feature: support to query wait on memory data size

Related-To: NEO-8145

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>

level_zero/core/source/device/device_imp.cpp

@@ -937,6 +937,9 @@ ze_result_t DeviceImp::getProperties(ze_device_properties_t *pDeviceProperties)
                         rtasProperties->rtasFormat = ZE_RTAS_FORMAT_EXP_INVALID;
                     }
                 }
+            } else if (extendedProperties->stype == ZE_INTEL_STRUCTURE_TYPE_DEVICE_COMMAND_LIST_WAIT_ON_MEMORY_DATA_SIZE_EXP_DESC) {
+                auto cmdListWaitOnMemDataSize = reinterpret_cast<ze_intel_device_command_list_wait_on_memory_data_size_exp_desc_t *>(extendedProperties);
+                cmdListWaitOnMemDataSize->cmdListWaitOnMemoryDataSizeInBytes = l0GfxCoreHelper.getCmdListWaitOnMemoryDataSize();
             }
             extendedProperties = static_cast<ze_base_properties_t *>(extendedProperties->pNext);
         }

level_zero/core/source/driver/driver_handle_imp_helper.cpp

@@ -31,5 +31,7 @@ const std::vector<std::pair<std::string, uint32_t>> DriverHandleImp::extensionsS
 
     // Driver experimental extensions
     {ZE_INTEL_DEVICE_MODULE_DP_PROPERTIES_EXP_NAME, ZE_INTEL_DEVICE_MODULE_DP_PROPERTIES_EXP_VERSION_CURRENT},
-    {ZE_EVENT_POOL_COUNTER_BASED_EXP_NAME, ZE_EVENT_POOL_COUNTER_BASED_EXP_VERSION_CURRENT}};
+    {ZE_EVENT_POOL_COUNTER_BASED_EXP_NAME, ZE_EVENT_POOL_COUNTER_BASED_EXP_VERSION_CURRENT},
+    {ZE_INTEL_COMMAND_LIST_MEMORY_SYNC, ZE_INTEL_COMMAND_LIST_MEMORY_SYNC_EXP_VERSION_CURRENT},
+};
 } // namespace L0

level_zero/core/source/gfx_core_helpers/l0_gfx_core_helper.h

@@ -87,6 +87,7 @@ class L0GfxCoreHelper : public NEO::ApiGfxCoreHelper {
     virtual ze_rtas_format_exp_t getSupportedRTASFormat() const = 0;
     virtual bool platformSupportsImmediateComputeFlushTask() const = 0;
     virtual zet_debug_regset_type_intel_gpu_t getRegsetTypeForLargeGrfDetection() const = 0;
+    virtual uint32_t getCmdListWaitOnMemoryDataSize() const = 0;
 
   protected:
     L0GfxCoreHelper() = default;
@@ -125,6 +126,7 @@ class L0GfxCoreHelperHw : public L0GfxCoreHelper {
     ze_rtas_format_exp_t getSupportedRTASFormat() const override;
     bool platformSupportsImmediateComputeFlushTask() const override;
     zet_debug_regset_type_intel_gpu_t getRegsetTypeForLargeGrfDetection() const override;
+    uint32_t getCmdListWaitOnMemoryDataSize() const override;
 
   protected:
     L0GfxCoreHelperHw() = default;

level_zero/core/source/gfx_core_helpers/l0_gfx_core_helper_base.inl

@@ -44,4 +44,13 @@ zet_debug_regset_type_intel_gpu_t L0GfxCoreHelperHw<Family>::getRegsetTypeForLar
     return ZET_DEBUG_REGSET_TYPE_INVALID_INTEL_GPU;
 }
 
+template <typename Family>
+uint32_t L0GfxCoreHelperHw<Family>::getCmdListWaitOnMemoryDataSize() const {
+    if constexpr (Family::isQwordInOrderCounter) {
+        return sizeof(uint64_t);
+    } else {
+        return sizeof(uint32_t);
+    }
+}
+
 } // namespace L0

level_zero/core/test/unit_tests/sources/cmdlist/test_cmdlist_append_launch_kernel_3.cpp

@@ -870,7 +870,7 @@ bool InOrderCmdListTests::verifyInOrderDependency(GenCmdList::iterator &cmd, uin
     cmd++;
     return true;
 }
-HWTEST2_F(InOrderCmdListTests, givenDriverHandleWhenAskingForExtensionsThenReturnCounterBasedEventExtension, IsAtLeastSkl) {
+HWTEST2_F(InOrderCmdListTests, givenDriverHandleWhenAskingForExtensionsThenReturnCorrectVersions, IsAtLeastSkl) {
     uint32_t count = 0;
     ze_result_t res = driverHandle->getExtensionProperties(&count, nullptr);
     EXPECT_NE(0u, count);
@@ -885,6 +885,10 @@ HWTEST2_F(InOrderCmdListTests, givenDriverHandleWhenAskingForExtensionsThenRetur
     auto it = std::find_if(extensionProperties.begin(), extensionProperties.end(), [](const auto &extension) { return (strcmp(extension.name, ZE_EVENT_POOL_COUNTER_BASED_EXP_NAME) == 0); });
     EXPECT_NE(it, extensionProperties.end());
     EXPECT_EQ((*it).version, ZE_EVENT_POOL_COUNTER_BASED_EXP_VERSION_CURRENT);
+
+    it = std::find_if(extensionProperties.begin(), extensionProperties.end(), [](const auto &extension) { return (strcmp(extension.name, ZE_INTEL_COMMAND_LIST_MEMORY_SYNC) == 0); });
+    EXPECT_NE(it, extensionProperties.end());
+    EXPECT_EQ((*it).version, ZE_INTEL_COMMAND_LIST_MEMORY_SYNC_EXP_VERSION_CURRENT);
 }
 
 HWTEST2_F(InOrderCmdListTests, givenCmdListWhenAskingForQwordDataSizeThenReturnFalse, IsAtLeastSkl) {

level_zero/core/test/unit_tests/sources/device/test_l0_device.cpp

@@ -4868,6 +4868,19 @@ TEST_F(DeviceTest, GivenValidDeviceWhenQueryingKernelTimestampsProptertiesThenCo
     EXPECT_NE(0u, tsProps.flags & ZE_EVENT_QUERY_KERNEL_TIMESTAMPS_EXT_FLAG_SYNCHRONIZED);
 }
 
+TEST_F(DeviceTest, givenDeviceWhenQueryingCmdListMemWaitOnMemDataSizeThenReturnValueFromHelper) {
+    ze_device_properties_t devProps;
+    ze_intel_device_command_list_wait_on_memory_data_size_exp_desc_t sizeProps = {ZE_INTEL_STRUCTURE_TYPE_DEVICE_COMMAND_LIST_WAIT_ON_MEMORY_DATA_SIZE_EXP_DESC};
+
+    devProps.stype = ZE_STRUCTURE_TYPE_DEVICE_PROPERTIES;
+    devProps.pNext = &sizeProps;
+
+    auto &l0GfxCoreHelper = this->neoDevice->getRootDeviceEnvironment().getHelper<L0GfxCoreHelper>();
+
+    EXPECT_EQ(ZE_RESULT_SUCCESS, zeDeviceGetProperties(device, &devProps));
+    EXPECT_EQ(l0GfxCoreHelper.getCmdListWaitOnMemoryDataSize(), sizeProps.cmdListWaitOnMemoryDataSizeInBytes);
+}
+
 struct RTASDeviceTest : public ::testing::Test {
     void SetUp() override {
         debugManager.flags.CreateMultipleRootDevices.set(numRootDevices);

level_zero/core/test/unit_tests/sources/helper/l0_gfx_core_helper_tests.cpp

@@ -58,6 +58,15 @@ HWTEST_F(L0GfxCoreHelperTest, givenL0GfxCoreHelperWhenAskingForImageCompressionS
     EXPECT_FALSE(l0GfxCoreHelper.imageCompressionSupported(*NEO::defaultHwInfo));
 }
 
+HWTEST_F(L0GfxCoreHelperTest, givenL0GfxCoreHelperWhenAskingForCmdListWaitOnMemDataSizeThenReturnCorrectSize) {
+    MockExecutionEnvironment executionEnvironment;
+    auto &l0GfxCoreHelper = executionEnvironment.rootDeviceEnvironments[0]->getHelper<L0GfxCoreHelper>();
+
+    uint32_t expectedSize = FamilyType::isQwordInOrderCounter ? sizeof(uint64_t) : sizeof(uint32_t);
+
+    EXPECT_EQ(expectedSize, l0GfxCoreHelper.getCmdListWaitOnMemoryDataSize());
+}
+
 HWTEST_F(L0GfxCoreHelperTest, givenL0GfxCoreHelperWhenAskingForUsmCompressionSupportThenReturnFalse) {
     DebugManagerStateRestore restore;
 

level_zero/include/ze_intel_gpu.h

@@ -87,7 +87,7 @@ typedef enum _ze_intel_command_list_memory_sync_exp_version_t {
 ///     - Implementation must support ::ZE_intel_experimental_command_list_memory_sync extension
 ///     - May be passed to ze_device_properties_t through pNext.
 typedef struct _ze_intel_device_command_list_wait_on_memory_data_size_exp_desc_t {
-    ze_structure_type_t stype = ZE_INTEL_STRUCTURE_TYPE_DEVICE_COMMAND_LIST_WAIT_ON_MEMORY_DATA_SIZE_EXP_DESC; ///< [in] type of this structure
+    ze_structure_type_t stype;                   ///< [in] type of this structure
     const void *pNext;                           ///< [in][optional] must be null or a pointer to an extension-specific
                                                  ///< structure (i.e. contains stype and pNext).
     uint32_t cmdListWaitOnMemoryDataSizeInBytes; /// <out> Defines supported data size for zexCommandListAppendWaitOnMemory[64] API