Open nikhilgupta10 opened 7 years ago
nikhilgupta10 imported these comments from Sourceforge: "dchassin": * status changed from new to accepted
The problem is related to the use of clock() as the basis for timing. The clock() function evaluate the cpu time used, not wall clock time elapsed. Recommend using a new exec_clock() function to evaluate elapsed wall clock time instead.
Another suggestion is to evaluate both times so that performance can be evaluated more carefully with respect to both wall clock and cpu time.
,
"dchassin": * owner changed from dchassin to jcfuller
The main problem is addressed by changeset:3487. We need to do manual inspection of profile reports to verify that this has resolved the problem. If the fix is confirmed, please reassign the ticket back to dchassin to address the second issue (outputting both cpu and wall time).
,
"dchassin": * priority changed from validate to minor
type changed from defect to validation ,
"jcfuller":- Description has changed:
Diff:
It looks like the profiler results for multithreading core runtimes are not correct.
Output for --profile -T 1 test_R3-12.47-3_NR.glm:
Output for --profile -T 2 test_R3-12.47-3_NR.glm:
,