Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!iuvax!cica!tut.cis.ohio-state.edu!gem.mps.ohio-state.edu!ginosko!uunet!iconsys!caeco!i-core!geo-works!bryan
From: bryan@geo-works.UUCP (Bryan Ford)
Newsgroups: comp.sys.amiga
Subject: Re: Software-only network protocol standards
Message-ID: <1789.AA1789@geo-works>
Date: 19 Jul 89 10:17:54 GMT
References: <1089@sas.UUCP> <1719.AA1719@geo-works> <18559@usc.edu>
Followup-To: comp.sys.amiga
Lines: 58

In article <18559@usc.edu>, papa@pollux.usc.edu (Marco Papa) writes:
>In article <1719.AA1719@geo-works> bryan@geo-works.UUCP (Bryan Ford) writes:
>>DNET is written completely in C, and doesn't have enough OSI layering built
>>into the code to make it easy to convert part to assembly language.  When
>>you're running at over 200KBPS, something in C just can't cut it.
>
>Oh, yea? What do you think is the language in which UNIX device drivers (serial
>or network, take your pick), TCP/IP, sockets, etc ... have been written in?
>Hint: the name of the language is one-letter long :-) And I can tell you that 
>that code supports data transfers > 200KBPS.  Don't blame the language. You got
>the wrong culprit.

Maybe you're right, and I have overestimated the difference in speed of C
and assembly language.  In any case, let's not get into a 'C vs Assembly'
war.  However, consider this:

This network has to run on a "lowest-level" 8 MHz Amiga (or is that 7.xx?).
This means I must program for 125 nanoseconds per clock cycle, and no 68000
instruction takes less than two cycles (maybe more, I can't remember).  I'm
assuming that most Unix boxes run at much higher speeds, and since they
inherently *require* at least a 68020 (if memory protection is supported),
many of the instructions will also execute in less clock cycles (the 68020
often uses less clock cycles per instruction).  At 200,000 baud, you have 5
milliseconds per bit and 50 milliseconds per byte (at 8N1).  That means
there are approximately 400 clock cycles per byte.  This may seem like a
lot, but remember that, in any single host, there at least *four* processes
going at this speed:  one sending and one receiving for each of the two
required ports.  And if there are more than two ports, the overhead is even
greater.  Since this is a multitasking machine, I don't want to bog down
the system too much because of network handling.

Only the raw transfer of raw data is handled in hardware in this case,
while (for example) most Ethernet cards often handle things like error
checking and resending bad packets.  In addition, with an Ethernet network,
packets go straight to the destination host without needing to be processed
and rerouted by other hosts.  In a mesh network, many hosts will have lots
of data going through them, so there will be more traffic through each
host.

What I'll probably do is implement the network and data-link layers in
assembly language, and do the rest in C.  This way, only the end source and
destination hosts of packets use C, while packets which just go 'through' a
host never get to the C part (the Network layer rerouts them first.)

Now, one question:  How many overhead cycles do the Amiga's interrupt
routines use (on a plain 68000)?  I know it's not much, but I'd like to
know exactly *how* much.  How about for software interrupts?

Thanks!
	
				Bryan

--

     _______________________________________
   _/   Bryan Ford - bryan@geo-works.uucp   \_
 _/  ..!utah-cs!caeco!i-core!geo-works!bryan  \_
/ ..!uunet!iconsys!caeco!i-core!geo-works!bryan \