Laz/FPC version of Velthuis arbitrary precision package for Dephi

[MathMan]

Hello all,

Is there, by any chance, somebody around here who has converted Rudy Velhuis arbitrary precision package for Delphi to a compilable & working Laz/FPC variant?

Src: https://github.com/rvelthuis/DelphiBigNumbers
Web: http://rvelthuis.de/programs/bigintegers.html

I would like to do some comparisons with my own package (in the making) and tried to convert it, with the help of the available converter tools from Laz/FPC. However I must admit I am not versed enough in Delphi & Laz/FPC to get it going.

Any feedback highly appreciated - even if it is just hints, what I need to look at to get it going myself.

My current environment is Laz 1.8.2 / FPC 3.0.4 on x86-64 Linux.

Kind regards,
MathMan

[Thaddy]

I was almost finished making it compatible, but I ran into an internal error 200706094, so I can't finish it before that is resolved.
In principle it was not much work. I reported the bug.

Code: Pascal  [Select][+][-]

1. {$MODE DELPHI}
2. {$RANGECHECKS OFF}
3. {$OVERFLOWCHECKS OFF}
4.  
5. function SplitMix64(var X: UInt64) : UInt64;
6. var
7.   Z: UInt64;
8. begin
9.   Inc(X, UInt64($9E3779B97F4A7C15)); // internal error
10.   Z := (X xor (X shr 30)) * UInt64($BF58476D1CE4E5B9);
11.   Z := (Z xor (Z shr 27)) * UInt64($94D049BB133111EB);
12.   Result := Z xor (Z shr 31);
13. end;
14.  
15. begin
16. end.

Much of the code already works, but I want the complete library to compile.

[Xor-el]

I was almost finished making it compatible, but I ran into an internal error 200706094, so I can't finish it before that is resolved.
In principle it was not much work. I reported the bug.

Code: Pascal  [Select][+][-]

1. {$MODE DELPHI}
2. {$RANGECHECKS OFF}
3. {$OVERFLOWCHECKS OFF}
4.  
5. function SplitMix64(var X: UInt64) : UInt64;
6. var
7.   Z: UInt64;
8. begin
9.   Inc(X, UInt64($9E3779B97F4A7C15)); // internal error
10.   Z := (X xor (X shr 30)) * UInt64($BF58476D1CE4E5B9);
11.   Z := (Z xor (Z shr 27)) * UInt64($94D049BB133111EB);
12.   Result := Z xor (Z shr 31);
13. end;
14.  
15. begin
16. end.

Much of the code already works, but I want the complete library to compile.

@Thaddy, I have encountered this problem quite a number of times and below is how I circumvented it. (by assigning that constant to a variable);

Code: Pascal  [Select][+][-]

1. function SplitMix64(var X: UInt64): UInt64;
2. var
3.   Z, K: UInt64;
4. begin
5.   K := UInt64($9E3779B97F4A7C15); // assign the constant to a variable first
6.   Inc(X, K);
7.   Z := (X xor (X shr 30)) * UInt64($BF58476D1CE4E5B9);
8.   Z := (Z xor (Z shr 27)) * UInt64($94D049BB133111EB);
9.   Result := Z xor (Z shr 31);
10. end;
11.  

[Thaddy]

I know the work-around, but the internal error should be fixed. It is legal code.

[MathMan]

I was almost finished making it compatible, but I ran into an internal error 200706094, so I can't finish it before that is resolved.
In principle it was not much work. I reported the bug.

Much of the code already works, but I want the complete library to compile.

Thanks Thaddy - that sounds very promising!

My objective is to do some benchmark comparison. If your current state of the library could support this objective I'd be glad to give it a try and report back my findings. In addition I'm also benchmarking against Wolfgang Ehrhardts arbitrary precision library.

Kind regards,
Jens

[MathMan]

I was almost finished making it compatible, but I ran into an internal error 200706094, so I can't finish it before that is resolved.
In principle it was not much work. I reported the bug.

Code: Pascal  [Select][+][-]

1. {$MODE DELPHI}
2. {$RANGECHECKS OFF}
3. {$OVERFLOWCHECKS OFF}
4.  
5. function SplitMix64(var X: UInt64) : UInt64;
6. var
7.   Z: UInt64;
8. begin
9.   Inc(X, UInt64($9E3779B97F4A7C15)); // internal error
10.   Z := (X xor (X shr 30)) * UInt64($BF58476D1CE4E5B9);
11.   Z := (Z xor (Z shr 27)) * UInt64($94D049BB133111EB);
12.   Result := Z xor (Z shr 31);
13. end;
14.  
15. begin
16. end.

Much of the code already works, but I want the complete library to compile.

Hi Thaddy - thanks again for looking into this subject. Any news on the bugfix and status of your porting activity?

Regards,
MathMan

[Thaddy]

The bug is still there. I will use the work-around and finish it.