Files
rocm-systems/rocclr/runtime/device/pal/palcounters.cpp
T
foreman c19449d03c P4 to Git Change 1415806 by jatang@jatang-opencl-hsa-stg3 on 2017/05/30 15:02:35
SWDEV-95919 - Correct PerfCounters BlockIndex for gfx9.

Affected files ...

... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palcounters.cpp#12 edit
... //depot/stg/opencl/drivers/opencl/tests/ocltst/module/runtime/OCLPerfCounters.cpp#37 edit
2017-05-30 15:22:13 -04:00

584 rivejä
16 KiB
C++

//
// Copyright (c) 2015 Advanced Micro Devices, Inc. All rights reserved.
//
#include "device/pal/palcounters.hpp"
#include "device/pal/palvirtual.hpp"
#include <array>
namespace pal {
PalCounterReference* PalCounterReference::Create(VirtualGPU& gpu) {
Pal::Result result;
// Create performance experiment
Pal::PerfExperimentCreateInfo createInfo = {};
createInfo.optionFlags.sampleInternalOperations = 1;
createInfo.optionFlags.cacheFlushOnCounterCollection = 1;
createInfo.optionFlags.sqShaderMask = 1;
createInfo.optionValues.sampleInternalOperations = true;
createInfo.optionValues.cacheFlushOnCounterCollection = true;
createInfo.optionValues.sqShaderMask = Pal::PerfShaderMaskCs;
size_t palExperSize = gpu.dev().iDev()->GetPerfExperimentSize(createInfo, &result);
if (result != Pal::Result::Success) {
return nullptr;
}
PalCounterReference* memRef = new (palExperSize) PalCounterReference(gpu);
if (memRef != nullptr) {
result = gpu.dev().iDev()->CreatePerfExperiment(createInfo, &memRef[1], &memRef->perfExp_);
if (result != Pal::Result::Success) {
memRef->release();
return nullptr;
}
}
return memRef;
}
PalCounterReference::~PalCounterReference() {
// The counter object is always associated with a particular queue,
// so we have to lock just this queue
amd::ScopedLock lock(gpu_.execution());
if (layout_ != nullptr) {
delete layout_;
}
if (memory_ != nullptr) {
delete memory_;
}
if (nullptr != iPerf()) {
iPerf()->Destroy();
}
}
uint64_t PalCounterReference::result(const std::vector<int>& index) {
if (index.size() == 0) {
// These are counters that have no corresponding PalSample created
return 0;
}
if (layout_ == nullptr) {
return 0;
}
uint64_t result = 0;
for (auto const& i : index) {
assert(i <= static_cast<int>(layout_->sampleCount) && "index not in range");
const Pal::GlobalSampleLayout& sample = layout_->samples[i];
if (sample.dataType == Pal::PerfCounterDataType::Uint32) {
uint32_t beginVal =
*reinterpret_cast<uint32_t*>(reinterpret_cast<char*>(cpuAddr_) + sample.beginValueOffset);
uint32_t endVal =
*reinterpret_cast<uint32_t*>(reinterpret_cast<char*>(cpuAddr_) + sample.endValueOffset);
result += (endVal - beginVal);
} else if (sample.dataType == Pal::PerfCounterDataType::Uint64) {
uint64_t beginVal =
*reinterpret_cast<uint64_t*>(reinterpret_cast<char*>(cpuAddr_) + sample.beginValueOffset);
uint64_t endVal =
*reinterpret_cast<uint64_t*>(reinterpret_cast<char*>(cpuAddr_) + sample.endValueOffset);
result += (endVal - beginVal);
} else {
assert(0 && "dataType should be either Uint32 or Uint64");
return 0;
}
}
return result;
}
bool PalCounterReference::finalize() {
Pal::Result result;
iPerf()->Finalize();
// Acquire GPU memory for the query from the pool and bind it.
Pal::GpuMemoryRequirements gpuMemReqs = {};
iPerf()->GetGpuMemoryRequirements(&gpuMemReqs);
memory_ = new Memory(gpu().dev(), amd::alignUp(gpuMemReqs.size, gpuMemReqs.alignment));
if (nullptr == memory_) {
return false;
}
if (!memory_->create(Resource::Remote)) {
return false;
}
cpuAddr_ = memory_->cpuMap(gpu_);
if (nullptr == cpuAddr_) {
return false;
}
gpu_.queue(gpu_.engineID_).addMemRef(memory_->iMem());
result = iPerf()->BindGpuMemory(memory_->iMem(), 0);
if (result == Pal::Result::Success) {
Pal::GlobalCounterLayout layout = {};
iPerf()->GetGlobalCounterLayout(&layout);
assert(layout.sampleCount == numExpCounters_);
size_t size = sizeof(Pal::GlobalCounterLayout) +
(sizeof(Pal::GlobalSampleLayout) * (layout.sampleCount - 1));
layout_ = reinterpret_cast<Pal::GlobalCounterLayout*>(new char[size]);
if (layout_ != nullptr) {
layout_->sampleCount = layout.sampleCount;
iPerf()->GetGlobalCounterLayout(layout_);
}
return true;
} else {
return false;
}
}
static const std::array<PCIndexSelect, 0x20> blockIdToIndexSelect = {{
PCIndexSelect::None, // CPF
PCIndexSelect::ShaderEngine, // IA
PCIndexSelect::ShaderEngine, // VGT
PCIndexSelect::ShaderEngine, // PA
PCIndexSelect::ShaderEngine, // SC
PCIndexSelect::ShaderEngine, // SPI
PCIndexSelect::ShaderEngine, // SQ
PCIndexSelect::ShaderEngine, // SX
PCIndexSelect::ShaderEngineAndInstance, // TA
PCIndexSelect::ShaderEngineAndInstance, // TD
PCIndexSelect::ShaderEngineAndInstance, // TCP
PCIndexSelect::Instance, // TCC
PCIndexSelect::Instance, // TCA
PCIndexSelect::ShaderEngineAndInstance, // DB
PCIndexSelect::ShaderEngineAndInstance, // CB
PCIndexSelect::None, // GDS
PCIndexSelect::None, // SRBM
PCIndexSelect::None, // GRBM
PCIndexSelect::None, // GRBMSE
PCIndexSelect::None, // RLC
PCIndexSelect::None, // DMA
PCIndexSelect::None, // MC
PCIndexSelect::None, // CPG
PCIndexSelect::None, // CPC
PCIndexSelect::None, // WD
PCIndexSelect::None, // TCS
PCIndexSelect::None, // UTC12
}};
// Converting from ORCA cmndefs.h to PAL palPerfExperiment.h
static const std::array<std::pair<int, int>, 83> ciBlockIdOrcaToPal = {{
{0x0E, 0}, // CB0
{0x0E, 1}, // CB1
{0x0E, 2}, // CB2
{0x0E, 3}, // CB3
{0x00, 0}, // CPF
{0x0D, 0}, // DB0
{0x0D, 1}, // DB1
{0x0D, 2}, // DB2
{0x0D, 3}, // DB3
{0x11, 0}, // GRBM
{0x12, 0}, // GRBMSE
{0x03, 0}, // PA_SU
{0x03, 0}, // PA_SC
{0x05, 0}, // SPI
{0x06, 0}, // SQ
{0x06, 0}, // SQ_ES
{0x06, 0}, // SQ_GS
{0x06, 0}, // SQ_VS
{0x06, 0}, // SQ_PS
{0x06, 0}, // SQ_LS
{0x06, 0}, // SQ_HS
{0x06, 0}, // SQ_CS
{0x07, 0}, // SX
{0x08, 0}, // TA0
{0x08, 1}, // TA1
{0x08, 2}, // TA2
{0x08, 3}, // TA3
{0x08, 4}, // TA4
{0x08, 5}, // TA5
{0x08, 6}, // TA6
{0x08, 7}, // TA7
{0x08, 8}, // TA8
{0x08, 9}, // TA9
{0x08, 0x0a}, // TA10
{0x0C, 0}, // TCA0
{0x0C, 1}, // TCA1
{0x0B, 0}, // TCC0
{0x0B, 1}, // TCC1
{0x0B, 2}, // TCC2
{0x0B, 3}, // TCC3
{0x0B, 4}, // TCC4
{0x0B, 5}, // TCC5
{0x0B, 6}, // TCC6
{0x0B, 7}, // TCC7
{0x0B, 8}, // TCC8
{0x0B, 9}, // TCC9
{0x0B, 0x0a}, // TCC10
{0x0B, 0x0b}, // TCC11
{0x0B, 0x0c}, // TCC12
{0x0B, 0x0d}, // TCC13
{0x0B, 0x0e}, // TCC14
{0x0B, 0x0f}, // TCC15
{0x09, 0}, // TD0
{0x09, 1}, // TD1
{0x09, 2}, // TD2
{0x09, 3}, // TD3
{0x09, 4}, // TD4
{0x09, 5}, // TD5
{0x09, 6}, // TD6
{0x09, 7}, // TD7
{0x09, 8}, // TD8
{0x09, 9}, // TD9
{0x09, 0x0a}, // TD10
{0x0A, 0}, // TCP0
{0x0A, 1}, // TCP1
{0x0A, 2}, // TCP2
{0x0A, 3}, // TCP3
{0x0A, 4}, // TCP4
{0x0A, 5}, // TCP5
{0x0A, 6}, // TCP6
{0x0A, 7}, // TCP7
{0x0A, 8}, // TCP8
{0x0A, 9}, // TCP9
{0x0A, 0x0a}, // TCP10
{0x0F, 0}, // GDS
{0x02, 0}, // VGT
{0x01, 0}, // IA
{0x15, 0}, // MC
{0x10, 0}, // SRBM
{0x19, 0}, // TCS
{0x18, 0}, // WD
{0x16, 0}, // CPG
{0x17, 0}, // CPC
}};
static const std::array<std::pair<int, int>, 97> viBlockIdOrcaToPal = {{
{0x0E, 0}, // CB0
{0x0E, 1}, // CB1
{0x0E, 2}, // CB2
{0x0E, 3}, // CB3
{0x00, 0}, // CPF
{0x0D, 0}, // DB0
{0x0D, 1}, // DB1
{0x0D, 2}, // DB2
{0x0D, 3}, // DB3
{0x11, 0}, // GRBM
{0x12, 0}, // GRBMSE
{0x03, 0}, // PA_SU
{0x03, 0}, // PA_SC
{0x05, 0}, // SPI
{0x06, 0}, // SQ
{0x06, 0}, // SQ_ES
{0x06, 0}, // SQ_GS
{0x06, 0}, // SQ_VS
{0x06, 0}, // SQ_PS
{0x06, 0}, // SQ_LS
{0x06, 0}, // SQ_HS
{0x06, 0}, // SQ_CS
{0x07, 0}, // SX
{0x08, 0}, // TA0
{0x08, 1}, // TA1
{0x08, 2}, // TA2
{0x08, 3}, // TA3
{0x08, 4}, // TA4
{0x08, 5}, // TA5
{0x08, 6}, // TA6
{0x08, 7}, // TA7
{0x08, 8}, // TA8
{0x08, 9}, // TA9
{0x08, 0x0a}, // TA10
{0x08, 0x0b}, // TA11
{0x08, 0x0c}, // TA12
{0x08, 0x0d}, // TA13
{0x08, 0x0e}, // TA14
{0x08, 0x0f}, // TA15
{0x0C, 0}, // TCA0
{0x0C, 1}, // TCA1
{0x0B, 0}, // TCC0
{0x0B, 1}, // TCC1
{0x0B, 2}, // TCC2
{0x0B, 3}, // TCC3
{0x0B, 4}, // TCC4
{0x0B, 5}, // TCC5
{0x0B, 6}, // TCC6
{0x0B, 7}, // TCC7
{0x0B, 8}, // TCC8
{0x0B, 9}, // TCC9
{0x0B, 0x0a}, // TCC10
{0x0B, 0x0b}, // TCC11
{0x0B, 0x0c}, // TCC12
{0x0B, 0x0d}, // TCC13
{0x0B, 0x0e}, // TCC14
{0x0B, 0x0f}, // TCC15
{0x09, 0}, // TD0
{0x09, 1}, // TD1
{0x09, 2}, // TD2
{0x09, 3}, // TD3
{0x09, 4}, // TD4
{0x09, 5}, // TD5
{0x09, 6}, // TD6
{0x09, 7}, // TD7
{0x09, 8}, // TD8
{0x09, 9}, // TD9
{0x09, 0x0a}, // TD10
{0x09, 0x0b}, // TD11
{0x09, 0x0c}, // TD12
{0x09, 0x0d}, // TD13
{0x09, 0x0e}, // TD14
{0x09, 0x0f}, // TD15
{0x0A, 0}, // TCP0
{0x0A, 1}, // TCP1
{0x0A, 2}, // TCP2
{0x0A, 3}, // TCP3
{0x0A, 4}, // TCP4
{0x0A, 5}, // TCP5
{0x0A, 6}, // TCP6
{0x0A, 7}, // TCP7
{0x0A, 8}, // TCP8
{0x0A, 9}, // TCP9
{0x0A, 0x0a}, // TCP10
{0x0A, 0x0b}, // TCP11
{0x0A, 0x0c}, // TCP12
{0x0A, 0x0d}, // TCP13
{0x0A, 0x0e}, // TCP14
{0x0A, 0x0f}, // TCP15
{0x0F, 0}, // GDS
{0x02, 0}, // VGT
{0x01, 0}, // IA
{0x15, 0}, // MC
{0x10, 0}, // SRBM
{0x18, 0}, // WD
{0x16, 0}, // CPG
{0x17, 0}, // CPC
}};
// The number of counters per block has been increased for gfx9 but this table may not reflect all
// of them
// as compute may not use all of them.
static const std::array<std::pair<int, int>, 97> gfx9BlockIdPal = {{
{0x0E, 0}, // CB0
{0x0E, 1}, // CB1
{0x0E, 2}, // CB2
{0x0E, 3}, // CB3
{0x00, 0}, // CPF
{0x0D, 0}, // DB0
{0x0D, 1}, // DB1
{0x0D, 2}, // DB2
{0x0D, 3}, // DB3
{0x11, 0}, // GRBM
{0x12, 0}, // GRBMSE
{0x03, 0}, // PA_SU
{0x03, 0}, // PA_SC
{0x05, 0}, // SPI
{0x06, 0}, // SQ
{0x06, 0}, // SQ_ES
{0x06, 0}, // SQ_GS
{0x06, 0}, // SQ_VS
{0x06, 0}, // SQ_PS
{0x06, 0}, // SQ_LS
{0x06, 0}, // SQ_HS
{0x06, 0}, // SQ_CS
{0x07, 0}, // SX
{0x08, 0}, // TA0
{0x08, 1}, // TA1
{0x08, 2}, // TA2
{0x08, 3}, // TA3
{0x08, 4}, // TA4
{0x08, 5}, // TA5
{0x08, 6}, // TA6
{0x08, 7}, // TA7
{0x08, 8}, // TA8
{0x08, 9}, // TA9
{0x08, 0x0a}, // TA10
{0x08, 0x0b}, // TA11
{0x08, 0x0c}, // TA12
{0x08, 0x0d}, // TA13
{0x08, 0x0e}, // TA14
{0x08, 0x0f}, // TA15
{0x0C, 0}, // TCA0
{0x0C, 1}, // TCA1
{0x0B, 0}, // TCC0
{0x0B, 1}, // TCC1
{0x0B, 2}, // TCC2
{0x0B, 3}, // TCC3
{0x0B, 4}, // TCC4
{0x0B, 5}, // TCC5
{0x0B, 6}, // TCC6
{0x0B, 7}, // TCC7
{0x0B, 8}, // TCC8
{0x0B, 9}, // TCC9
{0x0B, 0x0a}, // TCC10
{0x0B, 0x0b}, // TCC11
{0x0B, 0x0c}, // TCC12
{0x0B, 0x0d}, // TCC13
{0x0B, 0x0e}, // TCC14
{0x0B, 0x0f}, // TCC15
{0x09, 0}, // TD0
{0x09, 1}, // TD1
{0x09, 2}, // TD2
{0x09, 3}, // TD3
{0x09, 4}, // TD4
{0x09, 5}, // TD5
{0x09, 6}, // TD6
{0x09, 7}, // TD7
{0x09, 8}, // TD8
{0x09, 9}, // TD9
{0x09, 0x0a}, // TD10
{0x09, 0x0b}, // TD11
{0x09, 0x0c}, // TD12
{0x09, 0x0d}, // TD13
{0x09, 0x0e}, // TD14
{0x09, 0x0f}, // TD15
{0x0A, 0}, // TCP0
{0x0A, 1}, // TCP1
{0x0A, 2}, // TCP2
{0x0A, 3}, // TCP3
{0x0A, 4}, // TCP4
{0x0A, 5}, // TCP5
{0x0A, 6}, // TCP6
{0x0A, 7}, // TCP7
{0x0A, 8}, // TCP8
{0x0A, 9}, // TCP9
{0x0A, 0x0a}, // TCP10
{0x0A, 0x0b}, // TCP11
{0x0A, 0x0c}, // TCP12
{0x0A, 0x0d}, // TCP13
{0x0A, 0x0e}, // TCP14
{0x0A, 0x0f}, // TCP15
{0x0F, 0}, // GDS
{0x02, 0}, // VGT
{0x01, 0}, // IA
{0x15, 0}, // MC
{0x10, 0}, // SRBM
{0x18, 0}, // WD
{0x16, 0}, // CPG
{0x17, 0}, // CPC
}};
void PerfCounter::convertInfo() {
switch (dev().ipLevel()) {
case Pal::GfxIpLevel::GfxIp7:
if (info_.blockIndex_ < ciBlockIdOrcaToPal.size()) {
auto p = ciBlockIdOrcaToPal[info_.blockIndex_];
info_.blockIndex_ = std::get<0>(p);
info_.counterIndex_ = std::get<1>(p);
}
break;
case Pal::GfxIpLevel::GfxIp8:
if (info_.blockIndex_ < viBlockIdOrcaToPal.size()) {
auto p = viBlockIdOrcaToPal[info_.blockIndex_];
info_.blockIndex_ = std::get<0>(p);
info_.counterIndex_ = std::get<1>(p);
}
break;
case Pal::GfxIpLevel::GfxIp9:
if (info_.blockIndex_ < gfx9BlockIdPal.size()) {
auto p = gfx9BlockIdPal[info_.blockIndex_];
info_.blockIndex_ = std::get<0>(p);
info_.counterIndex_ = std::get<1>(p);
}
break;
default:
Unimplemented();
break;
}
assert(info_.blockIndex_ < blockIdToIndexSelect.size());
info_.indexSelect_ = blockIdToIndexSelect.at(info_.blockIndex_);
}
PerfCounter::~PerfCounter() {
if (palRef_ == nullptr) {
return;
}
// Release the counter reference object
palRef_->release();
}
bool PerfCounter::create() {
palRef_->retain();
// Initialize the counter
Pal::PerfCounterInfo counterInfo = {};
counterInfo.counterType = Pal::PerfCounterType::Global;
counterInfo.block = static_cast<Pal::GpuBlock>(info_.blockIndex_);
counterInfo.eventId = info_.eventIndex_;
Pal::PerfExperimentProperties perfExpProps;
Pal::Result result;
result = dev().iDev()->GetPerfExperimentProperties(&perfExpProps);
if (result != Pal::Result::Success) {
return false;
}
const auto& blockProps = perfExpProps.blocks[static_cast<uint32_t>(counterInfo.block)];
uint32_t counter_start, counter_step;
switch (info_.indexSelect_) {
case PCIndexSelect::ShaderEngine:
case PCIndexSelect::None:
counter_start = 0;
counter_step = 1;
break;
case PCIndexSelect::ShaderEngineAndInstance:
if (info_.counterIndex_ >=
dev().properties().gfxipProperties.shaderCore.maxCusPerShaderArray) {
return true;
}
counter_start = info_.counterIndex_;
counter_step = dev().properties().gfxipProperties.shaderCore.maxCusPerShaderArray;
break;
case PCIndexSelect::Instance:
counter_start = info_.counterIndex_;
counter_step = blockProps.instanceCount;
break;
default:
assert(0 && "Unknown indexSelect_");
return true;
}
for (uint32_t i = counter_start; i < blockProps.instanceCount; i += counter_step) {
counterInfo.instance = i;
result = iPerf()->AddCounter(counterInfo);
if (result == Pal::Result::Success) {
index_.push_back(palRef_->getPalCounterIndex());
} else {
// Get here when there's no HW PerfCounter matching the counterInfo
assert(0 && "AddCounter() failed");
}
}
return true;
}
uint64_t PerfCounter::getInfo(uint64_t infoType) const {
switch (infoType) {
case CL_PERFCOUNTER_GPU_BLOCK_INDEX: {
// Return the GPU block index
return info()->blockIndex_;
}
case CL_PERFCOUNTER_GPU_COUNTER_INDEX: {
// Return the GPU counter index
return info()->counterIndex_;
}
case CL_PERFCOUNTER_GPU_EVENT_INDEX: {
// Return the GPU event index
return info()->eventIndex_;
}
case CL_PERFCOUNTER_DATA: {
return palRef_->result(index_);
}
default:
LogError("Wrong PerfCounter::getInfo parameter");
}
return 0;
}
} // namespace pal