Computer Architecture Project #2 Cache Simulator

1. Computer Architecture Project #2Cache Simulator

2. Objectives • To understand cache memory • Organization • Set associativity • Operation • Cache Read & Write, Hit & Miss • LRU replacement policy • Performance • Hit/miss ratio, miss penalty • To develop your own cache simulator Memory Access Pattern Cache Organization Display Option Cache Simulator Hit/Miss Performance

3. General Cache Organization (S, E, B) E = 2e lines per set set line S = 2s sets If e = 1, “Direct Mapped Cache” else If s = 1, “Fully Associative Cache” else “E-Way Set Associative Cache” Cache size: C = S x E x B data bytes tag 0 1 2 B-1 v valid bit B = 2b bytes per cache block (the data)

4. E-way Set Associative Cache (Here: E = 2) E = 2: Two lines per set Assume that cache block size is 8 bytes Address of short int: t bits 0…01 • 100 v tag 0 1 2 3 4 5 6 7 v tag 0 1 2 3 4 5 6 7 find set v tag 0 1 2 3 4 5 6 7 v tag 0 1 2 3 4 5 6 7 v tag 0 1 2 3 4 5 6 7 v tag 0 1 2 3 4 5 6 7 v tag 0 1 2 3 4 5 6 7 v tag 0 1 2 3 4 5 6 7

5. E-way Set Associative Cache (Here: E = 2) E = 2: Two lines per set Assume that cache block size is 8 bytes Address of short int: t bits 0…01 • 100 compare both valid? + match: yes = hit v tag tag 0 1 2 3 4 5 6 7 v tag 0 1 2 3 4 5 6 7 block offset

6. E-way Set Associative Cache (Here: E = 2) E = 2: Two lines per set Assume that cache block size is 8 bytes Address of short int: t bits 0…01 • 100 compare both valid? + match: yes = hit v tag 0 1 2 3 4 5 6 7 v tag 0 1 2 3 4 5 6 7 block offset short int (2 Bytes) is here • No match : • One line in set is selected for eviction and replacement • Replacement policies: random, least recently used (LRU), …

7. LRU Replacement Policy • Theoretically… • Practically…

8. Performance • (Average Access Time) = (Hit Time) + (Miss Rate) × (Miss Penalty) = (Hit Time) + [1 – (Hit Rate)] × (Miss Penalty) • Example • Suppose cache hit time is 1 cycle, • Miss penalty is 100 cycles, • and hit rate is 97%. • Then average access time is:1 cycle + ( 1 – 0.97 ) × 100 cycles = 1 + 0.03 × 100 = 4 cycles.

9. Requirements of the cache simulator (1) • Cache simulator (hereinafter referred to CSIM) shall implementarbitrary numbers of sets and lines, and block size. • You should implement a way to provide the numbers of sets and lines, andblock size as inputs to CSIM. • CSIM shall a read trace file line by line and process it. • You should determine whether each memory operation is a cache hit or miss. • You should implement the LRU replacement policy • CSIM shall report the result of cache simulation. • You should report these three basic results: numbers of Hits, misses, and evicts • You should be able to report the average access time of cache simulation • You should be able to report whether each memory access in trace file results in a cache hit or miss

10. Restrictions & Advices • Implement method for input parameters. • You should implement it by argument passing. (full credit) • If you can’t, you can use standard input such as scanf(). (low credit) • Evaluate only data cache performance. • Therefore, you should ignore instruction load. • You should assume that the memory accesses are aligned properly.Therefore, you can ignore requested size in trace file. • You should evaluate your CSIM with, at least, 3 different trace data. You canuse one provided with this project. • Calculate average access time using below assumption: • Hit time = 1 cycle, miss penalty = 100 cycles. • Compile your CSIM without warnings.

11. How to trace memory accesses • “valgrind” • GPL licensed programming tool for memory debugging, memory leak detection, and profiling. (from http://en.wikipedia.org/wiki/Valgrind) • Usage: >> valgrind -log-fd=1 --tool=lackey -v --trace-mem=yes ls -l • Valgrind prints out memory accesses of “ls -l” on stdout, so you need to capture it by:>> valgrind -log-fd=1 --tool=lackey -v --trace-mem=yes ls -l > ls.trace • Output Format: [space]operation address,size

12. Reference Cache Simulator set • Usage: >>./csim [-v] -s <s> -E <E> -b <b> -t <trace file> • -v: Optional verbose flag that displays trace info • -s <s>: Number of set index bits (S = 2s is the number of sets) • -E <E>: Associativity (number of lines per set) • -b <b>: Number of block bits (B = 2b is the block size) • -t <trace file>: Name of the valgrind trace to replay line S = 2s sets Cache size: C = S x E x B data bytes tag 0 1 2 B-1 v valid bit B = 2b bytes per cache block (the data)

13. Cache Simulation Example (1) • Usage: >>./csim [-v] -s <s> -E <E> -b <b> -t <trace file> • Example: >>./csim -v -s 4 -E 1 -b 4 -t ./traces/yi.trace • Number of set index bits = 4 (16 sets) • Associativity = 1 (Direct Mapped Cache) • Number of block bits = 4 (16 blocks in a cache line) • Output L 10,1 miss M 20,1 miss hit …. hits: 4 misses:5 eviction: 3

14. Cache Simulation Example (2) • Example memory access pattern

15. Cache Simulation Example (3)

16. Cache Simulation Example (4)

17. Cache Simulation Example (5)

18. Cache Simulation Example (6)

19. Cache Simulation Example (7)

20. Cache Simulation Example (8)

21. Cache Simulation Example (9) Average Access Time = 1 + (5 / 9) * 100 = 56.5 Cycle

22. 보고서 작성요령 (1) • 아래의 내용을 포함할 것 • 설계 요구사항 • 제시된 CSIM의 설계 요구사항을 자신의 CSIM에 맞춰재정의 • 구현 • 자신의 CSIM이 어떤 식으로 동작하며 어떻게 설계 요구사항을반영하는지 서술 • 자신의 CSIM의 사용법과 시뮬레이션 결과 출력 방법에 대해 서술 • 시험 • CSIM의 요구사항을 어떤 방법으로 검증하였는지 서술 • 최소 3가지 Trace Data를 이용하여 검증 수행추가적으로, Trace Data를 어떤 방법으로 얻었는지를 서술 • CSIM 구현 내용을 알 수 있도록 캡쳐된 이미지를 첨부할 것 설계 시험 구현

23. 보고서 작성요령 (2) • 아래의 내용을 포함할 것 • 성능 평가 • 각각의 Cache 구조 (direct mapped, E-way setassociative 및 fully associative cache)별로 성능을 측정하고 각각을비교할 것 Design Testing Coding

24. 평가기준

25. 제출방법 • 아래 제출 목록의 산출물들을 메일로 제출 • E-mail address: yonghunlee@archi.snu.ac.kr • E-mail 제목: “[CSIM]학번_이름” • 산출물들은 “학번_이름.zip” 또는 “학번_이름.tar”으로 압축하여 제출 • 제출 목록 • CSIM source code • Project 보고서 • CSIM의 검증 시 사용한 Trace file • 제출 기한: ’13. 12. 18(수) 23:59 까지