AI/rocm-systems
2
0
Rozštěpit 0
Zdrojový kód Úkoly Pull requesty Akce Balíčky Projekty Vydání Wiki Aktivita

Report page migration events as start/end (#793)

Procházet zdrojové kódy 
* 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>
Tento commit je obsažen v:
Mythreya
2024-11-11 09:08:47 -08:00
odevzdal GitHub
rodič 472907a576
revize 363f85dc72
27 změnil soubory, kde provedl 1161 přidání a 1098 odebrání
Stáhněte soubor záplaty Stáhněte rozdílový soubor Rozbalit všechny soubory Sbalit všechny soubory
samples/api_buffered_tracing/client.cpp
+68 -25
Zobrazit soubor
@@ -86,6 +86,22 @@ rocprofiler_buffer_id_t client_buffer = {};
buffer_name_info client_name_info = {};
kernel_symbol_map_t client_kernels = {};
template <typename Tp>
std::string
as_hex(Tp _v, size_t _width = 16)
{
uintptr_t _vp = 0;
if constexpr(std::is_pointer<Tp>::value)
_vp = reinterpret_cast<uintptr_t>(_v);
else
_vp = _v;
auto _ss = std::stringstream{};
_ss.fill('0');
_ss << "0x" << std::hex << std::setw(_width) << _vp;
return _ss.str();
}
void
print_call_stack(const call_stack_t& _call_stack)
{
@@ -290,43 +306,71 @@ tool_tracing_callback(rocprofiler_context_id_t context,
auto info = std::stringstream{};
info << "kind=" << record->kind << ", operation=" << record->operation
<< ", pid=" << record->pid << ", start=" << record->start_timestamp
<< ", stop=" << record->end_timestamp
<< ", name=" << client_name_info.at(record->kind, record->operation);
<< ", pid=" << record->pid << ", timestamp=" << record->timestamp
<< ", name=" << client_name_info.at(record->kind, record->operation)
<< std::boolalpha;
switch(record->operation)
{
case ROCPROFILER_PAGE_MIGRATION_PAGE_MIGRATE:
case ROCPROFILER_PAGE_MIGRATION_PAGE_MIGRATE_START:
{
info << ", page_fault=(" << record->page_fault.read_fault << ", "
<< record->page_fault.migrated << ", " << record->page_fault.node_id
<< ", " << std::hex << "0x" << record->page_fault.address << ")";
const auto& arg = record->args;
info << ", page_migrate_start=(" << as_hex(arg.page_migrate_start.start_addr)
<< ", " << as_hex(arg.page_migrate_start.end_addr) << ", "
<< arg.page_migrate_start.from_agent.handle << ", "
<< arg.page_migrate_start.to_agent.handle << ", "
<< arg.page_migrate_start.prefetch_agent.handle << ", "
<< arg.page_migrate_start.preferred_agent.handle << ", "
<< arg.page_migrate_start.trigger << ")";
break;
}
case ROCPROFILER_PAGE_MIGRATION_PAGE_FAULT:
case ROCPROFILER_PAGE_MIGRATION_PAGE_MIGRATE_END:
{
info << ", page_migrate=(" << std::hex << "0x"
<< record->page_migrate.start_addr << ", 0x"
<< record->page_migrate.end_addr << ", " << std::dec
<< record->page_migrate.from_node << ", " << record->page_migrate.to_node
<< ", " << record->page_migrate.prefetch_node << ", "
<< record->page_migrate.preferred_node << ", "
<< record->page_migrate.trigger << ")";
const auto& arg = record->args;
info << ", page_migrate_end=(" << as_hex(arg.page_migrate_end.start_addr)
<< ", " << as_hex(arg.page_migrate_end.end_addr) << ", "
<< arg.page_migrate_end.from_agent.handle << ", "
<< arg.page_migrate_end.to_agent.handle << ", "
<< arg.page_migrate_end.trigger << ")";
break;
}
case ROCPROFILER_PAGE_MIGRATION_QUEUE_SUSPEND:
case ROCPROFILER_PAGE_MIGRATION_PAGE_FAULT_START:
{
info << ", queue_suspend=(" << record->queue_suspend.rescheduled << ", "
<< record->queue_suspend.node_id << ", " << record->queue_suspend.trigger
<< ")";
const auto& arg = record->args;
info << ", page_fault_start=(" << arg.page_fault_start.read_fault << ", "
<< arg.page_fault_start.agent_id.handle << ", "
<< as_hex(arg.page_fault_start.address) << ")";
break;
}
case ROCPROFILER_PAGE_MIGRATION_PAGE_FAULT_END:
{
const auto& arg = record->args;
info << ", page_fault_end=(" << arg.page_fault_end.migrated << ", "
<< arg.page_fault_end.agent_id.handle << ", "
<< as_hex(arg.page_fault_end.address) << ")";
break;
}
case ROCPROFILER_PAGE_MIGRATION_QUEUE_EVICTION:
{
const auto& arg = record->args;
info << ", queue_eviction=(" << arg.queue_eviction.agent_id.handle << ", "
<< arg.queue_eviction.trigger << ")";
break;
}
case ROCPROFILER_PAGE_MIGRATION_QUEUE_RESTORE:
{
const auto& arg = record->args;
info << ", queue_restore=(" << arg.queue_restore.rescheduled << ", "
<< arg.queue_restore.agent_id.handle << ")";
break;
}
case ROCPROFILER_PAGE_MIGRATION_UNMAP_FROM_GPU:
{
info << ", unmap_from_gpu=(" << record->unmap_from_gpu.node_id << std::hex
<< ", 0x" << record->unmap_from_gpu.start_addr << ", 0x"
<< record->unmap_from_gpu.end_addr << ", " << std::dec
<< record->unmap_from_gpu.trigger << ")";
const auto& arg = record->args;
info << ", unmap_from_gpu=(" << as_hex(arg.unmap_from_gpu.start_addr) << ", "
<< as_hex(arg.unmap_from_gpu.end_addr) << ", "
<< arg.unmap_from_gpu.agent_id.handle << ", " << arg.unmap_from_gpu.trigger
<< ")";
break;
}
case ROCPROFILER_PAGE_MIGRATION_NONE:
@@ -337,8 +381,7 @@ tool_tracing_callback(rocprofiler_context_id_t context,
}
}
if(record->start_timestamp > record->end_timestamp)
throw std::runtime_error("page migration: start > end");
if(record->timestamp == 0) throw std::runtime_error("page migration: timestamp == 0");
static_cast<call_stack_t*>(user_data)->emplace_back(
source_location{__FUNCTION__, __FILE__, __LINE__, kind_name + info.str()});