P4 to Git Change 2037301 by ssahasra@ssahasra-hip-vdi on 2019/11/26 22:42:25

SWDEV-204782 - introduce hostcall

	Hostcall is a service that allows a kernel to submit requests to the
	host using shared buffers, and block until a response is received. This
	will eventually replace the shared buffer currently used for printf, and
	repurposes the same hidden kernel argument.

	When the runtime launches a kernel that requires the hostcall service it
	performs the following actions:
	- Launch a hostcall listener thread if it is not already running.
	- Locate the hostcall buffer for the corresponding hardware queue, or
	  create a new one.
	- Register the new hostcall buffer with the listener thread.
	- Set the hostcall buffer pointer as an implicit argument to the kernel.

Affected files ...

... //depot/stg/opencl/drivers/opencl/make/hip.git/tests/Makefile#21 edit
... //depot/stg/opencl/drivers/opencl/make/hip.git/tests/build/Makefile.hip_tests#31 edit
... //depot/stg/opencl/drivers/opencl/make/hip.git/tests/scripts/hip_hostcall_tests.txt#1 add
... //depot/stg/opencl/drivers/opencl/make/hip.git/tests/scripts/run_all_tests.sh#22 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/devkernel.cpp#30 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/devkernel.hpp#19 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocdevice.cpp#143 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocdevice.hpp#45 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rochostcall.cpp#1 add
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rochostcall.hpp#1 add
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocvirtual.cpp#92 edit

rocclr/runtime/device/rocm/rocdevice.cpp

@@ -28,6 +28,8 @@
 #include "pro/prodriver.hpp"
 #endif
 #include "platform/sampler.hpp"
+#include "rochostcall.hpp"
+
 #include <cstring>
 #include <fstream>
 #include <sstream>
@@ -1916,10 +1918,52 @@ void Device::releaseQueue(hsa_queue_t* queue) {
   }
   ClPrint(amd::LOG_INFO, amd::LOG_QUEUE, "deleting hardware queue %p with refCount 0", queue);
 
+  if (qInfo.hostcallBuffer_) {
+    ClPrint(amd::LOG_INFO, amd::LOG_QUEUE, "deleting hostcall buffer %p for hardware queue %p",
+            qInfo.hostcallBuffer_, queue);
+    disableHostcalls(qInfo.hostcallBuffer_, queue);
+    context().svmFree(qInfo.hostcallBuffer_);
+  }
+
+  ClPrint(amd::LOG_INFO, amd::LOG_QUEUE, "deleting hardware queue %p with refCount 0", queue);
   hsa_queue_destroy(queue);
   queuePool_.erase(qIter);
 }
 
+void* Device::getOrCreateHostcallBuffer(hsa_queue_t* queue) {
+  auto qIter = queuePool_.find(queue);
+  assert(qIter != queuePool_.end());
+
+  auto& qInfo = qIter->second;
+  if (qInfo.hostcallBuffer_) {
+    return qInfo.hostcallBuffer_;
+  }
+
+  // The number of packets required in each buffer is at least equal to the
+  // maximum number of waves supported by the device.
+  auto wavesPerCu = info().maxThreadsPerCU_ / info().wavefrontWidth_;
+  auto numPackets = info().maxComputeUnits_ * wavesPerCu;
+
+  auto size = getHostcallBufferSize(numPackets);
+  auto align = getHostcallBufferAlignment();
+
+  void* buffer = context().svmAlloc(size, align, CL_MEM_SVM_FINE_GRAIN_BUFFER | CL_MEM_SVM_ATOMICS);
+  if (!buffer) {
+    ClPrint(amd::LOG_ERROR, amd::LOG_QUEUE,
+            "Failed to create hostcall buffer for hardware queue %p", queue);
+    return nullptr;
+  }
+  ClPrint(amd::LOG_INFO, amd::LOG_QUEUE, "Created hostcall buffer %p for hardware queue %p", buffer,
+          queue);
+  qInfo.hostcallBuffer_ = buffer;
+  if (!enableHostcalls(buffer, numPackets, queue)) {
+    ClPrint(amd::LOG_ERROR, amd::LOG_QUEUE, "Failed to register hostcall buffer %p with listener",
+            buffer);
+    return nullptr;
+  }
+  return buffer;
+}
+
 bool Device::findLinkTypeAndHopCount(amd::Device* other_device,
                                      uint32_t* link_type, uint32_t* hop_count) {
   hsa_amd_memory_pool_link_info_t link_info;