chatgpt can convert 32bit asm into 64bit

[RDP1974 40]

question

do you remember the good old https://fastcode.sourceforge.net/ ? (2007 project)

I asked chatgpt to convert asm32 to asm64 delphi, look at this example:

 

procedure MoveJOH_SSE2_10(const Source; var Dest; Count : Integer);
asm
  cmp     ecx, TINYSIZE
  ja      @@Large {Count > TINYSIZE or Count < 0}
  cmp     eax, edx
  jbe     @@SmallCheck
  add     eax, ecx
  add     edx, ecx
  jmp     SmallForwardMove_10
@@SmallCheck:
  jne     SmallBackwardMove_10
  ret {For Compatibility with Delphi's move for Source = Dest}
@@Large:
  jng     @@Done {For Compatibility with Delphi's move for Count < 0}
  cmp     eax, edx
  ja      Forwards_SSE2_10
  je      @@Done {For Compatibility with Delphi's move for Source = Dest}
  sub     edx, ecx
  cmp     eax, edx
  lea     edx, [edx+ecx]
  jna     Forwards_SSE2_10
  jmp     Backwards_SSE2_10 {Source/Dest Overlap}
@@Done:
end; {MoveJOH_SSE2}

 

chat gpt rewrite this in 64bit:

procedure MoveJOH_SSE2_10(const Source; var Dest; Count: Integer);
asm
  cmp     r8d, TINYSIZE        // Compare Count (R8D) with TINYSIZE
  ja      @@Large              // If Count > TINYSIZE or Count < 0, jump to @@Large
  cmp     rcx, rdx             // Compare Source (RCX) with Dest (RDX)
  jbe     @@SmallCheck
  add     rcx, r8              // Adjust Source address: Source + Count
  add     rdx, r8              // Adjust Dest address: Dest + Count
  jmp     SmallForwardMove_10  // Jump to small forward move
@@SmallCheck:
  jne     SmallBackwardMove_10 // If Source ≠ Dest, use backward move
  ret                           // If Source = Dest, exit (compatible with Delphi’s Move)
@@Large:
  jng     @@Done                // If Count < 0, exit (compatible with Delphi’s Move)
  cmp     rcx, rdx
  ja      Forwards_SSE2_10      // If Source > Dest, use forward move
  je      @@Done                // If Source = Dest, exit
  sub     rdx, r8               // Adjust Dest address: Dest - Count
  cmp     rcx, rdx
  lea     rdx, [rdx + r8]       // Restore original Dest address
  jna     Forwards_SSE2_10      // If no overlap, use forward move
  jmp     Backwards_SSE2_10     // If overlap, use backward move
@@Done:
end;
 

opinions? does this make sense to you? can we take the winners of fastcode 32bit and convert them in 64bit so easily?
chatgpt then owns the copyright?

[DelphiUdIT 218]

For something to be patentable it must have a creative and innovative character.
I don't think that a few lines of assembler can be considered as such.

[Brian Evans 112]

Note it is a 64-bit version of a 32-bit move function not a 64-bit move function. 

 

A 64-bit move function would have Count as: NativeInt and other changes to allow moves over the limits of a 32-bit integer.

Edited November 25, 2024 by Brian Evans

[Stefan Glienke 2083]

Converting source code that won any challenge over 15 years ago is questionable regardless of correctness.

 

Anyway, for this particular example, implementing a faster System.Move for Windows (other platforms use their libc memory) has been solved since Delphi 11.3.

 

I sincerely challenge everyone to come up with a faster/better implementation.

Edited November 25, 2024 by Stefan Glienke

[RDP1974 40]

indeed the quality of the system rtl of D12 is outstanding