Hi all,
I am not expert on the NUMA area and the scheduler but I was wondering the following:
On numa a process can define a memory policy to optimize access because remote memory access is suboptimal.
On a multicluster (aka multipackage), eg. Bi-Xeon 4 cores, the scheduler load balancing can spawn/balance a task to the second cluster even if the task is resulting from a fork or a pthread_create. If I am correct that could impact the performances due to L2 cache misses, no ?
Won't it be more efficient to tie a task to a cluster when it shares memory resources with another task on the same cluster ? and when the task does 'exec', this constraint is removed.
Thanks -- Daniel -- <http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs Follow Linaro: <http://www.facebook.com/pages/Linaro> Facebook | <http://twitter.com/#!/linaroorg> Twitter | <http://www.linaro.org/linaro-blog/> Blog -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
