Files
rocm-systems/projects/clr/rocclr/platform/commandqueue.hpp
T
SaleelK 340f3aa887 clr: Implement dynamic stream to HWq logic (#1958)
* clr: Implement dynamic stream to HW queue assignment

This change implements dynamic stream to hardware queue (HWq) mapping
with the following features:

* Queue depth heuristics with weights for optimal HWq assignment
* Make last used queue sticky for better locality
* Use pipe HWq to pipe mapping - gfx9 follows a round-robin queue to
  pipe mapping based on creation order (single process per device only,
  as pipe ID is statically assigned by runtime)
* More aggressive heuristic usage for better queue distribution
* Extend dynamic queues support for all stream priorities

Environment variables:
* DEBUG_HIP_DYNAMIC_QUEUE: 0 - disabled, 1 - Depth heuristics 2 -
  Depth+Pipe heuristics
* DEBUG_HIP_IGNORE_STREAM_PRIORITY=1: ignore priority stream creation

* clr: Clean up last_used_queue_
2026-01-23 10:40:54 -08:00

371 строка
12 KiB
C++

/* Copyright (c) 2012 - 2021 Advanced Micro Devices, Inc.
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. */
/*! \file commandqueue.hpp
* \brief Declarations CommandQueue object.
*
* \author Laurent Morichetti
* \date October 2008
*/
#ifndef COMMAND_QUEUE_HPP_
#define COMMAND_QUEUE_HPP_
#include "thread/thread.hpp"
#include "platform/object.hpp"
#include "platform/command.hpp"
/*! \brief Holds commands that will be executed on a specific device.
*
* \details A command queue is created on a specific device in
* a Context. A new virtual device will be instantiated from the given
* device and an execution environment (a thread) will be created to run
* the CommandQueue::loop() function.
*/
namespace amd {
class HostQueue;
class DeviceQueue;
class CommandQueue : public RuntimeObject {
public:
static constexpr uint RealTimeDisabled = 0xffffffff;
enum class Priority : uint { Low = 0, Normal, Medium, High };
struct Properties {
typedef cl_command_queue_properties value_type;
const value_type mask_;
value_type value_;
Properties(value_type mask, value_type value) : mask_(mask), value_(value & mask) {}
bool set(value_type bits) {
if ((mask_ & bits) != bits) {
return false;
}
value_ |= bits;
return true;
}
bool clear(value_type bits) {
if ((mask_ & bits) != bits) {
return false;
}
value_ &= ~bits;
return true;
}
bool test(value_type bits) const { return (value_ & bits) != 0; }
};
//! Return the context this command queue is part of.
Context& context() const { return context_(); }
//! Return the device for this command queue.
Device& device() const { return device_; }
//! Return the command queue properties.
Properties properties() const { return properties_; }
Properties& properties() { return properties_; }
//! Returns the base class object
CommandQueue* asCommandQueue() { return this; }
//! Returns TRUE if this queue requires a dedicated HW queue
bool isDedicatedQueue() const { return dedicated_queue_; }
virtual ~CommandQueue() {}
//! Returns TRUE if the object was successfully created
virtual bool create() = 0;
//! RTTI internal implementation
virtual ObjectType objectType() const { return ObjectTypeQueue; }
//! Rturns HostQueue object
virtual HostQueue* asHostQueue() { return NULL; }
//! Returns DeviceQueue object
virtual DeviceQueue* asDeviceQueue() { return NULL; }
//! Returns the number or requested real time CUs
uint rtCUs() const { return rtCUs_; }
//! Returns the queue priority
Priority priority() const { return priority_; }
//! Returns the CU mask array
const std::vector<uint32_t>& cuMask() const { return cuMask_; }
//! Returns the queue lock
Monitor& lock() { return queueLock_; }
protected:
//! CommandQueue constructor is protected
//! to keep the CommandQueue class as a virtual interface
CommandQueue(Context& context, //!< Context object
Device& device, //!< Device object
cl_command_queue_properties properties, //!< Queue properties
cl_command_queue_properties propMask, //!< Queue properties mask
uint rtCUs = RealTimeDisabled, //!< Avaialble real time compute units
Priority priority = Priority::Normal, //!< Queue priority
const std::vector<uint32_t>& cuMask = {}, //!< CU mask
bool dedicated_queue = false //!< TRUE if requires dedicated HW queue
)
: properties_(propMask, properties),
rtCUs_(rtCUs),
priority_(priority),
queueLock_(),
lastCmdLock_(),
device_(device),
context_(context),
cuMask_(cuMask),
dedicated_queue_(dedicated_queue) {}
Properties properties_; //!< Queue properties
uint rtCUs_; //!< The number of used RT compute units
Priority priority_; //!< Queue priority
Monitor queueLock_; //!< Lock protecting the queue
Monitor lastCmdLock_; //!< Lock protecting the last queued command
Device& device_; //!< The device
SharedReference<Context> context_; //!< The context of this command queue
const std::vector<uint32_t> cuMask_; //!< The CU mask
bool dedicated_queue_ = false; //!< TRUE if this queue requires a dedicated HW queue
private:
//! Disable copy constructor
CommandQueue(const CommandQueue&);
//! Disable assignment
CommandQueue& operator=(const CommandQueue&);
};
class HostQueue : public CommandQueue {
class Thread : public amd::Thread {
public:
//! True if this command queue thread is accepting commands.
volatile bool acceptingCommands_;
//! Create a new thread
Thread()
: amd::Thread("Command Queue Thread", CQ_THREAD_STACK_SIZE, !AMD_DIRECT_DISPATCH),
acceptingCommands_(false),
virtualDevice_(nullptr) {}
//! The command queue thread entry point.
void run(void* data) {
HostQueue* queue = static_cast<HostQueue*>(data);
virtualDevice_ = queue->device().createVirtualDevice(queue);
if (virtualDevice_ != nullptr) {
queue->loop(virtualDevice_);
Release();
} else {
acceptingCommands_ = false;
queue->flush();
}
}
void Init(HostQueue* queue) {
virtualDevice_ = queue->device().createVirtualDevice(queue);
if (virtualDevice_ != nullptr) {
acceptingCommands_ = true;
}
}
void Release() const { virtualDevice_->release(); }
//! Get virtual device for the current thread
device::VirtualDevice* vdev() const { return virtualDevice_; }
private:
device::VirtualDevice* virtualDevice_; //!< Virtual device for this thread
} thread_; //!< The command queue thread instance.
private:
ConcurrentLinkedQueue<Command*> queue_; //!< The queue.
Command* lastEnqueueCommand_; //!< The last submitted command
//! Await commands and execute them as they become ready.
void loop(device::VirtualDevice* virtualDevice);
protected:
virtual bool terminate();
public:
/*! \brief Construct a new host queue.
*
* \note A new virtual device instance will be created from the
* given device.
*/
HostQueue(Context& context, Device& device, cl_command_queue_properties properties,
uint queueRTCUs = 0, Priority priority = Priority::Normal,
const std::vector<uint32_t>& cuMask = {}, bool dedicated_queue = false);
//! Returns TRUE if this command queue can accept commands.
virtual bool create() { return thread_.acceptingCommands_; }
//! Append the given command to the queue.
void append(Command& command);
//! Return the thread object running the command loop.
const Thread& thread() const { return thread_; }
//! Signal to start processing the commands in the queue.
void flush() {
ScopedLock sl(queueLock_);
queueLock_.notify();
}
//! Finish all queued commands
void finish(bool cpu_wait = false);
//! Wait until finish of one command
void finishCommand(Command* command);
//! Check if hostQueue empty snapshot
bool isEmpty();
//! Get virtual device for the current command queue
device::VirtualDevice* vdev() const { return thread_.vdev(); }
//! Return the current queue as the HostQueue
virtual HostQueue* asHostQueue() { return this; }
//! Get last enqueued command
Command* getLastQueuedCommand(bool retain);
//! Get the submitted batch
Command* GetSubmissionBatch() const { return head_; }
//! Get the current batch size
size_t GetSubmissionBatchSize() const {
ScopedLock sl(vdev()->execution());
return size_;
}
//! Insert a command into the linked list of submitted commands
void FormSubmissionBatch(Command* command) {
// Insert the command to the linked list.
if (nullptr == head_) { // if the list is empty
head_ = tail_ = command;
} else {
tail_->setNext(command);
tail_ = command;
}
size_++;
command->setStatus(CL_SUBMITTED);
command->retain();
// @note: runtime needs double retain in order to maintain the batch,
// because setStatus(COMPLETE) releases command and batch update may have
// an invalid access
command->retain();
// Release the last command in the batch
if (lastEnqueueCommand_ != nullptr) {
lastEnqueueCommand_->release();
} else {
// The queue becomes active. Add it to the set of activeQueues.
device_.addToActiveQueues(this);
}
// Extra retain for the last command
command->retain();
lastEnqueueCommand_ = command;
}
//! Flushes submitted commands if the batch size significantly grew
void FlushSubmissionBatch(Command* command) {
if (size_ > DEBUG_CLR_MAX_BATCH_SIZE) {
command->notifyCmdQueue();
}
}
//! Reset the command batch list
void ResetSubmissionBatch() {
head_ = nullptr;
size_ = 0;
}
//! Set queue status
void SetQueueStatus() { isActive_ = true; }
//! Get queue status
bool GetQueueStatus() { return isActive_; }
//! Set the force destory to terminate queue without checking last command
void SetForceDestroy(bool forceDestroy) { forceDestroy_ = forceDestroy; }
uint64_t getQueueID() { return thread_.vdev()->getQueueID(); }
//! Returns Synchronization Policy for the current stream
amd::SyncPolicy GetSyncPolicy() const { return sync_policy_; }
//! Set Synchronization Policy used by Queue
void SetSyncPolicy(amd::SyncPolicy value) { sync_policy_ = value; }
private:
Command* head_; //!< Head of the batch list
Command* tail_; //!< Tail of the batch list
size_t size_ = 0; //!< The current batch size
//! True if this command queue is active
std::atomic<bool> isActive_;
bool forceDestroy_ = false; //!< Destroy the queue in the current state
amd::SyncPolicy sync_policy_; //!< Used for controlling stream synchronization
};
class DeviceQueue : public CommandQueue {
public:
DeviceQueue(Context& context, //!< Context object
Device& device, //!< Device object
cl_command_queue_properties properties, //!< Queue properties
uint size //!< Device queue size
)
: CommandQueue(context, device, properties,
device.info().queueOnDeviceProperties_ | CL_QUEUE_ON_DEVICE |
CL_QUEUE_ON_DEVICE_DEFAULT),
size_(size),
virtualDevice_(NULL) {}
virtual ~DeviceQueue();
//! Returns TRUE if device queue was successfully created
virtual bool create();
//! Return the current queue as the DeviceQueue
virtual DeviceQueue* asDeviceQueue() { return this; }
//! Returns the size of device queue
uint size() const { return size_; }
//! Returns virtual device for this device queue
device::VirtualDevice* vDev() const { return virtualDevice_; }
private:
uint size_; //!< Device queue size
device::VirtualDevice* virtualDevice_; //!< Virtual device for this queue
};
} // namespace amd
#endif // COMMAND_QUEUE_HPP_