search for: matrix_data

...erned with the > fixed-point accuracy. Maybe there's a way to keep the intermediate > results in 32 bits and rounding at the end of the sum. For example, > shifting only by 8 in the loop and doing a final shift by 7 outside the > loop. > > *Done. I've shifted >>4 to matrix_data and >>4 to input inside the loop > and >>= 7 to the output outside the loop. Let me know your thoughts.* I'm not sure I understand what you mean here (and I can't see where that change is in the code). > 9) For endianness conversion, you shouldn't have to even detect...

...mixing_matrix[2*i+1] << 8) (though you might want to check it too) 2) In mapping_matrix_multiply_short(), I would recommend something along these lines (untested) to improve accuracy: opus_int32 tmp = 0; for (col = 0; col < input_rows; col++) { tmp += (matrix_data[MATRIX_INDEX(matrix->rows, row, col)] * input[MATRIX_INDEX(input_rows, col, i)]) >> 8; } output[MATRIX_INDEX(output_rows, row, i)] = (tmp+64)>>7; 3) Looking at your get_*() function to compute offsets, I think you're missing some align() calls in the terms....

Hi Drew, Here's some comments on your patch (in no particular order): 1) I see that it's adding an #include of stdarg.h to opus_multistream.h Is that left over from the previous version? 2) Someone on this list might know better than I do on that one, but for the new _ctl_va_list() calls, I believe the convention is for va_start() and va_end() to appear in the caller rather than in in

...functions, I'm a bit concerned with the fixed-point accuracy. Maybe there's a way to keep the intermediate results in 32 bits and rounding at the end of the sum. For example, shifting only by 8 in the loop and doing a final shift by 7 outside the loop. *Done. I've shifted >>4 to matrix_data and >>4 to input inside the loop and >>= 7 to the output outside the loop. Let me know your thoughts.* 9) For endianness conversion, you shouldn't have to even detect it. I think something like this (untested) should work for both little-endian and big endian: for (i=0;i<...;i++...

...d-point accuracy. Maybe there's a way to keep the intermediate > > results in 32 bits and rounding at the end of the sum. For example, > > shifting only by 8 in the loop and doing a final shift by 7 outside the > > loop. > > > > *Done. I've shifted >>4 to matrix_data and >>4 to input inside the loop > > and >>= 7 to the output outside the loop. Let me know your thoughts.* > > I'm not sure I understand what you mean here (and I can't see where that > change is in the code). > > > 9) For endianness conversion, you shoul...

Hello Everyone,   Learning about R and its wonderful array of functions. If it's not obvious, I usually try to find out what a function stands for. I think this helps me remember better.   One function that has me stumped is "sink." Can anyone tell me if this stands for something?   Thanks,   Paul         __________________________________________________ [[alternative HTML