In the Radix sort example, I assume the size of the data set is held constant? The size of the data set would definitely have an impact on False sharing. If our data set was much larger than the size of a cache line, I think we would experience a much lower percentage of false sharing.
nemo
Why does an increase in cache line size decrease the number of cache misses due to true sharing?
pdp
@nemo: Because more of the working set of a program can be accommodated in a single cache line with higher cache line sizes thereby causing fewer cache misses (True sharing decreases)
Drizzle
Why does in increase in cache line size decrease the miss rate due to update?
In the Radix sort example, I assume the size of the data set is held constant? The size of the data set would definitely have an impact on False sharing. If our data set was much larger than the size of a cache line, I think we would experience a much lower percentage of false sharing.
Why does an increase in cache line size decrease the number of cache misses due to true sharing?
@nemo: Because more of the working set of a program can be accommodated in a single cache line with higher cache line sizes thereby causing fewer cache misses (True sharing decreases)
Why does in increase in cache line size decrease the miss rate due to update?
What do 'Capacity/Conflict' and 'Cold' mean here?
@achchange: Good time to review this slide!
The takeaway here is that if you don't have a lot of spatial locality, increasing cache line size will likely increase false sharing