flat assembler 1.67

Index > Main > flat assembler 1.67

Goto page 1, 2 Next

Author

Thread

Tomasz Grysztar

Joined: 16 Jun 2003
Posts: 7845
Location: Kraków, Poland

Tomasz Grysztar

The new development line has started with the 1.67.0 release, with the main trigger being this thread: http://board.flatassembler.net/topic.php?t=5458
Look there also for more information on the changes made in the new release.

The fact that PE files are now generated with non-relocatable addresses unless you put fixups directory somewhere allows now to do all the sophisticated things with the addresses that are not allowed in case of relocatable ones, like:

Code:

mov eax,-$

or:

Code:

mov eax,MessageBox xor 12345678h

30 Jun 2006, 11:58

Tomasz Grysztar

Joined: 16 Jun 2003
Posts: 7845
Location: Kraków, Poland

Tomasz Grysztar

1.67.1: added the few new instructions of the so-called AMD Pacifica.

30 Jun 2006, 20:14

LocoDelAssembly
Your code has a bug

Joined: 06 May 2005
Posts: 4633
Location: Argentina

LocoDelAssembly

Nice, time to contact Joanna Rutkowska to tell her of the new addition Wink

30 Jun 2006, 21:07

comrade

Joined: 16 Jun 2003
Posts: 1137
Location: Russian Federation

comrade

Thanks Privalov.

That xor thing is actually quite useful, and I used to even hack it in on a compiled executable (back when i was using masm) to do something like this:

Code:

if (font code = bold)
xor [fontPointer], (boldFont XOR regularFont)
end if

30 Jun 2006, 23:52

Adam Kachwalla

Joined: 01 Apr 2006
Posts: 150

Adam Kachwalla

What files you need to modify if you want to add extra prefixes such as 0b (binary numbers)?

locodelassembly wrote:

Nice, time to contact Joanna Rutkowska to tell her of the new addition

Who is Joanna Rutkowska?

02 Jul 2006, 10:48

HyperVista

Joined: 18 Apr 2005
Posts: 691
Location: Virginia, USA

HyperVista

@Adam Kachwalla - re: "who is Joanna Rutkowska"

http://board.flatassembler.net/topic.php?t=5473

02 Jul 2006, 13:58

LocoDelAssembly
Your code has a bug

Joined: 06 May 2005
Posts: 4633
Location: Argentina

LocoDelAssembly

Shouldn't be xchg al/eax/rax, al/eax/rax be assembled as NOP in all cases? Since the assembler encodes optimizing for size it should always use 90h, no?

Regards,
LocoDelAssembly

02 Jul 2006, 16:15

vid
Verbosity in development

Joined: 05 Sep 2003
Posts: 7105
Location: Slovakia

vid

loco: how about xor [something], 0 then. should it be assembled to nop too? or should it be assembled at all? how far do we want to go?

02 Jul 2006, 16:30

LocoDelAssembly
Your code has a bug

Joined: 06 May 2005
Posts: 4633
Location: Argentina

LocoDelAssembly

Vid, there is a big difference:

Software Optimization Guide for AMD64 Processors wrote:

XCHG AX/EAX/RAX, AX/EAX/RAX/(R8 ) 90h DirectPath ~0 5
(NOP)

5. These instructions have an effective latency as shown. They map to internal NOPs that can be issued at a rate of
three per cycle but do not occupy execution resources.

But the problem is that if you write xchg ax, ax under use32 it will assemble the prefix for operand size override and the same with xchg eax, eax under use16. About AL sorry, I was wrong Razz

Regards

02 Jul 2006, 17:11

vid
Verbosity in development

Joined: 05 Sep 2003
Posts: 7105
Location: Slovakia

vid

alright then, but then it is not question of size optimizing, but question of beheaving due to manuals

02 Jul 2006, 17:48

LocoDelAssembly
Your code has a bug

Joined: 06 May 2005
Posts: 4633
Location: Argentina

LocoDelAssembly

Why is not a question of size optimizing?

Current version do this:

Code:

use32
xchg ax, ax ; 66 90
xchg eax, eax ; 90
nop ; 90

use16
xchg ax, ax ; 90
xchg eax, eax ; 66 90
nop ; 90

But I think it should be:

Code:

use32
xchg ax, ax ; 90
xchg eax, eax ; 90
nop ; 90

use16
xchg ax, ax ; 90
xchg eax, eax ; 90
nop ; 90

Unless that there is a reason to keep the actual behavior (possibly there is one but I don't know).

02 Jul 2006, 18:04

Tomasz Grysztar

Joined: 16 Jun 2003
Posts: 7845
Location: Kraków, Poland

Tomasz Grysztar

Note that in long mode "xchg eax,eax" is not the same and NOP.
As for prefixes it's just a matter of consistency. I prefer to be able to generate 16-bit variant just for the sake of completness - it's actually a way of following the principles. And the agressive optimization is currently not fasm's principle - see also http://board.flatassembler.net/topic.php?t=1238 etc.

02 Jul 2006, 20:30

LocoDelAssembly
Your code has a bug

Joined: 06 May 2005
Posts: 4633
Location: Argentina

LocoDelAssembly

OK, however it's a little strange since the architecture considers 90H as NOP, so are you defining a 16-bit variant of NOP under use32 and viceversa? I posted this because I think it's not some kind of aggressive optimization like the suggested in the link you posted.

Regards,
Hernán

PS: Yes, I know I can use macros to produce my proposed behavior Razz

02 Jul 2006, 21:27

Tomasz Grysztar

Joined: 16 Jun 2003
Posts: 7845
Location: Kraków, Poland

Tomasz Grysztar

Well, perhaps I should encode "xor ax,ax" as 66-33-C0 in 32-bit mode... Wink

02 Jul 2006, 21:49

Adam Kachwalla

Joined: 01 Apr 2006
Posts: 150

Adam Kachwalla

In EXPRESSI.INC, in the number_begin label:

Code:

        mov     ebx,esi
        add     esi,ecx
        push    esi
        dec     esi
        mov     dword [edi],0
        mov     dword [edi+4],0
        cmp     byte [ebx],'$'
        je      pascal_hex_number
        cmp     word [ebx],'0x'
        je      get_hex_number

I have added this after the last line of the code above:

Code:

        cmp     word [ebx],'0b'
        je      get_bin_number

and it still isn't working. What do I have to do to make it work?

03 Jul 2006, 03:36

Tomasz Grysztar

Joined: 16 Jun 2003
Posts: 7845
Location: Kraków, Poland

Tomasz Grysztar

Look at the checks for 'x' and '$' after the "get_hex_number" label. Do analogously.

03 Jul 2006, 08:44

Adam Kachwalla

Joined: 01 Apr 2006
Posts: 150

Adam Kachwalla

All I have to modify is the EXPRESSI.INC file, right?

03 Jul 2006, 08:50

Tomasz Grysztar

Joined: 16 Jun 2003
Posts: 7845
Location: Kraków, Poland

Tomasz Grysztar

Yes, only the "get_number" routine.

03 Jul 2006, 09:21

Garthower

Joined: 21 Apr 2006
Posts: 158
Location: Ukraine

Garthower

2Tomasz Grysztar: Where I can download last build of FASM 1.67? On http: // flatassembler.net/download.php version 1.66 is laid out only...

03 Jul 2006, 10:36

Tomasz Grysztar

Joined: 16 Jun 2003
Posts: 7845
Location: Kraków, Poland

Tomasz Grysztar

Don't you have it cached somehow?

03 Jul 2006, 10:53

Goto page 1, 2 Next

< 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