From b06f6058957d6674233673d08af3698a63daaf6b Mon Sep 17 00:00:00 2001
From: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
Date: Thu, 20 Jul 2023 12:02:24 +0000
Subject: [PATCH] fix: Correctly flush DC when reuse heap

Related-To: NEO-8067

Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
---
 .../command_stream_receiver_flush_task_3_tests.cpp          | 6 ++++++
 .../command_stream/command_stream_receiver_hw_base.inl      | 2 +-
 2 files changed, 7 insertions(+), 1 deletion(-)

diff --git a/opencl/test/unit_test/command_stream/command_stream_receiver_flush_task_3_tests.cpp b/opencl/test/unit_test/command_stream/command_stream_receiver_flush_task_3_tests.cpp
index d6fca6eef3..046e373b41 100644
--- a/opencl/test/unit_test/command_stream/command_stream_receiver_flush_task_3_tests.cpp
+++ b/opencl/test/unit_test/command_stream/command_stream_receiver_flush_task_3_tests.cpp
@@ -615,6 +615,7 @@ HWTEST_F(CommandStreamReceiverFlushTaskTests, givenUpdateTaskCountFromWaitEnable
     DebugManager.flags.UpdateTaskCountFromWait.set(3);
 
     auto mockCsr = new MockCsrHw2<FamilyType>(*pDevice->executionEnvironment, pDevice->getRootDeviceIndex(), pDevice->getDeviceBitfield());
+    mockCsr->overrideDispatchPolicy(DispatchMode::ImmediateDispatch);
     pDevice->resetCommandStreamReceiver(mockCsr);
     mockCsr->useNewResourceImplicitFlush = false;
     mockCsr->useGpuIdleImplicitFlush = false;
@@ -635,7 +636,12 @@ HWTEST_F(CommandStreamReceiverFlushTaskTests, givenUpdateTaskCountFromWaitEnable
 
     parseCommands<FamilyType>(commandStream);
     auto itorPipeControl = find<typename FamilyType::PIPE_CONTROL *>(cmdList.begin(), cmdList.end());
+    if (MemorySynchronizationCommands<FamilyType>::isBarrierWaRequired(pDevice->getRootDeviceEnvironment())) {
+        itorPipeControl++;
+    }
     EXPECT_NE(itorPipeControl, cmdList.end());
+    auto cmd = genCmdCast<typename FamilyType::PIPE_CONTROL *>(*itorPipeControl);
+    EXPECT_EQ(cmd->getDcFlushEnable(), mockCsr->getDcFlushSupport());
 }
 
 HWTEST_F(CommandStreamReceiverFlushTaskTests, givenCsrInDefaultModeWhenFlushTaskIsCalledThenFlushedTaskCountIsModifed) {
diff --git a/shared/source/command_stream/command_stream_receiver_hw_base.inl b/shared/source/command_stream/command_stream_receiver_hw_base.inl
index 7bf9090d1b..2d75688781 100644
--- a/shared/source/command_stream/command_stream_receiver_hw_base.inl
+++ b/shared/source/command_stream/command_stream_receiver_hw_base.inl
@@ -377,7 +377,7 @@ CompletionStamp CommandStreamReceiverHw<GfxFamily>::flushTask(
             // for BatchedSubmissions it will be nooped and only last ppc in batch will be emitted.
             levelClosed = true;
             // if we guard with ppc, flush dc as well to speed up completion latency
-            if (dispatchFlags.guardCommandBufferWithPipeControl) {
+            if (dispatchFlags.guardCommandBufferWithPipeControl || this->heapStorageRequiresRecyclingTag) {
                 dispatchFlags.dcFlush = this->dcFlushSupport;
             }
         }