Page MenuHomePhabricator
Differential D109276 Diff 370752 openmp/libomptarget/src/rtl.cpp

Changeset View

Standalone View

View Options

openmp/libomptarget/src/rtl.cpp

Show First 20 Lines • Show All 243 Lines • ▼ Show 20 Lines
//////////////////////////////////////////////////////////////////////////////// ////////////////////////////////////////////////////////////////////////////////
// Functionality for registering Ctors/Dtors // Functionality for registering Ctors/Dtors
static void RegisterGlobalCtorsDtorsForImage(__tgt_bin_desc *desc, static void RegisterGlobalCtorsDtorsForImage(__tgt_bin_desc *desc,
__tgt_device_image *img, __tgt_device_image *img,
RTLInfoTy *RTL) { RTLInfoTy *RTL) {
for (int32_t i = 0; i < RTL->NumberOfDevices; ++i) { for (int32_t i = 0; i < RTL->NumberOfDevices; ++i) {
DeviceTy &Device = PM->Devices[RTL->Idx + i]; DeviceTy &Device = *PM->Devices[RTL->Idx + i];
Device.PendingGlobalsMtx.lock(); Device.PendingGlobalsMtx.lock();
Device.HasPendingGlobals = true; Device.HasPendingGlobals = true;
for (__tgt_offload_entry *entry = img->EntriesBegin; for (__tgt_offload_entry *entry = img->EntriesBegin;
entry != img->EntriesEnd; ++entry) { entry != img->EntriesEnd; ++entry) {
if (entry->flags & OMP_DECLARE_TARGET_CTOR) { if (entry->flags & OMP_DECLARE_TARGET_CTOR) {
DP("Adding ctor " DPxMOD " to the pending list.\n", DP("Adding ctor " DPxMOD " to the pending list.\n",
DPxPTR(entry->addr)); DPxPTR(entry->addr));
Device.PendingCtorsDtors[desc].PendingCtors.push_back(entry->addr); Device.PendingCtorsDtors[desc].PendingCtors.push_back(entry->addr);
▲ Show 20 Lines • Show All 53 Lines • ▼ Show 20 Linesvoid RTLsTy::RegisterRequires(int64_t flags) {
DP("New requires flags %" PRId64 " compatible with existing %" PRId64 "!\n", DP("New requires flags %" PRId64 " compatible with existing %" PRId64 "!\n",
flags, RequiresFlags); flags, RequiresFlags);
} }
void RTLsTy::initRTLonce(RTLInfoTy &R) { void RTLsTy::initRTLonce(RTLInfoTy &R) {
// If this RTL is not already in use, initialize it. // If this RTL is not already in use, initialize it.
if (!R.isUsed && R.NumberOfDevices != 0) { if (!R.isUsed && R.NumberOfDevices != 0) {
// Initialize the device information for the RTL we are about to use. // Initialize the device information for the RTL we are about to use.
DeviceTy device(&R); const size_t Start = PM->Devices.size();
size_t Start = PM->Devices.size(); PM->Devices.reserve(Start + R.NumberOfDevices);
PM->Devices.resize(Start + R.NumberOfDevices, device);
for (int32_t device_id = 0; device_id < R.NumberOfDevices; device_id++) { for (int32_t device_id = 0; device_id < R.NumberOfDevices; device_id++) {
PM->Devices.push_back(std::make_unique<DeviceTy>(&R));
tianshilei1992Unsubmitted
Not Done
ReplyInline Actions

what about emplace_back?

tianshilei1992: what about `emplace_back`?
ye-luoAuthorUnsubmitted
Done
ReplyInline Actions

Both push_back and emplace_back do the same thing of invoking the move constructor of unique_ptr. However, emplace_back is less verbose in readability. So push_back is preferred here.

ye-luo: Both push_back and emplace_back do the same thing of invoking the move constructor of…
// global device ID // global device ID
PM->Devices[Start + device_id].DeviceID = Start + device_id; PM->Devices[Start + device_id]->DeviceID = Start + device_id;
// RTL local device ID // RTL local device ID
PM->Devices[Start + device_id].RTLDeviceID = device_id; PM->Devices[Start + device_id]->RTLDeviceID = device_id;
} }
// Initialize the index of this RTL and save it in the used RTLs. // Initialize the index of this RTL and save it in the used RTLs.
R.Idx = (UsedRTLs.empty()) R.Idx = (UsedRTLs.empty())
? 0 ? 0
: UsedRTLs.back()->Idx + UsedRTLs.back()->NumberOfDevices; : UsedRTLs.back()->Idx + UsedRTLs.back()->NumberOfDevices;
assert((size_t)R.Idx == Start && assert((size_t)R.Idx == Start &&
"RTL index should equal the number of devices used so far."); "RTL index should equal the number of devices used so far.");
▲ Show 20 Lines • Show All 94 Lines • ▼ Show 20 Lines for (auto *R : UsedRTLs) {
DP("Image " DPxMOD " is compatible with RTL " DPxMOD "!\n", DP("Image " DPxMOD " is compatible with RTL " DPxMOD "!\n",
DPxPTR(img->ImageStart), DPxPTR(R->LibraryHandler)); DPxPTR(img->ImageStart), DPxPTR(R->LibraryHandler));
FoundRTL = R; FoundRTL = R;
// Execute dtors for static objects if the device has been used, i.e. // Execute dtors for static objects if the device has been used, i.e.
// if its PendingCtors list has been emptied. // if its PendingCtors list has been emptied.
for (int32_t i = 0; i < FoundRTL->NumberOfDevices; ++i) { for (int32_t i = 0; i < FoundRTL->NumberOfDevices; ++i) {
DeviceTy &Device = PM->Devices[FoundRTL->Idx + i]; DeviceTy &Device = *PM->Devices[FoundRTL->Idx + i];
Device.PendingGlobalsMtx.lock(); Device.PendingGlobalsMtx.lock();
if (Device.PendingCtorsDtors[desc].PendingCtors.empty()) { if (Device.PendingCtorsDtors[desc].PendingCtors.empty()) {
AsyncInfoTy AsyncInfo(Device); AsyncInfoTy AsyncInfo(Device);
for (auto &dtor : Device.PendingCtorsDtors[desc].PendingDtors) { for (auto &dtor : Device.PendingCtorsDtors[desc].PendingDtors) {
int rc = target(nullptr, Device, dtor, 0, nullptr, nullptr, nullptr, int rc = target(nullptr, Device, dtor, 0, nullptr, nullptr, nullptr,
nullptr, nullptr, nullptr, 1, 1, true /*team*/, nullptr, nullptr, nullptr, 1, 1, true /*team*/,
AsyncInfo); AsyncInfo);
if (rc != OFFLOAD_SUCCESS) { if (rc != OFFLOAD_SUCCESS) {
▲ Show 20 Lines • Show All 54 Lines • Show Last 20 Lines