Re: svn_client_proplist slow to get all of one dir

Here are the test measurements comparing recurse svn_client_proplist
with single file svn_client_proplist.
On all platforms there is an order of magnitude performance penalty
using single file vs recurse.
(I cannot use recursive in a GUI as svn will drill down into a huge
tree of files when not an a leaf node).

Why is one call to svn_client_proplist that returns details on 54 files
10 times faster then 54 calls to get
info on single files?

I have update the test program to allow testing of svn_client_proplist
with recurse=TRUE
Then I do two tests:

Test 1) svn_client_proplist recurse=TRUE ../../Workbench/Source
(contains 54 files and 1 dir with 1 file)
Test 2) svn_client_proplist recurse=FALSE
../../Workbench/Source/wb_main.py

I run the program in the Patches directory with the command line

----------------------------- Linux ----------------------------
Running on a 600MHz Mandrake 9.2 system I get:

test_propval ../../Workbench/Source 1 1

         Time for 1 calls 22ms recurse=1
         Time for 1 call 22ms recurse=1

test_propval ../../Workbench/Source/wb_main.py 54 0

         Time for 1 calls 506ms recurse=0
         Time for 1 call 9ms recurse=0

----------------------- Window AV enabled -----------------
Running on a 1GHz Windows XP SP1 systems with Norton
anti-virus enabled:

test_propval ..\..\Workbench/Source 1 1

         Time for 1 calls 60ms recurse=1
         Time for 1 call 60ms recurse=1

test_propval ..\..\Workbench/Source/wb_main.py 54 0

         Time for 1 calls 641ms recurse=0
         Time for 1 call 11ms recurse=0

----------------------- Window AV disabled -----------------
Running on a 1GHz Windows XP SP1 systems with Norton
anti-virus disabled:

test_propval ..\..\Workbench/Source 1 1

         Time for 1 calls 30ms recurse=1
         Time for 1 call 30ms recurse=1

test_propval ..\..\Workbench/Source/wb_main.py 54 0

         Time for 1 calls 370ms recurse=0
         Time for 1 call 6ms recurse=0

----------------------------------------------------------------------

On Oct 30, 2004, at 15:50, Ben Collins-Sussman wrote:

>
> On Oct 30, 2004, at 5:53 AM, Barry Scott wrote:
>> Why is the windows performance so poor?
>>
>
> As has been discussed on this list many times before, win32 disk i/o
> is well-known to be pretty slow, much worse than unix. It's
> especially bad at reading/writing lots of tiny files, which is what
> the .svn/ area is all about. So your test results are no surprise to
> me.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sun Oct 31 13:03:05 2004