Ioannis Assiouras
30a14a8a05
Use fetch_sub(std::memory_order_acq_rel) on release so the destroying thread acquires prior writes.
152 lines
4.5 KiB
C++
152 lines
4.5 KiB
C++
/* Copyright (c) 2008 - 2022 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. */
|
|
|
|
#include "platform/runtime.hpp"
|
|
#include "os/os.hpp"
|
|
#include "thread/thread.hpp"
|
|
#include "device/device.hpp"
|
|
#include "utils/flags.hpp"
|
|
#include "utils/options.hpp"
|
|
#include "platform/context.hpp"
|
|
#include "platform/agent.hpp"
|
|
|
|
#include "platform/interop_gl.hpp"
|
|
|
|
#ifdef _WIN32
|
|
#include <d3d10_1.h>
|
|
#include <dxgi.h>
|
|
#include "CL/cl_d3d10.h"
|
|
#endif //_WIN32
|
|
|
|
#if defined(_MSC_VER) // both Win32 and Win64
|
|
#include <intrin.h>
|
|
#endif
|
|
|
|
#include <atomic>
|
|
#include <cstdlib>
|
|
#include <iostream>
|
|
|
|
namespace amd {
|
|
|
|
volatile bool Runtime::initialized_ = false;
|
|
bool Runtime::LibraryDetached = false;
|
|
volatile int Runtime::pid_ = 0;
|
|
|
|
bool Runtime::init() {
|
|
if (initialized_) {
|
|
return true;
|
|
}
|
|
|
|
// Enter a very basic critical region. We want to prevent 2 threads
|
|
// from concurrently executing the init() routines. We can't use a
|
|
// Monitor since the system is not yet initialized.
|
|
|
|
static std::atomic_flag lock = ATOMIC_FLAG_INIT;
|
|
struct CriticalRegion {
|
|
std::atomic_flag& lock_;
|
|
CriticalRegion(std::atomic_flag& lock) : lock_(lock) {
|
|
while (lock.test_and_set(std::memory_order_acquire)) {
|
|
Os::yield();
|
|
}
|
|
}
|
|
~CriticalRegion() { lock_.clear(std::memory_order_release); }
|
|
} region(lock);
|
|
|
|
if (initialized_) {
|
|
return true;
|
|
}
|
|
|
|
if (!Flag::init() || !option::init() ||
|
|
!Device::init()
|
|
// Agent initializes last
|
|
|| (!amd::IS_HIP && !Agent::init())) {
|
|
ClPrint(LOG_ERROR, LOG_INIT, "Runtime initialization failed");
|
|
return false;
|
|
}
|
|
|
|
ClPrint(LOG_INFO, LOG_MISC && !amd::IS_HIP, "ROCclr version: %s", ROCCLR_VERSION_GITHASH);
|
|
|
|
initialized_ = true;
|
|
pid_ = amd::Os::getProcessId();
|
|
return true;
|
|
}
|
|
|
|
void Runtime::tearDown() {
|
|
if (!initialized_) {
|
|
return;
|
|
}
|
|
|
|
Agent::tearDown();
|
|
Device::tearDown();
|
|
option::teardown();
|
|
Flag::tearDown();
|
|
if (outFile != stderr && outFile != nullptr) {
|
|
fclose(outFile);
|
|
}
|
|
Command::ReleaseSysmemPool();
|
|
initialized_ = false;
|
|
}
|
|
|
|
// ~RuntimeTearDown() will reference listenerLock.
|
|
// listenerLock will be constructed ealier and destructed later than
|
|
// runtime_tear_down.
|
|
amd::Monitor listenerLock("Hostcall listener lock");
|
|
std::vector<ReferenceCountedObject*> RuntimeTearDown::external_;
|
|
|
|
RuntimeTearDown::~RuntimeTearDown() {
|
|
#if !defined(_WIN32) && !defined(BUILD_STATIC_LIBS)
|
|
// Only perform destruction if process matches the initialization,
|
|
// to avoid a call with the child process after fork()
|
|
if (amd::IS_HIP && amd::Os::getProcessId() == Runtime::pid()) {
|
|
for (auto it : external_) {
|
|
it->release();
|
|
}
|
|
Runtime::tearDown();
|
|
}
|
|
#endif
|
|
}
|
|
|
|
void RuntimeTearDown::RegisterObject(ReferenceCountedObject* obj) { external_.push_back(obj); }
|
|
|
|
class RuntimeTearDown runtime_tear_down;
|
|
|
|
uint ReferenceCountedObject::retain() {
|
|
uint prev = referenceCount_.fetch_add(1, std::memory_order_relaxed);
|
|
assert(prev != 0 && "An object with count==0 is invalid");
|
|
return prev + 1;
|
|
}
|
|
|
|
uint ReferenceCountedObject::release() {
|
|
uint newCount = referenceCount_.fetch_sub(1, std::memory_order_acq_rel) - 1;
|
|
if (newCount == 0) {
|
|
if (terminate()) {
|
|
// The destructor should be called with a count==1 for the last thread
|
|
// releasing the reference. Since an atomic load before the decrement
|
|
// would add more atomic operations, simply bump the count to 1 here
|
|
// before destructing the object.
|
|
referenceCount_.store(1, std::memory_order_relaxed);
|
|
delete this;
|
|
}
|
|
}
|
|
return newCount;
|
|
}
|
|
|
|
} // namespace amd
|