extremely inefficient piece of code ... bug?

2 次查看（过去 30 天）

显示更早的评论

Michal 2023-8-31

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2015336-extremely-inefficient-piece-of-code-bug

编辑： Matt J 2023-8-31

采纳的回答： Matt J

test.m

在 MATLAB Online 中打开

I have two codes that implement exactly the same task:

   idx(idx<=0) = 1;
   idx(idx>=n) = n-1;  

and

    if idx <= 0
        idx = 1;
    elseif idx >= n
        idx = n-1;
    end

where idx and n are a scalar integer values.

This piece of code is evaluating multiple times (N = 1e6, for example). During profiling of the code, I found very strange inefficiency of the first code, which is about ~ 100-1000x slower than the 2nd one. I use the 1st type of code to get more readable code by avoiding if-else construct, but the overall performance is on R2023a really terrible. See attached test code (test.m).

Is this behavior normal, or is it a BUG?

Please, could you verify this code on other (older) MATLAB releases to see, if the problem is only R2023a related or not?

24 个评论
显示 22更早的评论隐藏 22更早的评论

Bruno Luong 2023-8-31

Cheese: For the record: My timing figure is directed to Rik's comment, then you continue to draw me in with your justification about your montecarlo simulation, that I initially never ask for. I'm not really interested in knowing the topic to be honest. That's your work not mine. You don't need to attck me with "Especially if you don't know anything about the topic at hand."

Michal 2023-8-31

编辑：Michal 2023-8-31

@Bruno Luong I just react on your statement: "Don't tell me that you mainly do scalar clipping on your montecarlo simulation."

I tried to explain to you why I use "scalar clipping" in my MC simulation. You finally comment my explanation by: "Noise interpolation has no sense to start with...."

So, what ...?

请先登录，再进行评论。

请先登录，再回答此问题。

采纳的回答

Matt J 2023-8-31

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2015336-extremely-inefficient-piece-of-code-bug#answer_1298596

编辑：Matt J 2023-8-31

在 MATLAB Online 中打开

Well, I think the bottom line is just that the JIT does not have any optimization for indexing expressions like idx(idx<=0). That might be because such an optimization would need to know in advance that idx is a scalar, and remains a scalar throughout the loop.

Coversely, parsing an if-statement,

if idx>threshold
    
end

doesn't require nearly as much work, even when idx is a vector. The if test is done by looping through the vector elements idx(i)>threshold, and as soon as one of the elements is false, it bails out.

5 个评论
显示 3更早的评论隐藏 3更早的评论

Michal 2023-8-31

This is only a special (scalar) case of general array logical indexing... Why should be this special case penalized by absence of optimization?

Matt J 2023-8-31

编辑：Matt J 2023-8-31

No, I'm saying that to achieve the same performance you're seeing with if-else, the scalar indexing case would have to receive its own special treatment from the JIT, separate from what is normally done for general logical indexing.

请先登录，再进行评论。