GR32_LowLevel.Mirror PUREPASCAL does not produce expected values

[andersmelander]

GR32_LowLevel.Mirror PUREPASCAL does not produce expected values

[CuriousKit]

How goes the progress in fixing this?

[andersmelander]

I'm busy with other stuff so I'm not working on it (which is why I haven't assigned it to myself)

[CuriousKit]

That's fair. I saw that you made a commit related to the bug and thought initially that the bug had been fixed, but a closer look at the commit messages indicates that it's not all fixed yet. Still, these messages will indicate that the issue is still open.

[andersmelander]

For reference, what we need is a replacement for the current (broken) algorithm:

function MirrorPow2(Value, Max: Integer): Integer; overload;
begin
  if Value and (Max + 1) = 0 then
    Result := Value and Max
  else
    Result := Max - Value and Max;
end;

MirrorPow([0..13], 3)
Expected: 0123210123210
Actual: 0123210301232

[CuriousKit]

I'll take a study of the file myself at some point. I'm curious to know what the function is supposed to do exactly.

[andersmelander]

The Clamp, Wrap, and Mirror functions are primarily used when sampling bitmaps (e.g. resizing, rotating, transforming, etc). They specify how out of bounds pixel coordinates should be handled.
https://graphics32.github.io/Docs/Units/GR32/Types/TWrapMode.htm

[CuriousKit]

Ah I see, thank you. Having played with textures in OpenGL, I can definitely see the importance of getting this function correct then!

[andersmelander]

...and also used by some gradient functions: https://graphics32.github.io/Docs/Units/GR32_ColorGradients/Classes/TCustomGradientPolygonFiller/Properties/WrapMode.htm

[andersmelander]

I can definitely see the importance of getting this function correct then!

I would think so but the bug has been there since the functions introduction in 2007...

[lamdalili]

Sorry I didn't notice that it was resolved in new branch

graphics32/Source/GR32_LowLevel.pas

Lines 1255 to 1265 in a322add

	function Mirror(Value, Max: Integer): Integer;
	var
	Quotient: Integer;
	begin
	if Value < 0 then
	Value := -Value;
	Quotient := DivMod(Value, Max, Result);
	if ((Quotient and 1) = 1) then
	Result := Max - Result;
	end;

Test on GradFills example :
Original mirror :

After correction

[andersmelander]

@lamdalili Thanks but the Mirror function has already been fixed in the LowLevel.Mirror branch:

graphics32/Source/GR32_LowLevel.pas

Lines 1255 to 1265 in a322add

	function Mirror(Value, Max: Integer): Integer;
	var
	Quotient: Integer;
	begin
	if Value < 0 then
	Value := -Value;
	Quotient := DivMod(Value, Max, Result);
	if ((Quotient and 1) = 1) then
	Result := Max - Result;
	end;

What remains to be fixed is the MirrorPow2 function:

graphics32/Source/GR32_LowLevel.pas

Lines 1321 to 1330 in a322add

	function MirrorPow2(Value, Max: Integer): Integer; overload;
	begin
	if Value < 0 then
	Value := -Value;
	if ((Value and (Max + 1)) = 0) then
	Result := (Value ) and Max
	else
	Result := (Max - Value + Max) and Max;
	end;

See: #291 (comment)