-/* Copyright (c) 2004-2017. The SimGrid Team. All rights reserved. */
+/* Copyright (c) 2004-2019. 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. */
-/* \file callbacks.h */
-
#include "src/kernel/lmm/maxmin.hpp"
+#include "src/surf/surf_interface.hpp"
#include "xbt/backtrace.hpp"
-#include "xbt/log.h"
-#include "xbt/mallocator.h"
-#include "xbt/sysdep.h"
-#include "xbt/utility.hpp"
-#include <algorithm>
-#include <cmath>
-#include <cstdlib>
-#include <cxxabi.h>
-#include <limits>
-#include <vector>
XBT_LOG_NEW_DEFAULT_SUBCATEGORY(surf_maxmin, surf, "Logging specific to SURF (maxmin)");
typedef std::vector<int> dyn_light_t;
-int Variable::Global_debug_id = 1;
-int Constraint::Global_debug_id = 1;
+int Variable::next_rank_ = 1;
+int Constraint::next_rank_ = 1;
+
+System* make_new_maxmin_system(bool selective_update)
+{
+ return new System(selective_update);
+}
-int s_lmm_element_t::get_concurrency() const
+int Element::get_concurrency() const
{
// Ignore element with weight less than one (e.g. cross-traffic)
return (consumption_weight >= 1) ? 1 : 0;
// return (int)ceil(elem->weight);//Include element as the rounded-up integer value of the element weight
}
-void s_lmm_element_t::decrease_concurrency()
+void Element::decrease_concurrency()
{
- xbt_assert(constraint->concurrency_current >= get_concurrency());
- constraint->concurrency_current -= get_concurrency();
+ xbt_assert(constraint->concurrency_current_ >= get_concurrency());
+ constraint->concurrency_current_ -= get_concurrency();
}
-void s_lmm_element_t::increase_concurrency()
+void Element::increase_concurrency()
{
- constraint->concurrency_current += get_concurrency();
+ constraint->concurrency_current_ += get_concurrency();
- if (constraint->concurrency_current > constraint->concurrency_maximum)
- constraint->concurrency_maximum = constraint->concurrency_current;
+ if (constraint->concurrency_current_ > constraint->concurrency_maximum_)
+ constraint->concurrency_maximum_ = constraint->concurrency_current_;
xbt_assert(constraint->get_concurrency_limit() < 0 ||
- constraint->concurrency_current <= constraint->get_concurrency_limit(),
+ constraint->concurrency_current_ <= constraint->get_concurrency_limit(),
"Concurrency limit overflow!");
}
-void s_lmm_system_t::check_concurrency() const
+void System::check_concurrency() const
{
// These checks are very expensive, so do them only if we want to debug SURF LMM
if (not XBT_LOG_ISENABLED(surf_maxmin, xbt_log_priority_debug))
for (Constraint const& cnst : constraint_set) {
int concurrency = 0;
- for (s_lmm_element_t const& elem : cnst.enabled_element_set) {
- xbt_assert(elem.variable->sharing_weight > 0);
+ for (Element const& elem : cnst.enabled_element_set_) {
+ xbt_assert(elem.variable->sharing_penalty_ > 0);
concurrency += elem.get_concurrency();
}
- for (s_lmm_element_t const& elem : cnst.disabled_element_set) {
+ for (Element const& elem : cnst.disabled_element_set_) {
// We should have staged variables only if concurrency is reached in some constraint
- xbt_assert(cnst.get_concurrency_limit() < 0 || elem.variable->staged_weight == 0 ||
- elem.variable->get_min_concurrency_slack() < elem.variable->concurrency_share,
+ xbt_assert(cnst.get_concurrency_limit() < 0 || elem.variable->staged_penalty_ == 0 ||
+ elem.variable->get_min_concurrency_slack() < elem.variable->concurrency_share_,
"should not have staged variable!");
}
xbt_assert(cnst.get_concurrency_limit() < 0 || cnst.get_concurrency_limit() >= concurrency,
"concurrency check failed!");
- xbt_assert(cnst.concurrency_current == concurrency, "concurrency_current is out-of-date!");
+ xbt_assert(cnst.concurrency_current_ == concurrency, "concurrency_current is out-of-date!");
}
// Check that for each variable, all corresponding elements are in the same state (i.e. same element sets)
for (Variable const& var : variable_set) {
- if (var.cnsts.empty())
+ if (var.cnsts_.empty())
continue;
- const s_lmm_element_t& elem = var.cnsts[0];
- int belong_to_enabled = elem.enabled_element_set_hook.is_linked();
- int belong_to_disabled = elem.disabled_element_set_hook.is_linked();
- int belong_to_active = elem.active_element_set_hook.is_linked();
+ const Element& elem = var.cnsts_[0];
+ int belong_to_enabled = elem.enabled_element_set_hook.is_linked();
+ int belong_to_disabled = elem.disabled_element_set_hook.is_linked();
+ int belong_to_active = elem.active_element_set_hook.is_linked();
- for (s_lmm_element_t const& elem2 : var.cnsts) {
+ for (Element const& elem2 : var.cnsts_) {
xbt_assert(belong_to_enabled == elem2.enabled_element_set_hook.is_linked(),
"Variable inconsistency (1): enabled_element_set");
xbt_assert(belong_to_disabled == elem2.disabled_element_set_hook.is_linked(),
}
}
-void s_lmm_system_t::var_free(lmm_variable_t var)
+void System::var_free(Variable* var)
{
XBT_IN("(sys=%p, var=%p)", this, var);
- modified = true;
+ modified_ = true;
// TODOLATER Can do better than that by leaving only the variable in only one enabled_element_set, call
// update_modified_set, and then remove it..
- if (not var->cnsts.empty())
- update_modified_set(var->cnsts[0].constraint);
+ if (not var->cnsts_.empty())
+ update_modified_set(var->cnsts_[0].constraint);
- for (s_lmm_element_t& elem : var->cnsts) {
- if (var->sharing_weight > 0)
+ for (Element& elem : var->cnsts_) {
+ if (var->sharing_penalty_ > 0)
elem.decrease_concurrency();
if (elem.enabled_element_set_hook.is_linked())
- simgrid::xbt::intrusive_erase(elem.constraint->enabled_element_set, elem);
+ simgrid::xbt::intrusive_erase(elem.constraint->enabled_element_set_, elem);
if (elem.disabled_element_set_hook.is_linked())
- simgrid::xbt::intrusive_erase(elem.constraint->disabled_element_set, elem);
+ simgrid::xbt::intrusive_erase(elem.constraint->disabled_element_set_, elem);
if (elem.active_element_set_hook.is_linked())
- simgrid::xbt::intrusive_erase(elem.constraint->active_element_set, elem);
- int nelements = elem.constraint->enabled_element_set.size() + elem.constraint->disabled_element_set.size();
+ simgrid::xbt::intrusive_erase(elem.constraint->active_element_set_, elem);
+ int nelements = elem.constraint->enabled_element_set_.size() + elem.constraint->disabled_element_set_.size();
if (nelements == 0)
make_constraint_inactive(elem.constraint);
else
on_disabled_var(elem.constraint);
}
- var->cnsts.clear();
+ var->cnsts_.clear();
check_concurrency();
- xbt_mallocator_release(variable_mallocator, var);
+ xbt_mallocator_release(variable_mallocator_, var);
XBT_OUT();
}
-s_lmm_system_t::s_lmm_system_t(bool selective_update) : selective_update_active(selective_update)
+System::System(bool selective_update) : selective_update_active(selective_update)
{
- modified = false;
- visited_counter = 1;
-
XBT_DEBUG("Setting selective_update_active flag to %d", selective_update_active);
- keep_track = nullptr;
- variable_mallocator = xbt_mallocator_new(65536, s_lmm_system_t::variable_mallocator_new_f,
- s_lmm_system_t::variable_mallocator_free_f, nullptr);
- solve_fun = &lmm_solve;
+ if (selective_update)
+ modified_set_ = new kernel::resource::Action::ModifiedSet();
}
-s_lmm_system_t::~s_lmm_system_t()
+System::~System()
{
- lmm_variable_t var;
- lmm_constraint_t cnst;
+ Variable* var;
+ Constraint* cnst;
while ((var = extract_variable())) {
- auto demangled = simgrid::xbt::demangle(typeid(*var->id).name());
+ auto demangled = simgrid::xbt::demangle(var->id_ ? typeid(*var->id_).name() : "(unidentified)");
XBT_WARN("Probable bug: a %s variable (#%d) not removed before the LMM system destruction.", demangled.get(),
- var->id_int);
+ var->rank_);
var_free(var);
}
while ((cnst = extract_constraint()))
cnst_free(cnst);
- xbt_mallocator_free(variable_mallocator);
+ xbt_mallocator_free(variable_mallocator_);
+ delete modified_set_;
}
-void s_lmm_system_t::cnst_free(lmm_constraint_t cnst)
+void System::cnst_free(Constraint* cnst)
{
make_constraint_inactive(cnst);
delete cnst;
}
-Constraint::Constraint(void* id_value, double bound_value) : bound(bound_value), id(id_value)
+Constraint::Constraint(resource::Resource* id_value, double bound_value) : bound_(bound_value), id_(id_value)
{
- id_int = Global_debug_id++;
+ rank_ = next_rank_++;
- remaining = 0.0;
- usage = 0.0;
- concurrency_limit = sg_concurrency_limit;
- concurrency_current = 0;
- concurrency_maximum = 0;
- sharing_policy = 1; /* FIXME: don't hardcode the value */
+ remaining_ = 0.0;
+ usage_ = 0.0;
+ concurrency_limit_ = sg_concurrency_limit;
+ concurrency_current_ = 0;
+ concurrency_maximum_ = 0;
+ sharing_policy_ = s4u::Link::SharingPolicy::SHARED;
- lambda = 0.0;
- new_lambda = 0.0;
- cnst_light = nullptr;
+ lambda_ = 0.0;
+ new_lambda_ = 0.0;
+ cnst_light_ = nullptr;
}
-lmm_constraint_t s_lmm_system_t::constraint_new(void* id, double bound_value)
+Constraint* System::constraint_new(resource::Resource* id, double bound_value)
{
- lmm_constraint_t cnst = new Constraint(id, bound_value);
+ Constraint* cnst = new Constraint(id, bound_value);
insert_constraint(cnst);
return cnst;
}
-void* s_lmm_system_t::variable_mallocator_new_f()
+void* System::variable_mallocator_new_f()
{
return new Variable;
}
-void s_lmm_system_t::variable_mallocator_free_f(void* var)
+void System::variable_mallocator_free_f(void* var)
{
- delete static_cast<lmm_variable_t>(var);
+ delete static_cast<Variable*>(var);
}
-lmm_variable_t s_lmm_system_t::variable_new(simgrid::surf::Action* id, double sharing_weight, double bound,
- int number_of_constraints)
+Variable* System::variable_new(resource::Action* id, double sharing_penalty, double bound, size_t number_of_constraints)
{
- XBT_IN("(sys=%p, id=%p, weight=%f, bound=%f, num_cons =%d)", this, id, sharing_weight, bound, number_of_constraints);
+ XBT_IN("(sys=%p, id=%p, penalty=%f, bound=%f, num_cons =%zu)", this, id, sharing_penalty, bound,
+ number_of_constraints);
- lmm_variable_t var = static_cast<lmm_variable_t>(xbt_mallocator_get(variable_mallocator));
- var->initialize(id, sharing_weight, bound, number_of_constraints, visited_counter - 1);
- if (sharing_weight)
+ Variable* var = static_cast<Variable*>(xbt_mallocator_get(variable_mallocator_));
+ var->initialize(id, sharing_penalty, bound, number_of_constraints, visited_counter_ - 1);
+ if (sharing_penalty > 0)
variable_set.push_front(*var);
else
variable_set.push_back(*var);
return var;
}
-void s_lmm_system_t::variable_free(lmm_variable_t var)
+void System::variable_free(Variable* var)
{
remove_variable(var);
var_free(var);
}
-void s_lmm_system_t::expand(lmm_constraint_t cnst, lmm_variable_t var, double consumption_weight)
+void System::expand(Constraint* cnst, Variable* var, double consumption_weight)
{
- modified = true;
+ modified_ = true;
// Check if this variable already has an active element in this constraint
// If it does, substract it from the required slack
int current_share = 0;
- if (var->concurrency_share > 1) {
- for (s_lmm_element_t& elem : var->cnsts) {
+ if (var->concurrency_share_ > 1) {
+ for (Element& elem : var->cnsts_) {
if (elem.constraint == cnst && elem.enabled_element_set_hook.is_linked())
current_share += elem.get_concurrency();
}
}
// Check if we need to disable the variable
- if (var->sharing_weight > 0 && var->concurrency_share - current_share > cnst->get_concurrency_slack()) {
- double weight = var->sharing_weight;
+ if (var->sharing_penalty_ > 0 && var->concurrency_share_ - current_share > cnst->get_concurrency_slack()) {
+ double penalty = var->sharing_penalty_;
disable_var(var);
- for (s_lmm_element_t const& elem : var->cnsts)
+ for (Element const& elem : var->cnsts_)
on_disabled_var(elem.constraint);
consumption_weight = 0;
- var->staged_weight = weight;
- xbt_assert(not var->sharing_weight);
+ var->staged_penalty_ = penalty;
+ xbt_assert(not var->sharing_penalty_);
}
- xbt_assert(var->cnsts.size() < var->cnsts.capacity(), "Too much constraints");
+ xbt_assert(var->cnsts_.size() < var->cnsts_.capacity(), "Too much constraints");
- var->cnsts.resize(var->cnsts.size() + 1);
- s_lmm_element_t& elem = var->cnsts.back();
+ var->cnsts_.resize(var->cnsts_.size() + 1);
+ Element& elem = var->cnsts_.back();
elem.consumption_weight = consumption_weight;
elem.constraint = cnst;
elem.variable = var;
- if (var->sharing_weight) {
- elem.constraint->enabled_element_set.push_front(elem);
+ if (var->sharing_penalty_) {
+ elem.constraint->enabled_element_set_.push_front(elem);
elem.increase_concurrency();
} else
- elem.constraint->disabled_element_set.push_back(elem);
+ elem.constraint->disabled_element_set_.push_back(elem);
if (not selective_update_active) {
make_constraint_active(cnst);
- } else if (elem.consumption_weight > 0 || var->sharing_weight > 0) {
+ } else if (elem.consumption_weight > 0 || var->sharing_penalty_ > 0) {
make_constraint_active(cnst);
update_modified_set(cnst);
// TODOLATER: Why do we need this second call?
- if (var->cnsts.size() > 1)
- update_modified_set(var->cnsts[0].constraint);
+ if (var->cnsts_.size() > 1)
+ update_modified_set(var->cnsts_[0].constraint);
}
check_concurrency();
}
-void s_lmm_system_t::expand_add(lmm_constraint_t cnst, lmm_variable_t var, double value)
+void System::expand_add(Constraint* cnst, Variable* var, double value)
{
- modified = true;
+ modified_ = true;
check_concurrency();
// BEWARE: In case you have multiple elements in one constraint, this will always add value to the first element.
- auto elem_it = std::find_if(begin(var->cnsts), end(var->cnsts),
- [&cnst](s_lmm_element_t const& x) { return x.constraint == cnst; });
- if (elem_it != end(var->cnsts)) {
- s_lmm_element_t& elem = *elem_it;
- if (var->sharing_weight)
+ auto elem_it =
+ std::find_if(begin(var->cnsts_), end(var->cnsts_), [&cnst](Element const& x) { return x.constraint == cnst; });
+ if (elem_it != end(var->cnsts_)) {
+ Element& elem = *elem_it;
+ if (var->sharing_penalty_)
elem.decrease_concurrency();
- if (cnst->sharing_policy)
+ if (cnst->sharing_policy_ != s4u::Link::SharingPolicy::FATPIPE)
elem.consumption_weight += value;
else
elem.consumption_weight = std::max(elem.consumption_weight, value);
// We need to check that increasing value of the element does not cross the concurrency limit
- if (var->sharing_weight) {
+ if (var->sharing_penalty_) {
if (cnst->get_concurrency_slack() < elem.get_concurrency()) {
- double weight = var->sharing_weight;
+ double penalty = var->sharing_penalty_;
disable_var(var);
- for (s_lmm_element_t const& elem2 : var->cnsts)
+ for (Element const& elem2 : var->cnsts_)
on_disabled_var(elem2.constraint);
- var->staged_weight = weight;
- xbt_assert(not var->sharing_weight);
+ var->staged_penalty_ = penalty;
+ xbt_assert(not var->sharing_penalty_);
}
elem.increase_concurrency();
}
check_concurrency();
}
-lmm_variable_t Constraint::get_variable(const_lmm_element_t* elem) const
+Variable* Constraint::get_variable(const Element** elem) const
{
if (*elem == nullptr) {
// That is the first call, pick the first element among enabled_element_set (or disabled_element_set if
// enabled_element_set is empty)
- if (not enabled_element_set.empty())
- *elem = &enabled_element_set.front();
- else if (not disabled_element_set.empty())
- *elem = &disabled_element_set.front();
+ if (not enabled_element_set_.empty())
+ *elem = &enabled_element_set_.front();
+ else if (not disabled_element_set_.empty())
+ *elem = &disabled_element_set_.front();
else
*elem = nullptr;
} else {
// elem is not null, so we carry on
if ((*elem)->enabled_element_set_hook.is_linked()) {
// Look at enabled_element_set, and jump to disabled_element_set when finished
- auto iter = std::next(enabled_element_set.iterator_to(**elem));
- if (iter != std::end(enabled_element_set))
+ auto iter = std::next(enabled_element_set_.iterator_to(**elem));
+ if (iter != std::end(enabled_element_set_))
*elem = &*iter;
- else if (not disabled_element_set.empty())
- *elem = &disabled_element_set.front();
+ else if (not disabled_element_set_.empty())
+ *elem = &disabled_element_set_.front();
else
*elem = nullptr;
} else {
- auto iter = std::next(disabled_element_set.iterator_to(**elem));
- *elem = iter != std::end(disabled_element_set) ? &*iter : nullptr;
+ auto iter = std::next(disabled_element_set_.iterator_to(**elem));
+ *elem = iter != std::end(disabled_element_set_) ? &*iter : nullptr;
}
}
if (*elem)
// if we modify the list between calls, normal version may loop forever
// this safe version ensures that we browse the list elements only once
-lmm_variable_t Constraint::get_variable_safe(const_lmm_element_t* elem, const_lmm_element_t* nextelem,
- int* numelem) const
+Variable* Constraint::get_variable_safe(const Element** elem, const Element** nextelem, int* numelem) const
{
if (*elem == nullptr) {
- *numelem = enabled_element_set.size() + disabled_element_set.size() - 1;
- if (not enabled_element_set.empty())
- *elem = &enabled_element_set.front();
- else if (not disabled_element_set.empty())
- *elem = &disabled_element_set.front();
+ *numelem = enabled_element_set_.size() + disabled_element_set_.size() - 1;
+ if (not enabled_element_set_.empty())
+ *elem = &enabled_element_set_.front();
+ else if (not disabled_element_set_.empty())
+ *elem = &disabled_element_set_.front();
else
*elem = nullptr;
} else {
// elem is not null, so we carry on
if ((*elem)->enabled_element_set_hook.is_linked()) {
// Look at enabled_element_set, and jump to disabled_element_set when finished
- auto iter = std::next(enabled_element_set.iterator_to(**elem));
- if (iter != std::end(enabled_element_set))
+ auto iter = std::next(enabled_element_set_.iterator_to(**elem));
+ if (iter != std::end(enabled_element_set_))
*nextelem = &*iter;
- else if (not disabled_element_set.empty())
- *nextelem = &disabled_element_set.front();
+ else if (not disabled_element_set_.empty())
+ *nextelem = &disabled_element_set_.front();
else
*nextelem = nullptr;
} else {
- auto iter = std::next(disabled_element_set.iterator_to(**elem));
- *nextelem = iter != std::end(disabled_element_set) ? &*iter : nullptr;
+ auto iter = std::next(disabled_element_set_.iterator_to(**elem));
+ *nextelem = iter != std::end(disabled_element_set_) ? &*iter : nullptr;
}
return (*elem)->variable;
} else
}
static inline void saturated_variable_set_update(ConstraintLight* cnst_light_tab,
- const dyn_light_t& saturated_constraints, lmm_system_t sys)
+ const dyn_light_t& saturated_constraints, System* sys)
{
/* Add active variables (i.e. variables that need to be set) from the set of constraints to saturate
* (cnst_light_tab)*/
for (int const& saturated_cnst : saturated_constraints) {
ConstraintLight& cnst = cnst_light_tab[saturated_cnst];
- for (s_lmm_element_t const& elem : cnst.cnst->active_element_set) {
- // Visiting active_element_set, so, by construction, should never get a zero weight, correct?
- xbt_assert(elem.variable->sharing_weight > 0);
- if (elem.consumption_weight > 0 && not elem.variable->saturated_variable_set_hook.is_linked())
+ for (Element const& elem : cnst.cnst->active_element_set_) {
+ xbt_assert(elem.variable->sharing_penalty_ > 0); // All elements of active_element_set should be active
+ if (elem.consumption_weight > 0 && not elem.variable->saturated_variable_set_hook_.is_linked())
sys->saturated_variable_set.push_back(*elem.variable);
}
}
}
template <class ElemList>
-static void format_lmm_element_list(const ElemList& elem_list, int sharing_policy, double& sum, std::string& buf)
-{
- for (s_lmm_element_t const& elem : elem_list) {
- buf += std::to_string(elem.consumption_weight) + ".'" + std::to_string(elem.variable->id_int) + "'(" +
- std::to_string(elem.variable->value) + ")" + (sharing_policy ? " + " : " , ");
- if (sharing_policy)
- sum += elem.consumption_weight * elem.variable->value;
+static void format_element_list(const ElemList& elem_list, s4u::Link::SharingPolicy sharing_policy, double& sum,
+ std::string& buf)
+{
+ for (Element const& elem : elem_list) {
+ buf += std::to_string(elem.consumption_weight) + ".'" + std::to_string(elem.variable->rank_) + "'(" +
+ std::to_string(elem.variable->value_) + ")" +
+ (sharing_policy != s4u::Link::SharingPolicy::FATPIPE ? " + " : " , ");
+ if (sharing_policy != s4u::Link::SharingPolicy::FATPIPE)
+ sum += elem.consumption_weight * elem.variable->value_;
else
- sum = std::max(sum, elem.consumption_weight * elem.variable->value);
+ sum = std::max(sum, elem.consumption_weight * elem.variable->value_);
}
}
-void s_lmm_system_t::print() const
+void System::print() const
{
std::string buf = std::string("MAX-MIN ( ");
/* Printing Objective */
for (Variable const& var : variable_set)
- buf += "'" + std::to_string(var.id_int) + "'(" + std::to_string(var.sharing_weight) + ") ";
+ buf += "'" + std::to_string(var.rank_) + "'(" + std::to_string(var.sharing_penalty_) + ") ";
buf += ")";
XBT_DEBUG("%20s", buf.c_str());
buf.clear();
double sum = 0.0;
// Show the enabled variables
buf += "\t";
- buf += cnst.sharing_policy ? "(" : "max(";
- format_lmm_element_list(cnst.enabled_element_set, cnst.sharing_policy, sum, buf);
+ buf += cnst.sharing_policy_ != s4u::Link::SharingPolicy::FATPIPE ? "(" : "max(";
+ format_element_list(cnst.enabled_element_set_, cnst.sharing_policy_, sum, buf);
// TODO: Adding disabled elements only for test compatibility, but do we really want them to be printed?
- format_lmm_element_list(cnst.disabled_element_set, cnst.sharing_policy, sum, buf);
+ format_element_list(cnst.disabled_element_set_, cnst.sharing_policy_, sum, buf);
- buf += "0) <= " + std::to_string(cnst.bound) + " ('" + std::to_string(cnst.id_int) + "')";
+ buf += "0) <= " + std::to_string(cnst.bound_) + " ('" + std::to_string(cnst.rank_) + "')";
- if (not cnst.sharing_policy) {
+ if (cnst.sharing_policy_ == s4u::Link::SharingPolicy::FATPIPE) {
buf += " [MAX-Constraint]";
}
XBT_DEBUG("%s", buf.c_str());
buf.clear();
- xbt_assert(not double_positive(sum - cnst.bound, cnst.bound * sg_maxmin_precision),
- "Incorrect value (%f is not smaller than %f): %g", sum, cnst.bound, sum - cnst.bound);
+ xbt_assert(not double_positive(sum - cnst.bound_, cnst.bound_ * sg_maxmin_precision),
+ "Incorrect value (%f is not smaller than %f): %g", sum, cnst.bound_, sum - cnst.bound_);
}
XBT_DEBUG("Variables");
/* Printing Result */
for (Variable const& var : variable_set) {
- if (var.bound > 0) {
- XBT_DEBUG("'%d'(%f) : %f (<=%f)", var.id_int, var.sharing_weight, var.value, var.bound);
- xbt_assert(not double_positive(var.value - var.bound, var.bound * sg_maxmin_precision),
- "Incorrect value (%f is not smaller than %f", var.value, var.bound);
+ if (var.bound_ > 0) {
+ XBT_DEBUG("'%d'(%f) : %f (<=%f)", var.rank_, var.sharing_penalty_, var.value_, var.bound_);
+ xbt_assert(not double_positive(var.value_ - var.bound_, var.bound_ * sg_maxmin_precision),
+ "Incorrect value (%f is not smaller than %f", var.value_, var.bound_);
} else {
- XBT_DEBUG("'%d'(%f) : %f", var.id_int, var.sharing_weight, var.value);
+ XBT_DEBUG("'%d'(%f) : %f", var.rank_, var.sharing_penalty_, var.value_);
}
}
}
-void s_lmm_system_t::solve()
+void System::lmm_solve()
{
- if (modified) {
+ if (modified_) {
XBT_IN("(sys=%p)", this);
/* Compute Usage and store the variables that reach the maximum. If selective_update_active is true, only
* constraints that changed are considered. Otherwise all constraints with active actions are considered.
*/
if (selective_update_active)
- solve(modified_constraint_set);
+ lmm_solve(modified_constraint_set);
else
- solve(active_constraint_set);
+ lmm_solve(active_constraint_set);
XBT_OUT();
}
}
-template <class CnstList> void s_lmm_system_t::solve(CnstList& cnst_list)
+template <class CnstList> void System::lmm_solve(CnstList& cnst_list)
{
double min_usage = -1;
double min_bound = -1;
XBT_DEBUG("Active constraints : %zu", cnst_list.size());
/* Init: Only modified code portions: reset the value of active variables */
for (Constraint const& cnst : cnst_list) {
- for (s_lmm_element_t const& elem : cnst.enabled_element_set) {
- xbt_assert(elem.variable->sharing_weight > 0.0);
- elem.variable->value = 0.0;
+ for (Element const& elem : cnst.enabled_element_set_) {
+ xbt_assert(elem.variable->sharing_penalty_ > 0.0);
+ elem.variable->value_ = 0.0;
}
}
for (Constraint& cnst : cnst_list) {
/* INIT: Collect constraints that actually need to be saturated (i.e remaining and usage are strictly positive)
* into cnst_light_tab. */
- cnst.remaining = cnst.bound;
- if (not double_positive(cnst.remaining, cnst.bound * sg_maxmin_precision))
+ cnst.remaining_ = cnst.bound_;
+ if (not double_positive(cnst.remaining_, cnst.bound_ * sg_maxmin_precision))
continue;
- cnst.usage = 0;
- for (s_lmm_element_t& elem : cnst.enabled_element_set) {
- xbt_assert(elem.variable->sharing_weight > 0);
+ cnst.usage_ = 0;
+ for (Element& elem : cnst.enabled_element_set_) {
+ xbt_assert(elem.variable->sharing_penalty_ > 0);
if (elem.consumption_weight > 0) {
- if (cnst.sharing_policy)
- cnst.usage += elem.consumption_weight / elem.variable->sharing_weight;
- else if (cnst.usage < elem.consumption_weight / elem.variable->sharing_weight)
- cnst.usage = elem.consumption_weight / elem.variable->sharing_weight;
+ if (cnst.sharing_policy_ != s4u::Link::SharingPolicy::FATPIPE)
+ cnst.usage_ += elem.consumption_weight / elem.variable->sharing_penalty_;
+ else if (cnst.usage_ < elem.consumption_weight / elem.variable->sharing_penalty_)
+ cnst.usage_ = elem.consumption_weight / elem.variable->sharing_penalty_;
elem.make_active();
- simgrid::surf::Action* action = static_cast<simgrid::surf::Action*>(elem.variable->id);
- if (keep_track && not action->is_linked())
- keep_track->push_back(*action);
+ resource::Action* action = static_cast<resource::Action*>(elem.variable->id_);
+ if (modified_set_ && not action->is_within_modified_set())
+ modified_set_->push_back(*action);
}
}
- XBT_DEBUG("Constraint '%d' usage: %f remaining: %f concurrency: %i<=%i<=%i", cnst.id_int, cnst.usage,
- cnst.remaining, cnst.concurrency_current, cnst.concurrency_maximum, cnst.get_concurrency_limit());
+ XBT_DEBUG("Constraint '%d' usage: %f remaining: %f concurrency: %i<=%i<=%i", cnst.rank_, cnst.usage_,
+ cnst.remaining_, cnst.concurrency_current_, cnst.concurrency_maximum_, cnst.get_concurrency_limit());
/* Saturated constraints update */
- if (cnst.usage > 0) {
+ if (cnst.usage_ > 0) {
cnst_light_tab[cnst_light_num].cnst = &cnst;
- cnst.cnst_light = &cnst_light_tab[cnst_light_num];
- cnst_light_tab[cnst_light_num].remaining_over_usage = cnst.remaining / cnst.usage;
+ cnst.cnst_light_ = &cnst_light_tab[cnst_light_num];
+ cnst_light_tab[cnst_light_num].remaining_over_usage = cnst.remaining_ / cnst.usage_;
saturated_constraints_update(cnst_light_tab[cnst_light_num].remaining_over_usage, cnst_light_num,
saturated_constraints, &min_usage);
- xbt_assert(not cnst.active_element_set.empty(),
+ xbt_assert(not cnst.active_element_set_.empty(),
"There is no sense adding a constraint that has no active element!");
cnst_light_num++;
}
/* Fix the variables that have to be */
auto& var_list = saturated_variable_set;
for (Variable const& var : var_list) {
- if (var.sharing_weight <= 0.0)
+ if (var.sharing_penalty_ <= 0.0)
DIE_IMPOSSIBLE;
/* First check if some of these variables could reach their upper bound and update min_bound accordingly. */
- XBT_DEBUG("var=%d, var.bound=%f, var.weight=%f, min_usage=%f, var.bound*var.weight=%f", var.id_int, var.bound,
- var.sharing_weight, min_usage, var.bound * var.sharing_weight);
- if ((var.bound > 0) && (var.bound * var.sharing_weight < min_usage)) {
+ XBT_DEBUG("var=%d, var.bound=%f, var.penalty=%f, min_usage=%f, var.bound*var.penalty=%f", var.rank_, var.bound_,
+ var.sharing_penalty_, min_usage, var.bound_ * var.sharing_penalty_);
+ if ((var.bound_ > 0) && (var.bound_ * var.sharing_penalty_ < min_usage)) {
if (min_bound < 0)
- min_bound = var.bound * var.sharing_weight;
+ min_bound = var.bound_ * var.sharing_penalty_;
else
- min_bound = std::min(min_bound, (var.bound * var.sharing_weight));
+ min_bound = std::min(min_bound, (var.bound_ * var.sharing_penalty_));
XBT_DEBUG("Updated min_bound=%f", min_bound);
}
}
if (min_bound < 0) {
// If no variable could reach its bound, deal iteratively the constraints usage ( at worst one constraint is
// saturated at each cycle)
- var.value = min_usage / var.sharing_weight;
- XBT_DEBUG("Setting var (%d) value to %f\n", var.id_int, var.value);
+ var.value_ = min_usage / var.sharing_penalty_;
+ XBT_DEBUG("Setting var (%d) value to %f\n", var.rank_, var.value_);
} else {
// If there exist a variable that can reach its bound, only update it (and other with the same bound) for now.
- if (double_equals(min_bound, var.bound * var.sharing_weight, sg_maxmin_precision)) {
- var.value = var.bound;
- XBT_DEBUG("Setting %p (%d) value to %f\n", &var, var.id_int, var.value);
+ if (double_equals(min_bound, var.bound_ * var.sharing_penalty_, sg_maxmin_precision)) {
+ var.value_ = var.bound_;
+ XBT_DEBUG("Setting %p (%d) value to %f\n", &var, var.rank_, var.value_);
} else {
// Variables which bound is different are not considered for this cycle, but they will be afterwards.
- XBT_DEBUG("Do not consider %p (%d) \n", &var, var.id_int);
+ XBT_DEBUG("Do not consider %p (%d) \n", &var, var.rank_);
var_list.pop_front();
continue;
}
}
- XBT_DEBUG("Min usage: %f, Var(%d).weight: %f, Var(%d).value: %f ", min_usage, var.id_int, var.sharing_weight,
- var.id_int, var.value);
+ XBT_DEBUG("Min usage: %f, Var(%d).penalty: %f, Var(%d).value: %f ", min_usage, var.rank_, var.sharing_penalty_,
+ var.rank_, var.value_);
/* Update the usage of contraints where this variable is involved */
- for (s_lmm_element_t& elem : var.cnsts) {
- lmm_constraint_t cnst = elem.constraint;
- if (cnst->sharing_policy) {
+ for (Element& elem : var.cnsts_) {
+ Constraint* cnst = elem.constraint;
+ if (cnst->sharing_policy_ != s4u::Link::SharingPolicy::FATPIPE) {
// Remember: shared constraints require that sum(elem.value * var.value) < cnst->bound
- double_update(&(cnst->remaining), elem.consumption_weight * var.value, cnst->bound * sg_maxmin_precision);
- double_update(&(cnst->usage), elem.consumption_weight / var.sharing_weight, sg_maxmin_precision);
+ double_update(&(cnst->remaining_), elem.consumption_weight * var.value_, cnst->bound_ * sg_maxmin_precision);
+ double_update(&(cnst->usage_), elem.consumption_weight / var.sharing_penalty_, sg_maxmin_precision);
// If the constraint is saturated, remove it from the set of active constraints (light_tab)
- if (not double_positive(cnst->usage, sg_maxmin_precision) ||
- not double_positive(cnst->remaining, cnst->bound * sg_maxmin_precision)) {
- if (cnst->cnst_light) {
- int index = (cnst->cnst_light - cnst_light_tab);
+ if (not double_positive(cnst->usage_, sg_maxmin_precision) ||
+ not double_positive(cnst->remaining_, cnst->bound_ * sg_maxmin_precision)) {
+ if (cnst->cnst_light_) {
+ int index = (cnst->cnst_light_ - cnst_light_tab);
XBT_DEBUG("index: %d \t cnst_light_num: %d \t || usage: %f remaining: %f bound: %f ", index,
- cnst_light_num, cnst->usage, cnst->remaining, cnst->bound);
+ cnst_light_num, cnst->usage_, cnst->remaining_, cnst->bound_);
cnst_light_tab[index] = cnst_light_tab[cnst_light_num - 1];
- cnst_light_tab[index].cnst->cnst_light = &cnst_light_tab[index];
+ cnst_light_tab[index].cnst->cnst_light_ = &cnst_light_tab[index];
cnst_light_num--;
- cnst->cnst_light = nullptr;
+ cnst->cnst_light_ = nullptr;
}
} else {
- cnst->cnst_light->remaining_over_usage = cnst->remaining / cnst->usage;
+ cnst->cnst_light_->remaining_over_usage = cnst->remaining_ / cnst->usage_;
}
elem.make_inactive();
} else {
// Remember: non-shared constraints only require that max(elem.value * var.value) < cnst->bound
- cnst->usage = 0.0;
+ cnst->usage_ = 0.0;
elem.make_inactive();
- for (s_lmm_element_t& elem2 : cnst->enabled_element_set) {
- xbt_assert(elem2.variable->sharing_weight > 0);
- if (elem2.variable->value > 0)
+ for (Element& elem2 : cnst->enabled_element_set_) {
+ xbt_assert(elem2.variable->sharing_penalty_ > 0);
+ if (elem2.variable->value_ > 0)
continue;
if (elem2.consumption_weight > 0)
- cnst->usage = std::max(cnst->usage, elem2.consumption_weight / elem2.variable->sharing_weight);
+ cnst->usage_ = std::max(cnst->usage_, elem2.consumption_weight / elem2.variable->sharing_penalty_);
}
// If the constraint is saturated, remove it from the set of active constraints (light_tab)
- if (not double_positive(cnst->usage, sg_maxmin_precision) ||
- not double_positive(cnst->remaining, cnst->bound * sg_maxmin_precision)) {
- if (cnst->cnst_light) {
- int index = (cnst->cnst_light - cnst_light_tab);
+ if (not double_positive(cnst->usage_, sg_maxmin_precision) ||
+ not double_positive(cnst->remaining_, cnst->bound_ * sg_maxmin_precision)) {
+ if (cnst->cnst_light_) {
+ int index = (cnst->cnst_light_ - cnst_light_tab);
XBT_DEBUG("index: %d \t cnst_light_num: %d \t || \t cnst: %p \t cnst->cnst_light: %p "
"\t cnst_light_tab: %p usage: %f remaining: %f bound: %f ",
- index, cnst_light_num, cnst, cnst->cnst_light, cnst_light_tab, cnst->usage, cnst->remaining,
- cnst->bound);
+ index, cnst_light_num, cnst, cnst->cnst_light_, cnst_light_tab, cnst->usage_, cnst->remaining_,
+ cnst->bound_);
cnst_light_tab[index] = cnst_light_tab[cnst_light_num - 1];
- cnst_light_tab[index].cnst->cnst_light = &cnst_light_tab[index];
+ cnst_light_tab[index].cnst->cnst_light_ = &cnst_light_tab[index];
cnst_light_num--;
- cnst->cnst_light = nullptr;
+ cnst->cnst_light_ = nullptr;
}
} else {
- cnst->cnst_light->remaining_over_usage = cnst->remaining / cnst->usage;
- xbt_assert(not cnst->active_element_set.empty(),
+ cnst->cnst_light_->remaining_over_usage = cnst->remaining_ / cnst->usage_;
+ xbt_assert(not cnst->active_element_set_.empty(),
"Should not keep a maximum constraint that has no active"
" element! You want to check the maxmin precision and possible rounding effects.");
}
saturated_constraints.clear();
int pos;
for (pos = 0; pos < cnst_light_num; pos++) {
- xbt_assert(not cnst_light_tab[pos].cnst->active_element_set.empty(),
+ xbt_assert(not cnst_light_tab[pos].cnst->active_element_set_.empty(),
"Cannot saturate more a constraint that has"
" no active element! You may want to change the maxmin precision (--cfg=maxmin/precision:<new_value>)"
" because of possible rounding effects.\n\tFor the record, the usage of this constraint is %g while "
"the maxmin precision to which it is compared is %g.\n\tThe usage of the previous constraint is %g.",
- cnst_light_tab[pos].cnst->usage, sg_maxmin_precision, cnst_light_tab[pos - 1].cnst->usage);
+ cnst_light_tab[pos].cnst->usage_, sg_maxmin_precision, cnst_light_tab[pos - 1].cnst->usage_);
saturated_constraints_update(cnst_light_tab[pos].remaining_over_usage, pos, saturated_constraints, &min_usage);
}
} while (cnst_light_num > 0);
- modified = false;
+ modified_ = false;
if (selective_update_active)
remove_all_modified_set();
delete[] cnst_light_tab;
}
-void lmm_solve(lmm_system_t sys)
-{
- sys->solve();
-}
-
-/** \brief Attribute the value bound to var->bound.
+/** @brief Attribute the value bound to var->bound.
*
- * \param sys the lmm_system_t
- * \param var the lmm_variable_t
- * \param bound the new bound to associate with var
+ * @param var the Variable*
+ * @param bound the new bound to associate with var
*
* Makes var->bound equal to bound. Whenever this function is called a change is signed in the system. To
* avoid false system changing detection it is a good idea to test (bound != 0) before calling it.
*/
-void s_lmm_system_t::update_variable_bound(lmm_variable_t var, double bound)
+void System::update_variable_bound(Variable* var, double bound)
{
- modified = true;
- var->bound = bound;
+ modified_ = true;
+ var->bound_ = bound;
- if (not var->cnsts.empty())
- update_modified_set(var->cnsts[0].constraint);
+ if (not var->cnsts_.empty())
+ update_modified_set(var->cnsts_[0].constraint);
}
-void Variable::initialize(simgrid::surf::Action* id_value, double sharing_weight_value, double bound_value,
+void Variable::initialize(resource::Action* id_value, double sharing_penalty, double bound_value,
int number_of_constraints, unsigned visited_value)
{
- id = id_value;
- id_int = Global_debug_id++;
- cnsts.reserve(number_of_constraints);
- sharing_weight = sharing_weight_value;
- staged_weight = 0.0;
- bound = bound_value;
- concurrency_share = 1;
- value = 0.0;
- visited = visited_value;
- mu = 0.0;
- new_mu = 0.0;
- func_f = func_f_def;
- func_fp = func_fp_def;
- func_fpi = func_fpi_def;
-
- xbt_assert(not variable_set_hook.is_linked());
- xbt_assert(not saturated_variable_set_hook.is_linked());
+ id_ = id_value;
+ rank_ = next_rank_++;
+ cnsts_.reserve(number_of_constraints);
+ sharing_penalty_ = sharing_penalty;
+ staged_penalty_ = 0.0;
+ bound_ = bound_value;
+ concurrency_share_ = 1;
+ value_ = 0.0;
+ visited_ = visited_value;
+ mu_ = 0.0;
+
+ xbt_assert(not variable_set_hook_.is_linked());
+ xbt_assert(not saturated_variable_set_hook_.is_linked());
}
int Variable::get_min_concurrency_slack() const
{
int minslack = std::numeric_limits<int>::max();
- for (s_lmm_element_t const& elem : cnsts) {
+ for (Element const& elem : cnsts_) {
int slack = elem.constraint->get_concurrency_slack();
if (slack < minslack) {
// This is only an optimization, to avoid looking at more constraints when slack is already zero
return minslack;
}
-// Small remark: In this implementation of lmm_enable_var and lmm_disable_var, we will meet multiple times with var when
-// running sys->update_modified_set.
-// A priori not a big performance issue, but we might do better by calling sys->update_modified_set within the for loops
-// (after doing the first for enabling==1, and before doing the last for disabling==1)
-void s_lmm_system_t::enable_var(lmm_variable_t var)
+// Small remark: In this implementation of System::enable_var() and System::disable_var(), we will meet multiple times
+// with var when running System::update_modified_set().
+// A priori not a big performance issue, but we might do better by calling System::update_modified_set() within the for
+// loops (after doing the first for enabling==1, and before doing the last for disabling==1)
+void System::enable_var(Variable* var)
{
xbt_assert(not XBT_LOG_ISENABLED(surf_maxmin, xbt_log_priority_debug) || var->can_enable());
- var->sharing_weight = var->staged_weight;
- var->staged_weight = 0;
+ var->sharing_penalty_ = var->staged_penalty_;
+ var->staged_penalty_ = 0;
// Enabling the variable, move var to list head. Subtlety is: here, we need to call update_modified_set AFTER
// moving at least one element of var.
simgrid::xbt::intrusive_erase(variable_set, *var);
variable_set.push_front(*var);
- for (s_lmm_element_t& elem : var->cnsts) {
- simgrid::xbt::intrusive_erase(elem.constraint->disabled_element_set, elem);
- elem.constraint->enabled_element_set.push_front(elem);
+ for (Element& elem : var->cnsts_) {
+ simgrid::xbt::intrusive_erase(elem.constraint->disabled_element_set_, elem);
+ elem.constraint->enabled_element_set_.push_front(elem);
elem.increase_concurrency();
}
- if (not var->cnsts.empty())
- update_modified_set(var->cnsts[0].constraint);
+ if (not var->cnsts_.empty())
+ update_modified_set(var->cnsts_[0].constraint);
// When used within on_disabled_var, we would get an assertion fail, because transiently there can be variables
// that are staged and could be activated.
// Anyway, caller functions all call check_concurrency() in the end.
}
-void s_lmm_system_t::disable_var(lmm_variable_t var)
+void System::disable_var(Variable* var)
{
- xbt_assert(not var->staged_weight, "Staged weight should have been cleared");
+ xbt_assert(not var->staged_penalty_, "Staged penalty should have been cleared");
// Disabling the variable, move to var to list tail. Subtlety is: here, we need to call update_modified_set
// BEFORE moving the last element of var.
simgrid::xbt::intrusive_erase(variable_set, *var);
variable_set.push_back(*var);
- if (not var->cnsts.empty())
- update_modified_set(var->cnsts[0].constraint);
- for (s_lmm_element_t& elem : var->cnsts) {
- simgrid::xbt::intrusive_erase(elem.constraint->enabled_element_set, elem);
- elem.constraint->disabled_element_set.push_back(elem);
+ if (not var->cnsts_.empty())
+ update_modified_set(var->cnsts_[0].constraint);
+ for (Element& elem : var->cnsts_) {
+ simgrid::xbt::intrusive_erase(elem.constraint->enabled_element_set_, elem);
+ elem.constraint->disabled_element_set_.push_back(elem);
if (elem.active_element_set_hook.is_linked())
- simgrid::xbt::intrusive_erase(elem.constraint->active_element_set, elem);
+ simgrid::xbt::intrusive_erase(elem.constraint->active_element_set_, elem);
elem.decrease_concurrency();
}
- var->sharing_weight = 0.0;
- var->staged_weight = 0.0;
- var->value = 0.0;
+ var->sharing_penalty_ = 0.0;
+ var->staged_penalty_ = 0.0;
+ var->value_ = 0.0;
check_concurrency();
}
/* /brief Find variables that can be enabled and enable them.
*
- * Assuming that the variable has already been removed from non-zero weights
+ * Assuming that the variable has already been removed from non-zero penalties
* Can we find a staged variable to add?
* If yes, check that none of the constraints that this variable is involved in is at the limit of its concurrency
* And then add it to enabled variables
*/
-void s_lmm_system_t::on_disabled_var(lmm_constraint_t cnstr)
+void System::on_disabled_var(Constraint* cnstr)
{
if (cnstr->get_concurrency_limit() < 0)
return;
- int numelem = cnstr->disabled_element_set.size();
+ int numelem = cnstr->disabled_element_set_.size();
if (not numelem)
return;
- lmm_element_t elem = &cnstr->disabled_element_set.front();
+ Element* elem = &cnstr->disabled_element_set_.front();
- // Cannot use foreach loop, because lmm_enable_var will modify disabled_element_set.. within the loop
+ // Cannot use foreach loop, because System::enable_var() will modify disabled_element_set.. within the loop
while (numelem-- && elem) {
- lmm_element_t nextelem;
+ Element* nextelem;
if (elem->disabled_element_set_hook.is_linked()) {
- auto iter = std::next(cnstr->disabled_element_set.iterator_to(*elem));
- nextelem = iter != std::end(cnstr->disabled_element_set) ? &*iter : nullptr;
+ auto iter = std::next(cnstr->disabled_element_set_.iterator_to(*elem));
+ nextelem = iter != std::end(cnstr->disabled_element_set_) ? &*iter : nullptr;
} else {
nextelem = nullptr;
}
- if (elem->variable->staged_weight > 0 && elem->variable->can_enable()) {
+ if (elem->variable->staged_penalty_ > 0 && elem->variable->can_enable()) {
// Found a staged variable
// TODOLATER: Add random timing function to model reservation protocol fuzziness? Then how to make sure that
// staged variables will eventually be called?
enable_var(elem->variable);
}
- xbt_assert(cnstr->concurrency_current <= cnstr->get_concurrency_limit(), "Concurrency overflow!");
- if (cnstr->concurrency_current == cnstr->get_concurrency_limit())
+ xbt_assert(cnstr->concurrency_current_ <= cnstr->get_concurrency_limit(), "Concurrency overflow!");
+ if (cnstr->concurrency_current_ == cnstr->get_concurrency_limit())
break;
elem = nextelem;
// Anyway, caller functions all call check_concurrency() in the end.
}
-/* \brief update the weight of a variable, and enable/disable it.
- * @return Returns whether a change was made
- */
-void s_lmm_system_t::update_variable_weight(lmm_variable_t var, double weight)
+/** @brief update the penalty of a variable (disable it by passing 0 as a penalty) */
+void System::update_variable_penalty(Variable* var, double penalty)
{
- xbt_assert(weight >= 0, "Variable weight should not be negative!");
+ xbt_assert(penalty >= 0, "Variable penalty should not be negative!");
- if (weight == var->sharing_weight)
+ if (penalty == var->sharing_penalty_)
return;
- int enabling_var = (weight > 0 && var->sharing_weight <= 0);
- int disabling_var = (weight <= 0 && var->sharing_weight > 0);
+ int enabling_var = (penalty > 0 && var->sharing_penalty_ <= 0);
+ int disabling_var = (penalty <= 0 && var->sharing_penalty_ > 0);
- XBT_IN("(sys=%p, var=%p, weight=%f)", this, var, weight);
+ XBT_IN("(sys=%p, var=%p, penalty=%f)", this, var, penalty);
- modified = true;
+ modified_ = true;
// Are we enabling this variable?
if (enabling_var) {
- var->staged_weight = weight;
+ var->staged_penalty_ = penalty;
int minslack = var->get_min_concurrency_slack();
- if (minslack < var->concurrency_share) {
- XBT_DEBUG("Staging var (instead of enabling) because min concurrency slack %i, with weight %f and concurrency"
+ if (minslack < var->concurrency_share_) {
+ XBT_DEBUG("Staging var (instead of enabling) because min concurrency slack %i, with penalty %f and concurrency"
" share %i",
- minslack, weight, var->concurrency_share);
+ minslack, penalty, var->concurrency_share_);
return;
}
XBT_DEBUG("Enabling var with min concurrency slack %i", minslack);
enable_var(var);
} else if (disabling_var) {
- // Are we disabling this variable?
disable_var(var);
} else {
- var->sharing_weight = weight;
+ var->sharing_penalty_ = penalty;
}
check_concurrency();
XBT_OUT();
}
-void s_lmm_system_t::update_constraint_bound(lmm_constraint_t cnst, double bound)
+void System::update_constraint_bound(Constraint* cnst, double bound)
{
- modified = true;
+ modified_ = true;
update_modified_set(cnst);
- cnst->bound = bound;
+ cnst->bound_ = bound;
}
-/** \brief Update the constraint set propagating recursively to other constraints so the system should not be entirely
+/** @brief Update the constraint set propagating recursively to other constraints so the system should not be entirely
* computed.
*
- * \param sys the lmm_system_t
- * \param cnst the lmm_constraint_t affected by the change
+ * @param cnst the Constraint* affected by the change
*
* A recursive algorithm to optimize the system recalculation selecting only constraints that have changed. Each
* constraint change is propagated to the list of constraints for each variable.
*/
-void s_lmm_system_t::update_modified_set_rec(lmm_constraint_t cnst)
+void System::update_modified_set_rec(Constraint* cnst)
{
- for (s_lmm_element_t const& elem : cnst->enabled_element_set) {
- lmm_variable_t var = elem.variable;
- for (s_lmm_element_t const& elem2 : var->cnsts) {
- if (var->visited == visited_counter)
+ for (Element const& elem : cnst->enabled_element_set_) {
+ Variable* var = elem.variable;
+ for (Element const& elem2 : var->cnsts_) {
+ if (var->visited_ == visited_counter_)
break;
- if (elem2.constraint != cnst && not elem2.constraint->modified_constraint_set_hook.is_linked()) {
+ if (elem2.constraint != cnst && not elem2.constraint->modified_constraint_set_hook_.is_linked()) {
modified_constraint_set.push_back(*elem2.constraint);
update_modified_set_rec(elem2.constraint);
}
}
// var will be ignored in later visits as long as sys->visited_counter does not move
- var->visited = visited_counter;
+ var->visited_ = visited_counter_;
}
}
-void s_lmm_system_t::update_modified_set(lmm_constraint_t cnst)
+void System::update_modified_set(Constraint* cnst)
{
/* nothing to do if selective update isn't active */
- if (selective_update_active && not cnst->modified_constraint_set_hook.is_linked()) {
+ if (selective_update_active && not cnst->modified_constraint_set_hook_.is_linked()) {
modified_constraint_set.push_back(*cnst);
update_modified_set_rec(cnst);
}
}
-void s_lmm_system_t::remove_all_modified_set()
+void System::remove_all_modified_set()
{
// We cleverly un-flag all variables just by incrementing visited_counter
// In effect, the var->visited value will no more be equal to visited counter
// To be clean, when visited counter has wrapped around, we force these var->visited values so that variables that
// were in the modified a long long time ago are not wrongly skipped here, which would lead to very nasty bugs
// (i.e. not readibily reproducible, and requiring a lot of run time before happening).
- if (++visited_counter == 1) {
+ if (++visited_counter_ == 1) {
/* the counter wrapped around, reset each variable->visited */
for (Variable& var : variable_set)
- var.visited = 0;
+ var.visited_ = 0;
}
modified_constraint_set.clear();
}
/**
* Returns resource load (in flop per second, or byte per second, or similar)
*
- * If the resource is shared (the default case), the load is sum of resource usage made by every variables located on
- * this resource.
- *
- * If the resource is not shared (ie in FATPIPE mode), then the load is the max (not the sum) of all resource usages
- * located on this resource.
+ * If the resource is shared (the default case), the load is sum of resource usage made by
+ * every variables located on this resource.
*
- * \param cnst the lmm_constraint_t associated to the resource
+ * If the resource is not shared (ie in FATPIPE mode), then the load is the max (not the sum)
+ * of all resource usages located on this resource.
*/
double Constraint::get_usage() const
{
double result = 0.0;
- if (sharing_policy) {
- for (s_lmm_element_t const& elem : enabled_element_set)
+ if (sharing_policy_ != s4u::Link::SharingPolicy::FATPIPE) {
+ for (Element const& elem : enabled_element_set_)
if (elem.consumption_weight > 0)
- result += elem.consumption_weight * elem.variable->value;
+ result += elem.consumption_weight * elem.variable->value_;
} else {
- for (s_lmm_element_t const& elem : enabled_element_set)
+ for (Element const& elem : enabled_element_set_)
if (elem.consumption_weight > 0)
- result = std::max(result, elem.consumption_weight * elem.variable->value);
+ result = std::max(result, elem.consumption_weight * elem.variable->value_);
}
return result;
}
int Constraint::get_variable_amount() const
{
- return std::count_if(std::begin(enabled_element_set), std::end(enabled_element_set),
- [](const s_lmm_element_t& elem) { return elem.consumption_weight > 0; });
+ return std::count_if(std::begin(enabled_element_set_), std::end(enabled_element_set_),
+ [](const Element& elem) { return elem.consumption_weight > 0; });
}
}
}
