Do all of these mappings occur one after another? If so, is there any kind of communication that goes on in the background between these "layers". If not, what is stopping us from taking ownership of the entire process?
yuel1
@Berry, I believe we are not encouraged from taking ownership of the entire process because that would effectively lower the level of abstraction provided to the programmer and thus make reasoning over performance much much harder.
yuel1
Interestingly enough, the first mapping decision would probably lead to much better caching performance as suggested by the caching lectures.
Do all of these mappings occur one after another? If so, is there any kind of communication that goes on in the background between these "layers". If not, what is stopping us from taking ownership of the entire process?
@Berry, I believe we are not encouraged from taking ownership of the entire process because that would effectively lower the level of abstraction provided to the programmer and thus make reasoning over performance much much harder.
Interestingly enough, the first mapping decision would probably lead to much better caching performance as suggested by the caching lectures.