[clang] [NFC][Clang][OpenMP] Add helper functions/utils for finding/comparing attach base-ptrs. (PR #155625)
Krzysztof Parzyszek via cfe-commits
cfe-commits at lists.llvm.org
Mon Sep 22 03:10:41 PDT 2025
================
@@ -6765,12 +6765,255 @@ llvm::Value *CGOpenMPRuntime::emitNumThreadsForTargetDirective(
namespace {
LLVM_ENABLE_BITMASK_ENUMS_IN_NAMESPACE();
+/// Utility to compare expression locations.
+/// Returns true if expr-loc of LHS is less-than that of RHS.
+/// This function asserts that both expressions have valid expr-locations.
+static bool compareExprLocs(const Expr *LHS, const Expr *RHS) {
+ // Assert that neither LHS nor RHS can be null
+ assert(LHS && "LHS expression cannot be null");
+ assert(RHS && "RHS expression cannot be null");
+
+ // Get source locations
+ SourceLocation LocLHS = LHS->getExprLoc();
+ SourceLocation LocRHS = RHS->getExprLoc();
+
+ // Assert that we have valid source locations
+ assert(LocLHS.isValid() && "LHS expression must have valid source location");
+ assert(LocRHS.isValid() && "RHS expression must have valid source location");
+
+ // Compare source locations for deterministic ordering
+ return LocLHS < LocRHS;
+}
+
// Utility to handle information from clauses associated with a given
// construct that use mappable expressions (e.g. 'map' clause, 'to' clause).
// It provides a convenient interface to obtain the information and generate
// code for that information.
class MappableExprsHandler {
public:
+ /// Custom comparator for attach-pointer expressions that compares them by
+ /// complexity (i.e. their component-depth) first, then by their expr-locs if
+ /// they are semantically different.
----------------
kparzysz wrote:
Source locations aren't always present. Some expressions can be constructed by the compiler and will not have locations.
If the expressions of interest are limited to a certain set, you could "pre-traverse" them in a deterministic order (to give them numbers), and use that as a tie-breaker.
https://github.com/llvm/llvm-project/pull/155625
More information about the cfe-commits
mailing list