feature: support copy function post-sync [4/n]

Added EncodePostSync to BlitPostSyncProperties, and
applyAdditionalBlitProperties to BlitCommandsHelper.

Related-To: NEO-13003
Signed-off-by: Young Jin Yoon <young.jin.yoon@intel.com>

shared/source/gen12lp/command_stream_receiver_hw_gen12lp.cpp

@@ -16,6 +16,7 @@ using Family = NEO::Gen12LpFamily;
 #include "shared/source/gmm_helper/gmm.h"
 #include "shared/source/helpers/address_patch.h"
 #include "shared/source/helpers/blit_commands_helper_base.inl"
+#include "shared/source/helpers/blit_commands_helper_from_gen12lp_to_xe3.inl"
 #include "shared/source/helpers/populate_factory.h"
 #include "shared/source/helpers/state_base_address_tgllp_and_later.inl"
 
@@ -332,6 +333,8 @@ bool BlitCommandsHelper<Family>::preBlitCommandWARequired() {
 template class CommandStreamReceiverHw<Family>;
 template struct BlitCommandsHelper<Family>;
 
+template void BlitCommandsHelper<Family>::applyAdditionalBlitProperties<typename Family::XY_BLOCK_COPY_BLT>(const BlitProperties &blitProperties, typename Family::XY_BLOCK_COPY_BLT &blitCmd, const RootDeviceEnvironment &rootDeviceEnvironment, bool last);
+
 const Family::GPGPU_WALKER Family::cmdInitGpgpuWalker = Family::GPGPU_WALKER::sInit();
 const Family::INTERFACE_DESCRIPTOR_DATA Family::cmdInitInterfaceDescriptorData = Family::INTERFACE_DESCRIPTOR_DATA::sInit();
 const Family::MEDIA_INTERFACE_DESCRIPTOR_LOAD Family::cmdInitMediaInterfaceDescriptorLoad = Family::MEDIA_INTERFACE_DESCRIPTOR_LOAD::sInit();

shared/source/helpers/CMakeLists.txt

@@ -234,6 +234,12 @@ if(SUPPORT_XE3_AND_LATER)
   )
 endif()
 
+if(SUPPORT_XE3_AND_BEFORE)
+  list(APPEND NEO_CORE_HELPERS
+       ${CMAKE_CURRENT_SOURCE_DIR}/blit_commands_helper_from_gen12lp_to_xe3.inl
+  )
+endif()
+
 set(NEO_CORE_HELPERS_GMM_CALLBACKS_WINDOWS
     ${CMAKE_CURRENT_SOURCE_DIR}/windows/gmm_callbacks.cpp
     ${CMAKE_CURRENT_SOURCE_DIR}/windows/gmm_callbacks.h

shared/source/helpers/blit_commands_helper.h

@@ -96,5 +96,8 @@ struct BlitCommandsHelper {
 
     static void encodeWa(LinearStream &cmdStream, const BlitProperties &blitProperties, uint32_t &latestSentBcsWaValue);
     static size_t getWaCmdsSize(const BlitPropertiesContainer &blitPropertiesContainer);
+
+    template <typename CommandType>
+    static void applyAdditionalBlitProperties(const BlitProperties &blitProperties, CommandType &cmd, const RootDeviceEnvironment &rootDeviceEnvironment, bool last);
 };
 } // namespace NEO

shared/source/helpers/blit_commands_helper_base.inl

@@ -209,10 +209,13 @@ void BlitCommandsHelper<GfxFamily>::dispatchBlitCommandsForBufferPerRow(const Bl
 
     appendColorDepth(blitProperties, bltCmd);
 
+    const bool useAdditionalBlitProperties = rootDeviceEnvironment.getHelper<ProductHelper>().useAdditionalBlitProperties();
+
     for (uint64_t slice = 0; slice < blitProperties.copySize.z; slice++) {
         for (uint64_t row = 0; row < blitProperties.copySize.y; row++) {
             uint64_t offset = 0;
             uint64_t sizeToBlit = blitProperties.copySize.x;
+            bool lastIteration = (slice == blitProperties.copySize.z - 1) && (row == blitProperties.copySize.y - 1);
             while (sizeToBlit != 0) {
                 if (sizeToBlit > maxWidth) {
                     // dispatch 2D blit: maxBlitWidth x (1 .. maxBlitHeight)
@@ -223,6 +226,8 @@ void BlitCommandsHelper<GfxFamily>::dispatchBlitCommandsForBufferPerRow(const Bl
                     width = sizeToBlit;
                     height = 1;
                 }
+                auto blitSize = width * height;
+                auto lastCommand = lastIteration && (sizeToBlit - blitSize == 0);
 
                 bltCmd.setDestinationX2CoordinateRight(static_cast<uint32_t>(width));
                 bltCmd.setDestinationY2CoordinateBottom(static_cast<uint32_t>(height));
@@ -237,6 +242,9 @@ void BlitCommandsHelper<GfxFamily>::dispatchBlitCommandsForBufferPerRow(const Bl
 
                 bltCmd.setDestinationBaseAddress(dstAddr);
                 bltCmd.setSourceBaseAddress(srcAddr);
+                if (useAdditionalBlitProperties && lastCommand) {
+                    applyAdditionalBlitProperties(blitProperties, bltCmd, rootDeviceEnvironment, lastCommand);
+                }
 
                 appendBlitCommandsForBuffer(blitProperties, bltCmd, rootDeviceEnvironment);
 
@@ -245,7 +253,6 @@ void BlitCommandsHelper<GfxFamily>::dispatchBlitCommandsForBufferPerRow(const Bl
 
                 dispatchPostBlitCommand(linearStream, rootDeviceEnvironment);
 
-                auto blitSize = width * height;
                 sizeToBlit -= blitSize;
                 offset += blitSize;
             }

shared/source/helpers/blit_commands_helper_from_gen12lp_to_xe3.inl

@@ -0,0 +1,18 @@
+/*
+ * Copyright (C) 2025 Intel Corporation
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ */
+
+#include "shared/source/helpers/blit_properties.h"
+
+namespace NEO {
+
+template <typename GfxFamily>
+template <typename CommandType>
+void BlitCommandsHelper<GfxFamily>::applyAdditionalBlitProperties(const BlitProperties &blitProperties,
+                                                                  CommandType &cmd, const RootDeviceEnvironment &rootDeviceEnvironment, bool last) {
+}
+
+} // namespace NEO

shared/source/helpers/blit_properties.h

@@ -6,6 +6,7 @@
  */
 
 #pragma once
+#include "shared/source/command_container/command_encoder.h"
 #include "shared/source/command_stream/csr_deps.h"
 #include "shared/source/gmm_helper/gmm_lib.h"
 #include "shared/source/helpers/aux_translation.h"
@@ -17,6 +18,7 @@
 
 namespace NEO {
 struct TimestampPacketDependencies;
+struct BlitSyncPropertiesExt;
 class TagNodeBase;
 class TimestampPacketContainer;
 class GraphicsAllocation;
@@ -31,13 +33,13 @@ enum class BlitSyncMode {
 };
 
 struct BlitSyncProperties {
+    EncodePostSyncArgs postSyncArgs{};
     TagNodeBase *outputTimestampPacket = nullptr;
     BlitSyncMode syncMode = BlitSyncMode::none;
     uint64_t deviceGpuWriteAddress = 0;
     uint64_t hostGpuWriteAddress = 0;
     uint64_t timestampGpuWriteAddress = 0;
     uint64_t writeValue = 0;
-
     bool isTimestampMode() const {
         return (syncMode == BlitSyncMode::timestamp) || (syncMode == BlitSyncMode::timestampAndImmediate);
     }

shared/source/xe2_hpg_core/command_stream_receiver_hw_xe2_hpg_core.cpp

@@ -15,6 +15,7 @@ using Family = NEO::Xe2HpgCoreFamily;
 #include "shared/source/command_stream/command_stream_receiver_hw_heap_addressing.inl"
 #include "shared/source/command_stream/command_stream_receiver_hw_xehp_and_later.inl"
 #include "shared/source/gmm_helper/gmm.h"
+#include "shared/source/helpers/blit_commands_helper_from_gen12lp_to_xe3.inl"
 #include "shared/source/helpers/blit_commands_helper_pvc_and_later.inl"
 #include "shared/source/helpers/blit_commands_helper_xe2_and_later.inl"
 #include "shared/source/helpers/blit_commands_helper_xehp_and_later.inl"
@@ -271,6 +272,9 @@ template void BlitCommandsHelper<Family>::appendColorDepth<typename Family::XY_B
 template void BlitCommandsHelper<Family>::appendColorDepth<typename Family::XY_COPY_BLT>(const BlitProperties &blitProperties, typename Family::XY_COPY_BLT &blitCmd);
 template void BlitCommandsHelper<Family>::appendBlitCommandsForBuffer<typename Family::XY_COPY_BLT>(const BlitProperties &blitProperties, typename Family::XY_COPY_BLT &blitCmd, const RootDeviceEnvironment &rootDeviceEnvironment);
 
+template void BlitCommandsHelper<Family>::applyAdditionalBlitProperties<typename Family::XY_COPY_BLT>(const BlitProperties &blitProperties, typename Family::XY_COPY_BLT &blitCmd, const RootDeviceEnvironment &rootDeviceEnvironment, bool last);
+template void BlitCommandsHelper<Family>::applyAdditionalBlitProperties<typename Family::XY_BLOCK_COPY_BLT>(const BlitProperties &blitProperties, typename Family::XY_BLOCK_COPY_BLT &blitCmd, const RootDeviceEnvironment &rootDeviceEnvironment, bool last);
+
 const Family::COMPUTE_WALKER Family::cmdInitGpgpuWalker = Family::COMPUTE_WALKER::sInit();
 const Family::CFE_STATE Family::cmdInitCfeState = Family::CFE_STATE::sInit();
 const Family::INTERFACE_DESCRIPTOR_DATA Family::cmdInitInterfaceDescriptorData = Family::INTERFACE_DESCRIPTOR_DATA::sInit();

shared/source/xe3_core/command_stream_receiver_hw_xe3_core.cpp

@@ -14,6 +14,7 @@ using Family = NEO::Xe3CoreFamily;
 #include "shared/source/command_stream/command_stream_receiver_hw_heap_addressing.inl"
 #include "shared/source/command_stream/command_stream_receiver_hw_xehp_and_later.inl"
 #include "shared/source/gmm_helper/gmm.h"
+#include "shared/source/helpers/blit_commands_helper_from_gen12lp_to_xe3.inl"
 #include "shared/source/helpers/blit_commands_helper_pvc_and_later.inl"
 #include "shared/source/helpers/blit_commands_helper_xe2_and_later.inl"
 #include "shared/source/helpers/blit_commands_helper_xehp_and_later.inl"
@@ -225,6 +226,9 @@ template void BlitCommandsHelper<Family>::appendColorDepth<typename Family::XY_B
 template void BlitCommandsHelper<Family>::appendColorDepth<typename Family::XY_COPY_BLT>(const BlitProperties &blitProperties, typename Family::XY_COPY_BLT &blitCmd);
 template void BlitCommandsHelper<Family>::appendBlitCommandsForBuffer<typename Family::XY_COPY_BLT>(const BlitProperties &blitProperties, typename Family::XY_COPY_BLT &blitCmd, const RootDeviceEnvironment &rootDeviceEnvironment);
 
+template void BlitCommandsHelper<Family>::applyAdditionalBlitProperties<typename Family::XY_COPY_BLT>(const BlitProperties &blitProperties, typename Family::XY_COPY_BLT &blitCmd, const RootDeviceEnvironment &rootDeviceEnvironment, bool last);
+template void BlitCommandsHelper<Family>::applyAdditionalBlitProperties<typename Family::XY_BLOCK_COPY_BLT>(const BlitProperties &blitProperties, typename Family::XY_BLOCK_COPY_BLT &blitCmd, const RootDeviceEnvironment &rootDeviceEnvironment, bool last);
+
 const Family::COMPUTE_WALKER Family::cmdInitGpgpuWalker = Family::COMPUTE_WALKER::sInit();
 const Family::CFE_STATE Family::cmdInitCfeState = Family::CFE_STATE::sInit();
 const Family::INTERFACE_DESCRIPTOR_DATA Family::cmdInitInterfaceDescriptorData = Family::INTERFACE_DESCRIPTOR_DATA::sInit();

shared/source/xe_hpc_core/command_stream_receiver_hw_xe_hpc_core.cpp

@@ -16,6 +16,7 @@ using Family = NEO::XeHpcCoreFamily;
 #include "shared/source/command_stream/command_stream_receiver_hw_heap_addressing.inl"
 #include "shared/source/command_stream/command_stream_receiver_hw_xehp_and_later.inl"
 #include "shared/source/gmm_helper/gmm.h"
+#include "shared/source/helpers/blit_commands_helper_from_gen12lp_to_xe3.inl"
 #include "shared/source/helpers/blit_commands_helper_pvc_and_later.inl"
 #include "shared/source/helpers/blit_commands_helper_xehp_and_later.inl"
 #include "shared/source/helpers/blit_properties.h"
@@ -273,6 +274,8 @@ template class CommandStreamReceiverHw<Family>;
 template struct BlitCommandsHelper<Family>;
 template void BlitCommandsHelper<Family>::appendBlitCommandsForBuffer<typename Family::XY_COPY_BLT>(const BlitProperties &blitProperties, typename Family::XY_COPY_BLT &blitCmd, const RootDeviceEnvironment &rootDeviceEnvironment);
 
+template void BlitCommandsHelper<Family>::applyAdditionalBlitProperties<typename Family::XY_COPY_BLT>(const BlitProperties &blitProperties, typename Family::XY_COPY_BLT &blitCmd, const RootDeviceEnvironment &rootDeviceEnvironment, bool last);
+
 const Family::COMPUTE_WALKER Family::cmdInitGpgpuWalker = Family::COMPUTE_WALKER::sInit();
 const Family::CFE_STATE Family::cmdInitCfeState = Family::CFE_STATE::sInit();
 const Family::INTERFACE_DESCRIPTOR_DATA Family::cmdInitInterfaceDescriptorData = Family::INTERFACE_DESCRIPTOR_DATA::sInit();

shared/source/xe_hpg_core/command_stream_receiver_hw_xe_hpg_core.cpp

@@ -14,6 +14,7 @@ using Family = NEO::XeHpgCoreFamily;
 #include "shared/source/command_stream/command_stream_receiver_hw_dg2_and_later.inl"
 #include "shared/source/command_stream/command_stream_receiver_hw_heap_addressing.inl"
 #include "shared/source/command_stream/command_stream_receiver_hw_xehp_and_later.inl"
+#include "shared/source/helpers/blit_commands_helper_from_gen12lp_to_xe3.inl"
 #include "shared/source/helpers/blit_commands_helper_xehp_and_later.inl"
 #include "shared/source/helpers/populate_factory.h"
 #include "shared/source/helpers/state_base_address_xehp_and_later.inl"
@@ -162,6 +163,8 @@ template struct BlitCommandsHelper<Family>;
 template void BlitCommandsHelper<Family>::appendColorDepth<typename Family::XY_BLOCK_COPY_BLT>(const BlitProperties &blitProperties, typename Family::XY_BLOCK_COPY_BLT &blitCmd);
 template void BlitCommandsHelper<Family>::appendBlitCommandsForBuffer<typename Family::XY_BLOCK_COPY_BLT>(const BlitProperties &blitProperties, typename Family::XY_BLOCK_COPY_BLT &blitCmd, const RootDeviceEnvironment &rootDeviceEnvironment);
 
+template void BlitCommandsHelper<Family>::applyAdditionalBlitProperties<typename Family::XY_BLOCK_COPY_BLT>(const BlitProperties &blitProperties, typename Family::XY_BLOCK_COPY_BLT &blitCmd, const RootDeviceEnvironment &rootDeviceEnvironment, bool last);
+
 const Family::COMPUTE_WALKER Family::cmdInitGpgpuWalker = Family::COMPUTE_WALKER::sInit();
 const Family::CFE_STATE Family::cmdInitCfeState = Family::CFE_STATE::sInit();
 const Family::INTERFACE_DESCRIPTOR_DATA Family::cmdInitInterfaceDescriptorData = Family::INTERFACE_DESCRIPTOR_DATA::sInit();

shared/test/unit_test/helpers/blit_commands_helper_tests.cpp

@@ -15,6 +15,7 @@
 #include "shared/test/common/fixtures/device_fixture.h"
 #include "shared/test/common/helpers/debug_manager_state_restore.h"
 #include "shared/test/common/helpers/default_hw_info.h"
+#include "shared/test/common/helpers/mock_product_helper_hw.h"
 #include "shared/test/common/mocks/mock_device.h"
 #include "shared/test/common/mocks/mock_graphics_allocation.h"
 #include "shared/test/common/mocks/mock_timestamp_container.h"
@@ -710,3 +711,79 @@ HWTEST_F(BlitTests, givenPlatformWhenCallingDispatchPreBlitCommandThenNoneMiFlus
         blitPropertiesContainer2, false, true, false, false, pDevice->getRootDeviceEnvironment());
     EXPECT_NE(estimatedSizeWithoutNode, estimatedSizeWithNode);
 }
+
+HWTEST2_F(BlitTests, givenPlatformWithBlitSyncPropertiesWithAndWithoutTimestampModeWhenCallingDispatchBlitCommandForBufferThenTheResultsAreTheSame, MatchAny) {
+    uint32_t src[] = {1, 2, 3, 4};
+    uint32_t dst[] = {4, 3, 2, 1};
+    uint32_t clear[] = {5, 6, 7, 8};
+    uint64_t srcGpuAddr = 0x12345;
+    uint64_t dstGpuAddr = 0x54321;
+    uint64_t clearGpuAddr = 0x5678;
+    std::unique_ptr<MockGraphicsAllocation> srcAlloc(new MockGraphicsAllocation(src, srcGpuAddr, sizeof(src)));
+    std::unique_ptr<MockGraphicsAllocation> dstAlloc(new MockGraphicsAllocation(dst, dstGpuAddr, sizeof(dst)));
+    std::unique_ptr<GraphicsAllocation> clearColorAllocation(new MockGraphicsAllocation(clear, clearGpuAddr, sizeof(clear)));
+
+    Vec3<size_t> srcOffsets{1, 0, 0};
+    Vec3<size_t> dstOffsets{1, 0, 0};
+    Vec3<size_t> copySize{(BlitterConstants::maxBlitWidth * BlitterConstants::maxBlitHeight) + 1, 2, 2};
+
+    size_t srcRowPitch = 0;
+    size_t srcSlicePitch = 0;
+
+    size_t dstRowPitch = 0;
+    size_t dstSlicePitch = 0;
+
+    auto blitProperties = NEO::BlitProperties::constructPropertiesForCopy(dstAlloc.get(), srcAlloc.get(),
+                                                                          dstOffsets, srcOffsets, copySize, srcRowPitch, srcSlicePitch,
+                                                                          dstRowPitch, dstSlicePitch, clearColorAllocation.get());
+
+    uint32_t streamBuffer[400] = {};
+    LinearStream stream(streamBuffer, sizeof(streamBuffer));
+    NEO::BlitCommandsHelper<FamilyType>::dispatchBlitCommandsForBufferPerRow(blitProperties, stream, pDevice->getRootDeviceEnvironmentRef());
+
+    // change it into timestamp mode
+    blitProperties.blitSyncProperties.syncMode = NEO::BlitSyncMode::timestamp;
+
+    uint32_t streamBuffer2[400] = {};
+    LinearStream stream2(streamBuffer2, sizeof(streamBuffer2));
+    NEO::BlitCommandsHelper<FamilyType>::dispatchBlitCommandsForBufferPerRow(blitProperties, stream2, pDevice->getRootDeviceEnvironmentRef());
+
+    EXPECT_EQ(stream.getUsed(), stream2.getUsed());
+    EXPECT_EQ(0, memcmp(ptrOffset(stream.getCpuBase(), 0), ptrOffset(stream2.getCpuBase(), 0), std::min(stream.getUsed(), stream2.getUsed())));
+
+    // change productHelper to return true
+    pDevice->getRootDeviceEnvironmentRef().productHelper.reset(new MockProductHelperHw<productFamily>);
+    auto *mockProductHelper = static_cast<MockProductHelperHw<productFamily> *>(pDevice->getRootDeviceEnvironmentRef().productHelper.get());
+    mockProductHelper->enableAdditionalBlitProperties = true;
+
+    uint32_t streamBuffer3[400] = {};
+    LinearStream stream3(streamBuffer3, sizeof(streamBuffer2));
+    NEO::BlitCommandsHelper<FamilyType>::dispatchBlitCommandsForBufferPerRow(blitProperties, stream3, pDevice->getRootDeviceEnvironmentRef());
+
+    EXPECT_EQ(stream.getUsed(), stream3.getUsed());
+    EXPECT_EQ(0, memcmp(ptrOffset(stream.getCpuBase(), 0), ptrOffset(stream3.getCpuBase(), 0), std::min(stream.getUsed(), stream3.getUsed())));
+}
+
+HWTEST_F(BlitTests, givenBlitPropertieswithImageOperationWhenCallingEstimateBlitCommandSizeThenBlockCopySizeIsReturned) {
+    Vec3<size_t> copySize{BlitterConstants::maxBlitWidth - 1, 1, 1};
+    NEO::CsrDependencies csrDependencies{};
+
+    size_t totalSize = NEO::BlitCommandsHelper<FamilyType>::estimateBlitCommandSize(copySize, csrDependencies, false, false, true, pDevice->getRootDeviceEnvironmentRef(), false, false);
+
+    size_t expectedSize = sizeof(typename FamilyType::XY_BLOCK_COPY_BLT);
+    expectedSize += NEO::BlitCommandsHelper<FamilyType>::estimatePostBlitCommandSize();
+    expectedSize += NEO::BlitCommandsHelper<FamilyType>::estimatePreBlitCommandSize();
+    EXPECT_EQ(expectedSize, totalSize);
+}
+
+using IsWithinGen12LPAndXe3Core = IsWithinGfxCore<IGFX_GEN12LP_CORE, IGFX_XE3_CORE>;
+HWTEST2_F(BlitTests, givenXyCopyBltCommandWhenApplyBlitPropertiesIsCalledThenNothingChanged, IsWithinGen12LPAndXe3Core) {
+    using XY_COPY_BLT = typename FamilyType::XY_COPY_BLT;
+    auto bltCmd = FamilyType::cmdInitXyCopyBlt;
+    auto bltCmdBefore = bltCmd;
+    BlitProperties properties = {};
+    NEO::BlitCommandsHelper<FamilyType>::applyAdditionalBlitProperties(properties, bltCmd, pDevice->getRootDeviceEnvironment(), false);
+    EXPECT_EQ(memcmp(&bltCmd, &bltCmdBefore, sizeof(XY_COPY_BLT)), 0);
+    NEO::BlitCommandsHelper<FamilyType>::applyAdditionalBlitProperties(properties, bltCmd, pDevice->getRootDeviceEnvironment(), true);
+    EXPECT_EQ(memcmp(&bltCmd, &bltCmdBefore, sizeof(XY_COPY_BLT)), 0);
+}