cuda编程求助
a gpu architecture has 8 threads per warp,uses pipline alus requiring 5 circle per operation at frequency of 800 mhz,and has 10 processors with 640 registor words.
a:i:whati is the peak performance of gpu in alu operations per seconds;
ii:what is the mininum of the threads per blocks to fully use alu;
iii:a given kernel requires 7 registers-what is the maxnum block size that could be supported;
iiii:a differ kernal requries 13 registers and 100000000 alus operations per thread,contains no brances,and is launched using 32 threads per blocks ,and 27 blocks per grid. provide a low bound on execution time of the grid on gpu.