DIGITAL LIBRARY - ARITHMETIC CORES

Software models and tools:

VHDL IP Cores (home-made and half-baked):

The following VHDL IP cores are provided under the GPL license.
Documentation is included for some cores.

Expanded CORDIC (FX, DFX, FP):

  • Fixed Point (FX) and Floating Point (FP) hyperbolic engine, powering, logarithm: The fully parameterized iterative architectures are implemented in Fixed Point and Floating Point Arithmetic. The word size is customizable.
    • Architectures (FX, FP):
      • Expanded hyperbolic CORDIC engine
      • Natural Logarithm architecture (lnx) based on the expanded hyperbolic CORDIC engine
      • Powering architecture (x^y) based on the expanded hyperbolic CORDIC engine
    • Code/documentation:
  • Dual Fixed Point (DFX) CORDIC: circular, hyperbolic engines, powering and logarithm circuits: The following fully parameterized iterative architectures are implemented in Dual Fixed Point Arithmetic. The format (word size, p0,p1) is customizable.
    • Architectures (DFX):
      • Expanded circular CORDIC engine
      • Expanded hyperbolic CORDIC engine
      • Natural Logarithm architecture (lnx) based on the expanded hyperbolic CORDIC engine
      • Powering architecture (x^y) based on the expanded hyperbolic CORDIC engine
    • Code/Documentation:
      • Paper: DFX paper
      • VHDL implementation (stand-alone IP), Vivado projects, test files (testbenches, MATLAB scripts, TCL scripts, etc): DFX CORDIC

Parameterized CSA (Carry Save Adder) Units:

The following custom units (for unsigned integers) are fully parameterized architectures. Three units are presented:

  • csa_complete: N-bit addition of 3 numbers. It includes CSA + RCA adders.
  • csa_complete_cin: N-bit addition of 3 numbers with carry in. It includes CSA + RSA adders.
  • csa_mult: NxM multiplication of 2 numbers based on CSA adders - Fully Combinational.

VHDL implementations (stand-alone IPs): Generic CSA Units VHDL IPs

Parameterized Adder Tree with enable:

The following custom Pipelined Adder Tree core is a fully parameterized architecture.

VHDL implementations (stand-alone IPs): Generic Adder Tree with enable VHDL IP

Parameterized Integer Array Extrema Units:

The following custom units are fully pipelined architectures. Three units are presented:

  • array_extremum: Get the maximum or minimum of an n-element array. Indices generated internally.
  • array_extremum_id: Get the maximum or minimum of an n-element array. Indices fed externally.
  • array_minmax: Get the maximum and minimum of an n-element array. Indices generated internally.

VHDL implementations (stand-alone IPs): Generic Array Extrema Units VHDL IPs
Report on Parameterized Array Extrema Units: (pdf)

Parameterized Adder/Subtractor with carry in:

The following custom adder/subtractor core is a fully parameterized architecture.

VHDL implementations (stand-alone IPs): Generic Adder/Subtractor wtih carry in VHDL IP
Report on Parameterized Adder/Subtractor with carry in: (pdf)

Parameterized Unsigned Integer Divider:

The following custom divider core (for unsigned integers) is a fully parameterized architecture. Three versions are presented: one fully parallel for maximum performance, one fully combinatorial, and one iterative for minimum resources.

VHDL implementations (stand-alone IPs): Generic Divider VHDL IPs
Report on Parameterized divider: (pdf)

Parameterized Unsigned Integer Multipier:

The following custom multiplier core (for unsigned integers) is a fully parameterized architecture. Three versions are presented: one fully parallel for maximum performance, one fully combinational, and one iterative for minimum resources.

VHDL implementations (stand-alone IPs): Generic Multiplier VHDL IPs
Report on Parameterized multiplier: (pdf)

Parameterized Iterative Signed Multiplier:

The following custom signed multiplier core is a fully parameterized architecture. It is based on the unsigned integer multiplier.

VHDL implementations (stand-alone IP): Generic signed iterative multiplier VHDL IP

Parameterized Parallel LUT (Look-Up Table) core:

The following custom Parallel LUT core is a fully parameterized architecture.

VHDL implementation (stand-alone IP): Generic Parallel LUT core VHDL IP

 

 

 

 

Reconfigurable Computing Research Laboratory (RECRLab), Electrical and Computer Engineering Department, Oakland University