[ANN] NativeBoost meets Linux

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

[ANN] NativeBoost meets Linux

Igor Stasenko
Hello all,

I just uploaded a fresh snapshots of NB plugin and image code,
with necessary bits for using it under Linux.

I don't know how to package Linux VM binaries, so you won't find a prebuilt VM
and should build it yourself (anyone wants to help me with that, so i
can add a prebuilt VM at downloads?).

As usual, follow the installation instructions, located here:
http://code.google.com/p/nativeboost/wiki/Installation

Warning: recently, i found a bug in callback code.
So, please don't try using smalltalk callbacks until i fix it :)

Running tests should be safe (i disabled the test which causing VM crash).


--
Best regards,
Igor Stasenko AKA sig.

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: [ANN] NativeBoost meets Linux

Igor Stasenko
On 26 September 2010 21:09, Igor Stasenko <[hidden email]> wrote:

> Hello all,
>
> I just uploaded a fresh snapshots of NB plugin and image code,
> with necessary bits for using it under Linux.
>
> I don't know how to package Linux VM binaries, so you won't find a prebuilt VM
> and should build it yourself (anyone wants to help me with that, so i
> can add a prebuilt VM at downloads?).
>
> As usual, follow the installation instructions, located here:
> http://code.google.com/p/nativeboost/wiki/Installation
>
> Warning: recently, i found a bug in callback code.
> So, please don't try using smalltalk callbacks until i fix it :)
>
> Running tests should be safe (i disabled the test which causing VM crash).
>

Update: bug was fixed after two days sitting with debugger and trying
to convince gdb to let me debug and find the
place with problem, which finished by looking to own code once again
and finding what it doing wrong.
Thanks DDD+gdb developers with not letting me simply debug my stuff,
but instead train
my brain's analytic capabilities. :)


>
> --
> Best regards,
> Igor Stasenko AKA sig.
>



--
Best regards,
Igor Stasenko AKA sig.

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: [ANN] NativeBoost meets Linux

Schwab,Wilhelm K
Sig,

That's kinda scary coming from someone who thinks in assembly instructions :)  I generally have resorted to tracing when debugging code in my libraries.  So far, DebugView has been hard to beat for that.  Not the same level, I realize, just a plug for anything that helps us see what the machine is doing, especially when it something goes wrong.

Bill


________________________________________
From: [hidden email] [[hidden email]] On Behalf Of Igor Stasenko [[hidden email]]
Sent: Tuesday, September 28, 2010 1:12 AM
To: The general-purpose Squeak developers list; Pharo Development
Subject: Re: [Pharo-project] [ANN] NativeBoost meets Linux

On 26 September 2010 21:09, Igor Stasenko <[hidden email]> wrote:

> Hello all,
>
> I just uploaded a fresh snapshots of NB plugin and image code,
> with necessary bits for using it under Linux.
>
> I don't know how to package Linux VM binaries, so you won't find a prebuilt VM
> and should build it yourself (anyone wants to help me with that, so i
> can add a prebuilt VM at downloads?).
>
> As usual, follow the installation instructions, located here:
> http://code.google.com/p/nativeboost/wiki/Installation
>
> Warning: recently, i found a bug in callback code.
> So, please don't try using smalltalk callbacks until i fix it :)
>
> Running tests should be safe (i disabled the test which causing VM crash).
>

Update: bug was fixed after two days sitting with debugger and trying
to convince gdb to let me debug and find the
place with problem, which finished by looking to own code once again
and finding what it doing wrong.
Thanks DDD+gdb developers with not letting me simply debug my stuff,
but instead train
my brain's analytic capabilities. :)


>
> --
> Best regards,
> Igor Stasenko AKA sig.
>



--
Best regards,
Igor Stasenko AKA sig.

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: [ANN] NativeBoost meets Linux

Igor Stasenko
On 28 September 2010 14:27, Schwab,Wilhelm K <[hidden email]> wrote:
> Sig,
>
> That's kinda scary coming from someone who thinks in assembly instructions :)  I generally have resorted to tracing when debugging code in my libraries.  So far, DebugView has been hard to beat for that.  Not the same level, I realize, just a plug for anything that helps us see what the machine is doing, especially when it something goes wrong.
>
I tried many different debugging tools in recent months.
I found that all of them still inferior to examine and analysis of code :)

Some of them even refusing to show me anything userful, or making is
so hard, that its easier to simply read source code, rather than
debugging. And debugging VM is not the same as debugging usual C
program..
Now add a native code into the soup, and you'll understand why all of
these cool debuggers are incapable to help..
unless you spent 100++ hours with them and mastered everything about debugger.

From other side, there is not much complex things in code generator. I
isolating things into small pieces of code
(couple lines of smalltalk code per method), so there is nearly
impossible to do a mistake.
It is relatively easy to find a mistake at first run of the code.

The bug which i fixed recently was in callback code, which is quite
complex and is rather exception than common case
(a method contains maybe 100 lines of code). And bug was a logical
mistake, not something like code generator producing wrong
instructions etc :)

> Bill
>
>
> ________________________________________
> From: [hidden email] [[hidden email]] On Behalf Of Igor Stasenko [[hidden email]]
> Sent: Tuesday, September 28, 2010 1:12 AM
> To: The general-purpose Squeak developers list; Pharo Development
> Subject: Re: [Pharo-project] [ANN] NativeBoost meets Linux
>
> On 26 September 2010 21:09, Igor Stasenko <[hidden email]> wrote:
>> Hello all,
>>
>> I just uploaded a fresh snapshots of NB plugin and image code,
>> with necessary bits for using it under Linux.
>>
>> I don't know how to package Linux VM binaries, so you won't find a prebuilt VM
>> and should build it yourself (anyone wants to help me with that, so i
>> can add a prebuilt VM at downloads?).
>>
>> As usual, follow the installation instructions, located here:
>> http://code.google.com/p/nativeboost/wiki/Installation
>>
>> Warning: recently, i found a bug in callback code.
>> So, please don't try using smalltalk callbacks until i fix it :)
>>
>> Running tests should be safe (i disabled the test which causing VM crash).
>>
>
> Update: bug was fixed after two days sitting with debugger and trying
> to convince gdb to let me debug and find the
> place with problem, which finished by looking to own code once again
> and finding what it doing wrong.
> Thanks DDD+gdb developers with not letting me simply debug my stuff,
> but instead train
> my brain's analytic capabilities. :)
>
>
>>
>> --
>> Best regards,
>> Igor Stasenko AKA sig.
>>
>
>
>
> --
> Best regards,
> Igor Stasenko AKA sig.
>
> _______________________________________________
> Pharo-project mailing list
> [hidden email]
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
>
> _______________________________________________
> Pharo-project mailing list
> [hidden email]
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
>



--
Best regards,
Igor Stasenko AKA sig.

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: [ANN] NativeBoost meets Linux

Igor Stasenko
On 28 September 2010 16:14, Igor Stasenko <[hidden email]> wrote:
> On 28 September 2010 14:27, Schwab,Wilhelm K <[hidden email]> wrote:
>> Sig,
>>
>> That's kinda scary coming from someone who thinks in assembly instructions :)  I generally have resorted to tracing when debugging code in my libraries.  So far, DebugView has been hard to beat for that.  Not the same level, I realize, just a plug for anything that helps us see what the machine is doing, especially when it something goes wrong.
>>
> I tried many different debugging tools in recent months.
> I found that all of them still inferior to examine and analysis of code :)
>
Oh, i forgot to mention about smalltalk debugger. Well, i think you
know by youself:
it may miss some features, but in overall i never had any problems with
finding a bugs using it, which usually takes not more than a few minutes.
Using C debuggers is like an excursion to stone age, where you have
only one tool at
your disposal: stone hammer.

> Some of them even refusing to show me anything userful, or making is
> so hard, that its easier to simply read source code, rather than
> debugging. And debugging VM is not the same as debugging usual C
> program..
> Now add a native code into the soup, and you'll understand why all of
> these cool debuggers are incapable to help..
> unless you spent 100++ hours with them and mastered everything about debugger.
>
> From other side, there is not much complex things in code generator. I
> isolating things into small pieces of code
> (couple lines of smalltalk code per method), so there is nearly
> impossible to do a mistake.
> It is relatively easy to find a mistake at first run of the code.
>
> The bug which i fixed recently was in callback code, which is quite
> complex and is rather exception than common case
> (a method contains maybe 100 lines of code). And bug was a logical
> mistake, not something like code generator producing wrong
> instructions etc :)
>
>> Bill
>>
>>
>> ________________________________________
>> From: [hidden email] [[hidden email]] On Behalf Of Igor Stasenko [[hidden email]]
>> Sent: Tuesday, September 28, 2010 1:12 AM
>> To: The general-purpose Squeak developers list; Pharo Development
>> Subject: Re: [Pharo-project] [ANN] NativeBoost meets Linux
>>
>> On 26 September 2010 21:09, Igor Stasenko <[hidden email]> wrote:
>>> Hello all,
>>>
>>> I just uploaded a fresh snapshots of NB plugin and image code,
>>> with necessary bits for using it under Linux.
>>>
>>> I don't know how to package Linux VM binaries, so you won't find a prebuilt VM
>>> and should build it yourself (anyone wants to help me with that, so i
>>> can add a prebuilt VM at downloads?).
>>>
>>> As usual, follow the installation instructions, located here:
>>> http://code.google.com/p/nativeboost/wiki/Installation
>>>
>>> Warning: recently, i found a bug in callback code.
>>> So, please don't try using smalltalk callbacks until i fix it :)
>>>
>>> Running tests should be safe (i disabled the test which causing VM crash).
>>>
>>
>> Update: bug was fixed after two days sitting with debugger and trying
>> to convince gdb to let me debug and find the
>> place with problem, which finished by looking to own code once again
>> and finding what it doing wrong.
>> Thanks DDD+gdb developers with not letting me simply debug my stuff,
>> but instead train
>> my brain's analytic capabilities. :)
>>
>>
>>>
>>> --
>>> Best regards,
>>> Igor Stasenko AKA sig.
>>>
>>
>>
>>
>> --
>> Best regards,
>> Igor Stasenko AKA sig.
>>
>> _______________________________________________
>> Pharo-project mailing list
>> [hidden email]
>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
>>
>> _______________________________________________
>> Pharo-project mailing list
>> [hidden email]
>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
>>
>
>
>
> --
> Best regards,
> Igor Stasenko AKA sig.
>



--
Best regards,
Igor Stasenko AKA sig.

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project