Talk:Technical Reference:State Definition: Difference between revisions
m Pseudocode |
|||
| Line 67: | Line 67: | ||
[[User:Mellinger|Mellinger]] 15:00, 5 January 2010 (UTC) | [[User:Mellinger|Mellinger]] 15:00, 5 January 2010 (UTC) | ||
== Pseudocode == | |||
Thanks for pointing that out, but to someone (like me) who is not terribly familiar with C++ code, seeing the actual code is not very helpful. For example, I would have to start hunting through the code to find out what State::ValueType, mByteLength, mpData, et cetera mean. | |||
Also, I am proud to note that my algorithm is at least 8 times more efficient than the C++ function you pointed me to. | |||
Latest revision as of 23:07, 5 January 2010
State Vector
The section on state vectors is totally wrong. After reverse-engineering some data files with the help of the BCI2000Viewer, it appears that the state vectors are actually encoded in reverse-big-endian. In other words, if smaller numbers represent lower order bits, then the state vector layout mentioned in this section would actually be encoded as follows:
| State Vector Byte 1 | State Vector Byte 2 | State Vector Byte 3 | |||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
| Running | SourceTime | unused | |||||||||||||||||||||
| 0 | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | |||||||
Which on the physical machine actually maps to:
| State Vector Byte 1 | State Vector Byte 2 | State Vector Byte 3 | |||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 |
| SourceTime | Running | SourceTime | unused | SourceTime | |||||||||||||||||||
| 6 | 5 | 4 | 3 | 2 | 1 | 0 | 0 | 14 | 13 | 12 | 11 | 10 | 9 | 8 | 7 | 15 | |||||||
Fixed section on state vector encoding
Thanks for pointing out the error. State vector encoding is actually little-endian (LSB first) rather than big-endian, and its layout on a little-endian machine is straightforward as a concatenation of little-endian encoded state values. I fixed the description accordingly.
Pseudocode reference
Do you think it would be helpful to readers to put (in pseudocode) a way to read the State Definition?
int ExtractStateValue(unsigned char[] vector, int byteoffset, int bitoffset, int numberofbits){
int numberofbytes = numberofbits / 8 + 1;
char bytes[numbytes] = vector[ byteoffset : byteoffset + numbytes ];
int ret = bytes[0] >> bitoffset;
for ( int i=1; i<numbytes; i++ ){
ret += bytes[i] << ( 8 * i - bitoffset );
}
return ret % ( 1 << numbits );
}
This is not compilable C or C++ code, as far as I know. However, I think it might make things clearer.
Cstocks 21:15, 4 January 2010 (UTC)
Actual C++ code
The actual C++ code is here: http://www.bci2000.org/tracproj/browser/trunk/src/shared/types/StateVectorSample.cpp#L69 (log into trac if you get a message about browser view privileges)
Mellinger 15:00, 5 January 2010 (UTC)
Pseudocode
Thanks for pointing that out, but to someone (like me) who is not terribly familiar with C++ code, seeing the actual code is not very helpful. For example, I would have to start hunting through the code to find out what State::ValueType, mByteLength, mpData, et cetera mean.
Also, I am proud to note that my algorithm is at least 8 times more efficient than the C++ function you pointed me to.