How to get correct global variable with jupyter's timeit?
up vote
0
down vote
favorite
One first cell I have this:
from numba import cuda
@cuda.jit
def thread_counter_safe(global_counter):
cuda.atomic.add(global_counter, 0, 1) # Safely add 1 to offset 0 in global_counter array
On the next cell I have this:
global_counter = cuda.to_device(np.array([0], dtype=np.int32))
thread_counter_safe[64, 64](global_counter)
print('Should be %d:' % (64*64), global_counter.copy_to_host())
global_counter = cuda.to_device(np.array([0], dtype=np.int32))
%timeit thread_counter_safe[64, 64](global_counter)
print('Should be %d:' % (64*64), global_counter.copy_to_host())
Output of this second cell is something like this:
Should be 4096: [4096]
10000 loops, best of 3: 118 µs per loop
Should be 4096: [168390656]
Jupyter Notebook's timeit carries the global_counter accross its iteration test. How do it get it to give back global_counter correctly?
jupyter-notebook numba
edited Nov 9 at 15:10
talonmies
59k17128195
asked Nov 9 at 10:58
biocyberman
2,54922135
add a comment |
up vote
0
down vote
favorite
One first cell I have this:
from numba import cuda
@cuda.jit
def thread_counter_safe(global_counter):
cuda.atomic.add(global_counter, 0, 1) # Safely add 1 to offset 0 in global_counter array
On the next cell I have this:
global_counter = cuda.to_device(np.array([0], dtype=np.int32))
thread_counter_safe[64, 64](global_counter)
print('Should be %d:' % (64*64), global_counter.copy_to_host())
global_counter = cuda.to_device(np.array([0], dtype=np.int32))
%timeit thread_counter_safe[64, 64](global_counter)
print('Should be %d:' % (64*64), global_counter.copy_to_host())
Output of this second cell is something like this:
Should be 4096: [4096]
10000 loops, best of 3: 118 µs per loop
Should be 4096: [168390656]
Jupyter Notebook's timeit carries the global_counter accross its iteration test. How do it get it to give back global_counter correctly?
jupyter-notebook numba
edited Nov 9 at 15:10
talonmies
59k17128195
asked Nov 9 at 10:58
biocyberman
2,54922135
1
you would need to reset theglobal_counterto zero as part of the function that you are runningtimeiton. You can't safely/easily do that from thethread_counter_safecuda kernel, but you coulddefa new function that resets theglobal_counterto zero and then callsthread_counter_safe, and runtimeiton that.
– Robert Crovella
Nov 9 at 14:48
Got it. A wrapper function would do, but still prefer some flag or parameter when calling timeit.
– biocyberman
Nov 9 at 15:11
add a comment |
up vote
0
down vote
favorite
up vote
0
down vote
favorite
One first cell I have this:
from numba import cuda
@cuda.jit
def thread_counter_safe(global_counter):
cuda.atomic.add(global_counter, 0, 1) # Safely add 1 to offset 0 in global_counter array
On the next cell I have this:
global_counter = cuda.to_device(np.array([0], dtype=np.int32))
thread_counter_safe[64, 64](global_counter)
print('Should be %d:' % (64*64), global_counter.copy_to_host())
global_counter = cuda.to_device(np.array([0], dtype=np.int32))
%timeit thread_counter_safe[64, 64](global_counter)
print('Should be %d:' % (64*64), global_counter.copy_to_host())
Output of this second cell is something like this:
Should be 4096: [4096]
10000 loops, best of 3: 118 µs per loop
Should be 4096: [168390656]
Jupyter Notebook's timeit carries the global_counter accross its iteration test. How do it get it to give back global_counter correctly?
jupyter-notebook numba
edited Nov 9 at 15:10
talonmies
59k17128195
asked Nov 9 at 10:58
biocyberman
2,54922135
One first cell I have this:
from numba import cuda
@cuda.jit
def thread_counter_safe(global_counter):
cuda.atomic.add(global_counter, 0, 1) # Safely add 1 to offset 0 in global_counter array
On the next cell I have this:
global_counter = cuda.to_device(np.array([0], dtype=np.int32))
thread_counter_safe[64, 64](global_counter)
print('Should be %d:' % (64*64), global_counter.copy_to_host())
global_counter = cuda.to_device(np.array([0], dtype=np.int32))
%timeit thread_counter_safe[64, 64](global_counter)
print('Should be %d:' % (64*64), global_counter.copy_to_host())
Output of this second cell is something like this:
Should be 4096: [4096]
10000 loops, best of 3: 118 µs per loop
Should be 4096: [168390656]
Jupyter Notebook's timeit carries the global_counter accross its iteration test. How do it get it to give back global_counter correctly?
jupyter-notebook numba
jupyter-notebook numba
edited Nov 9 at 15:10
talonmies
59k17128195
asked Nov 9 at 10:58
biocyberman
2,54922135
edited Nov 9 at 15:10
talonmies
59k17128195
asked Nov 9 at 10:58
biocyberman
2,54922135
edited Nov 9 at 15:10
talonmies
59k17128195
edited Nov 9 at 15:10
talonmies
59k17128195
edited Nov 9 at 15:10
talonmies
59k17128195
59k17128195
asked Nov 9 at 10:58
biocyberman
2,54922135
asked Nov 9 at 10:58
biocyberman
2,54922135
asked Nov 9 at 10:58
biocyberman
2,54922135
2,54922135
1
you would need to reset theglobal_counterto zero as part of the function that you are runningtimeiton. You can't safely/easily do that from thethread_counter_safecuda kernel, but you coulddefa new function that resets theglobal_counterto zero and then callsthread_counter_safe, and runtimeiton that.
– Robert Crovella
Nov 9 at 14:48
Got it. A wrapper function would do, but still prefer some flag or parameter when calling timeit.
– biocyberman
Nov 9 at 15:11
add a comment |
1
you would need to reset theglobal_counterto zero as part of the function that you are runningtimeiton. You can't safely/easily do that from thethread_counter_safecuda kernel, but you coulddefa new function that resets theglobal_counterto zero and then callsthread_counter_safe, and runtimeiton that.
– Robert Crovella
Nov 9 at 14:48
Got it. A wrapper function would do, but still prefer some flag or parameter when calling timeit.
– biocyberman
Nov 9 at 15:11
1
1
you would need to reset the
global_counter to zero as part of the function that you are running timeiton. You can't safely/easily do that from the thread_counter_safe cuda kernel, but you could def a new function that resets the global_counter to zero and then calls thread_counter_safe, and run timeit on that.– Robert Crovella
Nov 9 at 14:48
you would need to reset the
global_counter to zero as part of the function that you are running timeiton. You can't safely/easily do that from the thread_counter_safe cuda kernel, but you could def a new function that resets the global_counter to zero and then calls thread_counter_safe, and run timeit on that.– Robert Crovella
Nov 9 at 14:48
Got it. A wrapper function would do, but still prefer some flag or parameter when calling timeit.
– biocyberman
Nov 9 at 15:11
Got it. A wrapper function would do, but still prefer some flag or parameter when calling timeit.
– biocyberman
Nov 9 at 15:11
add a comment |
active
oldest
votes
active
oldest
votes
active
oldest
votes
active
oldest
votes
active
oldest
votes
draft saved
draft discarded
Thanks for contributing an answer to Stack Overflow!
- Please be sure to answer the question. Provide details and share your research!
But avoid …
- Asking for help, clarification, or responding to other answers.
- Making statements based on opinion; back them up with references or personal experience.
To learn more, see our tips on writing great answers.
Some of your past answers have not been well-received, and you're in danger of being blocked from answering.
Please pay close attention to the following guidance:
- Please be sure to answer the question. Provide details and share your research!
But avoid …
- Asking for help, clarification, or responding to other answers.
- Making statements based on opinion; back them up with references or personal experience.
To learn more, see our tips on writing great answers.
draft saved
draft discarded
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function ()
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fstackoverflow.com%2fquestions%2f53224406%2fhow-to-get-correct-global-variable-with-jupyters-timeit%23new-answer', 'question_page');
);
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
1
you would need to reset the
global_counterto zero as part of the function that you are runningtimeiton. You can't safely/easily do that from thethread_counter_safecuda kernel, but you coulddefa new function that resets theglobal_counterto zero and then callsthread_counter_safe, and runtimeiton that.– Robert Crovella
Nov 9 at 14:48
Got it. A wrapper function would do, but still prefer some flag or parameter when calling timeit.
– biocyberman
Nov 9 at 15:11