gmane.comp.lang.slate.general

towards a 64 bit vm / image

Timmy Douglas — 2007-07-10T03:25:09

I put together some patches that redo the pidgin type system to use traits instead of rules, use sizeof instead of hard coded bit/byte/word sizes, and write slate images based on a given bitsize instead of assuming 32 bits. I'm doing this because i want to run slate on my amd64 system. Unfortunately it's broken now and nothing works, but hopefully I'll fix that in the next few years if I'm lucky. It includes some other stuff I think.

slate pidgin woes

Timmy Douglas — 2007-06-27T02:34:59

I'm trying to make some changes to the slate pidgin code to help make porting the VM to my 64 bit machine easier. I probably did a few too many search/replaces but basically I'm getting this error: Generating function PSObject_isImmutable Generating function PSObject_restrictsDelegation Generating function PSObject_addSlotNamed_valued_at_ The following condition was signaled: Error: A method translation wasn't found for copyBytes:into:. The following restarts are available: 0) Abort evaluation of expression 1) Quit Slate Enter 'help.' for instructions. Debug [0..1]: up. [tryHandlers] the method translation searcher is: g< at >(C SimpleGenerator traits) generateCFor: selector< at >(Symbol traits) on: arguments [...] g: ("SimpleGenerator" currentColumn: 0. currentLine: 0. level: 0. columnsPerLine: 1000. original: ("Stream" ...). inlining: True. asserting: False. verbose: False. useSymbolicConstants: True. useRightShiftForDivide: True. generateDeadCode: False. methodCalls: {"Bag" ...}.

darcs patch: Fixed bitShift: for BigIntegers for the case of 0.

2007-06-18T00:26:18

Sun Jun 17 17:25:23 PDT 2007 BrianTRice< at >gmail.com * Fixed bitShift: for BigIntegers for the case of 0.

Re: Slate on a message-passing machine

Stefan Plantikow — 2007-06-16T11:32:52

Am 15.06.2007 um 22:46 schrieb Jecel Assumpcao Jr: I wonder wether you are aware of Albert/Pepsi/Cola by Ian Piumarta? http://www.piumarta.com/pepsi http://www.piumarta.com/papers Hope this is useful to you, Stefan Plantikow PS: Sorry, I didn't keep the original message so I had to reply to this one.

Re: Slate on a message-passing machine

Jecel Assumpcao Jr — 2007-06-15T20:46:29

Matthew Fulmer wrote on Sat, 2 Jun 2007 09:19:18 -0700 It is a good idea to aim for the most general case if possible. It is also nice to have statistics which tell you what the most common case is and how important it is to the overall performance of the system. You need to add the concept of "role", but other than that RNA would be suitable for Slate. A message send would involve visiting several different objects (also the case for inheritance in a single dispatch version) but I like to think of that as several phases of a single message send rather than extra messages, though both interpretations are equally valid. The case where I think Slate would be equivalent to Smalltalk/Self is when an ancestor of the first argument implements the method with all of the other arguments having NoRole. But it has been a while and I am probably remembering the Slate terms wrong. Slate's dispatch is a little less symmetric than it would seem at first glance: there is a left-to-right bias. Another "quirk" is tha

64 bit slate VM images

Timmy Douglas — 2007-06-09T22:28:18

I just upgraded my computer to a dual core amd64 machine. I'd like to use this machine to work on slate, but I can't find any 64 bit images and I'm not sure how I could use the initial lisp sources to generate a new kernel image. Anyone have any ideas or experience here? I'd like to stay away from installing any 32 bit programs on this computer. Thanks.

Slate on a message-passing machine

Matthew Fulmer — 2007-06-06T17:32:05

I sent this four days ago, but it never yet arrived For my Master's thesis, I will probably be working on a hardware architecture for efficiently executing object-oriented systems. Slate is definitely the most generic of these types of systems, and so I believe I should make the hardware directly applicable to slate, since all other object oriented semantics are a subset of those provided by Slate. For an example of the kind of hardware I am talking about, consider Jecel's Ring Network Architecture: http://www.merlintec.com:8080/hardware/19 . According to the example, the RNA architecture can send, lookup, and execute a message that is installed on a single receiver without software assistance. I wonder if this design can accommodate some, if not all, of the multi-dispatched message sends in slate. In other words, is there (hopefully large) a subset of slate message-passing semantics that can be resolved without resorting to sending extra messages? I hazard to guess that a subset of slate does have the sa

Slate on a message-passing machine

Matthew Fulmer — 2007-06-02T16:19:18

For my Master's thesis, I will probably be working on a hardware architecture for efficiently executing object-oriented systems. Slate is definitely the most generic of these types of systems, and so I believe I should make the hardware directly applicable to slate, since all other object oriented semantics are a subset of those provided by Slate. For an example of the kind of hardware I am talking about, consider Jecel's Ring Network Architecture: http://www.merlintec.com:8080/hardware/19 . According to the example, the RNA architecture can send, lookup, and execute a message that is installed on a single receiver without software assistance. I wonder if this design can accommodate some, if not all, of the multi-dispatched message sends in slate. In other words, is there (hopefully large) a subset of slate message-passing semantics that can be resolved without resorting to sending extra messages? I hazard to guess that a subset of slate does have the same semantics as the operation natively handled by RNA

slate bootstrap bug status report

Timmy Douglas — 2007-05-14T16:19:59

I've been looking into a bug in the bootstrap process. After you make a new image with: Image bootstrapBasedOn: VM new, slate tries to load that image, and that's where it fails. The image file is made by parsing a lot of slate files which are defined in build.slate and then concatenating them into one function that initializes a normal slate environment. Since some of the functions used in earlier files depend on functions fully defined in later files, pre-bootstrap.slate is loaded first with some common functions defined in simpler terms to keep things from crashing because of missing dependancies. However, somewhere in numeric.slate, things go wrong. The first define: numerics define: #Comparable &parents: {Cloneable}. causes the Cloneable object to be transformed into a Compareable object, and things go spiraling out of control when Float is defined, which somehow causes the #traits3 = #traitsWindow to invoke CoercibleNumberMixin's specializer coerceTo (because Symbol's inheritance graph has change

Re: Some notes concening VM :)

sig — 2007-04-30T04:50:54

as i see the CompiledMethod structure holds a sourceTree object, which i suppose holds exact answers on what literals is selectors and what is symbols or something else. Im just against sacrificing the runtime model simplicity in favor of analysis, which can be simply accessible through other means. My approach is simple: minimalize number of operations required to evaluate single opcode instruction.The less we have - the more we gain. Using stack to run both CPU and Slate bytecode is dangerous. In case of exception it may be too hard to unwind stack to some safe point. If you know how this can be safely done, tell me. I don't think that its impossible, it just demands some research. Also GC at mark phase must know the set of its root objects , this means that it must scan stack to determine roots.. And it can be too hard and expensive to scan stack where real objects mixed with foreign data. But! If we use dual stack this can be quite possible. Because each stack frame have information of its starti

Re: Some notes concening VM :)

Brian Rice — 2007-04-29T17:53:35

On Apr 28, 2007, at 7:25 PM, sig wrote: Well, I can say that for introspection, it is easier to tell what selectors a method sends if they are in a separate array, so easier to track down a call graph without referring to source code. Granted, in the general case, plenty of those literal Symbols are also used by sendTo: and variants, or Symbols can be generated at run-time by the method, but this makes the base case easier. Also granted, just doing "method literals select: [| :each | each is: Symbol]" isn't that much harder (although is: is kind of expensive in aggregate). So I wouldn't fight much for this one. Perhaps the only other defense is how well the method format scales up, that is, how does method size grow with source length. The truly anomalous case is the bootstrap method... in an image bootstrap, what happens is that the Slate system compiles all the core source files' contents into a *single* bytecode method which is then marked as the starting point for kernel ima

Some notes concening VM :)

sig — 2007-04-29T02:25:31

Hello , i'm new here and have particular interests in developping a Slate VM. I didn't decided yet if i will participate in this project in nearest future, so i just want to discuss some of my ideas with you for now. Concerning OMM (object memory model) and stack organization. I'm refer to tables which you can see on this page: http://slate.tunes.org/doc/mobius/ 5.1.1 stack format. Table 3. Block format. I see little reason why selectors must be kept apart from literals. The only place i see where they handled differently is at interpretation stage to use different opcode 'Load Selector'. If there's no extra processing except that, we can merge selectors with literals slot and remove an opcode 'Load Selector'. This is the code that shows that they handled in same manner.. static INLINE void PSInterpreter_loadLiteral_(struct Interpreter * i, unsigned long int n) { PSInterpreter_stackPush_(i, (((i -> method) -> literals) -> elements)[n]); } static INLINE void PSInterpreter_loadSelector_(struct Interpr

Re: Long-overdue updates

Brian Rice — 2007-04-24T14:37:48

Quick update: On Apr 7, 2007, at 5:40 PM, Brian Rice wrote: I'm still working on this with Tim Douglas. We've at least made reasonable progress and re-learned some forgotten things. There's a new "stable" repository on the main site which may or may not be temporary, and isn't fully stable yet anyway (there seems to be one more tweak to make to be ready-to-go). -- -Brian http://briantrice.com

Re: Long-overdue updates

Brian Rice — 2007-04-08T19:57:14

On Apr 8, 2007, at 6:15 AM, Fredrik Johansson wrote: Slate is so slow right now due to heavy message-sending and no optimizations for it that I'll need to sort out how to limit CPU loading to do that. Maybe if I hosted it on my home server... -- -Brian http://briantrice.com

Re: Long-overdue updates

Brian Rice — 2007-04-08T19:55:46

On Apr 7, 2007, at 9:34 PM, Sam Phillips wrote: There were submitted patches which I did not sufficiently vet before putting into the alpha branch. As soon as I've sorted out what all those are, it'll be set up right. I'm sorry about the trouble; I just haven't felt motivated for a while. -- -Brian http://briantrice.com

Re: UI demo performance, GC

Brian Rice — 2007-04-08T19:51:57

On Apr 7, 2007, at 7:50 PM, Timmy Douglas wrote: Most projects have been for work. I've been doing contract work on two-month time periods for a *wide* variety of subjects. Some of these spawned for-fun projects, like some work for handheld UIs using Squeak (after an embedded systems prototyping project). The system I'm currently working on may be an open-source application by July, but it's not a technically-innovative project, just a good open-government framework. -- -Brian http://briantrice.com

Re: Long-overdue updates

Fredrik Johansson — 2007-04-08T13:15:00

A web-based interactive interpreter (like http://tryruby.hobix.com/) would be a nice feature. Fredrik

Re: Long-overdue updates

Sam Phillips — 2007-04-08T04:34:35

This in itself is great news for me. Throughout the past 6 months (off and on) I've been attempting to tweak my build environment in various small ways just to get it to bootstrap from a "pure" alpha branch check out. Cheers, Sam

Re: UI demo performance, GC

Timmy Douglas — 2007-04-08T02:50:27

work projects or for-fun projects? Hmm, I might have agreed with Lee there but he could have helped Tom more.

Long-overdue updates

Brian Rice — 2007-04-08T00:40:13

Hello all, I'd like to revive the Slate project from dormancy in the last very quiet year, and would just like everyone to know that major overhauls of the project or code are totally acceptable if it means making it viable again. Most importantly, run-time improvements for memory management (generational GC, weak pointers, or manual memory management support) and the VM (threaded-code vs. bytecode, and most importantly, dynamic method inlining). Also, if someone wants to just take the reins and go change things relatively drastically, I will support that. I've fixed up the main site homepage to correctly describe what we want to have environment-wise, compared to Self/Strongtalk: same optimizations, but self-hosted and in a simpler architecture. I've also hopefully more accurately described the current state of the project. I'm attempting to get the alpha repository to run all build targets without end-user tweaking. I'll set up some automation so that I can track this down wi

Re: UI demo performance, GC

Brian Rice — 2007-04-08T00:28:07

On Apr 7, 2007, at 12:32 PM, Timmy Douglas wrote: That's a simplified overview, yes. That's correct. I'm not sure what to advise at this point. I've got my attention on a lot of different projects these days and not much in Slate. A vaguely relevant email found via Googling: http://bespin.org/pipermail/slate/2005-January/001032.html I've decided to put enough effort into Slate that random interested people can try it out and contribute. I'll detail what I do in a separate email thread. -- -Brian http://briantrice.com