ARC-stuff / loda-rust /rust_project /loda-rust-cli /src /mine /histogram_instruction_constant.rs
Alignment-Lab-AI's picture
Upload folder using huggingface_hub
d5bfab8 verified
use super::random_indexes_with_distance;
use std::error::Error;
use std::fs::File;
use std::io::{BufRead, BufReader};
use std::path::Path;
use std::collections::{HashMap, HashSet};
use std::iter::FromIterator;
use serde::Deserialize;
use rand::Rng;
use rand::seq::SliceRandom;
use rand::SeedableRng;
use rand::rngs::StdRng;
use loda_rust_core::parser::{InstructionId, ParseInstructionId};
use crate::common::parse_csv_data;
type ValueAndWeight = (i32,u32);
type ValueAndWeightVector = Vec<ValueAndWeight>;
/// Instructions that takes a constant value.
///
/// The most used combo: `add $0,1` (addition by 1)
///
/// Almost as popular combo: `sub $0,1` (subtract by 1)
///
/// Usecase:
/// During mining, when mutating an `add` instruction.
///
/// It's a time waster making poor choices of constants.
/// The miner originally picked random integers, but it was excruciating slow.
///
/// Time can be saved this way.
///
/// Before mining: analyze all programs and build a histogram.
///
/// During mining: make weighted choices from the histogram.
#[derive(Clone, Debug)]
pub struct HistogramInstructionConstant {
instruction_and_valueweightvector: HashMap<InstructionId, ValueAndWeightVector>
}
impl HistogramInstructionConstant {
pub fn number_of_items(&self) -> usize {
self.instruction_and_valueweightvector.len()
}
#[allow(dead_code)]
pub fn load_csv_file(path: &Path) -> Result<HistogramInstructionConstant, Box<dyn Error>> {
let file = File::open(path)?;
let mut reader = BufReader::new(file);
Self::create(&mut reader)
}
const SHUFFLE_COUNT: usize = 0;
fn create(reader: &mut dyn BufRead) -> Result<HistogramInstructionConstant, Box<dyn Error>> {
let records_original: Vec<Record> = parse_csv_data::<Record>(reader)?;
// Shuffle the items slightly
let mut records: Vec<Record> = records_original.clone();
let seed: u64 = 1;
let mut rng = StdRng::seed_from_u64(seed);
let indexes: Vec<usize> = random_indexes_with_distance(&mut rng, records.len(), Self::SHUFFLE_COUNT);
for index in indexes {
records[index].count = records_original[index].count;
}
let instruction_and_valueweightvector: HashMap<InstructionId, ValueAndWeightVector> =
Record::instruction_and_valueweightvector(&records);
let result = Self {
instruction_and_valueweightvector: instruction_and_valueweightvector
};
Ok(result)
}
pub fn choose_weighted<R: Rng + ?Sized>(&self, rng: &mut R, instruction_id: InstructionId) -> Option<i32> {
let value_and_weight_vec: &ValueAndWeightVector =
match self.instruction_and_valueweightvector.get(&instruction_id) {
Some(value) => value,
None => {
return None;
}
};
let value: i32 = value_and_weight_vec.choose_weighted(rng, |item| item.1).unwrap().0;
Some(value)
}
}
#[derive(Clone, Debug, Deserialize)]
struct Record {
count: u32,
instruction: String,
constant: i32,
}
impl Record {
fn unique_instruction_ids(records: &Vec<Record>) -> HashSet<InstructionId> {
let mut instruction_ids = Vec::<InstructionId>::new();
for record in records {
match InstructionId::parse(&record.instruction, 0) {
Ok(instruction_id) => {
instruction_ids.push(instruction_id);
},
Err(_) => {}
}
}
HashSet::from_iter(instruction_ids.iter().cloned())
}
fn value_and_weight_vec(records: &Vec<Record>, instruction_id: InstructionId) -> ValueAndWeightVector {
let mut value_and_weight_vec: ValueAndWeightVector = vec!();
let needle: String = instruction_id.to_string();
let mut already_known = HashSet::<i32>::new();
for record in records {
if record.instruction != needle {
continue;
}
if already_known.contains(&record.constant) {
// Keep only the first value. Ignore following duplicates.
continue;
}
let value = (record.constant, record.count);
value_and_weight_vec.push(value);
already_known.insert(record.constant);
}
value_and_weight_vec
}
fn instruction_and_valueweightvector(records: &Vec<Record>) -> HashMap<InstructionId, ValueAndWeightVector> {
let instruction_ids: HashSet<InstructionId> = Record::unique_instruction_ids(&records);
let mut result: HashMap<InstructionId, ValueAndWeightVector> = HashMap::new();
for instruction_id in instruction_ids {
let value_and_weight_vec: ValueAndWeightVector =
Record::value_and_weight_vec(&records, instruction_id);
if value_and_weight_vec.is_empty() {
continue;
}
result.insert(instruction_id, value_and_weight_vec);
}
result
}
}
#[cfg(test)]
mod tests {
use super::*;
use rand::SeedableRng;
use rand::rngs::StdRng;
#[test]
fn test_10000_parse_csv_data() {
let data = "\
count;instruction;constant
36545;add;1
33648;sub;1
17147;mul;-2
";
let mut input: &[u8] = data.as_bytes();
let records: Vec<Record> = parse_csv_data(&mut input).unwrap();
let strings: Vec<String> = records.iter().map(|record| {
format!("{} {} {}", record.count, record.instruction, record.constant)
}).collect();
let strings_joined: String = strings.join(",");
assert_eq!(strings_joined, "36545 add 1,33648 sub 1,17147 mul -2");
}
#[test]
fn test_10001_unique_instruction_ids() {
let data = "\
count;instruction;constant
36545;add;1
9232;add;2
666;unknown;23
555;sub;1
171;mul;-2
92;add;3
";
let mut input: &[u8] = data.as_bytes();
let records: Vec<Record> = parse_csv_data(&mut input).unwrap();
let actual: HashSet<InstructionId> = Record::unique_instruction_ids(&records);
let v = vec![InstructionId::Add, InstructionId::Subtract, InstructionId::Multiply];
let expected: HashSet<InstructionId> = HashSet::from_iter(v);
assert_eq!(actual, expected);
}
#[test]
fn test_10002_value_and_weight_vec_typical_data() {
let data = "\
count;instruction;constant
36545;add;1
92;add;2
9232;add;3
100;add;4
";
let mut input: &[u8] = data.as_bytes();
let records: Vec<Record> = parse_csv_data(&mut input).unwrap();
let actual: ValueAndWeightVector = Record::value_and_weight_vec(&records, InstructionId::Add);
let expected: ValueAndWeightVector = vec![(1,36545),(2,92),(3,9232),(4,100)];
assert_eq!(actual, expected);
}
#[test]
fn test_10003_value_and_weight_vec_without_duplicates() {
let data = "\
count;instruction;constant
1000;add;1
1000;add;2
1000;add;1
1000;add;3
1001;add;1
";
let mut input: &[u8] = data.as_bytes();
let records: Vec<Record> = parse_csv_data(&mut input).unwrap();
let actual: ValueAndWeightVector = Record::value_and_weight_vec(&records, InstructionId::Add);
let expected: ValueAndWeightVector = vec![(1,1000),(2,1000),(3,1000)];
assert_eq!(actual, expected);
}
#[test]
fn test_10004_instruction_and_valueweightvector() {
let data = "\
count;instruction;constant
1001;add;1
1002;add;2
999;div;2
998;mul;-1
";
let mut input: &[u8] = data.as_bytes();
let records: Vec<Record> = parse_csv_data(&mut input).unwrap();
let actual: HashMap<InstructionId, ValueAndWeightVector> = Record::instruction_and_valueweightvector(&records);
assert_eq!(actual.len(), 3);
}
#[test]
fn test_20000_instruction_and_valueweightvector_add() {
let data = "\
count;instruction;constant
1000;add;1984
1000;mul;-1
";
let mut input: &[u8] = data.as_bytes();
let instance = HistogramInstructionConstant::create(&mut input).unwrap();
let mut rng = StdRng::seed_from_u64(0);
let actual = instance.choose_weighted(&mut rng, InstructionId::Add);
assert_eq!(actual, Some(1984));
}
#[test]
fn test_20001_instruction_and_valueweightvector_multiply() {
let data = "\
count;instruction;constant
1000;add;1984
1000;mul;-1
";
let mut input: &[u8] = data.as_bytes();
let instance = HistogramInstructionConstant::create(&mut input).unwrap();
let mut rng = StdRng::seed_from_u64(0);
let actual = instance.choose_weighted(&mut rng, InstructionId::Multiply);
assert_eq!(actual, Some(-1));
}
#[test]
fn test_20002_instruction_and_valueweightvector_no_such_instruction() {
let data = "\
count;instruction;constant
1000;add;1984
1000;mul;-1
";
let mut input: &[u8] = data.as_bytes();
let instance = HistogramInstructionConstant::create(&mut input).unwrap();
let mut rng = StdRng::seed_from_u64(0);
let actual = instance.choose_weighted(&mut rng, InstructionId::GCD);
assert_eq!(actual, None);
}
}