ControlFlowBranching+CrossBranchDataFlow.py

Go to the documentation of this file.

#!/usr/bin/env gaudirun.py
 
"""
Typically, an EVTREJECTED algorithm is used to "short-circuit" the execution flow within
the group of algorithms it belongs to. Within such early-terminated group, a false-negative
data dependency resolution can never occur since all downstream algorithms are skipped.
However, there might be other algorithms - outside of the early-terminated group - that depend
on data outputs of the EVTREJECTED algorithm. The test emulates this scenario.
"""
 
from Configurables import (
    AlgResourcePool,
    AvalancheSchedulerSvc,
    CPUCruncher,
    CPUCrunchSvc,
    GaudiSequencer,
    HiveSlimEventLoopMgr,
    HiveWhiteBoard,
)
from Gaudi.Configuration import *
 
# metaconfig
evtslots = 1
evtMax = 1
algosInFlight = 1
 
InertMessageSvc(OutputLevel=INFO)
 
whiteboard = HiveWhiteBoard("EventDataSvc", EventSlots=evtslots, OutputLevel=INFO)
 
slimeventloopmgr = HiveSlimEventLoopMgr(
    SchedulerName="AvalancheSchedulerSvc", OutputLevel=DEBUG
)
 
AvalancheSchedulerSvc(ThreadPoolSize=algosInFlight, OutputLevel=DEBUG)
 
AlgResourcePool(OutputLevel=DEBUG)
 
CPUCrunchSvc(shortCalib=True)
 
# Assemble data flow graph
# algorithm that triggers an early exit from "Branch2"
a1 = CPUCruncher("AlgA", InvertDecision=True)
a1.outKeys = ["/Event/A"]
 
a2 = CPUCruncher("AlgB")
a2.inpKeys = ["/Event/A"]
 
a3 = CPUCruncher("AlgC")
a3.inpKeys = ["/Event/A"]
 
for a in [a1, a2, a3]:
    a.avgRuntime = 0.01
 
# Assemble control flow graph
branch1 = GaudiSequencer("Branch1", ModeOR=False, ShortCircuit=False)
branch2 = GaudiSequencer("Branch2", ModeOR=False, ShortCircuit=True, Sequential=True)
 
branch2.Members = [a1, a2]
branch1.Members = [branch2, a3]
 
ApplicationMgr(
    EvtMax=evtMax,
    EvtSel="NONE",
    ExtSvc=[whiteboard],
    EventLoop=slimeventloopmgr,
    TopAlg=[branch1],
    MessageSvcType="InertMessageSvc",
    OutputLevel=INFO,
)