mirror of
https://github.com/SikongJueluo/Mini-Nav.git
synced 2026-07-12 20:15:31 +08:00
refactor(cam): remove read noise from noise architecture (Phase 2)
- Make cam_read_noise a pass-through module, removing all noise injection logic - Switch write noise to use noise_mask_bernoulli instead of noise_mask_grouped - Add state machine to cam_write_noise for mask generation timing - Remove noise_mask_grouped.sv (no longer needed) - Remove read noise parameters from cam_noisy and cam_top - Update simulation and benchmark code to reflect read noise removal - Sync documentation to reflect Phase 2 architecture
This commit is contained in:
@@ -1,94 +1,31 @@
|
||||
# -*- coding: utf-8 -*-
|
||||
"""
|
||||
参考模型（ref_model）的纯 Python 单元测试。
|
||||
参考模型（ref_model）的纯 Python 单元测试 — Phase 2 cleaned.
|
||||
|
||||
本文件不涉及任何 RTL / Verilator 仿真，仅验证 Python 参考模型的正确性。
|
||||
所有 RTL-vs-模型 的对比测试（如顶层 test_cam_basic.py）都依赖此参考模型，
|
||||
因此这里是整个测试体系的「基石」——参考模型如果有 bug，所有对比测试都将失效。
|
||||
Phase 2 后只保留 pure matching 函数；所有 grouped/read-noise helpers 已删除。
|
||||
|
||||
测试覆盖：
|
||||
1. 分组翻转掩码 — 完全速率 (rate=1/1) 的正确位翻转模式
|
||||
2. 分组翻转掩码 — 零速率 (rate=0/100) 不应产生任何翻转
|
||||
3. 评分函数语义 — 确认是「匹配位数」而非「汉明距离」
|
||||
4. 读取噪声模型 — 相同输入 + 相同种子 = 可复现结果
|
||||
1. XNOR 评分语义 — 确认是「匹配位数」而非「汉明距离」
|
||||
2. Top-1 matching — 纯匹配，正确选出最高分索引
|
||||
3. Top-K matching — 返回排序后的行索引列表
|
||||
4. Top-K 排序规则 — 分数降序、平局行号升序
|
||||
5. Top-K k 超过行数时 clamp
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
from model.ref_model import (
|
||||
generate_grouped_flip_mask,
|
||||
match_top1_with_read_noise,
|
||||
match_top1,
|
||||
match_topk,
|
||||
match_topk_from_scores,
|
||||
xnor_popcount_score,
|
||||
)
|
||||
import numpy as np
|
||||
|
||||
|
||||
# ==============================================================================
|
||||
# 测试 1：完全速率下的分组翻转掩码生成
|
||||
# ==============================================================================
|
||||
|
||||
|
||||
def test_grouped_flip_mask_full_rate_one_bit_per_64_bit_group():
|
||||
"""
|
||||
验证 generate_grouped_flip_mask 在 rate_num=1, rate_den=1 时的行为。
|
||||
|
||||
背景：
|
||||
- CAM 的 write noise 模块将 512-bit 哈希按 64-bit 分组，每组最多翻转 1 位。
|
||||
- random_value 的位域含义（每 group 14 bits）：
|
||||
bits [5:0] → sample（未使用）
|
||||
bits [13:6] → bit_idx（选择该组内翻转哪一位）
|
||||
|
||||
本测试：
|
||||
- 构造一个 random_value，使每个 group 的 bit_idx = group+1
|
||||
- 断言生成的 mask 恰好有 8 个位被置位（每 group 一个）
|
||||
- 断言每个被翻转的位位置与预期一致
|
||||
"""
|
||||
random_value = 0
|
||||
for group in range(8):
|
||||
bit_idx = group + 1
|
||||
sample = 0
|
||||
random_value |= bit_idx << (group * 14)
|
||||
random_value |= sample << (group * 14 + 6)
|
||||
|
||||
mask = generate_grouped_flip_mask(
|
||||
random_value=random_value,
|
||||
hash_bits=512, # 8 组 × 64 bits/组
|
||||
noise_bits=8, # 每组的 bit_idx 位宽
|
||||
rate_num=1, # 分子 = 1
|
||||
rate_den=1, # 分母 = 1 → 100% 概率，每组都翻转
|
||||
)
|
||||
|
||||
# 预期：每组的 bit_idx 位被翻转
|
||||
expected = 0
|
||||
for group in range(8):
|
||||
expected |= 1 << (group * 64 + group + 1)
|
||||
|
||||
assert mask == expected
|
||||
assert mask.bit_count() == 8 # 恰好 8 位被翻转（每组一位）
|
||||
|
||||
|
||||
# ==============================================================================
|
||||
# 测试 2：零速率下不应产生任何翻转
|
||||
# ==============================================================================
|
||||
|
||||
|
||||
def test_grouped_flip_mask_zero_rate_no_flips():
|
||||
"""
|
||||
验证 rate_num=0 时，无论 random_value 为何值，mask 都应为 0。
|
||||
|
||||
这是写入噪声的「零噪声」配置边界测试——
|
||||
确保 RTL 参数 WRITE_NOISE_RATE_NUM=0 能真正关闭噪声注入。
|
||||
"""
|
||||
mask = generate_grouped_flip_mask(
|
||||
random_value=(1 << 128) - 1, # 全 1 的 random_value
|
||||
hash_bits=512,
|
||||
noise_bits=8,
|
||||
rate_num=0, # 分子 = 0 → 翻转概率为 0
|
||||
rate_den=100,
|
||||
)
|
||||
assert mask == 0 # mask 必须全 0，一个位都不翻
|
||||
|
||||
|
||||
# ==============================================================================
|
||||
# 测试 3：评分函数语义 — 确认是「XNOR 匹配位数」而非「汉明距离」
|
||||
# 测试 1：评分函数语义 — 确认是「XNOR 匹配位数」而非「汉明距离」
|
||||
# ==============================================================================
|
||||
|
||||
|
||||
@@ -114,64 +51,27 @@ def test_score_is_bit_match_popcount_not_hamming_distance():
|
||||
|
||||
|
||||
# ==============================================================================
|
||||
# 测试 4：读取噪声模型的可复现性（确定性种子）
|
||||
# 测试 2：Top-1 matching
|
||||
# ==============================================================================
|
||||
|
||||
|
||||
def test_read_noise_model_is_reproducible_after_reset_seed():
|
||||
"""
|
||||
验证 match_top1_with_read_noise 在相同参数下产生相同结果。
|
||||
|
||||
为什么这个测试至关重要：
|
||||
- RTL 中的 read noise PRNG 使用固定种子 (0x6A09E667F3BCC909)
|
||||
- 参考模型必须使用相同的种子来复现 RTL 的噪声行为
|
||||
- 如果两次调用结果不同，说明模型存在非确定性 bug
|
||||
（如未重置 PRNG 状态、或使用了非确定性随机源）
|
||||
|
||||
测试数据：
|
||||
- 8 行不同模式的 512-bit 哈希（全0、全1、稀疏值）
|
||||
- 噪声配置：rate=1%, lanes=8, noise_bits=8
|
||||
"""
|
||||
rows = [0, (1 << 512) - 1, 0x1234, 0x5678, 0x9ABC, 0xDEF0, 0x1357, 0x2468]
|
||||
query = rows[2]
|
||||
kwargs = dict(
|
||||
query=query,
|
||||
rows=rows,
|
||||
width=512,
|
||||
lanes=8,
|
||||
noise_bits=8,
|
||||
rate_num=1,
|
||||
rate_den=100,
|
||||
seed=0x6A09_E667_F3BC_C909,
|
||||
)
|
||||
|
||||
first = match_top1_with_read_noise(**kwargs)
|
||||
second = match_top1_with_read_noise(**kwargs)
|
||||
|
||||
# 两次调用的 Top-1 结果和分数数组必须完全一致
|
||||
assert first.top1_index == second.top1_index
|
||||
assert first.top1_score == second.top1_score
|
||||
assert first.scores.tolist() == second.scores.tolist()
|
||||
def test_match_top1_selects_highest_xnor_score_with_row_index_tiebreak():
|
||||
"""Top-1 应选出 XNOR 分最高的行；平局时选最小行号。"""
|
||||
rows = [0b0000, 0b1111, 0b0011, 0b0101]
|
||||
query = 0b0000
|
||||
result = match_top1(query, rows, width=4)
|
||||
assert result.top1_index == 0
|
||||
assert result.top1_score == 4
|
||||
assert result.scores.tolist() == [4, 0, 2, 2]
|
||||
|
||||
|
||||
# ==============================================================================
|
||||
# 测试 5：Top-K 排序 — 分数降序、平局行号升序
|
||||
# 测试 3：Top-K matching
|
||||
# ==============================================================================
|
||||
|
||||
|
||||
def test_match_topk_from_scores_uses_score_desc_then_row_asc():
|
||||
"""Top-K 排序规则：分数越大越优先；分数相同时行号越小越优先。"""
|
||||
from model.ref_model import match_topk_from_scores
|
||||
import numpy as np
|
||||
|
||||
scores = np.array([7, 9, 9, 2, 7], dtype=np.int32)
|
||||
assert match_topk_from_scores(scores, 4) == [1, 2, 0, 4]
|
||||
|
||||
|
||||
def test_match_topk_scores_rows_by_xnor_popcount():
|
||||
"""match_topk 通过 xnor_popcount 计算分数，返回排序后的行索引和分数数组。"""
|
||||
from model.ref_model import match_topk
|
||||
|
||||
rows = [0b0000, 0b1111, 0b0011, 0b0101]
|
||||
query = 0b0000
|
||||
indices, scores = match_topk(query, rows, width=4, k=3)
|
||||
@@ -179,43 +79,24 @@ def test_match_topk_scores_rows_by_xnor_popcount():
|
||||
assert indices == [0, 2, 3]
|
||||
|
||||
|
||||
# ==============================================================================
|
||||
# 测试 4：Top-K 排序 — 分数降序、平局行号升序
|
||||
# ==============================================================================
|
||||
|
||||
|
||||
def test_match_topk_from_scores_uses_score_desc_then_row_asc():
|
||||
"""Top-K 排序规则：分数越大越优先；分数相同时行号越小越优先。"""
|
||||
scores = np.array([7, 9, 9, 2, 7], dtype=np.int32)
|
||||
assert match_topk_from_scores(scores, 4) == [1, 2, 0, 4]
|
||||
|
||||
|
||||
# ==============================================================================
|
||||
# 测试 5：Top-K k 超过行数时 clamp
|
||||
# ==============================================================================
|
||||
|
||||
|
||||
def test_match_topk_clamps_k_to_row_count():
|
||||
"""当 k 超过实际行数时，返回所有行（按排序）。"""
|
||||
from model.ref_model import match_topk
|
||||
|
||||
indices, scores = match_topk(0, [0, 1], width=1, k=5)
|
||||
assert scores.tolist() == [1, 0]
|
||||
assert indices == [0, 1]
|
||||
|
||||
|
||||
# ==============================================================================
|
||||
# 测试 6：读取噪声 stateful 评分助手的跨查询状态推进
|
||||
# ==============================================================================
|
||||
|
||||
|
||||
def test_score_rows_with_read_noise_stateful_across_queries():
|
||||
"""score_rows_with_read_noise 在多次调用间正确推进 lane PRNG 状态。
|
||||
|
||||
两次调用使用相同的 rows/query 和零噪声率：
|
||||
- 分数应一致（无噪声翻转）
|
||||
- 但 lane states 应该变化（PRNG 已推进）
|
||||
"""
|
||||
from model.ref_model import score_rows_with_read_noise
|
||||
|
||||
rows = [0, 0, 0, 0]
|
||||
query = 0
|
||||
lane_states = [1, 2]
|
||||
|
||||
scores_1, next_states_1 = score_rows_with_read_noise(
|
||||
query, rows, lane_states=lane_states, width=128, lanes=2,
|
||||
noise_bits=2, rate_num=0, rate_den=100,
|
||||
)
|
||||
scores_2, next_states_2 = score_rows_with_read_noise(
|
||||
query, rows, lane_states=next_states_1, width=128, lanes=2,
|
||||
noise_bits=2, rate_num=0, rate_den=100,
|
||||
)
|
||||
|
||||
assert scores_1.tolist() == [128, 128, 128, 128]
|
||||
assert scores_2.tolist() == [128, 128, 128, 128]
|
||||
assert next_states_1 != lane_states
|
||||
assert next_states_2 != next_states_1
|
||||
|
||||
Reference in New Issue
Block a user