Give the Reassociate pass a bit more flexibility and autonomy when optimizing expressions.

Particularly, it addresses cases where Reassociate breaks Subtracts but then fails to optimize combinations like I1 + -I2 where I1 and I2 have the same rank and are identical.

Patch by Dmitri Shtilman.

llvm-svn: 219092

2 files changed, 27 insertions, 2 deletions

diff --git a/llvm/lib/Transforms/Scalar/Reassociate.cpp b/llvm/lib/Transforms/Scalar/Reassociate.cpp
index d0cd48a3a26..b8fb25d0b9a 100644
--- a/llvm/lib/Transforms/Scalar/Reassociate.cpp
+++ b/llvm/lib/Transforms/Scalar/Reassociate.cpp
@@ -1034,13 +1034,23 @@ static unsigned FindInOperandList(SmallVectorImpl<ValueEntry> &Ops, unsigned i,
                                   Value *X) {
   unsigned XRank = Ops[i].Rank;
   unsigned e = Ops.size();
-  for (unsigned j = i+1; j != e && Ops[j].Rank == XRank; ++j)
+  for (unsigned j = i+1; j != e && Ops[j].Rank == XRank; ++j) {
     if (Ops[j].Op == X)
       return j;
+    if (Instruction *I1 = dyn_cast<Instruction>(Ops[j].Op))
+      if (Instruction *I2 = dyn_cast<Instruction>(X))
+        if (I1->isIdenticalTo(I2))
+          return j;
+  }
   // Scan backwards.
-  for (unsigned j = i-1; j != ~0U && Ops[j].Rank == XRank; --j)
+  for (unsigned j = i-1; j != ~0U && Ops[j].Rank == XRank; --j) {
     if (Ops[j].Op == X)
       return j;
+    if (Instruction *I1 = dyn_cast<Instruction>(Ops[j].Op))
+      if (Instruction *I2 = dyn_cast<Instruction>(X))
+        if (I1->isIdenticalTo(I2))
+          return j;
+  }
   return i;
 }
 
diff --git a/llvm/test/Transforms/Reassociate/negation1.ll b/llvm/test/Transforms/Reassociate/negation1.ll
new file mode 100644
index 00000000000..34b943cf496
--- /dev/null
+++ b/llvm/test/Transforms/Reassociate/negation1.ll
@@ -0,0 +1,15 @@
+; RUN: opt < %s -reassociate -instcombine -S | FileCheck %s
+
+; Test that we can turn things like A*B + X - A*B -> X.
+
+define i32 @test1(i32 %a, i32 %b, i32 %x) {
+; CHECK-LABEL: test1
+; CHECK: ret i32 %x
+
+  %c = mul i32 %a, %b
+  %d = add i32 %c, %x
+  %c1 = mul i32 %a, %b
+  %f = sub i32 %d, %c1
+  ret i32 %f
+}
+