[WIP] Task

[MenoIy]

Approach
Create new decorator task that will allow pyccel to recognize that the current function will be used in task parallelism and automatically calculate its dependencies.
task decorator have two optional arguments :

• type : type of the task if its 'master' that mean that this function is responsible for running other task functions, default 'child'
• threads_num : number of threads reserved for this task region

A task function can only run in parallel if the parent function is a task of type 'master'.

@task
def A():
    # do something

@task('master')
def main():
    a = A()

Block of task is identifies by the first occurrence of a task (currently : in function call / Assign) until the first non task element or the end of the program.

# Function  A and B are tasks

@task('master')
def main():
    x, y = # do something

    # this will run in parallel
    # ------
    a = A(x) # first task A
    b = B(y) # second task B
    # ------

    c = a + b # first non task block

Dependencies between tasks is calculated by collecting all the information of task function into a namespace list
that help to update dependencies(In, out) shared elements ( for openmp) and precoders function (for other library)
as we go on semantic stage.

[EmilyBourne]

Shouldn't these be in ast/omp.py ?

[EmilyBourne]

Typo complets->complete