|
@@ -1126,15 +1126,18 @@ lower bound for the total execution time of your tasks. If StarPU was compiled
|
|
|
with the glpk library installed, @code{starpu_bound_compute} can be used to solve it
|
|
with the glpk library installed, @code{starpu_bound_compute} can be used to solve it
|
|
|
immediately and get the optimized minimum.
|
|
immediately and get the optimized minimum.
|
|
|
|
|
|
|
|
-The @code{deps} parameter tells StarPU whether to take tasks and data
|
|
|
|
|
|
|
+The @code{deps} parameter tells StarPU whether to take tasks and implicit data
|
|
|
dependencies into account. It must be understood that the linear programming
|
|
dependencies into account. It must be understood that the linear programming
|
|
|
-problem size is quadratic with number of tasks and thus the time to solve it
|
|
|
|
|
|
|
+problem size is quadratic with the number of tasks and thus the time to solve it
|
|
|
will be very long, typically one minute for just 14 tasks. Setting @code{deps}
|
|
will be very long, typically one minute for just 14 tasks. Setting @code{deps}
|
|
|
-to 0 will only takes into account the actual computations on processing
|
|
|
|
|
|
|
+to 0 will only take into account the actual computations on processing
|
|
|
units. It however still properly takes into account the varying performances of
|
|
units. It however still properly takes into account the varying performances of
|
|
|
kernels and processing units, which is quite more accurate than just comparing
|
|
kernels and processing units, which is quite more accurate than just comparing
|
|
|
StarPU performances with the fastest of the kernels being used.
|
|
StarPU performances with the fastest of the kernels being used.
|
|
|
|
|
|
|
|
|
|
+Note that all this however doesn't take into account data transfer, which is
|
|
|
|
|
+assumed to be completely overlapped.
|
|
|
|
|
+
|
|
|
@node More examples
|
|
@node More examples
|
|
|
@section More examples
|
|
@section More examples
|
|
|
|
|
|
