Subject:
|
RE: Why java is (not) bad for Mindstorms
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Tue, 24 Jan 2006 08:31:27 GMT
|
Original-From:
|
Laurent Desnogues <laurent.desnogues@armNOMORESPAM.com>
|
Viewed:
|
1427 times
|
| |
|
|
In lugnet.robotics, pisymbol Clague wrote:
>
> I mean the NXT
> could already have a JIT type execution environment already
> whereby the
> graphical IDE generates pseudo-code ("bytecode") and the
> firmware on the NXT
> does the dynamic translation on the fly. I don't know, but
> the above *seems*
> reasonable to me. Moreover, future versions of the NXT
> compiler within the
> graphical IDE would be compatible with any platform changes
> (new JIT firmware).
>
> Right?
Wrong :) For two reasons:
- a JIT is quite complex if you want to really gain
something and requires too much memory (remember
the NXT only has 64 KB of RAM);
- JIT and real-time are not good friends...
On top of that, a carefully implemented bytecode interpreter
is not necessarily too slow. For instance, I have written
an ARM object code simulator which works by interpreting
opcodes and it simulates more than 40 million instructions
per second on a 2.4 GHz Opteron, that is about 60 host
cycles per simulated "bytecode". So if the NXT has some
low level bytecode to interpret and is running at about
16 MHz (wild guesses :) we could have an interpreter that
runs at several hundreds of thousands instructions per
second which should be enough for most tasks.
Of course, there are two assumptions here: that the
bytecode is low level and that the interpreter is fine
tuned for speed. But if the bytecode is not low level then
it means that each operation does complex tasks and so
compiling would not bring much anyway. This leaves us with
the potential problem of a too slow interpreter and to what
Steve Hassenplug talked about: a faster firmware.
Laurent
-- IMPORTANT NOTICE: The contents of this email and any attachments are
confidential and may also be privileged. If you are not the intended recipient,
please notify the sender immediately and do not disclose the contents to any
other person, use it for any purpose, or store or copy the information in any
medium. Thank you.
|
|
Message has 1 Reply:
 | | Re: Why java is (not) bad for Mindstorms
|
| (...) I believe there are JIT for J2ME environments which can be as low as 128KB of RAM. I'm not disagreeing with you, I'm just stating that the JIT technology is applicable to embedded systems and the argument of "its too complex" doesn't always (...) (18 years ago, 24-Jan-06, to lugnet.robotics)
|
2 Messages in This Thread:
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
