Problem with rept macro...

res equ 8 theta equ 6.283185307179f / res phi equ 3.141592653589f / res theta_array: rept res { dd % * theta ; Theta array } phi_array: rept res { dd % * phi ; Phi array }

Generally the "repeat" and "=" should be used instead for "rept" and "equ" where calculations are involved (see manual), but - as revolution already noted, fp calculations are not supported (for portability/SSSO reasons mainly, it agrees with NASM in this aspect).

This:

Code:

res = 8
theta = 6.283185307179f
repeat res
 dd % * (theta / res)
end repeat

is almost correct, but the calculation is performed on integers, the floating point value is by default 64-bit, thus the overflow happens. It you used:

Code:

theta = dword 6.283185307179f

the 32-bit represenation of fp value would be generated, so "theta" would get just the 40C90FDBh value - this code would even assemble, but such calculations are certainly not what you need.

It would be however possible to write some macroinstructions to perform fp operation using the integer ones. Just for the beginning, this can extract the mantissa and exponent from the 32-bit representation of floating point number:

Code:

macro get_fp value
{
 virtual at 0
  dd value
  load fp_value dword from 0
 end virtual
 mantissa = 800000h + fp_value and 07FFFFFh
 exponent = (fp_value shr 23) and 0FFh - 7Fh
 sign = fp_value shr 31
}

and this can use those components to make up the 32-bit representation of floating point value back:

Code:

macro store_fp name
{
 while mantissa >= 1000000h
  mantissa = mantissa shr 1
  exponent = exponent + 1
 end while
 while mantissa < 800000h
  mantissa = mantissa shl 1
  exponent = exponent - 1
 end while
 name = (mantissa - 800000h) + (exponent+7Fh) shl 23 + sign shl 31
}

I got it working this way:

Code:

theta equ 6.283185307179f
res = 8
theta_array:
repeat res
 get_fp theta
 mantissa = %*(mantissa/res)
 store_fp tmp
 dd tmp
end repeat

but it may loose some of the precision - the longer mantissa representation could be used to get better.