When marking a block as being unanalyzable, use "Clobber" on the terminator instead of the first instruction in the block. This is a bit of a hack; "Clobber" isn't really the right marking in the first place. memdep doesn't really have any way of properly expressing "unanalyzable" at the moment. Using it on the terminator is much less ambiguous than using it on an arbitrary instruction, though.

In the given testcase, the "Clobber" was pointing to a load, and GVN was incorrectly assuming that meant that the "Clobber" load overlapped the load being analyzed (when they are actually unrelated).

The included testcase tests both this commit and r132434.

Part two of rdar://9429882.  (r132434 was mislabeled.)



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132442 91177308-0d34-0410-b5e6-96231b3b80d8

2 files changed, 63 insertions, 2 deletions

diff --git a/lib/Analysis/MemoryDependenceAnalysis.cpp b/lib/Analysis/MemoryDependenceAnalysis.cpp
index e854f7c068..a412425770 100644
--- a/lib/Analysis/MemoryDependenceAnalysis.cpp
+++ b/lib/Analysis/MemoryDependenceAnalysis.cpp
@@ -1155,8 +1155,8 @@ getNonLocalPointerDepFromBB(const PHITransAddr &Pointer,
       
       assert(I->getResult().isNonLocal() &&
              "Should only be here with transparent block");
-      I->setResult(MemDepResult::getClobber(BB->begin()));
-      ReverseNonLocalPtrDeps[BB->begin()].insert(CacheKey);
+      I->setResult(MemDepResult::getClobber(BB->getTerminator()));
+      ReverseNonLocalPtrDeps[BB->getTerminator()].insert(CacheKey);
       Result.push_back(NonLocalDepResult(I->getBB(), I->getResult(),
                                          Pointer.getAddr()));
       break;
diff --git a/test/Transforms/GVN/2011-06-01-NonLocalMemdepMiscompile.ll b/test/Transforms/GVN/2011-06-01-NonLocalMemdepMiscompile.ll
new file mode 100644
index 0000000000..f24e956a2b
--- /dev/null
+++ b/test/Transforms/GVN/2011-06-01-NonLocalMemdepMiscompile.ll
@@ -0,0 +1,61 @@
+; RUN: opt < %s -basicaa -gvn -S | FileCheck %s
+; This test is checking that (a) this doesn't crash, and (b) we don't
+; conclude the value of %tmp17 is available in bb1.bb15_crit_edge.
+; rdar://9429882
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
+target triple = "x86_64-apple-macosx10.7.0"
+
+define i1 @rb_intern() nounwind ssp {
+; CHECK: @rb_intern
+
+bb:
+  %tmp = alloca i8*, align 8
+  store i8* null, i8** %tmp, align 8
+  store i8 undef, i8* null, align 536870912
+  br label %bb1
+
+bb1:
+  br i1 undef, label %bb3, label %bb15
+
+; CHECK: bb1:
+; CHECK: %tmp16 = phi i8* [ getelementptr (i8* null, i64 undef), %bb10 ], [ null, %bb ]
+
+; CHECK: bb1.bb15_crit_edge:
+; CHECK: %tmp17.pre = load i8* %tmp16, align 1
+
+bb3:
+  call void @isalnum()
+  br i1 undef, label %bb10, label %bb5
+
+bb5:
+  br i1 undef, label %bb10, label %bb6
+
+bb6:
+  %tmp7 = load i8** %tmp, align 8
+  %tmp8 = load i8* %tmp7, align 1
+  %tmp9 = zext i8 %tmp8 to i64
+  br i1 undef, label %bb15, label %bb10
+
+bb10:
+  %tmp11 = load i8** %tmp, align 8
+  %tmp12 = load i8* %tmp11, align 1
+  %tmp13 = zext i8 %tmp12 to i64
+  %tmp14 = getelementptr inbounds i8* null, i64 undef
+  store i8* %tmp14, i8** %tmp, align 8
+  br label %bb1
+
+bb15:
+  %tmp16 = load i8** %tmp, align 8
+  %tmp17 = load i8* %tmp16, align 1
+  %tmp18 = icmp eq i8 %tmp17, 0
+  br label %bb19
+
+; CHECK: bb15:
+; CHECK: %tmp17 = phi i8 [ %tmp17.pre, %bb1.bb15_crit_edge ], [ %tmp8, %bb6 ]
+
+bb19:                                             ; preds = %bb15
+  ret i1 %tmp18
+}
+
+declare void @isalnum() nounwind inlinehint ssp