Re: [fpc-devel] How to move this, but in FPC 32bit?

[Ralf Quint]

On 11/14/2018 4:45 AM, Marco van de Voort wrote:


Op 2018-11-14 om 13:44 schreef Ozz Nixon:
Thank you all ... I am on the road, when I get back, I will track 
down TUnitHeader, TObjectTypeDefinition, TOrdinalTypeDefinition - and 
post those structures. My problem was changing the above to simple:

 PreviousUnitAddressOffset:=TUnitHeader.PreviousUnitAddress;

does not work, I think they are constants - I will have to dig back 
through the code. (It's from 1995, written by Anders Hejlsberg, so it 
does a lot of odd ASM things).


Internal identifiers are not always the same between FPC and Delphi. 
It might not be possible at all, so you need to find out why these are 
needed. 


+1

I was not sure, having nothing but that little snippet of code, if this 
was something trying to do internal to the compiler or if that was part 
of some data conversion tool.


If this is indeed accessing some Delphi internal data structures, I 
think nobody can seriously expect FreePascal to be binary compatible. It 
is (tries to be) source code compatible, as much as feasible. But 
anything beyond that,...


Ralf


---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus

___
fpc-devel maillist  -  fpc-devel@lists.freepascal.org
http://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-devel

Re: [fpc-devel] The 15k bounty: Optimizing executable speed for Linux x86 / LLVM

[Sven Barth via fpc-devel]

Am Mi., 14. Nov. 2018, 14:46 hat Simon Kissel <
simon.kis...@nerdherrschaft.com> geschrieben:

> Hi Florian,
>
> you are a hero. In a very artificial benchmark which just consists
> of threads and exception handlers, a 32 bit Linux executable now
> is *twice as fast*!
>

Up to now only thread variables are improved, the exception handling not
yet.


> In a real-life scenario we are "only" seeing an improvement of about
> 10%. But really, this is huge progress. I think everyone will
> benefit from these improvements.
>
> We have not yet tested this on ARM (does it work on ARM?).
>

Currently it's i386-linux only.

Regards,
Sven
___
fpc-devel maillist  -  fpc-devel@lists.freepascal.org
http://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-devel

Re: [fpc-devel] The 15k bounty: Optimizing executable speed for Linux x86 / LLVM

[Simon Kissel]

Hi Florian,

you are a hero. In a very artificial benchmark which just consists
of threads and exception handlers, a 32 bit Linux executable now
is *twice as fast*!

In a real-life scenario we are "only" seeing an improvement of about
10%. But really, this is huge progress. I think everyone will
benefit from these improvements.

We have not yet tested this on ARM (does it work on ARM?).

Bero will do more testing in the next couple of days and report
back.

Cheers,

Simon

___
fpc-devel maillist  -  fpc-devel@lists.freepascal.org
http://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-devel

Re: [fpc-devel] How to move this, but in FPC 32bit?

[Marco van de Voort]

Op 2018-11-14 om 13:44 schreef Ozz Nixon:
Thank you all ... I am on the road, when I get back, I will track down 
TUnitHeader, TObjectTypeDefinition, TOrdinalTypeDefinition - and post 
those structures. My problem was changing the above to simple:

 PreviousUnitAddressOffset:=TUnitHeader.PreviousUnitAddress;

does not work, I think they are constants - I will have to dig back 
through the code. (It's from 1995, written by Anders Hejlsberg, so it 
does a lot of odd ASM things).


Internal identifiers are not always the same between FPC and Delphi. It 
might not be possible at all, so you need to find out why these are needed.



___
fpc-devel maillist  -  fpc-devel@lists.freepascal.org
http://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-devel

Re: [fpc-devel] How to move this, but in FPC 32bit?

[Ozz Nixon]

Thank you all ... I am on the road, when I get back, I will track down
TUnitHeader, TObjectTypeDefinition, TOrdinalTypeDefinition - and post those
structures. My problem was changing the above to simple:
   PreviousUnitAddressOffset:=TUnitHeader.PreviousUnitAddress;

does not work, I think they are constants - I will have to dig back through
the code. (It's from 1995, written by Anders Hejlsberg, so it does a lot of
odd ASM things).

Ozz

On Tue, Nov 13, 2018 at 1:31 PM Ralf Quint  wrote:

> On 11/12/2018 7:37 AM, Ozz Nixon wrote:
>
> I have these last couple of lines to port to FPC -Mdelphi, but everything
> I try fails...
>
> Asm
> MOV  PreviousUnitAddressOffset,
> TUnitHeader.PreviousUnitAddress
> MOV  PreviousLibraryUnitAddressOffset,
> TUnitHeader.PreviousLibraryUnitAddress
> MOV  AncestorTypeDefinitionOffset,
> TObjectTypeDefinition.AncestorTypeOffset
> MOV  OrdinalTypeDefLowerLimitOffset,
>  TOrdinalTypeDefinition.LowerLimit
> MOV  OrdinalTypeDefUpperLimitOffset,
>  TOrdinalTypeDefinition.UpperLimit
> end;
>
> Well, as others already mentioned, it is hard to give proper advice
> without having a reproducible piece of code.
>
> Just looking at the snippet above, I would start with questioning if there
> isn't a more "Pascal way" to do this, whatever this piece of code is trying
> to accomplish...
>
> Ralf
>
>
> 
>  Virus-free.
> www.avast.com
> 
> <#m_-1587466564115926972_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
> ___
> fpc-devel maillist  -  fpc-devel@lists.freepascal.org
> http://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-devel
>
___
fpc-devel maillist  -  fpc-devel@lists.freepascal.org
http://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-devel