Hi, all:
  I am doing fixed-point lame mp3 encoding, but there are some problems. Because the data in caclulating psycho model is very great, they exceed int, and the data of xmin in iteration is very little, so i have to use mantissa(int or short) and exponent to represent them, but it runs so slowly! Anyone would like to helping me to optimize it? i can open my fixed-point source code. Thank you very much!
 
best ragards
anne

Reply via email to