NVIDIA Grace (ARM Neoverse V2) Circuit Model
Scope
This document describes the architectural circuit blocks SLeeLa models reliably for the NVIDIA Grace (ARM Neoverse V2). It is not a transistor-for-transistor recreation.
Major blocks
- Register section
- AArch64 register model; many Neoverse-V2 cores with a coherent mesh and large shared cache
- ALU
- arithmetic;
- logic;
- shifts/rotates;
- flag generation.
- Address path
- 48-bit address formation;
- address latches;
- effective-address computation.
- Data path
- 64-bit data latch;
- internal data movement;
- external data bus buffer.
- Control
- instruction register;
- instruction decode;
- timing control;
- interrupt logic.
- Clock
- clock input;
- phase sequencing.
Accuracy boundary
Gate-level or transistor-level detail should be added only from sufficiently reliable die analysis or a validated hardware recreation. This profile stays at the architectural/behavioral/timing level.
Sources
- https://www.nvidia.com/en-us/data-center/grace-cpu/
- https://en.wikipedia.org/wiki/ARM_Neoverse