intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-12-18 13:54:58 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix: extract common logic for filling default gt system info

Browse Source 
gt system info should be queried from KMD

Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
This commit is contained in:
Mateusz Jablonski
2024-06-25 09:37:08 +00:00
committed by Compute-Runtime-Automation
parent 1702526c40
commit 66d4d141e1
23 changed files with 82 additions and 246 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
level_zero/core/source/gfx_core_helpers/l0_gfx_core_helper_pvc_and_later.inl
View File

@@ -19,7 +19,7 @@ namespace L0 {
template <typename Family>
void L0GfxCoreHelperHw<Family>::getAttentionBitmaskForSingleThreads(const std::vector<EuThread::ThreadId> &threads, const NEO::HardwareInfo &hwInfo, std::unique_ptr<uint8_t[]> &bitmask, size_t &bitmaskSize) const {
const uint32_t numSubslicesPerSlice = hwInfo.gtSystemInfo.MaxSubSlicesSupported / hwInfo.gtSystemInfo.MaxSlicesSupported;
const uint32_t numEuPerSubslice = hwInfo.gtSystemInfo.MaxEuPerSubSlice;
const uint32_t numEuPerSubslice = std::min(hwInfo.gtSystemInfo.MaxEuPerSubSlice, 8u);
const uint32_t numThreadsPerEu = (hwInfo.gtSystemInfo.ThreadCount / hwInfo.gtSystemInfo.EUCount);
const uint32_t bytesPerEu = alignUp(numThreadsPerEu, 8) / 8;
const uint32_t threadsSizePerSlice = numSubslicesPerSlice * numEuPerSubslice * bytesPerEu;

2
level_zero/core/source/gfx_core_helpers/l0_gfx_core_helper_skl_to_icllp.inl
View File

@@ -23,7 +23,7 @@ void L0GfxCoreHelperHw<Family>::setAdditionalGroupProperty(ze_command_queue_grou
template <typename Family>
void L0GfxCoreHelperHw<Family>::getAttentionBitmaskForSingleThreads(const std::vector<EuThread::ThreadId> &threads, const NEO::HardwareInfo &hwInfo, std::unique_ptr<uint8_t[]> &bitmask, size_t &bitmaskSize) const {
const uint32_t numSubslicesPerSlice = hwInfo.gtSystemInfo.MaxSubSlicesSupported / hwInfo.gtSystemInfo.MaxSlicesSupported;
const uint32_t numEuPerSubslice = hwInfo.gtSystemInfo.MaxEuPerSubSlice;
const uint32_t numEuPerSubslice = std::min(hwInfo.gtSystemInfo.MaxEuPerSubSlice, 8u);
const uint32_t numThreadsPerEu = (hwInfo.gtSystemInfo.ThreadCount / hwInfo.gtSystemInfo.EUCount);
const uint32_t bytesPerEu = alignUp(numThreadsPerEu, 8) / 8;
const uint32_t threadsSizePerSlice = numSubslicesPerSlice * numEuPerSubslice * bytesPerEu;

9
level_zero/core/test/unit_tests/sources/helper/l0_gfx_core_helper_tests.cpp
View File

@@ -128,7 +128,12 @@ HWTEST_F(L0GfxCoreHelperTest, givenSliceSubsliceEuAndThreadIdsWhenGettingBitmask
};
auto hwInfo = *NEO::defaultHwInfo.get();
MockExecutionEnvironment executionEnvironment;
const auto threadsPerEu = (hwInfo.gtSystemInfo.ThreadCount / hwInfo.gtSystemInfo.EUCount);
hwInfo.gtSystemInfo.MaxEuPerSubSlice = 16u;
hwInfo.gtSystemInfo.EUCount = hwInfo.gtSystemInfo.MaxEuPerSubSlice * hwInfo.gtSystemInfo.SubSliceCount;
hwInfo.gtSystemInfo.ThreadCount = hwInfo.gtSystemInfo.EUCount * threadsPerEu;
MockExecutionEnvironment executionEnvironment(&hwInfo);
auto &l0GfxCoreHelper = executionEnvironment.rootDeviceEnvironments[0]->getHelper<L0GfxCoreHelper>();
std::unique_ptr<uint8_t[]> bitmask;
@@ -137,8 +142,6 @@ HWTEST_F(L0GfxCoreHelperTest, givenSliceSubsliceEuAndThreadIdsWhenGettingBitmask
uint32_t subslicesPerSlice = hwInfo.gtSystemInfo.MaxSubSlicesSupported / hwInfo.gtSystemInfo.MaxSlicesSupported;
uint32_t subslice = subslicesPerSlice > 1 ? subslicesPerSlice - 1 : 0;
const auto threadsPerEu = (hwInfo.gtSystemInfo.ThreadCount / hwInfo.gtSystemInfo.EUCount);
auto releaseHelper = executionEnvironment.rootDeviceEnvironments[0]->getReleaseHelper();
auto bytesPerEu = releaseHelper ? Math::divideAndRoundUp(releaseHelper->getNumThreadsPerEu(), 8u) : 1u;

5
level_zero/tools/test/unit_tests/sources/debug/debug_session_thread_tests.cpp
View File

@@ -1,10 +1,11 @@
/*
* Copyright (C) 2021-2023 Intel Corporation
* Copyright (C) 2021-2024 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
*/
#include "shared/source/helpers/gfx_core_helper.h"
#include "shared/test/common/helpers/default_hw_info.h"
#include "shared/test/common/helpers/gtest_helpers.h"
#include "shared/test/common/mocks/mock_device.h"
@@ -297,7 +298,7 @@ TEST(DebugSession, givenAllSlicesWhenGettingSingleThreadsThenCorrectThreadsAreRe
auto debugSession = std::make_unique<DebugSessionMock>(config, &deviceImp);
ze_device_thread_t physicalThread = {UINT32_MAX, 0, 0, 0};
const uint32_t numSlices = hwInfo.gtSystemInfo.MaxSlicesSupported;
const uint32_t numSlices = neoDevice->getGfxCoreHelper().getHighestEnabledSlice(hwInfo);
auto threads = debugSession->getSingleThreadsForDevice(0, physicalThread, hwInfo);

6
level_zero/tools/test/unit_tests/sources/debug/linux/prelim/test_debug_api_linux.cpp
View File

@@ -5707,7 +5707,9 @@ TEST_F(DebugApiLinuxTest, GivenSliceALLWhenCallingResumeThenSliceIdIsNotRemapped
ze_device_thread_t thread = {};
thread.slice = UINT32_MAX;
for (uint32_t i = 0; i < device->getHwInfo().gtSystemInfo.MaxSlicesSupported; i++) {
const auto &hwInfo = device->getHwInfo();
auto numSlices = neoDevice->getGfxCoreHelper().getHighestEnabledSlice(hwInfo);
for (uint32_t i = 0; i < numSlices; i++) {
ze_device_thread_t singleThread = thread;
singleThread.slice = i;
sessionMock->allThreads[EuThread::ThreadId(0, singleThread)]->stopThread(1u);
@@ -6306,7 +6308,7 @@ TEST_F(DebugApiLinuxAttentionTest, GivenSentInterruptWhenHandlingAttEventThenAtt
sessionMock->handleEvent(reinterpret_cast<prelim_drm_i915_debug_event *>(data));
EXPECT_EQ(2u, sessionMock->newlyStoppedThreads.size());
auto expectedThreadsToCheck = (hwInfo.capabilityTable.fusedEuEnabled && hwInfo.gtSystemInfo.MaxEuPerSubSlice != 8) ? 4u : 2u;
auto expectedThreadsToCheck = hwInfo.capabilityTable.fusedEuEnabled ? 4u : 2u;
EXPECT_EQ(expectedThreadsToCheck, sessionMock->addThreadToNewlyStoppedFromRaisedAttentionCallCount);
EXPECT_EQ(expectedThreadsToCheck, sessionMock->readSystemRoutineIdentFromMemoryCallCount);
EXPECT_EQ(0u, sessionMock->readSystemRoutineIdentCallCount);

4
level_zero/tools/test/unit_tests/sources/debug/linux/xe/test_debug_api_linux_xe.cpp
View File

@@ -2093,7 +2093,7 @@ TEST_F(DebugApiLinuxTestXe, GivenSentInterruptWhenHandlingAttEventThenAttBitsAre
sessionMock->handleEvent(reinterpret_cast<drm_xe_eudebug_event *>(data));
EXPECT_EQ(1u, sessionMock->newlyStoppedThreads.size());
auto expectedThreadsToCheck = (hwInfo.capabilityTable.fusedEuEnabled && hwInfo.gtSystemInfo.MaxEuPerSubSlice != 8) ? 2u : 1u;
auto expectedThreadsToCheck = hwInfo.capabilityTable.fusedEuEnabled ? 2u : 1u;
EXPECT_EQ(expectedThreadsToCheck, sessionMock->addThreadToNewlyStoppedFromRaisedAttentionCallCount);
EXPECT_EQ(expectedThreadsToCheck, sessionMock->readSystemRoutineIdentFromMemoryCallCount);
EXPECT_EQ(0u, sessionMock->readSystemRoutineIdentCallCount);
@@ -2302,4 +2302,4 @@ TEST(DebugSessionLinuxXeTest, GivenRootDebugSessionWhenCreateTileSessionCalledTh
}
} // namespace ult
} // namespace L0
} // namespace L0

24
shared/source/helpers/hw_info.cpp
View File

@@ -1,5 +1,5 @@
/*
* Copyright (C) 2018-2023 Intel Corporation
* Copyright (C) 2018-2024 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
@@ -10,6 +10,7 @@
#include "shared/source/command_stream/linear_stream.h"
#include "shared/source/debug_settings/debug_settings_manager.h"
#include "shared/source/helpers/gfx_core_helper.h"
#include "shared/source/release_helper/release_helper.h"
#include <algorithm>
@@ -128,4 +129,25 @@ aub_stream::EngineType getChosenEngineType(const HardwareInfo &hwInfo) {
? hwInfo.capabilityTable.defaultEngineType
: static_cast<aub_stream::EngineType>(debugManager.flags.NodeOrdinal.get());
}
void setupDefaultGtSysInfo(HardwareInfo *hwInfo, const ReleaseHelper *releaseHelper) {
GT_SYSTEM_INFO *gtSysInfo = &hwInfo->gtSystemInfo;
gtSysInfo->L3CacheSizeInKb = 1;
gtSysInfo->CCSInfo.IsValid = 1;
gtSysInfo->CCSInfo.NumberOfCCSEnabled = 1;
gtSysInfo->CCSInfo.Instances.CCSEnableMask = 0b1;
// non-zero values for unit tests
if (gtSysInfo->SliceCount == 0) {
gtSysInfo->SliceCount = 2;
gtSysInfo->SubSliceCount = 8;
gtSysInfo->DualSubSliceCount = gtSysInfo->SubSliceCount;
gtSysInfo->EUCount = 64;
gtSysInfo->MaxEuPerSubSlice = gtSysInfo->EUCount / gtSysInfo->SubSliceCount;
gtSysInfo->MaxSlicesSupported = gtSysInfo->SliceCount;
gtSysInfo->MaxSubSlicesSupported = gtSysInfo->SubSliceCount;
gtSysInfo->MaxDualSubSlicesSupported = gtSysInfo->DualSubSliceCount;
gtSysInfo->L3BankCount = 1;
}
gtSysInfo->ThreadCount = gtSysInfo->EUCount * releaseHelper->getNumThreadsPerEu();
}
} // namespace NEO

1
shared/source/helpers/hw_info.h
View File

@@ -169,5 +169,6 @@ bool getHwInfoForPlatformString(std::string &platform, const HardwareInfo *&hwIn
void setHwInfoValuesFromConfig(const uint64_t hwInfoConfig, HardwareInfo &hwInfoIn);
bool parseHwInfoConfigString(const std::string &hwInfoConfigStr, uint64_t &hwInfoConfig);
aub_stream::EngineType getChosenEngineType(const HardwareInfo &hwInfo);
void setupDefaultGtSysInfo(HardwareInfo *hwInfo, const ReleaseHelper *releaseHelper);
} // namespace NEO

5
shared/source/xe_hpc_core/hw_cmds_pvc.h
View File

@@ -17,11 +17,6 @@ struct PVC : public XeHpcCoreFamily {
static const HardwareInfo hwInfo;
static FeatureTable featureTable;
static WorkaroundTable workaroundTable;
// Initial non-zero values for unit tests
static const uint32_t maxEuPerSubslice = 8;
static const uint32_t maxSlicesSupported = 8;
static const uint32_t maxSubslicesSupported = 64;
static const uint32_t maxDualSubslicesSupported = 64;
static const RuntimeCapabilityTable capabilityTable;
static constexpr uint32_t numberOfpartsInTileForConcurrentKernels = 8u;

50
shared/source/xe_hpc_core/hw_info_pvc.cpp
View File

@@ -135,21 +135,8 @@ void PVC::adjustHardwareInfo(HardwareInfo *hwInfo) {
}
void PVC::setupHardwareInfoBase(HardwareInfo *hwInfo, bool setupFeatureTableAndWorkaroundTable, const ReleaseHelper *releaseHelper) {
GT_SYSTEM_INFO *gtSysInfo = &hwInfo->gtSystemInfo;
gtSysInfo->ThreadCount = gtSysInfo->EUCount * releaseHelper->getNumThreadsPerEu();
gtSysInfo->MaxFillRate = 128;
gtSysInfo->TotalVsThreads = 336;
gtSysInfo->TotalHsThreads = 336;
gtSysInfo->TotalDsThreads = 336;
gtSysInfo->TotalGsThreads = 336;
gtSysInfo->TotalPsThreadsWindowerRange = 64;
gtSysInfo->CsrSizeInMb = 8;
gtSysInfo->MaxEuPerSubSlice = PVC::maxEuPerSubslice;
gtSysInfo->MaxSlicesSupported = PVC::maxSlicesSupported;
gtSysInfo->MaxSubSlicesSupported = PVC::maxSubslicesSupported;
gtSysInfo->MaxDualSubSlicesSupported = PVC::maxDualSubslicesSupported;
gtSysInfo->IsL3HashModeEnabled = false;
gtSysInfo->IsDynamicallyPopulated = false;
setupDefaultGtSysInfo(hwInfo, releaseHelper);
setupHardwareInfoMultiTileBase(hwInfo, true);
PVC::adjustHardwareInfo(hwInfo);
@@ -181,39 +168,6 @@ const HardwareInfo PvcHwConfig::hwInfo = {
GT_SYSTEM_INFO PvcHwConfig::gtSystemInfo = {0};
void PvcHwConfig::setupHardwareInfo(HardwareInfo *hwInfo, bool setupFeatureTableAndWorkaroundTable, const ReleaseHelper *releaseHelper) {
PVC::setupHardwareInfoBase(hwInfo, setupFeatureTableAndWorkaroundTable, releaseHelper);
GT_SYSTEM_INFO *gtSysInfo = &hwInfo->gtSystemInfo;
gtSysInfo->CsrSizeInMb = 8;
gtSysInfo->IsL3HashModeEnabled = false;
gtSysInfo->IsDynamicallyPopulated = false;
// non-zero values for unit tests
if (gtSysInfo->SliceCount == 0) {
setupHardwareInfoMultiTileBase(hwInfo, true);
gtSysInfo->SliceCount = 2;
gtSysInfo->SubSliceCount = 8;
gtSysInfo->DualSubSliceCount = gtSysInfo->SubSliceCount;
gtSysInfo->EUCount = 40;
gtSysInfo->MaxEuPerSubSlice = gtSysInfo->EUCount / gtSysInfo->SubSliceCount;
gtSysInfo->MaxSlicesSupported = gtSysInfo->SliceCount;
gtSysInfo->MaxSubSlicesSupported = gtSysInfo->SubSliceCount;
gtSysInfo->L3CacheSizeInKb = 1;
gtSysInfo->L3BankCount = 1;
gtSysInfo->CCSInfo.IsValid = true;
gtSysInfo->CCSInfo.NumberOfCCSEnabled = 1;
gtSysInfo->CCSInfo.Instances.CCSEnableMask = 0b1;
hwInfo->featureTable.ftrBcsInfo = 1;
gtSysInfo->IsDynamicallyPopulated = true;
for (uint32_t slice = 0; slice < gtSysInfo->SliceCount; slice++) {
gtSysInfo->SliceInfo[slice].Enabled = true;
}
}
if (setupFeatureTableAndWorkaroundTable) {
PVC::setupFeatureAndWorkaroundTable(hwInfo);
}
};
const HardwareInfo PVC::hwInfo = PvcHwConfig::hwInfo;

5
shared/source/xe_hpg_core/hw_cmds_dg2.h
View File

@@ -25,11 +25,6 @@ struct DG2 : public XeHpgCoreFamily {
static const HardwareInfo hwInfo;
static FeatureTable featureTable;
static WorkaroundTable workaroundTable;
// Initial non-zero values for unit tests
static const uint32_t maxEuPerSubslice = 16;
static const uint32_t maxSlicesSupported = 8;
static const uint32_t maxSubslicesSupported = 32;
static const uint32_t maxDualSubslicesSupported = 32;
static const RuntimeCapabilityTable capabilityTable;
static void (*setupHardwareInfo)(HardwareInfo *hwInfo, bool setupFeatureTableAndWorkaroundTable, uint64_t hwInfoConfig, const ReleaseHelper *releaseHelper);
static void setupFeatureAndWorkaroundTable(HardwareInfo *hwInfo);

7
shared/source/xe_hpg_core/hw_cmds_mtl.h
View File

@@ -1,5 +1,5 @@
/*
* Copyright (C) 2022-2023 Intel Corporation
* Copyright (C) 2022-2024 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
@@ -16,11 +16,6 @@ struct MTL : public XeHpgCoreFamily {
static const HardwareInfo hwInfo;
static FeatureTable featureTable;
static WorkaroundTable workaroundTable;
// Initial non-zero values for unit tests
static const uint32_t maxEuPerSubslice = 16;
static const uint32_t maxSlicesSupported = 8;
static const uint32_t maxSubslicesSupported = 32;
static const uint32_t maxDualSubslicesSupported = 32;
static const RuntimeCapabilityTable capabilityTable;
static void (*setupHardwareInfo)(HardwareInfo *hwInfo, bool setupFeatureTableAndWorkaroundTable, uint64_t hwInfoConfig, const ReleaseHelper *releaseHelper);
static void setupFeatureAndWorkaroundTable(HardwareInfo *hwInfo);

40
shared/source/xe_hpg_core/hw_info_arl.cpp
View File

@@ -110,26 +110,14 @@ void ARL::setupFeatureAndWorkaroundTable(HardwareInfo *hwInfo) {
featureTable->flags.ftrCCSNode = true;
featureTable->flags.ftrCCSRing = true;
featureTable->flags.ftrTile64Optimization = true;
featureTable->ftrBcsInfo = 1;
workaroundTable->flags.wa4kAlignUVOffsetNV12LinearSurface = true;
workaroundTable->flags.waUntypedBufferCompression = true;
};
void ARL::setupHardwareInfoBase(HardwareInfo *hwInfo, bool setupFeatureTableAndWorkaroundTable, const ReleaseHelper *releaseHelper) {
GT_SYSTEM_INFO *gtSysInfo = &hwInfo->gtSystemInfo;
gtSysInfo->ThreadCount = gtSysInfo->EUCount * releaseHelper->getNumThreadsPerEu();
gtSysInfo->TotalPsThreadsWindowerRange = 64;
gtSysInfo->CsrSizeInMb = 8;
gtSysInfo->IsL3HashModeEnabled = false;
gtSysInfo->IsDynamicallyPopulated = false;
gtSysInfo->TotalVsThreads = 336;
gtSysInfo->TotalHsThreads = 336;
gtSysInfo->TotalDsThreads = 336;
gtSysInfo->TotalGsThreads = 336;
gtSysInfo->CCSInfo.IsValid = true;
gtSysInfo->CCSInfo.NumberOfCCSEnabled = 1;
gtSysInfo->CCSInfo.Instances.CCSEnableMask = 0b1;
setupDefaultGtSysInfo(hwInfo, releaseHelper);
if (setupFeatureTableAndWorkaroundTable) {
setupFeatureAndWorkaroundTable(hwInfo);
@@ -146,30 +134,6 @@ const HardwareInfo ArlHwConfig::hwInfo = {
GT_SYSTEM_INFO ArlHwConfig::gtSystemInfo = {0};
void ArlHwConfig::setupHardwareInfo(HardwareInfo *hwInfo, bool setupFeatureTableAndWorkaroundTable, const ReleaseHelper *releaseHelper) {
ARL::setupHardwareInfoBase(hwInfo, setupFeatureTableAndWorkaroundTable, releaseHelper);
GT_SYSTEM_INFO *gtSysInfo = &hwInfo->gtSystemInfo;
gtSysInfo->CsrSizeInMb = 8;
gtSysInfo->IsL3HashModeEnabled = false;
gtSysInfo->IsDynamicallyPopulated = false;
// non-zero values for unit tests
if (gtSysInfo->SliceCount == 0) {
gtSysInfo->SliceCount = 1;
gtSysInfo->SubSliceCount = 2;
gtSysInfo->DualSubSliceCount = gtSysInfo->SubSliceCount;
gtSysInfo->EUCount = 4;
gtSysInfo->MaxEuPerSubSlice = gtSysInfo->EUCount / gtSysInfo->SubSliceCount;
gtSysInfo->MaxSlicesSupported = gtSysInfo->SliceCount;
gtSysInfo->MaxSubSlicesSupported = gtSysInfo->SubSliceCount;
gtSysInfo->L3BankCount = 1;
hwInfo->featureTable.ftrBcsInfo = 1;
gtSysInfo->IsDynamicallyPopulated = true;
for (uint32_t slice = 0; slice < gtSysInfo->SliceCount; slice++) {
gtSysInfo->SliceInfo[slice].Enabled = true;
}
}
gtSysInfo->L3CacheSizeInKb = 1;
if (setupFeatureTableAndWorkaroundTable) {
ARL::setupFeatureAndWorkaroundTable(hwInfo);

48
shared/source/xe_hpg_core/hw_info_dg2.cpp
View File

@@ -119,25 +119,13 @@ void DG2::setupFeatureAndWorkaroundTable(HardwareInfo *hwInfo) {
featureTable->flags.ftrUnified3DMediaCompressionFormats = true;
featureTable->flags.ftrTile64Optimization = true;
hwInfo->featureTable.ftrBcsInfo = 1;
workaroundTable->flags.wa4kAlignUVOffsetNV12LinearSurface = true;
};
void DG2::setupHardwareInfoBase(HardwareInfo *hwInfo, bool setupFeatureTableAndWorkaroundTable, const ReleaseHelper *releaseHelper) {
GT_SYSTEM_INFO *gtSysInfo = &hwInfo->gtSystemInfo;
gtSysInfo->ThreadCount = gtSysInfo->EUCount * releaseHelper->getNumThreadsPerEu();
gtSysInfo->TotalVsThreads = 336;
gtSysInfo->TotalHsThreads = 336;
gtSysInfo->TotalDsThreads = 336;
gtSysInfo->TotalGsThreads = 336;
gtSysInfo->TotalPsThreadsWindowerRange = 64;
gtSysInfo->CsrSizeInMb = 8;
gtSysInfo->MaxEuPerSubSlice = DG2::maxEuPerSubslice;
gtSysInfo->MaxSlicesSupported = DG2::maxSlicesSupported;
gtSysInfo->MaxSubSlicesSupported = DG2::maxSubslicesSupported;
gtSysInfo->MaxDualSubSlicesSupported = DG2::maxDualSubslicesSupported;
gtSysInfo->IsL3HashModeEnabled = false;
gtSysInfo->IsDynamicallyPopulated = false;
setupDefaultGtSysInfo(hwInfo, releaseHelper);
if (setupFeatureTableAndWorkaroundTable) {
setupFeatureAndWorkaroundTable(hwInfo);
@@ -153,42 +141,12 @@ const HardwareInfo Dg2HwConfig::hwInfo = {
GT_SYSTEM_INFO Dg2HwConfig::gtSystemInfo = {0};
void Dg2HwConfig::setupHardwareInfo(HardwareInfo *hwInfo, bool setupFeatureTableAndWorkaroundTable, const ReleaseHelper *releaseHelper) {
GT_SYSTEM_INFO *gtSysInfo = &hwInfo->gtSystemInfo;
gtSysInfo->CsrSizeInMb = 8;
gtSysInfo->IsL3HashModeEnabled = false;
gtSysInfo->IsDynamicallyPopulated = false;
// non-zero values for unit tests
if (gtSysInfo->SliceCount == 0) {
gtSysInfo->SliceCount = 2;
gtSysInfo->SubSliceCount = 8;
gtSysInfo->DualSubSliceCount = gtSysInfo->SubSliceCount;
gtSysInfo->EUCount = 40;
gtSysInfo->MaxEuPerSubSlice = gtSysInfo->EUCount / gtSysInfo->SubSliceCount;
gtSysInfo->MaxSlicesSupported = gtSysInfo->SliceCount;
gtSysInfo->MaxSubSlicesSupported = gtSysInfo->SubSliceCount;
gtSysInfo->IsDynamicallyPopulated = true;
gtSysInfo->L3BankCount = 1;
}
gtSysInfo->L3CacheSizeInKb = 1;
gtSysInfo->CCSInfo.IsValid = true;
gtSysInfo->CCSInfo.NumberOfCCSEnabled = 1;
hwInfo->featureTable.ftrBcsInfo = 1;
for (uint32_t slice = 0; slice < gtSysInfo->SliceCount; slice++) {
gtSysInfo->SliceInfo[slice].Enabled = true;
}
if (setupFeatureTableAndWorkaroundTable) {
DG2::setupFeatureAndWorkaroundTable(hwInfo);
}
DG2::setupHardwareInfoBase(hwInfo, setupFeatureTableAndWorkaroundTable, releaseHelper);
};
const HardwareInfo DG2::hwInfo = Dg2HwConfig::hwInfo;
void setupDG2HardwareInfoImpl(HardwareInfo *hwInfo, bool setupFeatureTableAndWorkaroundTable, uint64_t hwInfoConfig, const ReleaseHelper *releaseHelper) {
DG2::setupHardwareInfoBase(hwInfo, setupFeatureTableAndWorkaroundTable, releaseHelper);
Dg2HwConfig::setupHardwareInfo(hwInfo, setupFeatureTableAndWorkaroundTable, releaseHelper);
}

44
shared/source/xe_hpg_core/hw_info_mtl.cpp
View File

@@ -111,30 +111,14 @@ void MTL::setupFeatureAndWorkaroundTable(HardwareInfo *hwInfo) {
featureTable->flags.ftrCCSNode = true;
featureTable->flags.ftrCCSRing = true;
featureTable->flags.ftrTile64Optimization = true;
featureTable->ftrBcsInfo = 1;
workaroundTable->flags.wa4kAlignUVOffsetNV12LinearSurface = true;
workaroundTable->flags.waUntypedBufferCompression = true;
};
void MTL::setupHardwareInfoBase(HardwareInfo *hwInfo, bool setupFeatureTableAndWorkaroundTable, const ReleaseHelper *releaseHelper) {
GT_SYSTEM_INFO *gtSysInfo = &hwInfo->gtSystemInfo;
gtSysInfo->ThreadCount = gtSysInfo->EUCount * releaseHelper->getNumThreadsPerEu();
gtSysInfo->TotalVsThreads = 336;
gtSysInfo->TotalHsThreads = 336;
gtSysInfo->TotalDsThreads = 336;
gtSysInfo->TotalGsThreads = 336;
gtSysInfo->TotalPsThreadsWindowerRange = 64;
gtSysInfo->CsrSizeInMb = 8;
gtSysInfo->MaxEuPerSubSlice = MTL::maxEuPerSubslice;
gtSysInfo->MaxSlicesSupported = MTL::maxSlicesSupported;
gtSysInfo->MaxSubSlicesSupported = MTL::maxSubslicesSupported;
gtSysInfo->MaxDualSubSlicesSupported = MTL::maxDualSubslicesSupported;
gtSysInfo->IsL3HashModeEnabled = false;
gtSysInfo->IsDynamicallyPopulated = false;
gtSysInfo->CCSInfo.IsValid = true;
gtSysInfo->CCSInfo.NumberOfCCSEnabled = 1;
gtSysInfo->CCSInfo.Instances.CCSEnableMask = 0b1;
setupDefaultGtSysInfo(hwInfo, releaseHelper);
if (setupFeatureTableAndWorkaroundTable) {
setupFeatureAndWorkaroundTable(hwInfo);
@@ -151,30 +135,6 @@ const HardwareInfo MtlHwConfig::hwInfo = {
GT_SYSTEM_INFO MtlHwConfig::gtSystemInfo = {0};
void MtlHwConfig::setupHardwareInfo(HardwareInfo *hwInfo, bool setupFeatureTableAndWorkaroundTable, const ReleaseHelper *releaseHelper) {
MTL::setupHardwareInfoBase(hwInfo, setupFeatureTableAndWorkaroundTable, releaseHelper);
GT_SYSTEM_INFO *gtSysInfo = &hwInfo->gtSystemInfo;
gtSysInfo->CsrSizeInMb = 8;
gtSysInfo->IsL3HashModeEnabled = false;
gtSysInfo->IsDynamicallyPopulated = false;
// non-zero values for unit tests
if (gtSysInfo->SliceCount == 0) {
gtSysInfo->SliceCount = 2;
gtSysInfo->SubSliceCount = 8;
gtSysInfo->DualSubSliceCount = gtSysInfo->SubSliceCount;
gtSysInfo->EUCount = 40;
gtSysInfo->MaxEuPerSubSlice = gtSysInfo->EUCount / gtSysInfo->SubSliceCount;
gtSysInfo->MaxSlicesSupported = gtSysInfo->SliceCount;
gtSysInfo->MaxSubSlicesSupported = gtSysInfo->SubSliceCount;
gtSysInfo->L3BankCount = 1;
hwInfo->featureTable.ftrBcsInfo = 1;
gtSysInfo->IsDynamicallyPopulated = true;
for (uint32_t slice = 0; slice < gtSysInfo->SliceCount; slice++) {
gtSysInfo->SliceInfo[slice].Enabled = true;
}
}
gtSysInfo->L3CacheSizeInKb = 1;
if (setupFeatureTableAndWorkaroundTable) {
MTL::setupFeatureAndWorkaroundTable(hwInfo);

5
shared/test/unit_test/device/neo_device_tests.cpp
View File

@@ -260,7 +260,7 @@ HWTEST2_F(DeviceTest, whenAllocateRTDispatchGlobalsIsCalledAndRTStackAllocationF
std::unique_ptr<NEO::MemoryManager> otherMemoryManager;
otherMemoryManager = std::make_unique<NEO::MockMemoryManagerWithCapacity>(*device->executionEnvironment);
static_cast<NEO::MockMemoryManagerWithCapacity &>(*otherMemoryManager).capacity = 50000000;
static_cast<NEO::MockMemoryManagerWithCapacity &>(*otherMemoryManager).capacity = MemoryConstants::pageSize;
device->executionEnvironment->memoryManager.swap(otherMemoryManager);
device->initializeRayTracing(5);
@@ -1105,7 +1105,8 @@ HWCMDTEST_F(IGFX_XE_HP_CORE, DeviceTest, givenHwInfoWhenRequestedComputeUnitsUse
const uint32_t multiplyFactor = productHelper.getThreadEuRatioForScratch(hwInfo) / 8u;
const uint32_t numThreadsPerEu = (hwInfo.gtSystemInfo.ThreadCount / hwInfo.gtSystemInfo.EUCount) * multiplyFactor;
uint32_t expectedValue = hwInfo.gtSystemInfo.MaxSubSlicesSupported * hwInfo.gtSystemInfo.MaxEuPerSubSlice * numThreadsPerEu;
uint32_t expectedValue = productHelper.computeMaxNeededSubSliceSpace(hwInfo) * hwInfo.gtSystemInfo.MaxEuPerSubSlice * numThreadsPerEu;
EXPECT_EQ(expectedValue, gfxCoreHelper.getComputeUnitsUsedForScratch(pDevice->getRootDeviceEnvironment()));
EXPECT_EQ(expectedValue, pDevice->getDeviceInfo().computeUnitsUsedForScratch);

18
shared/test/unit_test/xe_hpc_core/pvc/hw_info_tests_pvc.cpp
View File

@@ -1,5 +1,5 @@
/*
* Copyright (C) 2022-2023 Intel Corporation
* Copyright (C) 2022-2024 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
@@ -88,15 +88,15 @@ PVCTEST_F(PvcConfigHwInfoTests, givenPvcConfigWhenSetupHardwareInfoBaseThenGtSys
GT_SYSTEM_INFO &gtSystemInfo = hwInfo.gtSystemInfo;
PVC::setupHardwareInfoBase(&hwInfo, false, releaseHelper.get());
EXPECT_EQ(128u, gtSystemInfo.MaxFillRate);
EXPECT_EQ(336u, gtSystemInfo.TotalVsThreads);
EXPECT_EQ(336u, gtSystemInfo.TotalHsThreads);
EXPECT_EQ(336u, gtSystemInfo.TotalDsThreads);
EXPECT_EQ(336u, gtSystemInfo.TotalGsThreads);
EXPECT_EQ(64u, gtSystemInfo.TotalPsThreadsWindowerRange);
EXPECT_EQ(8u, gtSystemInfo.CsrSizeInMb);
EXPECT_EQ(0u, gtSystemInfo.MaxFillRate);
EXPECT_EQ(0u, gtSystemInfo.TotalVsThreads);
EXPECT_EQ(0u, gtSystemInfo.TotalHsThreads);
EXPECT_EQ(0u, gtSystemInfo.TotalDsThreads);
EXPECT_EQ(0u, gtSystemInfo.TotalGsThreads);
EXPECT_EQ(0u, gtSystemInfo.TotalPsThreadsWindowerRange);
EXPECT_EQ(0u, gtSystemInfo.CsrSizeInMb);
EXPECT_FALSE(gtSystemInfo.IsL3HashModeEnabled);
EXPECT_FALSE(gtSystemInfo.IsDynamicallyPopulated);
EXPECT_TRUE(gtSystemInfo.IsDynamicallyPopulated);
}
PVCTEST_F(PvcConfigHwInfoTests, givenPvcConfigWhenSetupMultiTileInfoBaseThenGtSystemInfoIsCorrect) {

1
shared/test/unit_test/xe_hpg_core/arl/linux/product_helper_tests_arl.cpp
View File

@@ -82,7 +82,6 @@ ARLTEST_F(ArlHwInfoLinux, whenSetupHardwareInfoThenGtSetupIsCorrect) {
EXPECT_GT(gtSystemInfo.SubSliceCount, 0u);
EXPECT_GT(gtSystemInfo.DualSubSliceCount, 0u);
EXPECT_GT_VAL(gtSystemInfo.L3CacheSizeInKb, 0u);
EXPECT_EQ(gtSystemInfo.CsrSizeInMb, 8u);
EXPECT_FALSE(gtSystemInfo.IsDynamicallyPopulated);
EXPECT_GT(gtSystemInfo.MaxDualSubSlicesSupported, 0u);
EXPECT_GT(gtSystemInfo.MaxSlicesSupported, 0u);

25
shared/test/unit_test/xe_hpg_core/dg2/product_config_helper_tests_dg2.cpp
View File

@@ -152,28 +152,19 @@ DG2TEST_F(ProductHelperTestDg2, whenGettingAubstreamProductFamilyThenProperEnumV
EXPECT_EQ(aub_stream::ProductFamily::Dg2, productHelper->getAubStreamProductFamily());
}
DG2TEST_F(ProductHelperTestDg2, givenDg2ConfigWhenSetupHardwareInfoBaseThenGtSystemInfoIsCorrect) {
HardwareInfo hwInfo = *defaultHwInfo;
GT_SYSTEM_INFO &gtSystemInfo = hwInfo.gtSystemInfo;
Dg2HwConfig::setupHardwareInfoBase(&hwInfo, false, releaseHelper);
EXPECT_EQ(336u, gtSystemInfo.TotalVsThreads);
EXPECT_EQ(336u, gtSystemInfo.TotalHsThreads);
EXPECT_EQ(336u, gtSystemInfo.TotalDsThreads);
EXPECT_EQ(336u, gtSystemInfo.TotalGsThreads);
EXPECT_EQ(64u, gtSystemInfo.TotalPsThreadsWindowerRange);
EXPECT_EQ(8u, gtSystemInfo.CsrSizeInMb);
EXPECT_FALSE(gtSystemInfo.IsL3HashModeEnabled);
EXPECT_FALSE(gtSystemInfo.IsDynamicallyPopulated);
}
DG2TEST_F(ProductHelperTestDg2, givenDg2ConfigWhenSetupHardwareInfoThenGtSystemInfoIsCorrect) {
HardwareInfo hwInfo = *defaultHwInfo;
GT_SYSTEM_INFO &gtSystemInfo = hwInfo.gtSystemInfo;
Dg2HwConfig::setupHardwareInfo(&hwInfo, false, releaseHelper);
EXPECT_EQ(8u, gtSystemInfo.CsrSizeInMb);
EXPECT_EQ(0u, gtSystemInfo.TotalVsThreads);
EXPECT_EQ(0u, gtSystemInfo.TotalHsThreads);
EXPECT_EQ(0u, gtSystemInfo.TotalDsThreads);
EXPECT_EQ(0u, gtSystemInfo.TotalGsThreads);
EXPECT_EQ(0u, gtSystemInfo.TotalPsThreadsWindowerRange);
EXPECT_EQ(0u, gtSystemInfo.CsrSizeInMb);
EXPECT_FALSE(gtSystemInfo.IsL3HashModeEnabled);
EXPECT_TRUE(gtSystemInfo.IsDynamicallyPopulated);
}
DG2TEST_F(ProductHelperTestDg2, givenDg2ProductHelperWhenIsInitBuiltinAsyncSupportedThenReturnTrue) {

19
shared/test/unit_test/xe_hpg_core/os_agnostic_product_helper_xe_lpg_tests.cpp
View File

@@ -44,14 +44,14 @@ HWTEST2_F(XeLpgHwInfoTests, whenSetupHardwareInfoBaseThenGtSystemInfoIsCorrect,
GT_SYSTEM_INFO &gtSystemInfo = hwInfo.gtSystemInfo;
hardwareInfoSetup[hwInfo.platform.eProductFamily](&hwInfo, compilerProductHelper->getHwInfoConfig(hwInfo), false, releaseHelper.get());
EXPECT_EQ(336u, gtSystemInfo.TotalVsThreads);
EXPECT_EQ(336u, gtSystemInfo.TotalHsThreads);
EXPECT_EQ(336u, gtSystemInfo.TotalDsThreads);
EXPECT_EQ(336u, gtSystemInfo.TotalGsThreads);
EXPECT_EQ(64u, gtSystemInfo.TotalPsThreadsWindowerRange);
EXPECT_EQ(8u, gtSystemInfo.CsrSizeInMb);
EXPECT_EQ(0u, gtSystemInfo.TotalVsThreads);
EXPECT_EQ(0u, gtSystemInfo.TotalHsThreads);
EXPECT_EQ(0u, gtSystemInfo.TotalDsThreads);
EXPECT_EQ(0u, gtSystemInfo.TotalGsThreads);
EXPECT_EQ(0u, gtSystemInfo.TotalPsThreadsWindowerRange);
EXPECT_EQ(0u, gtSystemInfo.CsrSizeInMb);
EXPECT_FALSE(gtSystemInfo.IsL3HashModeEnabled);
EXPECT_FALSE(gtSystemInfo.IsDynamicallyPopulated);
EXPECT_TRUE(gtSystemInfo.IsDynamicallyPopulated);
}
HWTEST2_F(XeLpgHwInfoTests, whenCheckDirectSubmissionEnginesThenProperValuesAreSetToTrue, IsXeLpg) {
@@ -173,11 +173,6 @@ HWTEST2_F(XeLpgHwInfoTests, GivenEmptyHwInfoForUnitTestsWhenSetupHardwareInfoIsC
EXPECT_GT_VAL(gtSystemInfo.CCSInfo.NumberOfCCSEnabled, 0u);
EXPECT_NE_VAL(hwInfoToSet.featureTable.ftrBcsInfo, 0u);
EXPECT_TRUE(gtSystemInfo.IsDynamicallyPopulated);
for (uint32_t i = 0; i < gtSystemInfo.SliceCount; i++) {
EXPECT_TRUE(gtSystemInfo.SliceInfo[i].Enabled);
}
}
HWTEST2_F(XeLpgProductHelperTests, whenGettingAubstreamProductFamilyThenProperEnumValueIsReturned, IsXeLpg) {

4
target_unit_tests/xe_hpc_core/pvc/enable_pvc_testing.cmake
View File

@@ -1,10 +1,10 @@
#
# Copyright (C) 2021-2023 Intel Corporation
# Copyright (C) 2021-2024 Intel Corporation
#
# SPDX-License-Identifier: MIT
#
if(TESTS_PVC)
set(unit_test_config "pvc/2/4/5/47") # non-zero values for unit tests
set(unit_test_config "pvc/2/4/8/47") # non-zero values for unit tests
include(${NEO_SOURCE_DIR}/cmake/run_ult_target.cmake)
endif()

2
target_unit_tests/xe_hpg_core/arl/enable_arl_testing.cmake
View File

@@ -5,6 +5,6 @@
#
if(TESTS_ARL)
set(unit_test_config "arl/2/4/5/0/0x7D41") # non-zero values for unit tests
set(unit_test_config "arl/2/4/8/0/0x7D41") # non-zero values for unit tests
include(${NEO_SOURCE_DIR}/cmake/run_ult_target.cmake)
endif()

2
target_unit_tests/xe_hpg_core/mtl/enable_mtl_testing.cmake
View File

@@ -5,6 +5,6 @@
#
if(TESTS_MTL)
set(unit_test_config "mtl/2/4/5/0/0x7D40") # non-zero values for unit tests
set(unit_test_config "mtl/2/4/8/0/0x7D40") # non-zero values for unit tests
include(${NEO_SOURCE_DIR}/cmake/run_ult_target.cmake)
endif()