为么 Swift 里面自己实现的 concurrentMap() 会比 STL 自带的 map() 要慢？ 据称自带的 map() 还是单线程的

concurrentMap 实现如下：

extension Array { func concurrentMap<T>(_ mapper: (Element) -> T) -> [T] { if self.count == 0 { return [] } let q = DispatchQueue(label: "dispatch sync queue") var result = Array<T?>(repeating: nil, count: self.count) DispatchQueue.concurrentPerform(iterations: self.count) { (idx) in let element = self[idx] let transformed = mapper(element) q.sync { result[idx] = transformed } } return result.map{ $0! } } } 

测试用例如下：

let a = Array(0..<60000) let startTime = mach_absolute_time() a.map{ $0 / 2000 } let midTime = mach_absolute_time() a.concurrentMap{ $0 / 2000 } let endTime = mach_absolute_time() 

结果：

There are 60000 elements. STL map consumes: 8144985 concurrentMap consumes: 567991454