P4 to Git Change 1536925 by vsytchen@vsytchen-ocl-win10 on 2018/04/04 17:20:38
SWDEV-79445 - OCL generic changes and code clean-up 1. This change replaces the use of std::map with std::unordered_map to improve lookup/insert time. 2. Replace the use of std::make_pair and std::pair constructor with uniform initialization for cleaner code. 3. Replace the use of std::Container::iterator type with the auto keyword for cleaner code. 4. Use range based for loops where needed. ReviewBoardURL = http://ocltc.amd.com/reviews/r/14517/diff/ Affected files ... ... //depot/stg/opencl/drivers/opencl/api/hip/hip_platform.cpp#4 edit ... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_context.cpp#58 edit ... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_d3d10.cpp#16 edit ... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_d3d10_amd.hpp#9 edit ... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_d3d11.cpp#24 edit ... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_d3d11_amd.hpp#13 edit ... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_d3d9.cpp#34 edit ... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_d3d9_amd.hpp#17 edit ... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_gl.cpp#57 edit ... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_pipe.cpp#7 edit ... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_program.cpp#46 edit ... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_svm.cpp#23 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/appprofile.hpp#14 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/cpu/cpuprogram.cpp#72 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/cpu/cpuvirtual.cpp#27 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/device.cpp#216 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/device.hpp#297 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuappprofile.cpp#13 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpubinary.cpp#59 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpucompiler.cpp#158 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudevice.cpp#587 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpukernel.cpp#322 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuprintf.cpp#46 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuprogram.cpp#237 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuprogram.hpp#70 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuresource.cpp#242 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuvirtual.cpp#415 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuvirtual.hpp#143 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palappprofile.cpp#3 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palcompiler.cpp#22 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.cpp#79 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palprintf.cpp#9 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palprogram.cpp#59 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palresource.cpp#60 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.cpp#84 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.hpp#46 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/CMakeLists.txt#11 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/pro/prodevice.cpp#4 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/pro/prodevice.hpp#5 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocbinary.hpp#6 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/roccompiler.cpp#42 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/roccounters.cpp#3 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocprintf.cpp#10 edit ... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocprogram.cpp#81 edit ... //depot/stg/opencl/drivers/opencl/runtime/platform/command.cpp#81 edit ... //depot/stg/opencl/drivers/opencl/runtime/platform/command.hpp#89 edit ... //depot/stg/opencl/drivers/opencl/runtime/platform/commandqueue.cpp#24 edit ... //depot/stg/opencl/drivers/opencl/runtime/platform/context.cpp#49 edit ... //depot/stg/opencl/drivers/opencl/runtime/platform/context.hpp#29 edit ... //depot/stg/opencl/drivers/opencl/runtime/platform/memory.cpp#129 edit ... //depot/stg/opencl/drivers/opencl/runtime/platform/memory.hpp#102 edit ... //depot/stg/opencl/drivers/opencl/runtime/platform/perfctr.hpp#7 edit ... //depot/stg/opencl/drivers/opencl/runtime/platform/program.cpp#91 edit ... //depot/stg/opencl/drivers/opencl/runtime/platform/program.hpp#43 edit ... //depot/stg/opencl/drivers/opencl/runtime/platform/sampler.hpp#9 edit ... //depot/stg/opencl/drivers/opencl/runtime/utils/flags.cpp#17 edit
Esse commit está contido em:
foreman
@@ -370,11 +370,10 @@ void UnmapMemoryCommand::releaseResources() {
|
||||
|
||||
bool MigrateMemObjectsCommand::validateMemory() {
|
||||
if (queue()->device().info().type_ & CL_DEVICE_TYPE_GPU) {
|
||||
std::vector<amd::Memory*>::const_iterator itr;
|
||||
for (itr = memObjects_.begin(); itr != memObjects_.end(); itr++) {
|
||||
device::Memory* mem = (*itr)->getDeviceMemory(queue()->device());
|
||||
for (const auto& it : memObjects_) {
|
||||
device::Memory* mem = it->getDeviceMemory(queue()->device());
|
||||
if (NULL == mem) {
|
||||
LogPrintfError("Can't allocate memory size - 0x%08X bytes!", (*itr)->getSize());
|
||||
LogPrintfError("Can't allocate memory size - 0x%08X bytes!", it->getSize());
|
||||
return false;
|
||||
}
|
||||
}
|
||||
@@ -434,11 +433,10 @@ cl_int NDRangeKernelCommand::validateMemory() {
|
||||
bool ExtObjectsCommand::validateMemory() {
|
||||
bool retVal = true;
|
||||
if (queue()->device().info().type_ & CL_DEVICE_TYPE_GPU) {
|
||||
for (std::vector<amd::Memory*>::const_iterator itr = memObjects_.begin();
|
||||
itr != memObjects_.end(); itr++) {
|
||||
device::Memory* mem = (*itr)->getDeviceMemory(queue()->device());
|
||||
for (const auto& it : memObjects_) {
|
||||
device::Memory* mem = it->getDeviceMemory(queue()->device());
|
||||
if (NULL == mem) {
|
||||
LogPrintfError("Can't allocate memory size - 0x%08X bytes!", (*itr)->getSize());
|
||||
LogPrintfError("Can't allocate memory size - 0x%08X bytes!", it->getSize());
|
||||
return false;
|
||||
}
|
||||
retVal = processGLResource(mem);
|
||||
@@ -457,11 +455,10 @@ bool ReleaseExtObjectsCommand::processGLResource(device::Memory* mem) {
|
||||
|
||||
bool MakeBuffersResidentCommand::validateMemory() {
|
||||
if (queue()->device().info().type_ & CL_DEVICE_TYPE_GPU) {
|
||||
for (std::vector<amd::Memory*>::const_iterator itr = memObjects_.begin();
|
||||
itr != memObjects_.end(); itr++) {
|
||||
device::Memory* mem = (*itr)->getDeviceMemory(queue()->device());
|
||||
for (const auto& it : memObjects_) {
|
||||
device::Memory* mem = it->getDeviceMemory(queue()->device());
|
||||
if (NULL == mem) {
|
||||
LogPrintfError("Can't allocate memory size - 0x%08X bytes!", (*itr)->getSize());
|
||||
LogPrintfError("Can't allocate memory size - 0x%08X bytes!", it->getSize());
|
||||
return false;
|
||||
}
|
||||
}
|
||||
@@ -471,16 +468,14 @@ bool MakeBuffersResidentCommand::validateMemory() {
|
||||
}
|
||||
bool ThreadTraceMemObjectsCommand::validateMemory() {
|
||||
if (queue()->device().info().type_ & CL_DEVICE_TYPE_GPU) {
|
||||
for (std::vector<amd::Memory*>::const_iterator itr = memObjects_.begin();
|
||||
itr != memObjects_.end(); itr++) {
|
||||
device::Memory* mem = (*itr)->getDeviceMemory(queue()->device());
|
||||
for (auto& it = memObjects_.cbegin(); it != memObjects_.cend(); it++) {
|
||||
device::Memory* mem = (*it)->getDeviceMemory(queue()->device());
|
||||
if (NULL == mem) {
|
||||
std::vector<amd::Memory*>::const_iterator tmpItr;
|
||||
for (tmpItr = memObjects_.begin(); tmpItr != itr; tmpItr++) {
|
||||
device::Memory* tmpMem = (*tmpItr)->getDeviceMemory(queue()->device());
|
||||
for (auto& tmpIt = memObjects_.cbegin(); tmpIt != it; tmpIt++) {
|
||||
device::Memory* tmpMem = (*tmpIt)->getDeviceMemory(queue()->device());
|
||||
delete tmpMem;
|
||||
}
|
||||
LogPrintfError("Can't allocate memory size - 0x%08X bytes!", (*itr)->getSize());
|
||||
LogPrintfError("Can't allocate memory size - 0x%08X bytes!", (*it)->getSize());
|
||||
return false;
|
||||
}
|
||||
}
|
||||
|
||||
@@ -720,10 +720,9 @@ class MigrateMemObjectsCommand : public Command {
|
||||
const std::vector<amd::Memory*>& memObjects,
|
||||
cl_mem_migration_flags flags)
|
||||
: Command(queue, type, eventWaitList), migrationFlags_(flags) {
|
||||
std::vector<amd::Memory*>::const_iterator itr;
|
||||
for (itr = memObjects.begin(); itr != memObjects.end(); itr++) {
|
||||
(*itr)->retain();
|
||||
memObjects_.push_back(*itr);
|
||||
for (const auto& it : memObjects) {
|
||||
it->retain();
|
||||
memObjects_.push_back(it);
|
||||
}
|
||||
}
|
||||
|
||||
@@ -731,9 +730,8 @@ class MigrateMemObjectsCommand : public Command {
|
||||
|
||||
//! Release all resources associated with this command
|
||||
void releaseResources() {
|
||||
std::vector<amd::Memory*>::const_iterator itr;
|
||||
for (itr = memObjects_.begin(); itr != memObjects_.end(); itr++) {
|
||||
(*itr)->release();
|
||||
for (const auto& it : memObjects_) {
|
||||
it->release();
|
||||
}
|
||||
Command::releaseResources();
|
||||
}
|
||||
@@ -837,18 +835,16 @@ class ExtObjectsCommand : public Command {
|
||||
ExtObjectsCommand(HostQueue& queue, const EventWaitList& eventWaitList, cl_uint num_objects,
|
||||
const std::vector<amd::Memory*>& memoryObjects, cl_command_type type)
|
||||
: Command(queue, type, eventWaitList) {
|
||||
for (std::vector<amd::Memory*>::const_iterator itr = memoryObjects.begin();
|
||||
itr != memoryObjects.end(); itr++) {
|
||||
(*itr)->retain();
|
||||
memObjects_.push_back(*itr);
|
||||
for (const auto& it : memoryObjects) {
|
||||
it->retain();
|
||||
memObjects_.push_back(it);
|
||||
}
|
||||
}
|
||||
|
||||
//! Release all resources associated with this command
|
||||
void releaseResources() {
|
||||
for (std::vector<amd::Memory*>::const_iterator itr = memObjects_.begin();
|
||||
itr != memObjects_.end(); itr++) {
|
||||
(*itr)->release();
|
||||
for (const auto& it : memObjects_) {
|
||||
it->release();
|
||||
}
|
||||
Command::releaseResources();
|
||||
}
|
||||
@@ -954,9 +950,8 @@ class ThreadTraceMemObjectsCommand : public Command {
|
||||
//! Release all resources associated with this command
|
||||
void releaseResources() {
|
||||
threadTrace_.release();
|
||||
for (std::vector<amd::Memory*>::const_iterator itr = memObjects_.begin();
|
||||
itr != memObjects_.end(); itr++) {
|
||||
(*itr)->release();
|
||||
for (const auto& itr : memObjects_) {
|
||||
itr->release();
|
||||
}
|
||||
Command::releaseResources();
|
||||
}
|
||||
@@ -1067,19 +1062,17 @@ class MakeBuffersResidentCommand : public Command {
|
||||
const std::vector<amd::Memory*>& memObjects,
|
||||
cl_bus_address_amd* busAddr)
|
||||
: Command(queue, type, eventWaitList), busAddresses_(busAddr) {
|
||||
std::vector<amd::Memory*>::const_iterator itr;
|
||||
for (itr = memObjects.begin(); itr != memObjects.end(); itr++) {
|
||||
(*itr)->retain();
|
||||
memObjects_.push_back(*itr);
|
||||
for (const auto& it : memObjects) {
|
||||
it->retain();
|
||||
memObjects_.push_back(it);
|
||||
}
|
||||
}
|
||||
|
||||
virtual void submit(device::VirtualDevice& device) { device.submitMakeBuffersResident(*this); }
|
||||
|
||||
void releaseResources() {
|
||||
std::vector<amd::Memory*>::const_iterator itr;
|
||||
for (itr = memObjects_.begin(); itr != memObjects_.end(); itr++) {
|
||||
(*itr)->release();
|
||||
for (const auto& it : memObjects_) {
|
||||
it->release();
|
||||
}
|
||||
Command::releaseResources();
|
||||
}
|
||||
|
||||
@@ -96,15 +96,14 @@ void HostQueue::loop(device::VirtualDevice* virtualDevice) {
|
||||
|
||||
// Process the command's event wait list.
|
||||
const Command::EventWaitList& events = command->eventWaitList();
|
||||
Command::EventWaitList::const_iterator it;
|
||||
bool dependencyFailed = false;
|
||||
|
||||
for (it = events.begin(); it != events.end(); ++it) {
|
||||
for (const auto& it : events) {
|
||||
// Only wait if the command is enqueued into another queue.
|
||||
if ((*it)->command().queue() != this) {
|
||||
if (it->command().queue() != this) {
|
||||
virtualDevice->flush(head, true);
|
||||
tail = head = NULL;
|
||||
dependencyFailed |= !(*it)->awaitCompletion();
|
||||
dependencyFailed |= !it->awaitCompletion();
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@@ -62,10 +62,9 @@ Context::~Context() {
|
||||
|
||||
// Dissociate OCL context with any external device
|
||||
if (info_.flags_ & (GLDeviceKhr | D3D10DeviceKhr | D3D11DeviceKhr)) {
|
||||
std::vector<Device*>::const_iterator it;
|
||||
// Loop through all devices
|
||||
for (it = devices_.begin(); it != devices_.end(); it++) {
|
||||
(*it)->unbindExternalDevice(info_.flags_, info_.hDev_, info_.hCtx_, VALIDATE_ONLY);
|
||||
for (const auto& it : devices_) {
|
||||
it->unbindExternalDevice(info_.flags_, info_.hDev_, info_.hCtx_, VALIDATE_ONLY);
|
||||
}
|
||||
}
|
||||
|
||||
@@ -218,10 +217,9 @@ int Context::create(const intptr_t* properties) {
|
||||
// Check if OCL context can be associated with any external device
|
||||
if (info_.flags_ & (D3D10DeviceKhr | D3D11DeviceKhr | GLDeviceKhr | D3D9DeviceKhr |
|
||||
D3D9DeviceEXKhr | D3D9DeviceVAKhr)) {
|
||||
std::vector<Device*>::const_iterator it;
|
||||
// Loop through all devices
|
||||
for (it = devices_.begin(); it != devices_.end(); it++) {
|
||||
if (!(*it)->bindExternalDevice(info_.flags_, info_.hDev_, info_.hCtx_, VALIDATE_ONLY)) {
|
||||
for (const auto& it : devices_) {
|
||||
if (!it->bindExternalDevice(info_.flags_, info_.hDev_, info_.hCtx_, VALIDATE_ONLY)) {
|
||||
result = CL_INVALID_VALUE;
|
||||
}
|
||||
}
|
||||
@@ -331,10 +329,9 @@ void Context::svmFree(void* ptr) const {
|
||||
}
|
||||
|
||||
bool Context::containsDevice(const Device* device) const {
|
||||
std::vector<Device*>::const_iterator it;
|
||||
|
||||
for (it = devices_.begin(); it != devices_.end(); ++it) {
|
||||
if (device == *it || (*it)->isAncestor(device)) {
|
||||
for (const auto& it : devices_) {
|
||||
if (device == it || it->isAncestor(device)) {
|
||||
return true;
|
||||
}
|
||||
}
|
||||
@@ -342,8 +339,8 @@ bool Context::containsDevice(const Device* device) const {
|
||||
}
|
||||
|
||||
DeviceQueue* Context::defDeviceQueue(const Device& dev) const {
|
||||
std::map<const Device*, DeviceQueueInfo>::const_iterator it = deviceQueues_.find(&dev);
|
||||
if (it != deviceQueues_.end()) {
|
||||
const auto it = deviceQueues_.find(&dev);
|
||||
if (it != deviceQueues_.cend()) {
|
||||
return it->second.defDeviceQueue_;
|
||||
} else {
|
||||
return NULL;
|
||||
|
||||
@@ -11,7 +11,7 @@
|
||||
#include "platform/agent.hpp"
|
||||
|
||||
#include <vector>
|
||||
#include <map>
|
||||
#include <unordered_map>
|
||||
|
||||
namespace amd {
|
||||
|
||||
@@ -197,8 +197,8 @@ class Context : public RuntimeObject {
|
||||
GLFunctions* glenv_; //!< OpenGL context
|
||||
Device* customHostAllocDevice_; //!< Device responsible for host allocations
|
||||
std::vector<Device*> svmAllocDevice_; //!< Devices can support SVM allocations
|
||||
std::map<const Device*, DeviceQueueInfo> deviceQueues_; //!< Device queues mapping
|
||||
mutable Monitor ctxLock_; //!< Lock for the context access
|
||||
std::unordered_map<const Device*, DeviceQueueInfo> deviceQueues_; //!< Device queues mapping
|
||||
mutable Monitor ctxLock_; //!< Lock for the context access
|
||||
};
|
||||
|
||||
/*! @}
|
||||
|
||||
@@ -18,6 +18,7 @@
|
||||
#include <vector>
|
||||
#include <list>
|
||||
#include <map>
|
||||
#include <unordered_map>
|
||||
|
||||
namespace device {
|
||||
class Memory;
|
||||
@@ -124,7 +125,7 @@ class Memory : public amd::RuntimeObject {
|
||||
DeviceMemory* deviceMemories_;
|
||||
|
||||
//! The device alloced state
|
||||
std::map<const Device*, AllocState> deviceAlloced_;
|
||||
std::unordered_map<const Device*, AllocState> deviceAlloced_;
|
||||
|
||||
//! Linked list of destructor callbacks.
|
||||
std::atomic<DestructorCallBackEntry*> destructorCallbacks_;
|
||||
|
||||
@@ -24,7 +24,7 @@ namespace amd {
|
||||
*/
|
||||
class PerfCounter : public RuntimeObject {
|
||||
public:
|
||||
typedef std::map<cl_perfcounter_property, ulong> Properties;
|
||||
typedef std::unordered_map<cl_perfcounter_property, ulong> Properties;
|
||||
|
||||
//! Constructor of the performance counter object
|
||||
PerfCounter(const Device& device, //!< device object
|
||||
|
||||
@@ -21,13 +21,12 @@ namespace amd {
|
||||
|
||||
Program::~Program() {
|
||||
// Destroy all device programs
|
||||
deviceprograms_t::const_iterator it, itEnd;
|
||||
for (it = devicePrograms_.begin(), itEnd = devicePrograms_.end(); it != itEnd; ++it) {
|
||||
delete it->second;
|
||||
for (const auto& it : devicePrograms_) {
|
||||
delete it.second;
|
||||
}
|
||||
|
||||
for (devicebinary_t::const_iterator IT = binary_.begin(), IE = binary_.end(); IT != IE; ++IT) {
|
||||
const binary_t& Bin = IT->second;
|
||||
for (const auto& it : binary_) {
|
||||
const binary_t& Bin = it.second;
|
||||
if (Bin.first) {
|
||||
delete[] Bin.first;
|
||||
}
|
||||
@@ -43,8 +42,8 @@ const Symbol* Program::findSymbol(const char* kernelName) const {
|
||||
return NULL;
|
||||
}
|
||||
|
||||
symbols_t::const_iterator it = symbolTable_->find(kernelName);
|
||||
return (it == symbolTable_->end()) ? NULL : &it->second;
|
||||
const auto it = symbolTable_->find(kernelName);
|
||||
return (it == symbolTable_->cend()) ? NULL : &it->second;
|
||||
}
|
||||
|
||||
cl_int Program::addDeviceProgram(Device& device, const void* image, size_t length,
|
||||
@@ -151,8 +150,8 @@ cl_int Program::addDeviceProgram(Device& device, const void* image, size_t lengt
|
||||
}
|
||||
|
||||
device::Program* Program::getDeviceProgram(const Device& device) const {
|
||||
deviceprograms_t::const_iterator it = devicePrograms_.find(&device.rootDevice());
|
||||
if (it == devicePrograms_.end()) {
|
||||
const auto it = devicePrograms_.find(&device.rootDevice());
|
||||
if (it == devicePrograms_.cend()) {
|
||||
return NULL;
|
||||
}
|
||||
return it->second;
|
||||
@@ -198,16 +197,15 @@ cl_int Program::compile(const std::vector<Device*>& devices, size_t numHeaders,
|
||||
}
|
||||
|
||||
// Compile the program programs associated with the given devices.
|
||||
std::vector<Device*>::const_iterator it;
|
||||
for (it = devices.begin(); it != devices.end(); ++it) {
|
||||
device::Program* devProgram = getDeviceProgram(**it);
|
||||
for (const auto& it : devices) {
|
||||
device::Program* devProgram = getDeviceProgram(*it);
|
||||
if (devProgram == NULL) {
|
||||
const binary_t& bin = binary(**it);
|
||||
retval = addDeviceProgram(**it, bin.first, bin.second, &parsedOptions);
|
||||
const binary_t& bin = binary(*it);
|
||||
retval = addDeviceProgram(*it, bin.first, bin.second, &parsedOptions);
|
||||
if (retval != CL_SUCCESS) {
|
||||
return retval;
|
||||
}
|
||||
devProgram = getDeviceProgram(**it);
|
||||
devProgram = getDeviceProgram(*it);
|
||||
}
|
||||
|
||||
if (devProgram->type() == device::Program::TYPE_INTERMEDIATE || language_ == SPIRV) {
|
||||
@@ -277,8 +275,7 @@ cl_int Program::link(const std::vector<Device*>& devices, size_t numInputs,
|
||||
}
|
||||
|
||||
// Link the program programs associated with the given devices.
|
||||
std::vector<Device*>::const_iterator it;
|
||||
for (it = devices.begin(); it != devices.end(); ++it) {
|
||||
for (const auto& it : devices) {
|
||||
// find the corresponding device program in each input program
|
||||
std::vector<device::Program*> inputDevPrograms(numInputs);
|
||||
bool found = false;
|
||||
@@ -288,8 +285,8 @@ cl_int Program::link(const std::vector<Device*>& devices, size_t numInputs,
|
||||
parsedOptions.oVariables->BinaryIsSpirv = true;
|
||||
}
|
||||
deviceprograms_t inputDevProgs = inputProgram.devicePrograms();
|
||||
deviceprograms_t::const_iterator findIt = inputDevProgs.find(*it);
|
||||
if (findIt == inputDevProgs.end()) {
|
||||
const auto findIt = inputDevProgs.find(it);
|
||||
if (findIt == inputDevProgs.cend()) {
|
||||
if (found) break;
|
||||
continue;
|
||||
}
|
||||
@@ -328,14 +325,14 @@ cl_int Program::link(const std::vector<Device*>& devices, size_t numInputs,
|
||||
return CL_INVALID_VALUE;
|
||||
}
|
||||
|
||||
device::Program* devProgram = getDeviceProgram(**it);
|
||||
device::Program* devProgram = getDeviceProgram(*it);
|
||||
if (devProgram == NULL) {
|
||||
const binary_t& bin = binary(**it);
|
||||
retval = addDeviceProgram(**it, bin.first, bin.second, &parsedOptions);
|
||||
const binary_t& bin = binary(*it);
|
||||
retval = addDeviceProgram(*it, bin.first, bin.second, &parsedOptions);
|
||||
if (retval != CL_SUCCESS) {
|
||||
return retval;
|
||||
}
|
||||
devProgram = getDeviceProgram(**it);
|
||||
devProgram = getDeviceProgram(*it);
|
||||
}
|
||||
|
||||
// We only build a Device-Program once
|
||||
@@ -359,16 +356,14 @@ cl_int Program::link(const std::vector<Device*>& devices, size_t numInputs,
|
||||
}
|
||||
|
||||
// Rebuild the symbol table
|
||||
deviceprograms_t::iterator sit;
|
||||
for (sit = devicePrograms_.begin(); sit != devicePrograms_.end(); ++sit) {
|
||||
const Device& device = *sit->first;
|
||||
const device::Program& program = *sit->second;
|
||||
for (const auto& sit : devicePrograms_) {
|
||||
const Device& device = *(sit.first);
|
||||
const device::Program& program = *(sit.second);
|
||||
|
||||
const device::Program::kernels_t& kernels = program.kernels();
|
||||
device::Program::kernels_t::const_iterator kit;
|
||||
for (kit = kernels.begin(); kit != kernels.end(); ++kit) {
|
||||
const std::string& name = kit->first;
|
||||
const device::Kernel* devKernel = kit->second;
|
||||
for (const auto& it : kernels) {
|
||||
const std::string& name = it.first;
|
||||
const device::Kernel* devKernel = it.second;
|
||||
|
||||
Symbol& symbol = (*symbolTable_)[name];
|
||||
if (!symbol.setDeviceKernel(device, devKernel)) {
|
||||
@@ -379,9 +374,8 @@ cl_int Program::link(const std::vector<Device*>& devices, size_t numInputs,
|
||||
|
||||
// Create a string with all kernel names from the program
|
||||
if (kernelNames_.length() == 0) {
|
||||
amd::Program::symbols_t::const_iterator it;
|
||||
for (it = symbols().begin(); it != symbols().end(); ++it) {
|
||||
if (it != symbols().begin()) {
|
||||
for (auto it = symbols().cbegin(); it != symbols().cend(); ++it) {
|
||||
if (it != symbols().cbegin()) {
|
||||
kernelNames_.append(1, ';');
|
||||
}
|
||||
kernelNames_.append(it->first.c_str());
|
||||
@@ -474,20 +468,19 @@ cl_int Program::build(const std::vector<Device*>& devices, const char* options,
|
||||
}
|
||||
|
||||
// Build the program programs associated with the given devices.
|
||||
std::vector<Device*>::const_iterator it;
|
||||
for (it = devices.begin(); it != devices.end(); ++it) {
|
||||
device::Program* devProgram = getDeviceProgram(**it);
|
||||
for (const auto& it : devices) {
|
||||
device::Program* devProgram = getDeviceProgram(*it);
|
||||
if (devProgram == NULL) {
|
||||
const binary_t& bin = binary(**it);
|
||||
const binary_t& bin = binary(*it);
|
||||
if (sourceCode_.empty() && (bin.first == NULL)) {
|
||||
retval = false;
|
||||
continue;
|
||||
}
|
||||
retval = addDeviceProgram(**it, bin.first, bin.second, &parsedOptions);
|
||||
retval = addDeviceProgram(*it, bin.first, bin.second, &parsedOptions);
|
||||
if (retval != CL_SUCCESS) {
|
||||
return retval;
|
||||
}
|
||||
devProgram = getDeviceProgram(**it);
|
||||
devProgram = getDeviceProgram(*it);
|
||||
}
|
||||
|
||||
parsedOptions.oVariables->AssumeAlias = true;
|
||||
@@ -518,16 +511,14 @@ cl_int Program::build(const std::vector<Device*>& devices, const char* options,
|
||||
}
|
||||
|
||||
// Rebuild the symbol table
|
||||
deviceprograms_t::iterator sit;
|
||||
for (sit = devicePrograms_.begin(); sit != devicePrograms_.end(); ++sit) {
|
||||
const Device& device = *sit->first;
|
||||
const device::Program& program = *sit->second;
|
||||
for (const auto& it : devicePrograms_) {
|
||||
const Device& device = *(it.first);
|
||||
const device::Program& program = *(it.second);
|
||||
|
||||
const device::Program::kernels_t& kernels = program.kernels();
|
||||
device::Program::kernels_t::const_iterator kit;
|
||||
for (kit = kernels.begin(); kit != kernels.end(); ++kit) {
|
||||
const std::string& name = kit->first;
|
||||
const device::Kernel* devKernel = kit->second;
|
||||
for (const auto& kit : kernels) {
|
||||
const std::string& name = kit.first;
|
||||
const device::Kernel* devKernel = kit.second;
|
||||
|
||||
Symbol& symbol = (*symbolTable_)[name];
|
||||
if (!symbol.setDeviceKernel(device, devKernel)) {
|
||||
@@ -538,9 +529,8 @@ cl_int Program::build(const std::vector<Device*>& devices, const char* options,
|
||||
|
||||
// Create a string with all kernel names from the program
|
||||
if (kernelNames_.length() == 0) {
|
||||
amd::Program::symbols_t::const_iterator it;
|
||||
for (it = symbols().begin(); it != symbols().end(); ++it) {
|
||||
if (it != symbols().begin()) {
|
||||
for (auto it = symbols().cbegin(); it != symbols().cend(); ++it) {
|
||||
if (it != symbols().cbegin()) {
|
||||
kernelNames_.append(1, ';');
|
||||
}
|
||||
kernelNames_.append(it->first.c_str());
|
||||
@@ -555,12 +545,10 @@ cl_int Program::build(const std::vector<Device*>& devices, const char* options,
|
||||
}
|
||||
|
||||
void Program::clear() {
|
||||
deviceprograms_t::iterator sit;
|
||||
|
||||
// Destroy old programs if we have any
|
||||
for (sit = devicePrograms_.begin(); sit != devicePrograms_.end(); ++sit) {
|
||||
for (const auto& it : devicePrograms_) {
|
||||
// Destroy device program
|
||||
delete sit->second;
|
||||
delete it.second;
|
||||
}
|
||||
|
||||
devicePrograms_.clear();
|
||||
@@ -631,13 +619,13 @@ bool Symbol::setDeviceKernel(const Device& device, const device::Kernel* func, b
|
||||
|
||||
const device::Kernel* Symbol::getDeviceKernel(const Device& device, bool noAlias) const {
|
||||
const devicekernels_t* devKernels = (noAlias) ? &deviceKernels_ : &devKernelsNoOpt_;
|
||||
devicekernels_t::const_iterator itEnd = devKernels->end();
|
||||
devicekernels_t::const_iterator it = devKernels->find(&device);
|
||||
const auto itEnd = devKernels->cend();
|
||||
auto it = devKernels->find(&device);
|
||||
if (it != itEnd) {
|
||||
return it->second;
|
||||
}
|
||||
|
||||
for (it = devKernels->begin(); it != itEnd; ++it) {
|
||||
for (it = devKernels->cbegin(); it != itEnd; ++it) {
|
||||
if (it->first->isAncestor(&device)) {
|
||||
return it->second;
|
||||
}
|
||||
|
||||
@@ -35,7 +35,7 @@ namespace amd {
|
||||
//! A kernel function symbol
|
||||
class Symbol : public HeapObject {
|
||||
public:
|
||||
typedef std::map<const Device*, const device::Kernel*> devicekernels_t;
|
||||
typedef std::unordered_map<const Device*, const device::Kernel*> devicekernels_t;
|
||||
|
||||
private:
|
||||
devicekernels_t deviceKernels_; //! All device kernels objects.
|
||||
@@ -68,9 +68,9 @@ class Program : public RuntimeObject {
|
||||
public:
|
||||
typedef std::pair<uint8_t*, size_t> binary_t;
|
||||
typedef std::set<Device const*> devicelist_t;
|
||||
typedef std::map<Device const*, binary_t> devicebinary_t;
|
||||
typedef std::map<Device const*, device::Program*> deviceprograms_t;
|
||||
typedef std::map<std::string, Symbol> symbols_t;
|
||||
typedef std::unordered_map<Device const*, binary_t> devicebinary_t;
|
||||
typedef std::unordered_map<Device const*, device::Program*> deviceprograms_t;
|
||||
typedef std::unordered_map<std::string, Symbol> symbols_t;
|
||||
|
||||
enum Language {
|
||||
Binary = 0,
|
||||
|
||||
@@ -14,7 +14,7 @@ namespace amd {
|
||||
//! Abstraction layer sampler class
|
||||
class Sampler : public RuntimeObject {
|
||||
public:
|
||||
typedef std::map<Device const*, device::Sampler*> DeviceSamplers;
|
||||
typedef std::unordered_map<Device const*, device::Sampler*> DeviceSamplers;
|
||||
|
||||
//! \note the sampler states must match the compiler's defines.
|
||||
//! See amd_ocl_sys_predef.c
|
||||
|
||||
Referência em uma Nova Issue
Bloquear um usuário