User Guide

General-Purpose Programming 61

24592—Rev. 3.15—November 2009 AMD64 Technology

the CALL. When the called procedure finishes execution and is exited using a return instruction,

control is transferred to the return address saved on the stack.

The CALL instruction has the same forms as the JMP instruction, except that CALL lacks the short-

relative (1-byte offset) form.

• Relative Near Call—These specify an offset relative to the instruction following the CALL

instruction. The operand is an immediate 16-bit or 32-bit offset from the called procedure, within

the same code segment.

• Register-Indirect and Memory-Indirect Near Call—These specify a target address contained in a

• Direct Far Call—These specify a target address outside the current code segment. The address is

pointed to by a 32-bit or 48-bit far-pointer specified by the instruction, which consists of a 16-bit

code selector and a 16-bit or 32-bit offset. The direct far call form is invalid in 64-bit mode.

• Memory-Indirect Far Call—These specify a target address outside the current code segment. The

address is pointed to by a 32-bit or 48-bit far pointer in a specified memory location.

The size of the rIP is in all cases determined by the operand-size attribute of the CALL instruction.

CALLs push the return address to the stack. The data pushed on the stack depends on whether a near or

far call is performed, and whether a privilege change occurs. See Section 3.7.5, “Procedure Calls,” on

page 79 for further information.

For far CALLs, the selector portion of the target address can specify a code-segment selector (in which

case the selector is loaded into the CS register), or a call-gate selector, (used for calls that change

privilege level), or a task-state-segment (TSS) selector (used for task switches). In the latter two cases,

the offset portion of the CALL instruction’s target address is ignored, and the new values loaded into

CS and rIP are taken from the call gate or TSS.

Return

• RET—Return from Call

The RET instruction returns from a procedure originally called using the CALL instruction. CALL

places a return address (which points to the instruction following the CALL) on the stack. RET takes

the return address from the stack and transfers control to the instruction located at that address.

Like CALL instructions, RET instructions have both a near and far form. An optional immediate

operand for the RET specifies the number of bytes to be popped from the procedure stack for

parameters placed on the stack. See Section 3.7.6, “Returning from Procedures,” on page 81 for

additional information.

Interrupts and Exceptions.

• INT—Interrupt to Vector Number

• INTO—Interrupt to Overflow Vector

• IRET—Interrupt Return Word