https://github.com/dominiklindorfer/sycl-intelgpu-quickstart
Lightweight & simplified approach to SYCL development
https://github.com/dominiklindorfer/sycl-intelgpu-quickstart
c cpp dpp gpu gpu-programming intel oneapi oneapi-dpc sycl vscode
Last synced: 3 months ago
JSON representation
Lightweight & simplified approach to SYCL development
- Host: GitHub
- URL: https://github.com/dominiklindorfer/sycl-intelgpu-quickstart
- Owner: DominikLindorfer
- Created: 2024-11-09T14:23:59.000Z (7 months ago)
- Default Branch: master
- Last Pushed: 2024-11-09T14:32:16.000Z (7 months ago)
- Last Synced: 2025-03-16T07:14:36.496Z (3 months ago)
- Topics: c, cpp, dpp, gpu, gpu-programming, intel, oneapi, oneapi-dpc, sycl, vscode
- Language: C++
- Homepage:
- Size: 2.75 MB
- Stars: 2
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# SYCL Programming Setup and Execution
This project demonstrates a minimal setup for SYCL programming using Intel’s oneAPI DPC++ compiler, avoiding the complexity (and error-prone / misconfigured) use of the provided CMake-based setups in [Intel's official samples](https://github.com/oneapi-src/oneAPI-samples/). The `src` directory contains the example programs, and the `build` directory is used for compiling and running them.
## Project Structure
- **`src/`**: Contains source files with basic SYCL programs:
- `hello_world.cpp`: Simple device query and "Hello, World" message.
- `vector_add.cpp`: Performs vector addition on the device.
- `mat_mul.cpp`: Executes a matrix multiplication on the device.
- `PrefixSum.cpp`: Computes the inclusive prefix sum in parallel.
- **`build/`**: Folder for executing compilation commands listed below.
## Setup Instructions
To get the environment ready for compiling and running SYCL programs, I followed these steps:
### 1. Install Prerequisites
- **Visual Studio Build Tools and Visual Studio**: Downloaded from [here](https://visualstudio.microsoft.com/de/downloads/). Both were necessary to successfully set up the SYCL environment.
- **Intel oneAPI Base Toolkit**: Downloaded from [here](https://www.intel.com/content/www/us/en/developer/tools/oneapi/base-toolkit-download.html). This toolkit includes the DPC++ compiler (`dpcpp`) used to compile SYCL programs.
### 2. Set Up VSCode
In Visual Studio Code, select the correct setvars environment by navigating to the path where you installed the oneAPI toolkit into, e.g. `...\Intel\oneAPI\setvars.bat`. This script initializes the environment variables required by oneAPI tools:
which is followed by a success message:
### 3. Compilation and Execution
Intel's provided tutorials at [oneAPI-samples](https://github.com/oneapi-src/oneAPI-samples/tree/master) were overly complicated for me, using CMake extensively, which often failed to compile examples. I had also tried this [VSCode extension](https://marketplace.visualstudio.com/items?itemName=intel-corporation.oneapi-extension-pack) without luck.
Instead, I compiled the programs with these straightforward commands:
```bash
cd build
dpcpp -o hello_world.exe ..\src\hello_world.cpp
dpcpp -o vector_add.exe ..\src\vector_add.cpp
dpcpp -o mat_mul.exe ..\src\mat_mul.cpp
dpcpp -o PrefixSum.exe ..\src\PrefixSum.cpp
```
Each command compiles an individual source file to an executable, making the setup simple and quick.
### 4. Running the Programs
To run any of the executables, navigate to the `build` directory and execute them as follows:
```bash
./hello_world.exe
./vector_add.exe
./mat_mul.exe
./PrefixSum.exe
```
The output e.g. of `hello_world.cpp` will indicate which device the code is running on (CPU or GPU), along with a "Hello, World!" message:
```
Running on 13th Gen Intel(R) Core(TM) i5-13600K
Hello, World!
```
or
```
Running on Intel(R) UHD Graphics 770
Hello, World!
```