It writes to a GPU array named "vals" at its own thread index. ... CUDA Performance: Use Big Arrays! ... memcpy size 4194304: 0.38 ns/float (1574.1 us)
確定! 回上一頁