[Info-vax] VSI: "Official 8.4-1H1 Launch"

[Chris Scheers]

David Froble wrote:
> Chris Scheers wrote:
>> Bob Koehler wrote:
>>> In article <55734365$0$32453$c3e8da3$cc4fe22d at news.astraweb.com>, JF 
>>> Mezei <jfmezei.spamnot at vaxination.ca> writes:
>>>> When working on the early boot sequence/stages, I doubt that 
>>>> performance
>>>> really matters.  And if most operating systems can run just fin on a 
>>>> VM,
>>>> it means that the VM gives it adequate CPY time slots.
>>>>
>>>> Also, remember that VMS ran on the All Mighty Microvax II, where
>>>> performance wasn't measure in instructions per second, but seconds per
>>>> instruction :-)  Pretty sure a VM can offer better performance overall
>>>> than an MV II.
>>>
>>>    Yes, but a VM can intererfe with hard real-time to the extent that
>>>    your processor needs to be 1000 times faster.  You can probably get
>>>    processors 1000 times faster than an MV II now, but you can also
>>>    get VM interference tht's worse.
>>
>> Well, yes and no.
>>
>> Theoretically, in a VM, you actually run the native instruction set at 
>> full speed.  You only have issues if you need to enter/exit the 
>> hypervisor.
>>
>> Depending on how your real-time code is written, you may be able to 
>> stay out of the hypervisor, especially if you have enough cores 
>> available to dedicate one or more to your task.
>>
>> With an emulator, every instruction may be emulated and the factor of 
>> 1000 comes a lot closer.
>>
>> I do hard real-time conversions of Data General systems to an 
>> emulator, sometimes running under a VM, so I've seen these issues.
>>
>> It seems that each real-time system is unique and quite often requires 
>> special tuning, but it can be done.
>>
>> Interestingly, its not uncommon for the problem to be an emulation 
>> that runs too fast.
>>
> 
> Strictly curiosity.  Would that problem be rather easy to solve?  You 
> could say "do this, but don't give me the result for at least this 
> amount of time".  I know I'm being massively simplistic.
> 
> Not sure how much timer ASTs would interfere with real time.

The problems are on the order of: I need this result in 10us, not 15us, 
not 5us.

ASTs and OS scheduling are much to granular for some real-time problems.

But this has gotten a bit off topic.  Real-time solutions get to be very 
special purpose.

To get back on topic, ignoring real-time needs, I can see huge 
advantages to using a VM for porting an OS.

This gives you access to some debugging tools you don't have with bare 
metal.

Generally, there is also a simplified driver environment with "perfect" 
hardware that lets you cut corners on initial error handling.

-- 
-----------------------------------------------------------------------
Chris Scheers, Applied Synergy, Inc.

Voice: 817-237-3360            Internet: chris at applied-synergy.com
   Fax: 817-237-3074