Repository files navigation

University of Pittsburgh ECE 1195

Designing hardware components using Vivado/Vivado HLS and testing designs on a field programmable gate array. Coursework includes a multicycle MIPS CPU,hardware accelerators for matrix multiplication and 2d convolution, and various smaller components.

Spring 2023

Contains IPs for all top-level designs create in labs. Also contains various prewritten components to aid in the design process.

Intended for learning the basics of Vivado. A generic combinational hardware adder written in asynchronous VHDL.

A fully functional 32-bit ALU comprised of four components:

• Logic Unit for bitwise logic operations
• Arithmetic Unit for addition/subtraction operations (from Lab 1)
• Shift Unit for register shift operations
• Comparator for comparison operations

An unsigned 32-bit multiplier implemented using a shift registers, an adder, and finite state machine controller. The completion of an operation is signaled by a 'done' bit and the result is stored in a 64-bit register. Each operation takes 33 clock cycles to run.

A fully synchronous CPU that implements part of the MIPS instruction set. I designed the components and controller using behavioral VHDL. I combined these components in a Vivado block diagram to implement the top-level design (based on Von Neumann architecture). The top-level design is capable of executing MIPS assembly programs. Features include:

• ALU operations using ALU from Lab 2 (arithmetic, logic, comparison, and shift)
• Memory interfacing (lw, lh, sw)
• Program Counter functionality
• Branch/Jump instructions (j, jr, bne)
• Count leading ones instruction (clo)
• Multiplication using multiplier from Lab 3 (umult, mflo, mfhi)
• Immediate operands

A hardware accelerator for multiplying matrices implemented with Vivado HLS. Intended for learning how to use Vivado HLS.

A fully implemented hardware accelerator for grayscale image convolution. Runtime complexity is the same as the size of the input image. The onboard hardware implementation relies on a low-overhead Linux distribution. For results, see lab_6/lab6_report.pdf. Features both shift register and ring buffer implementations.