prime numbers

this is an exercies i set out to do in order to familiarise my self with linux and AT&T assembly syntax.

i worked a lot and read a lot before figureing out how each operator gets its paramaters and such...

the code doesn't compile with the following errors:

# as -gstabs sieve.s -o sieve.o
# gcc sieve.o -o sieve
sieve.o(.text+0x1):sieve.s:20: relocation truncated to fit: R_386_8 .data
sieve.o(.text+0xc):sieve.s:24: relocation truncated to fit: R_386_16 .data
sieve.o(.text+0x19):sieve.s:32: relocation truncated to fit: R_386_8 .data
sieve.o(.text+0x1b):sieve.s:33: relocation truncated to fit: R_386_8 .data
sieve.o(.text+0x2d):sieve.s:43: relocation truncated to fit: R_386_8 .data
sieve.o(.text+0x39):sieve.s:50: relocation truncated to fit: R_386_8 .data
collect2: ld returned 1 exit status

the source code is:

# assembly code for the calculation of the primes numbers in the range of [2-100]
# the code brute forces the numbers out, no formulas are taken in consideration

*.include* "defines.h" #system call numbers are stored here (i.e. __NR_<oper> = number)

*.global* main

*.data*

sieve:

	*.byte* 4
test:
	*.byte* 1

*.text*

main:

loop:
movb $test, %dl
incb %dl #increase test
movb %dl, (test)

movw $sieve, %ax #prepare for divb: move sieve in ax
divb %dl #div ax by dl

xor %cx, %cx #clear cx
movb %ah, %cl #move rest in cx

jcxz next #test if rest is 0 and increase if so

movb $sieve, %al
movb $test, %bl
cmpb %al, %bl #else test if "test" has reached sieve

jnz loop #if not go back and div by new test value

print: movl $__NR_write, %eax #else print the prime

	movl $1, %ebx		#STDOUT is $1
	xor %ecx, %ecx
	
	movb $sieve, %cl
	addb $30, %cl		#make ascii out of int
	
	movl $16,%edx		#buffer size
	int $0x80

next:
movb $sieve, %cl
incb %cl #increase sieve
movb %cl, (sieve)

movb $100, %al
cmpb %al, %cl #test if limit is reached
jz exit

movb $1, %al #if not start again with a new sieve and test = 1
movb %al, (test)

jmp loop

exit: movl $0, %eax
ret

please send me any critique and advices you can, i really want to make the complete switch from MASM to GAS as soon as possible....

cheers!

Paul Irofti wrote:

...

...

Pretty sure you want "movb test, %dl" - "$test" indicates that you're trying to move the *address* of your "test" variable (not the contents) into %dl... and it won't fit! (...etc.)

Best,
Frank

ok, so than what's the diffrence between:
a) movb $test, %al
b) movb test, %al
c) movb (test), %al
c) movb ($test), %al #if this exists!

and how come when i access kernel ops i use $<oper>

i.e. movl $__NR_write, %eax #i actually append the address 1?! not put the value 1?!

these are very confusing stuff for me. and i would greatly apreciate it if explained! thanks a lot Frank!

Paul Irofti wrote:

...

Hmmm, I'm actually Nasmist, not an as user, but...

I perhaps shouldn't have said "address"... As I understand it, the '$' indicates an "immediate" value. "movl $4, %eax" (or $__NR_write) moves the value 4 into %eax. "movl 4, %eax" would move the contents of address 0x00000004 into %eax - except that it segfaults because 0x00000004 isn't in "your" address space.

Maybe an actual (G)as user can clarify it better. It *is* confusing - if it were easy, everybody'd be doing it! :)

Best,
Frank

thanks man!

i had the same problems with syntax and address/value access in MASM when i started, now, i see myself hitting the same spot in the *nix world of assembly:))

well, i'm blown off, it's 6am here, i need some sleep, hopefully someone will make my morning..eh! afternoon... sweeter with a short lesson on the issue!

bye!