = phi i16 [ 0, %entry ], [ %11, %bb ] ; <i16> [#uses=1] + %x7.0 = phi i16 [ 0, %entry ], [ %10, %bb ] ; <i16> [#uses=1] + %x8.0 = phi i16 [ 0, %entry ], [ %9, %bb ] ; <i16> [#uses=1] + %x9.0 = phi i16 [ 0, %entry ], [ %8, %bb ] ; <i16> [#uses=1] + %x10.0 = phi i16 [ 0, %entry ], [ %7, %bb ] ; <i16> [#uses=1] + %x11.0 = phi i16 [ 0, %entry ], [ %6, %bb ] ; <i16> [#uses=1] + %x12.0 = phi i16 [ 0, %entry ], [ %5, %bb ] ; <i16> [#uses=1] + %x13.0 = phi i16 [ 0, %entry ], [ %4, %bb ] ; <i16> [#uses=1] + %x14.0 = phi i16 [ 0, %entry ], [ %3, %bb ] ; <i16> [#uses=1] + %x15.0 = phi i16 [ 0, %entry ], [ %2, %bb ] ; <i16> [#uses=1] + %x16.0 = phi i16 [ 0, %entry ], [ %1, %bb ] ; <i16> [#uses=1] + %x17.0 = phi i16 [ 0, %entry ], [ %0, %bb ] ; <i16> [#uses=1] + %i.0 = phi i32 [ 0, %entry ], [ %17, %bb ] ; <i32> [#uses=2] + %x1.0 = phi i16 [ 0, %entry ], [ %16, %bb ] ; <i16> [#uses=2] + icmp ult i32 %i.0, 8888 ; <i1>:18 [#uses=1] + br i1 %18, label %bb, label %bb2 + +bb2: ; preds = %bb1 + zext i16 %x1.0 to i32 ; <i32>:19 [#uses=1] + ret i32 %19 +} + +; CHECK: Exits: -19168 + diff --git a/test/Analysis/ScalarEvolution/2008-11-02-QuadraticCrash.ll b/test/Analysis/ScalarEvolution/2008-11-02-QuadraticCrash.ll new file mode 100644 index 0000000..7722122 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-11-02-QuadraticCrash.ll @@ -0,0 +1,21 @@ +; RUN: opt < %s -analyze -scalar-evolution +; PR1827 + +declare void @use(i32) + +define void @foo() { +entry: + br label %loop_1 + +loop_1: ; preds = %loop_1, %entry + %a = phi i32 [ 2, %entry ], [ %b, %loop_1 ] ; <i32> [#uses=2] + %c = phi i32 [ 5, %entry ], [ %d, %loop_1 ] ; <i32> [#uses=1] + %b = add i32 %a, 1 ; <i32> [#uses=1] + %d = add i32 %c, %a ; <i32> [#uses=3] + %A = icmp ult i32 %d, 50 ; <i1> [#uses=1] + br i1 %A, label %loop_1, label %endloop + +endloop: ; preds = %loop_1 + call void @use(i32 %d) + ret void +} diff --git a/test/Analysis/ScalarEvolution/2008-11-15-CubicOOM.ll b/test/Analysis/ScalarEvolution/2008-11-15-CubicOOM.ll new file mode 100644 index 0000000..2e2aabc --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-11-15-CubicOOM.ll @@ -0,0 +1,19 @@ +; RUN: opt < %s -analyze -scalar-evolution +; PR2602 + +define i32 @a() nounwind { +entry: + br label %bb + +bb: ; preds = %bb, %entry + %w.0 = phi i32 [ 0, %entry ], [ %tmp, %bb ] ; <i32> [#uses=2] + %e.0 = phi i32 [ 0, %entry ], [ %e.1, %bb ] ; <i32> [#uses=2] + %w.1 = add i32 0, %w.0 ; <i32>:0 [#uses=1] + %tmp = add i32 %e.0, %w.0 ; <i32>:1 [#uses=1] + %e.1 = add i32 %e.0, 1 ; <i32>:2 [#uses=1] + %cond = icmp eq i32 %w.1, -1 ; <i1>:3 [#uses=1] + br i1 %cond, label %return, label %bb + +return: ; preds = %bb + ret i32 undef +} diff --git a/test/Analysis/ScalarEvolution/2008-11-18-LessThanOrEqual.ll b/test/Analysis/ScalarEvolution/2008-11-18-LessThanOrEqual.ll new file mode 100644 index 0000000..06637b5 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-11-18-LessThanOrEqual.ll @@ -0,0 +1,32 @@ +; RUN: opt < %s -analyze -scalar-evolution |& \ +; RUN: grep {Loop %bb: backedge-taken count is (7 + (-1 \\* %argc))} +; XFAIL: * + +define i32 @main(i32 %argc, i8** %argv) nounwind { +entry: + %0 = icmp ugt i32 %argc, 7 ; <i1> [#uses=1] + br i1 %0, label %bb2, label %bb.nph + +bb.nph: ; preds = %entry + br label %bb + +bb: ; preds = %bb.nph, %bb1 + %indvar = phi i32 [ 0, %bb.nph ], [ %indvar.next, %bb1 ] ; <i32> [#uses=2] + %argc_addr.04 = add i32 %indvar, %argc ; <i32> [#uses=1] + tail call void (...)* @Test() nounwind + %1 = add i32 %argc_addr.04, 1 ; <i32> [#uses=1] + br label %bb1 + +bb1: ; preds = %bb + %phitmp = icmp ugt i32 %1, 7 ; <i1> [#uses=1] + %indvar.next = add i32 %indvar, 1 ; <i32> [#uses=1] + br i1 %phitmp, label %bb1.bb2_crit_edge, label %bb + +bb1.bb2_crit_edge: ; preds = %bb1 + br label %bb2 + +bb2: ; preds = %bb1.bb2_crit_edge, %entry + ret i32 0 +} + +declare void @Test(...) diff --git a/test/Analysis/ScalarEvolution/2008-11-18-Stride1.ll b/test/Analysis/ScalarEvolution/2008-11-18-Stride1.ll new file mode 100644 index 0000000..db527fe --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-11-18-Stride1.ll @@ -0,0 +1,35 @@ +; RUN: opt < %s -analyze -scalar-evolution \ +; RUN: | grep {Loop %bb: Unpredictable backedge-taken count\\.} + +; ScalarEvolution can't compute a trip count because it doesn't know if +; dividing by the stride will have a remainder. This could theoretically +; be teaching it how to use a more elaborate trip count computation. + +define i32 @f(i32 %x) nounwind readnone { +entry: + %0 = icmp ugt i32 %x, 4 ; <i1> [#uses=1] + br i1 %0, label %bb.nph, label %bb2 + +bb.nph: ; preds = %entry + br label %bb + +bb: ; preds = %bb.nph, %bb1 + %indvar = phi i32 [ 0, %bb.nph ], [ %indvar.next, %bb1 ] ; <i32> [#uses=2] + %tmp = mul i32 %indvar, -3 ; <i32> [#uses=1] + %x_addr.04 = add i32 %tmp, %x ; <i32> [#uses=1] + %1 = add i32 %x_addr.04, -3 ; <i32> [#uses=2] + br label %bb1 + +bb1: ; preds = %bb + %2 = icmp ugt i32 %1, 4 ; <i1> [#uses=1] + %indvar.next = add i32 %indvar, 1 ; <i32> [#uses=1] + br i1 %2, label %bb, label %bb1.bb2_crit_edge + +bb1.bb2_crit_edge: ; preds = %bb1 + %.lcssa = phi i32 [ %1, %bb1 ] ; <i32> [#uses=1] + br label %bb2 + +bb2: ; preds = %bb1.bb2_crit_edge, %entry + %x_addr.0.lcssa = phi i32 [ %.lcssa, %bb1.bb2_crit_edge ], [ %x, %entry ] ; <i32> [#uses=1] + ret i32 %x_addr.0.lcssa +} diff --git a/test/Analysis/ScalarEvolution/2008-11-18-Stride2.ll b/test/Analysis/ScalarEvolution/2008-11-18-Stride2.ll new file mode 100644 index 0000000..102acc6 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-11-18-Stride2.ll @@ -0,0 +1,31 @@ +; RUN: opt < %s -analyze -scalar-evolution |& grep {/u 3} +; XFAIL: * + +define i32 @f(i32 %x) nounwind readnone { +entry: + %0 = icmp ugt i32 %x, 999 ; <i1> [#uses=1] + br i1 %0, label %bb2, label %bb.nph + +bb.nph: ; preds = %entry + br label %bb + +bb: ; preds = %bb.nph, %bb1 + %indvar = phi i32 [ 0, %bb.nph ], [ %indvar.next, %bb1 ] ; <i32> [#uses=2] + %tmp = mul i32 %indvar, 3 ; <i32> [#uses=1] + %x_addr.04 = add i32 %tmp, %x ; <i32> [#uses=1] + %1 = add i32 %x_addr.04, 3 ; <i32> [#uses=2] + br label %bb1 + +bb1: ; preds = %bb + %2 = icmp ugt i32 %1, 999 ; <i1> [#uses=1] + %indvar.next = add i32 %indvar, 1 ; <i32> [#uses=1] + br i1 %2, label %bb1.bb2_crit_edge, label %bb + +bb1.bb2_crit_edge: ; preds = %bb1 + %.lcssa = phi i32 [ %1, %bb1 ] ; <i32> [#uses=1] + br label %bb2 + +bb2: ; preds = %bb1.bb2_crit_edge, %entry + %x_addr.0.lcssa = phi i32 [ %.lcssa, %bb1.bb2_crit_edge ], [ %x, %entry ] ; <i32> [#uses=1] + ret i32 %x_addr.0.lcssa +} diff --git a/test/Analysis/ScalarEvolution/2008-12-08-FiniteSGE.ll b/test/Analysis/ScalarEvolution/2008-12-08-FiniteSGE.ll new file mode 100644 index 0000000..226221b --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-12-08-FiniteSGE.ll @@ -0,0 +1,25 @@ +; RUN: opt < %s -analyze -scalar-evolution | grep {backedge-taken count is 255} +; XFAIL: * + +define i32 @foo(i32 %x, i32 %y, i32* %lam, i32* %alp) nounwind { +bb1.thread: + br label %bb1 + +bb1: ; preds = %bb1, %bb1.thread + %indvar = phi i32 [ 0, %bb1.thread ], [ %indvar.next, %bb1 ] ; <i32> [#uses=4] + %i.0.reg2mem.0 = sub i32 255, %indvar ; <i32> [#uses=2] + %0 = getelementptr i32* %alp, i32 %i.0.reg2mem.0 ; <i32*> [#uses=1] + %1 = load i32* %0, align 4 ; <i32> [#uses=1] + %2 = getelementptr i32* %lam, i32 %i.0.reg2mem.0 ; <i32*> [#uses=1] + store i32 %1, i32* %2, align 4 + %3 = sub i32 254, %indvar ; <i32> [#uses=1] + %4 = icmp slt i32 %3, 0 ; <i1> [#uses=1] + %indvar.next = add i32 %indvar, 1 ; <i32> [#uses=1] + br i1 %4, label %bb2, label %bb1 + +bb2: ; preds = %bb1 + %tmp10 = mul i32 %indvar, %x ; <i32> [#uses=1] + %z.0.reg2mem.0 = add i32 %tmp10, %y ; <i32> [#uses=1] + %5 = add i32 %z.0.reg2mem.0, %x ; <i32> [#uses=1] + ret i32 %5 +} diff --git a/test/Analysis/ScalarEvolution/2008-12-11-SMaxOverflow.ll b/test/Analysis/ScalarEvolution/2008-12-11-SMaxOverflow.ll new file mode 100644 index 0000000..33a7479 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-12-11-SMaxOverflow.ll @@ -0,0 +1,28 @@ +; RUN: opt < %s -analyze -scalar-evolution | grep {0 smax} +; XFAIL: * + +define i32 @f(i32 %c.idx.val) { + +bb2: + %k.018 = add i32 %c.idx.val, -1 ; <i32> [#uses=2] + %a14 = icmp slt i32 %k.018, 0 ; <i1> [#uses=1] + br i1 %a14, label %bb19, label %bb16.preheader + +bb16.preheader: + %k.019 = phi i32 [ %k.0, %bb18 ], [ %k.018, %bb2 ] ; <i32> [#uses=5] + %x = phi i32 [ 0, %bb2 ], [ %x.1, %bb18] + br label %bb18 + +bb18: ; preds = %bb18.loopexit + %x.1 = add i32 %x, 1 + %k.0 = add i32 %k.019, -1 ; <i32> [#uses=2] + %a107 = icmp slt i32 %k.0, 0 ; <i1> [#uses=1] + br i1 %a107, label %bb18.bb19_crit_edge, label %bb16.preheader + +bb18.bb19_crit_edge: + ret i32 %x + +bb19: + ret i32 0 + +} diff --git a/test/Analysis/ScalarEvolution/2008-12-14-StrideAndSigned.ll b/test/Analysis/ScalarEvolution/2008-12-14-StrideAndSigned.ll new file mode 100644 index 0000000..8152e98 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-12-14-StrideAndSigned.ll @@ -0,0 +1,22 @@ +; RUN: opt < %s -analyze -scalar-evolution |& \ +; RUN: grep {(((-1 \\* %i0) + (100005 smax %i0)) /u 5)} +; XFAIL: * + +define i32 @foo0(i32 %i0) nounwind { +entry: + br label %bb1 + +bb: ; preds = %bb1 + %0 = add i32 %j.0, 1 ; <i32> [#uses=1] + %1 = add i32 %i.0, 5 ; <i32> [#uses=1] + br label %bb1 + +bb1: ; preds = %bb, %entry + %j.0 = phi i32 [ 0, %entry ], [ %0, %bb ] ; <i32> [#uses=2] + %i.0 = phi i32 [ %i0, %entry ], [ %1, %bb ] ; <i32> [#uses=2] + %2 = icmp sgt i32 %i.0, 100000 ; <i1> [#uses=1] + br i1 %2, label %return, label %bb + +return: ; preds = %bb1 + ret i32 %j.0 +} diff --git a/test/Analysis/ScalarEvolution/2008-12-15-DontUseSDiv.ll b/test/Analysis/ScalarEvolution/2008-12-15-DontUseSDiv.ll new file mode 100644 index 0000000..3eaa492 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2008-12-15-DontUseSDiv.ll @@ -0,0 +1,21 @@ +; RUN: opt < %s -analyze -scalar-evolution |& grep {/u 5} +; XFAIL: * + +define i8 @foo0(i8 %i0) nounwind { +entry: + br label %bb1 + +bb: ; preds = %bb1 + %0 = add i8 %j.0, 1 ; <i8> [#uses=1] + %1 = add i8 %i.0, 5 ; <i8> [#uses=1] + br label %bb1 + +bb1: ; preds = %bb, %entry + %j.0 = phi i8 [ 0, %entry ], [ %0, %bb ] ; <i8> [#uses=2] + %i.0 = phi i8 [ %i0, %entry ], [ %1, %bb ] ; <i8> [#uses=2] + %2 = icmp sgt i8 %i.0, 100 ; <i1> [#uses=1] + br i1 %2, label %return, label %bb + +return: ; preds = %bb1 + ret i8 %j.0 +} diff --git a/test/Analysis/ScalarEvolution/2009-01-02-SignedNegativeStride.ll b/test/Analysis/ScalarEvolution/2009-01-02-SignedNegativeStride.ll new file mode 100644 index 0000000..cc2a2e4 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2009-01-02-SignedNegativeStride.ll @@ -0,0 +1,40 @@ +; RUN: opt < %s -analyze -scalar-evolution | not grep {/u -1} +; PR3275 + +@g_16 = external global i16 ; <i16*> [#uses=3] +@.str = external constant [4 x i8] ; <[4 x i8]*> [#uses=0] + +define void @func_15() nounwind { +entry: + %0 = load i16* @g_16, align 2 ; <i16> [#uses=1] + %1 = icmp sgt i16 %0, 0 ; <i1> [#uses=1] + br i1 %1, label %bb2, label %bb.nph + +bb.nph: ; preds = %entry + %g_16.promoted = load i16* @g_16 ; <i16> [#uses=1] + br label %bb + +bb: ; preds = %bb1, %bb.nph + %g_16.tmp.0 = phi i16 [ %g_16.promoted, %bb.nph ], [ %2, %bb1 ] ; <i16> [#uses=1] + %2 = add i16 %g_16.tmp.0, -1 ; <i16> [#uses=3] + br label %bb1 + +bb1: ; preds = %bb + %3 = icmp sgt i16 %2, 0 ; <i1> [#uses=1] + br i1 %3, label %bb1.bb2_crit_edge, label %bb + +bb1.bb2_crit_edge: ; preds = %bb1 + store i16 %2, i16* @g_16 + br label %bb2 + +bb2: ; preds = %bb1.bb2_crit_edge, %entry + br label %return + +return: ; preds = %bb2 + ret void +} + +declare i32 @main() nounwind + +declare i32 @printf(i8*, ...) nounwind + diff --git a/test/Analysis/ScalarEvolution/2009-04-22-TruncCast.ll b/test/Analysis/ScalarEvolution/2009-04-22-TruncCast.ll new file mode 100644 index 0000000..c2e108a --- /dev/null +++ b/test/Analysis/ScalarEvolution/2009-04-22-TruncCast.ll @@ -0,0 +1,37 @@ +; RUN: opt < %s -analyze -scalar-evolution | grep {(trunc i} | not grep ext + +define i16 @test1(i8 %x) { + %A = sext i8 %x to i32 + %B = trunc i32 %A to i16 + ret i16 %B +} + +define i8 @test2(i16 %x) { + %A = sext i16 %x to i32 + %B = trunc i32 %A to i8 + ret i8 %B +} + +define i16 @test3(i16 %x) { + %A = sext i16 %x to i32 + %B = trunc i32 %A to i16 + ret i16 %B +} + +define i16 @test4(i8 %x) { + %A = zext i8 %x to i32 + %B = trunc i32 %A to i16 + ret i16 %B +} + +define i8 @test5(i16 %x) { + %A = zext i16 %x to i32 + %B = trunc i32 %A to i8 + ret i8 %B +} + +define i16 @test6(i16 %x) { + %A = zext i16 %x to i32 + %B = trunc i32 %A to i16 + ret i16 %B +} diff --git a/test/Analysis/ScalarEvolution/2009-05-09-PointerEdgeCount.ll b/test/Analysis/ScalarEvolution/2009-05-09-PointerEdgeCount.ll new file mode 100644 index 0000000..dc7bd29 --- /dev/null +++ b/test/Analysis/ScalarEvolution/2009-05-09-PointerEdgeCount.ll @@ -0,0 +1,28 @@ +; RUN: opt < %s -analyze -scalar-evolution | grep {count is 2} +; PR3171 +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" + + %struct.Foo = type { i32 } + %struct.NonPod = type { [2 x %struct.Foo] } + +define void @_Z3foov() nounwind { +entry: + %x = alloca %struct.NonPod, align 8 ; <%struct.NonPod*> [#uses=2] + %0 = getelementptr %struct.NonPod* %x, i32 0, i32 0 ; <[2 x %struct.Foo]*> [#uses=1] + %1 = getelementptr [2 x %struct.Foo]* %0, i32 1, i32 0 ; <%struct.Foo*> [#uses=1] + br label %bb1.i + +bb1.i: ; preds = %bb2.i, %entry + %.0.i = phi %struct.Foo* [ %1, %entry ], [ %4, %bb2.i ] ; <%struct.Foo*> [#uses=2] + %2 = getelementptr %struct.NonPod* %x, i32 0, i32 0, i32 0 ; <%struct.Foo*> [#uses=1] + %3 = icmp eq %struct.Foo* %.0.i, %2 ; <i1> [#uses=1] + br i1 %3, label %_ZN6NonPodD1Ev.exit, label %bb2.i + +bb2.i: ; preds = %bb1.i + %4 = getelementptr %struct.Foo* %.0.i, i32 -1 ; <%struct.Foo*> [#uses=1] + br label %bb1.i + +_ZN6NonPodD1Ev.exit: ; preds = %bb1.i + ret void +} + diff --git a/test/Analysis/ScalarEvolution/2009-07-04-GroupConstantsWidthMismatch.ll b/test/Analysis/ScalarEvolution/2009-07-04-GroupConstantsWidthMismatch.ll new file mode 100644 index 0000000..a4358aa --- /dev/null +++ b/test/Analysis/ScalarEvolution/2009-07-04-GroupConstantsWidthMismatch.ll @@ -0,0 +1,16 @@ +; RUN: opt < %s -analyze -scalar-evolution +; PR4501 + +define void @test() { +entry: + %0 = load i16* undef, align 1 + %1 = lshr i16 %0, 8 + %2 = and i16 %1, 3 + %3 = zext i16 %2 to i32 + %4 = load i8* undef, align 1 + %5 = lshr i8 %4, 4 + %6 = and i8 %5, 1 + %7 = zext i8 %6 to i32 + %t1 = add i32 %3, %7 + ret void +} diff --git a/test/Analysis/ScalarEvolution/SolveQuadraticEquation.ll b/test/Analysis/ScalarEvolution/SolveQuadraticEquation.ll new file mode 100644 index 0000000..9573aed --- /dev/null +++ b/test/Analysis/ScalarEvolution/SolveQuadraticEquation.ll @@ -0,0 +1,32 @@ +; RUN: opt < %s -analyze -scalar-evolution \ +; RUN: -scalar-evolution-max-iterations=0 | grep {backedge-taken count is 100} +; PR1101 + +@A = weak global [1000 x i32] zeroinitializer, align 32 + + +define void @test(i32 %N) { +entry: + "alloca point" = bitcast i32 0 to i32 ; <i32> [#uses=0] + br label %bb3 + +bb: ; preds = %bb3 + %tmp = getelementptr [1000 x i32]* @A, i32 0, i32 %i.0 ; <i32*> [#uses=1] + store i32 123, i32* %tmp + %tmp2 = add i32 %i.0, 1 ; <i32> [#uses=1] + br label %bb3 + +bb3: ; preds = %bb, %entry + %i.0 = phi i32 [ 2, %entry ], [ %tmp2, %bb ] ; <i32> [#uses=3] + %SQ = mul i32 %i.0, %i.0 + %tmp4 = mul i32 %i.0, 2 + %tmp5 = sub i32 %SQ, %tmp4 + %tmp3 = icmp sle i32 %tmp5, 9999 ; <i1> [#uses=1] + br i1 %tmp3, label %bb, label %bb5 + +bb5: ; preds = %bb3 + br label %return + +return: ; preds = %bb5 + ret void +} diff --git a/test/Analysis/ScalarEvolution/and-xor.ll b/test/Analysis/ScalarEvolution/and-xor.ll new file mode 100644 index 0000000..1772573 --- /dev/null +++ b/test/Analysis/ScalarEvolution/and-xor.ll @@ -0,0 +1,8 @@ +; RUN: opt < %s -scalar-evolution -analyze \ +; RUN: | grep {\\--> (zext} | count 2 + +define i32 @foo(i32 %x) { + %n = and i32 %x, 255 + %y = xor i32 %n, 255 + ret i32 %y +} diff --git a/test/Analysis/ScalarEvolution/avoid-infinite-recursion-0.ll b/test/Analysis/ScalarEvolution/avoid-infinite-recursion-0.ll new file mode 100644 index 0000000..7eeb308 --- /dev/null +++ b/test/Analysis/ScalarEvolution/avoid-infinite-recursion-0.ll @@ -0,0 +1,30 @@ +; RUN: opt < %s -analyze -scalar-evolution +; PR4537 + +; ModuleID = 'b.bc' +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" +target triple = "x86_64-unknown-linux-gnu" + +define i32 @test() { +entry: + %0 = load i32** undef, align 8 ; <i32*> [#uses=1] + %1 = ptrtoint i32* %0 to i64 ; <i64> [#uses=1] + %2 = sub i64 undef, %1 ; <i64> [#uses=1] + %3 = lshr i64 %2, 3 ; <i64> [#uses=1] + %4 = trunc i64 %3 to i32 ; <i32> [#uses=2] + br i1 undef, label %bb10, label %bb4.i + +bb4.i: ; preds = %bb4.i, %entry + %i.0.i6 = phi i32 [ %8, %bb4.i ], [ 0, %entry ] ; <i32> [#uses=2] + %5 = sub i32 %4, %i.0.i6 ; <i32> [#uses=1] + %6 = sext i32 %5 to i64 ; <i64> [#uses=1] + %7 = udiv i64 undef, %6 ; <i64> [#uses=1] + %8 = add i32 %i.0.i6, 1 ; <i32> [#uses=2] + %phitmp = icmp eq i64 %7, 0 ; <i1> [#uses=1] + %.not.i = icmp sge i32 %8, %4 ; <i1> [#uses=1] + %or.cond.i = or i1 %phitmp, %.not.i ; <i1> [#uses=1] + br i1 %or.cond.i, label %bb10, label %bb4.i + +bb10: ; preds = %bb4.i, %entry + unreachable +} diff --git a/test/Analysis/ScalarEvolution/avoid-infinite-recursion-1.ll b/test/Analysis/ScalarEvolution/avoid-infinite-recursion-1.ll new file mode 100644 index 0000000..31b95e1 --- /dev/null +++ b/test/Analysis/ScalarEvolution/avoid-infinite-recursion-1.ll @@ -0,0 +1,354 @@ +; RUN: opt < %s -iv-users +; PR4538 + +; ModuleID = 'bugpoint-reduced-simplified.bc' +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" +target triple = "x86_64-unknown-freebsd8.0" +module asm ".ident\09\22$FreeBSD: head/sys/kern/vfs_subr.c 195285 2009-07-02 14:19:33Z jamie $\22" +module asm ".section set_pcpu, \22aw\22, @progbits" +module asm ".previous" + type <{ [40 x i8] }> ; type %0 + type <{ %struct.vm_object*, %struct.vm_object** }> ; type %1 + type <{ %struct.vm_object* }> ; type %2 + type <{ %struct.vm_page*, %struct.vm_page** }> ; type %3 + type <{ %struct.pv_entry*, %struct.pv_entry** }> ; type %4 + type <{ %struct.vm_reserv* }> ; type %5 + type <{ %struct.bufobj*, %struct.bufobj** }> ; type %6 + type <{ %struct.proc*, %struct.proc** }> ; type %7 + type <{ %struct.thread*, %struct.thread** }> ; type %8 + type <{ %struct.prison*, %struct.prison** }> ; type %9 + type <{ %struct.prison* }> ; type %10 + type <{ %struct.task* }> ; type %11 + type <{ %struct.osd*, %struct.osd** }> ; type %12 + type <{ %struct.proc* }> ; type %13 + type <{ %struct.ksiginfo*, %struct.ksiginfo** }> ; type %14 + type <{ %struct.pv_chunk*, %struct.pv_chunk** }> ; type %15 + type <{ %struct.pgrp*, %struct.pgrp** }> ; type %16 + type <{ %struct.knote*, %struct.knote** }> ; type %17 + type <{ %struct.ktr_request*, %struct.ktr_request** }> ; type %18 + type <{ %struct.mqueue_notifier* }> ; type %19 + type <{ %struct.turnstile* }> ; type %20 + type <{ %struct.namecache* }> ; type %21 + type <{ %struct.namecache*, %struct.namecache** }> ; type %22 + type <{ %struct.lockf*, %struct.lockf** }> ; type %23 + type <{ %struct.lockf_entry*, %struct.lockf_entry** }> ; type %24 + type <{ %struct.lockf_edge*, %struct.lockf_edge** }> ; type %25 + %struct.__siginfo = type <{ i32, i32, i32, i32, i32, i32, i8*, %union.sigval, %0 }> + %struct.__sigset = type <{ [4 x i32] }> + %struct.acl = type <{ i32, i32, [4 x i32], [254 x %struct.acl_entry] }> + %struct.acl_entry = type <{ i32, i32, i32, i16, i16 }> + %struct.au_mask = type <{ i32, i32 }> + %struct.au_tid_addr = type <{ i32, i32, [4 x i32] }> + %struct.auditinfo_addr = type <{ i32, %struct.au_mask, %struct.au_tid_addr, i32, i64 }> + %struct.bintime = type <{ i64, i64 }> + %struct.buf = type <{ %struct.bufobj*, i64, i8*, i8*, i32, i8, i8, i8, i8, i64, i64, void (%struct.buf*)*, i64, i64, %struct.buflists, %struct.buf*, %struct.buf*, i32, i8, i8, i8, i8, %struct.buflists, i16, i8, i8, i32, i8, i8, i8, i8, i8, i8, i8, i8, %struct.lock, i64, i64, i8*, i32, i8, i8, i8, i8, i64, %struct.vnode*, i32, i32, %struct.ucred*, %struct.ucred*, i8*, %union.pager_info, i8, i8, i8, i8, %union.anon, [32 x %struct.vm_page*], i32, i8, i8, i8, i8, %struct.workhead, i8*, i8*, i8*, i32, i8, i8, i8, i8 }> + %struct.buf_ops = type <{ i8*, i32 (%struct.buf*)*, void (%struct.bufobj*, %struct.buf*)*, i32 (%struct.bufobj*, i32)*, void (%struct.bufobj*, %struct.buf*)* }> + %struct.buflists = type <{ %struct.buf*, %struct.buf** }> + %struct.bufobj = type <{ %struct.mtx, %struct.bufv, %struct.bufv, i64, i32, i8, i8, i8, i8, %struct.buf_ops*, i32, i8, i8, i8, i8, %struct.vm_object*, %6, i8*, %struct.vnode* }> + %struct.bufv = type <{ %struct.buflists, %struct.buf*, i32, i8, i8, i8, i8 }> + %struct.callout = type <{ %union.anon, i32, i8, i8, i8, i8, i8*, void (i8*)*, %struct.lock_object*, i32, i32 }> + %struct.cdev_privdata = type opaque + %struct.cluster_save = type <{ i64, i64, i8*, i32, i8, i8, i8, i8, %struct.buf** }> + %struct.componentname = type <{ i64, i64, %struct.thread*, %struct.ucred*, i32, i8, i8, i8, i8, i8*, i8*, i64, i64 }> + %struct.cpuset = type opaque + %struct.cv = type <{ i8*, i32, i8, i8, i8, i8 }> + %struct.fid = type <{ i16, i16, [16 x i8] }> + %struct.file = type <{ i8*, %struct.fileops*, %struct.ucred*, %struct.vnode*, i16, i16, i32, i32, i32, i64, %struct.cdev_privdata*, i64, i8* }> + %struct.filedesc = type opaque + %struct.filedesc_to_leader = type opaque + %struct.fileops = type <{ i32 (%struct.file*, %struct.uio*, %struct.ucred*, i32, %struct.thread*)*, i32 (%struct.file*, %struct.uio*, %struct.ucred*, i32, %struct.thread*)*, i32 (%struct.file*, i64, %struct.ucred*, %struct.thread*)*, i32 (%struct.file*, i64, i8*, %struct.ucred*, %struct.thread*)*, i32 (%struct.file*, i32, %struct.ucred*, %struct.thread*)*, i32 (%struct.file*, %struct.knote*)*, i32 (%struct.file*, %struct.stat*, %struct.ucred*, %struct.thread*)*, i32 (%struct.file*, %struct.thread*)*, i32, i8, i8, i8, i8 }> + %struct.filterops = type <{ i32, i8, i8, i8, i8, i32 (%struct.knote*)*, void (%struct.knote*)*, i32 (%struct.knote*, i64)* }> + %struct.flock = type <{ i64, i64, i32, i16, i16, i32, i8, i8, i8, i8 }> + %struct.freelst = type <{ %struct.vnode*, %struct.vnode** }> + %struct.fsid = type <{ [2 x i32] }> + %struct.in6_addr = type opaque + %struct.in_addr = type opaque + %struct.inode = type opaque + %struct.iovec = type <{ i8*, i64 }> + %struct.itimers = type opaque + %struct.itimerval = type <{ %struct.bintime, %struct.bintime }> + %struct.kaioinfo = type opaque + %struct.kaudit_record = type opaque + %struct.kdtrace_proc = type opaque + %struct.kdtrace_thread = type opaque + %struct.kevent = type <{ i64, i16, i16, i32, i64, i8* }> + %struct.klist = type <{ %struct.knote* }> + %struct.knlist = type <{ %struct.klist, void (i8*)*, void (i8*)*, void (i8*)*, void (i8*)*, i8* }> + %struct.knote = type <{ %struct.klist, %struct.klist, %struct.knlist*, %17, %struct.kqueue*, %struct.kevent, i32, i32, i64, %union.sigval, %struct.filterops*, i8* }> + %struct.kqueue = type opaque + %struct.ksiginfo = type <{ %14, %struct.__siginfo, i32, i8, i8, i8, i8, %struct.sigqueue* }> + %struct.ktr_request = type opaque + %struct.label = type opaque + %struct.lock = type <{ %struct.lock_object, i64, i32, i32 }> + %struct.lock_list_entry = type opaque + %struct.lock_object = type <{ i8*, i32, i32, %struct.witness* }> + %struct.lock_owner = type opaque + %struct.lock_profile_object = type opaque + %struct.lockf = type <{ %23, %struct.mtx, %struct.lockf_entry_list, %struct.lockf_entry_list, i32, i8, i8, i8, i8 }> + %struct.lockf_edge = type <{ %25, %25, %struct.lockf_entry*, %struct.lockf_entry* }> + %struct.lockf_edge_list = type <{ %struct.lockf_edge* }> + %struct.lockf_entry = type <{ i16, i16, i8, i8, i8, i8, i64, i64, %struct.lock_owner*, %struct.vnode*, %struct.inode*, %struct.task*, %24, %struct.lockf_edge_list, %struct.lockf_edge_list, i32, i8, i8, i8, i8 }> + %struct.lockf_entry_list = type <{ %struct.lockf_entry* }> + %struct.lpohead = type <{ %struct.lock_profile_object* }> + %struct.md_page = type <{ %4 }> + %struct.mdproc = type <{ %struct.cv*, %struct.system_segment_descriptor }> + %struct.mdthread = type <{ i32, i8, i8, i8, i8, i64 }> + %struct.mntarg = type opaque + %struct.mntlist = type <{ %struct.mount*, %struct.mount** }> + %struct.mount = type <{ %struct.mtx, i32, i8, i8, i8, i8, %struct.mntlist, %struct.vfsops*, %struct.vfsconf*, %struct.vnode*, %struct.vnode*, i32, i8, i8, i8, i8, %struct.freelst, i32, i32, i32, i32, i32, i32, %struct.vfsoptlist*, %struct.vfsoptlist*, i32, i8, i8, i8, i8, %struct.statfs, %struct.ucred*, i8*, i64, i32, i8, i8, i8, i8, %struct.netexport*, %struct.label*, i32, i32, i32, i32, %struct.thread*, i8*, %struct.lock }> + %struct.mqueue_notifier = type opaque + %struct.mtx = type <{ %struct.lock_object, i64 }> + %struct.namecache = type opaque + %struct.netexport = type opaque + %struct.nlminfo = type opaque + %struct.osd = type <{ i32, i8, i8, i8, i8, i8**, %12 }> + %struct.p_sched = type opaque + %struct.pargs = type <{ i32, i32, [1 x i8], i8, i8, i8 }> + %struct.pcb = type opaque + %struct.pgrp = type <{ %16, %13, %struct.session*, %struct.sigiolst, i32, i32, %struct.mtx }> + %struct.plimit = type opaque + %struct.pmap = type <{ %struct.mtx, i64*, %15, i32, i8, i8, i8, i8, %struct.bintime, %struct.vm_page* }> + %struct.prison = type <{ %9, i32, i32, i32, i32, %10, %9, %struct.prison*, %struct.mtx, %struct.task, %struct.osd, %struct.cpuset*, %struct.vnet*, %struct.vnode*, i32, i32, %struct.in_addr*, %struct.in6_addr*, [4 x i8*], i32, i32, i32, i32, i32, [5 x i32], i64, [256 x i8], [1024 x i8], [256 x i8], [256 x i8], [64 x i8] }> + %struct.proc = type <{ %7, %8, %struct.mtx, %struct.ucred*, %struct.filedesc*, %struct.filedesc_to_leader*, %struct.pstats*, %struct.plimit*, %struct.callout, %struct.sigacts*, i32, i32, i32, i8, i8, i8, i8, %7, %7, %struct.proc*, %7, %13, %struct.mtx, %struct.ksiginfo*, %struct.sigqueue, i32, i8, i8, i8, i8, %struct.vmspace*, i32, i8, i8, i8, i8, %struct.itimerval, %struct.rusage, %struct.rusage_ext, %struct.rusage_ext, i32, i32, i32, i8, i8, i8, i8, %struct.vnode*, %struct.ucred*, %struct.vnode*, i32, i8, i8, i8, i8, %struct.sigiolst, i32, i32, i64, i32, i32, i8, i8, i8, i8, i8, i8, i8, i8, %struct.nlminfo*, %struct.kaioinfo*, %struct.thread*, i32, i8, i8, i8, i8, %struct.thread*, i32, i32, %struct.itimers*, i32, i32, [20 x i8], i8, i8, i8, i8, %struct.pgrp*, %struct.sysentvec*, %struct.pargs*, i64, i8, i8, i8, i8, i32, i16, i8, i8, i8, i8, i8, i8, %struct.knlist, i32, i8, i8, i8, i8, %struct.mdproc, %struct.callout, i16, i8, i8, i8, i8, i8, i8, %struct.proc*, %struct.proc*, i8*, %struct.label*, %struct.p_sched*, %18, %19, %struct.kdtrace_proc*, %struct.cv }> + %struct.pstats = type opaque + %struct.pv_chunk = type <{ %struct.pmap*, %15, [3 x i64], [2 x i64], [168 x %struct.pv_entry] }> + %struct.pv_entry = type <{ i64, %4 }> + %struct.rusage = type <{ %struct.bintime, %struct.bintime, i64, i64, i64, i64, i64, i64, i64, i64, i64, i64, i64, i64, i64, i64 }> + %struct.rusage_ext = type <{ i64, i64, i64, i64, i64, i64, i64 }> + %struct.selfd = type opaque + %struct.selfdlist = type <{ %struct.selfd*, %struct.selfd** }> + %struct.selinfo = type <{ %struct.selfdlist, %struct.knlist, %struct.mtx* }> + %struct.seltd = type opaque + %struct.session = type <{ i32, i8, i8, i8, i8, %struct.proc*, %struct.vnode*, %struct.tty*, i32, [24 x i8], i8, i8, i8, i8, %struct.mtx }> + %struct.shmmap_state = type opaque + %struct.sigacts = type <{ [128 x void (i32)*], [128 x %struct.__sigset], %struct.__sigset, %struct.__sigset, %struct.__sigset, %struct.__sigset, %struct.__sigset, %struct.__sigset, %struct.__sigset, %struct.__sigset, %struct.__sigset, %struct.__sigset, i32, i32, %struct.mtx }> + %struct.sigaltstack = type <{ i8*, i64, i32, i8, i8, i8, i8 }> + %struct.sigio = type <{ %union.sigval, %struct.sigiolst, %struct.sigio**, %struct.ucred*, i32, i8, i8, i8, i8 }> + %struct.sigiolst = type <{ %struct.sigio* }> + %struct.sigqueue = type <{ %struct.__sigset, %struct.__sigset, %14, %struct.proc*, i32, i8, i8, i8, i8 }> + %struct.sleepqueue = type opaque + %struct.sockaddr = type opaque + %struct.stat = type <{ i32, i32, i16, i16, i32, i32, i32, %struct.bintime, %struct.bintime, %struct.bintime, i64, i64, i32, i32, i32, i32, %struct.bintime }> + %struct.statfs = type <{ i32, i32, i64, i64, i64, i64, i64, i64, i64, i64, i64, i64, i64, i64, [10 x i64], i32, i32, %struct.fsid, [80 x i8], [16 x i8], [88 x i8], [88 x i8] }> + %struct.sysctl_req = type <{ %struct.thread*, i32, i8, i8, i8, i8, i8*, i64, i64, i32 (%struct.sysctl_req*, i8*, i64)*, i8*, i64, i64, i32 (%struct.sysctl_req*, i8*, i64)*, i64, i32, i8, i8, i8, i8 }> + %struct.sysentvec = type opaque + %struct.system_segment_descriptor = type <{ i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8 }> + %struct.task = type <{ %11, i16, i16, i8, i8, i8, i8, void (i8*, i32)*, i8* }> + %struct.td_sched = type opaque + %struct.thread = type <{ %struct.mtx*, %struct.proc*, %8, %8, %8, %8, %struct.cpuset*, %struct.seltd*, %struct.sleepqueue*, %struct.turnstile*, %struct.umtx_q*, i32, i8, i8, i8, i8, %struct.sigqueue, i32, i32, i32, i32, i32, i8, i8, i8, i8, i8*, i8*, i8, i8, i8, i8, i16, i16, i16, i8, i8, i8, i8, i8, i8, %struct.turnstile*, i8*, %20, %struct.lock_list_entry*, i32, i32, %struct.ucred*, i32, i32, %struct.rusage, i64, i64, i32, i32, i32, i32, i32, %struct.__sigset, %struct.__sigset, i32, %struct.sigaltstack, i32, i8, i8, i8, i8, i64, i32, [20 x i8], %struct.file*, i32, i32, %struct.osd, i8, i8, i8, i8, i8, i8, i8, i8, %struct.pcb*, i32, i8, i8, i8, i8, [2 x i64], %struct.callout, %struct.trapframe*, %struct.vm_object*, i64, i32, i8, i8, i8, i8, %struct.vm_object*, i64, i32, i32, %struct.mdthread, %struct.td_sched*, %struct.kaudit_record*, i32, i8, i8, i8, i8, [2 x %struct.lpohead], %struct.kdtrace_thread*, i32, i8, i8, i8, i8, %struct.vnet*, i8* }> + %struct.trapframe = type <{ i64, i64, i64, i64, i64, i64, i64, i64, i64, i64, i64, i64, i64, i64, i64, i32, i16, i16, i64, i32, i16, i16, i64, i64, i64, i64, i64, i64 }> + %struct.tty = type opaque + %struct.turnstile = type opaque + %struct.ucred = type <{ i32, i32, i32, i32, i32, i32, i32, i8, i8, i8, i8, %struct.uidinfo*, %struct.uidinfo*, %struct.prison*, %struct.vimage*, i32, i8, i8, i8, i8, [2 x i8*], %struct.label*, %struct.auditinfo_addr, i32*, i32, i8, i8, i8, i8 }> + %struct.uidinfo = type opaque + %struct.uio = type <{ %struct.iovec*, i32, i8, i8, i8, i8, i64, i64, i32, i32, %struct.thread* }> + %struct.umtx_q = type opaque + %struct.vattr = type <{ i32, i16, i16, i32, i32, i32, i8, i8, i8, i8, i64, i64, i64, %struct.bintime, %struct.bintime, %struct.bintime, %struct.bintime, i64, i64, i32, i8, i8, i8, i8, i64, i64, i32, i8, i8, i8, i8, i64 }> + %struct.vfsconf = type <{ i32, [16 x i8], i8, i8, i8, i8, %struct.vfsops*, i32, i32, i32, i8, i8, i8, i8, %struct.vfsoptdecl*, %struct.vfsconfhead }> + %struct.vfsconfhead = type <{ %struct.vfsconf*, %struct.vfsconf** }> + %struct.vfsops = type <{ i32 (%struct.mount*)*, i32 (%struct.mntarg*, i8*, i32)*, i32 (%struct.mount*, i32)*, i32 (%struct.mount*, i32, %struct.vnode**)*, i32 (%struct.mount*, i32, i32, i8*)*, i32 (%struct.mount*, %struct.statfs*)*, i32 (%struct.mount*, i32)*, i32 (%struct.mount*, i32, i32, %struct.vnode**)*, i32 (%struct.mount*, %struct.fid*, %struct.vnode**)*, i32 (%struct.mount*, %struct.sockaddr*, i32*, %struct.ucred**, i32*, i32**)*, i32 (%struct.vfsconf*)*, i32 (%struct.vfsconf*)*, i32 (%struct.mount*, i32, %struct.vnode*, i32, i8*)*, i32 (%struct.mount*, i32, %struct.sysctl_req*)*, void (%struct.mount*)* }> + %struct.vfsopt = type <{ %struct.vfsoptlist, i8*, i8*, i32, i32, i32, i8, i8, i8, i8 }> + %struct.vfsoptdecl = type opaque + %struct.vfsoptlist = type <{ %struct.vfsopt*, %struct.vfsopt** }> + %struct.vimage = type opaque + %struct.vm_map = type <{ %struct.vm_map_entry, %struct.mtx, %struct.mtx, i32, i8, i8, i8, i8, i64, i32, i8, i8, i8, i8, %struct.vm_map_entry*, %struct.pmap*, %struct.vm_map_entry* }> + %struct.vm_map_entry = type <{ %struct.vm_map_entry*, %struct.vm_map_entry*, %struct.vm_map_entry*, %struct.vm_map_entry*, i64, i64, i64, i64, i64, %union.sigval, i64, i32, i8, i8, i8, i8, i32, i8, i8, i8, i8, i64, %struct.uidinfo* }> + %struct.vm_object = type <{ %struct.mtx, %1, %2, %1, %3, %struct.vm_page*, i64, i32, i32, i32, i8, i8, i16, i16, i16, i32, %struct.vm_object*, i64, %1, %5, %struct.vm_page*, i8*, %union.anon, %struct.uidinfo*, i64 }> + %struct.vm_page = type <{ %3, %3, %struct.vm_page*, %struct.vm_page*, %struct.vm_object*, i64, i64, %struct.md_page, i8, i8, i16, i8, i8, i16, i32, i16, i16, i8, i8, i8, i8, i8, i8, i8, i8 }> + %struct.vm_reserv = type opaque + %struct.vmspace = type <{ %struct.vm_map, %struct.shmmap_state*, i64, i64, i64, i64, i8*, i8*, i8*, i32, i8, i8, i8, i8, %struct.pmap }> + %struct.vnet = type opaque + %struct.vnode = type <{ i32, i8, i8, i8, i8, i8*, %struct.vop_vector*, i8*, %struct.mount*, %struct.freelst, %union.sigval, %struct.freelst, i32, i8, i8, i8, i8, %21, %22, %struct.namecache*, i64, i64, i64, i32, i8, i8, i8, i8, %struct.lock, %struct.mtx, %struct.lock*, i32, i32, i64, i64, i32, i8, i8, i8, i8, %struct.freelst, %struct.bufobj, %struct.vpollinfo*, %struct.label*, %struct.lockf* }> + %struct.vnodeop_desc = type <{ i8*, i32, i8, i8, i8, i8, i32 (%struct.vop_generic_args*)*, i32*, i32, i32, i32, i32 }> + %struct.vop_access_args = type <{ %struct.vop_generic_args, %struct.vnode*, i32, i8, i8, i8, i8, %struct.ucred*, %struct.thread* }> + %struct.vop_aclcheck_args = type <{ %struct.vop_generic_args, %struct.vnode*, i32, i8, i8, i8, i8, %struct.acl*, %struct.ucred*, %struct.thread* }> + %struct.vop_advlock_args = type <{ %struct.vop_generic_args, %struct.vnode*, i8*, i32, i8, i8, i8, i8, %struct.flock*, i32, i8, i8, i8, i8 }> + %struct.vop_advlockasync_args = type <{ %struct.vop_generic_args, %struct.vnode*, i8*, i32, i8, i8, i8, i8, %struct.flock*, i32, i8, i8, i8, i8, %struct.task*, i8** }> + %struct.vop_bmap_args = type <{ %struct.vop_generic_args, %struct.vnode*, i64, %struct.bufobj**, i64*, i32*, i32* }> + %struct.vop_cachedlookup_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.vnode**, %struct.componentname* }> + %struct.vop_create_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.vnode**, %struct.componentname*, %struct.vattr* }> + %struct.vop_deleteextattr_args = type <{ %struct.vop_generic_args, %struct.vnode*, i32, i8, i8, i8, i8, i8*, %struct.ucred*, %struct.thread* }> + %struct.vop_fsync_args = type <{ %struct.vop_generic_args, %struct.vnode*, i32, i8, i8, i8, i8, %struct.thread* }> + %struct.vop_generic_args = type <{ %struct.vnodeop_desc* }> + %struct.vop_getattr_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.vattr*, %struct.ucred* }> + %struct.vop_getextattr_args = type <{ %struct.vop_generic_args, %struct.vnode*, i32, i8, i8, i8, i8, i8*, %struct.uio*, i64*, %struct.ucred*, %struct.thread* }> + %struct.vop_getpages_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.vm_page**, i32, i32, i64 }> + %struct.vop_getwritemount_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.mount** }> + %struct.vop_inactive_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.thread* }> + %struct.vop_ioctl_args = type <{ %struct.vop_generic_args, %struct.vnode*, i64, i8*, i32, i8, i8, i8, i8, %struct.ucred*, %struct.thread* }> + %struct.vop_islocked_args = type <{ %struct.vop_generic_args, %struct.vnode* }> + %struct.vop_kqfilter_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.knote* }> + %struct.vop_link_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.vnode*, %struct.componentname* }> + %struct.vop_listextattr_args = type <{ %struct.vop_generic_args, %struct.vnode*, i32, i8, i8, i8, i8, %struct.uio*, i64*, %struct.ucred*, %struct.thread* }> + %struct.vop_lock1_args = type <{ %struct.vop_generic_args, %struct.vnode*, i32, i8, i8, i8, i8, i8*, i32, i8, i8, i8, i8 }> + %struct.vop_open_args = type <{ %struct.vop_generic_args, %struct.vnode*, i32, i8, i8, i8, i8, %struct.ucred*, %struct.thread*, %struct.file* }> + %struct.vop_openextattr_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.ucred*, %struct.thread* }> + %struct.vop_pathconf_args = type <{ %struct.vop_generic_args, %struct.vnode*, i32, i8, i8, i8, i8, i64* }> + %struct.vop_putpages_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.vm_page**, i32, i32, i32*, i64 }> + %struct.vop_read_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.uio*, i32, i8, i8, i8, i8, %struct.ucred* }> + %struct.vop_readdir_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.uio*, %struct.ucred*, i32*, i32*, i64** }> + %struct.vop_readlink_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.uio*, %struct.ucred* }> + %struct.vop_reallocblks_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.cluster_save* }> + %struct.vop_rename_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.vnode*, %struct.componentname*, %struct.vnode*, %struct.vnode*, %struct.componentname* }> + %struct.vop_revoke_args = type <{ %struct.vop_generic_args, %struct.vnode*, i32, i8, i8, i8, i8 }> + %struct.vop_setextattr_args = type <{ %struct.vop_generic_args, %struct.vnode*, i32, i8, i8, i8, i8, i8*, %struct.uio*, %struct.ucred*, %struct.thread* }> + %struct.vop_setlabel_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.label*, %struct.ucred*, %struct.thread* }> + %struct.vop_strategy_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.buf* }> + %struct.vop_symlink_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.vnode**, %struct.componentname*, %struct.vattr*, i8* }> + %struct.vop_vector = type <{ %struct.vop_vector*, i32 (%struct.vop_generic_args*)*, i32 (%struct.vop_islocked_args*)*, i32 (%struct.vop_cachedlookup_args*)*, i32 (%struct.vop_cachedlookup_args*)*, i32 (%struct.vop_create_args*)*, i32 (%struct.vop_whiteout_args*)*, i32 (%struct.vop_create_args*)*, i32 (%struct.vop_open_args*)*, i32 (%struct.vop_access_args*)*, i32 (%struct.vop_access_args*)*, i32 (%struct.vop_access_args*)*, i32 (%struct.vop_getattr_args*)*, i32 (%struct.vop_getattr_args*)*, i32 (%struct.vop_islocked_args*)*, i32 (%struct.vop_read_args*)*, i32 (%struct.vop_read_args*)*, i32 (%struct.vop_ioctl_args*)*, i32 (%struct.vop_access_args*)*, i32 (%struct.vop_kqfilter_args*)*, i32 (%struct.vop_revoke_args*)*, i32 (%struct.vop_fsync_args*)*, i32 (%struct.vop_link_args*)*, i32 (%struct.vop_link_args*)*, i32 (%struct.vop_rename_args*)*, i32 (%struct.vop_create_args*)*, i32 (%struct.vop_link_args*)*, i32 (%struct.vop_symlink_args*)*, i32 (%struct.vop_readdir_args*)*, i32 (%struct.vop_readlink_args*)*, i32 (%struct.vop_inactive_args*)*, i32 (%struct.vop_inactive_args*)*, i32 (%struct.vop_lock1_args*)*, i32 (%struct.vop_revoke_args*)*, i32 (%struct.vop_bmap_args*)*, i32 (%struct.vop_strategy_args*)*, i32 (%struct.vop_getwritemount_args*)*, i32 (%struct.vop_islocked_args*)*, i32 (%struct.vop_pathconf_args*)*, i32 (%struct.vop_advlock_args*)*, i32 (%struct.vop_advlockasync_args*)*, i32 (%struct.vop_reallocblks_args*)*, i32 (%struct.vop_getpages_args*)*, i32 (%struct.vop_putpages_args*)*, i32 (%struct.vop_aclcheck_args*)*, i32 (%struct.vop_aclcheck_args*)*, i32 (%struct.vop_aclcheck_args*)*, i32 (%struct.vop_access_args*)*, i32 (%struct.vop_getextattr_args*)*, i32 (%struct.vop_listextattr_args*)*, i32 (%struct.vop_openextattr_args*)*, i32 (%struct.vop_deleteextattr_args*)*, i32 (%struct.vop_setextattr_args*)*, i32 (%struct.vop_setlabel_args*)*, i32 (%struct.vop_vptofh_args*)*, i32 (%struct.vop_vptocnp_args*)* }> + %struct.vop_vptocnp_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.vnode**, %struct.ucred*, i8*, i32* }> + %struct.vop_vptofh_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.fid* }> + %struct.vop_whiteout_args = type <{ %struct.vop_generic_args, %struct.vnode*, %struct.componentname*, i32, i8, i8, i8, i8 }> + %struct.vpollinfo = type <{ %struct.mtx, %struct.selinfo, i16, i16, i8, i8, i8, i8 }> + %struct.witness = type opaque + %struct.workhead = type <{ %struct.worklist* }> + %struct.worklist = type opaque + %union.anon = type <{ [16 x i8] }> + %union.pager_info = type <{ [4 x i8] }> + %union.sigval = type <{ [8 x i8] }> + +define i32 @vlrureclaim(%struct.mount* %mp) nounwind { +entry: + br i1 undef, label %if.then11, label %do.end + +if.then11: ; preds = %entry + br label %do.end + +do.end: ; preds = %if.then11, %entry + br label %while.cond.outer + +while.cond.outer: ; preds = %while.cond.outer.backedge, %do.end + %count.0.ph = phi i32 [ undef, %do.end ], [ undef, %while.cond.outer.backedge ] ; <i32> [#uses=1] + br label %while.cond + +while.cond: ; preds = %next_iter, %while.cond.outer + %count.0 = phi i32 [ %dec, %next_iter ], [ %count.0.ph, %while.cond.outer ] ; <i32> [#uses=2] + %cmp21 = icmp eq i32 %count.0, 0 ; <i1> [#uses=1] + br i1 %cmp21, label %do.body288.loopexit4, label %while.body + +while.body: ; preds = %while.cond + br label %while.cond27 + +while.cond27: ; preds = %while.body36, %while.body + br i1 undef, label %do.body288.loopexit, label %land.rhs + +land.rhs: ; preds = %while.cond27 + br i1 undef, label %while.body36, label %while.end + +while.body36: ; preds = %land.rhs + br label %while.cond27 + +while.end: ; preds = %land.rhs + br i1 undef, label %do.body288.loopexit4, label %do.body46 + +do.body46: ; preds = %while.end + br i1 undef, label %if.else64, label %if.then53 + +if.then53: ; preds = %do.body46 + br label %if.end72 + +if.else64: ; preds = %do.body46 + br label %if.end72 + +if.end72: ; preds = %if.else64, %if.then53 + %dec = add i32 %count.0, -1 ; <i32> [#uses=2] + br i1 undef, label %next_iter, label %if.end111 + +if.end111: ; preds = %if.end72 + br i1 undef, label %lor.lhs.false, label %do.body145 + +lor.lhs.false: ; preds = %if.end111 + br i1 undef, label %lor.lhs.false122, label %do.body145 + +lor.lhs.false122: ; preds = %lor.lhs.false + br i1 undef, label %lor.lhs.false128, label %do.body145 + +lor.lhs.false128: ; preds = %lor.lhs.false122 + br i1 undef, label %do.body162, label %land.lhs.true + +land.lhs.true: ; preds = %lor.lhs.false128 + br i1 undef, label %do.body145, label %do.body162 + +do.body145: ; preds = %land.lhs.true, %lor.lhs.false122, %lor.lhs.false, %if.end111 + br i1 undef, label %if.then156, label %next_iter + +if.then156: ; preds = %do.body145 + br label %next_iter + +do.body162: ; preds = %land.lhs.true, %lor.lhs.false128 + br i1 undef, label %if.then173, label %do.end177 + +if.then173: ; preds = %do.body162 + br label %do.end177 + +do.end177: ; preds = %if.then173, %do.body162 + br i1 undef, label %do.body185, label %if.then182 + +if.then182: ; preds = %do.end177 + br label %next_iter_mntunlocked + +do.body185: ; preds = %do.end177 + br i1 undef, label %if.then196, label %do.end202 + +if.then196: ; preds = %do.body185 + br label %do.end202 + +do.end202: ; preds = %if.then196, %do.body185 + br i1 undef, label %lor.lhs.false207, label %if.then231 + +lor.lhs.false207: ; preds = %do.end202 + br i1 undef, label %lor.lhs.false214, label %if.then231 + +lor.lhs.false214: ; preds = %lor.lhs.false207 + br i1 undef, label %do.end236, label %land.lhs.true221 + +land.lhs.true221: ; preds = %lor.lhs.false214 + br i1 undef, label %if.then231, label %do.end236 + +if.then231: ; preds = %land.lhs.true221, %lor.lhs.false207, %do.end202 + br label %next_iter_mntunlocked + +do.end236: ; preds = %land.lhs.true221, %lor.lhs.false214 + br label %next_iter_mntunlocked + +next_iter_mntunlocked: ; preds = %do.end236, %if.then231, %if.then182 + br i1 undef, label %yield, label %do.body269 + +next_iter: ; preds = %if.then156, %do.body145, %if.end72 + %rem2482 = and i32 %dec, 255 ; <i32> [#uses=1] + %cmp249 = icmp eq i32 %rem2482, 0 ; <i1> [#uses=1] + br i1 %cmp249, label %do.body253, label %while.cond + +do.body253: ; preds = %next_iter + br i1 undef, label %if.then264, label %yield + +if.then264: ; preds = %do.body253 + br label %yield + +yield: ; preds = %if.then264, %do.body253, %next_iter_mntunlocked + br label %do.body269 + +do.body269: ; preds = %yield, %next_iter_mntunlocked + br i1 undef, label %if.then280, label %while.cond.outer.backedge + +if.then280: ; preds = %do.body269 + br label %while.cond.outer.backedge + +while.cond.outer.backedge: ; preds = %if.then280, %do.body269 + br label %while.cond.outer + +do.body288.loopexit: ; preds = %while.cond27 + br label %do.body288 + +do.body288.loopexit4: ; preds = %while.end, %while.cond + br label %do.body288 + +do.body288: ; preds = %do.body288.loopexit4, %do.body288.loopexit + br i1 undef, label %if.then299, label %do.end303 + +if.then299: ; preds = %do.body288 + br label %do.end303 + +do.end303: ; preds = %if.then299, %do.body288 + ret i32 undef +} diff --git a/test/Analysis/ScalarEvolution/avoid-smax-0.ll b/test/Analysis/ScalarEvolution/avoid-smax-0.ll new file mode 100644 index 0000000..24275f9 --- /dev/null +++ b/test/Analysis/ScalarEvolution/avoid-smax-0.ll @@ -0,0 +1,35 @@ +; RUN: opt < %s -scalar-evolution -analyze | grep {Loop %bb3: backedge-taken count is (-1 + %n)} + +; We don't want to use a max in the trip count expression in +; this testcase. + +define void @foo(i32 %n, i32* %p, i32* %q) nounwind { +entry: + icmp sgt i32 %n, 0 + br i1 %0, label %bb, label %return + +bb: + load i32* %q, align 4 + icmp eq i32 %1, 0 + br i1 %2, label %return, label %bb3.preheader + +bb3.preheader: + br label %bb3 + +bb3: + %i.0 = phi i32 [ %7, %bb3 ], [ 0, %bb3.preheader ] + getelementptr i32* %p, i32 %i.0 + load i32* %3, align 4 + add i32 %4, 1 + getelementptr i32* %p, i32 %i.0 + store i32 %5, i32* %6, align 4 + add i32 %i.0, 1 + icmp slt i32 %7, %n + br i1 %8, label %bb3, label %return.loopexit + +return.loopexit: + br label %return + +return: + ret void +} diff --git a/test/Analysis/ScalarEvolution/avoid-smax-1.ll b/test/Analysis/ScalarEvolution/avoid-smax-1.ll new file mode 100644 index 0000000..0bc9ce8 --- /dev/null +++ b/test/Analysis/ScalarEvolution/avoid-smax-1.ll @@ -0,0 +1,236 @@ +; RUN: opt < %s -indvars -S > %t +; RUN: grep select %t | count 2 +; RUN: grep {icmp ne i32.\* %w } %t + +; Indvars should be able to insert a canonical induction variable +; for the bb6 loop without using a maximum calculation (icmp, select) +; because it should be able to prove that the comparison is guarded +; by an appropriate conditional branch. Unfortunately, indvars is +; not yet able to find the comparison for the other two loops in +; this testcase. + +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" + +define void @foo(i8* %r, i32 %s, i32 %w, i32 %x, i8* %j, i32 %d) nounwind { +entry: + %0 = mul i32 %x, %w ; <i32> [#uses=2] + %1 = mul i32 %x, %w ; <i32> [#uses=1] + %2 = sdiv i32 %1, 4 ; <i32> [#uses=1] + %.sum2 = add i32 %2, %0 ; <i32> [#uses=2] + %cond = icmp eq i32 %d, 1 ; <i1> [#uses=1] + br i1 %cond, label %bb29, label %bb10.preheader + +bb10.preheader: ; preds = %entry + %3 = icmp sgt i32 %x, 0 ; <i1> [#uses=1] + br i1 %3, label %bb.nph9, label %bb18.loopexit + +bb.nph7: ; preds = %bb7.preheader + %4 = mul i32 %y.08, %w ; <i32> [#uses=1] + %5 = mul i32 %y.08, %s ; <i32> [#uses=1] + %6 = add i32 %5, 1 ; <i32> [#uses=1] + br label %bb6 + +bb6: ; preds = %bb7, %bb.nph7 + %x.06 = phi i32 [ %13, %bb7 ], [ 0, %bb.nph7 ] ; <i32> [#uses=3] + %7 = add i32 %x.06, %4 ; <i32> [#uses=1] + %8 = shl i32 %x.06, 1 ; <i32> [#uses=1] + %9 = add i32 %6, %8 ; <i32> [#uses=1] + %10 = getelementptr i8* %r, i32 %9 ; <i8*> [#uses=1] + %11 = load i8* %10, align 1 ; <i8> [#uses=1] + %12 = getelementptr i8* %j, i32 %7 ; <i8*> [#uses=1] + store i8 %11, i8* %12, align 1 + %13 = add i32 %x.06, 1 ; <i32> [#uses=2] + br label %bb7 + +bb7: ; preds = %bb6 + %14 = icmp slt i32 %13, %w ; <i1> [#uses=1] + br i1 %14, label %bb6, label %bb7.bb9_crit_edge + +bb7.bb9_crit_edge: ; preds = %bb7 + br label %bb9 + +bb9: ; preds = %bb7.preheader, %bb7.bb9_crit_edge + %15 = add i32 %y.08, 1 ; <i32> [#uses=2] + br label %bb10 + +bb10: ; preds = %bb9 + %16 = icmp slt i32 %15, %x ; <i1> [#uses=1] + br i1 %16, label %bb7.preheader, label %bb10.bb18.loopexit_crit_edge + +bb10.bb18.loopexit_crit_edge: ; preds = %bb10 + br label %bb10.bb18.loopexit_crit_edge.split + +bb10.bb18.loopexit_crit_edge.split: ; preds = %bb.nph9, %bb10.bb18.loopexit_crit_edge + br label %bb18.loopexit + +bb.nph9: ; preds = %bb10.preheader + %17 = icmp sgt i32 %w, 0 ; <i1> [#uses=1] + br i1 %17, label %bb.nph9.split, label %bb10.bb18.loopexit_crit_edge.split + +bb.nph9.split: ; preds = %bb.nph9 + br label %bb7.preheader + +bb7.preheader: ; preds = %bb.nph9.split, %bb10 + %y.08 = phi i32 [ %15, %bb10 ], [ 0, %bb.nph9.split ] ; <i32> [#uses=3] + br i1 true, label %bb.nph7, label %bb9 + +bb.nph5: ; preds = %bb18.loopexit + %18 = sdiv i32 %w, 2 ; <i32> [#uses=1] + %19 = icmp slt i32 %w, 2 ; <i1> [#uses=1] + %20 = sdiv i32 %x, 2 ; <i32> [#uses=1] + br i1 %19, label %bb18.bb20_crit_edge.split, label %bb.nph5.split + +bb.nph5.split: ; preds = %bb.nph5 + br label %bb13 + +bb13: ; preds = %bb18, %bb.nph5.split + %y.14 = phi i32 [ %42, %bb18 ], [ 0, %bb.nph5.split ] ; <i32> [#uses=4] + %21 = mul i32 %18, %y.14 ; <i32> [#uses=2] + %22 = shl i32 %y.14, 1 ; <i32> [#uses=1] + %23 = srem i32 %y.14, 2 ; <i32> [#uses=1] + %24 = add i32 %23, %22 ; <i32> [#uses=1] + %25 = mul i32 %24, %s ; <i32> [#uses=2] + br i1 true, label %bb.nph3, label %bb17 + +bb.nph3: ; preds = %bb13 + %26 = add i32 %21, %0 ; <i32> [#uses=1] + %27 = add i32 %21, %.sum2 ; <i32> [#uses=1] + %28 = sdiv i32 %w, 2 ; <i32> [#uses=1] + br label %bb14 + +bb14: ; preds = %bb15, %bb.nph3 + %x.12 = phi i32 [ %40, %bb15 ], [ 0, %bb.nph3 ] ; <i32> [#uses=5] + %29 = shl i32 %x.12, 2 ; <i32> [#uses=1] + %30 = add i32 %29, %25 ; <i32> [#uses=1] + %31 = getelementptr i8* %r, i32 %30 ; <i8*> [#uses=1] + %32 = load i8* %31, align 1 ; <i8> [#uses=1] + %.sum = add i32 %26, %x.12 ; <i32> [#uses=1] + %33 = getelementptr i8* %j, i32 %.sum ; <i8*> [#uses=1] + store i8 %32, i8* %33, align 1 + %34 = shl i32 %x.12, 2 ; <i32> [#uses=1] + %35 = or i32 %34, 2 ; <i32> [#uses=1] + %36 = add i32 %35, %25 ; <i32> [#uses=1] + %37 = getelementptr i8* %r, i32 %36 ; <i8*> [#uses=1] + %38 = load i8* %37, align 1 ; <i8> [#uses=1] + %.sum6 = add i32 %27, %x.12 ; <i32> [#uses=1] + %39 = getelementptr i8* %j, i32 %.sum6 ; <i8*> [#uses=1] + store i8 %38, i8* %39, align 1 + %40 = add i32 %x.12, 1 ; <i32> [#uses=2] + br label %bb15 + +bb15: ; preds = %bb14 + %41 = icmp sgt i32 %28, %40 ; <i1> [#uses=1] + br i1 %41, label %bb14, label %bb15.bb17_crit_edge + +bb15.bb17_crit_edge: ; preds = %bb15 + br label %bb17 + +bb17: ; preds = %bb15.bb17_crit_edge, %bb13 + %42 = add i32 %y.14, 1 ; <i32> [#uses=2] + br label %bb18 + +bb18.loopexit: ; preds = %bb10.bb18.loopexit_crit_edge.split, %bb10.preheader + %43 = icmp slt i32 %x, 2 ; <i1> [#uses=1] + br i1 %43, label %bb20, label %bb.nph5 + +bb18: ; preds = %bb17 + %44 = icmp sgt i32 %20, %42 ; <i1> [#uses=1] + br i1 %44, label %bb13, label %bb18.bb20_crit_edge + +bb18.bb20_crit_edge: ; preds = %bb18 + br label %bb18.bb20_crit_edge.split + +bb18.bb20_crit_edge.split: ; preds = %bb18.bb20_crit_edge, %bb.nph5 + br label %bb20 + +bb20: ; preds = %bb18.bb20_crit_edge.split, %bb18.loopexit + switch i32 %d, label %return [ + i32 3, label %bb22 + i32 1, label %bb29 + ] + +bb22: ; preds = %bb20 + %45 = mul i32 %x, %w ; <i32> [#uses=1] + %46 = sdiv i32 %45, 4 ; <i32> [#uses=1] + %.sum3 = add i32 %46, %.sum2 ; <i32> [#uses=2] + %47 = add i32 %x, 15 ; <i32> [#uses=1] + %48 = and i32 %47, -16 ; <i32> [#uses=1] + %49 = add i32 %w, 15 ; <i32> [#uses=1] + %50 = and i32 %49, -16 ; <i32> [#uses=1] + %51 = mul i32 %48, %s ; <i32> [#uses=1] + %52 = icmp sgt i32 %x, 0 ; <i1> [#uses=1] + br i1 %52, label %bb.nph, label %bb26 + +bb.nph: ; preds = %bb22 + br label %bb23 + +bb23: ; preds = %bb24, %bb.nph + %y.21 = phi i32 [ %57, %bb24 ], [ 0, %bb.nph ] ; <i32> [#uses=3] + %53 = mul i32 %y.21, %50 ; <i32> [#uses=1] + %.sum1 = add i32 %53, %51 ; <i32> [#uses=1] + %54 = getelementptr i8* %r, i32 %.sum1 ; <i8*> [#uses=1] + %55 = mul i32 %y.21, %w ; <i32> [#uses=1] + %.sum5 = add i32 %55, %.sum3 ; <i32> [#uses=1] + %56 = getelementptr i8* %j, i32 %.sum5 ; <i8*> [#uses=1] + tail call void @llvm.memcpy.i32(i8* %56, i8* %54, i32 %w, i32 1) + %57 = add i32 %y.21, 1 ; <i32> [#uses=2] + br label %bb24 + +bb24: ; preds = %bb23 + %58 = icmp slt i32 %57, %x ; <i1> [#uses=1] + br i1 %58, label %bb23, label %bb24.bb26_crit_edge + +bb24.bb26_crit_edge: ; preds = %bb24 + br label %bb26 + +bb26: ; preds = %bb24.bb26_crit_edge, %bb22 + %59 = mul i32 %x, %w ; <i32> [#uses=1] + %.sum4 = add i32 %.sum3, %59 ; <i32> [#uses=1] + %60 = getelementptr i8* %j, i32 %.sum4 ; <i8*> [#uses=1] + %61 = mul i32 %x, %w ; <i32> [#uses=1] + %62 = sdiv i32 %61, 2 ; <i32> [#uses=1] + tail call void @llvm.memset.i32(i8* %60, i8 -128, i32 %62, i32 1) + ret void + +bb29: ; preds = %bb20, %entry + %63 = add i32 %w, 15 ; <i32> [#uses=1] + %64 = and i32 %63, -16 ; <i32> [#uses=1] + %65 = icmp sgt i32 %x, 0 ; <i1> [#uses=1] + br i1 %65, label %bb.nph11, label %bb33 + +bb.nph11: ; preds = %bb29 + br label %bb30 + +bb30: ; preds = %bb31, %bb.nph11 + %y.310 = phi i32 [ %70, %bb31 ], [ 0, %bb.nph11 ] ; <i32> [#uses=3] + %66 = mul i32 %y.310, %64 ; <i32> [#uses=1] + %67 = getelementptr i8* %r, i32 %66 ; <i8*> [#uses=1] + %68 = mul i32 %y.310, %w ; <i32> [#uses=1] + %69 = getelementptr i8* %j, i32 %68 ; <i8*> [#uses=1] + tail call void @llvm.memcpy.i32(i8* %69, i8* %67, i32 %w, i32 1) + %70 = add i32 %y.310, 1 ; <i32> [#uses=2] + br label %bb31 + +bb31: ; preds = %bb30 + %71 = icmp slt i32 %70, %x ; <i1> [#uses=1] + br i1 %71, label %bb30, label %bb31.bb33_crit_edge + +bb31.bb33_crit_edge: ; preds = %bb31 + br label %bb33 + +bb33: ; preds = %bb31.bb33_crit_edge, %bb29 + %72 = mul i32 %x, %w ; <i32> [#uses=1] + %73 = getelementptr i8* %j, i32 %72 ; <i8*> [#uses=1] + %74 = mul i32 %x, %w ; <i32> [#uses=1] + %75 = sdiv i32 %74, 2 ; <i32> [#uses=1] + tail call void @llvm.memset.i32(i8* %73, i8 -128, i32 %75, i32 1) + ret void + +return: ; preds = %bb20 + ret void +} + +declare void @llvm.memcpy.i32(i8*, i8*, i32, i32) nounwind + +declare void @llvm.memset.i32(i8*, i8, i32, i32) nounwind diff --git a/test/Analysis/ScalarEvolution/dg.exp b/test/Analysis/ScalarEvolution/dg.exp new file mode 100644 index 0000000..b65a250 --- /dev/null +++ b/test/Analysis/ScalarEvolution/dg.exp @@ -0,0 +1,3 @@ +load_lib llvm.exp + +RunLLVMTests [lsort [glob -nocomplain $srcdir/$subdir/*.ll]] diff --git a/test/Analysis/ScalarEvolution/div-overflow.ll b/test/Analysis/ScalarEvolution/div-overflow.ll new file mode 100644 index 0000000..4f6f1e2 --- /dev/null +++ b/test/Analysis/ScalarEvolution/div-overflow.ll @@ -0,0 +1,10 @@ +; RUN: opt < %s -scalar-evolution -analyze \ +; RUN: | grep {\\--> ((-128 \\* %a) /u -128)} + +; Don't let ScalarEvolution fold this div away. + +define i8 @foo(i8 %a) { + %t0 = shl i8 %a, 7 + %t1 = lshr i8 %t0, 7 + ret i8 %t1 +} diff --git a/test/Analysis/ScalarEvolution/do-loop.ll b/test/Analysis/ScalarEvolution/do-loop.ll new file mode 100644 index 0000000..6e3295a --- /dev/null +++ b/test/Analysis/ScalarEvolution/do-loop.ll @@ -0,0 +1,18 @@ +; RUN: opt < %s -analyze -scalar-evolution | grep smax +; PR1614 + +define i32 @f(i32 %x, i32 %y) { +entry: + br label %bb + +bb: ; preds = %bb, %entry + %indvar = phi i32 [ 0, %entry ], [ %indvar.next, %bb ] ; <i32> [#uses=2] + %x_addr.0 = add i32 %indvar, %x ; <i32> [#uses=1] + %tmp2 = add i32 %x_addr.0, 1 ; <i32> [#uses=2] + %tmp5 = icmp slt i32 %tmp2, %y ; <i1> [#uses=1] + %indvar.next = add i32 %indvar, 1 ; <i32> [#uses=1] + br i1 %tmp5, label %bb, label %bb7 + +bb7: ; preds = %bb + ret i32 %tmp2 +} diff --git a/test/Analysis/ScalarEvolution/max-trip-count.ll b/test/Analysis/ScalarEvolution/max-trip-count.ll new file mode 100644 index 0000000..a8966be --- /dev/null +++ b/test/Analysis/ScalarEvolution/max-trip-count.ll @@ -0,0 +1,34 @@ +; RUN: opt < %s -analyze -scalar-evolution \ +; RUN: | grep {\{%d,+,\[^\{\}\]\*\}<%bb>} + +; ScalarEvolution should be able to understand the loop and eliminate the casts. + +define void @foo(i32* nocapture %d, i32 %n) nounwind { +entry: + %0 = icmp sgt i32 %n, 0 ; <i1> [#uses=1] + br i1 %0, label %bb.nph, label %return + +bb.nph: ; preds = %entry + br label %bb + +bb: ; preds = %bb1, %bb.nph + %i.02 = phi i32 [ %5, %bb1 ], [ 0, %bb.nph ] ; <i32> [#uses=2] + %p.01 = phi i8 [ %4, %bb1 ], [ -1, %bb.nph ] ; <i8> [#uses=2] + %1 = sext i8 %p.01 to i32 ; <i32> [#uses=1] + %2 = sext i32 %i.02 to i64 ; <i64> [#uses=1] + %3 = getelementptr i32* %d, i64 %2 ; <i32*> [#uses=1] + store i32 %1, i32* %3, align 4 + %4 = add i8 %p.01, 1 ; <i8> [#uses=1] + %5 = add i32 %i.02, 1 ; <i32> [#uses=2] + br label %bb1 + +bb1: ; preds = %bb + %6 = icmp slt i32 %5, %n ; <i1> [#uses=1] + br i1 %6, label %bb, label %bb1.return_crit_edge + +bb1.return_crit_edge: ; preds = %bb1 + br label %return + +return: ; preds = %bb1.return_crit_edge, %entry + ret void +} diff --git a/test/Analysis/ScalarEvolution/nsw-offset.ll b/test/Analysis/ScalarEvolution/nsw-offset.ll new file mode 100644 index 0000000..4cd9a6d --- /dev/null +++ b/test/Analysis/ScalarEvolution/nsw-offset.ll @@ -0,0 +1,77 @@ +; RUN: opt < %s -S -analyze -scalar-evolution | FileCheck %s + +; ScalarEvolution should be able to fold away the sign-extensions +; on this loop with a primary induction variable incremented with +; a nsw add of 2. + +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" + +define void @foo(i32 %no, double* nocapture %d, double* nocapture %q) nounwind { +entry: + %n = and i32 %no, 4294967294 + %0 = icmp sgt i32 %n, 0 ; <i1> [#uses=1] + br i1 %0, label %bb.nph, label %return + +bb.nph: ; preds = %entry + br label %bb + +bb: ; preds = %bb.nph, %bb1 + %i.01 = phi i32 [ %16, %bb1 ], [ 0, %bb.nph ] ; <i32> [#uses=5] + +; CHECK: %1 = sext i32 %i.01 to i64 +; CHECK: --> {0,+,2}<%bb> + %1 = sext i32 %i.01 to i64 ; <i64> [#uses=1] + +; CHECK: %2 = getelementptr inbounds double* %d, i64 %1 +; CHECK: --> {%d,+,16}<%bb> + %2 = getelementptr inbounds double* %d, i64 %1 ; <double*> [#uses=1] + + %3 = load double* %2, align 8 ; <double> [#uses=1] + %4 = sext i32 %i.01 to i64 ; <i64> [#uses=1] + %5 = getelementptr inbounds double* %q, i64 %4 ; <double*> [#uses=1] + %6 = load double* %5, align 8 ; <double> [#uses=1] + %7 = or i32 %i.01, 1 ; <i32> [#uses=1] + +; CHECK: %8 = sext i32 %7 to i64 +; CHECK: --> {1,+,2}<%bb> + %8 = sext i32 %7 to i64 ; <i64> [#uses=1] + +; CHECK: %9 = getelementptr inbounds double* %q, i64 %8 +; CHECK: {(8 + %q),+,16}<%bb> + %9 = getelementptr inbounds double* %q, i64 %8 ; <double*> [#uses=1] + +; Artificially repeat the above three instructions, this time using +; add nsw instead of or. + %t7 = add nsw i32 %i.01, 1 ; <i32> [#uses=1] + +; CHECK: %t8 = sext i32 %t7 to i64 +; CHECK: --> {1,+,2}<%bb> + %t8 = sext i32 %t7 to i64 ; <i64> [#uses=1] + +; CHECK: %t9 = getelementptr inbounds double* %q, i64 %t8 +; CHECK: {(8 + %q),+,16}<%bb> + %t9 = getelementptr inbounds double* %q, i64 %t8 ; <double*> [#uses=1] + + %10 = load double* %9, align 8 ; <double> [#uses=1] + %11 = fadd double %6, %10 ; <double> [#uses=1] + %12 = fadd double %11, 3.200000e+00 ; <double> [#uses=1] + %13 = fmul double %3, %12 ; <double> [#uses=1] + %14 = sext i32 %i.01 to i64 ; <i64> [#uses=1] + %15 = getelementptr inbounds double* %d, i64 %14 ; <double*> [#uses=1] + store double %13, double* %15, align 8 + %16 = add nsw i32 %i.01, 2 ; <i32> [#uses=2] + br label %bb1 + +bb1: ; preds = %bb + %17 = icmp slt i32 %16, %n ; <i1> [#uses=1] + br i1 %17, label %bb, label %bb1.return_crit_edge + +bb1.return_crit_edge: ; preds = %bb1 + br label %return + +return: ; preds = %bb1.return_crit_edge, %entry + ret void +} + +; CHECK: Loop %bb: backedge-taken count is ((-1 + %n) /u 2) +; CHECK: Loop %bb: max backedge-taken count is 1073741822 diff --git a/test/Analysis/ScalarEvolution/nsw.ll b/test/Analysis/ScalarEvolution/nsw.ll new file mode 100644 index 0000000..456f3f0 --- /dev/null +++ b/test/Analysis/ScalarEvolution/nsw.ll @@ -0,0 +1,40 @@ +; RUN: opt < %s -analyze -scalar-evolution | grep { --> {.*,+,.*}<%bb>} | count 8 + +; The addrecs in this loop are analyzable only by using nsw information. + +target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64" + +define void @foo(double* %p) nounwind { +entry: + %tmp = load double* %p, align 8 ; <double> [#uses=1] + %tmp1 = fcmp ogt double %tmp, 2.000000e+00 ; <i1> [#uses=1] + br i1 %tmp1, label %bb.nph, label %return + +bb.nph: ; preds = %entry + br label %bb + +bb: ; preds = %bb1, %bb.nph + %i.01 = phi i32 [ %tmp8, %bb1 ], [ 0, %bb.nph ] ; <i32> [#uses=3] + %tmp2 = sext i32 %i.01 to i64 ; <i64> [#uses=1] + %tmp3 = getelementptr double* %p, i64 %tmp2 ; <double*> [#uses=1] + %tmp4 = load double* %tmp3, align 8 ; <double> [#uses=1] + %tmp5 = fmul double %tmp4, 9.200000e+00 ; <double> [#uses=1] + %tmp6 = sext i32 %i.01 to i64 ; <i64> [#uses=1] + %tmp7 = getelementptr double* %p, i64 %tmp6 ; <double*> [#uses=1] + store double %tmp5, double* %tmp7, align 8 + %tmp8 = add nsw i32 %i.01, 1 ; <i32> [#uses=2] + br label %bb1 + +bb1: ; preds = %bb + %phitmp = sext i32 %tmp8 to i64 ; <i64> [#uses=1] + %tmp9 = getelementptr double* %p, i64 %phitmp ; <double*> [#uses=1] + %tmp10 = load double* %tmp9, align 8 ; <double> [#uses=1] + %tmp11 = fcmp ogt double %tmp10, 2.000000e+00 ; <i1> [#uses=1] + br i1 %tmp11, label %bb, label %bb1.return_crit_edge + +bb1.return_crit_edge: ; preds = %bb1 + br label %return + +return: ; preds = %bb1.return_crit_edge, %entry + ret void +} diff --git a/test/Analysis/ScalarEvolution/pointer-sign-bits.ll b/test/Analysis/ScalarEvolution/pointer-sign-bits.ll new file mode 100644 index 0000000..b2cec2d --- /dev/null +++ b/test/Analysis/ScalarEvolution/pointer-sign-bits.ll @@ -0,0 +1,220 @@ +; RUN: opt < %s -analyze -scalar-evolution + +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-f80:32:32" + %JavaObject = type { [0 x i32 (...)*]*, i8* } + +define void @JnJVM_antlr_CSharpCodeGenerator_genBitSet__Lantlr_collections_impl_BitSet_2I(%JavaObject*, %JavaObject*, i32) { +start: + br i1 undef, label %"stack overflow", label %"no stack overflow" + +"GOTO or IF*2": ; preds = %"true verifyAndComputePtr89", %verifyNullCont84 + unreachable + +"GOTO or IF*5": ; preds = %"true verifyAndComputePtr127", %"GOTO or IF*6" + unreachable + +"GOTO or IF*6": ; preds = %"true verifyAndComputePtr131.GOTO or IF*6_crit_edge", %"true verifyAndComputePtr89" + %indvar = phi i32 [ %indvar.next, %"true verifyAndComputePtr131.GOTO or IF*6_crit_edge" ], [ 0, %"true verifyAndComputePtr89" ] ; <i32> [#uses=2] + %.0.in = add i32 %indvar, 0 ; <i32> [#uses=1] + %.0 = add i32 %.0.in, 1 ; <i32> [#uses=1] + %3 = icmp slt i32 %.0, %4 ; <i1> [#uses=1] + br i1 %3, label %verifyNullCont126, label %"GOTO or IF*5" + +end: ; preds = %"no exception block35" + ret void + +"stack overflow": ; preds = %start + ret void + +"no stack overflow": ; preds = %start + br i1 undef, label %verifyNullCont, label %"no stack overflow.end_crit_edge" + +"no stack overflow.end_crit_edge": ; preds = %"no stack overflow" + ret void + +verifyNullCont: ; preds = %"no stack overflow" + br i1 undef, label %verifyNullCont9, label %verifyNullCont.end_crit_edge + +verifyNullCont.end_crit_edge: ; preds = %verifyNullCont + ret void + +verifyNullCont9: ; preds = %verifyNullCont + br i1 undef, label %verifyNullCont12, label %verifyNullCont9.end_crit_edge + +verifyNullCont9.end_crit_edge: ; preds = %verifyNullCont9 + ret void + +verifyNullCont12: ; preds = %verifyNullCont9 + br i1 undef, label %"no exception block13", label %verifyNullCont12.end_crit_edge + +verifyNullCont12.end_crit_edge: ; preds = %verifyNullCont12 + ret void + +"no exception block13": ; preds = %verifyNullCont12 + br i1 undef, label %verifyNullExit14, label %verifyNullCont15 + +verifyNullExit14: ; preds = %"no exception block13" + ret void + +verifyNullCont15: ; preds = %"no exception block13" + br i1 undef, label %"no exception block16", label %verifyNullCont15.end_crit_edge + +verifyNullCont15.end_crit_edge: ; preds = %verifyNullCont15 + ret void + +"no exception block16": ; preds = %verifyNullCont15 + br i1 undef, label %verifyNullExit17, label %verifyNullCont18 + +verifyNullExit17: ; preds = %"no exception block16" + ret void + +verifyNullCont18: ; preds = %"no exception block16" + br i1 undef, label %"no exception block19", label %verifyNullCont18.end_crit_edge + +verifyNullCont18.end_crit_edge: ; preds = %verifyNullCont18 + ret void + +"no exception block19": ; preds = %verifyNullCont18 + br i1 undef, label %verifyNullExit20, label %verifyNullCont21 + +verifyNullExit20: ; preds = %"no exception block19" + ret void + +verifyNullCont21: ; preds = %"no exception block19" + br i1 undef, label %verifyNullCont24, label %verifyNullCont21.end_crit_edge + +verifyNullCont21.end_crit_edge: ; preds = %verifyNullCont21 + ret void + +verifyNullCont24: ; preds = %verifyNullCont21 + br i1 undef, label %verifyNullCont27, label %verifyNullCont24.end_crit_edge + +verifyNullCont24.end_crit_edge: ; preds = %verifyNullCont24 + ret void + +verifyNullCont27: ; preds = %verifyNullCont24 + br i1 undef, label %verifyNullCont32, label %verifyNullCont27.end_crit_edge + +verifyNullCont27.end_crit_edge: ; preds = %verifyNullCont27 + ret void + +verifyNullCont32: ; preds = %verifyNullCont27 + br i1 undef, label %verifyNullExit33, label %verifyNullCont34 + +verifyNullExit33: ; preds = %verifyNullCont32 + ret void + +verifyNullCont34: ; preds = %verifyNullCont32 + br i1 undef, label %"no exception block35", label %verifyNullCont34.end_crit_edge + +verifyNullCont34.end_crit_edge: ; preds = %verifyNullCont34 + ret void + +"no exception block35": ; preds = %verifyNullCont34 + br i1 undef, label %end, label %verifyNullCont60 + +verifyNullCont60: ; preds = %"no exception block35" + br i1 undef, label %verifyNullCont63, label %verifyNullCont60.end_crit_edge + +verifyNullCont60.end_crit_edge: ; preds = %verifyNullCont60 + ret void + +verifyNullCont63: ; preds = %verifyNullCont60 + br i1 undef, label %"no exception block64", label %verifyNullCont63.end_crit_edge + +verifyNullCont63.end_crit_edge: ; preds = %verifyNullCont63 + ret void + +"no exception block64": ; preds = %verifyNullCont63 + br i1 undef, label %verifyNullExit65, label %verifyNullCont66 + +verifyNullExit65: ; preds = %"no exception block64" + ret void + +verifyNullCont66: ; preds = %"no exception block64" + br i1 undef, label %"no exception block67", label %verifyNullCont66.end_crit_edge + +verifyNullCont66.end_crit_edge: ; preds = %verifyNullCont66 + ret void + +"no exception block67": ; preds = %verifyNullCont66 + br i1 undef, label %verifyNullExit68, label %verifyNullCont69 + +verifyNullExit68: ; preds = %"no exception block67" + ret void + +verifyNullCont69: ; preds = %"no exception block67" + br i1 undef, label %"no exception block70", label %verifyNullCont69.end_crit_edge + +verifyNullCont69.end_crit_edge: ; preds = %verifyNullCont69 + ret void + +"no exception block70": ; preds = %verifyNullCont69 + br i1 undef, label %verifyNullExit71, label %verifyNullCont72 + +verifyNullExit71: ; preds = %"no exception block70" + ret void + +verifyNullCont72: ; preds = %"no exception block70" + br i1 undef, label %verifyNullCont75, label %verifyNullCont72.end_crit_edge + +verifyNullCont72.end_crit_edge: ; preds = %verifyNullCont72 + ret void + +verifyNullCont75: ; preds = %verifyNullCont72 + br i1 undef, label %verifyNullCont78, label %verifyNullCont75.end_crit_edge + +verifyNullCont75.end_crit_edge: ; preds = %verifyNullCont75 + ret void + +verifyNullCont78: ; preds = %verifyNullCont75 + br i1 undef, label %"verifyNullCont78.GOTO or IF*4_crit_edge", label %verifyNullCont78.end_crit_edge + +"verifyNullCont78.GOTO or IF*4_crit_edge": ; preds = %verifyNullCont78 + br i1 undef, label %verifyNullExit80, label %verifyNullCont81 + +verifyNullCont78.end_crit_edge: ; preds = %verifyNullCont78 + ret void + +verifyNullExit80: ; preds = %"verifyNullCont78.GOTO or IF*4_crit_edge" + ret void + +verifyNullCont81: ; preds = %"verifyNullCont78.GOTO or IF*4_crit_edge" + %4 = ptrtoint i8* undef to i32 ; <i32> [#uses=2] + %5 = icmp slt i32 0, %4 ; <i1> [#uses=1] + br i1 %5, label %verifyNullCont84, label %verifyNullCont172 + +verifyNullCont84: ; preds = %verifyNullCont81 + br i1 undef, label %"GOTO or IF*2", label %verifyNullCont86 + +verifyNullCont86: ; preds = %verifyNullCont84 + br i1 undef, label %"true verifyAndComputePtr", label %"false verifyAndComputePtr" + +"true verifyAndComputePtr": ; preds = %verifyNullCont86 + br i1 undef, label %"true verifyAndComputePtr89", label %"false verifyAndComputePtr90" + +"false verifyAndComputePtr": ; preds = %verifyNullCont86 + ret void + +"true verifyAndComputePtr89": ; preds = %"true verifyAndComputePtr" + br i1 undef, label %"GOTO or IF*6", label %"GOTO or IF*2" + +"false verifyAndComputePtr90": ; preds = %"true verifyAndComputePtr" + ret void + +verifyNullCont126: ; preds = %"GOTO or IF*6" + br i1 undef, label %"true verifyAndComputePtr127", label %"false verifyAndComputePtr128" + +"true verifyAndComputePtr127": ; preds = %verifyNullCont126 + br i1 undef, label %"true verifyAndComputePtr131.GOTO or IF*6_crit_edge", label %"GOTO or IF*5" + +"false verifyAndComputePtr128": ; preds = %verifyNullCont126 + ret void + +"true verifyAndComputePtr131.GOTO or IF*6_crit_edge": ; preds = %"true verifyAndComputePtr127" + %indvar.next = add i32 %indvar, 1 ; <i32> [#uses=1] + br label %"GOTO or IF*6" + +verifyNullCont172: ; preds = %verifyNullCont81 + unreachable +} diff --git a/test/Analysis/ScalarEvolution/pr3909.ll b/test/Analysis/ScalarEvolution/pr3909.ll new file mode 100644 index 0000000..10e328d --- /dev/null +++ b/test/Analysis/ScalarEvolution/pr3909.ll @@ -0,0 +1,30 @@ +; RUN: opt < %s -indvars -disable-output +; PR 3909 + + + type { i32, %1* } ; type %0 + type { i32, i8* } ; type %1 + +define x86_stdcallcc i32 @_Dmain(%0 %unnamed) { +entry: + br label %whilebody + +whilebody: ; preds = %endwhile5, %entry + %i.0 = phi i64 [ 0, %entry ], [ %tmp11, %endwhile5 ] ; <i64> [#uses=1] + %m.0 = phi i64 [ 0, %entry ], [ %tmp11, %endwhile5 ] ; <i64> [#uses=2] + %tmp2 = mul i64 %m.0, %m.0 ; <i64> [#uses=1] + br label %whilecond3 + +whilecond3: ; preds = %whilebody4, %whilebody + %j.0 = phi i64 [ %tmp2, %whilebody ], [ %tmp9, %whilebody4 ] ; <i64> [#uses=2] + %tmp7 = icmp ne i64 %j.0, 0 ; <i1> [#uses=1] + br i1 %tmp7, label %whilebody4, label %endwhile5 + +whilebody4: ; preds = %whilecond3 + %tmp9 = add i64 %j.0, 1 ; <i64> [#uses=1] + br label %whilecond3 + +endwhile5: ; preds = %whilecond3 + %tmp11 = add i64 %i.0, 1 ; <i64> [#uses=2] + br label %whilebody +} diff --git a/test/Analysis/ScalarEvolution/scev-aa.ll b/test/Analysis/ScalarEvolution/scev-aa.ll new file mode 100644 index 0000000..e07aca2 --- /dev/null +++ b/test/Analysis/ScalarEvolution/scev-aa.ll @@ -0,0 +1,194 @@ +; RUN: opt < %s -scev-aa -aa-eval -print-all-alias-modref-info \ +; RUN: |& FileCheck %s + +; At the time of this writing, -basicaa only misses the example of the form +; A[i+(j+1)] != A[i+j], which can arise from multi-dimensional array references. + +target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64" + +; p[i] and p[i+1] don't alias. + +; CHECK: Function: loop: 3 pointers, 0 call sites +; CHECK: NoAlias: double* %pi, double* %pi.next + +define void @loop(double* nocapture %p, i64 %n) nounwind { +entry: + %j = icmp sgt i64 %n, 0 + br i1 %j, label %bb, label %return + +bb: + %i = phi i64 [ 0, %entry ], [ %i.next, %bb ] + %pi = getelementptr double* %p, i64 %i + %i.next = add i64 %i, 1 + %pi.next = getelementptr double* %p, i64 %i.next + %x = load double* %pi + %y = load double* %pi.next + %z = fmul double %x, %y + store double %z, double* %pi + %exitcond = icmp eq i64 %i.next, %n + br i1 %exitcond, label %return, label %bb + +return: + ret void +} + +; Slightly more involved: p[j][i], p[j][i+1], and p[j+1][i] don't alias. + +; CHECK: Function: nestedloop: 4 pointers, 0 call sites +; CHECK: NoAlias: double* %pi.j, double* %pi.next.j +; CHECK: NoAlias: double* %pi.j, double* %pi.j.next +; CHECK: NoAlias: double* %pi.j.next, double* %pi.next.j + +define void @nestedloop(double* nocapture %p, i64 %m) nounwind { +entry: + %k = icmp sgt i64 %m, 0 + br i1 %k, label %guard, label %return + +guard: + %l = icmp sgt i64 91, 0 + br i1 %l, label %outer.loop, label %return + +outer.loop: + %j = phi i64 [ 0, %guard ], [ %j.next, %outer.latch ] + br label %bb + +bb: + %i = phi i64 [ 0, %outer.loop ], [ %i.next, %bb ] + %i.next = add i64 %i, 1 + + %e = add i64 %i, %j + %pi.j = getelementptr double* %p, i64 %e + %f = add i64 %i.next, %j + %pi.next.j = getelementptr double* %p, i64 %f + %x = load double* %pi.j + %y = load double* %pi.next.j + %z = fmul double %x, %y + store double %z, double* %pi.j + + %o = add i64 %j, 91 + %g = add i64 %i, %o + %pi.j.next = getelementptr double* %p, i64 %g + %a = load double* %pi.j.next + %b = fmul double %x, %a + store double %b, double* %pi.j.next + + %exitcond = icmp eq i64 %i.next, 91 + br i1 %exitcond, label %outer.latch, label %bb + +outer.latch: + %j.next = add i64 %j, 91 + %h = icmp eq i64 %j.next, %m + br i1 %h, label %return, label %outer.loop + +return: + ret void +} + +; Even more involved: same as nestedloop, but with a variable extent. +; When n is 1, p[j+1][i] does alias p[j][i+1], and there's no way to +; prove whether n will be greater than 1, so that relation will always +; by MayAlias. The loop is guarded by a n > 0 test though, so +; p[j+1][i] and p[j][i] can theoretically be determined to be NoAlias, +; however the analysis currently doesn't do that. +; TODO: Make the analysis smarter and turn that MayAlias into a NoAlias. + +; CHECK: Function: nestedloop_more: 4 pointers, 0 call sites +; CHECK: NoAlias: double* %pi.j, double* %pi.next.j +; CHECK: MayAlias: double* %pi.j, double* %pi.j.next + +define void @nestedloop_more(double* nocapture %p, i64 %n, i64 %m) nounwind { +entry: + %k = icmp sgt i64 %m, 0 + br i1 %k, label %guard, label %return + +guard: + %l = icmp sgt i64 %n, 0 + br i1 %l, label %outer.loop, label %return + +outer.loop: + %j = phi i64 [ 0, %guard ], [ %j.next, %outer.latch ] + br label %bb + +bb: + %i = phi i64 [ 0, %outer.loop ], [ %i.next, %bb ] + %i.next = add i64 %i, 1 + + %e = add i64 %i, %j + %pi.j = getelementptr double* %p, i64 %e + %f = add i64 %i.next, %j + %pi.next.j = getelementptr double* %p, i64 %f + %x = load double* %pi.j + %y = load double* %pi.next.j + %z = fmul double %x, %y + store double %z, double* %pi.j + + %o = add i64 %j, %n + %g = add i64 %i, %o + %pi.j.next = getelementptr double* %p, i64 %g + %a = load double* %pi.j.next + %b = fmul double %x, %a + store double %b, double* %pi.j.next + + %exitcond = icmp eq i64 %i.next, %n + br i1 %exitcond, label %outer.latch, label %bb + +outer.latch: + %j.next = add i64 %j, %n + %h = icmp eq i64 %j.next, %m + br i1 %h, label %return, label %outer.loop + +return: + ret void +} + +; ScalarEvolution expands field offsets into constants, which allows it to +; do aggressive analysis. Contrast this with BasicAA, which works by +; recognizing GEP idioms. + +%struct.A = type { %struct.B, i32, i32 } +%struct.B = type { double } + +; CHECK: Function: foo: 7 pointers, 0 call sites +; CHECK: NoAlias: %struct.B* %B, i32* %Z +; CHECK: NoAlias: %struct.B* %B, %struct.B* %C +; CHECK: MustAlias: %struct.B* %C, i32* %Z +; CHECK: NoAlias: %struct.B* %B, i32* %X +; CHECK: MustAlias: i32* %X, i32* %Z +; CHECK: MustAlias: %struct.B* %C, i32* %Y +; CHECK: MustAlias: i32* %X, i32* %Y + +define void @foo() { +entry: + %A = alloca %struct.A + %B = getelementptr %struct.A* %A, i32 0, i32 0 + %Q = bitcast %struct.B* %B to %struct.A* + %Z = getelementptr %struct.A* %Q, i32 0, i32 1 + %C = getelementptr %struct.B* %B, i32 1 + %X = bitcast %struct.B* %C to i32* + %Y = getelementptr %struct.A* %A, i32 0, i32 1 + ret void +} + +; CHECK: Function: bar: 7 pointers, 0 call sites +; CHECK: NoAlias: %struct.B* %N, i32* %P +; CHECK: NoAlias: %struct.B* %N, %struct.B* %R +; CHECK: MustAlias: %struct.B* %R, i32* %P +; CHECK: NoAlias: %struct.B* %N, i32* %W +; CHECK: MustAlias: i32* %P, i32* %W +; CHECK: MustAlias: %struct.B* %R, i32* %V +; CHECK: MustAlias: i32* %V, i32* %W + +define void @bar() { + %M = alloca %struct.A + %N = getelementptr %struct.A* %M, i32 0, i32 0 + %O = bitcast %struct.B* %N to %struct.A* + %P = getelementptr %struct.A* %O, i32 0, i32 1 + %R = getelementptr %struct.B* %N, i32 1 + %W = bitcast %struct.B* %R to i32* + %V = getelementptr %struct.A* %M, i32 0, i32 1 + ret void +} + +; CHECK: 13 no alias responses +; CHECK: 26 may alias responses +; CHECK: 18 must alias responses diff --git a/test/Analysis/ScalarEvolution/sext-inreg.ll b/test/Analysis/ScalarEvolution/sext-inreg.ll new file mode 100644 index 0000000..23e1210 --- /dev/null +++ b/test/Analysis/ScalarEvolution/sext-inreg.ll @@ -0,0 +1,30 @@ +; RUN: opt < %s -analyze -scalar-evolution > %t +; RUN: grep {sext i57 \{0,+,199\}<%bb> to i64} %t | count 1 +; RUN: grep {sext i59 \{0,+,199\}<%bb> to i64} %t | count 1 + +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-darwin9.6" + +define i64 @foo(i64* nocapture %x, i64 %n) nounwind { +entry: + %t0 = icmp sgt i64 %n, 0 ; <i1> [#uses=1] + br i1 %t0, label %bb, label %return + +bb: ; preds = %bb, %entry + %i.01 = phi i64 [ 0, %entry ], [ %indvar.next, %bb ] ; <i32> [#uses=2] + %t1 = shl i64 %i.01, 7 ; <i32> [#uses=1] + %t2 = ashr i64 %t1, 7 ; <i32> [#uses=1] + %s1 = shl i64 %i.01, 5 ; <i32> [#uses=1] + %s2 = ashr i64 %s1, 5 ; <i32> [#uses=1] + %t3 = getelementptr i64* %x, i64 %i.01 ; <i64*> [#uses=1] + store i64 0, i64* %t3, align 1 + %indvar.next = add i64 %i.01, 199 ; <i32> [#uses=2] + %exitcond = icmp eq i64 %indvar.next, %n ; <i1> [#uses=1] + br i1 %exitcond, label %return, label %bb + +return: ; preds = %bb, %entry + %p = phi i64 [ 0, %entry ], [ %t2, %bb ] + %q = phi i64 [ 0, %entry ], [ %s2, %bb ] + %v = xor i64 %p, %q + ret i64 %v +} diff --git a/test/Analysis/ScalarEvolution/sext-iv-0.ll b/test/Analysis/ScalarEvolution/sext-iv-0.ll new file mode 100644 index 0000000..2af794f --- /dev/null +++ b/test/Analysis/ScalarEvolution/sext-iv-0.ll @@ -0,0 +1,31 @@ +; RUN: opt < %s -scalar-evolution -analyze \ +; RUN: | grep { --> \{-128,+,1\}<%bb1> Exits: 127} | count 5 + +; Convert (sext {-128,+,1}) to {sext(-128),+,sext(1)}, since the +; trip count is within range where this is safe. + +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" +target triple = "x86_64-unknown-linux-gnu" + +define void @foo(double* nocapture %x) nounwind { +bb1.thread: + br label %bb1 + +bb1: ; preds = %bb1, %bb1.thread + %i.0.reg2mem.0 = phi i64 [ -128, %bb1.thread ], [ %8, %bb1 ] ; <i64> [#uses=3] + %0 = trunc i64 %i.0.reg2mem.0 to i8 ; <i8> [#uses=1] + %1 = trunc i64 %i.0.reg2mem.0 to i9 ; <i8> [#uses=1] + %2 = sext i9 %1 to i64 ; <i64> [#uses=1] + %3 = getelementptr double* %x, i64 %2 ; <double*> [#uses=1] + %4 = load double* %3, align 8 ; <double> [#uses=1] + %5 = fmul double %4, 3.900000e+00 ; <double> [#uses=1] + %6 = sext i8 %0 to i64 ; <i64> [#uses=1] + %7 = getelementptr double* %x, i64 %6 ; <double*> [#uses=1] + store double %5, double* %7, align 8 + %8 = add i64 %i.0.reg2mem.0, 1 ; <i64> [#uses=2] + %9 = icmp sgt i64 %8, 127 ; <i1> [#uses=1] + br i1 %9, label %return, label %bb1 + +return: ; preds = %bb1 + ret void +} diff --git a/test/Analysis/ScalarEvolution/sext-iv-1.ll b/test/Analysis/ScalarEvolution/sext-iv-1.ll new file mode 100644 index 0000000..9063cbb --- /dev/null +++ b/test/Analysis/ScalarEvolution/sext-iv-1.ll @@ -0,0 +1,100 @@ +; RUN: opt < %s -scalar-evolution -analyze \ +; RUN: | grep { --> (sext i. \{.\*,+,.\*\}<%bb1> to i64)} | count 5 + +; Don't convert (sext {...,+,...}) to {sext(...),+,sext(...)} in cases +; where the trip count is not within range. + +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" +target triple = "x86_64-unknown-linux-gnu" + +define void @foo0(double* nocapture %x) nounwind { +bb1.thread: + br label %bb1 + +bb1: ; preds = %bb1, %bb1.thread + %i.0.reg2mem.0 = phi i64 [ -128, %bb1.thread ], [ %8, %bb1 ] ; <i64> [#uses=3] + %0 = trunc i64 %i.0.reg2mem.0 to i7 ; <i8> [#uses=1] + %1 = trunc i64 %i.0.reg2mem.0 to i9 ; <i8> [#uses=1] + %2 = sext i9 %1 to i64 ; <i64> [#uses=1] + %3 = getelementptr double* %x, i64 %2 ; <double*> [#uses=1] + %4 = load double* %3, align 8 ; <double> [#uses=1] + %5 = fmul double %4, 3.900000e+00 ; <double> [#uses=1] + %6 = sext i7 %0 to i64 ; <i64> [#uses=1] + %7 = getelementptr double* %x, i64 %6 ; <double*> [#uses=1] + store double %5, double* %7, align 8 + %8 = add i64 %i.0.reg2mem.0, 1 ; <i64> [#uses=2] + %9 = icmp sgt i64 %8, 127 ; <i1> [#uses=1] + br i1 %9, label %return, label %bb1 + +return: ; preds = %bb1 + ret void +} + +define void @foo1(double* nocapture %x) nounwind { +bb1.thread: + br label %bb1 + +bb1: ; preds = %bb1, %bb1.thread + %i.0.reg2mem.0 = phi i64 [ -128, %bb1.thread ], [ %8, %bb1 ] ; <i64> [#uses=3] + %0 = trunc i64 %i.0.reg2mem.0 to i8 ; <i8> [#uses=1] + %1 = trunc i64 %i.0.reg2mem.0 to i9 ; <i8> [#uses=1] + %2 = sext i9 %1 to i64 ; <i64> [#uses=1] + %3 = getelementptr double* %x, i64 %2 ; <double*> [#uses=1] + %4 = load double* %3, align 8 ; <double> [#uses=1] + %5 = fmul double %4, 3.900000e+00 ; <double> [#uses=1] + %6 = sext i8 %0 to i64 ; <i64> [#uses=1] + %7 = getelementptr double* %x, i64 %6 ; <double*> [#uses=1] + store double %5, double* %7, align 8 + %8 = add i64 %i.0.reg2mem.0, 1 ; <i64> [#uses=2] + %9 = icmp sgt i64 %8, 128 ; <i1> [#uses=1] + br i1 %9, label %return, label %bb1 + +return: ; preds = %bb1 + ret void +} + +define void @foo2(double* nocapture %x) nounwind { +bb1.thread: + br label %bb1 + +bb1: ; preds = %bb1, %bb1.thread + %i.0.reg2mem.0 = phi i64 [ -129, %bb1.thread ], [ %8, %bb1 ] ; <i64> [#uses=3] + %0 = trunc i64 %i.0.reg2mem.0 to i8 ; <i8> [#uses=1] + %1 = trunc i64 %i.0.reg2mem.0 to i9 ; <i8> [#uses=1] + %2 = sext i9 %1 to i64 ; <i64> [#uses=1] + %3 = getelementptr double* %x, i64 %2 ; <double*> [#uses=1] + %4 = load double* %3, align 8 ; <double> [#uses=1] + %5 = fmul double %4, 3.900000e+00 ; <double> [#uses=1] + %6 = sext i8 %0 to i64 ; <i64> [#uses=1] + %7 = getelementptr double* %x, i64 %6 ; <double*> [#uses=1] + store double %5, double* %7, align 8 + %8 = add i64 %i.0.reg2mem.0, 1 ; <i64> [#uses=2] + %9 = icmp sgt i64 %8, 127 ; <i1> [#uses=1] + br i1 %9, label %return, label %bb1 + +return: ; preds = %bb1 + ret void +} + +define void @foo3(double* nocapture %x) nounwind { +bb1.thread: + br label %bb1 + +bb1: ; preds = %bb1, %bb1.thread + %i.0.reg2mem.0 = phi i64 [ -128, %bb1.thread ], [ %8, %bb1 ] ; <i64> [#uses=3] + %0 = trunc i64 %i.0.reg2mem.0 to i8 ; <i8> [#uses=1] + %1 = trunc i64 %i.0.reg2mem.0 to i9 ; <i8> [#uses=1] + %2 = sext i9 %1 to i64 ; <i64> [#uses=1] + %3 = getelementptr double* %x, i64 %2 ; <double*> [#uses=1] + %4 = load double* %3, align 8 ; <double> [#uses=1] + %5 = fmul double %4, 3.900000e+00 ; <double> [#uses=1] + %6 = sext i8 %0 to i64 ; <i64> [#uses=1] + %7 = getelementptr double* %x, i64 %6 ; <double*> [#uses=1] + store double %5, double* %7, align 8 + %8 = add i64 %i.0.reg2mem.0, -1 ; <i64> [#uses=2] + %9 = icmp sgt i64 %8, 127 ; <i1> [#uses=1] + br i1 %9, label %return, label %bb1 + +return: ; preds = %bb1 + ret void +} diff --git a/test/Analysis/ScalarEvolution/sext-iv-2.ll b/test/Analysis/ScalarEvolution/sext-iv-2.ll new file mode 100644 index 0000000..97e252c --- /dev/null +++ b/test/Analysis/ScalarEvolution/sext-iv-2.ll @@ -0,0 +1,74 @@ +; RUN: opt < %s -analyze -scalar-evolution | FileCheck %s + +; CHECK: %tmp3 = sext i8 %tmp2 to i32 +; CHECK: --> (sext i8 {0,+,1}<%bb1> to i32) Exits: -1 +; CHECK: %tmp4 = mul i32 %tmp3, %i.02 +; CHECK: --> ((sext i8 {0,+,1}<%bb1> to i32) * {0,+,1}<%bb>) Exits: {0,+,-1}<%bb> + +; These sexts are not foldable. + +target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64" + +@table = common global [32 x [256 x i32]] zeroinitializer, align 32 ; <[32 x [256 x i32]]*> [#uses=2] + +define i32 @main() nounwind { +entry: + br i1 false, label %bb5, label %bb.nph3 + +bb.nph3: ; preds = %entry + br label %bb + +bb: ; preds = %bb4, %bb.nph3 + %i.02 = phi i32 [ %tmp10, %bb4 ], [ 0, %bb.nph3 ] ; <i32> [#uses=3] + br i1 false, label %bb3, label %bb.nph + +bb.nph: ; preds = %bb + br label %bb1 + +bb1: ; preds = %bb2, %bb.nph + %j.01 = phi i32 [ %tmp8, %bb2 ], [ 0, %bb.nph ] ; <i32> [#uses=3] + %tmp2 = trunc i32 %j.01 to i8 ; <i8> [#uses=1] + %tmp3 = sext i8 %tmp2 to i32 ; <i32> [#uses=1] + %tmp4 = mul i32 %tmp3, %i.02 ; <i32> [#uses=1] + %tmp5 = sext i32 %i.02 to i64 ; <i64> [#uses=1] + %tmp6 = sext i32 %j.01 to i64 ; <i64> [#uses=1] + %tmp7 = getelementptr [32 x [256 x i32]]* @table, i64 0, i64 %tmp5, i64 %tmp6 ; <i32*> [#uses=1] + store i32 %tmp4, i32* %tmp7, align 4 + %tmp8 = add i32 %j.01, 1 ; <i32> [#uses=2] + br label %bb2 + +bb2: ; preds = %bb1 + %phitmp1 = icmp sgt i32 %tmp8, 255 ; <i1> [#uses=1] + br i1 %phitmp1, label %bb2.bb3_crit_edge, label %bb1 + +bb2.bb3_crit_edge: ; preds = %bb2 + br label %bb3 + +bb3: ; preds = %bb2.bb3_crit_edge, %bb + %tmp10 = add i32 %i.02, 1 ; <i32> [#uses=2] + br label %bb4 + +bb4: ; preds = %bb3 + %phitmp = icmp sgt i32 %tmp10, 31 ; <i1> [#uses=1] + br i1 %phitmp, label %bb4.bb5_crit_edge, label %bb + +bb4.bb5_crit_edge: ; preds = %bb4 + br label %bb5 + +bb5: ; preds = %bb4.bb5_crit_edge, %entry + %tmp12 = load i32* getelementptr ([32 x [256 x i32]]* @table, i64 0, i64 9, i64 132), align 16 ; <i32> [#uses=1] + %tmp13 = icmp eq i32 %tmp12, -1116 ; <i1> [#uses=1] + br i1 %tmp13, label %bb7, label %bb6 + +bb6: ; preds = %bb5 + call void @abort() noreturn nounwind + unreachable + +bb7: ; preds = %bb5 + br label %return + +return: ; preds = %bb7 + ret i32 0 +} + +declare void @abort() noreturn nounwind diff --git a/test/Analysis/ScalarEvolution/smax.ll b/test/Analysis/ScalarEvolution/smax.ll new file mode 100644 index 0000000..15dd744 --- /dev/null +++ b/test/Analysis/ScalarEvolution/smax.ll @@ -0,0 +1,12 @@ +; RUN: opt < %s -analyze -scalar-evolution | grep smax | count 2 +; RUN: opt < %s -analyze -scalar-evolution | grep \ +; RUN: {%. smax %. smax %.} +; PR1614 + +define i32 @x(i32 %a, i32 %b, i32 %c) { + %A = icmp sgt i32 %a, %b + %B = select i1 %A, i32 %a, i32 %b + %C = icmp sle i32 %c, %B + %D = select i1 %C, i32 %B, i32 %c + ret i32 %D +} diff --git a/test/Analysis/ScalarEvolution/trip-count.ll b/test/Analysis/ScalarEvolution/trip-count.ll new file mode 100644 index 0000000..d750d4a --- /dev/null +++ b/test/Analysis/ScalarEvolution/trip-count.ll @@ -0,0 +1,29 @@ +; RUN: opt < %s -analyze -scalar-evolution \ +; RUN: -scalar-evolution-max-iterations=0 | grep {backedge-taken count is 10000} +; PR1101 + +@A = weak global [1000 x i32] zeroinitializer, align 32 + + +define void @test(i32 %N) { +entry: + "alloca point" = bitcast i32 0 to i32 ; <i32> [#uses=0] + br label %bb3 + +bb: ; preds = %bb3 + %tmp = getelementptr [1000 x i32]* @A, i32 0, i32 %i.0 ; <i32*> [#uses=1] + store i32 123, i32* %tmp + %tmp2 = add i32 %i.0, 1 ; <i32> [#uses=1] + br label %bb3 + +bb3: ; preds = %bb, %entry + %i.0 = phi i32 [ 0, %entry ], [ %tmp2, %bb ] ; <i32> [#uses=3] + %tmp3 = icmp sle i32 %i.0, 9999 ; <i1> [#uses=1] + br i1 %tmp3, label %bb, label %bb5 + +bb5: ; preds = %bb3 + br label %return + +return: ; preds = %bb5 + ret void +} diff --git a/test/Analysis/ScalarEvolution/trip-count2.ll b/test/Analysis/ScalarEvolution/trip-count2.ll new file mode 100644 index 0000000..79f3161 --- /dev/null +++ b/test/Analysis/ScalarEvolution/trip-count2.ll @@ -0,0 +1,35 @@ +; RUN: opt < %s -analyze -scalar-evolution | \ +; RUN: grep {backedge-taken count is 4} +; PR1101 + +@A = weak global [1000 x i32] zeroinitializer, align 32 + + +define void @test(i32 %N) { +entry: + "alloca point" = bitcast i32 0 to i32 ; <i32> [#uses=0] + br label %bb3 + +bb: ; preds = %bb3 + %tmp = getelementptr [1000 x i32]* @A, i32 0, i32 %i.0 ; <i32*> [#uses=1] + store i32 123, i32* %tmp + %tmp4 = mul i32 %i.0, 4 ; <i32> [#uses=1] + %tmp5 = or i32 %tmp4, 1 + %tmp61 = xor i32 %tmp5, -2147483648 + %tmp6 = trunc i32 %tmp61 to i16 + %tmp71 = shl i16 %tmp6, 2 + %tmp7 = zext i16 %tmp71 to i32 + %tmp2 = add i32 %tmp7, %i.0 + br label %bb3 + +bb3: ; preds = %bb, %entry + %i.0 = phi i32 [ 0, %entry ], [ %tmp2, %bb ] ; <i32> [#uses=3] + %tmp3 = icmp sle i32 %i.0, 9999 ; <i1> [#uses=1] + br i1 %tmp3, label %bb, label %bb5 + +bb5: ; preds = %bb3 + br label %return + +return: ; preds = %bb5 + ret void +} diff --git a/test/Analysis/ScalarEvolution/trip-count3.ll b/test/Analysis/ScalarEvolution/trip-count3.ll new file mode 100644 index 0000000..10b798b --- /dev/null +++ b/test/Analysis/ScalarEvolution/trip-count3.ll @@ -0,0 +1,78 @@ +; RUN: opt < %s -scalar-evolution -analyze \ +; RUN: | grep {Loop %bb3\\.i: Unpredictable backedge-taken count\\.} + +; ScalarEvolution can't compute a trip count because it doesn't know if +; dividing by the stride will have a remainder. This could theoretically +; be teaching it how to use a more elaborate trip count computation. + +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" +target triple = "x86_64-unknown-linux-gnu" + %struct.FILE = type { i32, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, %struct._IO_marker*, %struct.FILE*, i32, i32, i64, i16, i8, [1 x i8], i8*, i64, i8*, i8*, i8*, i8*, i64, i32, [20 x i8] } + %struct.SHA_INFO = type { [5 x i32], i32, i32, [16 x i32] } + %struct._IO_marker = type { %struct._IO_marker*, %struct.FILE*, i32 } +@_2E_str = external constant [26 x i8] ; <[26 x i8]*> [#uses=0] +@stdin = external global %struct.FILE* ; <%struct.FILE**> [#uses=0] +@_2E_str1 = external constant [3 x i8] ; <[3 x i8]*> [#uses=0] +@_2E_str12 = external constant [30 x i8] ; <[30 x i8]*> [#uses=0] + +declare void @sha_init(%struct.SHA_INFO* nocapture) nounwind + +declare fastcc void @sha_transform(%struct.SHA_INFO* nocapture) nounwind + +declare void @sha_print(%struct.SHA_INFO* nocapture) nounwind + +declare i32 @printf(i8* nocapture, ...) nounwind + +declare void @sha_final(%struct.SHA_INFO* nocapture) nounwind + +declare void @llvm.memset.i64(i8* nocapture, i8, i64, i32) nounwind + +declare void @sha_update(%struct.SHA_INFO* nocapture, i8* nocapture, i32) nounwind + +declare void @llvm.memcpy.i64(i8* nocapture, i8* nocapture, i64, i32) nounwind + +declare i64 @fread(i8* noalias nocapture, i64, i64, %struct.FILE* noalias nocapture) nounwind + +declare i32 @main(i32, i8** nocapture) nounwind + +declare noalias %struct.FILE* @fopen(i8* noalias nocapture, i8* noalias nocapture) nounwind + +declare i32 @fclose(%struct.FILE* nocapture) nounwind + +declare void @sha_stream(%struct.SHA_INFO* nocapture, %struct.FILE* nocapture) nounwind + +define void @sha_stream_bb3_2E_i(%struct.SHA_INFO* %sha_info, i8* %data1, i32, i8** %buffer_addr.0.i.out, i32* %count_addr.0.i.out) nounwind { +newFuncRoot: + br label %bb3.i + +sha_update.exit.exitStub: ; preds = %bb3.i + store i8* %buffer_addr.0.i, i8** %buffer_addr.0.i.out + store i32 %count_addr.0.i, i32* %count_addr.0.i.out + ret void + +bb2.i: ; preds = %bb3.i + %1 = getelementptr %struct.SHA_INFO* %sha_info, i64 0, i32 3 ; <[16 x i32]*> [#uses=1] + %2 = bitcast [16 x i32]* %1 to i8* ; <i8*> [#uses=1] + call void @llvm.memcpy.i64(i8* %2, i8* %buffer_addr.0.i, i64 64, i32 1) nounwind + %3 = getelementptr %struct.SHA_INFO* %sha_info, i64 0, i32 3, i64 0 ; <i32*> [#uses=1] + %4 = bitcast i32* %3 to i8* ; <i8*> [#uses=1] + br label %codeRepl + +codeRepl: ; preds = %bb2.i + call void @sha_stream_bb3_2E_i_bb1_2E_i_2E_i(i8* %4) + br label %byte_reverse.exit.i + +byte_reverse.exit.i: ; preds = %codeRepl + call fastcc void @sha_transform(%struct.SHA_INFO* %sha_info) nounwind + %5 = getelementptr i8* %buffer_addr.0.i, i64 64 ; <i8*> [#uses=1] + %6 = add i32 %count_addr.0.i, -64 ; <i32> [#uses=1] + br label %bb3.i + +bb3.i: ; preds = %byte_reverse.exit.i, %newFuncRoot + %buffer_addr.0.i = phi i8* [ %data1, %newFuncRoot ], [ %5, %byte_reverse.exit.i ] ; <i8*> [#uses=3] + %count_addr.0.i = phi i32 [ %0, %newFuncRoot ], [ %6, %byte_reverse.exit.i ] ; <i32> [#uses=3] + %7 = icmp sgt i32 %count_addr.0.i, 63 ; <i1> [#uses=1] + br i1 %7, label %bb2.i, label %sha_update.exit.exitStub +} + +declare void @sha_stream_bb3_2E_i_bb1_2E_i_2E_i(i8*) nounwind diff --git a/test/Analysis/ScalarEvolution/trip-count4.ll b/test/Analysis/ScalarEvolution/trip-count4.ll new file mode 100644 index 0000000..116f62d --- /dev/null +++ b/test/Analysis/ScalarEvolution/trip-count4.ll @@ -0,0 +1,24 @@ +; RUN: opt < %s -analyze -scalar-evolution \ +; RUN: | grep {sext.*trunc.*Exits: 11} + +; ScalarEvolution should be able to compute a loop exit value for %indvar.i8. + +define void @another_count_down_signed(double* %d, i64 %n) nounwind { +entry: + br label %loop + +loop: ; preds = %loop, %entry + %indvar = phi i64 [ %n, %entry ], [ %indvar.next, %loop ] ; <i64> [#uses=4] + %s0 = shl i64 %indvar, 8 ; <i64> [#uses=1] + %indvar.i8 = ashr i64 %s0, 8 ; <i64> [#uses=1] + %t0 = getelementptr double* %d, i64 %indvar.i8 ; <double*> [#uses=2] + %t1 = load double* %t0 ; <double> [#uses=1] + %t2 = fmul double %t1, 1.000000e-01 ; <double> [#uses=1] + store double %t2, double* %t0 + %indvar.next = sub i64 %indvar, 1 ; <i64> [#uses=2] + %exitcond = icmp eq i64 %indvar.next, 10 ; <i1> [#uses=1] + br i1 %exitcond, label %return, label %loop + +return: ; preds = %loop + ret void +} diff --git a/test/Analysis/ScalarEvolution/trip-count5.ll b/test/Analysis/ScalarEvolution/trip-count5.ll new file mode 100644 index 0000000..1194a1d --- /dev/null +++ b/test/Analysis/ScalarEvolution/trip-count5.ll @@ -0,0 +1,48 @@ +; RUN: opt < %s -analyze -scalar-evolution > %t +; RUN: grep sext %t | count 2 +; RUN: not grep {(sext} %t + +; ScalarEvolution should be able to compute a maximum trip count +; value sufficient to fold away both sext casts. + +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" + +define float @t(float* %pTmp1, float* %peakWeight, float* %nrgReducePeakrate, i32 %bim) nounwind { +entry: + %tmp3 = load float* %peakWeight, align 4 ; <float> [#uses=2] + %tmp2538 = icmp sgt i32 %bim, 0 ; <i1> [#uses=1] + br i1 %tmp2538, label %bb.nph, label %bb4 + +bb.nph: ; preds = %entry + br label %bb + +bb: ; preds = %bb1, %bb.nph + %distERBhi.036 = phi float [ %tmp10, %bb1 ], [ 0.000000e+00, %bb.nph ] ; <float> [#uses=1] + %hiPart.035 = phi i32 [ %tmp12, %bb1 ], [ 0, %bb.nph ] ; <i32> [#uses=2] + %peakCount.034 = phi float [ %tmp19, %bb1 ], [ %tmp3, %bb.nph ] ; <float> [#uses=1] + %tmp6 = sext i32 %hiPart.035 to i64 ; <i64> [#uses=1] + %tmp7 = getelementptr float* %pTmp1, i64 %tmp6 ; <float*> [#uses=1] + %tmp8 = load float* %tmp7, align 4 ; <float> [#uses=1] + %tmp10 = fadd float %tmp8, %distERBhi.036 ; <float> [#uses=3] + %tmp12 = add i32 %hiPart.035, 1 ; <i32> [#uses=3] + %tmp15 = sext i32 %tmp12 to i64 ; <i64> [#uses=1] + %tmp16 = getelementptr float* %peakWeight, i64 %tmp15 ; <float*> [#uses=1] + %tmp17 = load float* %tmp16, align 4 ; <float> [#uses=1] + %tmp19 = fadd float %tmp17, %peakCount.034 ; <float> [#uses=2] + br label %bb1 + +bb1: ; preds = %bb + %tmp21 = fcmp olt float %tmp10, 2.500000e+00 ; <i1> [#uses=1] + %tmp25 = icmp slt i32 %tmp12, %bim ; <i1> [#uses=1] + %tmp27 = and i1 %tmp21, %tmp25 ; <i1> [#uses=1] + br i1 %tmp27, label %bb, label %bb1.bb4_crit_edge + +bb1.bb4_crit_edge: ; preds = %bb1 + br label %bb4 + +bb4: ; preds = %bb1.bb4_crit_edge, %entry + %distERBhi.0.lcssa = phi float [ %tmp10, %bb1.bb4_crit_edge ], [ 0.000000e+00, %entry ] ; <float> [#uses=1] + %peakCount.0.lcssa = phi float [ %tmp19, %bb1.bb4_crit_edge ], [ %tmp3, %entry ] ; <float> [#uses=1] + %tmp31 = fdiv float %peakCount.0.lcssa, %distERBhi.0.lcssa ; <float> [#uses=1] + ret float %tmp31 +} diff --git a/test/Analysis/ScalarEvolution/trip-count6.ll b/test/Analysis/ScalarEvolution/trip-count6.ll new file mode 100644 index 0000000..956fb81 --- /dev/null +++ b/test/Analysis/ScalarEvolution/trip-count6.ll @@ -0,0 +1,37 @@ +; RUN: opt < %s -analyze -scalar-evolution \ +; RUN: | grep {max backedge-taken count is 1\$} + +@mode_table = global [4 x i32] zeroinitializer ; <[4 x i32]*> [#uses=1] + +define i8 @f() { +entry: + tail call i32 @fegetround( ) ; <i32>:0 [#uses=1] + br label %bb + +bb: ; preds = %bb4, %entry + %mode.0 = phi i8 [ 0, %entry ], [ %indvar.next, %bb4 ] ; <i8> [#uses=4] + zext i8 %mode.0 to i32 ; <i32>:1 [#uses=1] + getelementptr [4 x i32]* @mode_table, i32 0, i32 %1 ; <i32*>:2 [#uses=1] + load i32* %2, align 4 ; <i32>:3 [#uses=1] + icmp eq i32 %3, %0 ; <i1>:4 [#uses=1] + br i1 %4, label %bb1, label %bb2 + +bb1: ; preds = %bb + ret i8 %mode.0 + +bb2: ; preds = %bb + icmp eq i8 %mode.0, 1 ; <i1>:5 [#uses=1] + br i1 %5, label %bb5, label %bb4 + +bb4: ; preds = %bb2 + %indvar.next = add i8 %mode.0, 1 ; <i8> [#uses=1] + br label %bb + +bb5: ; preds = %bb2 + tail call void @raise_exception( ) noreturn + unreachable +} + +declare i32 @fegetround() + +declare void @raise_exception() noreturn diff --git a/test/Analysis/ScalarEvolution/trip-count7.ll b/test/Analysis/ScalarEvolution/trip-count7.ll new file mode 100644 index 0000000..a8b797e --- /dev/null +++ b/test/Analysis/ScalarEvolution/trip-count7.ll @@ -0,0 +1,150 @@ +; RUN: opt < %s -analyze -scalar-evolution \ +; RUN: | grep {Loop %bb7.i: Unpredictable backedge-taken count\\.} + +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" + + %struct.complex = type { float, float } + %struct.element = type { i32, i32 } + %struct.node = type { %struct.node*, %struct.node*, i32 } +@seed = external global i64 ; <i64*> [#uses=0] +@_2E_str = external constant [18 x i8], align 1 ; <[18 x i8]*> [#uses=0] +@_2E_str1 = external constant [4 x i8], align 1 ; <[4 x i8]*> [#uses=0] +@value = external global float ; <float*> [#uses=0] +@fixed = external global float ; <float*> [#uses=0] +@floated = external global float ; <float*> [#uses=0] +@permarray = external global [11 x i32], align 32 ; <[11 x i32]*> [#uses=0] +@pctr = external global i32 ; <i32*> [#uses=0] +@tree = external global %struct.node* ; <%struct.node**> [#uses=0] +@stack = external global [4 x i32], align 16 ; <[4 x i32]*> [#uses=0] +@cellspace = external global [19 x %struct.element], align 32 ; <[19 x %struct.element]*> [#uses=0] +@freelist = external global i32 ; <i32*> [#uses=0] +@movesdone = external global i32 ; <i32*> [#uses=0] +@ima = external global [41 x [41 x i32]], align 32 ; <[41 x [41 x i32]]*> [#uses=0] +@imb = external global [41 x [41 x i32]], align 32 ; <[41 x [41 x i32]]*> [#uses=0] +@imr = external global [41 x [41 x i32]], align 32 ; <[41 x [41 x i32]]*> [#uses=0] +@rma = external global [41 x [41 x float]], align 32 ; <[41 x [41 x float]]*> [#uses=0] +@rmb = external global [41 x [41 x float]], align 32 ; <[41 x [41 x float]]*> [#uses=0] +@rmr = external global [41 x [41 x float]], align 32 ; <[41 x [41 x float]]*> [#uses=0] +@piececount = external global [4 x i32], align 16 ; <[4 x i32]*> [#uses=0] +@class = external global [13 x i32], align 32 ; <[13 x i32]*> [#uses=0] +@piecemax = external global [13 x i32], align 32 ; <[13 x i32]*> [#uses=0] +@puzzl = external global [512 x i32], align 32 ; <[512 x i32]*> [#uses=0] +@p = external global [13 x [512 x i32]], align 32 ; <[13 x [512 x i32]]*> [#uses=0] +@n = external global i32 ; <i32*> [#uses=0] +@kount = external global i32 ; <i32*> [#uses=0] +@sortlist = external global [5001 x i32], align 32 ; <[5001 x i32]*> [#uses=0] +@biggest = external global i32 ; <i32*> [#uses=0] +@littlest = external global i32 ; <i32*> [#uses=0] +@top = external global i32 ; <i32*> [#uses=0] +@z = external global [257 x %struct.complex], align 32 ; <[257 x %struct.complex]*> [#uses=0] +@w = external global [257 x %struct.complex], align 32 ; <[257 x %struct.complex]*> [#uses=0] +@e = external global [130 x %struct.complex], align 32 ; <[130 x %struct.complex]*> [#uses=0] +@zr = external global float ; <float*> [#uses=0] +@zi = external global float ; <float*> [#uses=0] + +declare void @Initrand() nounwind + +declare i32 @Rand() nounwind + +declare void @Try(i32, i32*, i32*, i32*, i32*, i32*) nounwind + +declare i32 @puts(i8* nocapture) nounwind + +declare void @Queens(i32) nounwind + +declare i32 @printf(i8* nocapture, ...) nounwind + +declare i32 @main() nounwind + +declare void @Doit() nounwind + +declare void @Doit_bb7([15 x i32]*, [17 x i32]*, [9 x i32]*) nounwind + +define void @Doit_bb7_2E_i([9 x i32]* %x1, [15 x i32]* %c, [17 x i32]* %b, [9 x i32]* %a, i32* %q, i32* %x1.sub, i32* %b9, i32* %a10, i32* %c11) nounwind { +newFuncRoot: + br label %bb7.i + +Try.exit.exitStub: ; preds = %bb7.i + ret void + +bb.i: ; preds = %bb7.i + %tmp = add i32 %j.0.i, 1 ; <i32> [#uses=5] + store i32 0, i32* %q, align 4 + %tmp1 = sext i32 %tmp to i64 ; <i64> [#uses=1] + %tmp2 = getelementptr [9 x i32]* %a, i64 0, i64 %tmp1 ; <i32*> [#uses=1] + %tmp3 = load i32* %tmp2, align 4 ; <i32> [#uses=1] + %tmp4 = icmp eq i32 %tmp3, 0 ; <i1> [#uses=1] + br i1 %tmp4, label %bb.i.bb7.i.backedge_crit_edge, label %bb1.i + +bb1.i: ; preds = %bb.i + %tmp5 = add i32 %j.0.i, 2 ; <i32> [#uses=1] + %tmp6 = sext i32 %tmp5 to i64 ; <i64> [#uses=1] + %tmp7 = getelementptr [17 x i32]* %b, i64 0, i64 %tmp6 ; <i32*> [#uses=1] + %tmp8 = load i32* %tmp7, align 4 ; <i32> [#uses=1] + %tmp9 = icmp eq i32 %tmp8, 0 ; <i1> [#uses=1] + br i1 %tmp9, label %bb1.i.bb7.i.backedge_crit_edge, label %bb2.i + +bb2.i: ; preds = %bb1.i + %tmp10 = sub i32 7, %j.0.i ; <i32> [#uses=1] + %tmp11 = sext i32 %tmp10 to i64 ; <i64> [#uses=1] + %tmp12 = getelementptr [15 x i32]* %c, i64 0, i64 %tmp11 ; <i32*> [#uses=1] + %tmp13 = load i32* %tmp12, align 4 ; <i32> [#uses=1] + %tmp14 = icmp eq i32 %tmp13, 0 ; <i1> [#uses=1] + br i1 %tmp14, label %bb2.i.bb7.i.backedge_crit_edge, label %bb3.i + +bb3.i: ; preds = %bb2.i + %tmp15 = getelementptr [9 x i32]* %x1, i64 0, i64 1 ; <i32*> [#uses=1] + store i32 %tmp, i32* %tmp15, align 4 + %tmp16 = sext i32 %tmp to i64 ; <i64> [#uses=1] + %tmp17 = getelementptr [9 x i32]* %a, i64 0, i64 %tmp16 ; <i32*> [#uses=1] + store i32 0, i32* %tmp17, align 4 + %tmp18 = add i32 %j.0.i, 2 ; <i32> [#uses=1] + %tmp19 = sext i32 %tmp18 to i64 ; <i64> [#uses=1] + %tmp20 = getelementptr [17 x i32]* %b, i64 0, i64 %tmp19 ; <i32*> [#uses=1] + store i32 0, i32* %tmp20, align 4 + %tmp21 = sub i32 7, %j.0.i ; <i32> [#uses=1] + %tmp22 = sext i32 %tmp21 to i64 ; <i64> [#uses=1] + %tmp23 = getelementptr [15 x i32]* %c, i64 0, i64 %tmp22 ; <i32*> [#uses=1] + store i32 0, i32* %tmp23, align 4 + call void @Try(i32 2, i32* %q, i32* %b9, i32* %a10, i32* %c11, i32* %x1.sub) nounwind + %tmp24 = load i32* %q, align 4 ; <i32> [#uses=1] + %tmp25 = icmp eq i32 %tmp24, 0 ; <i1> [#uses=1] + br i1 %tmp25, label %bb5.i, label %bb3.i.bb7.i.backedge_crit_edge + +bb5.i: ; preds = %bb3.i + %tmp26 = sext i32 %tmp to i64 ; <i64> [#uses=1] + %tmp27 = getelementptr [9 x i32]* %a, i64 0, i64 %tmp26 ; <i32*> [#uses=1] + store i32 1, i32* %tmp27, align 4 + %tmp28 = add i32 %j.0.i, 2 ; <i32> [#uses=1] + %tmp29 = sext i32 %tmp28 to i64 ; <i64> [#uses=1] + %tmp30 = getelementptr [17 x i32]* %b, i64 0, i64 %tmp29 ; <i32*> [#uses=1] + store i32 1, i32* %tmp30, align 4 + %tmp31 = sub i32 7, %j.0.i ; <i32> [#uses=1] + %tmp32 = sext i32 %tmp31 to i64 ; <i64> [#uses=1] + %tmp33 = getelementptr [15 x i32]* %c, i64 0, i64 %tmp32 ; <i32*> [#uses=1] + store i32 1, i32* %tmp33, align 4 + br label %bb7.i.backedge + +bb7.i.backedge: ; preds = %bb3.i.bb7.i.backedge_crit_edge, %bb2.i.bb7.i.backedge_crit_edge, %bb1.i.bb7.i.backedge_crit_edge, %bb.i.bb7.i.backedge_crit_edge, %bb5.i + br label %bb7.i + +bb7.i: ; preds = %bb7.i.backedge, %newFuncRoot + %j.0.i = phi i32 [ 0, %newFuncRoot ], [ %tmp, %bb7.i.backedge ] ; <i32> [#uses=8] + %tmp34 = load i32* %q, align 4 ; <i32> [#uses=1] + %tmp35 = icmp eq i32 %tmp34, 0 ; <i1> [#uses=1] + %tmp36 = icmp ne i32 %j.0.i, 8 ; <i1> [#uses=1] + %tmp37 = and i1 %tmp35, %tmp36 ; <i1> [#uses=1] + br i1 %tmp37, label %bb.i, label %Try.exit.exitStub + +bb.i.bb7.i.backedge_crit_edge: ; preds = %bb.i + br label %bb7.i.backedge + +bb1.i.bb7.i.backedge_crit_edge: ; preds = %bb1.i + br label %bb7.i.backedge + +bb2.i.bb7.i.backedge_crit_edge: ; preds = %bb2.i + br label %bb7.i.backedge + +bb3.i.bb7.i.backedge_crit_edge: ; preds = %bb3.i + br label %bb7.i.backedge +} diff --git a/test/Analysis/ScalarEvolution/trip-count8.ll b/test/Analysis/ScalarEvolution/trip-count8.ll new file mode 100644 index 0000000..ac5ee60 --- /dev/null +++ b/test/Analysis/ScalarEvolution/trip-count8.ll @@ -0,0 +1,37 @@ +; RUN: opt < %s -analyze -scalar-evolution \ +; RUN: | grep {Loop %for\\.body: backedge-taken count is (-1 + \[%\]ecx)} +; PR4599 + +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" + +define i32 @foo(i32 %ecx) nounwind { +entry: + %cmp2 = icmp eq i32 %ecx, 0 ; <i1> [#uses=1] + br i1 %cmp2, label %for.end, label %bb.nph + +for.cond: ; preds = %for.inc + %cmp = icmp ult i32 %inc, %ecx ; <i1> [#uses=1] + br i1 %cmp, label %for.body, label %for.cond.for.end_crit_edge + +for.cond.for.end_crit_edge: ; preds = %for.cond + %phitmp = add i32 %i.01, 2 ; <i32> [#uses=1] + br label %for.end + +bb.nph: ; preds = %entry + br label %for.body + +for.body: ; preds = %bb.nph, %for.cond + %i.01 = phi i32 [ %inc, %for.cond ], [ 0, %bb.nph ] ; <i32> [#uses=3] + %call = call i32 @bar(i32 %i.01) nounwind ; <i32> [#uses=0] + br label %for.inc + +for.inc: ; preds = %for.body + %inc = add i32 %i.01, 1 ; <i32> [#uses=2] + br label %for.cond + +for.end: ; preds = %for.cond.for.end_crit_edge, %entry + %i.0.lcssa = phi i32 [ %phitmp, %for.cond.for.end_crit_edge ], [ 1, %entry ] ; <i32> [#uses=1] + ret i32 %i.0.lcssa +} + +declare i32 @bar(i32) diff --git a/test/Analysis/ScalarEvolution/trip-count9.ll b/test/Analysis/ScalarEvolution/trip-count9.ll new file mode 100644 index 0000000..9180f2b --- /dev/null +++ b/test/Analysis/ScalarEvolution/trip-count9.ll @@ -0,0 +1,408 @@ +; RUN: opt -analyze -scalar-evolution -S < %s | FileCheck %s + +; Every combination of +; - starting at 0, 1, or %x +; - steping by 1 or 2 +; - stopping at %n or %n*2 +; - using nsw, or not + +; Some of these represent missed opportunities. + +; CHECK: Determining loop execution counts for: @foo +; CHECK: Loop %loop: backedge-taken count is (-1 + %n) +; CHECK: Loop %loop: max backedge-taken count is 6 +define void @foo(i4 %n) { +entry: + %s = icmp sgt i4 %n, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ 0, %entry ], [ %i.next, %loop ] + %i.next = add i4 %i, 1 + %t = icmp slt i4 %i.next, %n + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @step2 +; CHECK: Loop %loop: Unpredictable backedge-taken count. +; CHECK: Loop %loop: Unpredictable max backedge-taken count. +define void @step2(i4 %n) { +entry: + %s = icmp sgt i4 %n, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ 0, %entry ], [ %i.next, %loop ] + %i.next = add i4 %i, 2 + %t = icmp slt i4 %i.next, %n + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @start1 +; CHECK: Loop %loop: backedge-taken count is (-2 + (2 smax %n)) +; CHECK: Loop %loop: max backedge-taken count is 5 +define void @start1(i4 %n) { +entry: + %s = icmp sgt i4 %n, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ 1, %entry ], [ %i.next, %loop ] + %i.next = add i4 %i, 1 + %t = icmp slt i4 %i.next, %n + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @start1_step2 +; CHECK: Loop %loop: Unpredictable backedge-taken count. +; CHECK: Loop %loop: Unpredictable max backedge-taken count. +define void @start1_step2(i4 %n) { +entry: + %s = icmp sgt i4 %n, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ 1, %entry ], [ %i.next, %loop ] + %i.next = add i4 %i, 2 + %t = icmp slt i4 %i.next, %n + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @startx +; CHECK: Loop %loop: backedge-taken count is (-1 + (-1 * %x) + ((1 + %x) smax %n)) +; CHECK: Loop %loop: max backedge-taken count is -1 +define void @startx(i4 %n, i4 %x) { +entry: + %s = icmp sgt i4 %n, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ %x, %entry ], [ %i.next, %loop ] + %i.next = add i4 %i, 1 + %t = icmp slt i4 %i.next, %n + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @startx_step2 +; CHECK: Loop %loop: Unpredictable backedge-taken count. +; CHECK: Loop %loop: Unpredictable max backedge-taken count. +define void @startx_step2(i4 %n, i4 %x) { +entry: + %s = icmp sgt i4 %n, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ %x, %entry ], [ %i.next, %loop ] + %i.next = add i4 %i, 2 + %t = icmp slt i4 %i.next, %n + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @nsw +; CHECK: Loop %loop: backedge-taken count is (-1 + %n) +; CHECK: Loop %loop: max backedge-taken count is 6 +define void @nsw(i4 %n) { +entry: + %s = icmp sgt i4 %n, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ 0, %entry ], [ %i.next, %loop ] + %i.next = add nsw i4 %i, 1 + %t = icmp slt i4 %i.next, %n + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; Be careful with this one. If %n is INT4_MAX, %i.next will wrap. The nsw bit +; says that the result is undefined, but ScalarEvolution must respect that +; subsequent passes may result the undefined behavior in predictable ways. +; CHECK: Determining loop execution counts for: @nsw_step2 +; CHECK: Loop %loop: Unpredictable backedge-taken count. +; CHECK: Loop %loop: Unpredictable max backedge-taken count. +define void @nsw_step2(i4 %n) { +entry: + %s = icmp sgt i4 %n, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ 0, %entry ], [ %i.next, %loop ] + %i.next = add nsw i4 %i, 2 + %t = icmp slt i4 %i.next, %n + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @nsw_start1 +; CHECK: Loop %loop: backedge-taken count is (-2 + (2 smax %n)) +; CHECK: Loop %loop: max backedge-taken count is 5 +define void @nsw_start1(i4 %n) { +entry: + %s = icmp sgt i4 %n, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ 1, %entry ], [ %i.next, %loop ] + %i.next = add nsw i4 %i, 1 + %t = icmp slt i4 %i.next, %n + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @nsw_start1_step2 +; CHECK: Loop %loop: Unpredictable backedge-taken count. +; CHECK: Loop %loop: Unpredictable max backedge-taken count. +define void @nsw_start1_step2(i4 %n) { +entry: + %s = icmp sgt i4 %n, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ 1, %entry ], [ %i.next, %loop ] + %i.next = add nsw i4 %i, 2 + %t = icmp slt i4 %i.next, %n + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @nsw_startx +; CHECK: Loop %loop: backedge-taken count is (-1 + (-1 * %x) + ((1 + %x) smax %n)) +; CHECK: Loop %loop: max backedge-taken count is -1 +define void @nsw_startx(i4 %n, i4 %x) { +entry: + %s = icmp sgt i4 %n, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ %x, %entry ], [ %i.next, %loop ] + %i.next = add nsw i4 %i, 1 + %t = icmp slt i4 %i.next, %n + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @nsw_startx_step2 +; CHECK: Loop %loop: Unpredictable backedge-taken count. +; CHECK: Loop %loop: Unpredictable max backedge-taken count. +define void @nsw_startx_step2(i4 %n, i4 %x) { +entry: + %s = icmp sgt i4 %n, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ %x, %entry ], [ %i.next, %loop ] + %i.next = add nsw i4 %i, 2 + %t = icmp slt i4 %i.next, %n + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @even +; CHECK: Loop %loop: backedge-taken count is (-1 + (2 * %n)) +; CHECK: Loop %loop: max backedge-taken count is 5 +define void @even(i4 %n) { +entry: + %m = shl i4 %n, 1 + %s = icmp sgt i4 %m, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ 0, %entry ], [ %i.next, %loop ] + %i.next = add i4 %i, 1 + %t = icmp slt i4 %i.next, %m + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @even_step2 +; CHECK: Loop %loop: Unpredictable backedge-taken count. +; CHECK: Loop %loop: max backedge-taken count is 2 +define void @even_step2(i4 %n) { +entry: + %m = shl i4 %n, 1 + %s = icmp sgt i4 %m, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ 0, %entry ], [ %i.next, %loop ] + %i.next = add i4 %i, 2 + %t = icmp slt i4 %i.next, %m + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @even_start1 +; CHECK: Loop %loop: backedge-taken count is (-2 + (2 smax (2 * %n))) +; CHECK: Loop %loop: max backedge-taken count is 4 +define void @even_start1(i4 %n) { +entry: + %m = shl i4 %n, 1 + %s = icmp sgt i4 %m, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ 1, %entry ], [ %i.next, %loop ] + %i.next = add i4 %i, 1 + %t = icmp slt i4 %i.next, %m + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @even_start1_step2 +; CHECK: Loop %loop: Unpredictable backedge-taken count. +; CHECK: Loop %loop: max backedge-taken count is 2 +define void @even_start1_step2(i4 %n) { +entry: + %m = shl i4 %n, 1 + %s = icmp sgt i4 %m, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ 1, %entry ], [ %i.next, %loop ] + %i.next = add i4 %i, 2 + %t = icmp slt i4 %i.next, %m + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @even_startx +; CHECK: Loop %loop: backedge-taken count is (-1 + (-1 * %x) + ((1 + %x) smax (2 * %n))) +; CHECK: Loop %loop: max backedge-taken count is -1 +define void @even_startx(i4 %n, i4 %x) { +entry: + %m = shl i4 %n, 1 + %s = icmp sgt i4 %m, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ %x, %entry ], [ %i.next, %loop ] + %i.next = add i4 %i, 1 + %t = icmp slt i4 %i.next, %m + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @even_startx_step2 +; CHECK: Loop %loop: Unpredictable backedge-taken count. +; CHECK: Loop %loop: max backedge-taken count is 7 +define void @even_startx_step2(i4 %n, i4 %x) { +entry: + %m = shl i4 %n, 1 + %s = icmp sgt i4 %m, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ %x, %entry ], [ %i.next, %loop ] + %i.next = add i4 %i, 2 + %t = icmp slt i4 %i.next, %m + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @even_nsw +; CHECK: Loop %loop: backedge-taken count is (-1 + (2 * %n)) +; CHECK: Loop %loop: max backedge-taken count is 5 +define void @even_nsw(i4 %n) { +entry: + %m = shl i4 %n, 1 + %s = icmp sgt i4 %m, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ 0, %entry ], [ %i.next, %loop ] + %i.next = add nsw i4 %i, 1 + %t = icmp slt i4 %i.next, %m + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @even_nsw_step2 +; CHECK: Loop %loop: backedge-taken count is ((-1 + (2 * %n)) /u 2) +; CHECK: Loop %loop: max backedge-taken count is 2 +define void @even_nsw_step2(i4 %n) { +entry: + %m = shl i4 %n, 1 + %s = icmp sgt i4 %m, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ 0, %entry ], [ %i.next, %loop ] + %i.next = add nsw i4 %i, 2 + %t = icmp slt i4 %i.next, %m + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @even_nsw_start1 +; CHECK: Loop %loop: backedge-taken count is (-2 + (2 smax (2 * %n))) +; CHECK: Loop %loop: max backedge-taken count is 4 +define void @even_nsw_start1(i4 %n) { +entry: + %m = shl i4 %n, 1 + %s = icmp sgt i4 %m, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ 1, %entry ], [ %i.next, %loop ] + %i.next = add nsw i4 %i, 1 + %t = icmp slt i4 %i.next, %m + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @even_nsw_start1_step2 +; CHECK: Loop %loop: backedge-taken count is ((-2 + (3 smax (2 * %n))) /u 2) +; CHECK: Loop %loop: max backedge-taken count is 2 +define void @even_nsw_start1_step2(i4 %n) { +entry: + %m = shl i4 %n, 1 + %s = icmp sgt i4 %m, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ 1, %entry ], [ %i.next, %loop ] + %i.next = add nsw i4 %i, 2 + %t = icmp slt i4 %i.next, %m + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @even_nsw_startx +; CHECK: Loop %loop: backedge-taken count is (-1 + (-1 * %x) + ((1 + %x) smax (2 * %n))) +; CHECK: Loop %loop: max backedge-taken count is -1 +define void @even_nsw_startx(i4 %n, i4 %x) { +entry: + %m = shl i4 %n, 1 + %s = icmp sgt i4 %m, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ %x, %entry ], [ %i.next, %loop ] + %i.next = add nsw i4 %i, 1 + %t = icmp slt i4 %i.next, %m + br i1 %t, label %loop, label %exit +exit: + ret void +} + +; CHECK: Determining loop execution counts for: @even_nsw_startx_step2 +; CHECK: Loop %loop: backedge-taken count is ((-1 + (-1 * %x) + ((2 + %x) smax (2 * %n))) /u 2) +; CHECK: Loop %loop: max backedge-taken count is 7 +define void @even_nsw_startx_step2(i4 %n, i4 %x) { +entry: + %m = shl i4 %n, 1 + %s = icmp sgt i4 %m, 0 + br i1 %s, label %loop, label %exit +loop: + %i = phi i4 [ %x, %entry ], [ %i.next, %loop ] + %i.next = add nsw i4 %i, 2 + %t = icmp slt i4 %i.next, %m + br i1 %t, label %loop, label %exit +exit: + ret void +} diff --git a/test/Analysis/ScalarEvolution/xor-and.ll b/test/Analysis/ScalarEvolution/xor-and.ll new file mode 100644 index 0000000..c0530bb --- /dev/null +++ b/test/Analysis/ScalarEvolution/xor-and.ll @@ -0,0 +1,12 @@ +; RUN: opt < %s -scalar-evolution -analyze \ +; RUN: | grep {\\--> (zext i4 (-8 + (trunc i64 (8 \\* %x) to i4)) to i64)} + +; ScalarEvolution shouldn't try to analyze %z into something like +; --> (zext i4 (-1 + (-1 * (trunc i64 (8 * %x) to i4))) to i64) + +define i64 @foo(i64 %x) { + %a = shl i64 %x, 3 + %t = and i64 %a, 8 + %z = xor i64 %t, 8 + ret i64 %z +} diff --git a/test/Analysis/ScalarEvolution/zext-wrap.ll b/test/Analysis/ScalarEvolution/zext-wrap.ll new file mode 100644 index 0000000..38d15ff --- /dev/null +++ b/test/Analysis/ScalarEvolution/zext-wrap.ll @@ -0,0 +1,24 @@ +; RUN: opt < %s -analyze -scalar-evolution \ +; RUN: | FileCheck %s +; PR4569 + +define i16 @main() nounwind { +entry: + br label %bb.i + +bb.i: ; preds = %bb1.i, %bb.nph + %l_95.0.i1 = phi i8 [ %tmp1, %bb.i ], [ 0, %entry ] + +; This cast shouldn't be folded into the addrec. +; CHECK: %tmp = zext i8 %l_95.0.i1 to i16 +; CHECK: --> (zext i8 {0,+,-1}<%bb.i> to i16) Exits: 2 + + %tmp = zext i8 %l_95.0.i1 to i16 + + %tmp1 = add i8 %l_95.0.i1, -1 + %phitmp = icmp eq i8 %tmp1, 1 + br i1 %phitmp, label %bb1.i.func_36.exit_crit_edge, label %bb.i + +bb1.i.func_36.exit_crit_edge: + ret i16 %tmp +} |