From a68adb0364109a398a1e350f3d9f32d0c5814085 Mon Sep 17 00:00:00 2001 From: Ichthyostega Date: Mon, 12 Feb 2024 17:53:47 +0100 Subject: [PATCH] Scheduler-test: need some instrumentation helper ...to sort out the interpretation of measurement results, the actual duration and concurrency of ComputationLoad invocations should be recorded, allowing to draw conclusions regarding the Scheduler's performance as opposed to further system and thread management effects due to concurrent operation under pressure. --- src/lib/incidence-count.hpp | 59 +++++++++ tests/15library.tests | 5 + tests/library/incidence-count-test.cpp | 96 ++++++++++++++ wiki/thinkPad.ichthyo.mm | 173 ++++++++++++++++++++++++- 4 files changed, 331 insertions(+), 2 deletions(-) create mode 100644 src/lib/incidence-count.hpp create mode 100644 tests/library/incidence-count-test.cpp diff --git a/src/lib/incidence-count.hpp b/src/lib/incidence-count.hpp new file mode 100644 index 000000000..5c2cde713 --- /dev/null +++ b/src/lib/incidence-count.hpp @@ -0,0 +1,59 @@ +/* + INCIDENCE-COUNT.hpp - instrumentation helper to watch possibly multithreaded task activations + + Copyright (C) Lumiera.org + 2024, Hermann Vosseler + + This program is free software; you can redistribute it and/or + modify it under the terms of the GNU General Public License as + published by the Free Software Foundation; either version 2 of + the License, or (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program; if not, write to the Free Software + Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + +*/ + +/** @file incidence-count.hpp + ** Record and evaluate concurrent activations. + ** This observation and instrumentation helper is typically used to capture + ** density of incidence and effective concurrency of performance critical tasks. + ** The overhead per measurement call amounts to one SystemClock invocation plus + ** some heap memory access, assuming that sufficient memory was pre-allocated + ** prior to the actual observation phase. Moreover, on first invocation per Thread, + ** a thread local ID is constructed, thereby incrementing an global atomic counter. + ** Statistics evaluation is comprised of integrating and sorting the captured + ** event log, followed by a summation pass. + ** @see IncidenceCount_test + ** @see vault::gear::TestChainLoad::ScheduleCtx + ** @see SchedulerStress_test + */ + + +#ifndef LIB_INCIDENCE_COUNT_H +#define LIB_INCIDENCE_COUNT_H + + +//#include "lib/meta/function.hpp" + +//#include + + +namespace lib { + +// using std::forward; + + class IncidenceCount + { + public: + }; + + +} // namespace lib +#endif /*LIB_INCIDENCE_COUNT_H*/ diff --git a/tests/15library.tests b/tests/15library.tests index aed131076..ecc06158f 100644 --- a/tests/15library.tests +++ b/tests/15library.tests @@ -424,6 +424,11 @@ return: 0 END +PLANNED "Instrumentation: concurrent calls" IncidenceCount_test < + + This program is free software; you can redistribute it and/or + modify it under the terms of the GNU General Public License as + published by the Free Software Foundation; either version 2 of + the License, or (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program; if not, write to the Free Software + Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + +* *****************************************************/ + +/** @file incidence-count-test.cpp + ** unit test \ref IncidenceCount_test + */ + + +#include "lib/test/run.hpp" +#include "lib/format-cout.hpp"//////////////TODO RLY? +#include "lib/incidence-count.hpp" + +//#include +#include + + +//using std::string; +using std::this_thread::sleep_for; + + +namespace lib { +namespace test{ + + + + + /***************************************************************************//** + * @test verifies capturing and restoring of std::ostream formatting state. + * @see ios-savepoint.hpp + */ + class IncidenceCount_test + : public Test + { + void + run (Arg) + { + demonstrate_usage(); + verify_incidentCount(); + verify_multithreadCount(); + } + + + + /** @test TODO + * @todo WIP 2/24 🔁 define ⟶ implement + */ + void + demonstrate_usage() + { + } + + + /** @test TODO verify proper counting of possibly overlapping incidences + * @todo WIP 2/24 🔁 define ⟶ implement + */ + void + verify_incidentCount() + { + UNIMPLEMENTED("verify proper counting of possibly overlapping incidences"); + } + + + /** @test TODO verify thread-safe operation under pressure + * @todo WIP 2/24 🔁 define ⟶ implement + */ + void + verify_multithreadCount() + { + UNIMPLEMENTED("verify thread-safe operation under pressure"); + } + }; + + LAUNCHER (IncidenceCount_test, "unit common"); + + +}} // namespace lib::test + diff --git a/wiki/thinkPad.ichthyo.mm b/wiki/thinkPad.ichthyo.mm index 726e3e1a2..4b454738e 100644 --- a/wiki/thinkPad.ichthyo.mm +++ b/wiki/thinkPad.ichthyo.mm @@ -110897,8 +110897,7 @@ Date:   Thu Apr 20 18:53:17 2023 +0200
Fokus ⟹ Ziel ⟹ es geht um Halten ⟷ Brechen

- - + @@ -110932,6 +110931,176 @@ Date:   Thu Apr 20 18:53:17 2023 +0200
+ + + + + + + + + + +

+ Zwar ist die ComputationalLoad auf die jeweilige Hardware dynamisch geeicht, aber in den Einzelbeobachtungen habe ich immer ganz erhebliche Abweichungen für Node-Aufrufe mit Load gesehen. Daher ist die Frage, wie verhält sich die insgesamt (kumuliert, d.h. integriert) verbrachte Zeit zu der nominell geeichten Zeit? Daraus ergäbe sich ein Korrektur-Faktor, den man aus dem StressFac herausrechnen könnte +

+ + + + + + + + + + + + +

+ Untersuchung per Godbolt.org : SystemClock-Aufrufe werden nicht wegoptimiert +

+ + + + + + + + +

+ ...das ist überraschend, denn sonst wird wirklich alles wie erwartet eingedampft und es bleibt nur die eigentliche Nutzfunktion übrig; der Grund ist aber leicht einzusehen: es handelt sich um einen externen Call, der Seiteneffekte haben könnte. +

+ + + + + + + + + + + + + + + + + + + + + + + + + + +

+ das bedeutet: jeder Thread hat seinen eigenen »Slot«, den er ohne jedwede Synchronisation zugreifen kann +

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +