Microprocessors, Part 2 Codexery

EnCore Processor

Compact 32-bit RISC family with high frequency and low energy.

EnCore Processor

The EnCore processor family is a configurable, extensible 32-bit RISC microprocessor design developed by the PASTA Research Group at the University of Edinburgh School of Informatics. It uses a compact instruction set architecture that supports freely mixed 16-bit and 32-bit encodings for high code density, with no overhead when switching between them. The processor features a five-stage pipeline, claims the highest operating frequency in its class, and achieves very low dynamic energy consumption—99% of its flip-flops are automatically clock-gated by standard synthesis tools. Most non-memory operations complete in a single cycle, and there is at most one load-delay slot. Cache architectures are easily configurable.

All EnCore test chips are named after hills in Edinburgh. The first, Calton, is the smallest. The second, Castle, is named after the rock on which Edinburgh Castle sits.

**EnCore Calton** Calton is the first silicon implementation of the EnCore processor, fabricated in a generic 130nm CMOS process using a standard ASIC flow. It implements the baseline configuration extended with a barrel shifter, multiplier, and a full set of 32 general-purpose registers. The chip includes bus interface and system control functions alongside the processor core, with 8KB direct-mapped instruction and data caches. The complete system-on-chip occupies 1 mm² of silicon at 75% utilization. Chip-level power consumption is 25 mW at 250 MHz, and first silicon samples operate above 375 MHz at typical voltage and temperature.

**EnCore Castle** Castle is the second silicon implementation of an extended EnCore processor, fabricated in a generic 90nm CMOS process. It contains an extended processor core with a 32KB 4-way set-associative instruction cache and a 32KB 4-way set-associative data cache, embedded in a system-on-chip that provides a generic 32-bit memory interface, interrupts, clocks, and reset signals. The design uses generic free foundry libraries and a stack of nine metal layers. The complete design occupies 2.25 mm² on a 1.875 × 1.875 mm die, including the baseline CPU, reconfigurable Configurable Flow Accelerator (CFA) extension logic, two 32KB caches, and off-chip interfaces. It operates on a core voltage of 0.9V to 1.1V with 2.5V LVCMOS I/O signals, and is packaged in a 68-pin Ceramic LCC.

Pipeline stages
5
Instruction encodings
Freely-mixed 16-bit and 32-bit
First test chip
Calton (130nm CMOS, 1 mm², 25 mW at 250 MHz, >375 MHz)
Second test chip
Castle (90nm CMOS, 2.25 mm², 70 mW at 600 MHz)
Cache (calton)
8KB direct-mapped instruction and data cache
Cache (castle)
32KB 4-way set-associative instruction and data cache
Clock gating (castle)
Over 97% of flip-flops automatically clock-gated

Lore & Background

The EnCore microprocessor family was developed by the PASTA Research Group at the University of Edinburgh School of Informatics. Its baseline instruction set architecture is compact, including freely-mixed 16-bit and 32-bit encodings for maximum code density, with no overhead for switching between them. Most non-memory operations achieve single-cycle latency, with no more than one load-delay slot, and cache architectures are easily configurable.

Reader's Guide

The EnCore family is notable for its two silicon test chips, both named after hills in Edinburgh. The first, Calton, was fabricated in a generic 130nm CMOS process using a standard ASIC flow, occupying 1 mm² at 75% utilization. It included a barrel shifter, multiplier, and a full set of 32 general purpose registers, with 8KB direct-mapped instruction and data caches. First silicon samples operated above 375 MHz, with chip-level power consumption of 25 mW at 250 MHz. The second test chip, Castle, was fabricated in a generic 90nm CMOS process with 9 metal layers, occupying 2.25 mm². It extended the processor with a Configurable Flow Accelerator (CFA), two 32KB 4-way set-associative caches, and off-chip interfaces. Castle operated at 600 MHz with 70 mW power consumption, and over 97% of its flip-flops were automatically clock-gated. The design flow from RTL to GDSII was performed in-house by the PASTA team using Synopsys Design Compiler and IC Compiler, with LVS and DRC checks via Calibre.

Did You Know?

More in Microprocessors, Part 2 1-24

Spotted an error? Know more?

Reader corrections go straight into our review queue. Suggest an edit · How this site is sourced

Comments

Loading…
Open in the interactive codex →