Path: utzoo!mnetor!uunet!husc6!bloom-beacon!gatech!hubcap!ncrcae!ncr-sd!crash!pnet01!haitex
From: haitex@pnet01.cts.com (Wade Bickel)
Newsgroups: comp.sys.amiga
Subject: Re: Practical task switch and software interrupt switching speed
Message-ID: <2742@crash.cts.com>
Date: 29 Mar 88 03:16:12 GMT
Sender: news@crash.cts.com
Organization: People-Net [pnet01], El Cajon CA
Lines: 51

dillon@CORY.BERKELEY.EDU (Matt Dillon) writes:
>
>	Here are some timing tests on the relative speed of message
>passing and software interrupt overhead.   The code is included.
>(It's too small to bother the moderators).
>
>						-Matt
>
>
>TEST 1:
>	Given two tasks running at priority 0 (the priority doesn't
>	really effect the results) with no other applications running,
>	how fast can these two tasks bandy a message back and forth?
>
>	Each cycle is defined by:  A-(msg)->B, B-(msg)->A 
>
>	results:	1000 cycles / second 
>
>TEST 2:
>	Given a task running at priority 1 (again, the priority could
>	just as well be 0), with no other applications running,
>	how fast can this task bandy a message to a software interrupt
>	handler (e.g. the port type is SOFTINT and the handler replies
>	the message):
>
>	Each cycle is defined by: A-(msg)->SoftHndlr, SoftHndlr-(msg)->A
>
>	results:	1736 cycles / second
>
>TEST Conditions:
>
>	The test programs are written in C.  They use standard calls for
>	everything.  NOTE: The software interrupt handler is also written
>	in C, and thus has a bit more overhead than one would normally
>	associate with a software interrupt.  However, the task-switch
>	tester has no more overhead than equivalent assembly.

        In my interrupt server I have been useing shared memory with a few
conventions applied to prevent conflicts.  Seems to me that 1700 cycles is
at least two orders of magnitude smaller than the possble communications rate
I can achieve.  Are you sure this is correct?


                                        Thanks,

                                                        Wade.


UUCP: {cbosgd, hplabs!hp-sdd, sdcsvax, nosc}!crash!pnet01!haitex
ARPA: crash!pnet01!haitex@nosc.mil
INET: haitex@pnet01.CTS.COM