From a864748284f4bab974ffb13b840a611c1f9bc7ac Mon Sep 17 00:00:00 2001 From: Chandler Carruth Date: Fri, 2 Nov 2012 08:33:25 +0000 Subject: Revert the switch of loop-idiom to use the new dependence analysis. The new analysis is not yet ready for prime time. It has a *critical* flawed assumption, and some troubling shortages of testing. Until it's been hammered into better shape, let's stick with the working code. This should be easy to revert itself when the analysis is ready. Fixes PR14241, a miscompile of any memcpy-able loop which uses a pointer as the induction mechanism. If you have been seeing miscompiles in this revision range, you really want to test with this backed out. The results of this miscompile are a bit subtle as they can lead to downstream passes concluding things are impossible which are in fact possible. Thanks to David Blaikie for the majority of the reduction of this miscompile. I'll be checking in the test case in a non-revert commit. Revesions reverted here: r167045: LoopIdiom: Fix a serious missed optimization: we only turned top-level loops into memmove. r166877: LoopIdiom: Add checks to avoid turning memmove into an infinite loop. r166875: LoopIdiom: Recognize memmove loops. r166874: LoopIdiom: Replace custom dependence analysis with DependenceAnalysis. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167286 91177308-0d34-0410-b5e6-96231b3b80d8 --- test/Transforms/LoopIdiom/basic.ll | 22 ----------- test/Transforms/LoopIdiom/memmove.ll | 42 -------------------- test/Transforms/LoopIdiom/memset_noidiom.ll | 53 +------------------------- test/Transforms/LoopIdiom/multi-dimensional.ll | 49 ------------------------ test/Transforms/LoopIdiom/sideeffect.ll | 53 -------------------------- 5 files changed, 1 insertion(+), 218 deletions(-) delete mode 100644 test/Transforms/LoopIdiom/memmove.ll delete mode 100644 test/Transforms/LoopIdiom/multi-dimensional.ll delete mode 100644 test/Transforms/LoopIdiom/sideeffect.ll (limited to 'test') diff --git a/test/Transforms/LoopIdiom/basic.ll b/test/Transforms/LoopIdiom/basic.ll index 5afc405fa6..46ab7e5542 100644 --- a/test/Transforms/LoopIdiom/basic.ll +++ b/test/Transforms/LoopIdiom/basic.ll @@ -383,26 +383,4 @@ for.end: ; preds = %for.inc } -@p = common global [1024 x i8] zeroinitializer, align 16 -define void @test15(i32 %n) nounwind { -entry: - %cmp6 = icmp eq i32 %n, 0 - br i1 %cmp6, label %for.end, label %for.body - -for.body: ; preds = %entry, %for.body - %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ] - %indvars.iv.next = add i64 %indvars.iv, 1 - %arrayidx = getelementptr inbounds [1024 x i8]* @p, i64 0, i64 %indvars.iv.next - %0 = load i8* %arrayidx, align 1 - %arrayidx2 = getelementptr inbounds [1024 x i8]* @p, i64 0, i64 %indvars.iv - store i8 %0, i8* %arrayidx2, align 1 - %lftr.wideiv = trunc i64 %indvars.iv.next to i32 - %exitcond = icmp eq i32 %lftr.wideiv, %n - br i1 %exitcond, label %for.end, label %for.body - -for.end: ; preds = %for.body, %entry - ret void -; CHECK: @test15 -; CHECK: call void @llvm.memmove.p0i8.p0i8.i64(i8* getelementptr inbounds ([1024 x i8]* @p, i32 0, i32 0), i8* getelementptr inbounds ([1024 x i8]* @p, i64 0, i64 1), -} diff --git a/test/Transforms/LoopIdiom/memmove.ll b/test/Transforms/LoopIdiom/memmove.ll deleted file mode 100644 index bfa0902a0d..0000000000 --- a/test/Transforms/LoopIdiom/memmove.ll +++ /dev/null @@ -1,42 +0,0 @@ -; RUN: opt -S -basicaa -loop-idiom < %s | 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-n8:16:32:64-S128" -target triple = "x86_64-apple-macosx10.8.0" - -declare i64 @foo() nounwind - -; Nested loops -define void @test1(i8* nocapture %A, i64 %n) nounwind { -entry: - %call8 = tail call i64 @foo() nounwind - %tobool9 = icmp eq i64 %call8, 0 - br i1 %tobool9, label %while.end, label %for.cond.preheader.lr.ph - -for.cond.preheader.lr.ph: ; preds = %entry - %cmp6 = icmp eq i64 %n, 0 - br label %for.cond.preheader - -while.cond.loopexit: ; preds = %for.body, %for.cond.preheader - %call = tail call i64 @foo() nounwind - %tobool = icmp eq i64 %call, 0 - br i1 %tobool, label %while.end, label %for.cond.preheader - -for.cond.preheader: ; preds = %for.cond.preheader.lr.ph, %while.cond.loopexit - br i1 %cmp6, label %while.cond.loopexit, label %for.body - -for.body: ; preds = %for.cond.preheader, %for.body - %i.07 = phi i64 [ %inc, %for.body ], [ 0, %for.cond.preheader ] - %add = add i64 %i.07, 10 - %arrayidx = getelementptr inbounds i8* %A, i64 %add - %0 = load i8* %arrayidx, align 1 - %arrayidx1 = getelementptr inbounds i8* %A, i64 %i.07 - store i8 %0, i8* %arrayidx1, align 1 - %inc = add i64 %i.07, 1 - %exitcond = icmp eq i64 %inc, %n - br i1 %exitcond, label %while.cond.loopexit, label %for.body - -while.end: ; preds = %while.cond.loopexit, %entry - ret void - -; CHECK: @test1 -; CHECK: call void @llvm.memmove.p0i8.p0i8.i64( -} diff --git a/test/Transforms/LoopIdiom/memset_noidiom.ll b/test/Transforms/LoopIdiom/memset_noidiom.ll index 27ef4f639c..168eb95357 100644 --- a/test/Transforms/LoopIdiom/memset_noidiom.ll +++ b/test/Transforms/LoopIdiom/memset_noidiom.ll @@ -1,4 +1,4 @@ -; RUN: opt -basicaa -loop-idiom < %s -S | FileCheck %s +; RUN: opt -loop-idiom < %s -S | 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-n8:16:32:64" target triple = "x86_64-apple-darwin10.0.0" @@ -28,54 +28,3 @@ for.end: ; preds = %for.cond.for.end_cr ret i8* %b } -; CHECK: @memcpy -; CHECK-NOT: llvm.memcpy -define i8* @memcpy(i8* noalias %dst, i8* noalias %src, i64 %n) nounwind { -entry: - %tobool3 = icmp eq i64 %n, 0 - br i1 %tobool3, label %while.end, label %while.body - -while.body: ; preds = %entry, %while.body - %c2.06 = phi i8* [ %incdec.ptr, %while.body ], [ %src, %entry ] - %c1.05 = phi i8* [ %incdec.ptr1, %while.body ], [ %dst, %entry ] - %n.addr.04 = phi i64 [ %dec, %while.body ], [ %n, %entry ] - %dec = add i64 %n.addr.04, -1 - %incdec.ptr = getelementptr inbounds i8* %c2.06, i64 1 - %0 = load i8* %c2.06, align 1 - %incdec.ptr1 = getelementptr inbounds i8* %c1.05, i64 1 - store i8 %0, i8* %c1.05, align 1 - %tobool = icmp eq i64 %dec, 0 - br i1 %tobool, label %while.end, label %while.body - -while.end: ; preds = %while.body, %entry - ret i8* %dst -} - -; CHECK: @memmove -; CHECK-NOT: llvm.memmove -define i8* @memmove(i8* %dst, i8* nocapture %src, i64 %count) nounwind { -entry: - %sub = add i64 %count, -1 - %tobool9 = icmp eq i64 %count, 0 - br i1 %tobool9, label %while.end, label %while.body.lr.ph - -while.body.lr.ph: ; preds = %entry - %add.ptr2 = getelementptr inbounds i8* %src, i64 %sub - %add.ptr = getelementptr inbounds i8* %dst, i64 %sub - br label %while.body - -while.body: ; preds = %while.body.lr.ph, %while.body - %b.012 = phi i8* [ %add.ptr2, %while.body.lr.ph ], [ %incdec.ptr, %while.body ] - %a.011 = phi i8* [ %add.ptr, %while.body.lr.ph ], [ %incdec.ptr3, %while.body ] - %count.addr.010 = phi i64 [ %count, %while.body.lr.ph ], [ %dec, %while.body ] - %dec = add i64 %count.addr.010, -1 - %incdec.ptr = getelementptr inbounds i8* %b.012, i64 -1 - %0 = load i8* %b.012, align 1 - %incdec.ptr3 = getelementptr inbounds i8* %a.011, i64 -1 - store i8 %0, i8* %a.011, align 1 - %tobool = icmp eq i64 %dec, 0 - br i1 %tobool, label %while.end, label %while.body - -while.end: ; preds = %while.body, %entry - ret i8* %dst -} diff --git a/test/Transforms/LoopIdiom/multi-dimensional.ll b/test/Transforms/LoopIdiom/multi-dimensional.ll deleted file mode 100644 index 991f2688d7..0000000000 --- a/test/Transforms/LoopIdiom/multi-dimensional.ll +++ /dev/null @@ -1,49 +0,0 @@ -; RUN: opt -basicaa -loop-idiom < %s -S | 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-n8:16:32:64" -target triple = "x86_64-apple-darwin10.0.0" - -%struct.ham = type { [2 x [2 x [2 x [16 x [8 x i32]]]]], i32, %struct.zot } -%struct.zot = type { i32, i16, i16, [2 x [1152 x i32]] } - -define void @test1(%struct.ham* nocapture %arg) nounwind { -bb: - br label %bb1 - -bb1: ; preds = %bb11, %bb - %tmp = phi i64 [ 0, %bb ], [ %tmp12, %bb11 ] - br label %bb2 - -bb2: ; preds = %bb2, %bb1 - %tmp3 = phi i64 [ 0, %bb1 ], [ %tmp8, %bb2 ] - %tmp4 = getelementptr inbounds %struct.ham* %arg, i64 0, i32 0, i64 0, i64 1, i64 1, i64 %tmp, i64 %tmp3 - store i32 0, i32* %tmp4, align 4 - %tmp5 = getelementptr inbounds %struct.ham* %arg, i64 0, i32 0, i64 0, i64 1, i64 0, i64 %tmp, i64 %tmp3 - store i32 0, i32* %tmp5, align 4 - %tmp6 = getelementptr inbounds %struct.ham* %arg, i64 0, i32 0, i64 0, i64 0, i64 1, i64 %tmp, i64 %tmp3 - store i32 0, i32* %tmp6, align 4 - %tmp7 = getelementptr inbounds %struct.ham* %arg, i64 0, i32 0, i64 0, i64 0, i64 0, i64 %tmp, i64 %tmp3 - store i32 0, i32* %tmp7, align 4 - %tmp8 = add i64 %tmp3, 1 - %tmp9 = trunc i64 %tmp8 to i32 - %tmp10 = icmp eq i32 %tmp9, 8 - br i1 %tmp10, label %bb11, label %bb2 - -bb11: ; preds = %bb2 - %tmp12 = add i64 %tmp, 1 - %tmp13 = trunc i64 %tmp12 to i32 - %tmp14 = icmp eq i32 %tmp13, 16 - br i1 %tmp14, label %bb15, label %bb1 - -bb15: ; preds = %bb11 - ret void - -; CHECK: @test1 -; CHECK: bb1: -; CHECK-NOT: store -; CHECK: call void @llvm.memset.p0i8.i64 -; CHECK-NEXT: call void @llvm.memset.p0i8.i64 -; CHECK-NEXT: call void @llvm.memset.p0i8.i64 -; CHECK-NEXT: call void @llvm.memset.p0i8.i64 -; CHECK-NOT: store -; CHECK: br -} diff --git a/test/Transforms/LoopIdiom/sideeffect.ll b/test/Transforms/LoopIdiom/sideeffect.ll deleted file mode 100644 index 460e5233f9..0000000000 --- a/test/Transforms/LoopIdiom/sideeffect.ll +++ /dev/null @@ -1,53 +0,0 @@ -; RUN: opt -basicaa -loop-idiom < %s -S | 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-n8:16:32:64" -target triple = "x86_64-apple-darwin10.0.0" - -; PR9481 -define i32 @test1() nounwind uwtable ssp { -entry: - %a = alloca [10 x i8], align 1 - br label %for.body - -for.cond1.preheader: ; preds = %for.body - %arrayidx5.phi.trans.insert = getelementptr inbounds [10 x i8]* %a, i64 0, i64 0 - %.pre = load i8* %arrayidx5.phi.trans.insert, align 1 - br label %for.body3 - -for.body: ; preds = %for.body, %entry - %indvars.iv29 = phi i64 [ 0, %entry ], [ %indvars.iv.next30, %for.body ] - call void (...)* @bar() nounwind - %arrayidx = getelementptr inbounds [10 x i8]* %a, i64 0, i64 %indvars.iv29 - store i8 23, i8* %arrayidx, align 1 - %indvars.iv.next30 = add i64 %indvars.iv29, 1 - %lftr.wideiv31 = trunc i64 %indvars.iv.next30 to i32 - %exitcond32 = icmp eq i32 %lftr.wideiv31, 1000000 - br i1 %exitcond32, label %for.cond1.preheader, label %for.body - -for.body3: ; preds = %for.body3, %for.cond1.preheader - %0 = phi i8 [ %.pre, %for.cond1.preheader ], [ %add, %for.body3 ] - %indvars.iv = phi i64 [ 1, %for.cond1.preheader ], [ %indvars.iv.next, %for.body3 ] - call void (...)* @bar() nounwind - %arrayidx7 = getelementptr inbounds [10 x i8]* %a, i64 0, i64 %indvars.iv - %1 = load i8* %arrayidx7, align 1 - %add = add i8 %1, %0 - store i8 %add, i8* %arrayidx7, align 1 - %indvars.iv.next = add i64 %indvars.iv, 1 - %lftr.wideiv = trunc i64 %indvars.iv.next to i32 - %exitcond = icmp eq i32 %lftr.wideiv, 1000000 - br i1 %exitcond, label %for.end12, label %for.body3 - -for.end12: ; preds = %for.body3 - %arrayidx13 = getelementptr inbounds [10 x i8]* %a, i64 0, i64 2 - %2 = load i8* %arrayidx13, align 1 - %conv14 = sext i8 %2 to i32 - %arrayidx15 = getelementptr inbounds [10 x i8]* %a, i64 0, i64 6 - %3 = load i8* %arrayidx15, align 1 - %conv16 = sext i8 %3 to i32 - %add17 = add nsw i32 %conv16, %conv14 - ret i32 %add17 - -; CHECK: @test1 -; CHECK-NOT: @llvm.memset -} - -declare void @bar(...) -- cgit v1.2.3