Question

我的时钟缓存无效。我在做什么：

如果key为self.items_to_index：将bit设置为1，则返回self.buffers中的值。

如果不是：

在输入上调用函数以获取newval。

调用self._increment（）直到我们在self.buffers中看到一个位设置为0的条目。

self.buffers [self._p]现在等于[newval，1]

删除_p等于该条目的字典条目。

然后执行self.items_to_index [k] = _p。

增加p。

我上面描述的代码是：[not working]

def __getitem__(self, k):
    """
    Returns func(k) using the buffer to cache limited results.

    :param k: Value to be evaluated

    >>> clock = ClockMap(4, lambda x: x ** 2)
    >>> clock[4]
    16
    >>> clock[3]
    9
    >>> clock._p
    2
    """
    #when clock[index] is queried, this _getitem__ function is called
    if k in self.items_to_index:
        #hit
        index = self.items_to_index[k]
        entry = self.buffers[index]
        entry[1] = 1
        return entry[0]
    else:
        val = self.fn(k)
        while self.buffers[self._p][1] == 1:
            self.buffers[self._p][1] = 0
            self._increment()
        self.buffers[self._p] = [val, 1]

        for inputx, bufflocation in self.items_to_index.iteritems():
            if bufflocation == self._p:
                self.items_to_index.pop(inputx)
                self.items_to_index[k] = bufflocation
                break

        self._increment()
        return val

整个班级都附上她以供参考：

class ClockMap:

def __init__(self, cacheSize, func):
    """
    Do not change existing variables.
    [Optional] You are free to add additional items and methods.
    """
    self.cacheSize = cacheSize #number of buffers
    self.fn = func #function whose results that you will be caching
    self._p = 0 # pointer
    self._increments = 0
    self._miss_count = 0
    self.buffers = [[None, 0] for x in range(cacheSize)] #actual output vals of the func passed in
    #hit: retrieve the value from the cache. Miss: reevaluate func with input, write to cache
    self.items_to_index = {} # dict,  input to buffer location, self.items_to_index[x] = buffer_index

def _increment(self):
    """
    Do not change this method.
    Updates the clock pointer. The modulo maintains the clock nature.
    """
    self._increments += 1
    self._p = (self._p + 1) % self.cacheSize

我如何测试这个和我的测试结果：

def test3ClockMap(ClockMap):
check_dir()
clock = ClockMap(4, lambda x: x ** 2)
requests = [1, 2, 3, 4, 1, 6, 1, 4, 7, 4, 7, 5, 4, 6]
with open(your_output + "task3ClockMap.txt", "wb") as f:
    writer = csv.writer(f)
    writer.writerow(["Request", "Result", "Pointer", "Increments"])
    for r in requests:
        writer.writerow([r, clock[r], clock._p, clock._increments])
diff_against_reference("task3ClockMap.txt", 3)

差异文件：

6,15c6,15
< 1,1,1,9
< 6,36,2,10
< 1,1,3,11
< 4,16,0,12
< 7,49,1,17
< 4,16,2,18
< 7,49,3,19
< 5,25,0,20
< 4,16,1,25
< 6,36,2,26
---
> 1,1,0,4
> 6,36,1,9
> 1,1,2,10
> 4,16,2,10
> 7,49,3,11
> 4,16,3,11
> 7,49,3,11
> 5,25,0,16
> 4,16,1,17
> 6,36,2,18

Answer 1

我相信它是因为行

for inputx, bufflocation in self.items_to_index.iteritems():

self.items_to_index被初始化为空，并且不会在上面的循环之外更新，因此循环中的任何代码都不会运行。

Answer 2

值永远不会添加到self.items_to_index字典中，因此永远不会从缓存中返回值。您可以通过更改ClockMap.__getitem__()来处理当前指针不在索引中的情况来解决这个问题：

        for inputx, bufflocation in self.items_to_index.iteritems():
            if bufflocation == self._p:
                self.items_to_index.pop(inputx)
                self.items_to_index[k] = bufflocation
                break
        else:
            self.items_to_index[k] = self._p

请注意添加到else循环的for块。这样可以处理当前指针尚未存在于索引中并添加它的情况。

您的缓存类似于LRU缓存。如果您使用的是Python 3，则可能需要查看functools.lru_cache装饰器，例如

from functools import lru_cache

@lru_cache()
def f(x):
    return x**3

>>> f.cache_info()
CacheInfo(hits=0, misses=0, maxsize=128, currsize=0)
>>> f(2)
8
>>> f.cache_info()
CacheInfo(hits=0, misses=1, maxsize=128, currsize=1)
>>> f(2)
8
>>> f.cache_info()
CacheInfo(hits=1, misses=1, maxsize=128, currsize=1)

时钟更换政策

2 个答案: