Hi everyone. I made ETuple a subclass of ERef and provided an ETupleResolver. I also repaired the TreeBrancher, which was returning a closure rather than scaling through th ETuple space. All fixed now, for your contemplation.
http://www.squeaksource.com/Cryptography/elinda-HenryHouse.13.mcz
- HH
On Thu, Oct 19, 2017 at 15:03, henry henry@callistohouse.club wrote:
This was my thinking on eLinda, whose combination of evaluating in Linda TupleSpace Subspaces and also the eventual sending based upon Erights.org ELib. I added event registrations of Tuples with ETupleSpace default eventuallyDo: [:tuple | count := count + 1]. Every eval afterwards would count. That eventual reactor runs on the PriorityVat' s Process. If tuple evals in the subspace started using the Vat's process, then it would be single-threaded, multitasking. I feel it could be a monitor of task dispersal.
Here is the latest eLinda, for Squeak and Pharo: http://www.squeaksource.com/Cryptography/elinda-HenryHouse.3.mcz
I also think it would make a good parsing engine for scripting ai computations. I was looking at PeitiParser and thinking of changing their baskends to generate Scripts that could run through this linda-vat engine.
As you have said, just thinking out loud.
- HH
-------- Original Message -------- Subject: Re: [Vm-dev] Multi-core system Squeak (was Re: Sista alternate bytecodes + Java Bytecode ?) Local Time: October 19, 2017 1:20 PM UTC Time: October 19, 2017 5:20 PM From: tblanchard@mac.com To: Squeak Virtual Machine Development Discussion vm-dev@lists.squeakfoundation.org
K, so if you were to marry that with Craig Latta"s micro images and "imprinting" strategy, offloading work to other "queues" might be a really cool way to chuck over a minimal amount of code associated with a task - where each work queue is just another micro worker image.
Just thinking out loud.
On Oct 19, 2017, at 10:09 AM, K K Subbu kksubbu.ml@gmail.com wrote:
On Thursday 19 October 2017 10:16 PM, tim Rowledge wrote:
On 19-10-2017, at 7:31 AM, Todd Blanchardtblanchard@mac.com wrote: It works very well and this idea of mini worker images sounds like it might be a good fit for that model.
One thing I noticed with surprise was how astonishingly fast Dave’s OSProcess could spawn an entire image, even on a Pi.
I second this idea. LatexMorph is a good example of that. It uses OSProcess to pipe LaTeX text through latex engine and Ghostscript into an image and import that image back into Squeak. The pipeline was fast enough for live rendering of image as the text was typed on laptops way back in 2010.
Regards .. Subbu
I meant rather than ‘eval’ing computations within the ETuple, through the TupleSpace.
- HH
On Sat, Oct 21, 2017 at 16:10, henry henry@callistohouse.club wrote:
Hi everyone. I made ETuple a subclass of ERef and provided an ETupleResolver. I also repaired the TreeBrancher, which was returning a closure rather than scaling through th ETuple space. All fixed now, for your contemplation.
http://www.squeaksource.com/Cryptography/elinda-HenryHouse.13.mcz
- HH
On Thu, Oct 19, 2017 at 15:03, henry henry@callistohouse.club wrote:
This was my thinking on eLinda, whose combination of evaluating in Linda TupleSpace Subspaces and also the eventual sending based upon Erights.org ELib. I added event registrations of Tuples with ETupleSpace default eventuallyDo: [:tuple | count := count + 1]. Every eval afterwards would count. That eventual reactor runs on the PriorityVat' s Process. If tuple evals in the subspace started using the Vat's process, then it would be single-threaded, multitasking. I feel it could be a monitor of task dispersal.
Here is the latest eLinda, for Squeak and Pharo: http://www.squeaksource.com/Cryptography/elinda-HenryHouse.3.mcz
I also think it would make a good parsing engine for scripting ai computations. I was looking at PeitiParser and thinking of changing their baskends to generate Scripts that could run through this linda-vat engine.
As you have said, just thinking out loud.
- HH
-------- Original Message -------- Subject: Re: [Vm-dev] Multi-core system Squeak (was Re: Sista alternate bytecodes + Java Bytecode ?) Local Time: October 19, 2017 1:20 PM UTC Time: October 19, 2017 5:20 PM From: tblanchard@mac.com To: Squeak Virtual Machine Development Discussion vm-dev@lists.squeakfoundation.org
K, so if you were to marry that with Craig Latta"s micro images and "imprinting" strategy, offloading work to other "queues" might be a really cool way to chuck over a minimal amount of code associated with a task - where each work queue is just another micro worker image.
Just thinking out loud.
On Oct 19, 2017, at 10:09 AM, K K Subbu kksubbu.ml@gmail.com wrote:
On Thursday 19 October 2017 10:16 PM, tim Rowledge wrote:
On 19-10-2017, at 7:31 AM, Todd Blanchardtblanchard@mac.com wrote: It works very well and this idea of mini worker images sounds like it might be a good fit for that model.
One thing I noticed with surprise was how astonishingly fast Dave’s OSProcess could spawn an entire image, even on a Pi.
I second this idea. LatexMorph is a good example of that. It uses OSProcess to pipe LaTeX text through latex engine and Ghostscript into an image and import that image back into Squeak. The pipeline was fast enough for live rendering of image as the text was typed on laptops way back in 2010.
Regards .. Subbu
vm-dev@lists.squeakfoundation.org
-
henry