/ Embedded · 2026
2-Way Out-of-Order LC2K Processor
SystemVerilog superscalar processor with dual-issue scheduling, dynamic branch prediction, and an SRAM instruction cache.

Overview
Built a 2-way superscalar out-of-order LC2K processor on a four-person team for the Fall 2026 EECS 470 final project. The design decouples the frontend and backend with a micro-op queue, executes independent instructions across two ALUs and one load/store unit, and preserves architectural correctness through hazard checks and ordered memory operations.
Key Features
- Priority selector issues up to two micro-ops per cycle out of order from a window of three or more operations
- RAW/WAW hazard checks and memory ordering around stores
- Split execute and writeback stages with pipeline-register and same-cycle ALU-to-ALU forwarding
- Two ALUs and one load/store unit let independent ALU operations pass stalled loads
- Banked branch target buffer, dynamic branch prediction, and wrong-path squashing
- SRAM instruction cache and prefetching ahead of single-ported, multi-cycle memory
- Self-checking micro-op queue testbench and design-space sweeps of cache, queue, and branch-predictor sizes
Technical Challenges
Coordinated dual-issue selection, forwarding, branch recovery, and single-port memory access while splitting register-offset loads and stores into single-unit micro-ops. Evaluated clock period, CPI, time per instruction, area, and power across configuration sweeps.
Outcome
Passed every project test program in simulation and synthesis, with writeback traces matching the reference implementation.
Technologies
- SystemVerilog
- Computer Architecture
- RTL
- Simulation
- Synthesis