Fast lookup table ideas

[rozick1]

I need a really fast lookup table:

-it will accept a two digit hex number
-the first digit will locate the row of the table (1-F) and the 2nd the column (1-F)
-returns the value of that column/row.

At the moment I have something like this and it is way too slow:

  int tolookup2 = tolookup; //find this number

    string inttostring;   
    stringstream hexout;
    hexout << hex<< tolookup2;
    inttostring = hexout.str();//convert to hex

   if (inttostring.length()==1)  //if less than 2 digits add 0 to start
   {


        inttostring.insert(0,"0");
   }

    int column,row;


    if (inttostring[0] >='0' && inttostring[0]<='9')  //if number 1-9
        row = inttostring[0] - 48;

    else
        row = inttostring[0] - 87;    //if a-f




     if (inttostring[1] >='0' && inttostring[1]<='9')
        column = inttostring[1] - 48;

    else
        column = inttostring[1] - 87;


        return table[row][column];  //table is the 2d array containing the table
  
}

There must be a faster way than this?

Any ideas would be great (hash tables?).

[Peter87]

row = (tolookup >> 4) & 0xF;
column = tolookup & 0xF;
return table[row][column];

[BlackSheep]

You're looking for a 2D array. The fastest "hash table" out there is just an array, and based on how you need to index it that's exactly what you need.

std::vector<Data> myData (0xFF);
//It doesn't matter if the vector is actually "2D", since every combination of 
//"rows" and "columns" has a different index. 

To read hex just use:

int index;
std::cin >> std::hex >> index;
myData[index] = 42;

[Duthomhas]

The fastest possible way to do it is a straight, 1D array as suggested.

That said, I cannot imagine what kind of requirements you are under to convert a two-digit hexadecimal value faster than a couple of arithmetic operations for each digit.

Have you really profiled your program or are you playing "human optimizer"?

[rozick1]

Stringstream is horrendously slow.

[rozick1]

Peter, that is about 230x faster than my code! Thank you!

[cire]

Stringstream is horrendously slow.

Let's not blame stringstream for your poor algorithmic choice.