| |
| |
| |
|
Comments:
<0> heheh
<1> When a process is started, can it ***ume the floating-point stack is empty? When an external library call is made?
<1> A better question.. I have a asm-coded object file that I want to link into a bigger C-coded project. It needs to use 6 floating-point registers per call. How can I be sure that six are available?
<2> http://x-os.homeip.net/x-os/hdir/pdf/sysv/psABI-i386.pdf
<2> you can be sure.
<1> reading...
<2> it's just that you shouldn't call any C code while stack is non-empty.
<1> noted. that's won't be a problem ;-)
<1> *that
<3> let's go the bed
<1> I have an object file compiled using 'as' under windows. I am linking it into a C project. I call one function in the asm object file (_asm_function) from C. I am getting an "undefined reference to _asm_function) from the linker. Under linux, all I had to do to export _asm_function was add .global _asm_function below .section .text Is there something else I need to do under win32?
<4> WHERE can i learn more about how/what a linker like ld does besides the gnu ld manual?
<4> (not sure ld manual goes into much details i'm interested in)
<5> I can tell you exactly what a linker does
<4> __mikem: great, thanks...the reason i'm asking is cuz i'm having problems combining my toy C kernel with my toy boot loader i wrote
<5> Do you want to know about dynamic linking or static linking
<4> __mikem: i guess static would be good enough...if my kernel was in asm then i could just append it to boot loader
<5> ok static linking is relatively streight forward
<4> __mikem: actually...i'd have to join the data/text/bss sections into one...etc.
<5> when the code is ***embled, symbols that represent certain functions are left behind. The linker looks at those symbols and tries to find the respective code in the static libraries.
<5> the code from the static libraries is "included" in the final output file
<4> __mikem: ok
<5> theres a bit more
<5> when the code is included, since there are a lot of sections in the file that come AFTER the executable code section, refferences to data in that section have to be updated. This is done in the relocation table
<5> actually that is an over simplistic explaination of the relocation table, but it really isn't important to know about that
<5> in reality, everything is moved around, not just the sections that have to move to make room for the additional code
<4> __mikem: doesn't every included piece have its own data/text/bss section? seems like quite a feat to combine all of 'em into one super text/data/bss chunk
<5> I am actually not sure how that works.
<5> Seb, it is posible to have more than one section marked 'executable' but I have never actually seen this when examening dis***embled code
<5> Dynamic linking is a bit different, in the PE format, its accomplished with a jump table
<4> __mikem: there is a ld linking script on 3 page of http://osdever.net/bkerndev/index.php?the_id=90
<5> THis jump table lists the name of each function in plain text. Each function name is listed under its library name
<4> __mikem: there are commands like . = ALIGN(4096) ...i'm not sure what is purpose of "aligning" code
<5> code that is aligned on addresses that are multiples of four require less time to execute
<4> __mikem: i like your good info....WHERE did you read/learn about this stuff? got any good books to recommend?
<5> seb- I have a lot of time on my hands
<4> __mikem: GOOD! aligning is just an optimization!! yay! something a newbie can ignore for now!
<4> __mikem: so do i but i still need a *source* to learn from during all that free time!
<5> lol, I really wouldn't say that. Optomization is one of the primary reasons to learn ***embly language in the first place now days
<5> seb- I learned about pe format off of documentation on msdn, I had to learn about linkers in order to understand what I was looking at. I was building my own dis***embler at the time
<4> __mikem: what is pe?
<5> Portable Executable Format. Its the binary format windows uses for executable files
<5> Mac also uses it
<1> really?
<5> Mac OS I mean
<1> I thought mac had their own
<5> nope
<5> or if they do, they switched after the documentation on msdn was writen
<1> I'd hazard a guess that Mac is ELF since OS X
<1> or whatever BSD uses
<5> It wouldn't surprise me
<5> But, in all fairness, PE is a prety elegent format
<1> anyways... I need to repost my question
<4> __mikem: are you sure PE doesn't stand for "pretty elegant" ?
<1> I have an object file compiled using 'as' under windows. I am linking it into a C project. I call one function in the asm object file (_asm_function) from C. I am getting an "undefined reference to _asm_function) from the linker. Under linux, all I had to do to export _asm_function was add .global _asm_function below .section .text Is there something else I need to do under win32?
<1> heh heh
<5> Probably a different object file format
<5> seb positive
<5> diodesign why are you doing that?
<6> appears to be xcode's default output with gcc
<5> but I can't see why mac osx would use Mach-0
<1> ah... I got it. Win32 has some stupid requirement about symbol names
<7> Hi
<7> I have an array, and a variable, the variable used as an offset in the array.
<7> How can I write to the array ?
<7> I guess mov DWORD PTR Array + DWORD PTR Offst, 42; won't work...
<4> Axioplase: you want lea a, [b]
<4> Axioplase: you can't do math outside of []'s
<4> such as addition
<7> lea [DWORD PTR Array + DWORD Ptr Offst], 42 ?
<7> I still guess there are too many memory operands...
<7> (I want to do it with as few instructions as possible)
<4> Axioplase: that is what lea is for...to use a few opcodes as possible
<4> Axioplase: if you were concerned with optimization you could add Array and Offset in another step first and store result in eax
<4> not*
<4> were not*
<7> If I were not, yes I would...
<7> hum... What if I use a real pointer and not an offset...
<4> Axioplase: are you trying to step thru an array?
<4> Axioplase: because you might like lodsb and friends if so
<7> seb-, well... yes.. filling an array/list, with a few bunch of elements at time...
<7> I ll get a look a lodsb then, thanks
<7> Hum... I only wish it could use something else than al/ax/eax ...
<4> Axioplase: why?
<7> seb-, because I'll have to save eax before using it...
<7> I didn't choose where data is sotred (reg or stack).
<7> I wrote a register allcoation program for that..
<7> And therefore, chances are there is some Important data in eax that I'll have to save/restore each time I wan't to append something to my array.
<4> Axioplase: then perhaps a loop with some lea's is preferred
<7> maybe...
<7> argh...
<7> I have defined an area to be used later....comm HEAP,4000000,32
<7> How can I get a pointer to this area ?
<7> mov eax, HEAP dosn't work... (it moves 0 in eax)
<8> lea eax, HEAP
<7> haa..
<7> seb-, err.. would you mind telling me if those (commented) 10 lines of code really do what I want them to do ? http://cpp.sourceforge.net/?show=19321
<7> because I ddidn't succeed in incrementing my pointer in the array..
<9> hello
<9> I wonder... in AT&T syntax, is jmpl == ljmp?
<9> I only know about ljmp (far jump) -- is jmpl the same (and with the same syntax)?
<10> ljmp? what ***embly is that?
<9> I try to understand this code: http://kernel25.sourcesdb.com/r.php?num=1094
<9> x86
<10> there is really no ljmp in x86 afaik, just jmp far, you mean that?
<9> yes, JUMP FAR in Intel syntax it is written as ljmp in gas syntax (gcc ***embler)
<9> JMP*
<10> hmm, k, sorry that I can't be of much help, not too familiar with GAS.
<9> I see this: jmpl$BOOTSEG, $start2 -- and from what I understand, is a far jump (because the code is 16 bit) to 0x07C0:start2
<9> ok, thanks anyway :)
<10> yeah, that's a far jump because it involves a segment selection
<11> http://rafb.net/paste/results/CPEVbE69.html the idct_row for coldfire line 327, I am not sure if the (int_16 * block) pointer is p***ed appropriately to the asm function. I use the "a" modifier, to force gcc to use an adress operand. I am also not sure the words are in the appropriate order (big endian or not)
<11> that's coldfire ***embly
<12> seems pretty much right
<2> however, you use quite a lot registers... gcc could run out of them if you use it without optimisation options
<2> also, i'm not quite sure if [block]"a"(block) is the right syntax... but then, i never used it, so don't rely on this info
<11> what do you mean ?
<11> mwk: I tell gcc to not use this register
<11> I have not other choice by the way
<2> ok.
<11> mwk there is a problem indeed
<11> the screen is garbeled
<11> though it's supposed ot be the copy of the C part in asm
<11> I believe I have a problem with bits order
<11> something like that
Return to
#asm
or
Go to some related
logs:
#oe
#perl
ubuntu bbo
yapgvb StringIO
error opening security policy ubuntu vncserver
#php
tombrt
kismet autoconfig
#perl
#web
|
|