[Coco] WiFi modem.

[Gene Heskett]

On Saturday 20 January 2018 11:25:54 Kandur wrote:

> You are fantastic Gene,
> your help and valuble contribution over many years
> to our Coco community is very much appreciated.
> God bless you, keep on going strong!
>
> An other old dog, Kandur
>
Thanks for the flowers, Kandur. At 83, they are appreciated.

> Saturday, January 20, 2018, 2:09:05 AM, you wrote:
> > On Saturday 20 January 2018 04:12:37 David Ladd wrote:
> >> On Sat, Jan 20, 2018 at 2:12 AM, Gene Heskett <gheskett at shentel.net
> >
> > wrote:
> >> > On Saturday 20 January 2018 01:30:46 RETRO Innovations wrote:
> >> >
> >> >
> >> > ​<cut>
> >> >
> >> > So it boils down that I am not in a position to advise, just warn
> >> > that all facets need to be checked. With the speed limits of the
> >> > coco in handling an error correction protocol such as rzsz, flow
> >> > controls are a must when the baud rate goes above 240. Even less
> >> > if its multitasking heavily.

One of the things I was going to do to rzsz, was, since the newer table 
lookup crc routine can do a whole sectors crc in one call, the per 
character call should be deprecated in favor of a final call to do the 
whole 256 byte block. This would likely result in at least a 50% speed 
increase, and to prevent a uart overrun, an xoff should be sent before 
this single ending call, and an xon sent as the first byte of the crc 
checks acknowledge transaction.

That would still require that the linux "window" be setup as 256 bytes 
though just to inform the linux end to wait for that ack after sending 
256 bytes.  The linux version assumes both a cleaner path, and that it 
can double the packet size starting from 512 bytes everytime it gets a 
good ack.  That has to be prevented by asserting the window size option 
otherwise it will go on up to 8192 byte packets on a clean circuit. 
Short of a complete redesign of the coco version so if on a coco3, it 
can ask for and hopefully get, a full 8192 byte memory block just for 
the data, keeping the rest of its vars on the local stack reg.u points 
at. Bear in mind we're talking about a program thats already a bit over 
36k of source compiled, that I would consider a major rewrite. I was, by 
3.36, pretty well burned out and took a break because there seemed to be 
little interest by the group in any further development. That break 
turned into a 20 year hiatius and today I would have to first modularize 
it, breaking it up into more easily editable sized pieces.

But why bother? Drivewire is at 115k, over 100x faster.

After all, I took that project on just to show Paul Jerkatis that his CS 
101 prof was almost criminally wrong about a certain pair of assembler 
commands that could speed up bit shifting by doing 8 bits not in a loop, 
but 2 assembly commands that did a full 8 bit shift either way in just 2 
commands. The compiler of course doesn't know about that, so the output 
of cc2 has to be pulled into an editor, calls for bit shifting located, 
and if an 8 bit shift, sub those 2 commands and remove the shifter loop.  
All that of course had to be done by hand. That was just one of the 
tricks that doubled its speed from 3.12 to 3.36. The fact that I did it 
and the result didn't crash the machine apparently upset Paul so much he 
bought a Sun workstation and left. Shrug.

Now I'm an old fart, past my code carving prime, so now I carve g-code to 
run metal carving machines, its a lot simpler than remembering the whole 
litany of 6x09 assembler.

I have written 90 LOC of g-code that took 3 days to run on a toy mill, 
but at the end I had the sharpest carbide tipped blade on my table saw 
ever. I cleaned the cherry oils off it a couple times but its made 
furniture out of several thousand $ worth of cherry, mahogany, and 
butternut. And cut 1/8" to 1/2" thick alu panels for numerous bits and 
pieces to cnc control older machinery. My latest conversion is to a 70 
yo 11 by 36 inch Sheldon lathe, putting in ball screws to drive it, and 
stepper motors to turn the screws, how far and how fast all determined 
by an r-py-3b credit card sized computer running the debian jessie 
version of linux.

Between that, careing for my missus of 29 years who's dying of COPD, and 
doing the housework that falls to me because she can't, it keeps me out 
of the bars. :)

So yeah, I appreciate the flowers, thank you.

Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Genes Web page <http://geneslinuxbox.net:6309/gene>