push dword ptr fs:[0] 64 67 FF 36 00 00
\/
how do i do it in fasm?
fasm
push dword ptr fs:[0] 64 FF 34 00 00 00 00
makes it a large dword ptr fs:0

any help would be appreciated

And if you are under 32bit there is no hint needed.
just push [fs:0]
^o) or is there sth I missed (I think there should be no harm even if FSelector is 16bit)

EDIT: There is a bug. According to manuals - FASM should choose the shortest one, but PUSH DWORD[FS:0] assembles to:

which is 7 bytes
The shortest form of this instruction is:

with 6 bytes (WOW - it uses TWO prefixes)
I think that assembler get fooled by this 2pre+2decode byte, but HERE the extra prefix pays off because the 0-data is 2 bytes shorter.
Strange thing is that I checked and even OllyDbg doesn't know this short form. It disassembles, when it meets it, but if you assemble with OllyDbg, it assembles to 7-byte version (I think its faster because of one less prefix).

fasm doesn't use 16-bit addressing in 32-bit mode unless you tell to do it (with "push [fs:word 0]"), because the prefixed instructions even if shorted, usually were decoded slower. The earliest versions of fasm did such optimization, but later I've decided that this was the bad choice.

However I have just noticed that in the latest releases, during making the changes to adapt the assembler for future integration of AMD64 support, I have introduced a small bug, which caused fasm to ignore the size operator for address (so "push [fs:word 0]" is by fasm 1.56 assembled exactly the same as "push [fs:0]). I am fixing it now, look out for the new release.

If I understood correctly - push [fs:word 0] should assemble to 6-byte operation/command, but it still pushes DWORD to the stack.
To my knowledge this command should push word (I mean written exactly as push [fs:word 0]), but I fear to argue here because I know that ADD, SUB, AND, ... have shorter forms but still dealing with 4-byte variables (m32/r32).

Just checking - did I understand correctly?

johnfound: i thought "pushd" is just a macro

As I wrote above, there is a bug in 1.56 causing it - check with some earlier release, or with the 1.57 - which is coming out.


To push word use the "push word [fs:0]" command, everything inside the square brackets define the address on which you operate, so if you put size operator there, it applies to address size.