rocr/dtif: add hsaKmtQueueRingDoorbell in thunk loader

hsaKmtQueueRingDoorbell is specfic to DTIF backend

Signed-off-by: Flora Cui <flora.cui@amd.com>
Signed-off-by: Jiadong Zhu <Jiadong.Zhu@amd.com>
Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Reviewed-by: Shane Xiao <shane.xiao@amd.com>
Reviewed-by: David Yat Sin <David.YatSin@amd.com>


[ROCm/ROCR-Runtime commit: e2d767879d]

projects/rocr-runtime/runtime/hsa-runtime/core/inc/thunk_loader.h

@@ -318,6 +318,7 @@ class ThunkLoader {
                                       HsaPcSamplingTraceId traceId);
     typedef HSAKMT_STATUS (HSAKMT_DEF(hsaKmtPcSamplingSupport))(void);
     typedef HSAKMT_STATUS (HSAKMT_DEF(hsaKmtModelEnabled))(bool* enable);
+    typedef HSAKMT_STATUS (HSAKMT_DEF(hsaKmtQueueRingDoorbell))(HSA_QUEUEID QueueId);
 
     /* drm API */
     typedef int (DRM_DEF(amdgpu_device_initialize))(int fd, \
@@ -454,6 +455,7 @@ class ThunkLoader {
     HSAKMT_DEF(hsaKmtPcSamplingStop)* HSAKMT_PFN(hsaKmtPcSamplingStop);
     HSAKMT_DEF(hsaKmtPcSamplingSupport)* HSAKMT_PFN(hsaKmtPcSamplingSupport);
     HSAKMT_DEF(hsaKmtModelEnabled)* HSAKMT_PFN(hsaKmtModelEnabled);
+    HSAKMT_DEF(hsaKmtQueueRingDoorbell)* HSAKMT_PFN(hsaKmtQueueRingDoorbell);
 
     DRM_DEF(amdgpu_device_initialize)* DRM_PFN(amdgpu_device_initialize);
     DRM_DEF(amdgpu_device_deinitialize)* DRM_PFN(amdgpu_device_deinitialize);

projects/rocr-runtime/runtime/hsa-runtime/core/runtime/amd_aql_queue.cpp

@@ -474,8 +474,12 @@ uint64_t AqlQueue::AddWriteIndexRelease(uint64_t value) {
 
 void AqlQueue::StoreRelaxed(hsa_signal_value_t value) {
   if (doorbell_type_ == 2) {
-    // Hardware doorbell supports AQL semantics.
-    atomic::Store(signal_.hardware_doorbell_ptr, uint64_t(value), std::memory_order_release);
+    if (core::Runtime::runtime_singleton_->flag().enable_dtif()) {
+      HSAKMT_CALL(hsaKmtQueueRingDoorbell(queue_id_));
+    } else {
+      // Hardware doorbell supports AQL semantics.
+      atomic::Store(signal_.hardware_doorbell_ptr, uint64_t(value), std::memory_order_release);
+    }
     return;
   }
 

projects/rocr-runtime/runtime/hsa-runtime/core/runtime/thunk_loader.cpp

@@ -348,6 +348,9 @@ namespace core {
       HSAKMT_PFN(hsaKmtModelEnabled) = (HSAKMT_DEF(hsaKmtModelEnabled)*)dlsym(dtif_handle, "hsaKmtModelEnabled");
       if (HSAKMT_PFN(hsaKmtModelEnabled) == NULL) goto ERROR;
 
+      HSAKMT_PFN(hsaKmtQueueRingDoorbell) = (HSAKMT_DEF(hsaKmtQueueRingDoorbell)*)dlsym(dtif_handle, "hsaKmtQueueRingDoorbell");
+      if (HSAKMT_PFN(hsaKmtQueueRingDoorbell) == NULL) goto ERROR;
+
       DRM_PFN(amdgpu_device_initialize) = (DRM_DEF(amdgpu_device_initialize)*)dlsym(dtif_handle, "amdgpu_device_initialize");
       if (DRM_PFN(amdgpu_device_initialize) == NULL) goto ERROR;