Remove deprecated code

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

opencl/source/command_queue/command_queue.h

@@ -36,7 +36,6 @@ class Kernel;
 class MemObj;
 class PerformanceCounters;
 struct CompletionStamp;
-struct DispatchGlobalsArgs;
 struct MultiDispatchInfo;
 
 enum class QueuePriority {
@@ -188,8 +187,6 @@ class CommandQueue : public BaseObject<_cl_command_queue> {
                                           const cl_event *eventWaitList, cl_event *event) = 0;
 
     virtual cl_int finish() = 0;
-    virtual cl_int enqueueInitDispatchGlobals(DispatchGlobalsArgs *dispatchGlobalsArgs, cl_uint numEventsInWaitList,
-                                              const cl_event *eventWaitList, cl_event *event) = 0;
 
     virtual cl_int flush() = 0;
 

opencl/source/command_queue/command_queue_hw.h

@@ -341,10 +341,6 @@ class CommandQueueHw : public CommandQueue {
                                   cl_event *event) override;
 
     cl_int finish() override;
-    cl_int enqueueInitDispatchGlobals(DispatchGlobalsArgs *dispatchGlobalsArgs,
-                                      cl_uint numEventsInWaitList,
-                                      const cl_event *eventWaitList,
-                                      cl_event *event) override;
     cl_int flush() override;
 
     template <uint32_t enqueueType>

opencl/source/command_queue/definitions/enqueue_init_dispatch_globals.h

@@ -1,23 +0,0 @@
-/*
- * Copyright (C) 2020-2021 Intel Corporation
- *
- * SPDX-License-Identifier: MIT
- *
- */
-
-#pragma once
-#include "opencl/source/command_queue/command_queue_hw.h"
-
-namespace NEO {
-
-struct DispatchGlobalsArgs {
-};
-
-template <typename GfxFamily>
-cl_int CommandQueueHw<GfxFamily>::enqueueInitDispatchGlobals(DispatchGlobalsArgs *dispatchGlobalsArgs,
-                                                             cl_uint numEventsInWaitList,
-                                                             const cl_event *eventWaitList,
-                                                             cl_event *event) {
-    return CL_INVALID_VALUE;
-}
-} // namespace NEO

opencl/source/command_queue/enqueue_common.h

@@ -794,7 +794,6 @@ CompletionStamp CommandQueueHw<GfxFamily>::enqueueNonBlocked(
     uint32_t numGrfRequired = GrfConfig::DefaultGrfNumber;
     auto specialPipelineSelectMode = false;
     Kernel *kernel = nullptr;
-    bool usePerDssBackedBuffer = false;
     bool auxTranslationRequired = false;
     bool useGlobalAtomics = false;
 
@@ -815,10 +814,6 @@ CompletionStamp CommandQueueHw<GfxFamily>::enqueueNonBlocked(
             anyUncacheableArgs = true;
         }
 
-        if (kernel->requiresPerDssBackedBuffer()) {
-            usePerDssBackedBuffer = true;
-        }
-
         if (kernel->getKernelInfo().kernelDescriptor.kernelAttributes.flags.useGlobalAtomics) {
             useGlobalAtomics = true;
         }
@@ -889,7 +884,7 @@ CompletionStamp CommandQueueHw<GfxFamily>::enqueueNonBlocked(
         implicitFlush,                                                                              //implicitFlush
         !eventBuilder.getEvent() || getGpgpuCommandStreamReceiver().isNTo1SubmissionModelEnabled(), //outOfOrderExecutionAllowed
         false,                                                                                      //epilogueRequired
-        usePerDssBackedBuffer,                                                                      //usePerDssBackedBuffer
+        false,                                                                                      //usePerDssBackedBuffer
         kernel->isSingleSubdevicePreferred(),                                                       //useSingleSubdevice
         useGlobalAtomics,                                                                           //useGlobalAtomics
         kernel->areMultipleSubDevicesInContext(),                                                   //areMultipleSubDevicesInContext

opencl/source/gen11/command_queue_gen11.cpp

@@ -13,8 +13,6 @@
 #include "opencl/source/command_queue/command_queue_hw_enabled_device_enqueue.inl"
 #include "opencl/source/command_queue/enqueue_resource_barrier.h"
 
-#include "enqueue_init_dispatch_globals.h"
-
 namespace NEO {
 
 typedef ICLFamily Family;

opencl/source/gen12lp/command_queue_gen12lp.cpp

@@ -12,7 +12,6 @@
 #include "opencl/source/command_queue/command_queue_hw_disabled_device_enqueue.inl"
 
 #include "command_queue_helpers_gen12lp.inl"
-#include "enqueue_init_dispatch_globals.h"
 
 namespace NEO {
 

opencl/source/gen8/command_queue_gen8.cpp

@@ -13,8 +13,6 @@
 #include "opencl/source/command_queue/command_queue_hw_disabled_device_enqueue.inl"
 #include "opencl/source/command_queue/enqueue_resource_barrier.h"
 
-#include "enqueue_init_dispatch_globals.h"
-
 namespace NEO {
 
 typedef BDWFamily Family;

opencl/source/gen9/command_queue_gen9.cpp

@@ -13,8 +13,6 @@
 #include "opencl/source/command_queue/command_queue_hw_enabled_device_enqueue.inl"
 #include "opencl/source/command_queue/enqueue_resource_barrier.h"
 
-#include "enqueue_init_dispatch_globals.h"
-
 namespace NEO {
 
 typedef SKLFamily Family;

opencl/source/helpers/task_information.cpp

@@ -247,7 +247,7 @@ CompletionStamp &CommandComputeKernel::submit(uint32_t taskLevel, bool terminate
         false,                                                                            //implicitFlush
         commandQueue.getGpgpuCommandStreamReceiver().isNTo1SubmissionModelEnabled(),      //outOfOrderExecutionAllowed
         false,                                                                            //epilogueRequired
-        kernel->requiresPerDssBackedBuffer(),                                             //usePerDssBackedBuffer
+        false,                                                                            //usePerDssBackedBuffer
         kernel->isSingleSubdevicePreferred(),                                             //useSingleSubdevice
         kernel->getKernelInfo().kernelDescriptor.kernelAttributes.flags.useGlobalAtomics, //useGlobalAtomics
         kernel->areMultipleSubDevicesInContext(),                                         //areMultipleSubDevicesInContext

opencl/source/kernel/kernel.h

@@ -384,7 +384,6 @@ class Kernel : public ReferenceTrackedObject<Kernel> {
         const bool kernelUsesLocalIds,
         const bool isCssUsed) const;
 
-    bool requiresPerDssBackedBuffer() const;
     bool requiresLimitedWorkgroupSize() const;
     bool isKernelDebugEnabled() const { return debugEnabled; }
     int32_t setAdditionalKernelExecInfoWithParam(uint32_t paramName, size_t paramValueSize, const void *paramValue);

opencl/source/kernel/kernel_extra.cpp

@@ -12,10 +12,6 @@
 
 namespace NEO {
 
-bool Kernel::requiresPerDssBackedBuffer() const {
-    return DebugManager.flags.ForcePerDssBackedBufferProgramming.get();
-}
-
 int32_t Kernel::setAdditionalKernelExecInfoWithParam(uint32_t paramName, size_t paramValueSize, const void *paramValue) {
     return CL_INVALID_VALUE;
 }

opencl/source/xe_hp_core/command_queue_xe_hp_core.cpp

@@ -11,8 +11,6 @@
 #include "opencl/source/command_queue/command_queue_hw.h"
 #include "opencl/source/command_queue/enqueue_resource_barrier.h"
 
-#include "enqueue_init_dispatch_globals.h"
-
 namespace NEO {
 
 using Family = XeHpFamily;

opencl/source/xe_hpc_core/command_queue_xe_hpc_core.cpp

@@ -11,8 +11,6 @@
 #include "opencl/source/command_queue/command_queue_hw.h"
 #include "opencl/source/command_queue/enqueue_resource_barrier.h"
 
-#include "enqueue_init_dispatch_globals.h"
-
 namespace NEO {
 
 using Family = XE_HPC_COREFamily;

opencl/source/xe_hpg_core/command_queue_xe_hpg_core.cpp

@@ -11,8 +11,6 @@
 #include "opencl/source/command_queue/command_queue_hw.h"
 #include "opencl/source/command_queue/enqueue_resource_barrier.h"
 
-#include "enqueue_init_dispatch_globals.h"
-
 namespace NEO {
 
 using Family = XE_HPG_COREFamily;

opencl/test/unit_test/command_queue/command_queue_hw_tests.cpp

@@ -1486,27 +1486,6 @@ HWTEST_F(CommandQueueHwTest, givenFinishWhenFlushBatchedSubmissionsFailsThenErro
     EXPECT_EQ(CL_OUT_OF_RESOURCES, errorCode);
 }
 
-HWTEST_F(CommandQueueHwTest, givenEmptyDispatchGlobalsArgsWhenEnqueueInitDispatchGlobalsCalledThenErrorIsReturned) {
-    EXPECT_EQ(CL_INVALID_VALUE, pCmdQ->enqueueInitDispatchGlobals(nullptr, 0, nullptr, nullptr));
-}
-
-HWTEST_F(CommandQueueHwTest, WhenForcePerDssBackedBufferProgrammingSetThenDispatchFlagsAreSetAccordingly) {
-    DebugManagerStateRestore restore;
-    DebugManager.flags.ForcePerDssBackedBufferProgramming = true;
-
-    MockKernelWithInternals mockKernelWithInternals(*pClDevice);
-    auto mockKernel = mockKernelWithInternals.mockKernel;
-    auto &csr = pDevice->getUltCommandStreamReceiver<FamilyType>();
-
-    size_t offset = 0;
-    size_t gws = 64;
-    size_t lws = 16;
-
-    cl_int status = pCmdQ->enqueueKernel(mockKernel, 1, &offset, &gws, &lws, 0, nullptr, nullptr);
-    EXPECT_EQ(CL_SUCCESS, status);
-    EXPECT_TRUE(csr.recordedDispatchFlags.usePerDssBackedBuffer);
-}
-
 template <bool ooq>
 struct CommandQueueHwBlitTest : ClDeviceFixture, ContextFixture, CommandQueueHwFixture, ::testing::Test {
     using ContextFixture::SetUp;

opencl/test/unit_test/command_queue/command_queue_tests.cpp

@@ -1251,18 +1251,6 @@ TEST(CommandQueue, GivenCommandQueueWhenCheckingIfIsCacheFlushCommandCalledThenF
     EXPECT_FALSE(isCommandCacheFlush);
 }
 
-TEST(CommandQueue, GivenCommandQueueWhenEnqueueInitDispatchGlobalsCalledThenSuccessReturned) {
-    MockContext context;
-    MockCommandQueue cmdQ(&context, nullptr, 0, false);
-
-    cl_int result = cmdQ.enqueueInitDispatchGlobals(
-        nullptr,
-        0,
-        nullptr,
-        nullptr);
-    EXPECT_EQ(CL_SUCCESS, result);
-}
-
 TEST(CommandQueue, givenBlitterOperationsSupportedWhenCreatingQueueThenTimestampPacketIsCreated) {
     DebugManagerStateRestore restore;
     DebugManager.flags.EnableTimestampPacket.set(0);

opencl/test/unit_test/command_stream/cache_flush_tests_dg2_and_later.cpp

@@ -6,7 +6,6 @@
  */
 
 #include "shared/source/helpers/l3_range.h"
-#include "shared/source/helpers/ray_tracing_helper.h"
 #include "shared/test/common/cmd_parse/hw_parse.h"
 #include "shared/test/common/helpers/debug_manager_state_restore.h"
 #include "shared/test/common/mocks/mock_csr.h"

opencl/test/unit_test/kernel/kernel_tests.cpp

@@ -3027,23 +3027,6 @@ TEST(KernelTest, givenKernelLocalIdGenerationByRuntimeFalseAndLocalIdsNotUsedWhe
     device->getMemoryManager()->freeGraphicsMemory(mockKernel.kernelInfo.getGraphicsAllocation());
 }
 
-TEST(KernelTest, givenKernelWhenForcePerDssBackedBufferProgrammingIsSetThenKernelRequiresPerDssBackedBuffer) {
-    DebugManagerStateRestore restore;
-    DebugManager.flags.ForcePerDssBackedBufferProgramming.set(true);
-
-    auto device = clUniquePtr(new MockClDevice(MockDevice::createWithNewExecutionEnvironment<MockDevice>(defaultHwInfo.get())));
-    MockKernelWithInternals kernel(*device);
-
-    EXPECT_TRUE(kernel.mockKernel->requiresPerDssBackedBuffer());
-}
-
-TEST(KernelTest, givenKernelWhenForcePerDssBackedBufferProgrammingIsNotSetThenKernelDoesntRequirePerDssBackedBuffer) {
-    auto device = clUniquePtr(new MockClDevice(MockDevice::createWithNewExecutionEnvironment<MockDevice>(defaultHwInfo.get())));
-    MockKernelWithInternals kernel(*device);
-
-    EXPECT_FALSE(kernel.mockKernel->requiresPerDssBackedBuffer());
-}
-
 TEST(KernelTest, whenKernelIsInitializedThenThreadArbitrationPolicyIsSetToDefaultValue) {
     UltClDeviceFactory deviceFactory{1, 0};
 

opencl/test/unit_test/mocks/mock_command_queue.h

@@ -198,9 +198,6 @@ class MockCommandQueue : public CommandQueue {
 
     cl_int finish() override { return CL_SUCCESS; }
 
-    cl_int enqueueInitDispatchGlobals(DispatchGlobalsArgs *dispatchGlobalsArgs, cl_uint numEventsInWaitList,
-                                      const cl_event *eventWaitList, cl_event *event) override { return CL_SUCCESS; }
-
     cl_int flush() override { return CL_SUCCESS; }
 
     bool obtainTimestampPacketForCacheFlush(bool isCacheFlushRequired) const override { return isCacheFlushRequired; }

opencl/test/unit_test/test_files/igdrcl.config

@@ -182,7 +182,6 @@ AllocateSharedAllocationsWithCpuAndGpuStorage = -1
 UseMaxSimdSizeToDeduceMaxWorkgroupSize = 0
 ReturnRawGpuTimestamps = 0
 EnableDeviceBasedTimestamps = 0
-ForcePerDssBackedBufferProgramming = 0
 MaxHwThreadsPercent = 0
 MinHwThreadsUnoccupied = 0
 LimitBlitterMaxWidth = -1

shared/source/debug_settings/debug_variables_base.inl

@@ -374,7 +374,6 @@ DECLARE_DEBUG_VARIABLE(int32_t, ForceHostPointerImport, -1, "-1: default, 0: dis
 DECLARE_DEBUG_VARIABLE(bool, UseMaxSimdSizeToDeduceMaxWorkgroupSize, false, "With this flag on, max workgroup size is deduced using SIMD32 instead of SIMD8, this causes the max wkg size to be 4 times bigger")
 DECLARE_DEBUG_VARIABLE(bool, ReturnRawGpuTimestamps, false, "Driver returns raw GPU tiemstamps instead of calculated ones.")
 DECLARE_DEBUG_VARIABLE(bool, EnableDeviceBasedTimestamps, false, "Driver returns timestamps in nanoseconds based on device timer.")
-DECLARE_DEBUG_VARIABLE(bool, ForcePerDssBackedBufferProgramming, false, "Always program per-DSS memory backed buffer in preamble")
 DECLARE_DEBUG_VARIABLE(bool, UseCommandBufferHeaderSizeForWddmQueueSubmission, true, "0: Page size (4096), 1: sizeof(COMMAND_BUFFER_HEADER)")
 DECLARE_DEBUG_VARIABLE(bool, DisableDeepBind, false, "Disable passing RTLD_DEEPBIND flag to all dlopen calls.")
 DECLARE_DEBUG_VARIABLE(bool, UseUmKmDataTranslator, false, "Use helper library for UMD<->KMD (WDDM) struct layout compatibility")