home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.lang.ada
- Path: sparky!uunet!wupost!sdd.hp.com!zaphod.mps.ohio-state.edu!think.com!linus!linus.mitre.org!mbunix.mitre.org!vtg
- From: vtg@mbunix.mitre.org (Giddings)
- Subject: Re: 1 millisecond tasking on a Mips
- Message-ID: <1992Jul21.174153.9722@linus.mitre.org>
- Followup-To: vtg@mitre.org
- Sender: vtg@mitre.org (Victor Giddings)
- Nntp-Posting-Host: mbunix.mitre.org
- Organization: The MITRE Corporation, Bedford, MA
- References: <ELLIOTTM.92Jul17225954@beach.csulb.edu>
- Date: Tue, 21 Jul 1992 17:41:53 GMT
- Lines: 50
-
- In article <ELLIOTTM.92Jul17225954@beach.csulb.edu>
- elliottm@beach.csulb.edu (Elliot Michael) writes:
- >We are considering an embedded application in Ada, running on a Mips
- >chip (speed yet unspecified) which will require at least one task to
- >be run at 1 KHz. Is this reasonable? I've heard lots of rumors about
- >how much overhead is involved in Ada tasking.
- >
- >
- >--
- >
- >
- > Mike Elliott
- > elliottm@beach.csulb.edu
-
- Yes, but be careful. (What a surprise!) I offer the following PIWG
- benchmark results:
- Test Description (abreviated) CPU Time Wall Time
- T000001 Minimum rendezvous, task ocedure 60.039 60.039
- T000002 One task, 1 entry, in package 59.961 59.961
- T000003 2 Tasks, 1 entry, in package 60.703 60.703
- T000004 1 Task, 2 entries, in package 70.976 70.976
- T000005 10 tasks, 1 entry, in package 63.562 63.562
- T000006 1 task, 10 entries, in package 81.500 81.500
- T000007 Minimum rendezvous 42.773 42.773
- T000008 Pass Int: Producer->Buffer->Consumer 179.687 179.687
- The time units are microseconds! These measurements were taken from
- a MIPS R3200 board (25MHz R3000, 64kB I cache, 64kB D cache, 1MB
- memory) using the MIPS ASAPP (Ada StandAlone Programmer's Product)
- Version 3.0 (derived from Verdix). Unfortunately, this compiler
- and run-time system are no longer available. About 6 months ago,
- we completed an evaluation of competing products from Verdix, Alsys,
- Telesoft, and InterAct. Our agreements with the vendors prevent me
- from discussing specifics and 6 months of additional development
- from the vendors may invalid our results, but let me assure you that
- similar and sometimes better performances are attainable from
- currently available compilers. Also, you can get 40MHZ clock rates
- in hardware.
-
- If you can use passive tasks, you can do even better. The following
- results are MODIFICATIONS of the PIWGs to use passive tasks. (Most
- modifications were to simply add pragma Passive.)
-
- Test CPU Time
- T000002P 6.226
- T000003P 6.938
- T000004P 7.246
- T000005P 6.203
- T000006P 7.875
- T000008P 25.723
-
-