def minimum_completion_time(tg, ag, shm, logging):
"""
:param tg: Task Graph
:param ag: Architecture Graph
:param shm: System Health Map
:param logging: logging File
:return: (TG, AG)
"""
# The difference with Min Min or Max Min is that we don't add priorities to
# tasks based on their WCET but we randomly choose a task and schedule it...
# Note :: This heuristic is not taking task ciriticality into account...
print ("===========================================")
print ("STARTING MIN COMPLETION TIME MAPPING")
for task_to_be_mapped in tg.nodes():
chosen_node = random.choice(Mapping_Functions.nodes_with_smallest_ct(ag, tg, shm, task_to_be_mapped))
tg.node[task_to_be_mapped]['task'].node = chosen_node
ag.node[chosen_node]['PE'].mapped_tasks.append(task_to_be_mapped)
ag.node[chosen_node]['PE'].utilization += tg.node[task_to_be_mapped]['task'].wcet
node_speed_down = 1+((100.0-shm.node[chosen_node]['NodeSpeed'])/100)
task_execution_on_node = tg.node[task_to_be_mapped]['task'].wcet*node_speed_down
completion_on_node = tg.node[task_to_be_mapped]['task'].release + task_execution_on_node
Scheduling_Functions_Nodes.add_tg_task_to_node(tg, ag, task_to_be_mapped, chosen_node,
tg.node[task_to_be_mapped]['task'].release,
completion_on_node, None)
print ("\tTASK "+str(task_to_be_mapped)+" MAPPED ON NODE: "+str(chosen_node))
print ("MIN COMPLETION TIME MAPPING FINISHED...")
Scheduling_Reports.report_mapped_tasks(ag, logging)
return tg, ag
def min_execution_time(tg, ag, shm, logging):
"""
:param tg: Task Graph
:param ag: Architecture Graph
:param shm: System Health Map
:param logging: logging file
:return: (TG, AG)
"""
# this sounds a little stupid because there are no job specific machines...
# we can Add Specific Accelerators or define different run time on different
# PEs so this becomes more interesting...
print ("===========================================")
print ("STARTING MIN EXECUTION TIME MAPPING")
for task_to_be_mapped in tg.nodes():
chosen_node = random.choice(Mapping_Functions.fastest_nodes(ag, shm))
tg.node[task_to_be_mapped]['task'].node = chosen_node
ag.node[chosen_node]['PE'].mapped_tasks.append(task_to_be_mapped)
ag.node[chosen_node]['PE'].utilization += tg.node[task_to_be_mapped]['task'].wcet
node_speed_down = 1+((100.0-shm.node[chosen_node]['NodeSpeed'])/100)
task_execution_on_node = tg.node[task_to_be_mapped]['task'].wcet*node_speed_down
completion_on_node = tg.node[task_to_be_mapped]['task'].release + task_execution_on_node
Scheduling_Functions_Nodes.add_tg_task_to_node(tg, ag, task_to_be_mapped, chosen_node,
tg.node[task_to_be_mapped]['task'].release,
completion_on_node, None)
print ("\tTASK "+str(task_to_be_mapped)+" MAPPED ON NODE: "+str(chosen_node))
print ("MIN EXECUTION TIME MAPPING FINISHED...")
Scheduling_Reports.report_mapped_tasks(ag, logging)
return tg, ag
def min_min_mapping(tg, ag, shm, logging):
"""
:param tg: Task Graph
:param ag: Architecture Graph
:param shm: System Health Map
:param logging: logging file
:return: (TG, AG)
"""
# this function finds the task with the smallest WCET and
# maps it on the machine that can offer smallest completion time...
# this means that the mapping algorithm has to take into account the mapping
# of the edges of the task graph on the links.
# Note:: this is a heuristic for independent tasks... so we are not going to
# schedule any link
# Note 2:: This heuristic is not taking task ciriticality into account...
print("===========================================")
print("STARTING MIN-MIN MAPPING")
shortest_tasks = Mapping_Functions.unmapped_task_with_smallest_wcet(
tg, logging)
while len(shortest_tasks) > 0:
task_to_be_mapped = shortest_tasks.pop()
# map the task on the Node that yields smallest Completion time
candidate_nodes = Mapping_Functions.nodes_with_smallest_ct(
ag, tg, shm, task_to_be_mapped)
print("\tCANDIDATE NODES FOR MAPPING: " + str(candidate_nodes))
if len(candidate_nodes) > 0:
chosen_node = random.choice(candidate_nodes)
print("\t\tMAPPING TASK " + str(task_to_be_mapped) +
" WITH RELEASE: " +
str(tg.node[task_to_be_mapped]['task'].release) +
" ---> NODE: " + str(chosen_node))
tg.node[task_to_be_mapped]['task'].node = chosen_node
ag.node[chosen_node]['PE'].mapped_tasks.append(task_to_be_mapped)
ag.node[chosen_node]['PE'].utilization += tg.node[
task_to_be_mapped]['task'].wcet
node_speed_down = 1 + (
(100.0 - shm.node[chosen_node]['NodeSpeed']) / 100)
task_execution_on_node = tg.node[task_to_be_mapped][
'task'].wcet * node_speed_down
completion_on_node = tg.node[task_to_be_mapped][
'task'].release + task_execution_on_node
Scheduling_Functions_Nodes.add_tg_task_to_node(
tg, ag, task_to_be_mapped, chosen_node,
tg.node[task_to_be_mapped]['task'].release, completion_on_node,
None)
if len(shortest_tasks) == 0:
shortest_tasks = Mapping_Functions.unmapped_task_with_smallest_wcet(
tg, logging)
print("MIN-MIN MAPPING FINISHED...")
Scheduling_Reports.report_mapped_tasks(ag, logging)
return tg, ag
def max_min_mapping(tg, ag, shm, logging):
"""
:param tg: Task Graph
:param ag: Architecture Graph
:param shm: System Health Map
:param logging: logging file
:return: (TG, AG)
"""
# this function finds the task with the biggest WCET and
# maps it on the machine that can offer smallest completion time...
# this means that the mapping algorithm has to take into account the mapping
# of the edges of the task graph on the links.
# Note:: this is a heuristic for independent tasks... so we are not going to
# schedule any link
# Note 2:: This heuristic is not taking task ciriticality into account...
print ("===========================================")
print ("STARTING MAX-MIN MAPPING")
longest_tasks = Mapping_Functions.unmapped_task_with_biggest_wcet(tg, logging)
while len(longest_tasks) > 0:
task_to_be_mapped = longest_tasks.pop()
# map the task on the Node that yields smallest Completion time
candidate_nodes = Mapping_Functions.nodes_with_smallest_ct(ag, tg, shm, task_to_be_mapped)
print ("CANDIDATE NODES FOR MAPPING: "+str(candidate_nodes))
if len(candidate_nodes) > 0:
chosen_node = random.choice(candidate_nodes)
if len(candidate_nodes) > 1:
print ("\tMAPPING TASK "+str(task_to_be_mapped)+" WITH RELEASE: " +
str(tg.node[task_to_be_mapped]['task'].release) +
" ---> NODE: "+str(chosen_node)+" (RANDOMLY CHOSEN FROM CANDIDATES)")
else:
print ("\tMAPPING TASK "+str(task_to_be_mapped)+" WITH RELEASE: " +
str(tg.node[task_to_be_mapped]['task'].release) +
" ---> NODE: "+str(chosen_node))
tg.node[task_to_be_mapped]['task'].node = chosen_node
ag.node[chosen_node]['PE'].mapped_tasks.append(task_to_be_mapped)
ag.node[chosen_node]['PE'].utilization += tg.node[task_to_be_mapped]['task'].wcet
node_speed_down = 1+((100.0-shm.node[chosen_node]['NodeSpeed'])/100)
task_execution_on_node = tg.node[task_to_be_mapped]['task'].wcet*node_speed_down
completion_on_node = tg.node[task_to_be_mapped]['task'].release + task_execution_on_node
Scheduling_Functions_Nodes.add_tg_task_to_node(tg, ag, task_to_be_mapped, chosen_node,
tg.node[task_to_be_mapped]['task'].release,
completion_on_node, None)
if len(longest_tasks) == 0:
longest_tasks = Mapping_Functions.unmapped_task_with_biggest_wcet(tg, logging)
print ("MIN-MAX MAPPING FINISHED...")
Scheduling_Reports.report_mapped_tasks(ag, logging)
return tg, ag
def minimum_completion_time(tg, ag, shm, logging):
"""
:param tg: Task Graph
:param ag: Architecture Graph
:param shm: System Health Map
:param logging: logging File
:return: (TG, AG)
"""
# The difference with Min Min or Max Min is that we don't add priorities to
# tasks based on their WCET but we randomly choose a task and schedule it...
# Note :: This heuristic is not taking task ciriticality into account...
print("===========================================")
print("STARTING MIN COMPLETION TIME MAPPING")
for task_to_be_mapped in tg.nodes():
chosen_node = random.choice(
Mapping_Functions.nodes_with_smallest_ct(ag, tg, shm,
task_to_be_mapped))
tg.node[task_to_be_mapped]['task'].node = chosen_node
ag.node[chosen_node]['PE'].mapped_tasks.append(task_to_be_mapped)
ag.node[chosen_node]['PE'].utilization += tg.node[task_to_be_mapped][
'task'].wcet
node_speed_down = 1 + (
(100.0 - shm.node[chosen_node]['NodeSpeed']) / 100)
task_execution_on_node = tg.node[task_to_be_mapped][
'task'].wcet * node_speed_down
completion_on_node = tg.node[task_to_be_mapped][
'task'].release + task_execution_on_node
Scheduling_Functions_Nodes.add_tg_task_to_node(
tg, ag, task_to_be_mapped, chosen_node,
tg.node[task_to_be_mapped]['task'].release, completion_on_node,
None)
print("\tTASK " + str(task_to_be_mapped) + " MAPPED ON NODE: " +
str(chosen_node))
print("MIN COMPLETION TIME MAPPING FINISHED...")
Scheduling_Reports.report_mapped_tasks(ag, logging)
return tg, ag
def min_execution_time(tg, ag, shm, logging):
"""
:param tg: Task Graph
:param ag: Architecture Graph
:param shm: System Health Map
:param logging: logging file
:return: (TG, AG)
"""
# this sounds a little stupid because there are no job specific machines...
# we can Add Specific Accelerators or define different run time on different
# PEs so this becomes more interesting...
print("===========================================")
print("STARTING MIN EXECUTION TIME MAPPING")
for task_to_be_mapped in tg.nodes():
chosen_node = random.choice(Mapping_Functions.fastest_nodes(ag, shm))
tg.node[task_to_be_mapped]['task'].node = chosen_node
ag.node[chosen_node]['PE'].mapped_tasks.append(task_to_be_mapped)
ag.node[chosen_node]['PE'].utilization += tg.node[task_to_be_mapped][
'task'].wcet
node_speed_down = 1 + (
(100.0 - shm.node[chosen_node]['NodeSpeed']) / 100)
task_execution_on_node = tg.node[task_to_be_mapped][
'task'].wcet * node_speed_down
completion_on_node = tg.node[task_to_be_mapped][
'task'].release + task_execution_on_node
Scheduling_Functions_Nodes.add_tg_task_to_node(
tg, ag, task_to_be_mapped, chosen_node,
tg.node[task_to_be_mapped]['task'].release, completion_on_node,
None)
print("\tTASK " + str(task_to_be_mapped) + " MAPPED ON NODE: " +
str(chosen_node))
print("MIN EXECUTION TIME MAPPING FINISHED...")
Scheduling_Reports.report_mapped_tasks(ag, logging)
return tg, ag