14.2.7 Direct-mapped Caches

MIT 6.004 Computation Structures, Spring 2017
Instructor: Chris Terman
View the complete course: ocw.mit.edu/6-...
RUclips Playlist: • MIT 6.004 Computation ...
14.2.7 Direct-mapped Caches
License: Creative Commons BY-NC-SA
More information at ocw.mit.edu/terms
More courses at ocw.mit.edu

[@chris7056]

First Cacheing video that really hit home and made sense. Thank you MIT from UCSD.

[@tharalpius778]

there is an Indian guy who explains better

[@integralcialbay4633]

@@tharalpius778 Who?

[@simonpedersen3734]

thanks

[@Saens406]

still don't understand offset

[@ahmedabouelnasr4673]

I think if the structure of the cache can hold more than 1 word of data to each tag, we use the offset to fetch the right word ... correct me if I am wrong

[@Saens406]

@@ahmedabouelnasr4673 such that you can put several words on a single data cache line even though they are not used for the same variable?

[@ahmedabouelnasr4673]

@@Saens406 Such large block size(more than one word) serves the principle of spatial locality which assumes that data next to each other are more likely to get fetched sooner ... I found this, I hope it helps ruclips.net/video/YOABS3tTHVc/видео.html

[@aviralsaxena5153]

@@Saens406 The point of a cache is to take advantage of both temporal and spatial locality. When you store into a cache, you're grabbing a bunch of bytes at once, the target byte is included in the block. Depending on what type of addressability you support, most cases it's byte addressability, you need to be able to index that block of bytes to grab the right one, that's where the offset comes in.

[@erikumble]

I agree with what the others have said. Just for clarification, the reason we are not using the offset is entirely a design decision for the Beta specification, since any piece of memory is the full 32 bit word. If we had allowed accessing just 8 or 16 bits, then we would need to use the offset to specify which ones.

[@notconnor1249]

DEADBEEF