Using parfor and parfeval in separate pools

4 次查看（过去 30 天）
Tyler Warner
Tyler Warner 2018-5-22
编辑： OCDER 2018-5-22
Is it possible, or even desirable, to make separate pools of workers that total to the number of cores on a system to speed up code processing?
Take the pseudo-code as an example for a 20-physical-core server:
parpool(16)
p = gcp();
parpool(4)
q = gcp()
parfor i = start:last % p
func1(q, i)
end
where
function func1(q, i)
for j = 0:i
parfeval(q,func2(j),...)
end
end

请先登录，再回答此问题。

采纳的回答

OCDER
OCDER 2018-5-22
编辑：OCDER 2018-5-22
I don't think it'll be beneficial nor possible. I get an error like this, since you can't have multiple pools running. Matlab won't know which pool to distribute which data:
parpool(16)
p = gcp();
parpool(4)
q = gcp()
Error using parpool (line 104)
Found an interactive session. You cannot have multiple interactive sessions
open simultaneously. To terminate the existing session, use 'delete(gcp('nocreate'))'.
This seems like the simplest use of parfor for 20 cores:
parpool(20)
parfor i = start:last % p
B(i) = func2(A(i));
end
If this was for working around the 16-core limit when using parpool, this limit was removed in R2014a. See notes here for R2014a, 1st line: https://www.mathworks.com/help/releases/R2017a/distcomp/release-notes.html?rntext=&startrelease=R2014a&endrelease=R2017a

更多回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Parallel Computing Fundamentals 的更多信息

标签

产品


版本

R2017b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by