[flang-commits] [flang] [flang] Simplify the comparison of characters (PR #154593)
Slava Zakharin via flang-commits
flang-commits at lists.llvm.org
Fri Sep 5 11:29:59 PDT 2025
================
@@ -0,0 +1,127 @@
+//===- ExpressionSimplification.cpp - Simplify HLFIR expressions ----------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "flang/Optimizer/Builder/FIRBuilder.h"
+#include "flang/Optimizer/HLFIR/HLFIROps.h"
+#include "flang/Optimizer/HLFIR/Passes.h"
+#include "llvm/Support/DebugLog.h"
+
+namespace hlfir {
+#define GEN_PASS_DEF_EXPRESSIONSIMPLIFICATION
+#include "flang/Optimizer/HLFIR/Passes.h.inc"
+} // namespace hlfir
+
+#define DEBUG_TYPE "expression-simplification"
+
+static void removeOp(mlir::Operation *op) {
+ op->dropAllReferences();
+ op->dropAllUses();
+ op->erase();
+}
+
+// Get the first user of `op`.
+// Note that we consider the first user to be the one on the lowest line of
+// the emitted HLFIR. The user iterator considers the opposite.
+template <typename UserOp>
+static UserOp getFirstUser(mlir::Operation *op) {
+ auto it = op->user_begin(), end = op->user_end(), prev = it;
+ for (; it != end; prev = it++)
+ ;
+ if (prev != end)
+ if (auto userOp = mlir::dyn_cast<UserOp>(*prev))
+ return userOp;
+ return {};
+}
+
+// Get the last user of `op`.
+// Note that we consider the last user to be the one on the highest line of
+// the emitted HLFIR. The user iterator considers the opposite.
+template <typename UserOp>
+static UserOp getLastUser(mlir::Operation *op) {
+ if (!op->getUsers().empty())
+ if (auto userOp = mlir::dyn_cast<UserOp>(*op->user_begin()))
+ return userOp;
+ return {};
+}
+
+namespace {
+
+// This class analyzes a trimmed character and removes the trim operation if
+// its result is not used elsewhere.
+class TrimRemover {
+public:
+ TrimRemover(mlir::Value charVal) : charVal(charVal) {}
+ TrimRemover(const TrimRemover &) = delete;
+
+ bool charWasTrimmed();
+ void removeTrim();
+
+private:
+ mlir::Value charVal;
+ hlfir::CharTrimOp trimOp;
+ hlfir::CmpCharOp cmpCharOp;
+ hlfir::DestroyOp destroyOp;
+};
+
+bool TrimRemover::charWasTrimmed() {
+ LDBG() << "charWasTrimmed: " << charVal;
+
+ trimOp = mlir::dyn_cast<hlfir::CharTrimOp>(charVal.getDefiningOp());
+ if (!trimOp)
+ return false;
+ int trimUses = std::distance(trimOp->use_begin(), trimOp->use_end());
+ cmpCharOp = getFirstUser<hlfir::CmpCharOp>(trimOp);
+ destroyOp = getLastUser<hlfir::DestroyOp>(trimOp);
+ return cmpCharOp && destroyOp && trimUses == 2;
+}
+
+void TrimRemover::removeTrim() {
+ LDBG() << "removeTrim: " << trimOp;
+
+ cmpCharOp->replaceUsesOfWith(trimOp.getResult(), trimOp.getChr());
+ removeOp(destroyOp);
+ removeOp(trimOp);
+}
+
+class ExpressionSimplification
+ : public hlfir::impl::ExpressionSimplificationBase<
+ ExpressionSimplification> {
+public:
+ using ExpressionSimplificationBase<
+ ExpressionSimplification>::ExpressionSimplificationBase;
+
+ void runOnOperation() override;
+
+private:
+ // Simplify character comparisons.
+ // Because character comparison appends spaces to the shorter character,
+ // calls to trim() that are used only in the comparison can be eliminated.
+ //
+ // Example:
+ // `trim(x) == trim(y)`
+ // can be simplified to
+ // `x == y`
+ void simplifyCmpChar(hlfir::CmpCharOp cmpChar);
+};
+
+void ExpressionSimplification::simplifyCmpChar(hlfir::CmpCharOp cmpChar) {
+ TrimRemover lhsTrimRem(cmpChar.getLchr());
+ TrimRemover rhsTrimRem(cmpChar.getRchr());
+
+ if (lhsTrimRem.charWasTrimmed())
+ lhsTrimRem.removeTrim();
+ if (rhsTrimRem.charWasTrimmed())
+ rhsTrimRem.removeTrim();
+}
+
+void ExpressionSimplification::runOnOperation() {
+ mlir::ModuleOp module = getOperation();
+ module.walk([&](hlfir::CmpCharOp cmpChar) { simplifyCmpChar(cmpChar); });
----------------
vzakhari wrote:
I would recommend walking all trim operations instead of the comparisons, e.g. consider the case where the same trim result is used in multiple character comparisons. We may not have such cases now, but the CSE may be able to optimize it in codes like `(trim(x) == trim(y)).and.(trim(x) == trim(z))`.
https://github.com/llvm/llvm-project/pull/154593
More information about the flang-commits
mailing list