Understanding the pop instruction in assembly

Question

I am a student studying computer systems for the first time (w/ Computer Systems: A Programmer's Perspective). We are working on assembly and I am starting to understand command suffixes for x86_64 such as using leaq in something like:

leaq (%rsp, %rdx), %rax

However, I am failing to understand using a suffix for pop . For example, using the same logic, it would make sense to me that we'd use popl for something like:

popl %edi

But, in the text and other examples online, I just see:

pop %edi

What is the difference? is popl even valid? Just looking for a little more insight. Anything helps, thank you.

Answer 1

What you can do in asm is limited by what the hardware can do. Implicit vs. explicit operand-size in the source (suffix or not) doesn't change the machine code it will assemble to.

So the right question to ask is whether the hardware can do a 32-bit push in 64-bit mode? No, it can't, therefore no asm source syntax exists that will get it to do exactly what you were trying to do with one instruction.

• Does each PUSH instruction push a multiple of 8 bytes on x64?
• x86 Assembly pushl/popl don't work with "Error: suffix or operands invalid"

That's why your assembler won't accept pop %edi or popl %edi . Those are exactly equivalent because the 32-bit register implies DWORD ( l ) operand-size. The examples you saw of popl or pop %edi are for 32-bit mode, where EDI is the full register instead of the low half of RDI, and that instruction is encodable.

You only need a size suffix when it's ambiguous, mov $1, (%rdi) . Your assembler will give an error for that instead of guessing one of b/w/l/q.

But push is a bit special: push $1 will default to a 64-bit push, even though pushw $1 is possible. How many bytes does the push instruction push onto the stack when I don't specify the operand size?