nz::CudaException Class Reference

A final class that represents CUDA exceptions, inheriting from std::runtime_error. More...

Detailed Description

A final class that represents CUDA exceptions, inheriting from std::runtime_error.

This class is designed to handle and provide detailed information about CUDA-related errors. It captures the file, line number, CUDA error code, and the expression that caused the error. By overriding the what() method, it can return a formatted error message with all these details.

Type Definitions:

• None

Key Features:

• Error Information Capture: Records the file, line number, CUDA error code, and the expression where the error occurred.
• Formatted Error Message: Generates a detailed and human - readable error message that includes all the captured information.

Usage Example:

#include <iostream>
#include <cuda_runtime.h>
 
// Assume DL_API is a valid macro for visibility
class DL_API CudaException final : public std::runtime_error {
    // Class definition as provided
};
 
void someCudaFunction() {
    cudaError_t err = cudaMalloc(nullptr, 1024);
    if (err!= cudaSuccess) {
        throw CudaException(__FILE__, __LINE__, err, "cudaMalloc(nullptr, 1024)");
    }
}
 
int main() {
    try {
        someCudaFunction();
    } catch (const CudaException& e) {
        std::cerr << e.what() << std::endl;
    }
    return 0;
}

Note

• Ensure that the buffer size of 1024 characters is sufficient for the error message. If the message exceeds this size, it will be truncated.
• The format_message() method should be called during object construction to ensure the error message is properly formatted.

Author

Mgepahmge(https://github.com/Mgepahmge)

Date

2024/07/24

Definition at line 62 of file NeuZephyrCudaErrorHandling.cuh.

---

The documentation for this class was generated from the following file:

• D:/Users/Mgepahmge/Documents/C Program/NeuZephyr/include/NeuZephyr/NeuZephyrCudaErrorHandling.cuh