Skip to guide

Workload to machine

Computer Architecture Guide

Computer architecture connects software-visible behavior to the structures that execute it. Good architectural reasoning starts with workloads and contracts, then evaluates latency, throughput, energy, bandwidth, complexity, and correctness together.

Reason from workload behavior to microarchitectural choices and validate those choices with models, counters, and experiments.

Reviewed and updated July 2026

Coverage map

What this guide covers

ISA and execution model

Understand the software-visible contract: instructions, registers, privilege, exceptions, memory behavior, and extension mechanisms.

  • ISA contracts
  • Privilege and exceptions
  • Memory model

Pipelines and hazards

Increase throughput while preserving dependencies, precise state, and correct recovery from control-flow changes.

  • Data hazards
  • Forwarding and stalls
  • Branch recovery

Caches and virtual memory

Exploit locality while managing misses, replacement, translation, protection, and interactions with devices.

  • Cache hierarchy
  • TLBs and page walks
  • AMAT

Interconnects and coherence

Move requests and data between agents while maintaining ordering, ownership, forward progress, and a coherent memory view.

  • On-chip networks
  • Coherence protocols
  • Ordering and deadlock

Performance, power, and validation

Evaluate designs with representative workloads, analytical models, simulation, counters, and disciplined bottleneck analysis.

  • CPI and throughput
  • Roofline reasoning
  • Performance per watt

Apply the concepts

Move from reading to reasoning.

Use the question bank to explain assumptions, debug failures, and validate tradeoffs in the language interviewers expect.