while trying to code for FPu, i saw that it is unpossible to make math expressions to declare FP datas.

You are telling this as a bug or feature request?

feature request.
it's not a bug because it's not provided by FASM

yes. FP data operations are not so easy to implement, and might be potential threat to SSSO principe.

My suggestion is to precompute value and use comment.

How? as an assembler which has an 80386 as minimun requirement then it would be implemented with no FPU so there is no room for minimun discrepancies that can be found between CPU (or emulators) vendors or even differencies between CPU models of same vendor.

Loco: FASM doesn't use FPU, it uses custom FP format, to assure result is precise. (I know because Tomasz's code "inspired" me in fp2a and a2fp routines for FASMLIB, and for bignum article too.). FASM has to use custom format, because you can't do bunch of operations with 64bit mantissa and then expect all 64bits of result to be valid.

You are right, I said it wrong. It would cause loss of precision. If all FPUs are quaranteed to give same result for same input, then SSSO wouldn't be broken.

I've never said it uses FPU and also discouraged the idea of using it because of the minimal FPU implementation differencies and the fact that it would add an extra minimun requirement. I have said implementing it with no FPU (no FPU instructions at all).

sorry, misread your post.

Implementing something like "custom FP format division" would be really hell i quess. Multiplication is feasible, addition and subtraction is easy.

I am not sure, maybe tomasz with his maths knows some way how to do division "easy way" too, but I don't. I would file this as feature request, but reality is that there are many more important things to come, so I would postpone this until those are done.