src/Vulkan/VkDevice.cpp - SwiftShader - Git at Google

 // Copyright 2018 The SwiftShader Authors. All Rights Reserved.
 //
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
 // You may obtain a copy of the License at
 //
 //    http://www.apache.org/licenses/LICENSE-2.0
 //
 // Unless required by applicable law or agreed to in writing, software
 // distributed under the License is distributed on an "AS IS" BASIS,
 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 // See the License for the specific language governing permissions and
 // limitations under the License.

 #include "VkDevice.hpp"

 #include "VkConfig.h"
 #include "VkDescriptorSetLayout.hpp"
 #include "VkFence.hpp"
 #include "VkQueue.hpp"
 #include "Debug/Context.hpp"
 #include "Debug/Server.hpp"
 #include "Device/Blitter.hpp"
 #include "System/Debug.hpp"

 #include <chrono>
 #include <climits>
 #include <new>  // Must #include this to use "placement new"

 namespace {

 std::chrono::time_point<std::chrono::system_clock, std::chrono::nanoseconds> now()
 {
 	return std::chrono::time_point_cast<std::chrono::nanoseconds>(std::chrono::system_clock::now());
 }

 }  // anonymous namespace

 namespace vk {

 void Device::SamplingRoutineCache::updateSnapshot()
 {
 	marl::lock lock(mutex);

 	if(snapshotNeedsUpdate)
 	{
 		snapshot.clear();

 		for(auto it : cache)
 		{
 			snapshot[it.key()] = it.data();
 		}

 		snapshotNeedsUpdate = false;
 	}
 }

 Device::SamplerIndexer::~SamplerIndexer()
 {
 	ASSERT(map.empty());
 }

 uint32_t Device::SamplerIndexer::index(const SamplerState &samplerState)
 {
 	marl::lock lock(mutex);

 	auto it = map.find(samplerState);

 	if(it != map.end())
 	{
 		it->second.count++;
 		return it->second.id;
 	}

 	nextID++;

 	map.emplace(samplerState, Identifier{ nextID, 1 });

 	return nextID;
 }

 void Device::SamplerIndexer::remove(const SamplerState &samplerState)
 {
 	marl::lock lock(mutex);

 	auto it = map.find(samplerState);
 	ASSERT(it != map.end());

 	auto count = --it->second.count;
 	if(count == 0)
 	{
 		map.erase(it);
 	}
 }

 Device::Device(const VkDeviceCreateInfo *pCreateInfo, void *mem, PhysicalDevice *physicalDevice, const VkPhysicalDeviceFeatures *enabledFeatures, const std::shared_ptr<marl::Scheduler> &scheduler)
     : physicalDevice(physicalDevice)
     , queues(reinterpret_cast<Queue *>(mem))
     , enabledExtensionCount(pCreateInfo->enabledExtensionCount)
     , enabledFeatures(enabledFeatures ? *enabledFeatures : VkPhysicalDeviceFeatures{})
     ,  // "Setting pEnabledFeatures to NULL and not including a VkPhysicalDeviceFeatures2 in the pNext member of VkDeviceCreateInfo is equivalent to setting all members of the structure to VK_FALSE."
     scheduler(scheduler)
 {
 	for(uint32_t i = 0; i < pCreateInfo->queueCreateInfoCount; i++)
 	{
 		const VkDeviceQueueCreateInfo &queueCreateInfo = pCreateInfo->pQueueCreateInfos[i];
 		queueCount += queueCreateInfo.queueCount;
 	}

 	uint32_t queueID = 0;
 	for(uint32_t i = 0; i < pCreateInfo->queueCreateInfoCount; i++)
 	{
 		const VkDeviceQueueCreateInfo &queueCreateInfo = pCreateInfo->pQueueCreateInfos[i];

 		for(uint32_t j = 0; j < queueCreateInfo.queueCount; j++, queueID++)
 		{
 			new(&queues[queueID]) Queue(this, scheduler.get());
 		}
 	}

 	extensions = reinterpret_cast<ExtensionName *>(static_cast<uint8_t *>(mem) + (sizeof(Queue) * queueCount));
 	for(uint32_t i = 0; i < enabledExtensionCount; i++)
 	{
 		strncpy(extensions[i], pCreateInfo->ppEnabledExtensionNames[i], VK_MAX_EXTENSION_NAME_SIZE);
 	}

 	if(pCreateInfo->enabledLayerCount)
 	{
 		// "The ppEnabledLayerNames and enabledLayerCount members of VkDeviceCreateInfo are deprecated and their values must be ignored by implementations."
 		UNSUPPORTED("enabledLayerCount");
 	}

 	// FIXME (b/119409619): use an allocator here so we can control all memory allocations
 	blitter.reset(new sw::Blitter());
 	samplingRoutineCache.reset(new SamplingRoutineCache());
 	samplerIndexer.reset(new SamplerIndexer());

 #ifdef ENABLE_VK_DEBUGGER
 	static auto port = getenv("VK_DEBUGGER_PORT");
 	if(port)
 	{
 		// Construct the debugger context and server - this may block for a
 		// debugger connection, allowing breakpoints to be set before they're
 		// executed.
 		debugger.context = vk::dbg::Context::create();
 		debugger.server = vk::dbg::Server::create(debugger.context, atoi(port));
 	}
 #endif  // ENABLE_VK_DEBUGGER
 }

 void Device::destroy(const VkAllocationCallbacks *pAllocator)
 {
 	for(uint32_t i = 0; i < queueCount; i++)
 	{
 		queues[i].~Queue();
 	}

 	vk::deallocate(queues, pAllocator);
 }

 size_t Device::ComputeRequiredAllocationSize(const VkDeviceCreateInfo *pCreateInfo)
 {
 	uint32_t queueCount = 0;
 	for(uint32_t i = 0; i < pCreateInfo->queueCreateInfoCount; i++)
 	{
 		queueCount += pCreateInfo->pQueueCreateInfos[i].queueCount;
 	}

 	return (sizeof(Queue) * queueCount) + (pCreateInfo->enabledExtensionCount * sizeof(ExtensionName));
 }

 bool Device::hasExtension(const char *extensionName) const
 {
 	for(uint32_t i = 0; i < enabledExtensionCount; i++)
 	{
 		if(strncmp(extensions[i], extensionName, VK_MAX_EXTENSION_NAME_SIZE) == 0)
 		{
 			return true;
 		}
 	}
 	return false;
 }

 VkQueue Device::getQueue(uint32_t queueFamilyIndex, uint32_t queueIndex) const
 {
 	ASSERT(queueFamilyIndex == 0);

 	return queues[queueIndex];
 }

 VkResult Device::waitForFences(uint32_t fenceCount, const VkFence *pFences, VkBool32 waitAll, uint64_t timeout)
 {
 	using time_point = std::chrono::time_point<std::chrono::system_clock, std::chrono::nanoseconds>;
 	const time_point start = now();
 	const uint64_t max_timeout = (LLONG_MAX - start.time_since_epoch().count());
 	bool infiniteTimeout = (timeout > max_timeout);
 	const time_point end_ns = start + std::chrono::nanoseconds(std::min(max_timeout, timeout));

 	if(waitAll != VK_FALSE)  // All fences must be signaled
 	{
 		for(uint32_t i = 0; i < fenceCount; i++)
 		{
 			if(timeout == 0)
 			{
 				if(Cast(pFences[i])->getStatus() != VK_SUCCESS)  // At least one fence is not signaled
 				{
 					return VK_TIMEOUT;
 				}
 			}
 			else if(infiniteTimeout)
 			{
 				if(Cast(pFences[i])->wait() != VK_SUCCESS)  // At least one fence is not signaled
 				{
 					return VK_TIMEOUT;
 				}
 			}
 			else
 			{
 				if(Cast(pFences[i])->wait(end_ns) != VK_SUCCESS)  // At least one fence is not signaled
 				{
 					return VK_TIMEOUT;
 				}
 			}
 		}

 		return VK_SUCCESS;
 	}
 	else  // At least one fence must be signaled
 	{
 		marl::containers::vector<marl::Event, 8> events;
 		for(uint32_t i = 0; i < fenceCount; i++)
 		{
 			events.push_back(Cast(pFences[i])->getEvent());
 		}

 		auto any = marl::Event::any(events.begin(), events.end());

 		if(timeout == 0)
 		{
 			return any.isSignalled() ? VK_SUCCESS : VK_TIMEOUT;
 		}
 		else if(infiniteTimeout)
 		{
 			any.wait();
 			return VK_SUCCESS;
 		}
 		else
 		{
 			return any.wait_until(end_ns) ? VK_SUCCESS : VK_TIMEOUT;
 		}
 	}
 }

 VkResult Device::waitIdle()
 {
 	for(uint32_t i = 0; i < queueCount; i++)
 	{
 		queues[i].waitIdle();
 	}

 	return VK_SUCCESS;
 }

 void Device::getDescriptorSetLayoutSupport(const VkDescriptorSetLayoutCreateInfo *pCreateInfo,
                                            VkDescriptorSetLayoutSupport *pSupport) const
 {
 	// From Vulkan Spec 13.2.1 Descriptor Set Layout, in description of vkGetDescriptorSetLayoutSupport:
 	// "This command does not consider other limits such as maxPerStageDescriptor*, and so a descriptor
 	// set layout that is supported according to this command must still satisfy the pipeline layout limits
 	// such as maxPerStageDescriptor* in order to be used in a pipeline layout."

 	// We have no "strange" limitations to enforce beyond the device limits, so we can safely always claim support.
 	pSupport->supported = VK_TRUE;
 }

 void Device::updateDescriptorSets(uint32_t descriptorWriteCount, const VkWriteDescriptorSet *pDescriptorWrites,
                                   uint32_t descriptorCopyCount, const VkCopyDescriptorSet *pDescriptorCopies)
 {
 	for(uint32_t i = 0; i < descriptorWriteCount; i++)
 	{
 		DescriptorSetLayout::WriteDescriptorSet(this, pDescriptorWrites[i]);
 	}

 	for(uint32_t i = 0; i < descriptorCopyCount; i++)
 	{
 		DescriptorSetLayout::CopyDescriptorSet(pDescriptorCopies[i]);
 	}
 }

 void Device::getRequirements(VkMemoryDedicatedRequirements *requirements) const
 {
 	requirements->prefersDedicatedAllocation = VK_FALSE;
 	requirements->requiresDedicatedAllocation = VK_FALSE;
 }

 Device::SamplingRoutineCache *Device::getSamplingRoutineCache() const
 {
 	return samplingRoutineCache.get();
 }

 void Device::updateSamplingRoutineSnapshotCache()
 {
 	samplingRoutineCache->updateSnapshot();
 }

 uint32_t Device::indexSampler(const SamplerState &samplerState)
 {
 	return samplerIndexer->index(samplerState);
 }

 void Device::removeSampler(const SamplerState &samplerState)
 {
 	samplerIndexer->remove(samplerState);
 }

 }  // namespace vk
	// Copyright 2018 The SwiftShader Authors. All Rights Reserved.
	//
	// Licensed under the Apache License, Version 2.0 (the "License");
	// you may not use this file except in compliance with the License.
	// You may obtain a copy of the License at
	//
	// http://www.apache.org/licenses/LICENSE-2.0
	//
	// Unless required by applicable law or agreed to in writing, software
	// distributed under the License is distributed on an "AS IS" BASIS,
	// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
	// See the License for the specific language governing permissions and
	// limitations under the License.

	#include "VkDevice.hpp"

	#include "VkConfig.h"
	#include "VkDescriptorSetLayout.hpp"
	#include "VkFence.hpp"
	#include "VkQueue.hpp"
	#include "Debug/Context.hpp"
	#include "Debug/Server.hpp"
	#include "Device/Blitter.hpp"
	#include "System/Debug.hpp"

	#include <chrono>
	#include <climits>
	#include <new> // Must #include this to use "placement new"

	namespace {

	std::chrono::time_point<std::chrono::system_clock, std::chrono::nanoseconds> now()
	{
	return std::chrono::time_point_cast<std::chrono::nanoseconds>(std::chrono::system_clock::now());
	}

	} // anonymous namespace

	namespace vk {

	void Device::SamplingRoutineCache::updateSnapshot()
	{
	marl::lock lock(mutex);

	if(snapshotNeedsUpdate)
	{
	snapshot.clear();

	for(auto it : cache)
	{
	snapshot[it.key()] = it.data();
	}

	snapshotNeedsUpdate = false;
	}
	}

	Device::SamplerIndexer::~SamplerIndexer()
	{
	ASSERT(map.empty());
	}

	uint32_t Device::SamplerIndexer::index(const SamplerState &samplerState)
	{
	marl::lock lock(mutex);

	auto it = map.find(samplerState);

	if(it != map.end())
	{
	it->second.count++;
	return it->second.id;
	}

	nextID++;

	map.emplace(samplerState, Identifier{ nextID, 1 });

	return nextID;
	}

	void Device::SamplerIndexer::remove(const SamplerState &samplerState)
	{
	marl::lock lock(mutex);

	auto it = map.find(samplerState);
	ASSERT(it != map.end());

	auto count = --it->second.count;
	if(count == 0)
	{
	map.erase(it);
	}
	}

	Device::Device(const VkDeviceCreateInfo pCreateInfo, void mem, PhysicalDevice physicalDevice, const VkPhysicalDeviceFeatures enabledFeatures, const std::shared_ptr<marl::Scheduler> &scheduler)
	: physicalDevice(physicalDevice)
	, queues(reinterpret_cast<Queue *>(mem))
	, enabledExtensionCount(pCreateInfo->enabledExtensionCount)
	, enabledFeatures(enabledFeatures ? *enabledFeatures : VkPhysicalDeviceFeatures{})
	, // "Setting pEnabledFeatures to NULL and not including a VkPhysicalDeviceFeatures2 in the pNext member of VkDeviceCreateInfo is equivalent to setting all members of the structure to VK_FALSE."
	scheduler(scheduler)
	{
	for(uint32_t i = 0; i < pCreateInfo->queueCreateInfoCount; i++)
	{
	const VkDeviceQueueCreateInfo &queueCreateInfo = pCreateInfo->pQueueCreateInfos[i];
	queueCount += queueCreateInfo.queueCount;
	}

	uint32_t queueID = 0;
	for(uint32_t i = 0; i < pCreateInfo->queueCreateInfoCount; i++)
	{
	const VkDeviceQueueCreateInfo &queueCreateInfo = pCreateInfo->pQueueCreateInfos[i];

	for(uint32_t j = 0; j < queueCreateInfo.queueCount; j++, queueID++)
	{
	new(&queues[queueID]) Queue(this, scheduler.get());
	}
	}

	extensions = reinterpret_cast<ExtensionName >(static_cast<uint8_t >(mem) + (sizeof(Queue) * queueCount));
	for(uint32_t i = 0; i < enabledExtensionCount; i++)
	{
	strncpy(extensions[i], pCreateInfo->ppEnabledExtensionNames[i], VK_MAX_EXTENSION_NAME_SIZE);
	}

	if(pCreateInfo->enabledLayerCount)
	{
	// "The ppEnabledLayerNames and enabledLayerCount members of VkDeviceCreateInfo are deprecated and their values must be ignored by implementations."
	UNSUPPORTED("enabledLayerCount");
	}

	// FIXME (b/119409619): use an allocator here so we can control all memory allocations
	blitter.reset(new sw::Blitter());
	samplingRoutineCache.reset(new SamplingRoutineCache());
	samplerIndexer.reset(new SamplerIndexer());

	#ifdef ENABLE_VK_DEBUGGER
	static auto port = getenv("VK_DEBUGGER_PORT");
	if(port)
	{
	// Construct the debugger context and server - this may block for a
	// debugger connection, allowing breakpoints to be set before they're
	// executed.
	debugger.context = vk::dbg::Context::create();
	debugger.server = vk::dbg::Server::create(debugger.context, atoi(port));
	}
	#endif // ENABLE_VK_DEBUGGER
	}

	void Device::destroy(const VkAllocationCallbacks *pAllocator)
	{
	for(uint32_t i = 0; i < queueCount; i++)
	{
	queues[i].~Queue();
	}

	vk::deallocate(queues, pAllocator);
	}

	size_t Device::ComputeRequiredAllocationSize(const VkDeviceCreateInfo *pCreateInfo)
	{
	uint32_t queueCount = 0;
	for(uint32_t i = 0; i < pCreateInfo->queueCreateInfoCount; i++)
	{
	queueCount += pCreateInfo->pQueueCreateInfos[i].queueCount;
	}

	return (sizeof(Queue) * queueCount) + (pCreateInfo->enabledExtensionCount * sizeof(ExtensionName));
	}

	bool Device::hasExtension(const char *extensionName) const
	{
	for(uint32_t i = 0; i < enabledExtensionCount; i++)
	{
	if(strncmp(extensions[i], extensionName, VK_MAX_EXTENSION_NAME_SIZE) == 0)
	{
	return true;
	}
	}
	return false;
	}

	VkQueue Device::getQueue(uint32_t queueFamilyIndex, uint32_t queueIndex) const
	{
	ASSERT(queueFamilyIndex == 0);

	return queues[queueIndex];
	}

	VkResult Device::waitForFences(uint32_t fenceCount, const VkFence *pFences, VkBool32 waitAll, uint64_t timeout)
	{
	using time_point = std::chrono::time_point<std::chrono::system_clock, std::chrono::nanoseconds>;
	const time_point start = now();
	const uint64_t max_timeout = (LLONG_MAX - start.time_since_epoch().count());
	bool infiniteTimeout = (timeout > max_timeout);
	const time_point end_ns = start + std::chrono::nanoseconds(std::min(max_timeout, timeout));

	if(waitAll != VK_FALSE) // All fences must be signaled
	{
	for(uint32_t i = 0; i < fenceCount; i++)
	{
	if(timeout == 0)
	{
	if(Cast(pFences[i])->getStatus() != VK_SUCCESS) // At least one fence is not signaled
	{
	return VK_TIMEOUT;
	}
	}
	else if(infiniteTimeout)
	{
	if(Cast(pFences[i])->wait() != VK_SUCCESS) // At least one fence is not signaled
	{
	return VK_TIMEOUT;
	}
	}
	else
	{
	if(Cast(pFences[i])->wait(end_ns) != VK_SUCCESS) // At least one fence is not signaled
	{
	return VK_TIMEOUT;
	}
	}
	}

	return VK_SUCCESS;
	}
	else // At least one fence must be signaled
	{
	marl::containers::vector<marl::Event, 8> events;
	for(uint32_t i = 0; i < fenceCount; i++)
	{
	events.push_back(Cast(pFences[i])->getEvent());
	}

	auto any = marl::Event::any(events.begin(), events.end());

	if(timeout == 0)
	{
	return any.isSignalled() ? VK_SUCCESS : VK_TIMEOUT;
	}
	else if(infiniteTimeout)
	{
	any.wait();
	return VK_SUCCESS;
	}
	else
	{
	return any.wait_until(end_ns) ? VK_SUCCESS : VK_TIMEOUT;
	}
	}
	}

	VkResult Device::waitIdle()
	{
	for(uint32_t i = 0; i < queueCount; i++)
	{
	queues[i].waitIdle();
	}

	return VK_SUCCESS;
	}

	void Device::getDescriptorSetLayoutSupport(const VkDescriptorSetLayoutCreateInfo *pCreateInfo,
	VkDescriptorSetLayoutSupport *pSupport) const
	{
	// From Vulkan Spec 13.2.1 Descriptor Set Layout, in description of vkGetDescriptorSetLayoutSupport:
	// "This command does not consider other limits such as maxPerStageDescriptor*, and so a descriptor
	// set layout that is supported according to this command must still satisfy the pipeline layout limits
	// such as maxPerStageDescriptor* in order to be used in a pipeline layout."

	// We have no "strange" limitations to enforce beyond the device limits, so we can safely always claim support.
	pSupport->supported = VK_TRUE;
	}

	void Device::updateDescriptorSets(uint32_t descriptorWriteCount, const VkWriteDescriptorSet *pDescriptorWrites,
	uint32_t descriptorCopyCount, const VkCopyDescriptorSet *pDescriptorCopies)
	{
	for(uint32_t i = 0; i < descriptorWriteCount; i++)
	{
	DescriptorSetLayout::WriteDescriptorSet(this, pDescriptorWrites[i]);
	}

	for(uint32_t i = 0; i < descriptorCopyCount; i++)
	{
	DescriptorSetLayout::CopyDescriptorSet(pDescriptorCopies[i]);
	}
	}

	void Device::getRequirements(VkMemoryDedicatedRequirements *requirements) const
	{
	requirements->prefersDedicatedAllocation = VK_FALSE;
	requirements->requiresDedicatedAllocation = VK_FALSE;
	}

	Device::SamplingRoutineCache *Device::getSamplingRoutineCache() const
	{
	return samplingRoutineCache.get();
	}

	void Device::updateSamplingRoutineSnapshotCache()
	{
	samplingRoutineCache->updateSnapshot();
	}

	uint32_t Device::indexSampler(const SamplerState &samplerState)
	{
	return samplerIndexer->index(samplerState);
	}

	void Device::removeSampler(const SamplerState &samplerState)
	{
	samplerIndexer->remove(samplerState);
	}

	} // namespace vk