General Computer Science 320201 GenCS I & II Lecture ... - Kwarc

Realizing arg
arg i pushes the i th argument from the current frame to the stack.
use the register IN3 for the frame pointer. (extend for first frame)
Realizing call
label instruction effect comment
〈arg〉 LOADIN 1 1 ACC : = P (VPC + 1) load i
STORE 0 P (0): = ACC cache i
MOVE IN3 ACC
STORE 1 P (1): = FP cache FP
SUBI 1
SUB 0 ACC : = FP − 1 − i load argument position
MOVE ACC IN3 FP: = ACC move it to FP
inc IN2 SP: = SP + 1 prepare push
LOADIN 3 0 ACC : = P (FP) load arg i
STOREIN 2 0 P (SP): = ACC push arg i
LOAD 1 ACC : = P (1) load FP
MOVE ACC IN3 FP: = ACC recover FP
MOVE IN1 ACC
ADDI 2
MOVE ACC IN1 VPC: = VPC + 2 next instruction
JUMP 〈jt〉 jump back
c○: Michael Kohlhase 314
call p pushes the current program address, and jumps to the program address p
(pushes the internal cells first!)
label instruction effect comment
〈call〉 MOVE IN1 ACC
STORE 0 P (0): = IN1 cache current VPC
inc IN2 SP: = SP + 1 prepare push for later
LOADIN 1 1 ACC : = P (VPC + 1) load argument
ADDI 2 24 + 3 ACC : = ACC +2 24 + 3 add displacement and skip proc a l
MOVE ACC IN1 VPC: = ACC point to the first instruction
LOADIN 1 − 2 ACC : = P (VPC − 2) stealing a from proc a l
STOREIN 2 0 P (SP): = ACC push the number of arguments
inc IN2 SP: = SP + 1 prepare push
MOVE IN3 ACC ACC : = IN3 load FP
STOREIN 2 0 P (SP): = ACC create anchor cell
MOVE IN2 IN3 FP: = SP update FP
inc IN2 SP: = SP + 1 prepare push
LOAD 0 ACC : = P (0) load VPC
ADDI 2 ACC : = ACC +2 point to next instruction
STOREIN 2 0 P (SP): = ACC push the return address
JUMP 〈jt〉 jump back
c○: Michael Kohlhase 315
Note that with these instructions we have maintained the linear quality. Thus the virtual machine
is still linear in the speed of the underlying register machine REMA.
Realizing return
return takes the current frame from the stack, jumps to previous program address.
(which is cached in the frame)
179