study-cuda

This repository is for studying NVIDIA CUDA C++

Contents

• 1_hello_kernel: what is kernel. kernel that prints "hello world!" twice using two threads.
• 2_grid_and_block: what is grid and block. prints "hello world!" using multiple blocks.
• 3_dim3: what is dim3. launch a kernel using dim3.
• 4_thread_id: kernel that prints the thread id and the block id.
• 5_warp: what is warp and branching.
• 6_global_memory: what is global memory. kernel that (element-wise) multiplies two arrays.

To be writen

• 7_shared_memory: what is shared memory. kernel that multiplies two arrays using shared memory.
• 8_constant_memory: what is constant memory. kernel that multiplies two arrays using constant memory.
• 9_register: what is register. kernel that multiplies an array and a value using register.
• 10_memory_coalescing: what is memory coalescing. kernel that multiplies two arrays making use of coalescing.
• 11_pinned_memory: what is pinned memory. kernel that loads data from pinned memory.
• 12_dynamic_parallelism: what is dynamic parallelism. kernel that calls another kernel.
• 13_stream: what is stream. run two kernels in parallel using one GPU.
• 14_multi_gpus: run two kernels in parallel using two GPUs.