averello/swift-mirror
1
0
Fork 0
mirror of https://github.com/apple/swift.git synced 2025-12-14 20:36:38 +01:00
Code Issues Packages Projects Releases Wiki Activity

[CSOptimizer] Allow only widening CGFloat->Double conversions while matching candidate arguments

Browse Source 
Allow CGFloat -> Double widening conversions between
candidate argument types and parameter types. This would
make sure that Double is always preferred over CGFloat
when using literals and ranking supported disjunction
choices. Narrowing conversion (Double -> CGFloat) should
be delayed as much as possible.
This commit is contained in:
Pavel Yaskevich
2024-11-25 21:55:57 -08:00
parent cb876cbd9e
commit bf8ae3bc1b
2 changed files with 59 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

39
lib/Sema/CSOptimizer.cpp
View File

@@ -530,18 +530,22 @@ static void determineBestChoicesInContext(
scoreCandidateMatch =
[&](GenericSignature genericSig, Type candidateType, Type paramType,
MatchOptions options) -> std::optional<double> {
auto areEqual = [&options](Type a, Type b) {
// Double<->CGFloat implicit conversion support for literals
// only since in this case the conversion might not result in
// score penalty.
if (options.contains(MatchFlag::Literal) &&
((a->isDouble() && b->isCGFloat()) ||
(a->isCGFloat() && b->isDouble())))
return true;
auto areEqual = [&](Type a, Type b) {
return a->getDesugaredType()->isEqual(b->getDesugaredType());
};
// Allow CGFloat -> Double widening conversions between
// candidate argument types and parameter types. This would
// make sure that Double is always preferred over CGFloat
// when using literals and ranking supported disjunction
// choices. Narrowing conversion (Double -> CGFloat) should
// be delayed as much as possible.
if (options.contains(MatchFlag::OnParam)) {
if (candidateType->isCGFloat() && paramType->isDouble()) {
return options.contains(MatchFlag::Literal) ? 0.2 : 0.9;
}
}
if (options.contains(MatchFlag::ExactOnly))
return areEqual(candidateType, paramType) ? 1 : 0;
@@ -559,12 +563,12 @@ static void determineBestChoicesInContext(
if (candidateType->isInt() &&
TypeChecker::conformsToKnownProtocol(
paramType, KnownProtocolKind::ExpressibleByIntegerLiteral))
return 0.2;
return paramType->isDouble() ? 0.2 : 0.3;
if (candidateType->isDouble() &&
TypeChecker::conformsToKnownProtocol(
paramType, KnownProtocolKind::ExpressibleByFloatLiteral))
return 0.2;
return 0.3;
}
return 0;
@@ -916,6 +920,19 @@ static void determineBestChoicesInContext(
// with a lot of favored overloads because on the result type alone.
if (decl->isOperator() && !isStandardComparisonOperator(decl)) {
if (llvm::any_of(resultTypes, [&](const Type candidateResultTy) {
// Avoid increasing weight based on CGFloat result type
// match because that could require narrowing conversion
// in the arguments and that is always detrimental.
//
// For example, `has_CGFloat_param(1.0 + 2.0)` should use
// `+(_: Double, _: Double) -> Double` instead of
// `+(_: CGFloat, _: CGFloat) -> CGFloat` which would match
// parameter of `has_CGFloat_param` exactly but use a
// narrowing conversion for both literals.
if (candidateResultTy->lookThroughAllOptionalTypes()
->isCGFloat())
return false;
return scoreCandidateMatch(genericSig,
overloadType->getResult(),
candidateResultTy,

31
validation-test/Sema/type_checker_perf/fast/array_literal_with_operators_and_double_literals.swift Normal file
View File

@@ -0,0 +1,31 @@
// RUN: %target-swift-frontend(mock-sdk: %clang-importer-sdk) %s -typecheck -solver-expression-time-threshold=1
// REQUIRES: asserts,no_asan
// REQUIRES: objc_interop
// FIXME: This should be a scale-test but it doesn't allow passing `%clang-importer-sdk`
// This import is important because it brings CGFloat and
// enables Double<->CGFloat implicit conversion that affects
// literals below.
import Foundation
let p/*: [(String, Bool, Bool, Double)]*/ = [
("", true, true, 0 * 0.0 * 0.0),
("", true, true, 0 * 0.0 * 0.0),
("", true, true, 0 * 0.0 * 0.0),
("", true, true, 0 * 0.0 * 0.0),
("", true, true, 0 * 0.0 * 0.0),
("", true, true, 0 * 0.0 * 0.0),
("", true, true, 0 * 0.0 * 0.0),
("", true, true, 0 * 0.0 * 0.0),
("", true, true, 0 * 0.0 * 0.0),
("", true, true, 0 * 0.0 * 0.0),
("", true, true, 0 * 0.0 * 0.0),
("", true, true, 0 * 0.0 * 0.0),
("", true, true, 0 * 0.0 * 0.0),
("", true, true, 0 * 0.0 * 0.0),
("", true, true, 0 * 0.0 * 0.0),
("", true, true, 0 * 0.0 * 0.0),
("", true, true, 0 * 0.0 * 0.0)
]