Snippet to find number of GPUs on system and pick one for CUDA use

100%

Snippet to find number of GPUs on system and pick one for CUDA use

✓ Published0🌍 Public

Last edited Jul 15, 2016

Created on Jul 14, 2016

This snippet demonstrates how to query the local system for the number of available CUDA-enabled GPUs and assign one to a specific thread or MPI rank. It shows a two-step workflow: first, `getGPUCount()` uses `cudaGetDeviceCount` to retrieve the GPU count, returning zero on failure; second, `pickGPU()` maps a rank to a GPU via modulo arithmetic and activates it with `cudaSetDevice`, returning -1 if the assignment fails. The code relies on the CUDA runtime API and is intended for multithreaded or MPI environments to avoid device conflicts.

AI-generated description

Similar vizzes

Loading thumbnail…

Python venv and Nvidia Driver installation instruction (https://bit.ly/2E3H9Rk)

This example documents the setup of Python virtual environments and Nvidia GPU computing prerequisites for TensorFlow. It provides step-by-step instructions for installing the Nvidia driver, CUDA, cuDNN, and tensorflow-gpu, along with commands for creating virtual environments using Python’s `venv` module, Anaconda’s `conda create`, and Pipenv. The text includes platform-specific shell commands for Windows, Mac OS, and Linux, such as `python3 -m venv`, `source activate`, and `conda install tensorflow-gpu`, and references verification tools like `nvidia-smi` and `nvcc --version`.

VVioletVivirand

44% match

Loading thumbnail…

HeatMap

This heatmap visualizes the monthly popularity of GPUs on eBay, using color intensity to represent the number of units sold for each GPU model across several months. The darker orange cells indicate higher sales volumes, while lighter cells show lower activity. The data is loaded from a CSV file using D3 v7, and the visualization is rendered as an SVG. The code employs `d3.scaleBand` for the x and y axes, `d3.scaleSequential` with an orange interpolator for color encoding, and tooltips to display the exact sold count on hover.

TTerence Lin

41% match

Loading thumbnail…

Gist 8be3539eabe0a1c2e07e749fb3ee6372

This example compares the performance of spatial joins using Dask-GeoPandas, PostGIS, and GPU-based RAPIDS.AI. It downloads Philadelphia parking violation data from CARTO and neighborhood shapefiles, then benchmarks how quickly Dask-GeoPandas processes the spatial join on a local cluster with eight workers. The notebook uses `dask_geopandas`, `dask.dataframe`, and `distributed.Client` to set up parallel processing. It reproduces Paul Ramsey's original benchmark to evaluate Dask-based execution speed against the other technologies.

TThomasG77

39% match

Loading thumbnail…

Counting the number of cores in Azure Analysis Services

This example visualizes how many processor cores are available in an Azure Analysis Services instance by querying Windows performance counters. It shows the processor frequency for each individual core, from 0 to 31, revealing the total core count and their activity. The visualization uses a DAX query with the `SYSTEMRESTRICTSCHEMA` function to retrieve data from `DISCOVER_PERFORMANCE_COUNTERS`, explicitly listing each `Processor Information` counter. The rendering relies on the data returned by this query, presenting the per-core frequencies as a direct measure of the allocated compute resources.

HHugoberry

38% match

Loading thumbnail…

Gist 8131253

This example simulates a perfect out-shuffle on a deck of cards and tracks how many shuffles are needed for the deck to return to its original order. The code implements two approaches: a fast solution that uses greatest common divisor and least common multiple calculations to determine the shuffle count, and a naive solution that iteratively compares the shuffled deck against a sorted version until they match. The visualization relies on JavaScript and the Underscore.js library for utility functions like memoization, array manipulation, and uniqueness checks, with shuffle counts logged to the console.

YYonet

38% match

Loading thumbnail…

Mandelbrot Set in Julia

This example renders the Mandelbrot set in Julia by leveraging a custom GPU-based implementation through CUDA.jl. The visualization computes the classic fractal in the complex plane, mapping iteration counts to colors using the ColorSchemes package. It uses CUDAnative and CUDAdrv for parallel computation, with the final image displayed via the Julia plotting ecosystem. The code includes a dynamic zoom interaction, letting viewers explore the fractal’s fine-scale detail.

WWikunia

38% match

Loading thumbnail…

Python snippets for quickly finding the platform type, cpu and memory stats of a Linux machine.

This example shows how to quickly retrieve platform type, CPU model names, and memory statistics from a Linux machine. It parses the `/proc/cpuinfo` and `/proc/meminfo` files using Python’s `collections.OrderedDict` to preserve field order, printing each processor’s model name and total/free memory. The third snippet uses the `platform` module’s `uname()` and `linux_distribution()` functions to display system and distribution details. Together, these scripts demonstrate lightweight, file-based system introspection without external dependencies.

PP7h

38% match

Loading thumbnail…

n/2 matching numbers : pseudocode

This example demonstrates a pseudocode representation of the n/2 majority element problem, showing how to identify a number that appears more than half the time in an array. The page displays a hardcoded test array `[1, 2, 4, 4, 4, 4, 4, 13, 10, 6]` and a target number `4`, then logs the indices where the number matches via `console.log` in a `for` loop. The visualization presents the pseudocode alongside the algorithm’s logic, though it renders no graphical output or charts, relying solely on HTML body text and JavaScript execution.

NNita

38% match