Add capability to csr to allow N:1 aggregation when ooq is created.

- This allows applications to force the N:1 aggregation by creating out
of order queue.
- That switches csr to N:1 submission model where commands from multiple
command streams may be aggregated.
- That forces scenarios returning an event to be aggregated as well.

Change-Id: I8fd8d7f88bb2665234ee90870133120b206710a8

runtime/command_queue/command_queue_hw.h

@@ -66,6 +66,7 @@ class CommandQueueHw : public CommandQueue {
 
         if (getCmdQueueProperties<cl_queue_properties>(properties, CL_QUEUE_PROPERTIES) & static_cast<cl_queue_properties>(CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE)) {
             device->getCommandStreamReceiver().overrideDispatchPolicy(DispatchMode::BatchedDispatch);
+            device->getCommandStreamReceiver().enableNTo1SubmissionModel();
         }
     }
 

runtime/command_queue/enqueue_common.h

@@ -556,7 +556,7 @@ CompletionStamp CommandQueueHw<GfxFamily>::enqueueNonBlocked(
     dispatchFlags.implicitFlush = implicitFlush;
     dispatchFlags.flushStampReference = this->flushStamp->getStampReference();
     dispatchFlags.preemptionMode = PreemptionHelper::taskPreemptionMode(*device, multiDispatchInfo);
-    dispatchFlags.outOfOrderExecutionAllowed = !eventBuilder.getEvent() || this->isOOQEnabled();
+    dispatchFlags.outOfOrderExecutionAllowed = !eventBuilder.getEvent() || commandStreamReceiver.isNTo1SubmissionModelEnabled();
 
     DEBUG_BREAK_IF(taskLevel >= Event::eventNotReady);
 

runtime/command_stream/command_stream_receiver.h

@@ -104,6 +104,8 @@ class CommandStreamReceiver {
 
     uint32_t peekLatestFlushedTaskCount() const { return latestFlushedTaskCount; }
 
+    void enableNTo1SubmissionModel() { this->nTo1SubmissionModelEnabled = true; }
+    bool isNTo1SubmissionModelEnabled() const { return this->nTo1SubmissionModelEnabled; }
     void overrideDispatchPolicy(DispatchMode overrideValue) { this->dispatchMode = overrideValue; }
 
     virtual void overrideMediaVFEStateDirty(bool dirty) { mediaVfeStateDirty = dirty; }
@@ -173,6 +175,7 @@ class CommandStreamReceiver {
     std::unique_ptr<OSInterface> osInterface;
     std::unique_ptr<SubmissionAggregator> submissionAggregator;
 
+    bool nTo1SubmissionModelEnabled = false;
     DispatchMode dispatchMode = DispatchMode::ImmediateDispatch;
     bool disableL3Cache = false;
     uint32_t requiredScratchSize = 0;

unit_tests/api/cl_create_command_queue_tests.cpp

@@ -84,4 +84,16 @@ HWTEST_F(clCreateCommandQueueTest, givenOoqParametersWhenQueueIsCreatedThenComma
     EXPECT_EQ(DispatchMode::BatchedDispatch, csr.dispatchMode);
     retVal = clReleaseCommandQueue(cmdq);
 }
+
+HWTEST_F(clCreateCommandQueueTest, givenOoqParametersWhenQueueIsCreatedThenCommandStreamReceiverSwitchesToNTo1SubmissionModel) {
+    cl_int retVal = CL_SUCCESS;
+    cl_queue_properties ooq = CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE;
+    auto &csr = reinterpret_cast<UltCommandStreamReceiver<FamilyType> &>(pContext->getDevice(0)->getCommandStreamReceiver());
+    EXPECT_FALSE(csr.isNTo1SubmissionModelEnabled());
+
+    auto cmdq = clCreateCommandQueue(pContext, devices[0], ooq, &retVal);
+    EXPECT_TRUE(csr.isNTo1SubmissionModelEnabled());
+    retVal = clReleaseCommandQueue(cmdq);
+}
+
 } // namespace ULT

unit_tests/command_queue/enqueue_kernel_tests.cpp

@@ -1469,10 +1469,34 @@ HWTEST_F(EnqueueKernelTest, givenInOrderCommandQueueWhenEnqueueKernelReturningEv
     clReleaseEvent(event);
 }
 
-HWTEST_F(EnqueueKernelTest, givenOutOfOrderCommandQueueWhenEnqueueKernelReturningEventIsMadeThenPipeControlPositionIsRecorded) {
-    const cl_queue_properties props[3] = {CL_QUEUE_PROPERTIES, CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE, 0};
+HWTEST_F(EnqueueKernelTest, givenInOrderCommandQueueWhenEnqueueKernelReturningEventIsMadeAndCommandStreamReceiverIsInNTo1ModeThenPipeControlPositionIsRecorded) {
+    const cl_queue_properties props[] = {0};
     auto inOrderQueue = clCreateCommandQueueWithProperties(context, pDevice, props, nullptr);
 
+    auto mockCsr = new MockCsrHw2<FamilyType>(pDevice->getHardwareInfo());
+    mockCsr->overrideDispatchPolicy(DispatchMode::BatchedDispatch);
+    pDevice->resetCommandStreamReceiver(mockCsr);
+    mockCsr->enableNTo1SubmissionModel();
+
+    auto mockedSubmissionsAggregator = new mockSubmissionsAggregator();
+    mockCsr->overrideSubmissionAggregator(mockedSubmissionsAggregator);
+
+    MockKernelWithInternals mockKernel(*pDevice);
+    size_t gws[3] = {1, 0, 0};
+    cl_event event;
+
+    clEnqueueNDRangeKernel(inOrderQueue, mockKernel.mockKernel, 1, nullptr, gws, nullptr, 0, nullptr, &event);
+
+    EXPECT_FALSE(mockedSubmissionsAggregator->peekCmdBufferList().peekIsEmpty());
+    auto cmdBuffer = mockedSubmissionsAggregator->peekCmdBufferList().peekHead();
+    EXPECT_NE(nullptr, cmdBuffer->pipeControlThatMayBeErasedLocation);
+    EXPECT_NE(nullptr, cmdBuffer->epiloguePipeControlLocation);
+
+    clReleaseCommandQueue(inOrderQueue);
+    clReleaseEvent(event);
+}
+
+HWTEST_F(EnqueueKernelTest, givenOutOfOrderCommandQueueWhenEnqueueKernelReturningEventIsMadeThenPipeControlPositionIsRecorded) {
     auto mockCsr = new MockCsrHw2<FamilyType>(pDevice->getHardwareInfo());
     mockCsr->overrideDispatchPolicy(DispatchMode::BatchedDispatch);
     pDevice->resetCommandStreamReceiver(mockCsr);
@@ -1480,6 +1504,9 @@ HWTEST_F(EnqueueKernelTest, givenOutOfOrderCommandQueueWhenEnqueueKernelReturnin
     auto mockedSubmissionsAggregator = new mockSubmissionsAggregator();
     mockCsr->overrideSubmissionAggregator(mockedSubmissionsAggregator);
 
+    const cl_queue_properties props[3] = {CL_QUEUE_PROPERTIES, CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE, 0};
+    auto inOrderQueue = clCreateCommandQueueWithProperties(context, pDevice, props, nullptr);
+
     MockKernelWithInternals mockKernel(*pDevice);
     size_t gws[3] = {1, 0, 0};
     cl_event event;