Vm-dev May 2011

vm-dev@lists.squeakfoundation.org

36 participants
89 discussions

doesNotUnderstand:/createActualMessageTo: for zero-argument sends appears to imply the need for an extra stack slot in methods
by Eliot Miranda 06 May '11

06 May '11

Hi All, could you check my reasoning here. I think I've found a serious bug in the VM with doesNotUnderstand: processing for zero-argument sends and the calculation of the necessary stack size in MethodNode>>generate:* (senders of CompiledMethod>>needsFrameSize:). The bug causes a push beyond the end of a context, overwriting the header of the following object, when an MNU occurs for a zero-argument send when the stack is full. If you look at Interpreter>>createActualMessageTo: it does the following to push the message created as the argument of the doesNotUnderstand: self pop: argumentCount thenPush: message If argumentCount is zero this pushes an item on the stack. If the stack is full at this point (the receiver of the not-understood message is the last item in the context) then the push will be out of bound of the activeContext's stack and overwrite the first word of the following object's header. We observe this with a Newspeak application at Cadence and see a subsequent VM crash. 1. am I right? 2. is this an old bug? 3. did this, or did this not, affect Smalltalk-80 v2 and e.g. the Dorado implementation, and if so why (e.g. was the message of an MNU not pushed onto the stack, and instead MNUs activated by special code?, or was the activeContext not on the heap? or...) 4. what's the right fix? It seems to me that the right fix is simply to add one to the needed stack depth, with an explanatory comment, but in senders of CompiledMethod>>needsFrameSize: or in CompiledMethod>>needsFrameSize:? Trying to avoid pushing the argument of an MNU is going to involve lots of changes to the VM and will be tricky and likely have a bug tail. best, Eliot

2 1

VM Maker: VMMaker-dtl.231.mcz
by squeak-dev-noreply＠lists.squeakfoundation.org 06 May '11

06 May '11

Dave Lewis uploaded a new version of VMMaker to project VM Maker: http://www.squeaksource.com/VMMaker/VMMaker-dtl.231.mcz ==================== Summary ==================== Name: VMMaker-dtl.231 Author: dtl Time: 6 May 2011, 12:04:23 pm UUID: f21c38a5-01c6-46d1-ad3a-c3050ca8c29a Ancestors: VMMaker-dtl.230 VMMaker 4.4.18 Changes to allow 64-bit image interpreter simulator to run on little endian (e.g. Intel) platforms. Implement missing methods #longAt: #longAt:put: and #byteSwapped: in InterpreterSimulatorLSB64. Fix InterpreterSimulatorLSB>>charsOfLong: to use bytesPerWord instead of 4, required for running InterpreterSimulatorLSB64. Fix the inSmalltalk: block in #primitiveNewWithArg (fix affects simulator only, no effect on generated code). Use "self bytesPerWord" in Interpreter (required for C translation) but use direct variable ref in the simulator because it is noticeably faster in simulation. Fix typo in InterpreterSimulator class comment. Note, original 64-bit image work was done on a big endian Mac, so a few LSB64 updates may have been missed at that time.

1 0

SystemTracing: SystemTracing-dtl.21.mcz
by squeak-dev-noreply＠lists.squeakfoundation.org 06 May '11

06 May '11

Dave Lewis uploaded a new version of SystemTracing to project SystemTracing: http://www.squeaksource.com/SystemTracing/SystemTracing-dtl.21.mcz ==================== Summary ==================== Name: SystemTracing-dtl.21 Author: dtl Time: 5 May 2011, 11:29:07 am UUID: 08000000-1508-d21f-1508-d21f14000000 Ancestors: SystemTracing-dtl.20 Remove redundant class side methods

1 0

Using symbols in slang?
by Igor Stasenko 06 May '11

06 May '11

hello, just a thought, that it would be nice if we could use symbols in translated code. They don't have to carry any associated value(s) like classvars, but instead could be used as unique identifiers. struct sqSymbol { }; #define declareSymbol(name) struct sqSymbol sqSym_##name #define sqSym(name) &sqSym_##name declareSymbol(foo); each time translator sees code like: (a == #foo ) ifTrue: [ ] it could generate a following code: if ( ((sqSymbol*)a) == sqSym(foo)) { .... } Just a single example of application of it: Currently a CogRTLOpcodes defines a lot of symbolic constants , like ... Jump JumpAbove JumpAboveOrEqual JumpBelow JumpBelowOrEqual ... and the values of these constants carry no any additional semantic load, and could be replaced by simple unique values (which is symbol objects in smalltalk). If i remember correctly , currently a symbol literals in method frame are translated to strings. But in practice i think there are very few of such uses, because using string literals is less confusing and more intention revealing. While using symbols in a way which i propose could simplify some code. There is of course a different way how to translate symbols: - a code generator could contain an incrementing counter for each symbol it meets in code and then and the end it generates a symbols.h file which looks like: ... #define foo 1 #define bar 2 .. and so on. -- Best regards, Igor Stasenko AKA sig.

1 0

SystemTracing: SystemTracing-dtl.20.mcz
by squeak-dev-noreply＠lists.squeakfoundation.org 05 May '11

05 May '11

Dave Lewis uploaded a new version of SystemTracing to project SystemTracing: http://www.squeaksource.com/SystemTracing/SystemTracing-dtl.20.mcz ==================== Summary ==================== Name: SystemTracing-dtl.20 Author: dtl Time: 4 May 2011, 10:22:06 am UUID: 08000000-1508-7013-1508-701314000000 Ancestors: SystemTracing-ul.19 PositionableStream>>nextLongPut: required a class definition override, so replace it with SystemTracer>>putLong:to: based on the equivalent methods in the interpreter simulator.

1 0

CocoaVMs DO NOT automatically recognize proxy settings of MacOS
by Mariano Martinez Peck 04 May '11

04 May '11

Hi guys. The VMs for Carbon, both build with CMakeVMMaker or the ones from Eliot, do automatically recognize the proxy settings in the network preferences of MacOS. So, the images work out of the when going to internet. However, CocoaVMs (tested building with CMakeVMMAker) don't do that, and in fact, you have to explicitly set the proxy, by using HTTPSocket proxyUser: userName password: password or using the Pharo Settings. Here at work I have a proxy...setting manually everytime I open an image is not funny. Even more, if I then try to open it at home (where I don't have a proxy), of course it doesn't work, so I need to reset the settings. So....anyone has a clue where such differences can be? what can be done? thanks in advance -- Mariano http://marianopeck.wordpress.com

3 3

Integer comparition primitive failures on Cog
by Henrik Sperre Johansen 04 May '11

04 May '11

As per my comment on the recent Morphic performance graphs,(http://blog.openinworld.com/2011/03/morphic-flavour-performance/#co… on a Cog VM these primitives fail with a Float parameter, which leads to a huge performance hit when comparing Ints to Floats by doing silly things in the fallback code. Cog: [1 to: 2000000 do: [:i | i < 2354.234. i <= 2354.234. i >= 2354.234. i > 2354.234.]] timeToRun 26594 Trunk: [1 to: 2000000 do: [:i | i < 2354.234. i <= 2354.234. i >= 2354.234. i > 2354.234.]] timeToRun 229 Should this be changed in the VM, or is the difference for pure int/int comparition large enough that we should instead change Float>>adaptToInteger: andCompare: to use the corresponding comparitions with Float as receiver, which does work? That's still slower than trunk in my image, but quite a bit better, above test takes about 1s. Cheers, Henry

6 24

Fwd: [Pharo-project] System time
by stephane ducasse 04 May '11

04 May '11

Begin forwarded message: > From: Jimmie Houchin <jlhouchin(a)gmail.com> > Date: May 1, 2011 4:15:40 AM GMT+02:00 > To: Pharo-project(a)lists.gforge.inria.fr > Subject: Re: [Pharo-project] System time > Reply-To: Pharo-project(a)lists.gforge.inria.fr > > Hello, > > My OS for this application is Windows. On a freshly opened image, the image and OS time are in sync. Over a period of time the image drifted 34 seconds away from OS time. I do not know how long I had this image open, but it has been a number of days, possibly a week plus. > > I saved the image and quit. I reopened the image and did DateAndTime now and it was in sync with the OS once again. Same image, vm and OS that was out of sync by 34 seconds only one minute before. > > I have opened an image on my server which is also running the latest cog vm, the current Pharo 1.3 and Windows Vista. I am going to periodically do a DateAndTime now to log the drift if any. Prior to opening this image I updated the OS system clock with time.nist.gov which is about 1000 miles away. > > I know that within a second or so it can be subject to my coordination of doing DateAndTime now while watching my OS system clocks seconds. But 34 seconds is beyond any subjective standards. > > The only reason I discovered this was while exploring how I want to retrieve my data, act upon my data and then pause (Delay wait) until time to repeat. > dtnow := DateAndTime now. > dtnext := DateAndTime year: 2011 month: 4 day: 30 hour: 11 minute: i second: 5. "The next minute in the future" > (Delay forDuration: (dtnext - dtnow)) wait. > > I got an error, that Duration cannot be a negative number. As I processed the debugger, I saw that dtnow was greater than dtnext. I watched the OS clock and manually selected the next minute before I executed. However, I was within that 34 second window. > > So now I am exploring the problem. > > Thanks. > > Jimmie > > On 4/30/2011 5:06 PM, Alexandre Bergel wrote: >> 'Time now' gives me the same result than OSX >> >> Alexandre >> >> >> On 30 Apr 2011, at 16:15, Jimmie Houchin wrote: >> >>> On 4/30/2011 11:50 AM, Stéphane Ducasse wrote: >>>>> I am developing a time sensitive application. I poll a server every minute for new data. The data is time sensitive and the sooner I can get the data to the release of the data the better. >>>>> >>>>> I have noticed that Pharo's system time drifts away from the OS system time. I would really like the Pharo system time to be the correct time. >>>> do you have a repeatable scenario? >>>> VM >>>> Image information. >>> Pharo 1.3 and the latest Cog. >>> >>>>> I have not been able to find where in Pharo I can call to update the System clock to the correct time. >>>> you can use OSProcess (only mac and linux) >>> No, that's not the problem. The OS clock is right. It's Pharo that has drifted away from correct time. Regardless it would be nice if I have to correct the OS clock, it would still be nice to resync Pharo with the OS, keep both in sync and correct. What I am wanting is simply to resync the Pharo clock with the OS clock. >>> >>> It seems that when the Pharo image is started that it gets the OS System time, but keeps its own time after that. Now I do not know the consequences of Pharo adjusting its time backward. If this isn't possible, then I might have to do some workarounds to insure that I get my data in a timely manner. >>> >>> The image I have been developing in is currently about 30 seconds ahead of the OS clock. It is on my laptop. It might be due to the laptops sleep and hibernation. I have not tested this on my server. >>> >>> That is the situation as best I know it. >>> >>> Thanks. >>> >>> Jimmie >>> > >

4 8

Hudson Status?
by Casey Ransberger 04 May '11

04 May '11

Hello all, I have chunk of free time coming up, and a bunch of professional experience with CI. Seems I end up being the guy who sets it up and drives the effort wherever I go. Curious where the status for the Hudson effort is. Is the blocker mostly time? I should be able to get it done, if so. I'm interested in doing both VM-level and image-level CI. If you're currently working on the Hudson effort for Squeak, please let me know so that I don't have to duplicate your efforts. Thank you!

4 4

Questions about Cog internals
by Mariano Martinez Peck 03 May '11

03 May '11

Hi Eliot. I am really trying (with all my lack of knowledge) to understand a little about how Cog works internally. I am also reading your posts, and I have a couple of (probably newbie) questions. If any of them are answered in the blog, please point me to them (I couldn't ream all of them yet): 1) Suppose you have a CompiledMethod XXX that you JIT, and you get a CogMethod YYY. While doing the GC (#lastPointersOf:, #lastPointerWhileForwarding:, etc), you need to check whether XXX is a CogMethodReference because if so, you need to fetch XXX header from YYY. Perfect. To avoid the GC to look in the CogMethod "objects" you put a header with the special format for empty objects, hence the GC doesn't follow the non-existent "instVars" of CogMethod. Perfect. CogMethod has a pointer to its original CM (back-pointer), called 'methodObject'. In this case, YYY has a pointer to XXX. So....my question is, during a GC compaction or a #become, where the address of XXX is changed, how do you update YYY so that to point to the new address of XXX? because if you flag YYY as an empy object, then the GC doesn't update it. 2) As far as I understand, CogMethod doesn't "store/duplicate" the literals of the CompiledMethod. Hence, even when you have a jitted method, when you need a special literal, you ask it to the CM, using the backPointer 'methodObject'. Is this correct ? 3) This is the most stupid question, but I don't see WHERE the machine code is kept. When I jit a method, I get a structure CogMethod, perfect. What where is the generated machine code? where is it kept? how can I know from a CogMethod which is the associated machine code? 4) I guess that my thought of 2) is not correct, because otherwise, I don't understand why you need CoInterpreter >>markAndTraceOrFreeMachineCode:. The comments says "Deal with a fulGC's effects on machine code. Either mark and trace oops in machine code or free machine-code methds that refer to freed oops. The stack pages have already been traced so any method of live stack activations have already been marked and traced." which oops do you mean by "oops in machine code" ? literals? the back-poiner to the CM? and by " free machine-code methods that refer to freed oops" what do you mean? literals or oops as the back pointer? I can think you refer to the backpointer since the original CM could have been garbage collected and since you flag the CogMethod as empty... 5) This is not a question, but rather that I would like to know whether I understood correctly or not. You Jit a method when it is secondly used, that is, when you find it in the cache. To know how to generate the machine code or a particular bytecode, you check in the table that you generate wth #initializeBytecodeTableForClosureV3 where you basically map bytecodes to methods that generates the machine code of such bytecode. If it is a primitive you use instead #compilePrimitive which cecks in a similar table, but for primitives, which is set in #initializePrimitiveTableForSqueakV3. Now, I have compiled method XXX (selector xxx) which sends #foo. XXX was jitted to CogMethod YYY (selector yyy). When xxx is executed, YYY is executed. When YYY was jitted, you defined in #initializeBytecodeTableForClosureV3 that it just be a specific method, which at the end, for normal messages it is: #genSend:numArgs:. That method to generate the machine code includes the "trampoline" (which is searched in 'sendTrampolines', and in #generateSendTrampolines we can see how you map from one to the other one) and sends the associated message, in this case, #ceSend:super:to:numArgs:. So...the #foo will be finally "handle" in ceSend:super:to:numArgs:. This is ONLY true if the send was "unlinked". If #foo in fact was jitted also, then you try to link it (to avoid searching in cache next times???). Suppose you could link both of them,so next time YYY is executed, it will call DIRECTLY the CogMethod of #foo. In this case, the method to be executed in the VM is #executeCogMethodFromLinkedSend:withReceiver: instead of #ceSend:super:to:numArgs: So..I am delirious or that is more or less correct ? Thanks a lot in advance, -- Mariano http://marianopeck.wordpress.com

2 1

← Newer
1
2
3
4
5
6
7
8
9
Older →

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

Vm-dev May 2011