Welcome!
-
Robust CUDA programs require systematic error checking since GPU operations can fail silently. When you start a kernel on the GPU, it runs immediately without giving an error code if something goes wrong. Using cudaError_t, cudaGetLastError(), and error-checking macros helps…
-
CUDA: Compilation and Execution
CUDA programs require special compilation to generate both CPU and GPU code. The nvcc tool helps by splitting the code into two parts: host (C++) and device (PTX/SASS). Then it combines them. Using the right compiler flags is important, especially…
-
CUDA: Thread Indexing and IDs
Thread indexing is how each parallel thread determines which data element to process. Computing a unique global thread ID from threadIdx, blockIdx, and blockDim enables thousands of threads to safely access different array elements without conflicts. This way of connecting…
-
CUDA: Hello World Kernel
Our first CUDA kernel helps connect CPU and GPU programming. It runs a simple function using many parallel threads. This is different from normal “Hello World” programs because it shows true parallelism, where hundreds or thousands of threads work at…
-
CUDA Programming Model
The CUDA programming model splits work between two parts: the CPU (host) and the GPU (device). The CPU controls what happens in the program and sends tasks called kernels to the GPU for processing. To write good CUDA programs, you…