write fixed point data via fwrite to tcp

Question

Tom DeLonge 2019-3-14

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/450152-write-fixed-point-data-via-fwrite-to-tcp

回答： Tom DeLonge 2019-3-26

I have a mx7 array of doubles which I would like to transmit via tcpip using fwrite. However, I would like to transmit each column in it's own format, e.g.

int16 - fixed point - fixed point - double - int16 - fixed point - fixed point

The fixed points are 64bit long (16 + 48).

Now, how can I do this using fwrite? I would like to send the whole array at once and not each row/value on it's own.

Also, I should note that m can be rather large (on the order of 50k)... Can anyone suggest a solution?

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Tom DeLonge 2019-3-14

编辑：Tom DeLonge 2019-3-14

my idea so far was to

rearrange the columns a bit,
typecast the whole thing down to a mx11 single array,
transmit via fwrite as 'single'

the only problem with this approach is that I cannot use typecast on fixed point objects in matlab. I would have to write my own conversion routine for that I guess?

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Tom DeLonge 2019-3-26

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/450152-write-fixed-point-data-via-fwrite-to-tcp#answer_367480

在 MATLAB Online 中打开

Just wanted to answer my own question. I used my own routine to convert a double number to a fixed point, which is then being sent in disguise as a uint64. For a big endian receiver and a 64 bit fixed-point number with a word length of 16 bits, the code is simply

fxpInt64 = swapbytes(int64(dblIn*2^48));
fxpInt64 = swapbytes(typecast(fxpInt64(:),'uint16'));

Then, as can be seen, I in turn typecasted everything to uint16, which was sort of the largest common denominator for my classes.This is then being sent using fwrite and uint16 in it's class argument.