[openfabrics-ewg] Re: Minor issues with resource tracking test

[Dotan Barak]

On Friday 26 May 2006 09:44, Sharma, Karun wrote:

> While testing, i observed that we are using -m option 2 times as per help output (see below). One is to specify "MR TYPE" and other is to specify "Multicast" option. This is quite misleading.  
You are absolutely right, this issue was fixed (the multicast short flag is now '-l').

> Moreover, -m option doesn't affect Multicast setting.  "-m YES" or "-m NO" doesn't switch on/off multicast. It execute the test with default MR Type and default Multicast setting.
You get this response because of the previous issue (same letter is being used for both options)

> If we specify -m to any other number from 6 onwards, we get an error saying "Bad option". But if we specify any alphabet to -m, the test executes with default settings. For eg. 
> "./resource_tracking_test -m 6" gives error but "./resource_tracking_test -m a" doesn't give any error.
There are only several MR types that are being supported by the test ([0..5]).  please review the readme.txt for more details.

> One more thing, please let me know if there is any upper limit to -x (number of threads) option. It fails if i specify -x 20. 
The value of X is multiplied by the number of objects, in your example 120 threads will be created in the machine, and everyone of them will try to create the number all the resources.
The limitations are the total number of threads that one process can open and the amount of memory that you have in your host.


Thanks
Dotan