Cache

Question 1

Describe 1) what cache memory can improve in computing system and how 2) is this achieved?

Answer 1

• Cache memory can improve the average computational performance of a microprocessor system by having a parcial copy of the data from the main memory stored in a local, faster accessible cache memory.
• Cache is useful due the spatial locality and temporal locality properties.

Question 2

What is spatial locality

Answer 2

data/ code you access most likely has data you will access soon in its vicinity.

Question 3

What is temporal locality

Answer 3

data / code you access now, you will likely access soon again

Question 4

compare and contrast direct mapping and associative cache mapping policies.

Answer 4

Direct Mapping
Advantage:
- temporal behavior can be analyzed
- simple implementation
Disadvantage:
- inefficient utilization of cache resources, if the program accesses multiple addresses frequently which are mapped to the same cache-line (conflict miss), many cache misses occur slowing down the execution.
- many unneccessary read/write operations

Associative mapping
Advantage:
- better use of cache space. less on-chip consuming
- simple for timing analysis
Disadvantage: more complex cache control

Question 5

What is write through

Answer 5

done synchronously both to cache and main memory (next level cache)

Question 6

What is write back

Answer 6

initially writing only to cache, write to main memory (next level cache) when data is replaced by new content
- more complex

Question 7

Pro and cons for Fully associative cache

Answer 7

Adbvantages
- Simple and consumes less chip area
- Reasonably mimics LRU, hence, better performance than direct mapped
Drawbacks
- Slightly more complicated to predict
- slightly less performance compared to LRU

Question 8

Set associative cache

Answer 8

Hybrid of gully associative and direct mapping
cache lines are grouped in sets
Data can go to any cache line within the set

Question 9

Compare and contrast write through and write back

Answer 9

Write through
- A lower level memory is also updated on a write access
- Number of writes are higher;
- Simple but more energy consuming; - Depending on application, execution time may be longer;
- Generally employed if lower level memory is fast enough

Write back
- A lower memory is updated only when dirty line is evicted;
- less number of writes;
- Complex but more efficient;
- Execution time will be mostly shorter;
- Generally employed if lower level memory is slow (last level cache)