Files
compute-runtime/runtime/command_stream/command_stream_receiver.cpp
Milczarek, Slawomir 7c42353c4c Add support for batched dispatch to userspace AUBs
This commit as aimed to add support for batched dispatch,
but doesn't make it the default mode for AubCSR yet.

Change-Id: I4dc366ec5f01adf2c4793009da2100ba0230c60a
2018-01-12 15:40:56 +01:00

217 lines
8.4 KiB
C++

/*
* Copyright (c) 2018, Intel Corporation
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included
* in all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*/
#include "runtime/built_ins/built_ins.h"
#include "runtime/command_stream/command_stream_receiver.h"
#include "runtime/command_stream/preemption.h"
#include "runtime/device/device.h"
#include "runtime/memory_manager/memory_manager.h"
#include "runtime/helpers/cache_policy.h"
#include "runtime/os_interface/os_interface.h"
#include "runtime/event/event.h"
#include "runtime/event/event_builder.h"
namespace OCLRT {
// Global table of CommandStreamReceiver factories for HW and tests
CommandStreamReceiverCreateFunc commandStreamReceiverFactory[2 * IGFX_MAX_CORE] = {};
CommandStreamReceiver::CommandStreamReceiver() {
latestSentStatelessMocsConfig = CacheSettings::unknownMocs;
submissionAggregator.reset(new SubmissionAggregator());
if (DebugManager.flags.CsrDispatchMode.get()) {
this->dispatchMode = (DispatchMode)DebugManager.flags.CsrDispatchMode.get();
}
flushStamp.reset(new FlushStampTracker(true));
}
CommandStreamReceiver::~CommandStreamReceiver() {
cleanupResources();
}
void CommandStreamReceiver::makeResident(GraphicsAllocation &gfxAllocation) {
auto submissionTaskCount = this->taskCount + 1;
if (gfxAllocation.residencyTaskCount < (int)submissionTaskCount) {
getMemoryManager()->pushAllocationForResidency(&gfxAllocation);
gfxAllocation.taskCount = submissionTaskCount;
if (gfxAllocation.residencyTaskCount == ObjectNotResident) {
this->totalMemoryUsed += gfxAllocation.getUnderlyingBufferSize();
}
}
gfxAllocation.residencyTaskCount = submissionTaskCount;
}
void CommandStreamReceiver::processEviction() {
getMemoryManager()->clearEvictionAllocations();
}
void CommandStreamReceiver::makeNonResident(GraphicsAllocation &gfxAllocation) {
if (gfxAllocation.residencyTaskCount != ObjectNotResident) {
makeCoherent(gfxAllocation.getUnderlyingBuffer(), gfxAllocation.getUnderlyingBufferSize());
getMemoryManager()->pushAllocationForEviction(&gfxAllocation);
}
gfxAllocation.residencyTaskCount = ObjectNotResident;
}
void CommandStreamReceiver::makeSurfacePackNonResident(ResidencyContainer *allocationsForResidency) {
auto &residencyAllocations = allocationsForResidency ? *allocationsForResidency : this->getMemoryManager()->getResidencyAllocations();
for (auto &surface : residencyAllocations) {
this->makeNonResident(*surface);
}
if (allocationsForResidency) {
residencyAllocations.clear();
} else {
this->getMemoryManager()->clearResidencyAllocations();
}
this->processEviction();
}
GraphicsAllocation *CommandStreamReceiver::createAllocationAndHandleResidency(const void *address, size_t size, bool addToDefferedDeleteList) {
GraphicsAllocation *graphicsAllocation = getMemoryManager()->allocateGraphicsMemory(size, address);
makeResident(*graphicsAllocation);
if (addToDefferedDeleteList) {
getMemoryManager()->storeAllocation(std::unique_ptr<GraphicsAllocation>(graphicsAllocation), TEMPORARY_ALLOCATION);
}
if (!graphicsAllocation->isL3Capable()) {
disableL3Cache = true;
}
return graphicsAllocation;
}
void CommandStreamReceiver::cleanAllocationList(uint32_t requiredTaskCount, uint32_t allocationType) {
auto address = getTagAddress();
if (address && requiredTaskCount != (unsigned int)-1) {
while (*address < requiredTaskCount)
;
}
getMemoryManager()->cleanAllocationList(requiredTaskCount, allocationType);
}
MemoryManager *CommandStreamReceiver::getMemoryManager() {
return memoryManager;
}
void CommandStreamReceiver::setMemoryManager(MemoryManager *mm) {
memoryManager = mm;
}
LinearStream &CommandStreamReceiver::getCS(size_t minRequiredSize) {
auto memoryManager = this->getMemoryManager();
DEBUG_BREAK_IF(nullptr == memoryManager);
if (commandStream.getAvailableSpace() < minRequiredSize) {
// Make sure we have enough room for a MI_BATCH_BUFFER_END and any padding.
// Currently reserving 64bytes (cacheline) which should be more than enough.
static const size_t sizeForSubmission = MemoryConstants::cacheLineSize;
minRequiredSize += sizeForSubmission;
// If not, allocate a new block. allocate full pages
minRequiredSize = alignUp(minRequiredSize, MemoryConstants::pageSize);
auto requiredSize = minRequiredSize + CSRequirements::csOverfetchSize;
auto allocation = memoryManager->obtainReusableAllocation(requiredSize).release();
if (!allocation) {
allocation = memoryManager->allocateGraphicsMemory(requiredSize, MemoryConstants::pageSize);
}
//pass current allocation to reusable list
if (commandStream.getBase()) {
memoryManager->storeAllocation(std::unique_ptr<GraphicsAllocation>(commandStream.getGraphicsAllocation()), REUSABLE_ALLOCATION);
}
commandStream.replaceBuffer(allocation->getUnderlyingBuffer(), minRequiredSize - sizeForSubmission);
commandStream.replaceGraphicsAllocation(allocation);
}
return commandStream;
}
void CommandStreamReceiver::cleanupResources() {
auto memoryManager = this->getMemoryManager();
if (!memoryManager)
return;
if (scratchAllocation) {
memoryManager->freeGraphicsMemory(scratchAllocation);
scratchAllocation = nullptr;
}
if (preemptionCsrAllocation) {
memoryManager->freeGraphicsMemory(preemptionCsrAllocation);
}
cleanAllocationList(-1, TEMPORARY_ALLOCATION);
cleanAllocationList(-1, REUSABLE_ALLOCATION);
if (commandStream.getBase()) {
memoryManager->freeGraphicsMemory(commandStream.getGraphicsAllocation());
commandStream.replaceGraphicsAllocation(nullptr);
commandStream.replaceBuffer(nullptr, 0);
}
}
bool CommandStreamReceiver::waitForCompletionWithTimeout(bool enableTimeout, int64_t timeoutMs, uint32_t taskCountToWait) {
std::chrono::high_resolution_clock::time_point time1, time2;
int64_t timeDiff = 0;
uint32_t latestSentTaskCount = this->latestFlushedTaskCount;
if (latestSentTaskCount < taskCountToWait) {
this->flushBatchedSubmissions();
}
time1 = std::chrono::high_resolution_clock::now();
while (*getTagAddress() < taskCountToWait && timeDiff <= timeoutMs) {
if (enableTimeout) {
time2 = std::chrono::high_resolution_clock::now();
timeDiff = std::chrono::duration_cast<std::chrono::milliseconds>(time2 - time1).count();
}
}
if (*getTagAddress() >= taskCountToWait) {
return true;
}
return false;
}
void CommandStreamReceiver::setTagAllocation(GraphicsAllocation *allocation) {
this->tagAllocation = allocation;
this->tagAddress = reinterpret_cast<uint32_t *>(allocation->getUnderlyingBuffer());
}
void CommandStreamReceiver::setRequiredScratchSize(uint32_t newRequiredScratchSize) {
if (newRequiredScratchSize > requiredScratchSize) {
requiredScratchSize = newRequiredScratchSize;
}
}
size_t CommandStreamReceiver::getInstructionHeapCmdStreamReceiverReservedSize() const {
return PreemptionHelper::getInstructionHeapSipKernelReservedSize(*memoryManager->device);
}
void CommandStreamReceiver::initializeInstructionHeapCmdStreamReceiverReservedBlock(LinearStream &ih) const {
return PreemptionHelper::initializeInstructionHeapSipKernelReservedBlock(ih, *memoryManager->device);
}
} // namespace OCLRT