- move Gaudi::Funtional to GaudiFunctional subdir
  - headers are now under "Gaudi/Functional"
  - legacy GaudiAlg corresponding header redirect to the new ones
- replace GaudiAlgorithm with Algorithm wherever possible in tests and
  utilities
- add Gaudi::Sequencer basic implementation to replace GaudiSequencer
- move GaudiAlg specific tests, examples, dictionaries and Python modules
  to GaudiAlg subdir
- add CMake option GAUDI_ENABLE_GAUDIALG to allow skipping GaudiAlg
  (default to true)
- minor fixes and clean ups

Benjamin Michael Wynne authored 3 years ago and Marco Clemencic committed 3 years ago

patch generated by https://gitlab.cern.ch/bwynne/Gaudi_gaudi/-/jobs/12974967

extend the category of tasks that can be scheduled preemptively to 'CPU-blocking' from just 'I/O-bound' (terminological change only)

Gitlab CI authored 6 years ago and Marco Clemencic committed 6 years ago

patch generated by standalone job

This reverts merge request !462

The original data handle holder was solely focused on event data, and
did not acknowledge the existence of conditions. Introducing conditions
in Gaudi therefore requires clarifying the distinction.

Since it was agreed at a Gaudi workshop that the distinction should not
be made at the DataObjID level, it must be made in functions that emit
or accept DataObjIDs.

add a call to setFilterPassed in the CPUCruncher and add GaudiHive test for ControlFlowBranching+CrossBranchDataFlow

This reverts commit d92ff23f.

The use of boost::string_ref (to be replaced by std::string_view) reduce
significantly the space required by Property, avoiding pointless
duplications in memory (see GAUDI-1229).

With the `profile_Property` executable I measured a reduction of a
factor 2 in both size (272 -> 128 bytes) and time (0.2 -> 0.1 us) for
each created Property (always the same name and doc).

The drawback of the approach is that if we pass a temporary string to the
Property constructor we get undefined behaviour.

CPUCruncher extended to replace part of crunching time with plain sleeping (needed to simulate I/O-bound behaviourof an algorithm)

draft implementation of IO-bound behaviour of algorithms and their scheduling, somewhat messy here and there..

new declareInput/declareOutput with DataObjectHandles directly in the class, DataObjectHandle is no longer an IInterface