compute-runtime/shared/source/helpers/timestamp_packet.h

227 lines
10 KiB
C++

/*
* Copyright (C) 2018-2021 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
*/
#pragma once
#include "shared/source/command_container/command_encoder.h"
#include "shared/source/command_stream/command_stream_receiver.h"
#include "shared/source/command_stream/csr_deps.h"
#include "shared/source/helpers/aux_translation.h"
#include "shared/source/helpers/hw_helper.h"
#include "shared/source/helpers/non_copyable_or_moveable.h"
#include "shared/source/helpers/pipe_control_args.h"
#include "shared/source/helpers/string.h"
#include "shared/source/utilities/tag_allocator.h"
#include <cstdint>
#include <vector>
namespace NEO {
class CommandStreamReceiver;
class LinearStream;
namespace TimestampPacketSizeControl {
constexpr uint32_t preferredPacketCount = 16u;
}
#pragma pack(1)
template <typename TSize>
class TimestampPackets : public TagTypeBase {
protected:
struct Packet {
TSize contextStart = 1u;
TSize globalStart = 1u;
TSize contextEnd = 1u;
TSize globalEnd = 1u;
};
public:
static constexpr GraphicsAllocation::AllocationType getAllocationType() {
return GraphicsAllocation::AllocationType::TIMESTAMP_PACKET_TAG_BUFFER;
}
static constexpr TagNodeType getTagNodeType() { return TagNodeType::TimestampPacket; }
static constexpr size_t getSinglePacketSize() { return sizeof(Packet); }
void initialize() {
for (auto &packet : packets) {
packet.contextStart = 1u;
packet.globalStart = 1u;
packet.contextEnd = 1u;
packet.globalEnd = 1u;
}
}
void assignDataToAllTimestamps(uint32_t packetIndex, void *source) {
memcpy_s(&packets[packetIndex], sizeof(Packet), source, sizeof(Packet));
}
static constexpr size_t getGlobalStartOffset() { return offsetof(Packet, globalStart); }
static constexpr size_t getContextStartOffset() { return offsetof(Packet, contextStart); }
static constexpr size_t getContextEndOffset() { return offsetof(Packet, contextEnd); }
static constexpr size_t getGlobalEndOffset() { return offsetof(Packet, globalEnd); }
uint64_t getContextStartValue(uint32_t packetIndex) const { return static_cast<uint64_t>(packets[packetIndex].contextStart); }
uint64_t getGlobalStartValue(uint32_t packetIndex) const { return static_cast<uint64_t>(packets[packetIndex].globalStart); }
uint64_t getContextEndValue(uint32_t packetIndex) const { return static_cast<uint64_t>(packets[packetIndex].contextEnd); }
uint64_t getGlobalEndValue(uint32_t packetIndex) const { return static_cast<uint64_t>(packets[packetIndex].globalEnd); }
void const *getContextEndAddress(uint32_t packetIndex) const { return static_cast<void const *>(&packets[packetIndex].contextEnd); }
protected:
Packet packets[TimestampPacketSizeControl::preferredPacketCount];
};
#pragma pack()
static_assert(((4 * TimestampPacketSizeControl::preferredPacketCount) * sizeof(uint32_t)) == sizeof(TimestampPackets<uint32_t>),
"This structure is consumed by GPU and has to follow specific restrictions for padding and size");
class TimestampPacketContainer : public NonCopyableClass {
public:
TimestampPacketContainer() = default;
TimestampPacketContainer(TimestampPacketContainer &&) = default;
TimestampPacketContainer &operator=(TimestampPacketContainer &&) = default;
MOCKABLE_VIRTUAL ~TimestampPacketContainer();
const std::vector<TagNodeBase *> &peekNodes() const { return timestampPacketNodes; }
void add(TagNodeBase *timestampPacketNode);
void swapNodes(TimestampPacketContainer &timestampPacketContainer);
void assignAndIncrementNodesRefCounts(const TimestampPacketContainer &inputTimestampPacketContainer);
void makeResident(CommandStreamReceiver &commandStreamReceiver);
void moveNodesToNewContainer(TimestampPacketContainer &timestampPacketContainer);
protected:
std::vector<TagNodeBase *> timestampPacketNodes;
};
struct TimestampPacketDependencies : public NonCopyableClass {
TimestampPacketContainer cacheFlushNodes;
TimestampPacketContainer previousEnqueueNodes;
TimestampPacketContainer barrierNodes;
TimestampPacketContainer auxToNonAuxNodes;
TimestampPacketContainer nonAuxToAuxNodes;
void moveNodesToNewContainer(TimestampPacketContainer &timestampPacketContainer);
};
struct TimestampPacketHelper {
static uint64_t getContextEndGpuAddress(const TagNodeBase &timestampPacketNode) {
return timestampPacketNode.getGpuAddress() + timestampPacketNode.getContextEndOffset();
}
static uint64_t getContextStartGpuAddress(const TagNodeBase &timestampPacketNode) {
return timestampPacketNode.getGpuAddress() + timestampPacketNode.getContextStartOffset();
}
static uint64_t getGlobalEndGpuAddress(const TagNodeBase &timestampPacketNode) {
return timestampPacketNode.getGpuAddress() + timestampPacketNode.getGlobalEndOffset();
}
static uint64_t getGlobalStartGpuAddress(const TagNodeBase &timestampPacketNode) {
return timestampPacketNode.getGpuAddress() + timestampPacketNode.getGlobalStartOffset();
}
template <typename GfxFamily>
static void programSemaphore(LinearStream &cmdStream, TagNodeBase &timestampPacketNode) {
using COMPARE_OPERATION = typename GfxFamily::MI_SEMAPHORE_WAIT::COMPARE_OPERATION;
using MI_SEMAPHORE_WAIT = typename GfxFamily::MI_SEMAPHORE_WAIT;
auto compareAddress = getContextEndGpuAddress(timestampPacketNode);
for (uint32_t packetId = 0; packetId < timestampPacketNode.getPacketsUsed(); packetId++) {
uint64_t compareOffset = packetId * timestampPacketNode.getSinglePacketSize();
EncodeSempahore<GfxFamily>::addMiSemaphoreWaitCommand(cmdStream, compareAddress + compareOffset, 1, COMPARE_OPERATION::COMPARE_OPERATION_SAD_NOT_EQUAL_SDD);
}
}
template <typename GfxFamily>
static void programCsrDependenciesForTimestampPacketContainer(LinearStream &cmdStream, const CsrDependencies &csrDependencies) {
for (auto timestampPacketContainer : csrDependencies.timestampPacketContainer) {
for (auto &node : timestampPacketContainer->peekNodes()) {
TimestampPacketHelper::programSemaphore<GfxFamily>(cmdStream, *node);
}
}
}
template <typename GfxFamily>
static void programCsrDependenciesForForTaskCountContainer(LinearStream &cmdStream, const CsrDependencies &csrDependencies) {
auto taskCountContainer = csrDependencies.taskCountContainer;
for (auto &[taskCountPreviousRootDevice, tagAddressPreviousRootDevice] : taskCountContainer) {
using COMPARE_OPERATION = typename GfxFamily::MI_SEMAPHORE_WAIT::COMPARE_OPERATION;
using MI_SEMAPHORE_WAIT = typename GfxFamily::MI_SEMAPHORE_WAIT;
EncodeSempahore<GfxFamily>::addMiSemaphoreWaitCommand(cmdStream,
static_cast<uint64_t>(tagAddressPreviousRootDevice),
taskCountPreviousRootDevice,
COMPARE_OPERATION::COMPARE_OPERATION_SAD_GREATER_THAN_OR_EQUAL_SDD);
}
}
template <typename GfxFamily, AuxTranslationDirection auxTranslationDirection>
static void programSemaphoreForAuxTranslation(LinearStream &cmdStream,
const TimestampPacketDependencies *timestampPacketDependencies,
const HardwareInfo &hwInfo) {
auto &container = (auxTranslationDirection == AuxTranslationDirection::AuxToNonAux)
? timestampPacketDependencies->auxToNonAuxNodes
: timestampPacketDependencies->nonAuxToAuxNodes;
// cache flush after NDR, before NonAuxToAux
if (auxTranslationDirection == AuxTranslationDirection::NonAuxToAux && timestampPacketDependencies->cacheFlushNodes.peekNodes().size() > 0) {
UNRECOVERABLE_IF(timestampPacketDependencies->cacheFlushNodes.peekNodes().size() != 1);
auto cacheFlushTimestampPacketGpuAddress = getContextEndGpuAddress(*timestampPacketDependencies->cacheFlushNodes.peekNodes()[0]);
PipeControlArgs args;
args.dcFlushEnable = MemorySynchronizationCommands<GfxFamily>::getDcFlushEnable(true, hwInfo);
MemorySynchronizationCommands<GfxFamily>::addPipeControlAndProgramPostSyncOperation(
cmdStream, GfxFamily::PIPE_CONTROL::POST_SYNC_OPERATION::POST_SYNC_OPERATION_WRITE_IMMEDIATE_DATA,
cacheFlushTimestampPacketGpuAddress, 0, hwInfo, args);
}
for (auto &node : container.peekNodes()) {
TimestampPacketHelper::programSemaphore<GfxFamily>(cmdStream, *node);
}
}
template <typename GfxFamily, AuxTranslationDirection auxTranslationDirection>
static size_t getRequiredCmdStreamSizeForAuxTranslationNodeDependency(size_t count, const HardwareInfo &hwInfo, bool cacheFlushForBcsRequired) {
size_t size = count * TimestampPacketHelper::getRequiredCmdStreamSizeForNodeDependencyWithBlitEnqueue<GfxFamily>();
if (auxTranslationDirection == AuxTranslationDirection::NonAuxToAux && cacheFlushForBcsRequired) {
size += MemorySynchronizationCommands<GfxFamily>::getSizeForPipeControlWithPostSyncOperation(hwInfo);
}
return size;
}
template <typename GfxFamily>
static size_t getRequiredCmdStreamSizeForNodeDependencyWithBlitEnqueue() {
return sizeof(typename GfxFamily::MI_SEMAPHORE_WAIT);
}
template <typename GfxFamily>
static size_t getRequiredCmdStreamSizeForNodeDependency(TagNodeBase &timestampPacketNode) {
return (timestampPacketNode.getPacketsUsed() * sizeof(typename GfxFamily::MI_SEMAPHORE_WAIT));
}
template <typename GfxFamily>
static size_t getRequiredCmdStreamSize(const CsrDependencies &csrDependencies) {
size_t totalCommandsSize = 0;
for (auto timestampPacketContainer : csrDependencies.timestampPacketContainer) {
for (auto &node : timestampPacketContainer->peekNodes()) {
totalCommandsSize += getRequiredCmdStreamSizeForNodeDependency<GfxFamily>(*node);
}
}
return totalCommandsSize;
}
template <typename GfxFamily>
static size_t getRequiredCmdStreamSizeForTaskCountContainer(const CsrDependencies &csrDependencies) {
return csrDependencies.taskCountContainer.size() * sizeof(typename GfxFamily::MI_SEMAPHORE_WAIT);
}
};
} // namespace NEO