On Nov 16, 2011, at 1:25 PM, Christian Schmitz wrote:

> 
> Am 16.11.2011 um 19:24 schrieb Michael Diehr:
> 
>> On the server side, the memoryblock returned by FileMappingView.create can't 
>> be set to an existing memoryblock (thus, requiring a copy operation)
> 
> Right. Some memory is mapped and returned as memory block.
> 
> I can extend that, so you can pass in a memoryblock which is used for this 
> avoiding the copy.

That would probably be nice, but not urgent.

>> On the client side, the "The memoryblock has no known size" - does this mean 
>> you can't use it for operations such as MemoryblockRGBAtoPictureMBS? and 
>> thus a 2nd copy of the memoryblock would need to be made?
> 
> The memory block here is a Ptr. It has an address, but Real Studio doesn't 
> know the size of the memoryblock.
> I don't think that MemoryblockRGBAtoPictureMBS needs the size property here.

I'll test.

> 
>> My worry is that if there are two extra copy operations (one on server end, 
>> one on client end) that the shared memory may not be any faster than IPC 
>> sockets.
> 
> 
> In general IPCSocket uses memory mapping behind the scences, so the question 
> is whether it's much faster to do it yourself.

I have the same concern.

Final question:  I have a XCode /Cocoa app which I want to talk to a RB app.   
I've already set up a cocoa socket that works with RB IPC sockets.   Are you 
just using shm_open and mmap?  If so I should be able to hook into that from 
Cocoa.  Anything I should be aware of?

Thanks!
_______________________________________________
Mbsplugins_monkeybreadsoftware.info mailing list
[email protected]
https://ml01.ispgateway.de/mailman/listinfo/mbsplugins_monkeybreadsoftware.info

Reply via email to