abi.txt

		AMD64 C ABI

rax | r0  | ret#1
rcx | r1  | arg#4
rdx | r2  | arg#3 ret#2
rbx | r3  | callee saved
rsp | r4  |
rbp | r5  | callee saved
rsi | r6  | arg#2
rdi | r7  | arg#1
    | r8  | arg#5
    | r9  | arg#6
    | r10 |
    | r11 |
    | r12 | callee saved
    | r13 | callee saved
    | r14 | callee saved
    | r15 | callee saved



r3 r5 r12-r15 "belong" to the caller
(i.e., are callee-saved)

DF must be clear on function entry and return

once registers are assigned, the arguments passed in memory are pushed on
the stack in right-to-left order (i.e., the first vararg is the closest to
the top of the stack).

in a varargs call, 'al' specifies the number of vector registers used



None of these are preserved across function calls:
mmx0-mmx7 : temporaries
xmm0-xmm1 : pass and return floating point args
xmm2-xmm7 : pass floating point args
xmm8-xmm15 : temporaries

If the return type has class MEMORY, then the caller allocates space for it
on the stack and passes a pointer to it in RDI as a "hidden" first argument.
On return, RAX contains the address that has been passed by the caller
in RDI.
If the return type is of class INTEGER, then the value is returned in the
next available register in the sequence RAX RDX.
If the return type is SSE, then the next available register in the sequence
XMM0 XMM1 is used.