I have some message strings that I define in my program and display to the user at various times.

I'm programming on Linux using the ELF format.

I'm wondering if I should define my string messages at the end of my program to make sure that my program is aligned on a 4 byte boundary. But I'm wondering if this is necessary.

Does it matter performance-wise if I define my string messages in various places throughout my source code, or should I define them at the end to ensure that the program is on a 4 byte boundary?

I noticed that in the FASM source code that string messages are defined near the end of fasm.asm, and then a bunch of files are included, the last one being messages.inc which consists of a bunch of string messages. Was it does like this for performance? Why are the string messages in the fasm.asm file not also in the messages.asm file?

If you are using message string then performance should be the last of your concerns. And aligning your "program" on a 4-byte boundary does not really make sense. The x86 opcodes have variable alignment so aligning any code, except the start of a procedure, will have no effect.

If your strings are bytes then no alignment is needed. If they are words (like in unicode) then you might like to use word alignment, but it is not needed, just tidier.

Thanks. When and why would the start of a procedure be aligned? Or do you mean that aligning the start of a procedure would have an effect until the procedure encountered an instruction that wasn't a multiple of 4?

What do you mean by message string?

I define the messages in my program like this:
usage db 'This program takes 2 arguments',0xA,0
Is this what you mean by message string?

So it makes sense to align data, like in a data segment. But it doesn't make sense to align code because the instructions are variable length anyway?

Yes, that is a message string. Aligning data: Yes, a good thing.
Aligning code: Doesn't hurt, but mostly pointless except for some very exotic situations.

Why would someone need alignment? I'm inexperienced.

The main reason I can think of is for SIMD instructions like SSE's movaps (and probably others). Also some of the FPU instructions require aligned memory..

http://flatassembler.net/docs.php?article=manual

The manual above mentions some of the instructions that require aligned memory. Normally, its SSE instructions, and normally its aligned to 16 bytes.

For example:


This is not just a performance enhancement. its a requirement, if you try to execute those instructions on memory locations which are not aligned to 16 bytes, they WILL fail.

As far as I know, no "standard" instruction REQUIRES alignment, only the extended sets like FPU, MMX, SSE, etc require it.. However you may be able to improve performance (whether you use SSE or not) by aligning data correctly.

Here's a good article by intel on the subject:
http://software.intel.com/en-us/articles/data-alignment-when-migrating-to-64-bit-intel-architecture/

Might also be worth grabbing a copy of the Intel Software Development Manual (if you dont already have it), as that will most likely mention alignment, or more specifically, grab the Optimization manual available on this page:
http://www.intel.com/content/www/us/en/architecture-and-technology/64-ia-32-architectures-optimization-manual.html

That mentions both Code alignment and Memory alignment (page 3-12 and page 3-61 respectively).

Wow! Thanks for such a detailed answer!