From e4ba75f43e2ab1480d119d2d4eb878256274e0fb Mon Sep 17 00:00:00 2001 From: Chandler Carruth Date: Mon, 7 Jan 2013 14:41:08 +0000 Subject: Switch the SCEV expander and LoopStrengthReduce to use TargetTransformInfo rather than TargetLowering, removing one of the primary instances of the layering violation of Transforms depending directly on Target. This is a really big deal because LSR used to be a "special" pass that could only be tested fully using llc and by looking at the full output of it. It also couldn't run with any other loop passes because it had to be created by the backend. No longer is this true. LSR is now just a normal pass and we should probably lift the creation of LSR out of lib/CodeGen/Passes.cpp and into the PassManagerBuilder. =] I've not done this, or updated all of the tests to use opt and a triple, because I suspect someone more familiar with LSR would do a better job. This change should be essentially without functional impact for normal compilations, and only change behvaior of targetless compilations. The conversion required changing all of the LSR code to refer to the TTI interfaces, which fortunately are very similar to TargetLowering's interfaces. However, it also allowed us to *always* expect to have some implementation around. I've pushed that simplification through the pass, and leveraged it to simplify code somewhat. It required some test updates for one of two things: either we used to skip some checks altogether but now we get the default "no" answer for them, or we used to have no information about the target and now we do have some. I've also started the process of removing AddrMode, as the TTI interface doesn't use it any longer. In some cases this simplifies code, and in others it adds some complexity, but I think it's not a bad tradeoff even there. Subsequent patches will try to clean this up even further and use other (more appropriate) abstractions. Yet again, almost all of the formatting changes brought to you by clang-format. =] git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@171735 91177308-0d34-0410-b5e6-96231b3b80d8 --- .../LoopStrengthReduce/2008-08-14-ShadowIV.ll | 99 ---------------------- .../LoopStrengthReduce/2011-07-20-DoubleIV.ll | 43 ---------- .../2012-07-18-LimitReassociate.ll | 13 +-- .../LoopStrengthReduce/X86/2008-08-14-ShadowIV.ll | 99 ++++++++++++++++++++++ .../LoopStrengthReduce/X86/2011-07-20-DoubleIV.ll | 43 ++++++++++ .../LoopStrengthReduce/post-inc-icmpzero.ll | 8 +- 6 files changed, 153 insertions(+), 152 deletions(-) delete mode 100644 test/Transforms/LoopStrengthReduce/2008-08-14-ShadowIV.ll delete mode 100644 test/Transforms/LoopStrengthReduce/2011-07-20-DoubleIV.ll create mode 100644 test/Transforms/LoopStrengthReduce/X86/2008-08-14-ShadowIV.ll create mode 100644 test/Transforms/LoopStrengthReduce/X86/2011-07-20-DoubleIV.ll (limited to 'test/Transforms/LoopStrengthReduce') diff --git a/test/Transforms/LoopStrengthReduce/2008-08-14-ShadowIV.ll b/test/Transforms/LoopStrengthReduce/2008-08-14-ShadowIV.ll deleted file mode 100644 index c650d8cf76..0000000000 --- a/test/Transforms/LoopStrengthReduce/2008-08-14-ShadowIV.ll +++ /dev/null @@ -1,99 +0,0 @@ -; RUN: opt < %s -loop-reduce -S | grep "phi double" | count 1 - -define void @foobar(i32 %n) nounwind { -entry: - icmp eq i32 %n, 0 ; :0 [#uses=2] - br i1 %0, label %return, label %bb.nph - -bb.nph: ; preds = %entry - %umax = select i1 %0, i32 1, i32 %n ; [#uses=1] - br label %bb - -bb: ; preds = %bb, %bb.nph - %i.03 = phi i32 [ 0, %bb.nph ], [ %indvar.next, %bb ] ; [#uses=3] - tail call void @bar( i32 %i.03 ) nounwind - uitofp i32 %i.03 to double ; :1 [#uses=1] - tail call void @foo( double %1 ) nounwind - %indvar.next = add i32 %i.03, 1 ; [#uses=2] - %exitcond = icmp eq i32 %indvar.next, %umax ; [#uses=1] - br i1 %exitcond, label %return, label %bb - -return: ; preds = %bb, %entry - ret void -} - -; Unable to eliminate cast because the mantissa bits for double are not enough -; to hold all of i64 IV bits. -define void @foobar2(i64 %n) nounwind { -entry: - icmp eq i64 %n, 0 ; :0 [#uses=2] - br i1 %0, label %return, label %bb.nph - -bb.nph: ; preds = %entry - %umax = select i1 %0, i64 1, i64 %n ; [#uses=1] - br label %bb - -bb: ; preds = %bb, %bb.nph - %i.03 = phi i64 [ 0, %bb.nph ], [ %indvar.next, %bb ] ; [#uses=3] - trunc i64 %i.03 to i32 ; :1 [#uses=1] - tail call void @bar( i32 %1 ) nounwind - uitofp i64 %i.03 to double ; :2 [#uses=1] - tail call void @foo( double %2 ) nounwind - %indvar.next = add i64 %i.03, 1 ; [#uses=2] - %exitcond = icmp eq i64 %indvar.next, %umax ; [#uses=1] - br i1 %exitcond, label %return, label %bb - -return: ; preds = %bb, %entry - ret void -} - -; Unable to eliminate cast due to potentional overflow. -define void @foobar3() nounwind { -entry: - tail call i32 (...)* @nn( ) nounwind ; :0 [#uses=1] - icmp eq i32 %0, 0 ; :1 [#uses=1] - br i1 %1, label %return, label %bb - -bb: ; preds = %bb, %entry - %i.03 = phi i32 [ 0, %entry ], [ %3, %bb ] ; [#uses=3] - tail call void @bar( i32 %i.03 ) nounwind - uitofp i32 %i.03 to double ; :2 [#uses=1] - tail call void @foo( double %2 ) nounwind - add i32 %i.03, 1 ; :3 [#uses=2] - tail call i32 (...)* @nn( ) nounwind ; :4 [#uses=1] - icmp ugt i32 %4, %3 ; :5 [#uses=1] - br i1 %5, label %bb, label %return - -return: ; preds = %bb, %entry - ret void -} - -; Unable to eliminate cast due to overflow. -define void @foobar4() nounwind { -entry: - br label %bb.nph - -bb.nph: ; preds = %entry - br label %bb - -bb: ; preds = %bb, %bb.nph - %i.03 = phi i8 [ 0, %bb.nph ], [ %indvar.next, %bb ] ; [#uses=3] - %tmp2 = sext i8 %i.03 to i32 ; :0 [#uses=1] - tail call void @bar( i32 %tmp2 ) nounwind - %tmp3 = uitofp i8 %i.03 to double ; :1 [#uses=1] - tail call void @foo( double %tmp3 ) nounwind - %indvar.next = add i8 %i.03, 1 ; [#uses=2] - %tmp = sext i8 %indvar.next to i32 - %exitcond = icmp eq i32 %tmp, 32767 ; [#uses=1] - br i1 %exitcond, label %return, label %bb - -return: ; preds = %bb, %entry - ret void -} - -declare void @bar(i32) - -declare void @foo(double) - -declare i32 @nn(...) - diff --git a/test/Transforms/LoopStrengthReduce/2011-07-20-DoubleIV.ll b/test/Transforms/LoopStrengthReduce/2011-07-20-DoubleIV.ll deleted file mode 100644 index 5d9ed64ef4..0000000000 --- a/test/Transforms/LoopStrengthReduce/2011-07-20-DoubleIV.ll +++ /dev/null @@ -1,43 +0,0 @@ -; RUN: opt < %s -loop-reduce -S | FileCheck %s -; -; Test LSR's OptimizeShadowIV. Handle a floating-point IV with a -; nonzero initial value. -; rdar://9786536 - -; First, make sure LSR doesn't crash on an empty IVUsers list. -; CHECK: @dummyIV -; CHECK-NOT: phi -; CHECK-NOT: sitofp -; CHECK: br -define void @dummyIV() nounwind { -entry: - br label %loop - -loop: - %i.01 = phi i32 [ -39, %entry ], [ %inc, %loop ] - %conv = sitofp i32 %i.01 to double - %inc = add nsw i32 %i.01, 1 - br i1 undef, label %loop, label %for.end - -for.end: - unreachable -} - -; Now check that the computed double constant is correct. -; CHECK: @doubleIV -; CHECK: phi double [ -3.900000e+01, %entry ] -; CHECK: br -define void @doubleIV() nounwind { -entry: - br label %loop - -loop: - %i.01 = phi i32 [ -39, %entry ], [ %inc, %loop ] - %conv = sitofp i32 %i.01 to double - %div = fdiv double %conv, 4.000000e+01 - %inc = add nsw i32 %i.01, 1 - br i1 undef, label %loop, label %for.end - -for.end: - unreachable -} diff --git a/test/Transforms/LoopStrengthReduce/2012-07-18-LimitReassociate.ll b/test/Transforms/LoopStrengthReduce/2012-07-18-LimitReassociate.ll index 0aa2787620..53da462716 100644 --- a/test/Transforms/LoopStrengthReduce/2012-07-18-LimitReassociate.ll +++ b/test/Transforms/LoopStrengthReduce/2012-07-18-LimitReassociate.ll @@ -5,16 +5,17 @@ ; PR13361: LSR + SCEV "hangs" on reasonably sized test with sequence of loops ; ; Without limits on CollectSubexpr, we have thousands of formulae for -; the use that crosses loops. With limits we have five. +; the use that crosses loops. With limits we have six. ; CHECK: LSR on loop %bb221: ; CHECK: After generating reuse formulae: ; CHECK: LSR is examining the following uses: ; CHECK: LSR Use: Kind=Special -; CHECK: {{.*reg\(\{\{\{\{\{\{\{\{\{}} -; CHECK: {{.*reg\(\{\{\{\{\{\{\{\{\{}} -; CHECK: {{.*reg\(\{\{\{\{\{\{\{\{\{}} -; CHECK: {{.*reg\(\{\{\{\{\{\{\{\{\{}} -; CHECK: {{.*reg\(\{\{\{\{\{\{\{\{\{}} +; CHECK: {{.*reg\(\{.*\{.*\{.*\{.*\{.*\{.*\{.*\{.*\{}} +; CHECK: {{.*reg\(\{.*\{.*\{.*\{.*\{.*\{.*\{.*\{.*\{}} +; CHECK: {{.*reg\(\{.*\{.*\{.*\{.*\{.*\{.*\{.*\{.*\{}} +; CHECK: {{.*reg\(\{.*\{.*\{.*\{.*\{.*\{.*\{.*\{.*\{}} +; CHECK: {{.*reg\(\{.*\{.*\{.*\{.*\{.*\{.*\{.*\{.*\{}} +; CHECK: {{.*reg\(\{.*\{.*\{.*\{.*\{.*\{.*\{.*\{.*\{}} ; CHECK-NOT:reg ; CHECK: Filtering for use 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" diff --git a/test/Transforms/LoopStrengthReduce/X86/2008-08-14-ShadowIV.ll b/test/Transforms/LoopStrengthReduce/X86/2008-08-14-ShadowIV.ll new file mode 100644 index 0000000000..9a7f4865c5 --- /dev/null +++ b/test/Transforms/LoopStrengthReduce/X86/2008-08-14-ShadowIV.ll @@ -0,0 +1,99 @@ +; RUN: opt < %s -loop-reduce -S -mtriple=x86_64-unknown-unknown | grep "phi double" | count 1 + +define void @foobar(i32 %n) nounwind { +entry: + icmp eq i32 %n, 0 ; :0 [#uses=2] + br i1 %0, label %return, label %bb.nph + +bb.nph: ; preds = %entry + %umax = select i1 %0, i32 1, i32 %n ; [#uses=1] + br label %bb + +bb: ; preds = %bb, %bb.nph + %i.03 = phi i32 [ 0, %bb.nph ], [ %indvar.next, %bb ] ; [#uses=3] + tail call void @bar( i32 %i.03 ) nounwind + uitofp i32 %i.03 to double ; :1 [#uses=1] + tail call void @foo( double %1 ) nounwind + %indvar.next = add i32 %i.03, 1 ; [#uses=2] + %exitcond = icmp eq i32 %indvar.next, %umax ; [#uses=1] + br i1 %exitcond, label %return, label %bb + +return: ; preds = %bb, %entry + ret void +} + +; Unable to eliminate cast because the mantissa bits for double are not enough +; to hold all of i64 IV bits. +define void @foobar2(i64 %n) nounwind { +entry: + icmp eq i64 %n, 0 ; :0 [#uses=2] + br i1 %0, label %return, label %bb.nph + +bb.nph: ; preds = %entry + %umax = select i1 %0, i64 1, i64 %n ; [#uses=1] + br label %bb + +bb: ; preds = %bb, %bb.nph + %i.03 = phi i64 [ 0, %bb.nph ], [ %indvar.next, %bb ] ; [#uses=3] + trunc i64 %i.03 to i32 ; :1 [#uses=1] + tail call void @bar( i32 %1 ) nounwind + uitofp i64 %i.03 to double ; :2 [#uses=1] + tail call void @foo( double %2 ) nounwind + %indvar.next = add i64 %i.03, 1 ; [#uses=2] + %exitcond = icmp eq i64 %indvar.next, %umax ; [#uses=1] + br i1 %exitcond, label %return, label %bb + +return: ; preds = %bb, %entry + ret void +} + +; Unable to eliminate cast due to potentional overflow. +define void @foobar3() nounwind { +entry: + tail call i32 (...)* @nn( ) nounwind ; :0 [#uses=1] + icmp eq i32 %0, 0 ; :1 [#uses=1] + br i1 %1, label %return, label %bb + +bb: ; preds = %bb, %entry + %i.03 = phi i32 [ 0, %entry ], [ %3, %bb ] ; [#uses=3] + tail call void @bar( i32 %i.03 ) nounwind + uitofp i32 %i.03 to double ; :2 [#uses=1] + tail call void @foo( double %2 ) nounwind + add i32 %i.03, 1 ; :3 [#uses=2] + tail call i32 (...)* @nn( ) nounwind ; :4 [#uses=1] + icmp ugt i32 %4, %3 ; :5 [#uses=1] + br i1 %5, label %bb, label %return + +return: ; preds = %bb, %entry + ret void +} + +; Unable to eliminate cast due to overflow. +define void @foobar4() nounwind { +entry: + br label %bb.nph + +bb.nph: ; preds = %entry + br label %bb + +bb: ; preds = %bb, %bb.nph + %i.03 = phi i8 [ 0, %bb.nph ], [ %indvar.next, %bb ] ; [#uses=3] + %tmp2 = sext i8 %i.03 to i32 ; :0 [#uses=1] + tail call void @bar( i32 %tmp2 ) nounwind + %tmp3 = uitofp i8 %i.03 to double ; :1 [#uses=1] + tail call void @foo( double %tmp3 ) nounwind + %indvar.next = add i8 %i.03, 1 ; [#uses=2] + %tmp = sext i8 %indvar.next to i32 + %exitcond = icmp eq i32 %tmp, 32767 ; [#uses=1] + br i1 %exitcond, label %return, label %bb + +return: ; preds = %bb, %entry + ret void +} + +declare void @bar(i32) + +declare void @foo(double) + +declare i32 @nn(...) + diff --git a/test/Transforms/LoopStrengthReduce/X86/2011-07-20-DoubleIV.ll b/test/Transforms/LoopStrengthReduce/X86/2011-07-20-DoubleIV.ll new file mode 100644 index 0000000000..a932b47925 --- /dev/null +++ b/test/Transforms/LoopStrengthReduce/X86/2011-07-20-DoubleIV.ll @@ -0,0 +1,43 @@ +; RUN: opt < %s -loop-reduce -S -mtriple=x86_64-unknown-unknown | FileCheck %s +; +; Test LSR's OptimizeShadowIV. Handle a floating-point IV with a +; nonzero initial value. +; rdar://9786536 + +; First, make sure LSR doesn't crash on an empty IVUsers list. +; CHECK: @dummyIV +; CHECK-NOT: phi +; CHECK-NOT: sitofp +; CHECK: br +define void @dummyIV() nounwind { +entry: + br label %loop + +loop: + %i.01 = phi i32 [ -39, %entry ], [ %inc, %loop ] + %conv = sitofp i32 %i.01 to double + %inc = add nsw i32 %i.01, 1 + br i1 undef, label %loop, label %for.end + +for.end: + unreachable +} + +; Now check that the computed double constant is correct. +; CHECK: @doubleIV +; CHECK: phi double [ -3.900000e+01, %entry ] +; CHECK: br +define void @doubleIV() nounwind { +entry: + br label %loop + +loop: + %i.01 = phi i32 [ -39, %entry ], [ %inc, %loop ] + %conv = sitofp i32 %i.01 to double + %div = fdiv double %conv, 4.000000e+01 + %inc = add nsw i32 %i.01, 1 + br i1 undef, label %loop, label %for.end + +for.end: + unreachable +} diff --git a/test/Transforms/LoopStrengthReduce/post-inc-icmpzero.ll b/test/Transforms/LoopStrengthReduce/post-inc-icmpzero.ll index 96904c66e6..9e02d92a6f 100644 --- a/test/Transforms/LoopStrengthReduce/post-inc-icmpzero.ll +++ b/test/Transforms/LoopStrengthReduce/post-inc-icmpzero.ll @@ -4,12 +4,12 @@ ; LSR should properly handle the post-inc offset when folding the ; non-IV operand of an icmp into the IV. -; CHECK: %4 = sub i64 %sub.ptr.lhs.cast, %sub.ptr.rhs.cast -; CHECK: %5 = lshr i64 %4, 1 -; CHECK: %6 = mul i64 %5, 2 +; CHECK: %3 = sub i64 %sub.ptr.lhs.cast, %sub.ptr.rhs.cast +; CHECK: %4 = lshr i64 %3, 1 +; CHECK: %5 = mul i64 %4, 2 ; CHECK: br label %for.body ; CHECK: for.body: -; CHECK: %lsr.iv2 = phi i64 [ %lsr.iv.next, %for.body ], [ %6, %for.body.lr.ph ] +; CHECK: %lsr.iv2 = phi i64 [ %lsr.iv.next, %for.body ], [ %5, %for.body.lr.ph ] ; CHECK: %lsr.iv.next = add i64 %lsr.iv2, -2 ; CHECK: %lsr.iv.next3 = inttoptr i64 %lsr.iv.next to i16* ; CHECK: %cmp27 = icmp eq i16* %lsr.iv.next3, null -- cgit v1.2.3