Discussion :

[millepattes]

Bonjour,
Je dois lire un format binaire non connu des classiques float32 ou float64. Je suis en échec sur la routine suivante:

Code :

Sélectionner tout - Visualiser dans une fenêtre à part

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
function [ floatCAP ] = uint48le_to_CAP( mots )
% récupération de 3 mots uint16 en little endian avec un mots=fread(fid,3,uint16) classique
% routine inspirée de ci dessous:
% UINT64LE_TO_VAXD Converts from IEEE-LE (UINT32) to VAXD (double)
%  This function takes a raw 32bit unsigned integer (little endian)
%  and converts it into the equivalent floating point number it represents
%  in the VAXF file format for floating point numbers. The VAXF format is
%  the single precision format used with both the VAXD and VAXG 
%  double precision formats and files. 
%  <a href="http://www.opengroup.org/onlinepubs/9629399/chap14.htm#tagfcjh_20" target="_blank">http://www.opengroup.org/onlinepubs/...htm#tagfcjh_20</a>
%
%   See also UINT64LE_TO_VAXG, UINT64LE_TO_VAXD, FREADVAXD, FREADVAXG
 
 
 
%% Define floating value properties for VAX architecture
% The generic equation for a floating point number is:
% (-1)^double(S) * (F+C) * A^(double(E)-B);
% Different operating systems and file formats utilize different values
% for A, B, and C. F, E, and S are computed from the appropriate bits in
% the number as stored on disk.
 
    A = 2   ;%exposant base 2
    B = 128 ;%Complément à 2 de l'exposant
    C = 0.5 ;%biais de normalisation de la mantisse
 
%% Convert raw unsigned number into right answer
% Flip the upper and lower bits (based on how Vax data storage format)
% CAP      <-----mot1-----><-----mot2-----><-----mot3----->
% IEEE-LE  <-----WORD3-----><-----WORD2-----><-----WORD1----->
 
    mot1  = mots(3);% 
    mot2  = mots(2);%
    mot3  = mots(1);%
 
    CAPInt = bitor( bitshift(mot2, 16), bitshift(mot1, 0) );
    CAPInt = bitor( bitshift(mot3, 32), bitshift(CAPInt,0));%reconstruction des 48 bits
 
% Pull out the sign, exponent, and fractional component
% VAX FLOAT BYTES  <-----WORD3----><-----WORD2-----><-----WORD1----->
% VAX FLOAT BITS   0123456789ABCDEF 0123456789ABCDEF 0123456789ABCDEF
% Sign Exp Fract   SFFFFFFFFFFFFFFF FFFFFFFFFFFFFFFF FFFFFFFFEEEEEEEE
 
    S = bitshift(bitshift(CAPInt , 0), -47); %)signe
    E = bitshift(bitshift(CAPInt , 40,48), -40);%)exposant
    F = bitshift(bitshift(CAPInt , 1,48),  -9); %   mantisse
 
% Construct the floating point number from SEF (Sign, Exp, Fract)
% Formula valid for non-zero values only (zeros fixed in next step)
% <a href="http://www.codeproject.com/KB/applications/libnumber.aspx" target="_blank">http://www.codeproject.com/KB/applic...libnumber.aspx</a>
% <a href="http://www.opengroup.org/onlinepubs/9629399/chap14.htm#tagfcjh_20" target="_blank">http://www.opengroup.org/onlinepubs/...htm#tagfcjh_20</a>
 
    M = C+double(F)./1.099511627776e+12;%DACS Specific 1099511627776=2^40 pour normalisation de la mantisse
    floatCAP = (-1).^double(S) .* M .* A.^(double(E)-B);%Generic
 
% Add in zeros (if E and S are zero, doubleVAXF should be set to zero)
 
    zerosIndex = (E == 0 & S == 0);%logical index of all zeros
    floatCAP(zerosIndex) = 0;
 
    zerosIndex = (E == 0 & S == 1);%cas non défini
    floatCAP(zerosIndex) = NaN;
 
end

je joints la description dès que je pourrais mettre une PJ