https://github.com/mnicely/nvml_examples

Examples showing how to utilize the NVML library for GPU monitoring
https://github.com/mnicely/nvml_examples

cublas cuda nvidia nvml

Last synced: 4 months ago
JSON representation

Examples showing how to utilize the NVML library for GPU monitoring

• Host: GitHub
• URL: https://github.com/mnicely/nvml_examples
• Owner: mnicely
• License: bsd-3-clause
• Created: 2019-05-09T22:57:45.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2022-05-31T20:18:02.000Z (over 2 years ago)
• Last Synced: 2023-09-20T01:03:58.329Z (over 1 year ago)
• Topics: cublas, cuda, nvidia, nvml
• Language: C++
• Size: 256 KB
• Stars: 16
• Watchers: 0
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# nvml_example

Example showing how to utilize NVML library for GPU monitoring with high sampling rate.

This example utilizing the NVML Library and C++11 mutlithreading to provide GPU monitoring with a high sampling rate. Normally, one would pipe nvidia-smi to a file, but this can cause excessive I/O usage.

Also provided is a Matlab script used to plot the data.

### Prerequisites

Example requires Tesla card because some calls are not supported on Geforce family.

### Built With

This example utilizes the following toolsets:

* cuBLAS

* Thrust

* C++11 multithreading

* NVIDIA Management Library (NVML)

### Deployment

Include header class

```#include nvmlClass.h```

Create object of nvmlClass

```nvmlClass nvml( dev, filename );```

Create thread to run in parallel, gathering data

```std::thread threadStart( &nvmlClass::getStats, &nvml );```

Create thread to kill getStats function

```std::thread threadKill( &nvmlClass::killThread, &nvml );```

Join all threads

```threadStart.join( );```

```threadKill.join( );```