Motorola DSP Developer's Kit

conv-c.asm

C = mot###_conv( A,B )

Description

This function convolves vectors A and B, where both A and B are complex vectors

Input/Output

Input: Complex vector A (Ar is the real part, and Ai is the imaginary part), and complex vector B (Br is the real part, and Bi is the imaginary part)

Output: Complex vector C (Cr is the real part, and Ci is the imaginary part)

Algorithm

LengthC = LengthB + LengthA - 1
for ( i = 1; i  <= LengthC; i ++ ) {
if ( i <= LengthA ) {
   for ( j = 1; j <= i; j ++ ) {
       Cr[i] += Ar[j] * Br[i - j + 1] - Ai[j] * Bi[i - j  + 1];
       Ci[i] += Ai[j] * Br[i - j + 1] + Ar[j] * Bi[i - j  + 1];
   }
}
else {
   for ( j = i - LengthA; j <= LengthA - 1; j ++ ) {
       Cr[i] += Ar[j + 1] * Br[i - j] - Ai[j + 1] * Bi[i - j];
       Ci[i] += Ai[j + 1] * Br[i - j] + Ar[j + 1] * Bi[i - j];
    }
}
}

Memory & Register

Memory allocation:

In X memory:

• X:(INA) stores the start address of the real part of vector A
• X:(INA + LENGTH(A)) stores the start address of the imaginary part of vector A
• X:(INA + LENGTH(A)*2) stores the start address of the real part of vector C

In Y memory:

• Y:(INB) stores the start address of the real part of vector B
• Y:(INB + LENGTH(B)) stores the start address of the imaginary part of vector B
• Y:INB + LENGTH(B)*2 stores the start address of the imaginary part of vector C

Register usage:

• R0 is used for the index of the real part of vector A
• R0+N0 is used for the index of the imaginary part of vector A
• R4 and R5 are used for the index of the real part of vector B
• R4 +N4 are used for the index of the imaginary part of vector B
• R3 is used for the index of the real part of vector C
• R7 is used for the index of the imaginary part of vector C
• R1, R2, and R6 are used as loop control
• R2-1 -> M5, assume M{0,1,2,3,4,6,7} = $ffffff

Status Register

The assembly function conv-c.asm does not set any status registers/bits during function execution.

Data Size Limit

The maximum length of vector A and B can't be larger than 1/4 of the continuous available data memory size.

Data Range Limit

The input data vector range is [-1.0, +1.0].

Precision

In the case of DSP563, precision is the full 23 bits.

In the case of DSP566, precision is the full 15 bits.

Performance Limit

DSP563:

DSP566:

conv-r.asm

decimate-fir-r.asm