| Line 446... |
Line 446... |
| 446 |
The parameter \code{affinity.list} of the \code{mclapply} function
|
446 |
The parameter \code{affinity.list} of the \code{mclapply} function
|
| 447 |
can be used to run elements of the input vector \code{X} of \code{mclapply} on
|
447 |
can be used to run elements of the input vector \code{X} of \code{mclapply} on
|
| 448 |
specific CPUs.
|
448 |
specific CPUs.
|
| 449 |
\code{affinity.list} is a vector (atomic or list) containing the CPU
|
449 |
\code{affinity.list} is a vector (atomic or list) containing the CPU
|
| 450 |
affinity mask for each element of \code{X}, it describes on which CPU
|
450 |
affinity mask for each element of \code{X}, it describes on which CPU
|
| 451 |
(core or hyperthread unit) a given item is allowed to run, see
|
451 |
(core or hyper-thread unit) a given item is allowed to run, see
|
| 452 |
\code{? mcaffinity}.
|
452 |
\code{? mcaffinity}.
|
| 453 |
This can be helpful, if the elements of \code{X} (parallel jobs) have a
|
453 |
This can be helpful, if the elements of \code{X} (parallel jobs) have a
|
| 454 |
high variance of completion time or if the hardware architecture
|
454 |
high variance of completion time or if the hardware architecture
|
| 455 |
is heterogeneous. It also enables the development of scheduling
|
455 |
is heterogeneous. It also enables the development of scheduling
|
| 456 |
strategies for optimizing the overall runtime of independent parallel jobs.
|
456 |
strategies for optimizing the overall runtime of independent parallel jobs.
|
The R Project SVN
R