From 785fad32025ac29b3169a3c9e25b01b2835869d4 Mon Sep 17 00:00:00 2001
From: Johannes Doerfert <jdoerfert@anl.gov>
Date: Fri, 23 Aug 2019 17:29:23 +0000
Subject: [PATCH] [Attributor] Deal with shrinking dereferenceability in a loop

Summary:
If we have a loop in which the dereferenceability of a pointer decreases
we did slowly decrease it iteration by iteration, leading to a timeout.
With this patch we detect such circular reasoning and indicate a
fixpoint early.

Reviewers: uenoku, sstefan1

Subscribers: hiraditya, bollu, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D66558

llvm-svn: 369784
---
 llvm/lib/Transforms/IPO/Attributor.cpp        | 23 ++++++++----
 .../FunctionAttrs/dereferenceable.ll          | 35 +++++++++++++++++--
 2 files changed, 50 insertions(+), 8 deletions(-)

diff --git a/llvm/lib/Transforms/IPO/Attributor.cpp b/llvm/lib/Transforms/IPO/Attributor.cpp
index e518994f1532..9367f482cc98 100644
--- a/llvm/lib/Transforms/IPO/Attributor.cpp
+++ b/llvm/lib/Transforms/IPO/Attributor.cpp
@@ -1988,16 +1988,27 @@ struct AADereferenceableFloating : AADereferenceableImpl {
       // For now we do not try to "increase" dereferenceability due to negative
       // indices as we first have to come up with code to deal with loops and
       // for overflows of the dereferenceable bytes.
-      if (Offset.getSExtValue() < 0)
+      int64_t OffsetSExt = Offset.getSExtValue();
+      if (OffsetSExt < 0)
         Offset = 0;
 
       T.takeAssumedDerefBytesMinimum(
-          std::max(int64_t(0), DerefBytes - Offset.getSExtValue()));
+          std::max(int64_t(0), DerefBytes - OffsetSExt));
 
-      if (!Stripped && this == &AA) {
-        T.takeKnownDerefBytesMaximum(
-            std::max(int64_t(0), DerefBytes - Offset.getSExtValue()));
-        T.indicatePessimisticFixpoint();
+      if (this == &AA) {
+        if (!Stripped) {
+          // If nothing was stripped IR information is all we got.
+          T.takeKnownDerefBytesMaximum(
+              std::max(int64_t(0), DerefBytes - OffsetSExt));
+          T.indicatePessimisticFixpoint();
+        } else if (OffsetSExt > 0) {
+          // If something was stripped but there is circular reasoning we look
+          // for the offset. If it is positive we basically decrease the
+          // dereferenceable bytes in a circluar loop now, which will simply
+          // drive them down to the known value in a very slow way which we
+          // can accelerate.
+          T.indicatePessimisticFixpoint();
+        }
       }
 
       return T.isValidState();
diff --git a/llvm/test/Transforms/FunctionAttrs/dereferenceable.ll b/llvm/test/Transforms/FunctionAttrs/dereferenceable.ll
index f61b62e9e239..3ed32d3002fc 100644
--- a/llvm/test/Transforms/FunctionAttrs/dereferenceable.ll
+++ b/llvm/test/Transforms/FunctionAttrs/dereferenceable.ll
@@ -1,6 +1,8 @@
 ; RUN: opt -attributor --attributor-disable=false -attributor-max-iterations=4 -S < %s | FileCheck %s --check-prefixes=ATTRIBUTOR
 
 
+declare void @deref_phi_user(i32* %a);
+
 ; TEST 1
 ; take mininimum of return values
 ;
@@ -52,8 +54,7 @@ define dereferenceable(4) i32* @test4(i32* dereferenceable(8) %0) local_unnamed_
 
 ; TEST 5
 ; loop in which dereferenceabily "grows"
-declare void @deref_phi_user(i32* %a);
-define void @deref_phi(i32* dereferenceable(4000) %a) {
+define void @deref_phi_growing(i32* dereferenceable(4000) %a) {
 entry:
   br label %for.cond
 
@@ -80,3 +81,33 @@ for.inc:                                          ; preds = %for.body
 for.end:                                          ; preds = %for.cond.cleanup
   ret void
 }
+
+; TEST 6
+; loop in which dereferenceabily "shrinks"
+define void @deref_phi_shrinking(i32* dereferenceable(4000) %a) {
+entry:
+  br label %for.cond
+
+for.cond:                                         ; preds = %for.inc, %entry
+  %i.0 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
+  %a.addr.0 = phi i32* [ %a, %entry ], [ %incdec.ptr, %for.inc ]
+; CHECK: call void @deref_phi_user(i32* %a.addr.0)
+  call void @deref_phi_user(i32* %a.addr.0)
+  %tmp = load i32, i32* %a.addr.0, align 4
+  %cmp = icmp slt i32 %i.0, %tmp
+  br i1 %cmp, label %for.body, label %for.cond.cleanup
+
+for.cond.cleanup:                                 ; preds = %for.cond
+  br label %for.end
+
+for.body:                                         ; preds = %for.cond
+  br label %for.inc
+
+for.inc:                                          ; preds = %for.body
+  %incdec.ptr = getelementptr inbounds i32, i32* %a.addr.0, i64 1
+  %inc = add nuw nsw i32 %i.0, 1
+  br label %for.cond
+
+for.end:                                          ; preds = %for.cond.cleanup
+  ret void
+}