文件列表:

hello.cu (336, 2020-10-19)
hello2.cu (724, 2020-10-19)
histogram.cu (8698, 2020-10-19)
matrix.cu (6049, 2020-10-19)
performance_histogram.png (51971, 2020-10-19)
share_mem.cu (10257, 2020-10-19)

# CUDA C Programming demos ## Dependencies - CUDA ## Items - *hello.cu*: hello world from GPU! - *hello2.cu*: understanding the thread index (1D, 2D, 3D). - *share_mem.cu*: understanding the memory hierachy, specifically, the power of shared memory compared with the global memory! - *matrix.cu*: an example of matrix multiplication. to **compile** these files, run this command ``` nvcc hello.cu -o hello ``` or ``` nvcc share_mem.cu -o share -std=c++11 ``` ## Performance analysis to eval the performance of different kernels, run the command like ``` nvprof ./share ``` or ``` nvprof --metrics gld_throughput --metrics gst_throughput./share ``` where `gld_throughput` means global memory load throughput, `gst_througput` refers to global memory store throughput. More metrics can be found via `nvprof --query-metrics`. ## Demo performance comparions on calculating the histogram : 

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：