Lecture 18. Caches - Carnegie Mellon - Comp. Arch. 2015 - Onur Mutlu

preview_player
Показать описание
Lecture 19: High Performance Caches
Date: Feb 27th, 2015

Рекомендации по теме
Комментарии
Автор

gosh this is so useful. thanks for uploading!

ishaankulkarni
Автор

Amazing Lectures!!! Thank you for making it public. Hope some day will listen to your live lecture and Interact with you

tvcharan
Автор

Regarding sectored caches, "With a sector cache, it is possible to fetch only a portion of a block (or sector), and thus the time to handle a miss, and the bus traffic, can both be significantly reduced"

adeshshetty
Автор

Regarding write back caches saving on bandwidth, its a little less than what would naively assume, as writing to a block on cache N still has to tell cache N+1 to invalidate the block, which has to tell cache N+1 the same. Its less traffic than write through, but some bandwidth is still used.

momoanddudu
Автор

Actually shaded one is like 1-9-25 numbered blocks so when you are writing address corresponding to 256 Bytes memory at Time 14:08, 9th block should have been 01001, please correct me if i misunderstood :) Excellent videos though, , Thanks a lot for sharing your knowledge

jeetendraahuja
Автор

кто от wina лукасы ставим и лайкаем видос

glebfadeev
Автор

Excellent Video!!!...Is the lab assignments can be for the outside students???

purushothramalingam
Автор

I have one more question on cache… 
As per my knowledge unto know, cache is standing on its two pillars spatial and temporal locality, but when  we store data, its not necessary that all data fits serially in memory, so I am little worried about how spatial locality come to help. What I mean is, array I can say that it will be stored in serial manner, but when it comes to Linked List and other data structures, I don’t think spatial locality comes to any help, If we work with spatial locality on linked list then obviously we end up lot of cache miss and unnecessary cache writes if I am not wrong. Adding to that array has advantage when it comes to cache write, i.e when there is a miss then, we can bring that whole block and place in cache with possible future hits. But LinkedList I don’t think we can make use of spatial locality. Above statement may holds good for instructions too. Does it mean that we should be careful when we use LinkedList in design? Do cache follows some other methods to handle linked list and other data structures.

Автор

Why it is 0% hit rate if processor hits on B, C, D and misses on A and E? 58:00

Folsets
Автор

excellent videos playlist but not affordable to watch them all :(.
They are too long to learn before an exam

shadowangel