Prev: Create a vector out of other two vectors
Next: there is an error when using lp_solve to solve MIP in matlab
From: dpb on 16 Jun 2010 11:04
utab wrote:
> On Wed, 16 Jun 2010 09:24:34 -0500, dpb wrote:
>
>> Matlab "assumes" the word order is consistent with the natural order on
>> the platform on which it is running.
>>
>> You can handle foreign files with non-native ordering thru the options
>> in the fopen() call but there's nothing specific done different
>> automagically, no.
>
> Then there is still sth fishy because I should swap bytes in C to get the
> right information, with native C standard library functions and as far as
> I know MATLAB also uses these functions, but you are right that it should
> be consistent on the same system, however it does not seem so.

Indeed. That implies there's something else not described in the
equation. Some option somewhere that's being set that isn't explicit or
the file(s) aren't what think they are or...

One way to confirm for certain would be to have a look at the internals
of the file w/ a binary file viewer and see what is actually there in
which order.

Possible mixing up word length and/or not seeing effects owing to
magnitude issues or any other such perturbations?

But, I'm quite sure ML doesn't have some crystal ball in the fopen/fread
that swaps things sometimes and doesn't others w/o the fopen() options
on the same platform.

--
From: James Tursa on 16 Jun 2010 11:20
"Steven Lord" <slord(a)mathworks.com> wrote in message <hvanj3$dl4$1(a)fred.mathworks.com>...
>
> If for some reason you can't handle this through the machineformat input
> argument to FOPEN (which is what dpb suggested), you could use SWAPBYTES
> (assuming you have a sufficiently recent version of MATLAB.)
>
> http://www.mathworks.com/access/helpdesk/help/techdoc/ref/fopen.html
>
> http://www.mathworks.com/access/helpdesk/help/techdoc/ref/swapbytes.html

Or download swapbytes from the FEX if you have an older version of MATLAB:

http://www.mathworks.com/matlabcentral/fileexchange/17661-swapbytes-c-mex-function

James Tursa
From: James Tursa on 16 Jun 2010 11:22
utab <utabak(a)tudelft.nl> wrote in message <23d53$4c18e1c0$82a112b1$11227(a)news1.tudelft.nl>...
> On Wed, 16 Jun 2010 09:24:34 -0500, dpb wrote:
>
> > Matlab "assumes" the word order is consistent with the natural order on
> > the platform on which it is running.
> >
> > You can handle foreign files with non-native ordering thru the options
> > in the fopen() call but there's nothing specific done different
> > automagically, no.
>
> Then there is still sth fishy because I should swap bytes in C to get the
> right information, with native C standard library functions and as far as
> I know MATLAB also uses these functions, but you are right that it should
> be consistent on the same system, however it does not seem so.

How, exactly (i.e. post your code), are you doing the reading in MATLAB vs in C?

James Tursa
From: utab on 16 Jun 2010 12:07
On Wed, 16 Jun 2010 15:22:04 +0000, James Tursa wrote:

> How, exactly (i.e. post your code), are you doing the reading in MATLAB
> vs in C?
>
> James Tursa
Hi James,

A little mistake, I am using C++ ifstream objects to read.

I did first in MATLAB to really understand the structure of the binary
files. Then I did the same in C++ where I had to swap the bytes. The
program is ANSYS btw, commercial FE package

In MATLAB

% Set 1: Standard ANSYS file header
rlength=fread(fid,1,format1); % record length
dum=fread(fid,1,format1);
ANSYS_file_header=fread(fid,rlength,format1); % read ANSYS file header

in this read rlength is 100, and read correctly,

if I use ifstream read in C++

int record;
readInt(ifs, &record);

where readInt is

20 int readInt(ifstream &ifs,int *n)
21 {
22 unsigned char *cptr,tmp;
23 if (!ifs.read(as_bytes(*n),sizeof(int)))
24 return -1;
25 cptr = (unsigned char *)n;
26 tmp = cptr[0];
27 cptr[0] = cptr[3];
28 cptr[3] = tmp;
29 tmp = cptr[1];
30 cptr[1] = cptr[2];
31 cptr[2] = tmp;
32
33 return 0;
34 }

13 template<class T>
14 char* as_bytes(T& i)
15 {
16 void* addr=&i;
17 return static_cast<char*>(addr);
18 }

If I do not do the byte swapping in C, I get garbage.

uhh, the whole story.

Best,
Umut
From: dpb on 16 Jun 2010 12:25
utab wrote:
> On Wed, 16 Jun 2010 15:22:04 +0000, James Tursa wrote:
>
>> How, exactly (i.e. post your code), are you doing the reading in MATLAB
>> vs in C?
>>
>> James Tursa
> Hi James,
>
> A little mistake, I am using C++ ifstream objects to read.
>
> I did first in MATLAB to really understand the structure of the binary
> files. Then I did the same in C++ where I had to swap the bytes. The
> program is ANSYS btw, commercial FE package
>
> In MATLAB
>
> % Set 1: Standard ANSYS file header
> rlength=fread(fid,1,format1); % record length
> dum=fread(fid,1,format1);
> ANSYS_file_header=fread(fid,rlength,format1); % read ANSYS file header
>
> in this read rlength is 100, and read correctly,
>
> if I use ifstream read in C++
>
> int record;
> readInt(ifs, &record);
>
> where readInt is
>
> 20 int readInt(ifstream &ifs,int *n)
> 21 {
....

>
> If I do not do the byte swapping in C, I get garbage.
>
> uhh, the whole story.

Not quite... :)

fid=fopen(????....
format1 = ????

I think your read lengths are different but w/o "the rest of the story"...

--

First  |  Prev  |  Next  |  Last
Pages: 1 2 3
Prev: Create a vector out of other two vectors
Next: there is an error when using lp_solve to solve MIP in matlab