AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[SDK][rocprofv3] MI300 Stochastic PC sampling (#92)

Browse Source 
* MI300 Stochastic PC sampling SDK API implementation

* ROCProfV3: Stochastic PC sampling Support (#94)

* ROCProfV3: MI300 Stochastic PC sampling initial draft

* ROCProfV3: Initial Stochastic PC sampling Tests (#95)

ROCProfV3: Initial Stochastic PC sampling tests

* Update rocprofiler_pc_sampling_record_stochastic_v0_t

- update doxygen docs for members
- replace rocprofiler_correlation_id_t with rocprofiler_async_correlation_id_t

* Relax the check in JSON tests

* drain PC sampling buffer during finalize_rocprofv3

* Increase timeout for "Test Install Build" step

- 10 minutes -> 20 minutes
- "Test Installed Packages" has 20 minutes so "Test Install Build" should also

---------

Co-authored-by: Jonathan R. Madsen <jonathanrmadsen@gmail.com>
This commit is contained in:
Indic, Vladimir
2025-03-21 20:40:45 +01:00
committed by GitHub
parent c06feccf2a
commit 49ce79a5b5
98 changed files with 5266 additions and 1031 deletions
Download Patch File Download Diff File Expand all files Collapse all files
source/lib/rocprofiler-sdk/pc_sampling/service.cpp
+40 -1
View File
@@ -263,6 +263,7 @@ flush_internal_agent_buffers(rocprofiler_buffer_id_t buffer_id)
auto* service = get_configured_pc_sampling_service().load();
if(service && ctx->pc_sampler.get() == service)
{
rocprofiler_status_t status = ROCPROFILER_STATUS_SUCCESS;
// The context `ctx` (that holds the buffer with `buffer_id`)
// is the one containing PC sampling service.
// The HSA interception table is registered.
@@ -272,7 +273,10 @@ flush_internal_agent_buffers(rocprofiler_buffer_id_t buffer_id)
if(agent_session->buffer_id.handle == buffer_id.handle)
{
// Flush internal PC sampling buffers filled by the agent
return hsa::flush_internal_agent_buffers(agent_session.get());
// NOTE: one rocprofiler-SDK PC sampling buffer can be tied
// to multiple agent (agent sessions).
status = hsa::flush_internal_agent_buffers(agent_session.get());
if(status != ROCPROFILER_STATUS_SUCCESS) return status;
}
}
}
@@ -281,6 +285,41 @@ flush_internal_agent_buffers(rocprofiler_buffer_id_t buffer_id)
return ROCPROFILER_STATUS_SUCCESS;
}
rocprofiler_status_t
flush_all_agent_buffers()
{
auto* service = get_configured_pc_sampling_service().load();
if(!service) return ROCPROFILER_STATUS_ERROR;
rocprofiler_status_t status = ROCPROFILER_STATUS_SUCCESS;
// Loop over all agents that have PC sampling service configured
// and drain their internal buffers.
// NOTE: one SDK buffer can consume data from multiple agents
// (multiple HSA runtime buffers)
for(const auto& [_, agent_session] : service->agent_sessions)
{
status = flush_internal_agent_buffers(agent_session->buffer_id);
if(status != ROCPROFILER_STATUS_SUCCESS)
{
ROCP_ERROR << "Failed to flush internal HSA buffers tied to rocp buffer "
<< agent_session->buffer_id.handle;
}
}
return status;
}
void
service_sync()
{
flush_all_agent_buffers();
}
void
service_fini()
{
flush_all_agent_buffers();
}
} // namespace pc_sampling
} // namespace rocprofiler