von a8w » Mi 16. Mär 2016, 06:36
dl7ukk hat geschrieben:Geht es denn auch ohne (Hardware-) Handshake? Nur mit Software- Handshake ?
When hardware handshaking is used (DSR/RI or CTS) performance is much worse compared to "NO handshake". AspeQt does not have a "Software Handshake". "NO handshake" simply monitors the data line (SIO5 from Atari) and acts upon bytes arriving to the port. For Bluetooth devices the speed is limited to 19,200 bps, for wired SIO2PC devices that provide no hardware hanshaking, the speed can go up to 125,000 bps.
So the problem is worse when a device with hardware handshake is used. I've done a test tonight and the results surprised me:
MY PC SETUP: HP Netbook with 1.6Ghz ATOM CPU, with Ubuntu 15.04, and 1GB of RAM
SIO2PC SETUP: My own SIO2PC/10502PC Dual-USB v3.1
I ran RWTEST.COM with AspeQt set for 125,000 bps (divisor 0) SIO Speed:
WITH "DSR" handshaking, I had lots of speed adjustments (stutters) and some I/O errors. RWTEST result is 1,682 KBs and CPU load was 25%.
WITH NO Handshaking, I had some speed adjustments (stutters) and 1 I/O error. RWTEST result is 4,179 KBs and CPU load was 40%.
So, as you see, even with almost twice as much CPU load "NO handshake" performed better. The reason for higher CPU load was due to constant polling of the DATA IN line by AspeQt. So my conclusion is, the speed problem in Linux is not due to CPU load, but perhaps related to driver issues or some other factor, which needs more detailed investigation.