Reference

A searchable map of the supported instructions and the calling convention.

filter

80 instructions

mov

Copy register or wide immediate into Rd.

try in playground

Data processing

mov xd, xn / mov xd, #imm / mov xd, sp / mov vd.T, vn.T

c equivalent

// uint64_t Rd, Rm (an immediate works the same way)Rd = Rm;// .16b: uint8_t Vd[16], Vn[16]for (int i = 0; i < 16; i++) Vd[i] = Vn[i];

flags

NZCVdoes not set flags

mov     x9, 42                  // x9 = 42mov     x10, x9                 // copy: x10 = 42 too

the immediate form only takes a value that fits one shifted 16-bit field; a wider constant needs a move-wide then keep sequence.

#mov

register file · aapcs64

x0 – x7arguments · results
x8struct result address
x9 – x15caller-saved temps
x16 – x18reserved · avoid
x19 – x28callee-saved
x29 · x30fp · lr (the frame record)
d0 – d7float args · results
d8 – d15callee-saved
d16 – d31caller-saved float temps
v0 – v7q0 – q7vector args · results
v8 – v15q8 – q15callee-saved: low 64 bits only
v16 – v31q16 – q31caller-saved vector temps

Colour marks the registers every function call depends on: one colour for the argument rows and one for the callee-saved rows, where the callee is the function being called. The callee must preserve the callee-saved rows and the frame record, and the argument rows are yours to pass and receive. Each x row is one register with a w view of its low 32 bits, and each d row is one register with an s view of its low 32 bits. The v rows are those same registers at their full 128 bits, named q when read as one value.