Clear Filters
Clear Filters

Mandelbrot Set with CUDA

9 views (last 30 days)
Anton Kogios
Anton Kogios on 6 Jul 2023
Edited: Anton Kogios on 9 Aug 2023
I am attempting to work through the three methods to compute the Mandelbrot Set outlined on this page. I am stuck on the third method, which I think is also demonstrated here. In particular, I cannot compile the .cu file using mexcuda:
mexcuda -ptx pctdemo_processMandelbrotElement.cu
I have installed MATLAB Support for MinGW-w64 C/C++ Compiler, NVIDIA CUDA Toolkit 12.2 and Microsoft Visual Studio Community 2022 (but I'm not completely sure if I installed the C++ Compiler with Visual Studio).
Is there something else I am missing? I am on a Windows 10 PC with an NVIDIA GeForce GTX 1080 and MATLAB R2023a Update 3. See below for the error(s) I am getting:

Answers (1)

Ty
Ty on 6 Jul 2023
NVidia CUDA requires the Visual C++ compiler. It looks like MATLAB is showing you only have MinGW available. I would double check that you selected the "Desktop development with C++" workflow during the Visual Studio 2022 install. Just re-run the installer which should allow you to add or remove packages.
After you've verified that, see if it now lists the Microsoft Visual C++ compiler when you run the "mex -setup C++" command. You can add a verbose flag with "-v" to see which program folders it's checking.
Lastly, once you've been able to switch the compilers, make sure you've followed the instructions for Setting Up the Prerequisite Products. There are a few system environment variables that need to get set.
  1 Comment
Anton Kogios
Anton Kogios on 9 Aug 2023
Edited: Anton Kogios on 9 Aug 2023
Hi Ty, thank you so much for your reply, and apologies for taking a long time to respond.
I finally got around to looking at this again, and have since gotten Microsoft Visual C++ Compiler to work. However, I am now stuck at this line:
kernel = parallel.gpu.CUDAKernel(ptxFilename,cudaFilename);
What I believe is the source of the issue is the PTX code not compiling correctly. I have attached it as a TXT file, and it appears to not have any actual code in it.
I attempted the first example on the CUDAKernel documentation page, and that seemed to compile fine, not that I would know what a correct PTX file would look like (resulting PTX file also attached as TXT file).
In terms of setting up the prerequisite products as you mentioned, a couple of the coder.checkGpuInstall tests fail on my system. However, I am not too sure how to address them.
gpuEnvObj = coder.gpuEnvConfig;
gpuEnvObj.BasicCodegen = 1;
gpuEnvObj.BasicCodeexec = 1;
gpuEnvObj.DeepLibTarget = 'tensorrt';
gpuEnvObj.DeepCodeexec = 1;
gpuEnvObj.DeepCodegen = 1;
results = coder.checkGpuInstall(gpuEnvObj)
I also attempted to run the GPU Environment Check app, and it all seems to be fine (results attached as PDF).
Edit: I just checked this, and it appears that when I follow the example on this page, it works fine.

Sign in to comment.

Categories

Find more on Get Started with GPU Coder in Help Center and File Exchange

Products


Release

R2023a

Community Treasure Hunt

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

Start Hunting!