If it is possible to read the input file more than once (your problem statement doesn't say it can't), the following should work. It is described in Benchley's book "Programming Perls." If we store each number in 8 bytes we can store 250,000 numbers in one megabyte. Use a program that makes 40 passes over the input file. On the first pass it reads into memory any integer between 0 and 249,999, sorts the (at most) 250,000 integers and writes them to the output file. The second pass sorts the integers from 250,000 to 499,999 and so on to the 40th pass, which sorts 9,750,000 to 9,999,999.

If the numbers are evenly distributed we can use Counting sort. We should keep the number of times that each number is repeated in an array. Available space is: 1 MB - 3 KB = 1045504 B or 8364032 bits Number of bits per number= 8364032/1000000 = 8 Therefore, we can store the number of times each number is repeated to the maximum of 2^8-1=255. Using this approach we have an extra 364032 bits unused that can be used to handle cases where a number is repeated more than 255 times. For example we can say a number 255 indicates a repetition greater than or equal to 255. In this case we should store a sequence of numbers+repetitions. We can handle 7745 special cases as shown bellow:

364032/(表示每个数字所需的位数+表示100万所需的位数)= 364032 / (27+20)=7745

你最多要数到99,999,999，并在沿途标明1,000,000个站点。因此，可以使用位流进行解释，即1表示递增计数器，0表示输出数字。如果流中的前8位是00110010，到目前为止我们将有0,0,2,2,3。

Log (99,999,999 + 1,000,000) / Log(2) = 26.59。你的内存中有2^28位。你只需要用一半!

To represent the sorted array one can just store the first element and the difference between adjacent elements. In this way we are concerned with encoding 10^6 elements that can sum up to at most 10^8. Let's call this D. To encode the elements of D one can use a Huffman code. The dictionary for the Huffman code can be created on the go and the array updated every time a new item is inserted in the sorted array (insertion sort). Note that when the dictionary changes because of a new item the whole array should be updated to match the new encoding.

如果每个唯一元素的数量相等，则编码D中每个元素的平均比特数将最大化。比如元素d1 d2…， dN在D中各出现F次。在这种情况下(最坏的情况是输入序列中同时有0和10^8)我们有

sum（1<=i<=N） F. di = 10^8

在哪里

sum(1<=i<=N) F=10^6，或F=10^6/N，归一化频率将是p= F/10^=1/N

平均比特数为-log2(1/P) = log2(N)。在这种情况下，我们应该找到使n最大化的情况，这发生在di从0开始的连续数，或者di= i-1时

10 ^ 8 =(1 < =我< = N) f . di =(1 < =我< = N) (10 ^ 6 / N)(张)= (10 ^ 6 / N) N (N - 1) / 2

i.e.

N <= 201。在这种情况下，平均比特数是log2(201)=7.6511，这意味着我们将需要大约1字节的每个输入元素来保存排序的数组。注意，这并不意味着D一般不能有超过201个元素。它只是说明，如果D的元素是均匀分布的，那么D的唯一值不可能超过201个。

如果输入流可以接收几次，这就容易多了(没有关于这方面的信息，想法和时间性能问题)。然后，我们可以数小数。有了计数值，就很容易生成输出流。通过计算值来压缩。 这取决于输入流中的内容。

你用的是哪种电脑?它可能没有任何其他“正常”的本地存储，但它是否有视频RAM，例如?100万像素x每像素32位(比如说)非常接近你所需的数据输入大小。

(我主要是问旧的Acorn RISC PC的内存，如果你选择低分辨率或低颜色深度的屏幕模式，它可以“借用”VRAM来扩展可用的系统RAM !)这在只有几MB普通RAM的机器上非常有用。