Hi Patrick, some comments in line below.
On 02/03/2017 14:24, f1...@free.fr wrote: > The following is only to help the team, so do not see any negative idea in it. > > I am using an Intel Core 2 Duo CPU (E8400) at 3GHz, this is not the latest > I7, but not too old... Not an AMD anyway. The Core 2 Duo does not have Intel Hyper-threading which means that it only supports two concurrent threads of execution, one per core. This does mean that the architecture of WSJT-X, which uses three separate threads for the user interface, the audio synthesis and capture and the decoding, will have to share one CPU thread at times. Basically three tasks sharing two threads of execution. In theory the user interface thread is not very busy other than updating the waterfall so that should allow one of the other tasks to share reasonably well with it. One thing is important, when running MSK144 on a legacy system, you should take care to stop any other CPU intensive software. E.g. a logging program is fine but video streaming would probably have a negative impact. > > FYI my test conditions are : > - WSJT-X : r7405 > - FTol = 200Hz > - Decode : DEEP. > > With these conditions, the CPU load is around 40% so I have the feeling that > I have plenty. But, I wanted to try George's recommandations, so I reduced > the FTol to 100Hz. > Now, the CPU load dropped to 32%, but the graphics fills up to 30s. Reducing the decoding depth only has a minor impact on decoding sensitivity, it eliminates some compute intensive tries at decoding that might yield a successful decode on weaker signals. > > I have never tried to build from sources in Windows, but I did it with Linux > (quite a while ago). So do you think that this problem could be solved by > using Linux ? Or is there a newer version available under Linux that could > overcome this problem ? The platform you run on has little impact although MSK144 does have some technical restrictions on Mac OS X that limit its performance, these are due to tool chain restrictions. I would not expect any detectable performance difference between Windows and Linux on equal hardware. Linux probably has the edge on application performance due to the operating system and services normally being less needy of CPU and other resources, but not enough difference to switch operating systems just for performance. > > Another question : using FTol = 100Hz does not leave much room for the > transceiver frequency error (I am using a FT-2000 on 6m so it should be ok). See my comment above about decoding depth. Obviously the FTol setting depends upon both your frequency accuracy and that of your QSO partners. As most rigs are pretty accurate and stable these days, starting with a narrow FTol and perhaps widening it if pings are there but not decoded is a good policy if CPU capacity is limited. 73 Bill G4WJS. ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, SlashDot.org! http://sdm.link/slashdot _______________________________________________ wsjt-devel mailing list wsjt-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wsjt-devel
