Karan Diwan

FIFO Architectures

Three FIFOs in Chisel, showing how the architecture changes with clocking, clock-domain crossing, and control flow.

A FIFO looks like one component, but its right design depends on what the system around it demands. This project implements three parameterized FIFOs in Chisel, a Scala-embedded hardware construction language: a synchronous FIFO in a single clock domain, an asynchronous FIFO that crosses between independent read and write clocks, and an elastic FIFO with valid/ready flow control for backpressure. Each is elaborated into SystemVerilog and taken through simulation, verification, synthesis, and timing analysis.

Synchronous FIFO simulation waveform
Synchronous FIFO simulation waveform.

Architectures built

Post-synthesis characterization, Sky130HD

All three are characterized at the same configuration, 16 × 8-bit (128 bits of storage).

ArchitectureAreaCrit. pathFmaxPower
Synchronous FIFO5,915.7 µm²3.03 ns~330 MHz865 µW
Asynchronous FIFO6,730.2 µm²2.94 ns~340 MHz2,130 µW
Elastic FIFO5,620.4 µm²2.94 ns~340 MHz793 µW
The asynchronous FIFO pays for clock-domain crossing in both area and power, with roughly 2.5× the power of the synchronous design. Its worst-case path is on the read clock (2.94 ns); the write clock path is 2.29 ns.

Verification

Each architecture is verified at three levels: Chisel-native directed and randomized tests, an independent SystemVerilog testbench on the generated RTL, and gate-level simulation of the post-synthesis netlist using Sky130HD functional cell models.

Toolchain: Chisel, Scala, SystemVerilog, Icarus Verilog, Yosys, OpenSTA, GTKWave, Sky130HD
View repository on GitHub →
← Gate-Level Cybernetic Classifier Next: Repeated Arithmetic Machine →