[MERGE] Mystery of progress is solved
jelmer at samba.org
Sat Nov 4 18:07:45 GMT 2006
On Fri, 2006-11-03 at 17:34 -0600, John Arbash Meinel wrote:
> I'm not sure how to test this, so for now, I'm just going to ask that we
> merge this patch.
> Basically, I was investigating why progress bars take so long to start
> up, because I can see that the updates are happening. It took me a while
> to remember, but then I remembered... 'time.clock()' measures CPU time
> on Linux platforms (on Win32 it measures high-resolution wall-clock time).
> So what is happening is that we are spending all of our time in Suspend
> mode because of latency, and time.clock() isn't incrementing at all. So
> when the progress bar shows up, we have spent 1 second of *processor*
> time, and likely much much more than that in wall time.
> The attached patch reverts the code back to the old 'time.time()'
> values. Now, I know one of the reasons we switched to time.clock() was
> because we were trying to decrease the overhead of calling time.time().
> But I think we just need to live with it. Because this was just broken.
> This makes a *lot* of remote operations feel snappier. Because it
> doesn't just sit there with no information.
This does indeed seem to have a significant impact when cloning remote
+1 from me. Are there guidelines on how much +1's are required (and from
whom) before submitting to pqm?
Jelmer Vernooij <jelmer at samba.org> - http://samba.org/~jelmer/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 189 bytes
Desc: This is a digitally signed message part
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20061104/4fa198ea/attachment.pgp
More information about the bazaar