What's the difference between lea and mov in assembly?

abefroman · 09-14-2015, 04:19 PM

Code:

        
lea rax, [sample]
mov rbx, sample
mov rcx, [sample]
...
sample: db 0xaa, 0xbb, 0xcc, 0xdd, 0xee, 0xff, 0x11, 0x22

When viewing the registers, it seems the first 2 commands are the same, and then 3rd one puts the value of sample into rcx. I understand there is supposed to be some difference between the first 2 commands though.

GDB output:

Code:

print $rax
$1 = 6291708 (hex: 0x6000fc)

print $rbx
$7 = 6291708 (hex: 0x6000fc)

x/8xb 6291708
0x6000fc:       0xaa    0xbb    0xcc    0xdd    0xee    0xff    0x11    0x22

print $rcx
$8 = 2455024673282112426 (hex: 2211ffeeddccbbaa)

print &sample
$9 = (<data variable, no debug info> *) 0x6000fc

x/8xb 0x6000fc
0x6000fc:       0xaa    0xbb    0xcc    0xdd    0xee    0xff    0x11    0x22

x/8xb &sample
0x6000fc:       0xaa    0xbb    0xcc    0xdd    0xee    0xff    0x11    0x22

x/8xb $eax
0x6000fc:       0xaa    0xbb    0xcc    0xdd    0xee    0xff    0x11    0x22

x/8xb $ebx
0x6000fc:       0xaa    0xbb    0xcc    0xdd    0xee    0xff    0x11    0x22

x/8xb $ecx
0xffffffffddccbbaa:     Cannot access memory at address 0xffffffffddccbbaa

TIA

Keith Hedger · 09-14-2015, 07:18 PM

to start you dont say what processor you are using which would help, anyway been a few years since i did any assembly programming but 'lea' stands for 'load efective address' and 'mov' is just short fore move, lea usualy takes an address plus a register and combines them to give the final or 'effective ' address which is where your data will be stored/loaded from, mov just uses a simple address, read up on indexed addressing for a full exxplanation.

abefroman · 09-14-2015, 08:30 PM

Thanks for the reply, the CPU is Intel x86_64

Quote:

Originally Posted by Keith Hedger

lea usualy takes an address plus a register and combines them to give the final or 'effective ' address which is where your data will be stored/loaded from

What do you mean combines them?

Code:

lea rax, [sample]

This is putting the address of sample into rax, no?

My question is how is that different from (besides the fact I'm using rbx instead of rax):

Code:

mov rbx, sample

Which also puts the address of sample into rbx.

NevemTeve · 09-15-2015, 12:04 AM

The Assembler-writers don't agree on this. You might want to get the address of a field or the content of a field. The following syntaxes are used (and perhaps more):

#1 (address)

Code:

LEA    AX,field
MOV    AX,field
MOV    AX,OFFSET field

#2 (content)

Code:

MOV    AX,field
MOV    AX,[field]

It doesn't look very clear; my suggestion is using the very first for the address and the very last for the content.

NevemTeve · 09-15-2015, 03:26 AM

nasm, for example, accepts these versions:

Code:

        mov     eax,[4*ecx+0x12] ; content
        lea     eax,[4*ecx+0x12] ; address

gdb-disass says this (set disassembly-flavor att; default):

Code:

   0x080483c0 <+0>:	mov    0x12(,%ecx,4),%eax
   0x080483c7 <+7>:	lea    0x12(,%ecx,4),%eax

or this (set disassembly-flavor intel):

Code:

   0x080483c0 <+0>:	mov    eax,DWORD PTR [ecx*4+0x12]
   0x080483c7 <+7>:	lea    eax,[ecx*4+0x12]

GNU-as expects the "att flavor":

Code:

        mov    0x12(,%ecx,4),%eax
        lea    0x12(,%ecx,4),%eax

sundialsvcs · 09-15-2015, 07:57 AM

There are many, many online descriptions of the various flavors of the x86 instruction set, with examples . . .