Disable PC sampling service if counter collection service is configur…

…ed (#899)

source/include/rocprofiler-sdk/pc_sampling.h

@@ -90,6 +90,9 @@ ROCPROFILER_EXTERN_C_INIT
  *
  * Constraint4: PC sampling feature is not available within the ROCgdb.
  *
+ * Constraint5: PC sampling service cannot be used simultaneously with
+ * counter collection service.
+ *
  * @param [in] context_id - id of the context used for starting/stopping PC sampling service
  * @param [in] agent_id   - id of the agent on which caller tries using PC sampling capability
  * @param [in] method     - the type of PC sampling the caller tries to use on the agent.
@@ -105,7 +108,8 @@ ROCPROFILER_EXTERN_C_INIT
  * @retval ::ROCPROFILER_STATUS_ERROR_INCOMPATIBLE_KERNEL the amdgpu driver installed on the system
  * does not support the PC sampling feature
  * @retval ::ROCPROFILER_STATUS_ERROR a general error caused by the amdgpu driver
- *
+ * @retval ::ROCPROFILER_STATUS_ERROR_CONTEXT_CONFLICT counter collection service already
+ * setup in the context
  */
 rocprofiler_status_t ROCPROFILER_API
 rocprofiler_configure_pc_sampling_service(rocprofiler_context_id_t         context_id,

source/lib/rocprofiler-sdk/pc_sampling/service.cpp

@@ -162,6 +162,20 @@ configure_pc_sampling_service(context::context*                ctx,
                               uint64_t                         interval,
                               rocprofiler_buffer_id_t          buffer_id)
 {
+    // FIXME: PC Sampling cannot be used simultaneously with counter collection.
+    // PC sampling requires clock gating to be disabled on MI2xx and MI3xx,
+    // otherwise a weird GPU hang might appear and a machine must be rebooted.
+    // Current implementation of (dispatch) counter collection service assumes disabling
+    // the clock gating before dispatching a kernel and reenabling the clock gating
+    // after kernel completion. Consequently, if PC sampling is active, (dispatch)
+    // counter collection service can enable clock gating and hang might appear.
+    // As a workaround, PC sampling and (dispatch) counter collection service
+    // cannot coexist in the same context.
+    if(ctx->counter_collection || ctx->agent_counter_collection)
+    {
+        return ROCPROFILER_STATUS_ERROR_CONTEXT_CONFLICT;
+    }
+
     if(!ctx->pc_sampler)
     {
         ctx->pc_sampler = std::make_unique<context::pc_sampling_service>();

source/lib/rocprofiler-sdk/pc_sampling/tests/CMakeLists.txt

@@ -5,7 +5,7 @@ include(GoogleTest)
 set(ROCPROFILER_LIB_PC_SAMPLING_TEST_SOURCES
     configure_service.cpp cid_manager.cpp
     # samples_processing.cpp
-    query_configuration.cpp)
+    pc_sampling_vs_counter_collection.cpp query_configuration.cpp)
 set(ROCPROFILER_LIB_PC_SAMPLING_TEST_HEADERS pc_sampling_internals.hpp)
 
 add_executable(pcs-test)