일 | 월 | 화 | 수 | 목 | 금 | 토 |
---|---|---|---|---|---|---|
1 | 2 | 3 | 4 | 5 | ||
6 | 7 | 8 | 9 | 10 | 11 | 12 |
13 | 14 | 15 | 16 | 17 | 18 | 19 |
20 | 21 | 22 | 23 | 24 | 25 | 26 |
27 | 28 | 29 | 30 |
- transactional memory
- cache coherence miss
- structural hazard
- store conditional
- pipline hazards
- relational model
- theta join
- speculative execution
- multibanekd cache
- ISA
- dependence
- pipelined cache
- Subquery
- dynamic scheduling
- directory based coherence protocol
- sequential consistency
- cache optimization
- atomic exchange
- nonblocking cache
- cache coherence
- register renaming
- load linked
- branch prediction
- moesi
- Cache
- 관계형 모델
- pipelined
- sql
- mesi
- way prediction
- Today
- Total
목록분류 전체보기 (32)
공대생의 공부흔적
참고: Computer Architecture: A Quantitative Approach (5th edition) - 5.5.이번 글에서는 동기화에 대해 알아볼 것이다.목차LockLock을 위한 하드웨어 지원: exchLock을 위한 하드웨어 지원: LL&SCTransaction Memory1. Lock프로그래밍 시 Lock을 사용하는 이유는, 병렬 프로그램에서 *data race condition을 회피하기 위함이다. *data race condition: 공유 메모리 시스템에서, 여러 쓰레드가 정해지지 않은 순서로 공유 메모리 위치에 접근하는 경우. 이때 적어도 하나의 접근은 쓰기인 경우.예시) total_count가 global variable인 경우, 적절한 동기화 없이 모든 쓰레드가 total..

참고: Computer Architecture: A Quantitative Approach (5th edition) - 5.4.지난 글에 이어, 캐시 일관성에 대한 추가적인 주제들을 다룰 것이다.목차캐시 일관성 미스다층 캐시디렉토리 기반 프로토콜1. 캐시 일관성 미스일반적으로 캐시 미스에는 cold, capacity, conflict 세 가지의 미스가 존재한다.하지만 invalidation 기반 멀티프로세서에만 있는 새로운 캐시 미스의 종류가 존재한다. 예를 들면 다음과 같다.P1 read address A (S state)P2 write to address A (P1은 I state, P2는 M state)P1 read address A → 이때 invalidation으로 인해 캐시 미스가 발생한다.이..

참고: Computer Architecture: A Quantitative Approach (5th edition) - 5.2.이번 글에서는 캐시 일관성에 대한 개념 소개 및 3가지 프로토콜에 대해서 다루고자 한다.목차캐시 일관성 문제MSI 프로토콜MESI 프로토콜MOESI 프로토콜1. 캐시 일관성 문제소프트웨어는 캐시 일관성에 대해 신경쓰지 않는다. 하드웨어는 여러 캐시나 메인 메모리 내의 같은 주소에 대해 값들을 유지해야 하며, 하나의 공유 메모리 공간에 대한 추상화를 지원해야 한다. → 캐시 일관성이란, 공유 메모리 멀티프로세서에서의 통신 메커니즘을 뜻한다.프로세서는 데이터를 공유하기 위해 로컬 캐시를 읽거나/에 쓴다.캐시 일관성은 여러 프로세서에 공유되는 데이터 값을 제공하는 데 그 책임이 있다...
주제1. Thread-Level Parallelism2. Simultaneous Multithreading
주제Branch PredictionLoad/Store QueueSuperscalar issues
지난 글의 ILP에 이어, 이번에는 out of order execution에 더해 speculative execution을 구현하는 방법에 대해 살펴볼 것이다. 2024.04.14 - [Computer Architecture] - [컴퓨터구조#7-2] ILP (Instruction-Level Parallelism) - Tomasulo Algorithm 목차 Speculation Reorder Buffer (ROB) Example Future File Physical Register File 1. Speculation 이전 Tomasulo 알고리즘에서 사용된 Reservaton Station을 잠깐 살펴보자. RS entry는 각 명령어가 완료되는 시점까지 할당된다. 연속적인 data-dependent ..