ValueTracking: ComputeNumSignBits should tolerate misshapen phi nodes

PHI nodes can have zero operands in the middle of a transform.  It is
expected that utilities in Analysis don't freak out when this happens.

Note that it is considered invalid to allow these misshapen phi nodes to
make it to another pass.

This fixes PR22086.

llvm-svn: 225126

1 files changed, 28 insertions, 0 deletions

diff --git a/llvm/test/Transforms/JumpThreading/pr22086.ll b/llvm/test/Transforms/JumpThreading/pr22086.ll
new file mode 100644
index 00000000000..35d9aa5b184
--- /dev/null
+++ b/llvm/test/Transforms/JumpThreading/pr22086.ll
@@ -0,0 +1,28 @@
+; RUN: opt -S -jump-threading < %s | FileCheck %s
+
+
+; CHECK-LABEL: @f(
+; CHECK-LABEL: entry:
+; CHECK-NEXT:  br label %[[loop:.*]]
+; CHECK:       [[loop]]:
+; CHECK-NEXT:  br label %[[loop]]
+
+define void @f() {
+entry:
+  br label %for.cond1
+
+if.end16:
+  %phi1 = phi i32 [ undef, %for.cond1 ]
+  %g.3 = phi i32 [ %g.1, %for.cond1 ]
+  %sext = shl i32 %g.3, 16
+  %conv20 = ashr exact i32 %sext, 16
+  %tobool21 = icmp eq i32 %phi1, 0
+  br i1 %tobool21, label %lor.rhs, label %for.cond1
+
+for.cond1:
+  %g.1 = phi i32 [ 0, %entry ], [ 0, %lor.rhs ], [ %g.3, %if.end16 ]
+  br i1 undef, label %lor.rhs, label %if.end16
+
+lor.rhs:
+  br label %for.cond1
+}