[PATCH] D109528: [AArch64][GlobalISel] Add a new reassociation for G_PTR_ADDs.

[Jay Foad via Phabricator via llvm-commits]

foad added a comment.

Instead of optimizing three specific cases do you think it's worth doing something more general, like traversing a whole tree of G_PTR_ADD and G_ADD expressions, collecting the pointers / ints / constant-offsets, and then reassembling them in an optimal order?



================
Comment at: llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp:4188
+  // 3) Move a constant from the LHS of an inner op to the RHS of the outer.
+  // G_PTR_ADD (G_PTR_ADD X, C), Y) -> (G_PTR_ADD (G_PTR_ADD(X, Y), C)
+  // iif (G_PTR_ADD X, C) has one use.
----------------
Mismatched parentheses on this line.


================
Comment at: llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp:4201
 
-      auto NewBase =
-          Builder.buildPtrAdd(PtrTy, Src1Reg, RHS->getOperand(1).getReg());
-      Observer.changingInstr(MI);
-      MI.getOperand(1).setReg(NewBase.getReg(0));
-      MI.getOperand(2).setReg(RHS->getOperand(2).getReg());
-      Observer.changedInstr(MI);
-    };
-  } else {
-    // Try to match example 2.
-    Register LHSSrc1 = LHS->getOperand(1).getReg();
-    Register LHSSrc2 = LHS->getOperand(2).getReg();
-    auto C1 = getConstantVRegVal(LHSSrc2, MRI);
-    if (!C1)
-      return false;
-    auto C2 = getConstantVRegVal(Src2Reg, MRI);
-    if (!C2)
-      return false;
+  // Try to match example 1).
+  if (matchReassocConstantInnerRHS(PtrAdd, RHS, MatchInfo))
----------------
Stray parenthesis.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D109528/new/

https://reviews.llvm.org/D109528