Created
September 14, 2021 01:22
-
-
Save alanvgreen/53420020cfbe979ad6835eed75015254 to your computer and use it in GitHub Desktop.
Better filter loading
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
google/CFU-Playground at 75aabdfa | |
Summary: obvious priority is to move filter as far outside loop as possible | |
# Measure Main loop | |
Counter | Total | Starts | Average | Raw | |
---------+--------+--------+---------+-------------- | |
0 | 1475M | 27 | 55M | 1474982499 | |
1 | 0 | 0 | n/a | 0 | |
2 | 0 | 0 | n/a | 0 | |
3 | 0 | 0 | n/a | 0 | |
4 | 0 | 0 | n/a | 0 | |
5 | 0 | 0 | n/a | 0 | |
6 | 0 | 0 | n/a | 0 | |
7 | 0 | 0 | n/a | 0 | |
1495M ( 1494824995) cycles total | |
Profiler reports 1484M for 2D conv, so 9M cycles are not measured here - about 300K/op | |
# Detailed Measurements | |
Counter | Total | Starts | Average | Raw | |
---------+--------+------- -+---------+-------------- | |
0 | 1501M | 27 | 56M | 1501428675 | |
1 | 789M | 31898 | 25k | 789199190 | |
2 | 208k | 18 | 12k | 208038 | |
3 | 40M | 69566 | 571 | 39726988 | |
4 | 90M | 401808 | 64 | 90145055 | |
5 | 0 | 0 | n/a | 0 | |
6 | 0 | 0 | n/a | 0 | |
7 | 0 | 0 | n/a | 0 | |
1521M ( 1521252055) cycles total | |
Counter 0 - Overall | |
Counter 1 - filter loads inside main loop > 50% of overall | |
counter 2 - filter loads outside main loop | |
Counter 3 - input loads | |
Counter 4 - post processing | |
With the change | |
Counter | Total | Starts | Average | Raw | |
---------+--------+--------+---------+-------------- | |
0 | 704M | 27 | 26M | 703967503 | |
1 | 1383k | 62 | 22k | 1383146 | |
2 | 0 | 0 | n/a | 0 | |
3 | 68M | 92846 | 733 | 68074384 | |
4 | 83M | 1401808 | 59 | 83122496 | |
5 | 0 | 0 | n/a | 0 | |
6 | 0 | 0 | n/a | 0 | |
7 | 0 | 0 | n/a | 0 | |
Counter 0 - Overall | |
Counter 1 - total filter loads | |
Counter 3 - input loads | |
Counter 4 - post processing |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment