Main Memory Bandwidth test/testing ?
Main Memory Bandwidth Measurement
Most of the High Performance Computing related Developers face problems while optimising their code
some times we may find memory bandwidth limitations also
how to test memory bandwidth of a Linux system?
we can check the memory bandwidth on Linux by downloading rpm or source from following link
http://rpmfind.net/linux/rpm2html/search.php?query=bandwidth&submit=Search+...&system=&arch=
or we can search for bandwidth in rpmfind.net/linux to find the rpm for memory bandwidth test
why we need to test memory bandwidth of system?
Most of the modern system are coming with more processors and each processor
or sub cores having separate memory channels.
when process is executing on one processor that process will yield more memory bandwidth only when the process is accessing the memory directly connected to the processor. if the process executing on one processor if tries access memory which not directly connected to the same processor and connected via another processor will yield less memory bandwidth because it has to access the memory through the second processor either HyperTransport (HT) or Quick Path Interconnect (QPI) (for example)
these system will feature Non-Uniform Memory Access (NUMA)
Main Memory Bandwidth Measurement
Most of the High Performance Computing related Developers face problems while optimising their code
some times we may find memory bandwidth limitations also
how to test memory bandwidth of a Linux system?
we can check the memory bandwidth on Linux by downloading rpm or source from following link
http://rpmfind.net/linux/rpm2html/search.php?query=bandwidth&submit=Search+...&system=&arch=
or we can search for bandwidth in rpmfind.net/linux to find the rpm for memory bandwidth test
why we need to test memory bandwidth of system?
Most of the modern system are coming with more processors and each processor
or sub cores having separate memory channels.
when process is executing on one processor that process will yield more memory bandwidth only when the process is accessing the memory directly connected to the processor. if the process executing on one processor if tries access memory which not directly connected to the same processor and connected via another processor will yield less memory bandwidth because it has to access the memory through the second processor either HyperTransport (HT) or Quick Path Interconnect (QPI) (for example)
these system will feature Non-Uniform Memory Access (NUMA)
