cuda-flux-stdlib

The Deeper Connection.

Standard library for FLUX VM programs — a collection of useful bytecode subroutine patterns defined as Rust constants.

Overview

FLUX VM is a register-based virtual machine with a compact instruction set. This crate provides pre-built bytecode sequences for common operations, ready to be loaded into any FLUX VM instance.

FLUX Opcode Reference

Hex	Mnemonic	Format	Description
0x00	NOP	—	No operation
0x01	HALT	—	Halt execution
0x10	ADD	A: rd, rs	Add rs to rd
0x11	SUB	A: rd, rs	Subtract rs from rd
0x12	MUL	A: rd, rs	Multiply rd by rs
0x13	DIV	A: rd, rs	Divide rd by rs
0x14	MOD	A: rd, rs	Modulo rd by rs
0x15	NEG	A: rd	Negate rd
0x18	AND	A: rd, rs	Bitwise AND
0x19	XOR	A: rd, rs	Bitwise XOR
0x1A	OR	A: rd, rs	Bitwise OR
0x1B	NOT	A: rd	Bitwise NOT
0x20	CMP	A: rd, rs	Compare rd and rs
0x30	LOAD	A: rd, rs	Load from memory
0x31	STORE	A: rd, rs	Store to memory
0x40	PUSH	A: rd	Push rd to stack
0x41	POP	A: rd	Pop stack into rd
0x48	MOV	B: rd, imm8	Move immediate
0x50	JMP	B: offset	Unconditional jump
0x51	JZ	B: offset	Jump if zero
0x52	JNZ	B: offset	Jump if not zero
0x53	JLT	B: offset	Jump if less than
0x54	JGT	B: offset	Jump if greater than
0x55	JEQ	B: offset	Jump if equal
0x56	JNE	B: offset	Jump if not equal
0x58	CALL	B: addr	Call subroutine
0x59	RET	—	Return from subroutine
0x5A	LOOP	B: count	Loop with counter

Function Reference

Math (src/math.rs)

Name	Bytes	Params	Returns	Description
abs	30 00 48 01 00 20 00 01 53 03 59 15 00 59	1	1	Absolute value: negate if negative
max	20 00 01 54 02 40 00 48 00 01 41 01 59	2	1	Maximum of two values
min	20 00 01 53 02 40 00 48 00 01 41 01 59	2	1	Minimum of two values
clamp	20 00 01 53 02 50 06 48 00 01 20 00 02 54 02 50 02 48 00 02 59	3	1	Clamp value to [min, max]
swap	19 00 01 19 01 00 19 00 01 59	2	2	XOR swap two registers
power	48 03 01 48 02 00 20 02 01 55 04 12 03 00 10 02 48 01 50 F6 48 00 03 59	2	1	Exponentiation via loop multiply
gcd	20 01 48 00 51 06 48 02 00 14 00 01 48 01 02 50 F5 59	2	1	GCD via Euclidean algorithm

Memory (src/memory.rs)

Name	Params	Returns	Description
memset	3 (addr, count, value)	0	Fill memory region with value
memcpy	3 (src, dst, count)	0	Copy bytes between regions
memcmp	3 (a, b, count)	1	Compare regions byte-by-byte
reverse	2 (addr, count)	0	Reverse bytes in place

Search (src/search.rs)

Name	Params	Returns	Description
linear_search	3 (arr, val, len)	1 (index or 0xFF)	Linear scan for value
binary_search	3 (arr, val, len)	1 (index or 0xFF)	Binary search on sorted array

Sorting (src/sorting.rs)

Name	Params	Returns	Description
bubble_sort	2 (arr, len)	0	Bubble sort ascending
selection_sort	2 (arr, len)	0	Selection sort ascending

Agent (src/agent.rs)

Name	Params	Returns	Description
trust_check	1 (agent_id)	1 (trust score)	Read from trust memory region (0xFE00+)
energy_check	0	1 (energy)	Read energy register (0xFF00)
should_delegate	3 (energy, threshold, confidence)	1 (bool)	Decision: delegate or not
report_status	4 (buf, id, state, energy)	0	Prepare A2A TELL message buffer

Usage

use cuda_flux_stdlib::Stdlib;

let lib = Stdlib::new();

if let Some(abs_fn) = lib.get("abs") {
    println!("ABS: {} bytes, {} params", abs_fn.bytecodes.len(), abs_fn.param_count);
    // Load abs_fn.bytecodes into your FLUX VM
}

for f in lib.all() {
    println!("{}: {:?}", f.name, f.bytecodes);
}

Testing

cargo test

License

MIT