GPU Performance Analyzer does not accept Hardware Config argument

26 次查看（过去 30 天）

Ryan 2024-8-1

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2142171-gpu-performance-analyzer-does-not-accept-hardware-config-argument

评论： Ryan 2024-8-2，16:04

Hello, I am getting an error when passing hardware configuration to the MATLAB GPU Performance Analyzer.

I am using the Mandelbrot Count example along with this tutorial: https://www.mathworks.com/help/gpucoder/ug/gpu-profiling-nvidia-jetson.html to run the GPU Performance Analyzer with my connected NVIDIA Jetson Orin Nano. I can add hardware options to the configuration in two ways:

1: as the tutorial says,

2: after using the config editor GUI when you double click in the variables list.

The ways are commented below:

%% This is an excerpt from my overall script
cfg = coder.gpuConfig('dll','ecoder',true);
% cfg.Hardware = coder.Hardware('NVIDIA Jetson'); % 1. from tutorial
% cfg.Hardware = coder.hardware( "NVIDIA Jetson" ); % 2. directly from cfg GUI
inputs = {maxIterations,xGrid,yGrid};
designFileName = 'mandelbrot_count';
gpuPerformanceAnalyzer(designFileName, {maxIterations, xGrid, yGrid}, NumIterations=2, Config=cfg)

If I uncomment one of those lines, I get an error:

In mandelbrot_count_dll_script (line 36) 
Error using gpuPerformanceAnalyzer
Index exceeds the number of array elements. Index must not exceed 0.
Error in mandelbrot_count_dll_script (line 36)
gpuPerformanceAnalyzer(designFileName, {maxIterations, xGrid, yGrid}, NumIterations=2, Config=cfg)

When I run without this hardware configuration it runs without error but defaults to the host machine GPU which defeats the purpose - I need this to run on the NVIDIA Jetson target board. Any help would be much appreciated. Thank you!

2 个评论
显示无隐藏无

Chao Luo 2024-8-2，12:48

在 MATLAB Online 中打开

Did you setup the hwobj before running gpuPerformanceAnalyzer, like this:

hwobj = jetson('your_board_name_or_IP','user','passwd');

Ryan 2024-8-2，16:04

在 MATLAB Online 中打开

Yes I did. However I figured out the issue. The solution ended up involving your answer to my previous question: https://www.mathworks.com/matlabcentral/answers/2140006-nvidia-jetson-setup-issue-coder-checkgpuinstall-can-t-find-nvcc

Once I added 'cfg.GpuConfig.ComputeCapability = '8.7', the error went away.

%% 3) Code generation
cfg = coder.gpuConfig('dll','ecoder',true);
cfg.Hardware = coder.hardware( "NVIDIA Jetson" );
cfg.GpuConfig.ComputeCapability = '8.7';  % <-- this line fixed the error
inputs = {maxIterations,xGrid,yGrid};
designFileName = 'mandelbrot_count';
gpuPerformanceAnalyzer(designFileName, {maxIterations, xGrid, yGrid}, NumIterations=2, Config=cfg)