Performance becomes pretty bad with this time profiling default on.
This should be due to err_msgprintf which is used to print profile
output along with input command string.
I wrote this patch as an aid to developer working on performance
intensive project, to see which part of code is taking what time
etc. I do not think, this should be default on. We can however make user
timer to get timing information. Developer working
on performance related projects, should turn it on themselves. (Maybe
./configure can be modified to take extra param for this).
Hi,
> Please ask monty (or Lee) to set you up with hudson param build to see how
> this patch does with performance regression on the benchmarks.
I ran few sysbench related test to see how it performs with this patch
Here are results.
With time profiling default on /lists. launchpad. net/drizzle- benchmark/ msg03728. html
https:/
Without time profiling default off /lists. launchpad. net/drizzle- benchmark/ msg03729. html
https:/
Performance becomes pretty bad with this time profiling default on.
This should be due to err_msgprintf which is used to print profile
output along with input command string.
I wrote this patch as an aid to developer working on performance
intensive project, to see which part of code is taking what time
etc. I do not think, this should be default on. We can however make user
timer to get timing information. Developer working
on performance related projects, should turn it on themselves. (Maybe
./configure can be modified to take extra param for this).
Let me know what you think
--
Thanks
Prafulla