From a5f0824eb518ba82e4395b11e4207eaf5197556d Mon Sep 17 00:00:00 2001
From: Roman Lebedev <lebedev.ri@gmail.com>
Date: Fri, 19 Jul 2019 08:25:43 +0000
Subject: [InstCombine] Dropping redundant masking before left-shift [0/5]
 (PR42563)

Summary:
If we have some pattern that leaves only some low bits set, and then performs
left-shift of those bits, if none of the bits that are left after the final
shift are modified by the mask, we can omit the mask.

There are many variants to this pattern:
a. `(x & ((1 << MaskShAmt) - 1)) << ShiftShAmt`
All these patterns can be simplified to just:
`x << ShiftShAmt`
iff:
a. `(MaskShAmt+ShiftShAmt) u>= bitwidth(x)`

alive proof:
a: https://rise4fun.com/Alive/wi9

Indeed, not all of these patterns are canonical.
But since this fold will only produce a single instruction
i'm really interested in handling even uncanonical patterns,
since i have this general kind of pattern in hotpaths,
and it is not totally outlandish for bit-twiddling code.

For now let's start with patterns where both shift amounts are variable,
with trivial constant "offset" between them, since i believe this is
both simplest to handle and i think this is most common.
But again, there are likely other variants where we could use
ValueTracking/ConstantRange to handle more cases.

https://bugs.llvm.org/show_bug.cgi?id=42563

Reviewers: spatel, nikic, huihuiz, xbolva00

Reviewed By: xbolva00

Subscribers: efriedma, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D64512

llvm-svn: 366535
---
 .../Transforms/InstCombine/InstCombineShifts.cpp   | 50 ++++++++++++++++++++++
 1 file changed, 50 insertions(+)

(limited to 'llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp')

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp b/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp
index c821292400c..bac4c07cf82 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp
@@ -65,6 +65,53 @@ reassociateShiftAmtsOfTwoSameDirectionShifts(BinaryOperator *Sh0,
   return NewShift;
 }
 
+// If we have some pattern that leaves only some low bits set, and then performs
+// left-shift of those bits, if none of the bits that are left after the final
+// shift are modified by the mask, we can omit the mask.
+//
+// There are many variants to this pattern:
+//   a)  (x & ((1 << MaskShAmt) - 1)) << ShiftShAmt
+// All these patterns can be simplified to just:
+//   x << ShiftShAmt
+// iff:
+//   a) (MaskShAmt+ShiftShAmt) u>= bitwidth(x)
+static Instruction *
+dropRedundantMaskingOfLeftShiftInput(BinaryOperator *OuterShift,
+                                     const SimplifyQuery &SQ) {
+  assert(OuterShift->getOpcode() == Instruction::BinaryOps::Shl &&
+         "The input must be 'shl'!");
+
+  Value *Masked = OuterShift->getOperand(0);
+  Value *ShiftShAmt = OuterShift->getOperand(1);
+
+  Value *MaskShAmt;
+
+  // ((1 << MaskShAmt) - 1)
+  auto MaskA = m_Add(m_Shl(m_One(), m_Value(MaskShAmt)), m_AllOnes());
+
+  Value *X;
+  if (!match(Masked, m_c_And(MaskA, m_Value(X))))
+    return nullptr;
+
+  // Can we simplify (MaskShAmt+ShiftShAmt) ?
+  Value *SumOfShAmts =
+      SimplifyAddInst(MaskShAmt, ShiftShAmt, /*IsNSW=*/false, /*IsNUW=*/false,
+                      SQ.getWithInstruction(OuterShift));
+  if (!SumOfShAmts)
+    return nullptr; // Did not simplify.
+  // Is the total shift amount *not* smaller than the bit width?
+  // FIXME: could also rely on ConstantRange.
+  unsigned BitWidth = X->getType()->getScalarSizeInBits();
+  if (!match(SumOfShAmts, m_SpecificInt_ICMP(ICmpInst::Predicate::ICMP_UGE,
+                                             APInt(BitWidth, BitWidth))))
+    return nullptr;
+  // All good, we can do this fold.
+
+  // No 'NUW'/'NSW'!
+  // We no longer know that we won't shift-out non-0 bits.
+  return BinaryOperator::Create(OuterShift->getOpcode(), X, ShiftShAmt);
+}
+
 Instruction *InstCombiner::commonShiftTransforms(BinaryOperator &I) {
   Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
   assert(Op0->getType() == Op1->getType());
@@ -629,6 +676,9 @@ Instruction *InstCombiner::visitShl(BinaryOperator &I) {
   if (Instruction *V = commonShiftTransforms(I))
     return V;
 
+  if (Instruction *V = dropRedundantMaskingOfLeftShiftInput(&I, SQ))
+    return V;
+
   Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
   Type *Ty = I.getType();
   unsigned BitWidth = Ty->getScalarSizeInBits();
-- 
cgit v1.2.3