[hpsdr] VMs and USB2

[Bob Cowdery]

Hi Bill

I'm using 2nd Feb at the moment so I guess that doesn't have the extra
thread. I'm going to do some experiments and see if I can figure out
what the problem is - no guarantee - I'm not good at this low level
stuff!

Regards
Bob
G3UKB

On Fri, 2009-02-06 at 00:18 -0600, Bill Tracey wrote:
> Hi Bob -- disappointed to hear of your VMWare issue -- I was hoping 
> to bring up a big quad core box with VMWare running different OS's to 
> play HPSDR on.
> 
> As to what may be the cause of the problem -- latency and  perhaps 
> turnaround time.  For 192khz our inbound data rate is about 12.4 
> Mbits/sec (192ksamp/sec * 64 bits/sample) + 1.6% overheard for sync 
> and command and control (8/512).  Outbound the data rate is about 3.1 
> Mbits/sec, so it would appear we should have plenty of bandwidth.
> 
> There is another possibility -- if you're using the 4 Feb PowerSDR 
> code I've got a thread reading the spectrum data from EP4.  At the 
> moment that thread is reading data from EP4 as fast as it can be 
> delivered.  I don't think there's all that much data coming up the 
> EP4 pipe, but I've not measured it yet.  If you did have the 4 Feb 
> drop I can build one with the EP4 stuff turned off and see if that 
> makes any difference.
> 
> Regards,
> 
> Bill (kd5tfd)
> 
> 
> At 04:23 PM 2/5/2009, Bob Cowdery wrote:
> >***** High Performance Software Defined Radio Discussion List *****
> >
> >I should have done these very simple tests before I started but
> >hindsight is a wonderful thing.
> >
> >I copied a 288MB file from a USB stick to my hard drive. All systems are
> >reporting a USB device at 480Mbps (although it's hard to get info out of
> >XP).
> >
> >On my host Ubuntu system this takes 12 seconds.
> >On the VM's both Ubuntu and XP it took 45 seconds.
> >
> >This is 4 x slower so it should equate to about 120Mbps. In practise we
> >seem to be going about 1/2 speed anyway accounting for protocol, the
> >stick itself and disc write times I suppose that's reasonable.
> >
> >Why would this not be way more than is needed to transfer 48k at 24 bits
> >and 48k back at 16 bits. What am I missing here?
> >
> >73
> >Bob
> >G3UKB
> 
> 


 1233914604.0