Performance Evaluation of CPU and GPU in Matrix Transpose Operation

University:
University of Cambridge
Course:
ACC 241 | Accounting
Academic year:

2024
Views:

331

Pages:

2
Author:

Kamis Berhasil

Find and analyze the time difference between using CPU (Numpy) and GPU (Cupy) in performing matrix transpose operations. import time import numpy as np import cupy as cp stime.time() A np.random.randint(10, size (1000,1000)) C= A.transpose e time.time() print("CPU time:", es, "seconds") Then the output is: CPU time: 0.013174057006835938 seconds stime.time() A cp.random.randint(10, size (1000,1000)) C= A.transpose cp.cuda.Stream.null.synchronize() e= time.time() print("GPU Time:", e-s, "seconds") Then the output is: GPU time: 0.004245758056640625 seconds In the Numpy library, the time obtained for using the numpy matrix transpose operation with 1000 rows and 1000 columns is 0.013174057006835938 seconds. While in the Cupy library, the time obtained for the use of Cupy matrix transpose operation with 1000 rows and 1000 columns is 0.004245758056640625 seconds, much faster than CPU time. The use of GPU greatly affects the speed and performance of the operations we perform.