In source I wrote

but in debugger (OllyDebugger) I see:

but why?
Other compillers find an error in this instruction...

in fasm must write
push dword 3.2

the op code 66 68 CD CC 4C 40 is:
push dword (66 68 ) 3.2 (404CCCCD)

see the floating point format to understand the encoding!

that is not true, and not what he was asking about

necromancer: FASM accepts floating point value in place of dword, and encodes it as single precision floating point number. "3.2" corresponds to 404CCCCD in single precision format.

it was what i was saying

push dword ( 66 68 ) push ( 68 ) dword, 32 bits (66 prefix)
3.2 ( 404CCCCD ) single precision format

bit 31 = sign
bits 0 to 22 = mantissa
bits 23 to 30 = exponent

1. why do you mention 66h prefix? that has nothing to do with problem. He didn't understood why FASM accepts floating point at place of dword.

2. you was talking about 16 bit code. In 32bit code you don'y have to use "dword", and no "66" prefix is generated. We are in "WINDOWS" section... why do you mix 66h prefix and 16 bit code here?

3. When describing format of number, you shouldn't forget about normalization, and about format of exponenet. Otherwise this info isn't very useful.

because for me it was very late

very tired

sorry

ok

ok... Thanks:)
I think I must read about floating point format...

Necromancer: why? you don't need to understand how are floating point numbers encoded to work with them using FPU

a good technician need to know the more possible about his tools and techniques

don't you better use the FPU and SSEx instructions since you know how it is encoded??

Yeah, understanding how it works is good, but NOT for someone who wants to learn how to use it.


i doubt understanding encoding of FP would affect how you write FPU code

No, I need to know how are they encoded.... I want to know it:(

yes, encoding is not important to write good code

but with binary ( integer ) it is very usefull to know the different values of bits
to make xor eax,eax you need to know the encoding of dwords

as to make fpn manipulation well, you need to know that exists NANs QNANs etc
and knowing the fp encoding is good to manipulate the exponents, the sign, the mantissa

no?

PS: i never code in fp because it's too hard for me
but a young beginer can probably use that really good by learning these kind of things!

first of all: why would you need to manipulate other way than by using FPU? If you use FPU, you don't need to know encoding. Encoding of floating point numbers is quite complicated.

Of course you need to know there is NAN, etc.

I'll try to find some information about it on the Net...

Necro: my advice: forget about how are FPs encoded, and learn how to work with them

necro: listen to vid, first learn how to use the fpu
vid: maybe one day he will want to know fp encoding?

Let's see: SSE doesn't have FABS, FCHS, FSCALE, nor FXTRACT instructions so to synthesize them you have to know the format of floating point numbers in XMM registers, whether single or double, and slice and dice them via logical and shift operations.

Knowing IEEE 754 formats is essential for SSE and one of the fundamental pieces of knowledge one should have if one wants to write x87 code competently. Just think about how any introductory book on assembly language has a chapter on binary, octal, 2's complement arithmetic and the like. Introductory treatises on the x87 FPU similarly have a discussion on IEEE 754 formats. Look in 253665.pdf, for example in section 4.8 Real Numbers and Floating-Point Formats.

I've already understood the encoding:)