AST's CDSP-k5 Benchmarks for some Common DSP Algorithms
(preliminary information)
The following table presents a performance estimation for some common DSP algorithms.
The estimation is based on a fully-featured architecture that includes the complete MAC unit and supports the post-increment and bit-reverse indexing facilities.
|
Operation
|
Description
|
Conditions
|
Parameters
|
Cycle-Count
|
|
Complex FIR
|
Finite Impulse Response Filter with Constant or Variable Complex Coefficients
|
Arbitrary number of Coefficients and Samples
|
N=number of coefficients
M=number of samples
|
M*(N+p)+q
|
|
Real FIR
|
Finite Impulse Response Filter with Constant or Variable Real Coefficients
|
Even number of Coefficients
|
N=number of coefficients
M=number of samples
|
M/2*(N+p)+q
|
|
Complex IIR
|
Direct Form II IIR Filter, Constant or Variable Complex Coefficients
|
Cascaded biquad sections, Samples multiple of 8
|
N=number of samples
M=number of cells
|
N/A
|
|
LMS Complex FIR Coefficients Update
|
Update the Complex FIR coefficients using the Least Mean SquareAlgorithm
|
Even number of coefficients
|
N=Number of coefficients
|
2*N+p
|
|
Complex Autocorrelation
|
Complex Autocorrelation
|
Arbitrary input and output vector size
|
N=number of input samples
M=number of output samples
|
M*(N+p)+q
|
|
Real Autocorrelation
|
Real Autocorrelation
|
Even number of samples
|
N=number of input samples
M=number of output samples
|
M/2*(N+p)+q
|
|
Complex Energy
|
Square each complex element in a vector and sum all the squared values
|
Arbitrary vector size
|
N=vectors size
|
N+p
|
|
Real Energy
|
Square each real element in a vector and sum all the squared values
|
Even number of samples
|
N=vectors size
|
N/2+p
|
|
Complex FFT
|
Radix 2 Fast Fourier Transform
|
Number of samples is a power of 2
|
N=number of samples
|
Log2(N)*(2*N+p)+q
|
|
Complex Dot Product
|
Point-by-point multiplication of two equal-size complex vectors
|
Arbitrary vector size
|
N=vector size
|
2*N+p
|
|
Real Dot Product
|
Point-by-point multiplication of two equal-size real vectors
|
Even number of samples
|
N=vector size
|
N+p
|
|
Complex Weighted Vector Sum
|
Add two complex vectors elements having one of them multiplied by a complex constant
|
Even number of elements in the vectors
|
N=vectors size
|
3*N+p
|
|
Real Weighted Vector Sum
|
Add two real vectors elements having one of them multiplied by a real constant
|
Even number of elements in the vectors
|
N=vectors size
|
3*N/2+p
|
|
Real Max/Min
|
Find the minimum/maximum value in a vector and retainits position
|
Arbitrary vector size
|
N=vectors size
|
N+p
|
|
Real or Complex Search/Skip
|
Search until/while a value is found in a vector and retain the (last) position where it was found
|
Arbitrary vector size
|
N=vectors size
|
N+p
|
|
Complex Move
|
Move a complex vector from one position to another in memory
|
Arbitrary vector size
|
N=vectors size
|
2*N+p
|
|
Real Move
|
Move a real vector from one position to another in memory
|
Even number of samples
|
N=vectors size
|
N+p
|
Note 1: "p" and "q" count for the overhead loop-initializationinstructions. Performance seriously degrades when "N" is relatively small(<32) as compared with "p".