postpone
{
        repeat $%% shr 1
                load a byte from :%-1
                load b byte from :$%%-%
                store byte a at :$%%-%
                store byte b at :%-1
        end repeat
}    

It adds $% and $%% built-in variables ...

I found yet few new instructions that are still missing, such as VDPBF16PS and similar.
https://software.intel.com/en-us/download/intel-architecture-instruction-set-extensions-programming-reference

For those that don't know fasmg's workings, what do these variables do? From the example code it looks like $%% is the emitted code length.

What is the difference between the "current offset" from $% and the "actual offset" from $%%? How does "current" or "actual" differ from the $ value?

So in most circumstances $% and $%% give the same value? But when there are uninitialised data at the current position then $% and $%% will have different values?

It adds $% and $%% built-in variables, and a syntax variant for LOAD/STORE that allows to use output file offsets instead of addresses.

These new features can only be used with (default) binary output format. If you use FORMAT directive with anything other than BINARY setting, you are not going to be able to use them.

As a quick demonstration, this snippet reverses the order of all bytes in the assembled output file:

Code:

postpone
{
        repeat $%% shr 1
                load a byte from :%-1
                load b byte from :$%%-%
                store byte a at :$%%-%
                store byte b at :%-1
        end repeat
}    

Well, for certain definitions of “readable”, ’cause, y’know, $%%, but there aren’t much alternatives to choose from anyway that would not break some existing code somewhere. So, thanks a lot.

DimonSoft wrote:

Well, for certain definitions of “readable”, ’cause, y’know, $%%, but there aren’t much alternatives to choose from anyway that would not break some existing code somewhere. So, thanks a lot.

Keep in mind that if you need just an offset in file where your piece of code/data is going to be put, you should use $%. And $%% is needed only for very specific purposes when you need to know the size of already defined data.

Yes, I looked through the docs. This one just looked a bit more bizarre.

DimonSoft wrote:

Yes, I looked through the docs. This one just looked a bit more bizarre.

For syntactical symmetry it probably should have been $@% or $%@, but these just look like expletives. So staying with no more than two symbols per name felt like a safer choice.

What is the policy for using symbols like that? Things related to addresses and generated data seem to start with $. As far as I can tell, we only have two things starting with %: % itself in loops and %t for current timestamp—some general purpose integer stuff. Am I missing something else? @ is used in anonymous labels. But what are the rules for new features?

"Here is some data" Wrong.
"Here are some data" Correct.

[/offtopic]