Fast bit packing?

Question

Kevin 2011-6-8

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/9054-fast-bit-packing

Hi everyone,

I am looking for a fast way to do bit packing in Matlab (my data set is pretty big). Here is what I mean by bit packing.

>> bits = [0 0 0 0 1 1 1 0]; % bits = vector of 1's or 0's.

I want each set of 4 bits got packed into a unsigned 4-bit integer, i.e.

>> x = [0 15];

bits(1) = most significant bit of the first integer bits(4) = least significant bit of the first integer

bits(5) = most significant bit of the second integer bits(8) = least significant bit of the second integer

Thanks

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Jan 2011-6-8

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/9054-fast-bit-packing#answer_12457

在 MATLAB Online 中打开

bits = [0 0 0 0 1 1 1 0];
factor = [8;4;2;1]; 
x = [bits(1:4) * factor, bits(5:8) * factor]

I assume, this is a very efficient task for a C-mex function, because the Matlab version creates large temporary arrays, if "bits" is large.

4 个评论
显示 2更早的评论隐藏 2更早的评论

Jan 2011-6-9

Andrei's "[8,4,2,1]*reshape(bits,4,[])" and Walter's "32bits and TYPECAST" are efficient. But bit-packing seems to be more straight in C, especially for large data.

Walter Roberson 2011-6-9

If you are going to be packing in to 32 bits, you would reshape() to group 32 bits at a time, and you would use the appropriate list of powers of 2 depending on the bit order you wanted, with (e.g.)

P2 = 2.^(31:-1:0);

in your initialization and

typecast(uint32(P2*reshape(bits,32,[])),'uint64')

Unless, that is, some of the bits from the second group of 32 bits need to be mixed with the first group. If you do need that kind of mixing, reshape to columns of 64, index the resulting matrix at a permutation matrix, reshape to columns of 32, matrix multiply, uint32, typecast.

请先登录，再进行评论。