Can I still rely on the order of the output elements when using par_unseq?

user_106064 0
Viewed 464

After reading the documentation I'm still confused about the usage of par_unseq. I know I can't tell anything about the order of execution because of threading and vectorization but can I still rely on the order of the outputs?

transform([x0, x1, x2], f) == [f(x0), f(x1), f(x2)]]

In order words, is this test ever going to fail?

std::vector<int> xs = {1, 2, 3, 4};
std::vector<int> ys(xs.size());

std::transform(
    std::execution::par_unseq,
    cbegin(xs), cend(xs),
    begin(ys),
    [](int x) { return x*x; });

std::vector<int> expected = {1, 4, 9, 16};
ASSERT_EQ(expected , ys);
c++17 parallel-processing algorithm c++
2 Answers

The Standard, [alg.transform], reads:

Effects: Assigns through every iterator i in the range [result,result + (last1 - first1)) a new corresponding value equal to op(*(first1 + (i - result)) or binary_op(*(first1 + (i - result)), *(first2 + (i - result))).

and (thanks, @Caleth), [algorithms.parallel.overloads]:

Unless otherwise specified, the semantics of ExecutionPolicy algorithm overloads are identical to their overloads without.

So, yes, you can rely on the order in the output.

user_16251870

No, your test is never going to fail because even if order of execution changes, ys[0...3] = xs[0...3] * xs[0...3] = {1*1, 2*2, 3*3, 4*4}; won't change.

user_39751770
Related