I want to know your opinion about this library? (If it is useful to improve, neurons cache efficacy further)while L1 cache in a modern CPU is random access, access to the main memory
(DRAM) is fastest when sequential. Morever, on a PC access to cache is much, much faster than
access to main memory, even when main memory is on the same circuit board. Consequently,
a program developer should attempt to partition calculations into blocks of data that will fit
into cache. This is done automatically for some optimized libraries such as ATLAS, used for
linear algebra.
Also, I have decided to buy a new Multiprocessor workstation, to work on my neuron project. In that book, there is another phrase that say
I have 2 options available -->For high-performance computing (HPC) applications, better performance has often been obtained
when SMT was disabled. It remains to be determined the efficiency of SMT on the Intel Core
i7 processor introduced at the end of 2008.
1) to buy a core i7 950, --> (High End CPU's benchmark) OR
2) buy a 2 xeon 5405 processors (Multiple CPU Systems benchmark)
both of them are in the same range of price. I want to know which one is recommended for neuron.