Comments
2 comments
-
You can only do accurate *absolute* timings when the profiler is in low-overhead mode. This discounts using line-level profiling. Can you repeat the experiment with sampling mode? Also, wall-clock times are going to be heavily influenced by external latencies. Try CPU timing first - again if its comparing *absolute* time you are interested in. -
Chris.Allen wrote:You can only do accurate *absolute* timings when the profiler is in low-overhead mode. This discounts using line-level profiling. Can you repeat the experiment with sampling mode? Also, wall-clock times are going to be heavily influenced by external latencies. Try CPU timing first - again if its comparing *absolute* time you are interested in.
Thank you for the information: I'll try it with those settings and get back to you if I am still confused.
Add comment
Please sign in to leave a comment.
The testing is isolated on a development workstation and all external devices; [SQL Server, etc.] are the same.
When comparing wall-clock time for the top level process in a given call stack I am do not understand the results I am seeing.
Application A shows a time of 250.764 and application B shows 229.300.
In this scenario I would expect that application B would complete the task quicker than application A.
When run outside the profiler application A completes the task in 70% of the time taken by application B.
Any help in explaining what I am misunderstanding appreciated.