Report page migration events as start/end (#793)
* Squashed commit of the following: commit b76f2635f4b65599f03812a73d0cf410f5ada213 Author: Mythreya <mythreya.kuricheti@amd.com> Date: Fri Apr 26 00:29:09 2024 +0000 Changed for PR feedback commit bedb8ad566ff42fbf117b19202c26c507abcf8ac Author: Jonathan R. Madsen <jonathanrmadsen@gmail.com> Date: Thu Apr 25 19:20:06 2024 -0500 Fix installation commit a98f8a69459a1450a1be9c98e20b3c1e7f2568c2 Author: Jonathan R. Madsen <jonathanrmadsen@gmail.com> Date: Thu Apr 25 19:16:35 2024 -0500 Restructure the headers commit 46489a020ffafdd5f4ce3f580469ff233ef67fe1 Author: Mythreya <mythreya.kuricheti@amd.com> Date: Tue Apr 23 23:31:10 2024 +0000 Update hsa include commit 8e795282cce348fc6aa736b7857b21aeb32aa20a Author: Mythreya <mythreya.kuricheti@amd.com> Date: Tue Apr 23 23:02:32 2024 +0000 Report page migration events as start/end * Updated tests accordingly * Page migration events are reported independently commit 8784e5ad4895a626a2a8e4ac12f8021b34172bd4 Author: Mythreya <mythreya.kuricheti@amd.com> Date: Tue Apr 16 17:01:57 2024 +0000 Update handling of dropped page migration events Previously, we dropped all locally buffered events when we detect that KFD has dropped some events. This may drop too many pending events too eagerly. When we receive an end event and cannot find the corresponding start, we can be sure that KFD has dropped some events in the immediate past. When this happens, we look through all locally buffered events and report the start events that are older than 10s as partial events --- they have no "end" information (we expect that the end events have been dropped). We also set the polling timeout to 10s to prevent the local buffer from getting too large with events waiting to be paired up. Updated tests commit 2e8e0b07eeda9b5990e1ae8d28dcd3a035ce38e1 Author: Mythreya <mythreya.kuricheti@amd.com> Date: Tue Apr 16 17:01:31 2024 +0000 Docs for triggers * Fix page migration sample * Fix hasher, kfd install * Add hsa include * Install KFD include dir * Updates from code review - single timestamp field - node_id -> agent_id - from_node -> from_agent - to_node -> to_agent * Misc revisions * Remove page-migration install target * Update page-migration pytest * Tweak to serialization * Address PR comments * Update page-migration test * Add cli args, update iterations * Address PR comments * Add abi.cpp for static_asserts * Update page_migration gtest with only runtime tests * Moved helpers into utils.hpp --------- Co-authored-by: Jonathan R. Madsen <jonathanrmadsen@gmail.com>
This commit is contained in:
@@ -24,6 +24,7 @@
|
||||
#pragma once
|
||||
|
||||
#include <rocprofiler-sdk/fwd.h>
|
||||
#include <rocprofiler-sdk/hsa.h>
|
||||
#include <rocprofiler-sdk/internal_threading.h>
|
||||
|
||||
namespace rocprofiler
|
||||
|
||||
@@ -59,6 +59,7 @@
|
||||
|
||||
#include <string>
|
||||
#include <string_view>
|
||||
#include <utility>
|
||||
#include <vector>
|
||||
|
||||
#define ROCP_SDK_SAVE_DATA_FIELD(FIELD) ar(make_nvp(#FIELD, data.FIELD))
|
||||
@@ -460,6 +461,122 @@ save(ArchiveT& ar, rocprofiler_buffer_tracing_memory_copy_record_t data)
|
||||
ROCP_SDK_SAVE_DATA_FIELD(bytes);
|
||||
}
|
||||
|
||||
template <typename ArchiveT>
|
||||
void
|
||||
save(ArchiveT& ar, const rocprofiler_page_migration_page_fault_start_t& data)
|
||||
{
|
||||
ROCP_SDK_SAVE_DATA_BITFIELD("read_fault", read_fault);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(agent_id);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(address);
|
||||
}
|
||||
|
||||
template <typename ArchiveT>
|
||||
void
|
||||
save(ArchiveT& ar, const rocprofiler_page_migration_page_fault_end_t& data)
|
||||
{
|
||||
ROCP_SDK_SAVE_DATA_BITFIELD("migrated", migrated);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(agent_id);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(address);
|
||||
}
|
||||
|
||||
template <typename ArchiveT>
|
||||
void
|
||||
save(ArchiveT& ar, const rocprofiler_page_migration_page_migrate_start_t& data)
|
||||
{
|
||||
ROCP_SDK_SAVE_DATA_FIELD(start_addr);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(end_addr);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(from_agent);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(to_agent);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(prefetch_agent);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(preferred_agent);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(trigger);
|
||||
}
|
||||
|
||||
template <typename ArchiveT>
|
||||
void
|
||||
save(ArchiveT& ar, const rocprofiler_page_migration_page_migrate_end_t& data)
|
||||
{
|
||||
ROCP_SDK_SAVE_DATA_FIELD(start_addr);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(end_addr);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(from_agent);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(to_agent);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(trigger);
|
||||
}
|
||||
|
||||
template <typename ArchiveT>
|
||||
void
|
||||
save(ArchiveT& ar, const rocprofiler_page_migration_queue_eviction_t& data)
|
||||
{
|
||||
ROCP_SDK_SAVE_DATA_FIELD(agent_id);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(trigger);
|
||||
}
|
||||
|
||||
template <typename ArchiveT>
|
||||
void
|
||||
save(ArchiveT& ar, const rocprofiler_page_migration_queue_restore_t& data)
|
||||
{
|
||||
ROCP_SDK_SAVE_DATA_BITFIELD("rescheduled", rescheduled);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(agent_id);
|
||||
}
|
||||
|
||||
template <typename ArchiveT>
|
||||
void
|
||||
save(ArchiveT& ar, const rocprofiler_page_migration_unmap_from_gpu_t& data)
|
||||
{
|
||||
ROCP_SDK_SAVE_DATA_FIELD(start_addr);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(end_addr);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(agent_id);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(trigger);
|
||||
}
|
||||
|
||||
namespace details
|
||||
{
|
||||
template <size_t Idx>
|
||||
struct save_page_migration_arg;
|
||||
|
||||
#define ROCP_SDK_SPECIALIZE_PAGE_MIGRATION_ARG(ENUM_VALUE, UNION_ARG) \
|
||||
template <> \
|
||||
struct save_page_migration_arg<ROCPROFILER_PAGE_MIGRATION_##ENUM_VALUE> \
|
||||
{ \
|
||||
static constexpr auto value = ROCPROFILER_PAGE_MIGRATION_##ENUM_VALUE; \
|
||||
template <typename ArchiveT> \
|
||||
void operator()(ArchiveT& ar, rocprofiler_page_migration_args_t args) \
|
||||
{ \
|
||||
ar(make_nvp(#UNION_ARG, args.UNION_ARG)); \
|
||||
} \
|
||||
};
|
||||
|
||||
ROCP_SDK_SPECIALIZE_PAGE_MIGRATION_ARG(NONE, none)
|
||||
ROCP_SDK_SPECIALIZE_PAGE_MIGRATION_ARG(PAGE_MIGRATE_START, page_migrate_start)
|
||||
ROCP_SDK_SPECIALIZE_PAGE_MIGRATION_ARG(PAGE_MIGRATE_END, page_migrate_end)
|
||||
ROCP_SDK_SPECIALIZE_PAGE_MIGRATION_ARG(PAGE_FAULT_START, page_fault_start)
|
||||
ROCP_SDK_SPECIALIZE_PAGE_MIGRATION_ARG(PAGE_FAULT_END, page_fault_end)
|
||||
ROCP_SDK_SPECIALIZE_PAGE_MIGRATION_ARG(QUEUE_EVICTION, queue_eviction)
|
||||
ROCP_SDK_SPECIALIZE_PAGE_MIGRATION_ARG(QUEUE_RESTORE, queue_restore)
|
||||
ROCP_SDK_SPECIALIZE_PAGE_MIGRATION_ARG(UNMAP_FROM_GPU, unmap_from_gpu)
|
||||
|
||||
#undef ROCP_SDK_SPECIALIZE_PAGE_MIGRATION_ARG
|
||||
|
||||
template <typename ArchiveT, size_t Idx, size_t... IdxTail>
|
||||
void
|
||||
save_page_migration_args(ArchiveT& ar,
|
||||
rocprofiler_page_migration_operation_t op,
|
||||
rocprofiler_page_migration_args_t args,
|
||||
std::index_sequence<Idx, IdxTail...>)
|
||||
{
|
||||
using save_page_migration_type = save_page_migration_arg<Idx>;
|
||||
if(op == save_page_migration_type::value)
|
||||
{
|
||||
if constexpr(save_page_migration_type::value != ROCPROFILER_PAGE_MIGRATION_NONE)
|
||||
save_page_migration_type{}(ar, args);
|
||||
}
|
||||
else if constexpr(sizeof...(IdxTail) > 0)
|
||||
{
|
||||
save_page_migration_args(ar, op, args, std::index_sequence<IdxTail...>{});
|
||||
}
|
||||
}
|
||||
} // namespace details
|
||||
|
||||
template <typename ArchiveT>
|
||||
void
|
||||
save(ArchiveT& ar, const rocprofiler_buffer_tracing_page_migration_record_t& data)
|
||||
@@ -467,81 +584,10 @@ save(ArchiveT& ar, const rocprofiler_buffer_tracing_page_migration_record_t& dat
|
||||
ROCP_SDK_SAVE_DATA_FIELD(size);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(kind);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(operation);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(start_timestamp);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(end_timestamp);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(timestamp);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(pid);
|
||||
|
||||
switch(data.operation)
|
||||
{
|
||||
case ROCPROFILER_PAGE_MIGRATION_PAGE_FAULT:
|
||||
{
|
||||
ar(make_nvp("page_fault", data.page_fault));
|
||||
break;
|
||||
}
|
||||
case ROCPROFILER_PAGE_MIGRATION_PAGE_MIGRATE:
|
||||
{
|
||||
ar(make_nvp("page_migrate", data.page_migrate));
|
||||
break;
|
||||
}
|
||||
case ROCPROFILER_PAGE_MIGRATION_QUEUE_SUSPEND:
|
||||
{
|
||||
ar(make_nvp("queue_suspend", data.queue_suspend));
|
||||
break;
|
||||
}
|
||||
case ROCPROFILER_PAGE_MIGRATION_UNMAP_FROM_GPU:
|
||||
{
|
||||
ar(make_nvp("unmap_from_gpu", data.unmap_from_gpu));
|
||||
break;
|
||||
}
|
||||
case ROCPROFILER_PAGE_MIGRATION_NONE:
|
||||
case ROCPROFILER_PAGE_MIGRATION_LAST:
|
||||
{
|
||||
throw std::runtime_error{"unsupported page migration operation type"};
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <typename ArchiveT>
|
||||
void
|
||||
save(ArchiveT& ar, const rocprofiler_buffer_tracing_page_migration_page_fault_record_t& data)
|
||||
{
|
||||
ROCP_SDK_SAVE_DATA_FIELD(node_id);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(address);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(read_fault);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(migrated);
|
||||
}
|
||||
|
||||
template <typename ArchiveT>
|
||||
void
|
||||
save(ArchiveT& ar, const rocprofiler_buffer_tracing_page_migration_page_migrate_record_t& data)
|
||||
{
|
||||
ROCP_SDK_SAVE_DATA_FIELD(start_addr);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(end_addr);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(from_node);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(to_node);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(prefetch_node);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(preferred_node);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(trigger);
|
||||
}
|
||||
|
||||
template <typename ArchiveT>
|
||||
void
|
||||
save(ArchiveT& ar, const rocprofiler_buffer_tracing_page_migration_queue_suspend_record_t& data)
|
||||
{
|
||||
ROCP_SDK_SAVE_DATA_FIELD(node_id);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(trigger);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(rescheduled);
|
||||
}
|
||||
|
||||
template <typename ArchiveT>
|
||||
void
|
||||
save(ArchiveT& ar, const rocprofiler_buffer_tracing_page_migration_unmap_from_gpu_record_t& data)
|
||||
{
|
||||
ROCP_SDK_SAVE_DATA_FIELD(node_id);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(start_addr);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(end_addr);
|
||||
ROCP_SDK_SAVE_DATA_FIELD(trigger);
|
||||
details::save_page_migration_args(
|
||||
ar, data.operation, data.args, std::make_index_sequence<ROCPROFILER_PAGE_MIGRATION_LAST>{});
|
||||
}
|
||||
|
||||
template <typename ArchiveT>
|
||||
|
||||
Reference in New Issue
Block a user