ternary-vm

A balanced-ternary virtual machine and assembler.

A tryte is 9 trits (each in {-1, 0, +1}), holding values in [-9841, +9841]. This is a clean reference runtime for the Multi-Phase Ternary Computing paradigm introduced in MPTC, realised here as a small interpreter plus a two-pass assembler with labels.

Why ternary?

Balanced ternary is the most efficient integer base for its symbol count (Information Theory, 1950): it needs the fewest trits to represent a given range, and the symmetric digit set makes negation and sign handling trivial. It is the natural fit for MPTC's wave/superposition model.

Assembler

One instruction per line; # starts a comment. Operands are register names R0..R7 or balanced-ternary integer literals (may be negative).

LOAD  R0  5
LOAD  R1  7
ADD   R0  R1      # R0 = R0 + R1
OUT   R0          # print R0
HALT

Labels and branches:

LOAD  R0  0
LOAD  R1  1
LOAD  R2  10
LOOP:
OUT   R1
ADD   R0  R1
MOV   R4  R1
MOV   R1  R0
MOV   R0  R4
SUB   R2  R3
JNZ   LOOP R2
HALT

Run

python vm/ternary_vm.py programs/fib.tasm
# output: [1, 1, 2, 3, 5, 8, 13, 21, 34, 55]

[![CI](https://github.com/kostyk348/ternary-vm/actions/workflows/ci.yml/badge.svg)](https://github.com/kostyk348/ternary-vm/actions/workflows/ci.yml)

Programs in programs/: add.tasm (12), mul.tasm (6*8 = 48 via repeated addition), fib.tasm (Fibonacci).

Instruction set

Opcode	Operands	Semantics
HALT	—	stop
LOAD	R dst, imm	R[dst] = imm
STORE	R src, addr	mem[addr] = R[src]
ADD	R d, R s	R[d] = R[d] + R[s]
SUB	R d, R s	R[d] = R[d] - R[s]
MUL	R d, R s	R[d] = R[d] * R[s]
NEG	R d	R[d] = -R[d]
MOV	R d, R s	R[d] = R[s]
JMP	target	pc = target
JZ	target, R c	if R[c]==0 pc = target
JNZ	target, R c	if R[c]!=0 pc = target
OUT	R s	append R[s] to output
NOP	—	no-op

All arithmetic clamps to the 9-trit range (saturation). A future C port (ternary_vm.h) is planned to run the same bytecode on FPU-less MCUs.

License

Apache 2.0