From: Martin Quinson Date: Sun, 17 Mar 2019 16:49:58 +0000 (+0100) Subject: Simplify the exec-ptask example X-Git-Tag: v3_22~74 X-Git-Url: http://info.iut-bm.univ-fcomte.fr/pub/gitweb/simgrid.git/commitdiff_plain/096bef11ea84253450578d54bfa97f36c1d5984e Simplify the exec-ptask example - Remove the energy thing out of the picture - Don't use the trick of using ptask to do remote exec now that we have a much cleaner way of doing remote exec - Don't test the homogeneous compute-only case. The heterogeneous compute-only is sufficient. Everybody wants the examples to be short and instructive. --- diff --git a/examples/s4u/README.rst b/examples/s4u/README.rst index 1cb6ee033a..8610665eb2 100644 --- a/examples/s4u/README.rst +++ b/examples/s4u/README.rst @@ -208,18 +208,19 @@ Executions on the CPU on which they occur during their execution. |br| `examples/s4u/exec-remote/s4u-exec-remote.cpp `_ + - **Parallel executions:** + These objects are convenient abstractions of parallel + computational kernels that span over several machines, such as a + PDGEM and the other ScaLAPACK routines. Note that this only works + with the "ptask_L07" host model (`--cfg=host/model:ptask_L07`). + |br| `examples/s4u/exec-ptask/s4u-exec-ptask.cpp `_ + - **Using Pstates on a host:** Shows how define a set of pstatesfor a host in the XML, and how the current pstate can be accessed/changed with :cpp:func:`simgrid::s4u::Host::get_pstate_speed` and :cpp:func:`simgrid::s4u::Host::set_pstate`. |br| `examples/s4u/exec-dvfs/s4u-exec-dvfs.cpp `_ |br| `examples/platforms/energy_platform.xml `_ - - **Parallel executions:** - These objects are convenient abstractions of parallel - computational kernels that span over several machines, such as a - PDGEM and the other ScaLAPACK routines. - |br| `examples/s4u/exec-ptask/s4u-exec-ptask.cpp `_ - I/O on Disks and Files ---------------------- diff --git a/examples/s4u/exec-ptask/s4u-exec-ptask.cpp b/examples/s4u/exec-ptask/s4u-exec-ptask.cpp index c243494d00..c3f804bea1 100644 --- a/examples/s4u/exec-ptask/s4u-exec-ptask.cpp +++ b/examples/s4u/exec-ptask/s4u-exec-ptask.cpp @@ -18,10 +18,9 @@ * Please note that you must have the LV07 platform model enabled to use such constructs. */ -#include "simgrid/plugins/energy.h" #include -XBT_LOG_NEW_DEFAULT_CATEGORY(s4u_energyptask, "Messages specific for this s4u example"); +XBT_LOG_NEW_DEFAULT_CATEGORY(s4u_ptask, "Messages specific for this s4u example"); static void runner() { @@ -36,8 +35,8 @@ static void runner() std::vector communication_amounts; /* ------[ test 1 ]----------------- */ - computation_amounts.assign(hosts.size(), 1e9 /*1Gflop*/); - communication_amounts.assign(hosts.size() * hosts.size(), 0); + computation_amounts.assign(hosts_count, 1e9 /*1Gflop*/); + communication_amounts.assign(hosts_count * hosts_count, 0); for (size_t i = 0; i < hosts_count; i++) for (size_t j = i + 1; j < hosts_count; j++) communication_amounts[i * hosts_count + j] = 1e7; // 10 MB @@ -45,51 +44,33 @@ static void runner() simgrid::s4u::this_actor::parallel_execute(hosts, computation_amounts, communication_amounts); /* ------[ test 2 ]----------------- */ - XBT_INFO("We can do the same with a timeout of one second enabled."); - computation_amounts.assign(hosts.size(), 1e9 /*1Gflop*/); - communication_amounts.assign(hosts.size() * hosts.size(), 0); + XBT_INFO("We can do the same with a timeout of 10 seconds enabled."); + computation_amounts.assign(hosts_count, 1e9 /*1Gflop*/); + communication_amounts.assign(hosts_count * hosts_count, 0); for (size_t i = 0; i < hosts_count; i++) for (size_t j = i + 1; j < hosts_count; j++) communication_amounts[i * hosts_count + j] = 1e7; // 10 MB try { simgrid::s4u::this_actor::parallel_execute(hosts, computation_amounts, communication_amounts, - 1.0 /* timeout (in seconds)*/); - XBT_WARN("Woops, this did not timeout as expected... Please report that bug."); - } catch (xbt_ex& e) { - /* Do nothing this exception on timeout was expected */ - XBT_DEBUG("Caught expected exception: %s", e.what()); + 10.0 /* timeout (in seconds)*/); + xbt_die("Woops, this did not timeout as expected... Please report that bug."); + } catch (simgrid::TimeoutError& e) { + XBT_INFO("Caught the expected timeout exception."); } /* ------[ test 3 ]----------------- */ - XBT_INFO("Then, build a parallel task involving only computations and no communication (1 Gflop per node)"); - computation_amounts.assign(hosts.size(), 1e9 /*1Gflop*/); - communication_amounts.clear(); /* no comm */ + XBT_INFO("Then, build a parallel task involving only computations (of different amounts) and no communication"); + computation_amounts = {3e8, 6e8, 1e9}; // 300Mflop, 6Mflop, 1Gflop + communication_amounts.clear(); // no comm simgrid::s4u::this_actor::parallel_execute(hosts, computation_amounts, communication_amounts); /* ------[ test 4 ]----------------- */ - XBT_INFO("Then, build a parallel task involving only heterogeneous computations and no communication"); - computation_amounts.resize(hosts.size()); - for (size_t i = 0; i < hosts_count; i++) - computation_amounts[i] = 5 * (i + 1) * 1e8; // 500Mflop, 1Gflop, 1.5Gflop - communication_amounts.clear(); /* no comm */ - simgrid::s4u::this_actor::parallel_execute(hosts, computation_amounts, communication_amounts); - - /* ------[ test 5 ]----------------- */ XBT_INFO("Then, build a parallel task with no computation nor communication (synchro only)"); computation_amounts.clear(); communication_amounts.clear(); simgrid::s4u::this_actor::parallel_execute(hosts, computation_amounts, communication_amounts); - /* ------[ test 6 ]----------------- */ - XBT_INFO("Finally, trick the ptask to do a 'remote execution', on host %s", hosts[1]->get_cname()); - std::vector remote; - remote.push_back(hosts[1]); - computation_amounts.assign(1, 1e9); - communication_amounts.clear(); - - simgrid::s4u::this_actor::parallel_execute(remote, computation_amounts, communication_amounts); - XBT_INFO("Goodbye now!"); } @@ -97,15 +78,9 @@ int main(int argc, char* argv[]) { simgrid::s4u::Engine e(&argc, argv); - xbt_assert(argc <= 3, "1Usage: %s [--energy]", argv[0]); - xbt_assert(argc >= 2, "2Usage: %s [--energy]", argv[0]); - - if (argc == 3 && argv[2][2] == 'e') - sg_host_energy_plugin_init(); + xbt_assert(argc == 2, "Usage: %s ", argv[0]); e.load_platform(argv[1]); - - /* Pick a process, no matter which, from the platform file */ simgrid::s4u::Actor::create("test", simgrid::s4u::Host::by_name("MyHost1"), runner); e.run(); diff --git a/examples/s4u/exec-ptask/s4u-exec-ptask.tesh b/examples/s4u/exec-ptask/s4u-exec-ptask.tesh index 40d34e6ccf..e8dde985c7 100644 --- a/examples/s4u/exec-ptask/s4u-exec-ptask.tesh +++ b/examples/s4u/exec-ptask/s4u-exec-ptask.tesh @@ -1,6 +1,6 @@ #!/usr/bin/env tesh -$ ${bindir:=.}/s4u-exec-ptask$EXEEXT ${platfdir}/energy_platform.xml --energy --cfg=host/model:ptask_L07 --cfg=tracing:yes --cfg=tracing/uncategorized:yes --log=instr_resource.t:debug --log=no_loc "--log=root.fmt:[%10.6r]%e(%i:%P@%h)%e%m%n" +$ ${bindir:=.}/s4u-exec-ptask$EXEEXT ${platfdir}/energy_platform.xml --cfg=host/model:ptask_L07 --cfg=tracing:yes --cfg=tracing/uncategorized:yes --log=instr_resource.t:debug --log=no_loc "--log=root.fmt:[%10.6r]%e(%i:%P@%h)%e%m%n" > [ 0.000000] (0:maestro@) Configuration change: Set 'host/model' to 'ptask_L07' > [ 0.000000] (0:maestro@) Configuration change: Set 'tracing' to 'yes' > [ 0.000000] (0:maestro@) Configuration change: Set 'tracing/uncategorized' to 'yes' @@ -10,21 +10,12 @@ $ ${bindir:=.}/s4u-exec-ptask$EXEEXT ${platfdir}/energy_platform.xml --energy -- > [300.000000] (0:maestro@) UNCAT HOST [0.000000 - 300.000000] MyHost2 speed_used 3333333.333333 > [300.000000] (0:maestro@) UNCAT HOST [0.000000 - 300.000000] MyHost3 speed_used 3333333.333333 > [300.000000] (0:maestro@) UNCAT LINK [0.000000 - 300.000000] bus bandwidth_used 100000.000000 -> [300.000000] (1:test@MyHost1) We can do the same with a timeout of one second enabled. -> [301.000000] (1:test@MyHost1) Then, build a parallel task involving only computations and no communication (1 Gflop per node) -> [311.000000] (0:maestro@) UNCAT HOST [301.000000 - 311.000000] MyHost1 speed_used 100000000.000000 -> [311.000000] (0:maestro@) UNCAT HOST [301.000000 - 311.000000] MyHost2 speed_used 100000000.000000 -> [311.000000] (0:maestro@) UNCAT HOST [301.000000 - 311.000000] MyHost3 speed_used 100000000.000000 -> [311.000000] (1:test@MyHost1) Then, build a parallel task involving only heterogeneous computations and no communication -> [326.000000] (0:maestro@) UNCAT HOST [311.000000 - 326.000000] MyHost1 speed_used 33333333.333333 -> [326.000000] (0:maestro@) UNCAT HOST [311.000000 - 326.000000] MyHost2 speed_used 66666666.666667 -> [326.000000] (0:maestro@) UNCAT HOST [311.000000 - 326.000000] MyHost3 speed_used 100000000.000000 -> [326.000000] (1:test@MyHost1) Then, build a parallel task with no computation nor communication (synchro only) -> [326.000000] (1:test@MyHost1) Finally, trick the ptask to do a 'remote execution', on host MyHost2 -> [336.000000] (0:maestro@) UNCAT HOST [326.000000 - 336.000000] MyHost2 speed_used 100000000.000000 -> [336.000000] (1:test@MyHost1) Goodbye now! -> [336.000000] (0:maestro@) Total energy consumption: 165494.222222 Joules (used hosts: 165494.222222 Joules; unused/idle hosts: 0.000000) -> [336.000000] (0:maestro@) Simulation done. -> [336.000000] (0:maestro@) Energy consumption of host MyHost1: 32094.222222 Joules -> [336.000000] (0:maestro@) Energy consumption of host MyHost2: 67200.000000 Joules -> [336.000000] (0:maestro@) Energy consumption of host MyHost3: 66200.000000 Joules +> [300.000000] (1:test@MyHost1) We can do the same with a timeout of 10 seconds enabled. +> [310.000000] (1:test@MyHost1) Caught the expected timeout exception. +> [310.000000] (1:test@MyHost1) Then, build a parallel task involving only computations (of different amounts) and no communication +> [320.000000] (0:maestro@) UNCAT HOST [310.000000 - 320.000000] MyHost1 speed_used 30000000.000000 +> [320.000000] (0:maestro@) UNCAT HOST [310.000000 - 320.000000] MyHost2 speed_used 60000000.000000 +> [320.000000] (0:maestro@) UNCAT HOST [310.000000 - 320.000000] MyHost3 speed_used 100000000.000000 +> [320.000000] (1:test@MyHost1) Then, build a parallel task with no computation nor communication (synchro only) +> [320.000000] (1:test@MyHost1) Goodbye now! +> [320.000000] (0:maestro@) Simulation done.