#include <boost/algorithm/string.hpp>
#include <boost/algorithm/string/classification.hpp>
#include <boost/algorithm/string/split.hpp>
-#include <simgrid/msg.h>
#include <simgrid/s4u/Engine.hpp>
#include <string>
#include <utility>
XBT_LOG_NEW_DEFAULT_SUBCATEGORY(surf_plugin_dvfs, surf, "Logging specific to the SURF HostDvfs plugin");
+static const char* property_sampling_rate = "plugin/dvfs/sampling_rate";
+static const char* property_governor = "plugin/dvfs/governor";
+
namespace simgrid {
namespace plugin {
void init()
{
- const char* local_sampling_rate_config = host->getProperty("plugin/dvfs/sampling_rate");
- double global_sampling_rate_config = xbt_cfg_get_double("plugin/dvfs/sampling_rate");
+ const char* local_sampling_rate_config = host->getProperty(property_sampling_rate);
+ double global_sampling_rate_config = xbt_cfg_get_double(property_sampling_rate);
if (local_sampling_rate_config != nullptr) {
sampling_rate = std::stod(local_sampling_rate_config);
} else {
}
}
- virtual void update() {}
+ virtual void update() = 0;
+ virtual std::string getName() = 0;
double samplingRate() { return sampling_rate; }
};
+/**
+ * The linux kernel doc describes this governor as follows:
+ * https://www.kernel.org/doc/Documentation/cpu-freq/governors.txt
+ *
+ * > The CPUfreq governor "performance" sets the CPU statically to the
+ * > highest frequency within the borders of scaling_min_freq and
+ * > scaling_max_freq.
+ *
+ * We do not support scaling_min_freq/scaling_max_freq -- we just pick the lowest frequency.
+ */
class Performance : public Governor {
public:
explicit Performance(simgrid::s4u::Host* ptr) : Governor(ptr) {}
- void update() { host->setPstate(0); }
+ void update() override { host->setPstate(0); }
+ std::string getName() override { return "Performance"; }
};
+/**
+ * The linux kernel doc describes this governor as follows:
+ * https://www.kernel.org/doc/Documentation/cpu-freq/governors.txt
+ *
+ * > The CPUfreq governor "powersave" sets the CPU statically to the
+ * > lowest frequency within the borders of scaling_min_freq and
+ * > scaling_max_freq.
+ *
+ * We do not support scaling_min_freq/scaling_max_freq -- we just pick the lowest frequency.
+ */
class Powersave : public Governor {
public:
explicit Powersave(simgrid::s4u::Host* ptr) : Governor(ptr) {}
- void update() { host->setPstate(host->getPstatesCount() - 1); }
+ void update() override { host->setPstate(host->getPstatesCount() - 1); }
+ std::string getName() override { return "Powersave"; }
};
+/**
+ * The linux kernel doc describes this governor as follows:
+ * https://www.kernel.org/doc/Documentation/cpu-freq/governors.txt
+ *
+ * > The CPUfreq governor "ondemand" sets the CPU frequency depending on the
+ * > current system load. [...] when triggered, cpufreq checks
+ * > the CPU-usage statistics over the last period and the governor sets the
+ * > CPU accordingly.
+ */
class OnDemand : public Governor {
- double freq_up_threshold = 0.95;
+ /**
+ * See https://elixir.bootlin.com/linux/v4.15.4/source/drivers/cpufreq/cpufreq_ondemand.c
+ * DEF_FREQUENCY_UP_THRESHOLD and od_update()
+ */
+ double freq_up_threshold = 0.80;
public:
explicit OnDemand(simgrid::s4u::Host* ptr) : Governor(ptr) {}
- void update()
+ std::string getName() override { return "OnDemand"; }
+ void update() override
{
- double load = sg_host_get_current_load(host);
+ double load = host->getCoreCount() * sg_host_get_avg_load(host);
+ sg_host_load_reset(host); // Only consider the period between two calls to this method!
- // FIXME I don't like that we multiply with the getCoreCount() just here...
- if (load*host->getCoreCount() > freq_up_threshold) {
+ if (load > freq_up_threshold) {
host->setPstate(0); /* Run at max. performance! */
- XBT_INFO("Changed to pstate %f", 0.0);
+ XBT_INFO("Load: %f > threshold: %f --> changed to pstate %i", load, freq_up_threshold, 0);
} else {
/* The actual implementation uses a formula here: (See Kernel file cpufreq_ondemand.c:158)
*
* lowest_pstate - load*pstatesCount()
*/
int max_pstate = host->getPstatesCount() - 1;
- int new_pstate = max_pstate - load * max_pstate;
+ // Load is now < freq_up_threshold; exclude pstate 0 (the fastest)
+ // because pstate 0 can only be selected if load > freq_up_threshold
+ int new_pstate = max_pstate - load * (max_pstate + 1);
host->setPstate(new_pstate);
- XBT_DEBUG("Load: %f --> changed to pstate %i", load*host->getCoreCount(), new_pstate);
+ XBT_DEBUG("Load: %f < threshold: %f --> changed to pstate %i", load, freq_up_threshold, new_pstate);
}
}
};
+/**
+ * This is the conservative governor, which is very similar to the
+ * OnDemand governor. The Linux Kernel Documentation describes it
+ * very well, see https://www.kernel.org/doc/Documentation/cpu-freq/governors.txt:
+ *
+ * > The CPUfreq governor "conservative", much like the "ondemand"
+ * > governor, sets the CPU frequency depending on the current usage. It
+ * > differs in behaviour in that it gracefully increases and decreases the
+ * > CPU speed rather than jumping to max speed the moment there is any load
+ * > on the CPU. This behaviour is more suitable in a battery powered
+ * > environment.
+ */
class Conservative : public Governor {
double freq_up_threshold = .8;
double freq_down_threshold = .2;
public:
explicit Conservative(simgrid::s4u::Host* ptr) : Governor(ptr) {}
- void update()
+ virtual std::string getName() override { return "Conservative"; }
+ virtual void update() override
{
- double load = sg_host_get_current_load(host)*host->getCoreCount();
+ double load = host->getCoreCount() * sg_host_get_avg_load(host);
int pstate = host->getPstate();
+ sg_host_load_reset(host); // Only consider the period between two calls to this method!
if (load > freq_up_threshold) {
if (pstate != 0) {
else {
XBT_DEBUG("Load: %f > threshold: %f -> but cannot speed up even more, already in highest pstate %d", load, freq_up_threshold, pstate);
}
- }
-
- if (load < freq_down_threshold) {
+ } else if (load < freq_down_threshold) {
int max_pstate = host->getPstatesCount() - 1;
if (pstate != max_pstate) { // Are we in the slowest pstate already?
host->setPstate(pstate + 1);
}
}
};
-}
+/**
+ * Add this to your host tag:
+ * - <prop id="plugin/dvfs/governor" value="performance" />
+ *
+ * Valid values as of now are: performance, powersave, ondemand, conservative
+ * It doesn't matter if you use uppercase or lowercase.
+ *
+ * For the sampling rate, use this:
+ *
+ * - <prop id="plugin/dvfs/sampling_rate" value="2" />
+ *
+ * This will run the update() method of the specified governor every 2 seconds
+ * on that host.
+ *
+ * These properties can also be used within the <config> tag to configure
+ * these values globally. Using them within the <host> will overwrite this
+ * global configuration
+ */
class HostDvfs {
public:
static simgrid::xbt::Extension<simgrid::s4u::Host, HostDvfs> EXTENSION_ID;
HostDvfs::~HostDvfs() = default;
}
}
+}
-using simgrid::plugin::HostDvfs;
+using simgrid::plugin::dvfs::HostDvfs;
/* **************************** events callback *************************** */
static void on_host_added(simgrid::s4u::Host& host)
XBT_DEBUG("DVFS process on %s is a daemon: %d", daemonProc->getHost()->getName().c_str(), daemonProc->isDaemon());
std::string dvfs_governor;
- const char* host_conf = daemonProc->getHost()->getProperty("plugin/dvfs/governor");
+ const char* host_conf = daemonProc->getHost()->getProperty(property_governor);
if (host_conf != nullptr) {
- dvfs_governor = std::string(daemonProc->getHost()->getProperty("plugin/dvfs/governor"));
+ dvfs_governor = std::string(daemonProc->getHost()->getProperty(property_governor));
boost::algorithm::to_lower(dvfs_governor);
} else {
- dvfs_governor = xbt_cfg_get_string("plugin/dvfs/governor");
+ dvfs_governor = xbt_cfg_get_string(property_governor);
boost::algorithm::to_lower(dvfs_governor);
}
- simgrid::plugin::dvfs::Governor governor(daemonProc->getHost());
- if (dvfs_governor == "conservative") {
- governor = simgrid::plugin::dvfs::Conservative(daemonProc->getHost());
- }
+ auto governor = [&dvfs_governor, &daemonProc]() {
+ if (dvfs_governor == "conservative") {
+ return std::unique_ptr<simgrid::plugin::dvfs::Governor>(
+ new simgrid::plugin::dvfs::Conservative(daemonProc->getHost()));
+ } else if (dvfs_governor == "ondemand") {
+ return std::unique_ptr<simgrid::plugin::dvfs::Governor>(
+ new simgrid::plugin::dvfs::OnDemand(daemonProc->getHost()));
+ } else if (dvfs_governor == "performance") {
+ return std::unique_ptr<simgrid::plugin::dvfs::Governor>(
+ new simgrid::plugin::dvfs::Performance(daemonProc->getHost()));
+ } else if (dvfs_governor == "powersave") {
+ return std::unique_ptr<simgrid::plugin::dvfs::Governor>(
+ new simgrid::plugin::dvfs::Powersave(daemonProc->getHost()));
+ } else {
+ XBT_CRITICAL("No governor specified for host %s, falling back to Performance",
+ daemonProc->getHost()->getCname());
+ return std::unique_ptr<simgrid::plugin::dvfs::Governor>(
+ new simgrid::plugin::dvfs::Performance(daemonProc->getHost()));
+ }
+ }();
while (1) {
// Sleep *before* updating; important for startup (i.e., t = 0).
// In the beginning, we want to go with the pstates specified in the platform file
// (so we sleep first)
- simgrid::s4u::this_actor::sleep_for(governor.samplingRate());
- governor.update();
- XBT_INFO("Governor just updated!");
+ simgrid::s4u::this_actor::sleep_for(governor->samplingRate());
+ governor->update();
+ XBT_DEBUG("Governor (%s) just updated!", governor->getName().c_str());
}
XBT_WARN("I should have never reached this point: daemons should be killed when all regular processes are done");
sg_host_load_plugin_init();
simgrid::s4u::Host::onCreation.connect(&on_host_added);
- xbt_cfg_register_double("plugin/dvfs/sampling_rate", 0.1, nullptr,
+ xbt_cfg_register_double(property_sampling_rate, 0.1, nullptr,
"How often should the dvfs plugin check whether the frequency needs to be changed?");
- xbt_cfg_register_string("plugin/dvfs/governor", "performance", nullptr,
+ xbt_cfg_register_string(property_governor, "performance", nullptr,
"Which Governor should be used that adapts the CPU frequency?");
}
}
