fread and float
in [Matlab]
|
From: Oleg Komarov on 13 Jul 2010 03:50 "James Tursa" <aclassyguy_with_a_k_not_a_c(a)hotmail.com> wrote in message <i1g7vl$4h3$1(a)fred.mathworks.com>... > "James Tursa" <aclassyguy_with_a_k_not_a_c(a)hotmail.com> wrote in message <i1g1u0$g2l$1(a)fred.mathworks.com>... > > dpb <none(a)non.net> wrote in message <i1g0ar$hfh$1(a)news.eternal-september.org>... > > > > > > <http://support.microsoft.com/kb/42980> > > > > > > for the MBF tutorial. > > > > That was a very useful link, in which it is stated: > > > > Sign Bits Exponent Bits Mantissa Bits > > --------- ------------- ------------- > > IEEE 1 11 52 + 1 (Implied) > > MBF 1 8 56 > > > > So it looks like MBF does *not* have a hidden leading mantissa bit. Bottom line is the 'VAXD' options stuff will not work with this format as is. So custom code will be required. However, it should be fairly easy to modify the custom 'VAXD' code to work with this. > > But now that I look at it again, it is confusing. 1 + 8 + 56 = 65. This doesn't add up to 64, so something is wrong with the description. Maybe the 56 should have been 55 + 1 ?? I need to find a more descriptive link ... > > James Tursa In this blog the explanations match my case: single float MBF to IEEE 754. http://www.boyet.com/Articles/MBFSinglePrecision.html Oleg
From: Oleg Komarov on 13 Jul 2010 09:14 > In this blog the explanations match my case: single float MBF to IEEE 754. > http://www.boyet.com/Articles/MBFSinglePrecision.html > > Oleg But I still don't get it. I use fread(fid, 4, 'uint32') and it gives me a 4 by 1 array of uint32 values. Should I use fread(fid, 4, 'uint8') instead? How can I merge the array into a single number, at least theoretically? Oleg
From: us on 13 Jul 2010 09:23 "Oleg Komarov" <oleg.komarovRemove.this(a)hotmail.it> wrote in message <i1hoqt$e55$1(a)fred.mathworks.com>... > > In this blog the explanations match my case: single float MBF to IEEE 754. > > http://www.boyet.com/Articles/MBFSinglePrecision.html > > > > Oleg > > But I still don't get it. > I use fread(fid, 4, 'uint32') and it gives me a 4 by 1 array of uint32 values. > > Should I use fread(fid, 4, 'uint8') instead? > > How can I merge the array into a single number, at least theoretically? > > Oleg a hint: - did you try... help typecast; us
From: Oleg Komarov on 13 Jul 2010 09:57 > a hint: > - did you try... > > help typecast; > > us Thanks for the hint! Now what I do is: input = double(typecast(fread(fid,4,'*uint8'),'uint32')); output = bitset(bitand(input,16777215),24)... ..*(power(2,(bitshift(bitand(input,4278190080),-24)-152)))... ..*(input~=0).*(((bitget(input,24)==0)*2)-1) output = 960628 Which is what I was looking for: YYMMDD --> 96/06/28 --> 1996 may 28 One last question...is there any good reference to understand the conversion outlined above?? I read a lot of stuff on the internet but my knowledge is messy... Thank you for support Oleg
From: dpb on 13 Jul 2010 13:11
Oleg Komarov wrote: .... > Which is what I was looking for: YYMMDD --> 96/06/28 --> 1996 may 28 That would be June 28, 1996 in these here parts... :) > One last question...is there any good reference to understand the > conversion outlined above?? I read a lot of stuff on the internet but my > knowledge is messy... .... I don't know what to suggest other than again doc bitshift % and friends The description of each operator seems pretty straightforward to me. One thing I'd recommend to make parsing the above conversion somewhat easier would be to print/display the coefficients (like 4278190080) as hex or binary numbers. doc sprintf % '%x' might be of interest...??? -- |