sure, tackeling it on a higher level is a fundamentaly good idea, but it do require some extra work.
Oh, and about the cost of virtual functions for a matrix multiplicaion it''s neglible the cost of dereferencing the pointer is what 1-2 clock cycles having SSE code doing the muls does help.
Stategy design pattern with different processor implementations
Thought about using dll''s ?
you could create a dll for each processor and at the start up of the app load the correct dll for the processor.
you could create a dll for each processor and at the start up of the app load the correct dll for the processor.
quote:Original post by Anonymous Poster
Thought about using dll''s ?
you could create a dll for each processor and at the start up of the app load the correct dll for the processor.
That would have the same effect as my suggestion with the only differance that the could would also get dynamicly linked instead of staticly linked.
But sure a DLL does offer some advantages...
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement