Path: utzoo!mnetor!uunet!husc6!hao!ames!pasteur!ucbvax!CORY.BERKELEY.EDU!dillon
From: dillon@CORY.BERKELEY.EDU (Matt Dillon)
Newsgroups: comp.sys.amiga
Subject: Re: CLtd (Was: 2090 SCSI driver needed)
Message-ID: <8803100838.AA13156@cory.Berkeley.EDU>
Date: 10 Mar 88 08:38:03 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Lines: 28

>AmigaTerm, written entirely in C, works at 9200 BAUD *without* flow control
>(this is faster than a VT101)  With flow control it works fine at 19.2KB.
> ..
>loosely on the baud rate.
> ..
>>losing characters.  Commodore says they can go MIDI speeds and still multi-task
>>...this may be true, but the driver's passing stuff to me at about 2400 baud
>>or less, depending on system loading.
>
>Only if you use it poorly.

	Yah, really.  Anybody who passes 1 byte read requests to the serial
device deserves what he gets!  What you *do* get with 1 byte read requests
is about 1500 baud on the read.  On the otherhand, with MIN(256,SCMD_QUERY)
byte requests, you get it at full speed.  Every program I know uses the
SCMD_QUERY method.

	Now, assuming C-A fixes (in 1.4?) the problems with Disabling for
too long within the kernel, the serial.device shouldn't have any further
problems getting overruns reading at high speeds.

	BTW: BUG in the serial.device ... you cannot have multiple 
outstanding CMD_WRITE requests.  There is a timing window in the 
serial.device (as far as I can tell) which can cause it to fail to begin
a CMD_WRITE and never return the request.  That is, you should have only one
asynchronous CMD_WRITE active at a time.  I found that one working on DNET.

			-Matt