On 2016-02-28 22:20 , Uwe Hermann wrote: > Hi, > > On Sun, Feb 21, 2016 at 08:35:26PM +0100, Henk Hendricks wrote: >> I was trying to continuously log some signals using an FX2 demo-board. >> And I noticed a big difference the number of samples recorded on Win7 >> and Fedora23, 250kS and 200MS respectively. >> >> This was at a sample frequency of 200kHz, using a sample frequency of >> 500kHz, around 650kS were received. >> Trying the same on a i3 Win7 laptop resulted in roughly the same the >> numbers. A 7inch Acer with an Atom cpu wasn't able to compete. >> Then I tried an old i7 desktop, it receives 410kS, and 1MS respectively. >> >> This was all recorded using a python script which repeatedly called >> sigrok-cli with a new filename. >> So this is where I thought the problem was maybe my python script, so I >> tried manually on the old i7. >> But manually was worse!, 320kS and 810kS respectively. >> Note that I just reinstalled sigrok-cli and the zadig driver. So I would >> think the new libusb patch would be included. > The most recent Windows installers we provide for sigrok-cli and > PulseView are fine, yeah. Both include the special libusb branch that's > currently required and an additional patch on top of that to improve > fx2lafw performance. > > >> So is 't possible to get the performance I got using Fedora 23 on Windows? > I don't think this specific issue is related to Windows here. It's more > likely to be a difference between sigrok-cli and PulseView in general.
> I opened bug #762 for this and attached some of my own measurements there. > > http://sigrok.org/bugzilla/show_bug.cgi?id=762 > > I haven't verified this, but my guess is that the difference is due to > PulseView storing samples directly in RAM (which is fast), while > sigrok-cli (with "-o foo.sr") will try to save every chunk into an .sr > file immediately when it arrives. This takes a lot longer due to file > access in general and compressing via libzip, so samples may be missed > eventually. > > If that's the case, we'll need to add some buffering (in RAM) so that no > matter how long the .sr save takes, samples will not be "lost" by the > hardware. All measurements I've done, were done using sigrok-cli, not using PulseView. So does this mean the difference between Fedora and Windows can than be explained by some caching/disk management differences? Just checked, and PulseView seems to work better than sigrok-cli, as you said. > >> Maybe some simpler questions: >> Why is a sample limit of 1T invalid using the command-line, but it is >> selectable in PulseView? > Good point, that's missing support in libsigrok. Filed a bug here: > http://sigrok.org/bugzilla/show_bug.cgi?id=763 > > The max number of samples (for devices where there's no limit) is pretty > arbitrary in PulseView, might as well add higher ones too, and most > certainly a "sample continously" option (I think there's a bug for that > already, haven't checked though). > > >> And is 't possible to 'continiously' measure (like my script) with >> sigrok-cli, i.e. without restarting sigrok and overhead associated with it? > Yep, use --continous instead of --samples or --time. I didn't know that parameter, but is would sure come in handy. But what I really meant by measure continuously, was that a long recording (like a day) will probably be interrupted a few times (i.e. Device only sent X samples errors). I don't mind gaps in the recording, but I want a recording for the whole duration. So I want sigrok-cli to restart automatically on a 'Device only sent X samples' error. And not come back after the duration, only to find a small partial recording. > Until the buffering mentioned above is implemented PulseView will > currently yield better results with fx2lafw devices, though. For quick measurements I shall use PulseView, and for the longer measurements I'll use Fedora in conjunction with my script. Many thanks and -- Greetings, Henk Hendricks ------------------------------------------------------------------------------ Site24x7 APM Insight: Get Deep Visibility into Application Performance APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month Monitor end-to-end web transactions and take corrective actions now Troubleshoot faster and improve end-user experience. Signup Now! http://pubads.g.doubleclick.net/gampad/clk?id=272487151&iu=/4140 _______________________________________________ sigrok-devel mailing list sigrok-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sigrok-devel
