Thread Subject: fwrite and fread with uint64

I am creating binary files for a camera where 3 by 21-bit data are stored in 64 bit words.

My code creates uint64 words which parse correctly.

However, when I write these uint64 words, and then read them, they are not correct.

For example, if I create a uint64 word (using bitshift and bitor) which reports as
q =

  9147373404693901452

(note that this *can't* naively be entered in Matlab as q = uint64(9147373404693901452) because of the precision limitations of double precision numbers)
then issue the following fwrite and fread:

fid = fopen('test.hex','w');fwrite(fid,q,'uint64'),fclose(fid);
fid = fopen('test.hex','r');t = fread(fid,'uint64=>uint64'),fclose(fid)

I get
t =

  9147373404693901312

Not sure if the error is in the write or the read (I fear the write), but it does appear related to an invisible intermediate double precision step.

Is there any advice as to a fix or an elegant solution/work around?

I am running into this same problem now. Did anyone ever figure out a solution?

Thanks,
Sarah

"Philip Arcuni" wrote in message <h6she1$5h7$1@fred.mathworks.com>...
> I am creating binary files for a camera where 3 by 21-bit data are stored in 64 bit words.
>
> My code creates uint64 words which parse correctly.
>
> However, when I write these uint64 words, and then read them, they are not correct.
>
> For example, if I create a uint64 word (using bitshift and bitor) which reports as
> q =
>
> 9147373404693901452
>
> (note that this *can't* naively be entered in Matlab as q = uint64(9147373404693901452) because of the precision limitations of double precision numbers)
> then issue the following fwrite and fread:
>
> fid = fopen('test.hex','w');fwrite(fid,q,'uint64'),fclose(fid);
> fid = fopen('test.hex','r');t = fread(fid,'uint64=>uint64'),fclose(fid)
>
> I get
> t =
>
> 9147373404693901312
>
> Not sure if the error is in the write or the read (I fear the write), but it does appear related to an invisible intermediate double precision step.
>
> Is there any advice as to a fix or an elegant solution/work around?

"Sarah" wrote in message <jf50l7$mvf$1@newscl01ah.mathworks.com>...
> I am running into this same problem now. Did anyone ever figure out a solution?
>
> Thanks,
> Sarah

I think MATLAB fixed this with later versions. However, in the meantime you can try typecasting. E.g.

- typecast your variable to UINT32
- write that out as a UINT32
- read it in as a UINT32
- typecast it back to UINT64

James Tursa