ok

tetrarun.py

@@ -43,9 +43,9 @@ def main(argv):
    res = start_kernel(Cx, Cy, Dx, Dy, Dz, v, w)
    #printresults(res)
    print("Exact one 3-3")
-   print(exact_one_gpu(res, 3-1, 3-1))
+   exact_one33 = exact_one_gpu(res, 3-1, 3-1)
    print("Filter 2-1")
-   print(filter_gpu(res, 2-1, 1-1))
+   filter21 = filter_gpu(res, 2-1, 1-1)
 
 if __name__ == "__main__":
    main(sys.argv[1:])
\ No newline at end of file

utils.py

 import cupy as cp
 import numpy as np
+from functools import wraps
+import time
 
 with open('filtering.cu') as f:
     code = f.read()
@@ -9,6 +11,17 @@ ep_pontok_module = cp.RawModule(code=code, options=('--std=c++11',), name_expres
 exact_one_cuda = ep_pontok_module.get_function(kers[0])
 filter_cuda = ep_pontok_module.get_function(kers[1])
 
+def timeit(func):
+    @wraps(func)
+    def timeit_wrapper(*args, **kwargs):
+        start_time = time.perf_counter()
+        result = func(*args, **kwargs)
+        end_time = time.perf_counter()
+        total_time = end_time - start_time
+        print(f'Function {func.__name__}{args} {kwargs} Took {total_time:.4f} seconds')
+        return result
+    return timeit_wrapper
+
 def expSpace(min, max, N, exponentialliness = 20.0):
     LinVec = cp.linspace(0, cp.log10(exponentialliness+1, dtype=cp.float64),N, dtype=cp.float64)
     return (max-min)/exponentialliness * (10.0**LinVec - 1) + min
@@ -71,7 +84,6 @@ def exact_one_gpu(egyensulyi_mtx, S, U):
     exact_one_cuda((numBlock,), (256,), (egyensulyi_mtx, size, indexes, S, U))
     return egyensulyi_mtx[indexes]
 
-
 def filter_gpu(egyensulyi_mtx, S, U):
     size = int(egyensulyi_mtx.size / 16)
     indexes = cp.zeros((size,), dtype=cp.bool)