flat assembler
Message board for the users of flat assembler.

Index > Compiler Internals > label vs virtual directive

Author
Thread Post new topic Reply to topic
BiDark



Joined: 22 Jun 2003
Posts: 109
Location: .th
BiDark
which is consumes more cpu time to finish the job (when the labels is fit in range). for example


Code:
  label a dword at ...

  and

  virtual at ... 
    a dd ?
  end virtual
    


which one?
Post 30 Nov 2003, 10:01
View user's profile Send private message Reply with quote
vid
Verbosity in development


Joined: 05 Sep 2003
Posts: 7105
Location: Slovakia
vid
i am not provalov so maybe i am wrong, but label-at only defines displaced label, while second thing starts, virtual section, defines displaced label too (displaced by argument of virtual), defines virtual data, and end virtual section.

btw: privalov: how did you do 'load' directive. Is whole virtual section compiled to some other buffer from where it is loaded, or is memory area which should be loaded marked somehow and assembled in next step or how?
Post 30 Nov 2003, 11:51
View user's profile Send private message Visit poster's website AIM Address MSN Messenger ICQ Number Reply with quote
BiDark



Joined: 22 Jun 2003
Posts: 109
Location: .th
BiDark
Thx.
Post 02 Dec 2003, 02:17
View user's profile Send private message Reply with quote
Tomasz Grysztar



Joined: 16 Jun 2003
Posts: 7797
Location: Kraków, Poland
Tomasz Grysztar
"virtual" itself only changes the addressings for the following code, as it would the "org" do, and code is assembled normally until the "end virtual" is reached - then assembler moves code generation pointer back to the state at which it was when the "virtual" block was started - so it data will be overwritten by the new, real code.
And "load" directive allows you to load from addresses from what I called "current addressing space", that is from the area beginning at the last addressings change, like "org", "virtual", "section", etc.
Post 02 Dec 2003, 09:14
View user's profile Send private message Visit poster's website Reply with quote
vid
Verbosity in development


Joined: 05 Sep 2003
Posts: 7105
Location: Slovakia
vid
smart Idea
but what about virtual at ebx? Can you also do "label name at ebx"?
Post 02 Dec 2003, 10:33
View user's profile Send private message Visit poster's website AIM Address MSN Messenger ICQ Number Reply with quote
Tomasz Grysztar



Joined: 16 Jun 2003
Posts: 7797
Location: Kraków, Poland
Tomasz Grysztar
Yes. Why don't you just try? Wink
Post 02 Dec 2003, 11:33
View user's profile Send private message Visit poster's website Reply with quote
vid
Verbosity in development


Joined: 05 Sep 2003
Posts: 7105
Location: Slovakia
vid
it is easier to ask, you know... Embarassed

i am just curious how you made it. I can imagine you can remember offset for each label - of course - but how can you remember value like 4*ebx + eax + 3? I just find it quite more dificult, so i wasnt sure.
Post 02 Dec 2003, 17:56
View user's profile Send private message Visit poster's website AIM Address MSN Messenger ICQ Number Reply with quote
Tomasz Grysztar



Joined: 16 Jun 2003
Posts: 7797
Location: Kraków, Poland
Tomasz Grysztar
OK, perhaps you'll have to wait long for it to be covered in the official fasm internals documentation, so here is a little gift - the structure of fasm's symbol entry (this is the 24-byte data structure used commonly for all labels, numerical constants and segments):
Code:
Offset     Size    Description

    +0    qword    value of symbol (signed)
    +8    word     flags, any combination of the following values:
                     1 - symbol was defined
                     2 - symbol can be redefined (set only for symbols
                          defined with "=")
                     4 - symbol was redefined (when set, no forward
                          references to this symbol are allowed)
                     8 - symbol was used
                     10h - symbol was checked for being used
                            with prediction in current pass
                     20h - result of last predicted check for being used
                     40h - symbol was checked for being defined
                            with prediction in current pass
                     80h - result of last predicted check for being defined
   +10    byte     size of data labelled by this symbol
   +11    byte     type of symbol, any of the following values:
                     0 - absolute value
                     1 - relocatable segment address (only with MZ output)
                     2 - relocatable 32-bit address
                     3 - relocatable relative 32-bit address
                          (only valid for symbol that is used in the same
                          section where it was calculated, so it should not 
                          occur in the regular label structure)
   +12    dword    extended SIB, first two bytes are register codes
                   and second two bytes are corresponding scales
   +16    word     number of pass in which symbol was defined last time
   +18    word     number of pass in which symbol was used last time
   +20    dword    address of object section or external symbol to which
                    this symbol is relative (only for relocatable symbols
                    with object output)
    

This is the most actual data for the 1.50 release.


Last edited by Tomasz Grysztar on 04 Dec 2003, 10:27; edited 2 times in total
Post 02 Dec 2003, 19:05
View user's profile Send private message Visit poster's website Reply with quote
vid
Verbosity in development


Joined: 05 Sep 2003
Posts: 7105
Location: Slovakia
vid
really thanks, by the way si it problem to do something like
+24 pointer to symbol name
????
Post 03 Dec 2003, 18:28
View user's profile Send private message Visit poster's website AIM Address MSN Messenger ICQ Number Reply with quote
Tomasz Grysztar



Joined: 16 Jun 2003
Posts: 7797
Location: Kraków, Poland
Tomasz Grysztar
The size of the structure would have to be changed.
Post 03 Dec 2003, 20:52
View user's profile Send private message Visit poster's website Reply with quote
JohnFound



Joined: 16 Jun 2003
Posts: 3502
Location: Bulgaria
JohnFound
Hi Privalov.
I have a question about label data. On which stage this data is created and what fields of it are filled initially?

Regards.
Post 04 Dec 2003, 05:03
View user's profile Send private message Visit poster's website ICQ Number Reply with quote
Tomasz Grysztar



Joined: 16 Jun 2003
Posts: 7797
Location: Kraków, Poland
Tomasz Grysztar
Parser allocates addresses for those structures, and those addresses become the IDs for the labels (parser put the appropriate ID for each label into the fasm code, so assembler doesn't have to deal with names anymore). At this stage the memory that is reserved for them is used for other purposes, but parser only allocates the address. The 24-byte blocks are allocated from the top of main memory block, and assembler, before starting the first pass, fills all that area with zeros.
Post 04 Dec 2003, 10:02
View user's profile Send private message Visit poster's website Reply with quote
JohnFound



Joined: 16 Jun 2003
Posts: 3502
Location: Bulgaria
JohnFound
Thank you Privalov.
Another question: Is there some reference to the char name of the label, maybe not in this table but in preprocessed/parsed code?

Regards.
Post 04 Dec 2003, 10:41
View user's profile Send private message Visit poster's website ICQ Number Reply with quote
Tomasz Grysztar



Joined: 16 Jun 2003
Posts: 7797
Location: Kraków, Poland
Tomasz Grysztar
At the parser stage the table of labels consists of 8-byte records, each containing the address of name string and the hash for label. When parser finds a record in this table that corresponds to the label for which ID it has to retrieve, it calculates the offset in table and multiplies it by 3, so the offset in the table of 24-byte records is obtained - and it makes an ID being the address of the 24-byte structure in table, which at assembly time replaces the old table of 8-byte records. Smaller record size is used at parser stage, because it makes hash scanning faster with data cache.
As the same memory is used for both tables, those references are of course lost at the assembly stage.
Post 04 Dec 2003, 10:50
View user's profile Send private message Visit poster's website Reply with quote
Display posts from previous:
Post new topic Reply to topic

Jump to:  


< Last Thread | Next Thread >
Forum Rules:
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You cannot attach files in this forum
You can download files in this forum


Copyright © 1999-2020, Tomasz Grysztar. Also on GitHub, YouTube, Twitter.

Website powered by rwasa.