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

Put cached perfetto traces as default one (#2138)

Browse Source 
* Put cached perfetto traces as default one

* Improve cached data and perfetto traces in order to be more aligned with E2E tests

* Addressing PR comments and findings

* Force early instrumentation bundle instantiation

* Sync-up insturumented containers with thread growth data

* Revert ompvv number of host threads to default 8

* Fixed counter track namings for amd-smi

* AIPROFSYST-34 [rocprof-sys] Update documentation describing newly introduced changes to default tracing mechanism
This commit is contained in:
marantic-amd
2025-12-22 12:47:35 +01:00
committed by GitHub
parent 7da3275b42
commit ba1380a75d
36 changed files with 557 additions and 183 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rocprofiler-systems/source/lib/rocprof-sys/library.cpp
+9 -3
View File
@@ -899,11 +899,12 @@ rocprofsys_finalize_hidden(void)
#endif
ROCPROFSYS_DEBUG_F("Stopping and destroying instrumentation bundles...\n");
for(size_t i = 0; i < thread_info::get_peak_num_threads(); ++i)
auto* _bundles = instrumentation_bundles::get();
for(size_t i = 0; _bundles && i < thread_info::get_peak_num_threads(); ++i)
{
if(!instrumentation_bundles::get()) continue;
if(i >= _bundles->size()) continue;
const auto& _info = thread_info::get(i, SequentTID);
auto& itr = instrumentation_bundles::get()->at(i);
auto& itr = _bundles->at(i);
while(itr != nullptr && !itr->empty())
{
int _lvl = 1;
@@ -1026,6 +1027,11 @@ rocprofsys_finalize_hidden(void)
tracing::copy_timemory_hash_ids();
// Flush any pending region cache entries (e.g., main entry point that wasn't
// explicitly stopped before finalization)
ROCPROFSYS_DEBUG_F("Flushing pending region cache entries...\n");
rocprofsys_flush_pending_region_cache_hidden();
bool _perfetto_output_error = false;
if(get_use_perfetto())
{