#2387: Gather hashed trace user events at the end of run #2395
+164
−43
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Pipelines resultsPR tests (gcc-12, ubuntu, mpich, verbose, kokkos) Build for 56b67a9 (2025-02-14 16:19:23 UTC) PR tests (intel icpx, ubuntu, mpich, verbose) Build for 56b67a9 (2025-02-14 16:19:23 UTC) PR tests (clang-9, ubuntu, mpich) Build for 56b67a9 (2025-02-14 16:19:23 UTC) PR tests (clang-12, ubuntu, mpich) Build for 56b67a9 (2025-02-14 16:19:23 UTC) PR tests (gcc-9, ubuntu, mpich, zoltan) Build for 56b67a9 (2025-02-14 16:19:23 UTC) PR tests (clang-11, ubuntu, mpich) Build for 56b67a9 (2025-02-14 16:19:23 UTC) PR tests (gcc-10, ubuntu, openmpi, no LB) Build for 56b67a9 (2025-02-14 16:19:23 UTC) PR tests (clang-13, ubuntu, mpich) Build for 56b67a9 (2025-02-14 16:19:23 UTC) PR tests (clang-14, ubuntu, mpich, verbose) Build for 56b67a9 (2025-02-14 16:19:23 UTC) PR tests (clang-10, ubuntu, mpich) Build for 56b67a9 (2025-02-14 16:19:23 UTC) PR tests (nvidia cuda 11.2, gcc-9, ubuntu, mpich) Build for 56b67a9 (2025-02-14 16:19:23 UTC) PR tests (intel icpc, ubuntu, mpich) Build for 56b67a9 (2025-02-14 16:19:23 UTC) PR tests (gcc-11, ubuntu, mpich, trace runtime, coverage) Build for 56b67a9 (2025-02-14 16:19:23 UTC) PR tests (nvidia cuda 12.2.0, gcc-9, ubuntu, mpich, verbose) Build for 56b67a9 (2025-02-14 16:19:23 UTC) PR tests (clang-13, alpine, mpich) Build for 56b67a9 (2025-02-14 16:19:23 UTC) |
cwschilly
changed the title
cwschilly
force-pushed
the
2387-hashed-trace-user-events-should-do-an-all-reduce-at-the-end-instead-of-sending-pt2pts-to-rank-0-during-the-run
branch
from
1742430 to
7ee3b85
Compare
cwschilly
force-pushed
the
2387-hashed-trace-user-events-should-do-an-all-reduce-at-the-end-instead-of-sending-pt2pts-to-rank-0-during-the-run
branch
from
ef7edca to
9e40158
Compare
cwschilly
force-pushed
the
2387-hashed-trace-user-events-should-do-an-all-reduce-at-the-end-instead-of-sending-pt2pts-to-rank-0-during-the-run
branch
from
067d4a3 to
27b7d4c
Compare
cwschilly
commented
Feb 11, 2025
cwschilly
commented
Feb 11, 2025
|
This PR does not compile in Is it possible to use |
No, because in |
cwschilly
force-pushed
the
2387-hashed-trace-user-events-should-do-an-all-reduce-at-the-end-instead-of-sending-pt2pts-to-rank-0-during-the-run
branch
from
8665c22 to
444d54e
Compare
cz4rs
reviewed
Feb 18, 2025
There was a problem hiding this comment.
For what it's worth, I can reproduce the segfault easily:
Thread 1 "hello_world" received signal SIGSEGV, Segmentation fault.
vt::objgroup::ObjGroupManager::makeCollectiveImpl (this=0x0, label="Trace", base=std::unique_ptr<vt::objgroup::holder::HolderBase> = {...}, obj_ptr=0x3096ed0)
at /home/cz4rs/code/vt/src/vt/objgroup/manager.cc:89
89 auto const id = cur_obj_id_++;
(gdb) bt
#0 vt::objgroup::ObjGroupManager::makeCollectiveImpl (this=0x0, label="Trace", base=std::unique_ptr<vt::objgroup::holder::HolderBase> = {...}, obj_ptr=0x3096ed0)
at /home/cz4rs/code/vt/src/vt/objgroup/manager.cc:89
#1 0x0000000001c2dcf6 in vt::objgroup::ObjGroupManager::makeCollectiveObj<vt::trace::Trace> (this=0x0, label="Trace", obj=0x3096ed0,
holder=std::unique_ptr<vt::objgroup::holder::HolderBase> = {...}) at /home/cz4rs/code/vt/src/vt/objgroup/manager.impl.h:107
#2 0x0000000001c2a022 in vt::objgroup::ObjGroupManager::makeCollective<vt::trace::Trace> (this=0x0, obj=0x3096ed0, label="Trace")
at /home/cz4rs/code/vt/src/vt/objgroup/manager.impl.h:82
#3 0x0000000001bb309f in vt::trace::Trace::construct (in_prog_name="hello_world") at /home/cz4rs/code/vt/src/vt/trace/trace.cc:133
#4 0x00000000016a6f30 in vt::runtime::component::ComponentConstructor<vt::trace::Trace, void, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&>::apply<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&> (args="hello_world")
at /home/cz4rs/code/vt/src/vt/runtime/component/component.h:73
#5 0x00000000016a6ee0 in vt::runtime::component::Component<vt::trace::Trace>::staticInit<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&> (args="hello_world") at /home/cz4rs/code/vt/src/vt/runtime/component/component.h:123
#6 0x0000000001672470 in vt::runtime::component::(anonymous namespace)::tupleConsImpl<vt::trace::Trace, std::tuple<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&>, 0ul> (tup=..., seq=...) at /home/cz4rs/code/vt/src/vt/runtime/component/component_pack.impl.h:57
#7 0x0000000001672430 in vt::runtime::component::(anonymous namespace)::tupleCons<vt::trace::Trace, std::tuple<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&> > (tup=...) at /home/cz4rs/code/vt/src/vt/runtime/component/component_pack.impl.h:64
#8 0x00000000016a6aec in vt::runtime::component::ComponentPack::registerComponent<vt::trace::Trace, vt::ctx::Context, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&>(vt::trace::Trace**, vt::runtime::component::BaseComponent::DepsPack<vt::ctx::Context>, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)::{lambda()#1}::operator()() (this=0x3075488) at /home/cz4rs/code/vt/src/vt/runtime/component/component_pack.impl.h:93
#9 0x00000000016a6ab9 in vt::runtime::component::MovableFnTyped<vt::runtime::component::ComponentPack::registerComponent<vt::trace::Trace, vt::ctx::Context, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&>(vt::trace::Trace**, vt::runtime::component::BaseComponent::DepsPack<vt::ctx::Context>, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)::{lambda()#1}>::invoke() (this=0x3075480)
at /home/cz4rs/code/vt/src/vt/runtime/component/movable_fn.h:78
#10 0x000000000179d6af in vt::runtime::component::ComponentPack::construct (this=0x3067d20) at /home/cz4rs/code/vt/src/vt/runtime/component/component_pack.cc:63
#11 0x000000000165facb in vt::runtime::Runtime::initializeComponents (this=0x2f1f510) at /home/cz4rs/code/vt/src/vt/runtime/runtime.cc:963
#12 0x000000000165eca8 in vt::runtime::Runtime::initialize (this=0x2f1f510, force_now=true) at /home/cz4rs/code/vt/src/vt/runtime/runtime.cc:453
#13 0x000000000165ec5e in vt::runtime::Runtime::tryInitialize (this=0x2f1f510) at /home/cz4rs/code/vt/src/vt/runtime/runtime.cc:402
#14 0x000000000165ee4a in vt::runtime::Runtime::initialize (this=0x2f1f510, force_now=false) at /home/cz4rs/code/vt/src/vt/runtime/runtime.cc:491
#15 0x00000000016586f6 in vt::CollectiveAnyOps<(vt::runtime::eRuntimeInstance)0>::initialize (argc=@0x7fffffffc988: 1, argv=@0x7fffffffc980: 0x307dea0, is_interop=false,
comm=0x0, appConfig=0x0) at /home/cz4rs/code/vt/src/vt/collective/collective_ops.cc:238
#16 0x000000000165a35d in vt::initialize (argc=@0x7fffffffc988: 1, argv=@0x7fffffffc980: 0x307dea0, comm=0x0, appConfig=0x0)
at /home/cz4rs/code/vt/src/vt/collective/startup.cc:78
#17 0x0000000001609692 in main (argc=1, argv=0x307dea0) at /home/cz4rs/code/vt/examples/hello_world/hello_world.cc:52
|
@cwschilly Let me know if you need help debugging the failures. I'll try to do a proper review soon-ish. |
cwschilly
force-pushed
the
2387-hashed-trace-user-events-should-do-an-all-reduce-at-the-end-instead-of-sending-pt2pts-to-rank-0-during-the-run
branch
from
b55a195 to
9d4a193
Compare
lifflander
approved these changes
Feb 26, 2025
…tempt allreduce at end of run
lifflander
force-pushed
the
2387-hashed-trace-user-events-should-do-an-all-reduce-at-the-end-instead-of-sending-pt2pts-to-rank-0-during-the-run
branch
from
9d4a193 to
a7343f4
Compare
|
diff --git a/src/vt/pipe/callback/cb_union/cb_raw_base.fwd.h b/src/vt/pipe/callback/cb_union/cb_raw_base.fwd.h
index 416fb3403..974ac0695 100644
--- a/src/vt/pipe/callback/cb_union/cb_raw_base.fwd.h
+++ b/src/vt/pipe/callback/cb_union/cb_raw_base.fwd.h
@@ -44,9 +44,11 @@
#if !defined INCLUDED_VT_PIPE_CALLBACK_CB_UNION_CB_RAW_BASE_FWD_H
#define INCLUDED_VT_PIPE_CALLBACK_CB_UNION_CB_RAW_BASE_FWD_H
-namespace vt { namespace pipe { namespace callback { namespace cbunion {
+namespace vt {
+namespace pipe { namespace callback { namespace cbunion {
-template <typename... Args> struct CallbackTyped;
+template <typename... Args>
+struct CallbackTyped;
struct CallbackRawBaseSingle;
diff --git a/src/vt/runtime/component/component_pack.cc b/src/vt/runtime/component/component_pack.cc
index 6775c83bb..00a0f8941 100644
--- a/src/vt/runtime/component/component_pack.cc
+++ b/src/vt/runtime/component/component_pack.cc
@@ -143,7 +143,9 @@ std::list<int> ComponentPack::topoSort() {
return order;
}
-void ComponentPack::topoSortImpl(int v, std::list<int>& order, bool* visited, bool* visiting) {
+void ComponentPack::topoSortImpl(
+ int v, std::list<int>& order, bool* visited, bool* visiting
+) {
//fmt::print("impl v={}\n",v);
vtAbortIf(visiting[v] == true, "Already visiting this node, cycle detected");
visiting[v] = true;
diff --git a/src/vt/runtime/component/component_pack.h b/src/vt/runtime/component/component_pack.h
index 2b2fa8e41..32a49bef9 100644
--- a/src/vt/runtime/component/component_pack.h
+++ b/src/vt/runtime/component/component_pack.h
@@ -162,7 +162,8 @@ private:
* \param[in] visited array of visited vertices
* \param[in] visiting array of currently visiting vertices
*/
- void topoSortImpl(int v, std::list<int>& order, bool* visited, bool* visiting);
+ void
+ topoSortImpl(int v, std::list<int>& order, bool* visited, bool* visiting);
/**
* \internal \brief Detect cycles in the dependence graph
diff --git a/src/vt/runtime/runtime.cc b/src/vt/runtime/runtime.cc
index 083468bc9..cb330adee 100644
--- a/src/vt/runtime/runtime.cc
+++ b/src/vt/runtime/runtime.cc
@@ -749,10 +749,11 @@ void Runtime::initializeComponents() {
# if vt_check_enabled(trace_enabled)
// The Trace and Scheduler components have a co-dependency. However,
// the lifetime of theTrace should be longer than that of theSched.
- p_->registerComponent<trace::Trace>(&theTrace, Deps<
- ctx::Context, // Everything depends on theContext
- objgroup::ObjGroupManager
- >{},
+ p_->registerComponent<trace::Trace>(
+ &theTrace,
+ Deps<
+ ctx::Context, // Everything depends on theContext
+ objgroup::ObjGroupManager>{},
prog_name
);
# endif
@@ -768,12 +769,12 @@ void Runtime::initializeComponents() {
#endif
p_->registerComponent<objgroup::ObjGroupManager>(
- &theObjGroup, Deps<
- ctx::Context // Everything depends on theContext
+ &theObjGroup,
+ Deps<ctx::Context // Everything depends on theContext
- // Break this dependency for startup ordering
- // messaging::ActiveMessenger // Depends on active messenger to send
- >{}
+ // Break this dependency for startup ordering
+ // messaging::ActiveMessenger // Depends on active messenger to send
+ >{}
);
p_->registerComponent<messaging::ActiveMessenger>(
diff --git a/src/vt/trace/trace.cc b/src/vt/trace/trace.cc
index e98bae813..675c68e32 100644
--- a/src/vt/trace/trace.cc
+++ b/src/vt/trace/trace.cc
@@ -127,14 +127,13 @@ void Trace::setProxy(objgroup::proxy::Proxy<Trace> in_proxy) {
}
#endif
-/*static*/ std::unique_ptr<Trace> Trace::construct(std::string const& in_prog_name) {
+/*static*/ std::unique_ptr<Trace>
+Trace::construct(std::string const& in_prog_name) {
auto ptr = std::make_unique<Trace>(in_prog_name);
- #if !vt_check_enabled(trace_only)
- auto proxy = theObjGroup()->makeCollective<Trace>(
- ptr.get(), "Trace"
- );
+#if !vt_check_enabled(trace_only)
+ auto proxy = theObjGroup()->makeCollective<Trace>(ptr.get(), "Trace");
proxy.get()->setProxy(proxy);
- #endif
+#endif
return ptr;
}
@@ -219,9 +218,9 @@ void Trace::setUserEvents(const UserEventRegistry& events) {
}
void Trace::gatherUserEvents() {
- #if !vt_check_enabled(trace_only)
+#if !vt_check_enabled(trace_only)
proxy_.reduce<&reducedEventsHan, vt::collective::PlusOp>(0, user_event_);
- #endif
+#endif
}
UserEventIDType Trace::registerUserEventRoot(std::string const& name) {
@@ -241,10 +240,12 @@ void Trace::registerUserEventManual(
void reducedEventsHan(
[[maybe_unused]] const UserEventRegistry& gathered_user_events
) {
- #if vt_check_enabled(trace_enabled)
- vtAssert(theContext()->getNode() == 0, "User events must be gathered on node 0");
+#if vt_check_enabled(trace_enabled)
+ vtAssert(
+ theContext()->getNode() == 0, "User events must be gathered on node 0"
+ );
theTrace()->setUserEvents(gathered_user_events);
- #endif
+#endif
}
void insertNewUserEvent(
diff --git a/src/vt/trace/trace.h b/src/vt/trace/trace.h
index 88fc1ef3a..393cb1a97 100644
--- a/src/vt/trace/trace.h
+++ b/src/vt/trace/trace.h
@@ -135,9 +135,9 @@ struct Trace : runtime::component::Component<Trace>, TraceLite {
void startup() override;
void finalize() override;
- #if !vt_check_enabled(trace_only)
+#if !vt_check_enabled(trace_only)
void setProxy(objgroup::proxy::Proxy<Trace> in_proxy);
- #endif
+#endif
/**
* \brief Initiate a paired processing event.
@@ -407,31 +407,15 @@ struct Trace : runtime::component::Component<Trace>, TraceLite {
template <typename SerializerT>
void serialize(SerializerT& s) {
- s | incremental_flush_mode_
- | traces_
- | open_events_
- | event_holds_
- | cur_event_
- | enabled_
- | idle_begun_
- | start_time_
- | user_event_
- | prog_name_
- | trace_name_
- | full_trace_name_
- | full_sts_name_
- | full_dir_name_
- | wrote_sts_file_
- | trace_write_count_
- | spec_proxy_
- #if !vt_check_enabled(trace_only)
+ s | incremental_flush_mode_ | traces_ | open_events_ | event_holds_ |
+ cur_event_ | enabled_ | idle_begun_ | start_time_ | user_event_ |
+ prog_name_ | trace_name_ | full_trace_name_ | full_sts_name_ |
+ full_dir_name_ | wrote_sts_file_ | trace_write_count_ | spec_proxy_
+#if !vt_check_enabled(trace_only)
| proxy_
- #endif
- | trace_enabled_cur_phase_
- | flush_event_
- | between_sched_event_type_
- | between_sched_event_
- | inside_invoke_context_;
+#endif
+ | trace_enabled_cur_phase_ | flush_event_ | between_sched_event_type_ |
+ between_sched_event_ | inside_invoke_context_;
s.skip(log_file_); // definition unavailable
}
@@ -445,10 +429,10 @@ private:
ObjGroupProxyType spec_proxy_ = vt::no_obj_group;
- #if !vt_check_enabled(trace_only)
+#if !vt_check_enabled(trace_only)
// Objgroup proxy
objgroup::proxy::Proxy<Trace> proxy_;
- #endif
+#endif
// Processing event between top-level loops.
TraceEntryIDType between_sched_event_type_ = no_trace_entry_id;
diff --git a/src/vt/trace/trace_user_event.cc b/src/vt/trace/trace_user_event.cc
index e0a13179d..269794027 100644
--- a/src/vt/trace/trace_user_event.cc
+++ b/src/vt/trace/trace_user_event.cc
@@ -126,15 +126,13 @@ bool UserEventRegistry::insertEvent(
std::forward_as_tuple(name)
);
return true;
- } else if (user_event_[event] != name){
+ } else if (user_event_[event] != name) {
user_event_[event] += " COLLISION " + name;
}
return false;
}
-UserEventRegistry operator+(
- UserEventRegistry r1, UserEventRegistry const& r2
-) {
+UserEventRegistry operator+(UserEventRegistry r1, UserEventRegistry const& r2) {
for (auto& [hash, event_str] : r2.getEvents()) {
r1.insertEvent(hash, event_str);
}
diff --git a/src/vt/trace/trace_user_event.h b/src/vt/trace/trace_user_event.h
index de4f0bad5..07f10553e 100644
--- a/src/vt/trace/trace_user_event.h
+++ b/src/vt/trace/trace_user_event.h
@@ -123,7 +123,8 @@ struct UserEventRegistry {
friend void insertNewUserEvent(UserEventIDType event, std::string const& name);
- friend UserEventRegistry operator+(UserEventRegistry r1, UserEventRegistry const& r2);
+ friend UserEventRegistry
+ operator+(UserEventRegistry r1, UserEventRegistry const& r2);
template <typename Serializer>
void serialize(Serializer& s) { |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixes #2387