MasmLib FloatToStr does not work for me

[jj2007]

You are provided with a clean FPU at the beginning of your program. However, the precision control is only set for 64-bit double-precision floats by Windows. If you need extended double-precision (full 80 bits precision), either you modify the Control Word accordingly or use finit.

I tested the state of the FPU at program start. To my surprise, I found this (using OllyDbg 2.0):

EDIT: I found the answer:
53 means the precision that can be applied to a Real8, 64 bits (this is what Windows gives you)
64 means the precision that can be applied to a Real10, 80 bits (this is what finit gives you)
So there is no "full 80 bits precision"  :wink

- at ModuleEntryPoint : FCW 027F, NEAR, 53
- after performing finit : FCW 037F, NEAR, 64

5432109876543210
0000001001111111 027F
0000001101111111 037F

This is Windows XP SP2...
Any idea why this differs from your para, and/or where the exact behaviour is documented?
Thanks, JJ

Code Select Expand

include \masm32\include\masm32rt.inc
.code
start:
int 3 ; set a breakpoint in OllyDbg
finit
exit
end start


[raymond]

80-bit floats is the maximum precision you can get with the current FPU architecture. It consists of:
- 64 bits for the mantissa
- 15 bits for the biased exponent
- 1 bit for the sign.