diff --git a/shared/source/command_container/command_encoder.h b/shared/source/command_container/command_encoder.h
index ab2b96ca8d..4c580e9641 100644
--- a/shared/source/command_container/command_encoder.h
+++ b/shared/source/command_container/command_encoder.h
@@ -43,6 +43,7 @@ struct PipelineSelectArgs;
 struct RootDeviceEnvironment;
 struct StateBaseAddressProperties;
 struct StateComputeModeProperties;
+struct ImplicitArgs;
 
 struct EncodeDispatchKernelArgs {
     uint64_t eventAddress = 0;
@@ -187,7 +188,7 @@ struct EncodeDispatchKernel {
     static void adjustWalkOrder(WalkerType &walkerCmd, uint32_t requiredWorkGroupOrder, const RootDeviceEnvironment &rootDeviceEnvironment);
 
     template <bool heaplessModeEnabled>
-    static void programInlineDataHeapless(uint8_t *inlineDataPtr, EncodeDispatchKernelArgs &args, CommandContainer &container, uint64_t offsetThreadData);
+    static void programInlineDataHeapless(uint8_t *inlineDataPtr, EncodeDispatchKernelArgs &args, CommandContainer &container, uint64_t offsetThreadData, uint64_t scratchPtr);
 
     static size_t getSizeRequiredDsh(const KernelDescriptor &kernelDescriptor, uint32_t iddCount);
     static size_t getSizeRequiredSsh(const KernelInfo &kernelInfo);
@@ -201,6 +202,10 @@ struct EncodeDispatchKernel {
 
     template <bool isHeapless>
     static void setScratchAddress(uint64_t &scratchAddress, uint32_t requiredScratchSlot0Size, uint32_t requiredScratchSlot1Size, IndirectHeap *ssh, CommandStreamReceiver &csr);
+    template <bool isHeapless>
+    static uint64_t getScratchAddressForImmediatePatching(CommandContainer &container, EncodeDispatchKernelArgs &args);
+    template <bool isHeapless>
+    static void patchScratchAddressInImplicitArgs(ImplicitArgs &implicitArgs, uint64_t scratchAddress, bool scratchPtrPatchingRequired);
 
     static size_t getInlineDataOffset(EncodeDispatchKernelArgs &args);
 };
diff --git a/shared/source/command_container/command_encoder_enablers.inl b/shared/source/command_container/command_encoder_enablers.inl
index 2c40fcea90..a7c38ce3d7 100644
--- a/shared/source/command_container/command_encoder_enablers.inl
+++ b/shared/source/command_container/command_encoder_enablers.inl
@@ -24,8 +24,10 @@ template void NEO::EncodeDispatchKernel<Family>::adjustWalkOrder<Family::Default
 template void NEO::EncodeDispatchKernel<Family>::programBarrierEnable<Family::INTERFACE_DESCRIPTOR_DATA>(Family::INTERFACE_DESCRIPTOR_DATA &interfaceDescriptor, uint32_t value, const HardwareInfo &hwInfo);
 template void NEO::EncodeDispatchKernel<Family>::setScratchAddress<false>(uint64_t &scratchAddress, uint32_t requiredScratchSlot0Size, uint32_t requiredScratchSlot1Size, IndirectHeap *ssh, CommandStreamReceiver &csr);
 template void NEO::EncodeDispatchKernel<Family>::setScratchAddress<true>(uint64_t &scratchAddress, uint32_t requiredScratchSlot0Size, uint32_t requiredScratchSlot1Size, IndirectHeap *ssh, CommandStreamReceiver &csr);
-template void NEO::EncodeDispatchKernel<Family>::programInlineDataHeapless<false>(uint8_t *inlineDataPtr, EncodeDispatchKernelArgs &args, CommandContainer &container, uint64_t offsetThreadData);
+template void NEO::EncodeDispatchKernel<Family>::programInlineDataHeapless<false>(uint8_t *inlineDataPtr, EncodeDispatchKernelArgs &args, CommandContainer &container, uint64_t offsetThreadData, uint64_t scratchPtr);
 template void NEO::EncodeDispatchKernel<Family>::encodeEuSchedulingPolicy<Family::INTERFACE_DESCRIPTOR_DATA>(Family::INTERFACE_DESCRIPTOR_DATA *pInterfaceDescriptor, const KernelDescriptor &kernelDesc, int32_t defaultPipelinedThreadArbitrationPolicy);
+template uint64_t NEO::EncodeDispatchKernel<Family>::getScratchAddressForImmediatePatching<false>(CommandContainer &container, EncodeDispatchKernelArgs &args);
+template void NEO::EncodeDispatchKernel<Family>::patchScratchAddressInImplicitArgs<false>(ImplicitArgs &implicitArgs, uint64_t scratchAddress, bool scratchPtrPatchingRequired);
 
 template struct NEO::EncodeStates<Family>;
 template struct NEO::EncodeMath<Family>;
diff --git a/shared/source/command_container/command_encoder_heap_addressing.inl b/shared/source/command_container/command_encoder_heap_addressing.inl
index 154f514820..21dcbbe9cd 100644
--- a/shared/source/command_container/command_encoder_heap_addressing.inl
+++ b/shared/source/command_container/command_encoder_heap_addressing.inl
@@ -10,7 +10,18 @@
 namespace NEO {
 template <typename Family>
 template <bool heaplessModeEnabled>
-void EncodeDispatchKernel<Family>::programInlineDataHeapless(uint8_t *inlineDataPtr, EncodeDispatchKernelArgs &args, CommandContainer &container, uint64_t offsetThreadData) {
+void EncodeDispatchKernel<Family>::programInlineDataHeapless(uint8_t *inlineDataPtr, EncodeDispatchKernelArgs &args, CommandContainer &container, uint64_t offsetThreadData, uint64_t scratchPtr) {
+}
+
+template <typename Family>
+template <bool heaplessModeEnabled>
+uint64_t EncodeDispatchKernel<Family>::getScratchAddressForImmediatePatching(CommandContainer &container, EncodeDispatchKernelArgs &args) {
+    return 0u;
+}
+
+template <typename Family>
+template <bool heaplessModeEnabled>
+void EncodeDispatchKernel<Family>::patchScratchAddressInImplicitArgs(ImplicitArgs &implicitArgs, uint64_t scratchAddress, bool scratchPtrRequired) {
 }
 
 } // namespace NEO
diff --git a/shared/source/command_container/command_encoder_xehp_and_later.inl b/shared/source/command_container/command_encoder_xehp_and_later.inl
index 3bbd851d96..8c7de7a692 100644
--- a/shared/source/command_container/command_encoder_xehp_and_later.inl
+++ b/shared/source/command_container/command_encoder_xehp_and_later.inl
@@ -239,6 +239,7 @@ void EncodeDispatchKernel<Family>::encode(CommandContainer &container, EncodeDis
         inlineDataProgramming = inlineDataProgrammingOffset != 0;
     }
 
+    auto scratchAddressForImmediatePatching = EncodeDispatchKernel<Family>::getScratchAddressForImmediatePatching<heaplessModeEnabled>(container, args);
     uint32_t sizeThreadData = sizePerThreadDataForWholeGroup + sizeCrossThreadData;
     uint32_t sizeForImplicitArgsPatching = NEO::ImplicitArgsHelper::getSizeForImplicitArgsPatching(pImplicitArgs, kernelDescriptor, !localIdsGenerationByRuntime, rootDeviceEnvironment);
     uint32_t iohRequiredSize = sizeThreadData + sizeForImplicitArgsPatching + args.reserveExtraPayloadSpace;
@@ -258,6 +259,8 @@ void EncodeDispatchKernel<Family>::encode(CommandContainer &container, EncodeDis
         if (pImplicitArgs) {
             offsetThreadData -= ImplicitArgs::getSize();
             pImplicitArgs->localIdTablePtr = heap->getGraphicsAllocation()->getGpuAddress() + heap->getUsed() - iohRequiredSize;
+            EncodeDispatchKernel<Family>::patchScratchAddressInImplicitArgs<heaplessModeEnabled>(*pImplicitArgs, scratchAddressForImmediatePatching, args.immediateScratchAddressPatching);
+
             ptr = NEO::ImplicitArgsHelper::patchImplicitArgs(ptr, *pImplicitArgs, kernelDescriptor, std::make_pair(localIdsGenerationByRuntime, requiredWorkgroupOrder), rootDeviceEnvironment);
         }
 
@@ -322,7 +325,7 @@ void EncodeDispatchKernel<Family>::encode(CommandContainer &container, EncodeDis
     }
 
     uint8_t *inlineData = reinterpret_cast<uint8_t *>(walkerCmd.getInlineDataPointer());
-    EncodeDispatchKernel<Family>::programInlineDataHeapless<heaplessModeEnabled>(inlineData, args, container, offsetThreadData);
+    EncodeDispatchKernel<Family>::programInlineDataHeapless<heaplessModeEnabled>(inlineData, args, container, offsetThreadData, scratchAddressForImmediatePatching);
 
     if constexpr (heaplessModeEnabled == false) {
         walkerCmd.setIndirectDataStartAddress(static_cast<uint32_t>(offsetThreadData));