ftol and the 64bit Delphi inline asm changes
April 9, 2019 | Delphi Programming | No Comments
|
1 2 3 4 5 6 7 8 |
function _ftol( f: double) : Integer; cdecl; begin asm lea eax, f fstp qword ptr [eax] end; result := Trunc(f); end; |
Above is a Delphi inline asm version of the ftol (Floating Point to Integer Conversion) function used by our HotPDF Delphi PDF component.
When compiling above Delphi codes in the 64-bit target platform, Delphi will raise below error:
[dcc64 Error] E1025 Unsupported language feature: ‘ASM’
The reason is: The 64-bit Delphi compiler DCC64 does not support mixed Pascal/assembly code.
DCC64.EXE supports 64-bit assembly code, but the routines must be written completely in assembler.
In addition, the 64-bit asm codes have many different than the 32-bit ones, you can not copy the 32-bit asm codes to the 64-bit assembler routines directly, you must re-write them from scratch, I found below codes from the internet:
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 |
function _ftol: Integer; cdecl; // Assumes double value is in FPU stack on entry // Make a truncation to integer and put it into function result var TmpVal: Int64; SaveCW, ScratchCW: word; asm .NOFRAME fnstcw word ptr [SaveCW] fnstcw word ptr [ScratchCW] or word ptr [ScratchCW], 0F00h ;// trunc toward zero, full precision fldcw word ptr [ScratchCW] fistp qword ptr [TmpVal] fldcw word ptr [SaveCW] mov rax, TmpVal end; |
But maybe we do not really need assembly this routine,
|
1 2 3 4 |
function _ftol(f: double): integer; cdecl; begin Result := Trunc(f); end; |
The trunc function of Delphi just returns the integer part of a floating point number, that’s the conversion from floating point to integer. The speed of this version should be not bad.