Page 9 of 9

Re: uip-tool

Posted: Sun Apr 28, 2019 10:04 pm
by sqward
czietz: I don't really understand why you'd like to convert it to gas as installing vasm takes no time at all. But if you feel strongly about that then I'd rather not loose a contributor over that.

Let's move it out of this thread.

Re: uip-tool

Posted: Mon Apr 29, 2019 6:53 pm
by czietz
Another question: I assume the binaries you release are build with libcmini for size, right? Which version of libcmini do you use?

In my builds I have some issues when linking with libcmini, e.g., files are only partially downloaded (the Content-Length header is already wrong). Everything works when using MiNTLib, but of course the binary is bigger then.

Re: uip-tool

Posted: Mon Apr 29, 2019 6:55 pm
by sqward
I have some patches to libcmini I yet have to upstream. In the meantime just don't use it :)

Re: uip-tool

Posted: Mon Apr 29, 2019 7:49 pm
by czietz
Looks like I found one them myself: ;-)
https://github.com/mfro0/libcmini/issues/32

Re: uip-tool

Posted: Fri May 17, 2019 9:21 pm
by jonord
Just got around to test the latest version on my Falcon. Awesome. Great work! I experienced that the old version was just moderately faster than using Ghostlink over serial and 115200 baud. Now it's blindingly fast!

Re: uip-tool

Posted: Fri May 17, 2019 9:36 pm
by sqward
Surely old version wasn't that slow? ;)

Also, bear in mind that if you're using USB drivers then that's at least 30% penality on uiptool performance. Not sure what's going on in the USB driver but even if you're not touching your mouse it slows the system quite badly. It would be great if that could be fixed somehow.

Re: uip-tool

Posted: Sat May 18, 2019 2:09 am
by Perdrix24
sqward wrote:Surely old version wasn't that slow? ;)

Also, bear in mind that if you're using USB drivers then that's at least 30% penality on uiptool performance. Not sure what's going on in the USB driver but even if you're not touching your mouse it slows the system quite badly. It would be great if that could be fixed somehow.


It may slow down the transfer rate, but surely not the CPU time that much?

Also Sting and MagxNet conflict with the USB storage driver, when present and configured, file transfers abort then the whole USB susbsystem goes down. There is surely a conflict that needs investigating.

Uip-tool is the only networking I can reliably use with the USB drivers under NetUSBee.

Re: uip-tool

Posted: Sat May 18, 2019 5:43 am
by jonord
Mind you it was quite some time since I last used Uip-tool, probably the very first versions or so. And with the proper serial drivers Ghostlink isn't that bad, plus it's very convenient to use on the Atari as it shows up as an extra device.

Funnily enough the Ghostlink device shows up in the web browser so I could if I download a file from the PC via the Atari.... :lol:

Re: uip-tool

Posted: Sat May 18, 2019 4:19 pm
by sqward
Perdrix24 wrote:It may slow down the transfer rate, but surely not the CPU time that much?


Not sure what you mean. The only way to slow down transfere rate is to eat the CPU time. Not sure what the USB driver is doing but I'd guess it hooks itself up to VBL or 200Hz timer to periodically poll USB hardware.

Re: uip-tool

Posted: Sun May 19, 2019 8:22 am
by susher
The USB interface is quite CPU intensive as there's no hardware off-load and it's a very simple/cheap system. This is why it won over Firewire, it was slower, slowed the machine (as the CPU had to do all the work) but was cheap.

Re: uip-tool

Posted: Thu May 23, 2019 7:38 pm
by Perdrix24
sqward wrote:
Perdrix24 wrote:It may slow down the transfer rate, but surely not the CPU time that much?


Not sure what you mean. The only way to slow down transfere rate is to eat the CPU time. Not sure what the USB driver is doing but I'd guess it hooks itself up to VBL or 200Hz timer to periodically poll USB hardware.


I did some benchmarks, on 68000 the mouse driver takes 10% of the CPU time, while on the 68030 Falcon, it's 5%.

The worse case scenario is having the mouse and keyboard drivers, in addition to usbtool.acc, on a 68000, that resulted in a 21% slowdown of the CPU.

Re: uip-tool

Posted: Thu May 23, 2019 7:58 pm
by sqward
Thanks, that's in line with my observation.