-/* Copyright (c) 2009-2019. The SimGrid Team. All rights reserved. */
+/* Copyright (c) 2009-2020. The SimGrid Team. All rights reserved. */
/* This program is free software; you can redistribute it and/or modify it
* under the terms of the license (GNU LGPL) which comes with this package. */
#include "cpu_cas01.hpp"
#include "simgrid/sg_config.hpp"
+#include "src/kernel/resource/profile/Event.hpp"
#include "src/surf/cpu_ti.hpp"
#include "src/surf/surf_interface.hpp"
#include "surf/surf.hpp"
xbt_assert(surf_cpu_model_vm == nullptr, "CPU model already initialized. This should not happen.");
if (cpu_optim_opt == "TI") {
- simgrid::surf::CpuTiModel::create_pm_vm_models();
+ simgrid::kernel::resource::CpuTiModel::create_pm_vm_models();
return;
}
else
algo = simgrid::kernel::resource::Model::UpdateAlgo::FULL;
- surf_cpu_model_pm = new simgrid::surf::CpuCas01Model(algo);
- surf_cpu_model_vm = new simgrid::surf::CpuCas01Model(algo);
+ surf_cpu_model_pm = new simgrid::kernel::resource::CpuCas01Model(algo);
+ surf_cpu_model_vm = new simgrid::kernel::resource::CpuCas01Model(algo);
}
namespace simgrid {
-namespace surf {
+namespace kernel {
+namespace resource {
-CpuCas01Model::CpuCas01Model(kernel::resource::Model::UpdateAlgo algo) : simgrid::surf::CpuModel(algo)
+CpuCas01Model::CpuCas01Model(Model::UpdateAlgo algo) : CpuModel(algo)
{
all_existing_models.push_back(this);
- bool select = simgrid::config::get_value<bool>("cpu/maxmin-selective-update");
+ bool select = config::get_value<bool>("cpu/maxmin-selective-update");
- if (algo == Model::UpdateAlgo::LAZY) {
- xbt_assert(select || simgrid::config::is_default("cpu/maxmin-selective-update"),
+ if (is_update_lazy()) {
+ xbt_assert(select || config::is_default("cpu/maxmin-selective-update"),
"You cannot disable cpu selective update when using the lazy update mechanism");
select = true;
}
- set_maxmin_system(new simgrid::kernel::lmm::System(select));
+ set_maxmin_system(new lmm::System(select));
}
CpuCas01Model::~CpuCas01Model()
surf_cpu_model_pm = nullptr;
}
-Cpu* CpuCas01Model::create_cpu(simgrid::s4u::Host* host, const std::vector<double>& speed_per_pstate, int core)
+Cpu* CpuCas01Model::create_cpu(s4u::Host* host, const std::vector<double>& speed_per_pstate, int core)
{
return new CpuCas01(this, host, speed_per_pstate, core);
}
/************
* Resource *
************/
-CpuCas01::CpuCas01(CpuCas01Model* model, simgrid::s4u::Host* host, const std::vector<double>& speed_per_pstate,
- int core)
+CpuCas01::CpuCas01(CpuCas01Model* model, s4u::Host* host, const std::vector<double>& speed_per_pstate, int core)
: Cpu(model, host, model->get_maxmin_system()->constraint_new(this, core * speed_per_pstate.front()),
speed_per_pstate, core)
{
/** @brief take into account changes of speed (either load or max) */
void CpuCas01::on_speed_change()
{
- kernel::lmm::Variable* var = nullptr;
- const kernel::lmm::Element* elem = nullptr;
+ const lmm::Variable* var;
+ const lmm::Element* elem = nullptr;
get_model()->get_maxmin_system()->update_constraint_bound(get_constraint(),
get_core_count() * speed_.scale * speed_.peak);
while ((var = get_constraint()->get_variable(&elem))) {
- CpuCas01Action* action = static_cast<CpuCas01Action*>(var->get_id());
+ const CpuCas01Action* action = static_cast<CpuCas01Action*>(var->get_id());
get_model()->get_maxmin_system()->update_variable_bound(action->get_variable(),
action->requested_core() * speed_.scale * speed_.peak);
Cpu::on_speed_change();
}
-void CpuCas01::apply_event(kernel::profile::Event* event, double value)
+void CpuCas01::apply_event(profile::Event* event, double value)
{
if (event == speed_.event) {
/* TODO (Hypervisor): do the same thing for constraint_core[i] */
if (value > 0) {
if (not is_on()) {
- XBT_VERB("Restart processes on host %s", get_host()->get_cname());
+ XBT_VERB("Restart actors on host %s", get_host()->get_cname());
get_host()->turn_on();
}
} else {
- kernel::lmm::Constraint* cnst = get_constraint();
- kernel::lmm::Variable* var = nullptr;
- const kernel::lmm::Element* elem = nullptr;
+ const lmm::Constraint* cnst = get_constraint();
+ const lmm::Variable* var;
+ const lmm::Element* elem = nullptr;
double date = surf_get_clock();
get_host()->turn_off();
while ((var = cnst->get_variable(&elem))) {
- kernel::resource::Action* action = static_cast<kernel::resource::Action*>(var->get_id());
+ Action* action = var->get_id();
- if (action->get_state() == kernel::resource::Action::State::INITED ||
- action->get_state() == kernel::resource::Action::State::STARTED ||
- action->get_state() == kernel::resource::Action::State::IGNORED) {
+ if (action->get_state() == Action::State::INITED || action->get_state() == Action::State::STARTED ||
+ action->get_state() == Action::State::IGNORED) {
action->set_finish_time(date);
- action->set_state(kernel::resource::Action::State::FAILED);
+ action->set_state(Action::State::FAILED);
}
}
}
duration = std::max(duration, sg_surf_precision);
XBT_IN("(%s,%g)", get_cname(), duration);
- CpuCas01Action* action =
- new CpuCas01Action(get_model(), 1.0, not is_on(), speed_.scale * speed_.peak, get_constraint());
+ auto* action = new CpuCas01Action(get_model(), 1.0, not is_on(), speed_.scale * speed_.peak, get_constraint());
// FIXME: sleep variables should not consume 1.0 in System::expand()
action->set_max_duration(duration);
- action->set_suspend_state(kernel::resource::Action::SuspendStates::SLEEPING);
+ action->set_suspend_state(Action::SuspendStates::SLEEPING);
if (duration == NO_MAX_DURATION)
- action->set_state(simgrid::kernel::resource::Action::State::IGNORED);
+ action->set_state(Action::State::IGNORED);
- get_model()->get_maxmin_system()->update_variable_weight(action->get_variable(), 0.0);
- if (get_model()->get_update_algorithm() == kernel::resource::Model::UpdateAlgo::LAZY) { // remove action from the heap
+ get_model()->get_maxmin_system()->update_variable_penalty(action->get_variable(), 0.0);
+ if (get_model()->is_update_lazy()) { // remove action from the heap
get_model()->get_action_heap().remove(action);
// this is necessary for a variable with weight 0 since such variables are ignored in lmm and we need to set its
// max_duration correctly at the next call to share_resources
/**********
* Action *
**********/
-CpuCas01Action::CpuCas01Action(kernel::resource::Model* model, double cost, bool failed, double speed,
- kernel::lmm::Constraint* constraint, int requested_core)
+CpuCas01Action::CpuCas01Action(Model* model, double cost, bool failed, double speed, lmm::Constraint* constraint,
+ int requested_core)
: CpuAction(model, cost, failed,
model->get_maxmin_system()->variable_new(this, 1.0 / requested_core, requested_core * speed, 1))
, requested_core_(requested_core)
{
- if (model->get_update_algorithm() == kernel::resource::Model::UpdateAlgo::LAZY)
+ if (model->is_update_lazy())
set_last_update();
model->get_maxmin_system()->expand(constraint, get_variable(), 1.0);
}
-CpuCas01Action::CpuCas01Action(kernel::resource::Model* model, double cost, bool failed, double speed,
- kernel::lmm::Constraint* constraint)
+CpuCas01Action::CpuCas01Action(Model* model, double cost, bool failed, double speed, lmm::Constraint* constraint)
: CpuCas01Action(model, cost, failed, speed, constraint, /* requested_core */ 1)
{
}
-int CpuCas01Action::requested_core()
+int CpuCas01Action::requested_core() const
{
return requested_core_;
}
CpuCas01Action::~CpuCas01Action()=default;
-}
-}
+} // namespace resource
+} // namespace kernel
+} // namespace simgrid