[netperf-talk] RE: netperf reporting 0 bytes/s

[Rick Jones]

Srinivasan S wrote:
> Hi Rick,
> 
> I replaced in  send_tcp_stream() and in send_tcp_maerts() the following line 
> 
>           bytes_sent  = ntohd(tcp_stream_result->bytes_received);
> 
> with 
> 
>           bytes_sent  = tcp_stream_result->bytes_received; //removed ntohd
> 
> with this I am able to get what appears to be correct values. Just want a 
> confirmation, that the calculation would be theoritically correct.

For TCP, and the way the TCP_STREAM test works - with that shutdown;read;close 
bit at the "end" there in send_tcp_stream(), that should be OK.

It would _NOT_ be OK for say the UDP_STREAM test, but then it is already 
reporting both, and one has to expect that the sending side stats may be rather 
higher than the receiving side.

> 
> Thank you so much Rick.
> 
> -Srini
> 
> -----Original Message-----
> From: Rick Jones [mailto:rick.jones2 at hp.com]
> Sent: Monday, December 19, 2005 11:33 PM
> To: Srinivasan S; netperf-talk at netperf.org
> Subject: netperf reporting 0 bytes/s 
> 
> 
> Srinivasan S wrote:
> 
>>Hi Rick,
>>after adding few prints in netperf I observer some value in bytes_received
>>and in thruput. when printing in %llx format, but in %f it is all zeros. Is
>>it got to do any thing with endian and network to host conversions?
> 
> 
> Likely as not differences in the format of floatingpoint numbers.  Sadly, 
> netperf echanges floating point numbers because "way back when" there wasn't 
> much support for 64-bit integers.  Fixing your problem may require changing 
> netperf to move 64-bit integers around.
> 
> Short-term you can hack around the problem for the TCP tests at least - but NOT 
> the UDP tests - by using the netperf side stats rather than the netserver side 
> stats.
> 
> rick
> 
> I'm going to cc this to netperf-talk because it would be useful for that to be 
> in the record and perhaps it will help find someone willing and able to add the 
> 64-bit integer stuff to the messaging over the holidays :)
> _______________________________________________
> netperf-talk mailing list
> netperf-talk at netperf.org
> http://www.netperf.org/cgi-bin/mailman/listinfo/netperf-talk