NVIDIA CUDA-GDB is a powerful debugger designed for developing massively parallel applications on the GPU. It allows developers to debug both the CPU and GPU portions of their applications simultaneously, handling thousands of threads running on each GPU. CUDA-GDB provides a console-based debugging interface that can be used from the command line on local or remote systems with Telnet or SSH access. It also supports integration with GUI frontends like DDD, EMACS, Nsight Eclipse Edition, and Nsight Visual Studio Code Edition.

Features

• Full Control Over Execution: Debug both 32 and 64-bit CUDA C/C++ applications with breakpoints, single-stepping, and variable inspection.
• Simultaneous Debugging: Debug CPU and GPU portions of applications simultaneously, with support for multiple GPUs.
• Memory and Register Inspection: Examine variables, read/write memory and registers, and inspect GPU state when the application is suspended.
• Conditional Breakpoints: Use conditional breakpoints to identify and correct errors in CUDA code.
• Dynamic Parallelism Support: Debug multiple applications simultaneously using multiple sessions, with support for dynamic parallelism and inlined subroutines.
• Integrated CUDA-MEMCHECK: Run CUDA-MEMCHECK in integrated mode to detect precise exceptions.

Benefits

• Seamless Debugging: Provides a seamless debugging experience for both CPU and GPU portions of applications.
• Enhanced Error Detection: Identify and correct errors with advanced debugging features like conditional breakpoints and integrated CUDA-MEMCHECK.
• Comprehensive Control: Gain full control over the execution of CUDA applications, including memory and register inspection.
• Flexible Integration: Supports integration with various GUI frontends and remote debugging for x86 and ARM systems.
• Improved Productivity: Streamline the debugging process and improve productivity with powerful debugging tools and features.