Doesn't the GPU Coder support singleton expansion of ARRAYFUN?
1 Ansicht (letzte 30 Tage)
Ältere Kommentare anzeigen
Shogo Muramatsu
am 25 Sep. 2021
Kommentiert: Shogo Muramatsu
am 2 Okt. 2021
I have a question about SINGLETON EXPANSION for ARRAYFUN on GPU.
For gpuArrays, @times with ARRAYFUN and direct calculation on M-file both work fine.
However, MEX generated by GPU Coder from a code using ARRAYFUN does not seem to perform SINGLETON EXPANSION well.
Doesn't the GPU Coder support SINGLETON EXPANSION of ARRAYFUN?
Shogo MURAMATSU
Sample code
function Y = arrayfuntest(m,X) %#codegen
Y = arrayfun(@times,m,X);
%Y = m.*X;
end
Code generation for ARRAYFUN
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/749309/image.png)
Code generation for a direct operation
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/749314/image.png)
Preparation of arrays
m = gpuArray(single([1;-1;1]))
X = gpuArray(rand(3,'single'))
Execution with M-file
Y = arrayfuntest(m,X)
MEX Generation
cfg = coder.gpuConfig('mex');
nChs = 64;
datatype = 'single';
aM = coder.typeof(gpuArray(cast(0,datatype)),[nChs 1],[1 0]);
aX = coder.typeof(gpuArray(cast(0,datatype)),[nChs nChs],[1 1]);
cfg.DynamicMemoryAllocation = 'Off';
cfg.GenerateReport = true;
codegen -config cfg -o arrayfuntest_mex arrayfuntest.m -args { aM, aX }
Execution with MEX-file
Y = arrayfuntest_mex(m,X)
0 Kommentare
Akzeptierte Antwort
Joss Knight
am 2 Okt. 2021
Support for dimension expansion is actually restricted to the GPU version of arrayfun, and GPU Coder supports the CPU version's syntax. So the answer is no, it doesn't support dimension expansion.
Weitere Antworten (0)
Siehe auch
Kategorien
Mehr zu Get Started with GPU Coder finden Sie in Help Center und File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!