X-Git-Url: http://info.iut-bm.univ-fcomte.fr/pub/gitweb/simgrid.git/blobdiff_plain/7cd237f9659b2df47fca65a4340ca7b5407f46a0..HEAD:/src/smpi/mpi/smpi_win.cpp diff --git a/src/smpi/mpi/smpi_win.cpp b/src/smpi/mpi/smpi_win.cpp index 41c17e7aa8..af63fe62d0 100644 --- a/src/smpi/mpi/smpi_win.cpp +++ b/src/smpi/mpi/smpi_win.cpp @@ -1,8 +1,9 @@ -/* Copyright (c) 2007-2019. The SimGrid Team. All rights reserved. */ +/* Copyright (c) 2007-2023. The SimGrid Team. All rights reserved. */ /* This program is free software; you can redistribute it and/or modify it * under the terms of the license (GNU LGPL) which comes with this package. */ +#include #include "smpi_win.hpp" #include "private.hpp" @@ -13,21 +14,39 @@ #include "smpi_keyvals.hpp" #include "smpi_request.hpp" #include "src/smpi/include/smpi_actor.hpp" +#include "src/mc/mc_replay.hpp" + +#include +#include // std::scoped_lock XBT_LOG_NEW_DEFAULT_SUBCATEGORY(smpi_rma, smpi, "Logging specific to SMPI (RMA operations)"); +#define CHECK_RMA_REMOTE_WIN(fun, win)\ + if(target_count*target_datatype->get_extent()>win->size_){\ + XBT_WARN("%s: Trying to move %zd, which exceeds the window size on target process %d : %zd - Bailing out.",\ + fun, target_count*target_datatype->get_extent(), target_rank, win->size_);\ + simgrid::smpi::utils::set_current_buffer(1,"win_base",win->base_);\ + return MPI_ERR_RMA_RANGE;\ + } + +#define CHECK_WIN_LOCKED(win) \ + if (opened_ == 0) { /*check that post/start has been done*/ \ + bool locked = std::any_of(begin(win->lockers_), end(win->lockers_), [this](int it) { return it == this->rank_; }); \ + if (not locked) \ + return MPI_ERR_WIN; \ + } -namespace simgrid{ -namespace smpi{ +namespace simgrid::smpi { std::unordered_map Win::keyvals_; int Win::keyval_id_=0; -Win::Win(void* base, MPI_Aint size, int disp_unit, MPI_Info info, MPI_Comm comm, int allocated, int dynamic) +Win::Win(void* base, MPI_Aint size, int disp_unit, MPI_Info info, MPI_Comm comm, bool allocated, bool dynamic) : base_(base) , size_(size) , disp_unit_(disp_unit) , info_(info) , comm_(comm) + , connected_wins_(comm->size()) , rank_(comm->rank()) , allocated_(allocated) , dynamic_(dynamic) @@ -35,70 +54,65 @@ Win::Win(void* base, MPI_Aint size, int disp_unit, MPI_Info info, MPI_Comm comm, XBT_DEBUG("Creating window"); if(info!=MPI_INFO_NULL) info->ref(); - int comm_size = comm->size(); - name_ = nullptr; - opened_ = 0; - group_ = MPI_GROUP_NULL; - requests_ = new std::vector(); - mut_ = s4u::Mutex::create(); - lock_mut_ = s4u::Mutex::create(); - atomic_mut_ = s4u::Mutex::create(); - connected_wins_ = new MPI_Win[comm_size]; connected_wins_[rank_] = this; - count_ = 0; - if(rank_==0){ - bar_ = new s4u::Barrier(comm_size); - } - mode_=0; - errhandler_=MPI_ERRORS_ARE_FATAL; errhandler_->ref(); comm->add_rma_win(this); comm->ref(); - colls::allgather(&(connected_wins_[rank_]), sizeof(MPI_Win), MPI_BYTE, connected_wins_, sizeof(MPI_Win), MPI_BYTE, - comm); - - colls::bcast(&(bar_), sizeof(s4u::Barrier*), MPI_BYTE, 0, comm); - - colls::barrier(comm); + colls::allgather(&connected_wins_[rank_], sizeof(MPI_Win), MPI_BYTE, connected_wins_.data(), sizeof(MPI_Win), + MPI_BYTE, comm); + if (MC_is_active() || MC_record_replay_is_active()){ + s4u::Barrier* bar_ptr; + if (rank_ == 0) { + bar_ = s4u::Barrier::create(comm->size()); + bar_ptr = bar_.get(); + } + colls::bcast(&bar_ptr, sizeof(s4u::Barrier*), MPI_BYTE, 0, comm); + if (rank_ != 0) + bar_ = s4u::BarrierPtr(bar_ptr); + } + this->add_f(); } -Win::~Win(){ +int Win::del(Win* win){ //As per the standard, perform a barrier to ensure every async comm is finished - bar_->wait(); - - int finished = finish_comms(); - XBT_DEBUG("Win destructor - Finished %d RMA calls", finished); - - delete requests_; - delete[] connected_wins_; - if (name_ != nullptr){ - xbt_free(name_); + if (MC_is_active() || MC_record_replay_is_active()) + win->bar_->wait(); + else + colls::barrier(win->comm_); + win->flush_local_all(); + + if (win->info_ != MPI_INFO_NULL) + simgrid::smpi::Info::unref(win->info_); + if (win->errhandler_ != MPI_ERRHANDLER_NULL) + simgrid::smpi::Errhandler::unref(win->errhandler_); + + win->comm_->remove_rma_win(win); + + colls::barrier(win->comm_); + Comm::unref(win->comm_); + if (not win->lockers_.empty() || win->opened_ < 0) { + XBT_WARN("Freeing a locked or opened window"); + return MPI_ERR_WIN; } - if (info_ != MPI_INFO_NULL) - simgrid::smpi::Info::unref(info_); - if (errhandler_ != MPI_ERRHANDLER_NULL) - simgrid::smpi::Errhandler::unref(errhandler_); + if (win->allocated_) + xbt_free(win->base_); + for (auto m : {win->mut_, win->lock_mut_, win->atomic_mut_}) + if (m->get_owner() != nullptr) + m->unlock(); - comm_->remove_rma_win(this); + F2C::free_f(win->f2c_id()); + win->cleanup_attr(); - colls::barrier(comm_); - Comm::unref(comm_); - - if (rank_ == 0) - delete bar_; - - if(allocated_ !=0) - xbt_free(base_); - - cleanup_attr(); + delete win; + return MPI_SUCCESS; } int Win::attach(void* /*base*/, MPI_Aint size) { - if (not(base_ == MPI_BOTTOM || base_ == 0)) + if (not(base_ == MPI_BOTTOM || base_ == nullptr)) return MPI_ERR_ARG; - base_=0;//actually the address will be given in the RMA calls, as being the disp. + base_ = nullptr; // actually the address will be given in the RMA calls, as being the disp. size_+=size; return MPI_SUCCESS; } @@ -110,14 +124,13 @@ int Win::detach(const void* /*base*/) return MPI_SUCCESS; } -void Win::get_name(char* name, int* length){ - if(name_==nullptr){ - *length=0; - name=nullptr; - return; +void Win::get_name(char* name, int* length) const +{ + *length = static_cast(name_.length()); + if (not name_.empty()) { + name_.copy(name, *length); + name[*length] = '\0'; } - *length = strlen(name_); - strncpy(name, name_, *length+1); } void Win::get_group(MPI_Group* group){ @@ -130,29 +143,36 @@ void Win::get_group(MPI_Group* group){ MPI_Info Win::info() { - if (info_ == MPI_INFO_NULL) - info_ = new Info(); - info_->ref(); return info_; } -int Win::rank(){ +int Win::rank() const +{ return rank_; } -MPI_Aint Win::size(){ +MPI_Comm Win::comm() const +{ + return comm_; +} + +MPI_Aint Win::size() const +{ return size_; } -void* Win::base(){ +void* Win::base() const +{ return base_; } -int Win::disp_unit(){ +int Win::disp_unit() const +{ return disp_unit_; } -int Win::dynamic(){ +bool Win::dynamic() const +{ return dynamic_; } @@ -166,38 +186,30 @@ void Win::set_info(MPI_Info info) } void Win::set_name(const char* name){ - name_ = xbt_strdup(name); + name_ = name; } int Win::fence(int assert) { XBT_DEBUG("Entering fence"); - if (opened_ == 0) - opened_=1; - if (assert != MPI_MODE_NOPRECEDE) { + opened_++; + if (not (assert & MPI_MODE_NOPRECEDE)) { // This is not the first fence => finalize what came before - bar_->wait(); - mut_->lock(); - // This (simulated) mutex ensures that no process pushes to the vector of requests during the waitall. - // Without this, the vector could get redimensioned when another process pushes. - // This would result in the array used by Request::waitall() to be invalidated. - // Another solution would be to copy the data and cleanup the vector *before* Request::waitall - std::vector *reqs = requests_; - int size = static_cast(reqs->size()); - // start all requests that have been prepared by another process - if (size > 0) { - MPI_Request* treqs = &(*reqs)[0]; - Request::waitall(size, treqs, MPI_STATUSES_IGNORE); - } + if (MC_is_active() || MC_record_replay_is_active()) + bar_->wait(); + else + colls::barrier(comm_); + flush_local_all(); count_=0; - mut_->unlock(); } - if(assert==MPI_MODE_NOSUCCEED)//there should be no ops after this one, tell we are closed. + if (assert & MPI_MODE_NOSUCCEED) // there should be no ops after this one, tell we are closed. opened_=0; assert_ = assert; - - bar_->wait(); + if (MC_is_active() || MC_record_replay_is_active()) + bar_->wait(); + else + colls::barrier(comm_); XBT_DEBUG("Leaving fence"); return MPI_SUCCESS; @@ -207,35 +219,23 @@ int Win::put(const void *origin_addr, int origin_count, MPI_Datatype origin_data MPI_Aint target_disp, int target_count, MPI_Datatype target_datatype, MPI_Request* request) { //get receiver pointer - const Win* recv_win = connected_wins_[target_rank]; - - if(opened_==0){//check that post/start has been done - // no fence or start .. lock ok ? - int locked=0; - for (auto const& it : recv_win->lockers_) - if (it == comm_->rank()) - locked = 1; - if(locked != 1) - return MPI_ERR_WIN; - } + Win* recv_win = connected_wins_[target_rank]; - if(target_count*target_datatype->get_extent()>recv_win->size_) - return MPI_ERR_ARG; + CHECK_WIN_LOCKED(recv_win) + CHECK_RMA_REMOTE_WIN("MPI_Put", recv_win) - void* recv_addr = static_cast ( static_cast(recv_win->base_) + target_disp * recv_win->disp_unit_); + void* recv_addr = static_cast(recv_win->base_) + target_disp * recv_win->disp_unit_; - if (target_rank != comm_->rank()) { // This is not for myself, so we need to send messages + if (target_rank != rank_) { // This is not for myself, so we need to send messages XBT_DEBUG("Entering MPI_Put to remote rank %d", target_rank); // prepare send_request MPI_Request sreq = - // TODO cheinrich Check for rank / pid conversion - Request::rma_send_init(origin_addr, origin_count, origin_datatype, comm_->rank(), target_rank, SMPI_RMA_TAG + 1, - comm_, MPI_OP_NULL); + Request::rma_send_init(origin_addr, origin_count, origin_datatype, rank_, target_rank, SMPI_RMA_TAG + 1, comm_, + MPI_OP_NULL); //prepare receiver request - // TODO cheinrich Check for rank / pid conversion - MPI_Request rreq = Request::rma_recv_init(recv_addr, target_count, target_datatype, recv_win->comm_->rank(), - target_rank, SMPI_RMA_TAG + 1, recv_win->comm_, MPI_OP_NULL); + MPI_Request rreq = Request::rma_recv_init(recv_addr, target_count, target_datatype, rank_, target_rank, + SMPI_RMA_TAG + 1, recv_win->comm_, MPI_OP_NULL); //start send sreq->start(); @@ -243,16 +243,14 @@ int Win::put(const void *origin_addr, int origin_count, MPI_Datatype origin_data if(request!=nullptr){ *request=sreq; }else{ - mut_->lock(); - requests_->push_back(sreq); - mut_->unlock(); + const std::scoped_lock lock(*mut_); + requests_.push_back(sreq); } //push request to receiver's win - recv_win->mut_->lock(); - recv_win->requests_->push_back(rreq); + const std::scoped_lock recv_lock(*recv_win->mut_); + recv_win->requests_.push_back(rreq); rreq->start(); - recv_win->mut_->unlock(); } else { XBT_DEBUG("Entering MPI_Put from myself to myself, rank %d", target_rank); Datatype::copy(origin_addr, origin_count, origin_datatype, recv_addr, target_count, target_datatype); @@ -267,41 +265,29 @@ int Win::get( void *origin_addr, int origin_count, MPI_Datatype origin_datatype, MPI_Aint target_disp, int target_count, MPI_Datatype target_datatype, MPI_Request* request) { //get sender pointer - const Win* send_win = connected_wins_[target_rank]; - - if(opened_==0){//check that post/start has been done - // no fence or start .. lock ok ? - int locked=0; - for (auto const& it : send_win->lockers_) - if (it == comm_->rank()) - locked = 1; - if(locked != 1) - return MPI_ERR_WIN; - } + Win* send_win = connected_wins_[target_rank]; - if(target_count*target_datatype->get_extent()>send_win->size_) - return MPI_ERR_ARG; + CHECK_WIN_LOCKED(send_win) + CHECK_RMA_REMOTE_WIN("MPI_Get", send_win) const void* send_addr = static_cast(static_cast(send_win->base_) + target_disp * send_win->disp_unit_); XBT_DEBUG("Entering MPI_Get from %d", target_rank); - if(target_rank != comm_->rank()){ + if (target_rank != rank_) { //prepare send_request - MPI_Request sreq = Request::rma_send_init(send_addr, target_count, target_datatype, target_rank, - send_win->comm_->rank(), SMPI_RMA_TAG + 2, send_win->comm_, MPI_OP_NULL); + MPI_Request sreq = Request::rma_send_init(send_addr, target_count, target_datatype, target_rank, rank_, + SMPI_RMA_TAG + 2, send_win->comm_, MPI_OP_NULL); //prepare receiver request - MPI_Request rreq = Request::rma_recv_init( - origin_addr, origin_count, origin_datatype, target_rank, - comm_->rank(), // TODO cheinrich Check here if comm_->rank() and above send_win->comm_->rank() are correct - SMPI_RMA_TAG + 2, comm_, MPI_OP_NULL); + MPI_Request rreq = Request::rma_recv_init(origin_addr, origin_count, origin_datatype, target_rank, rank_, + SMPI_RMA_TAG + 2, comm_, MPI_OP_NULL); //start the send, with another process than us as sender. sreq->start(); - //push request to receiver's win - send_win->mut_->lock(); - send_win->requests_->push_back(sreq); - send_win->mut_->unlock(); + // push request to sender's win + if (const std::scoped_lock send_lock(*send_win->mut_); true) { + send_win->requests_.push_back(sreq); + } //start recv rreq->start(); @@ -309,9 +295,8 @@ int Win::get( void *origin_addr, int origin_count, MPI_Datatype origin_datatype, if(request!=nullptr){ *request=rreq; }else{ - mut_->lock(); - requests_->push_back(rreq); - mut_->unlock(); + const std::scoped_lock lock(*mut_); + requests_.push_back(rreq); } } else { Datatype::copy(send_addr, target_count, target_datatype, origin_addr, origin_count, origin_datatype); @@ -326,53 +311,45 @@ int Win::accumulate(const void *origin_addr, int origin_count, MPI_Datatype orig { XBT_DEBUG("Entering MPI_Win_Accumulate"); //get receiver pointer - const Win* recv_win = connected_wins_[target_rank]; - - if(opened_==0){//check that post/start has been done - // no fence or start .. lock ok ? - int locked=0; - for (auto const& it : recv_win->lockers_) - if (it == comm_->rank()) - locked = 1; - if(locked != 1) - return MPI_ERR_WIN; - } - //FIXME: local version + Win* recv_win = connected_wins_[target_rank]; - if(target_count*target_datatype->get_extent()>recv_win->size_) - return MPI_ERR_ARG; + //FIXME: local version + CHECK_WIN_LOCKED(recv_win) + CHECK_RMA_REMOTE_WIN("MPI_Accumulate", recv_win) - void* recv_addr = static_cast(static_cast(recv_win->base_) + target_disp * recv_win->disp_unit_); + void* recv_addr = static_cast(recv_win->base_) + target_disp * recv_win->disp_unit_; XBT_DEBUG("Entering MPI_Accumulate to %d", target_rank); // As the tag will be used for ordering of the operations, subtract count from it (to avoid collisions with other // SMPI tags, SMPI_RMA_TAG is set below all the other ones we use) // prepare send_request - MPI_Request sreq = Request::rma_send_init(origin_addr, origin_count, origin_datatype, comm_->rank(), target_rank, + MPI_Request sreq = Request::rma_send_init(origin_addr, origin_count, origin_datatype, rank_, target_rank, SMPI_RMA_TAG - 3 - count_, comm_, op); // prepare receiver request - MPI_Request rreq = Request::rma_recv_init(recv_addr, target_count, target_datatype, recv_win->comm_->rank(), - recv_win->comm_->group()->rank(comm_->group()->actor(target_rank)), SMPI_RMA_TAG - 3 - count_, recv_win->comm_, op); + MPI_Request rreq = Request::rma_recv_init(recv_addr, target_count, target_datatype, rank_, target_rank, + SMPI_RMA_TAG - 3 - count_, recv_win->comm_, op); count_++; // start send sreq->start(); // push request to receiver's win - recv_win->mut_->lock(); - recv_win->requests_->push_back(rreq); - rreq->start(); - recv_win->mut_->unlock(); + if (const std::scoped_lock recv_lock(*recv_win->mut_); true) { + recv_win->requests_.push_back(rreq); + rreq->start(); + } if (request != nullptr) { *request = sreq; } else { - mut_->lock(); - requests_->push_back(sreq); - mut_->unlock(); + const std::scoped_lock lock(*mut_); + requests_.push_back(sreq); } + // FIXME: The current implementation fails to ensure the correct ordering of the accumulate requests. The following + // 'flush' is a workaround to fix that. + flush(target_rank); XBT_DEBUG("Leaving MPI_Win_Accumulate"); return MPI_SUCCESS; } @@ -384,23 +361,13 @@ int Win::get_accumulate(const void* origin_addr, int origin_count, MPI_Datatype //get sender pointer const Win* send_win = connected_wins_[target_rank]; - if(opened_==0){//check that post/start has been done - // no fence or start .. lock ok ? - int locked=0; - for (auto const& it : send_win->lockers_) - if (it == comm_->rank()) - locked = 1; - if(locked != 1) - return MPI_ERR_WIN; - } - - if(target_count*target_datatype->get_extent()>send_win->size_) - return MPI_ERR_ARG; + CHECK_WIN_LOCKED(send_win) + CHECK_RMA_REMOTE_WIN("MPI_Get_Accumulate", send_win) XBT_DEBUG("Entering MPI_Get_accumulate from %d", target_rank); //need to be sure ops are correctly ordered, so finish request here ? slow. - MPI_Request req; - send_win->atomic_mut_->lock(); + MPI_Request req = MPI_REQUEST_NULL; + const std::scoped_lock lock(*send_win->atomic_mut_); get(result_addr, result_count, result_datatype, target_rank, target_disp, target_count, target_datatype, &req); if (req != MPI_REQUEST_NULL) @@ -410,7 +377,6 @@ int Win::get_accumulate(const void* origin_addr, int origin_count, MPI_Datatype target_disp, target_count, target_datatype, op, &req); if (req != MPI_REQUEST_NULL) Request::wait(&req, MPI_STATUS_IGNORE); - send_win->atomic_mut_->unlock(); return MPI_SUCCESS; } @@ -420,19 +386,11 @@ int Win::compare_and_swap(const void* origin_addr, const void* compare_addr, voi //get sender pointer const Win* send_win = connected_wins_[target_rank]; - if(opened_==0){//check that post/start has been done - // no fence or start .. lock ok ? - int locked=0; - for (auto const& it : send_win->lockers_) - if (it == comm_->rank()) - locked = 1; - if(locked != 1) - return MPI_ERR_WIN; - } + CHECK_WIN_LOCKED(send_win) XBT_DEBUG("Entering MPI_Compare_and_swap with %d", target_rank); MPI_Request req = MPI_REQUEST_NULL; - send_win->atomic_mut_->lock(); + const std::scoped_lock lock(*send_win->atomic_mut_); get(result_addr, 1, datatype, target_rank, target_disp, 1, datatype, &req); if (req != MPI_REQUEST_NULL) @@ -441,7 +399,6 @@ int Win::compare_and_swap(const void* origin_addr, const void* compare_addr, voi put(origin_addr, 1, datatype, target_rank, target_disp, 1, datatype); } - send_win->atomic_mut_->unlock(); return MPI_SUCCESS; } @@ -460,30 +417,24 @@ int Win::start(MPI_Group group, int /*assert*/) must complete, without further dependencies. */ //naive, blocking implementation. - int i = 0; - int j = 0; - int size = group->size(); - MPI_Request* reqs = xbt_new0(MPI_Request, size); - XBT_DEBUG("Entering MPI_Win_Start"); - while (j != size) { - int src = comm_->group()->rank(group->actor(j)); - if (src != rank_ && src != MPI_UNDEFINED) { // TODO cheinrich: The check of MPI_UNDEFINED should be useless here - reqs[i] = Request::irecv_init(nullptr, 0, MPI_CHAR, src, SMPI_RMA_TAG + 4, comm_); - i++; - } - j++; + std::vector reqs; + for (int i = 0; i < group->size(); i++) { + int src = comm_->group()->rank(group->actor(i)); + xbt_assert(src != MPI_UNDEFINED); + if (src != rank_) + reqs.emplace_back(Request::irecv_init(nullptr, 0, MPI_CHAR, src, SMPI_RMA_TAG + 4, comm_)); } - size = i; - Request::startall(size, reqs); - Request::waitall(size, reqs, MPI_STATUSES_IGNORE); - for (i = 0; i < size; i++) { - Request::unref(&reqs[i]); - } - xbt_free(reqs); - opened_++; //we're open for business ! - group_=group; + int size = static_cast(reqs.size()); + + Request::startall(size, reqs.data()); + Request::waitall(size, reqs.data(), MPI_STATUSES_IGNORE); + for (auto& req : reqs) + Request::unref(&req); + group->ref(); + dst_group_ = group; + opened_--; // we're open for business ! XBT_DEBUG("Leaving MPI_Win_Start"); return MPI_SUCCESS; } @@ -491,100 +442,78 @@ int Win::start(MPI_Group group, int /*assert*/) int Win::post(MPI_Group group, int /*assert*/) { //let's make a synchronous send here - int i = 0; - int j = 0; - int size = group->size(); - MPI_Request* reqs = xbt_new0(MPI_Request, size); - XBT_DEBUG("Entering MPI_Win_Post"); - while(j!=size){ - int dst = comm_->group()->rank(group->actor(j)); - if (dst != rank_ && dst != MPI_UNDEFINED) { - reqs[i] = Request::send_init(nullptr, 0, MPI_CHAR, dst, SMPI_RMA_TAG + 4, comm_); - i++; - } - j++; + std::vector reqs; + for (int i = 0; i < group->size(); i++) { + int dst = comm_->group()->rank(group->actor(i)); + xbt_assert(dst != MPI_UNDEFINED); + if (dst != rank_) + reqs.emplace_back(Request::send_init(nullptr, 0, MPI_CHAR, dst, SMPI_RMA_TAG + 4, comm_)); } - size=i; + int size = static_cast(reqs.size()); + + Request::startall(size, reqs.data()); + Request::waitall(size, reqs.data(), MPI_STATUSES_IGNORE); + for (auto& req : reqs) + Request::unref(&req); - Request::startall(size, reqs); - Request::waitall(size, reqs, MPI_STATUSES_IGNORE); - for(i=0;iref(); + src_group_ = group; + opened_--; // we're open for business ! XBT_DEBUG("Leaving MPI_Win_Post"); return MPI_SUCCESS; } int Win::complete(){ - if(opened_==0) - xbt_die("Complete called on already opened MPI_Win"); + xbt_assert(opened_ != 0, "Complete called on already opened MPI_Win"); XBT_DEBUG("Entering MPI_Win_Complete"); - int i = 0; - int j = 0; - int size = group_->size(); - MPI_Request* reqs = xbt_new0(MPI_Request, size); - - while(j!=size){ - int dst = comm_->group()->rank(group_->actor(j)); - if (dst != rank_ && dst != MPI_UNDEFINED) { - reqs[i] = Request::send_init(nullptr, 0, MPI_CHAR, dst, SMPI_RMA_TAG + 5, comm_); - i++; - } - j++; + std::vector reqs; + for (int i = 0; i < dst_group_->size(); i++) { + int dst = comm_->group()->rank(dst_group_->actor(i)); + xbt_assert(dst != MPI_UNDEFINED); + if (dst != rank_) + reqs.emplace_back(Request::send_init(nullptr, 0, MPI_CHAR, dst, SMPI_RMA_TAG + 5, comm_)); } - size=i; - XBT_DEBUG("Win_complete - Sending sync messages to %d processes", size); - Request::startall(size, reqs); - Request::waitall(size, reqs, MPI_STATUSES_IGNORE); + int size = static_cast(reqs.size()); - for(i=0;isize(); - MPI_Request* reqs = xbt_new0(MPI_Request, size); - - while(j!=size){ - int src = comm_->group()->rank(group_->actor(j)); - if (src != rank_ && src != MPI_UNDEFINED) { - reqs[i] = Request::irecv_init(nullptr, 0, MPI_CHAR, src, SMPI_RMA_TAG + 5, comm_); - i++; - } - j++; + std::vector reqs; + for (int i = 0; i < src_group_->size(); i++) { + int src = comm_->group()->rank(src_group_->actor(i)); + xbt_assert(src != MPI_UNDEFINED); + if (src != rank_) + reqs.emplace_back(Request::irecv_init(nullptr, 0, MPI_CHAR, src, SMPI_RMA_TAG + 5, comm_)); } - size=i; + int size = static_cast(reqs.size()); + XBT_DEBUG("Win_wait - Receiving sync messages from %d processes", size); - Request::startall(size, reqs); - Request::waitall(size, reqs, MPI_STATUSES_IGNORE); - for(i=0;imode_ == MPI_LOCK_SHARED && lock_type == MPI_LOCK_EXCLUSIVE)) target_win->mode_ += lock_type; // don't set to exclusive if it's already shared - target_win->lockers_.push_back(comm_->rank()); + target_win->lockers_.push_back(rank_); - int finished = finish_comms(rank); - XBT_DEBUG("Win_lock %d - Finished %d RMA calls", rank, finished); - finished = target_win->finish_comms(rank_); - XBT_DEBUG("Win_lock target %d - Finished %d RMA calls", rank, finished); + flush(rank); return MPI_SUCCESS; } int Win::lock_all(int assert){ - int i=0; int retval = MPI_SUCCESS; - for (i=0; isize();i++){ - int ret = this->lock(MPI_LOCK_SHARED, i, assert); - if(ret != MPI_SUCCESS) - retval = ret; + for (int i = 0; i < comm_->size(); i++) { + int ret = this->lock(MPI_LOCK_SHARED, i, assert); + if (ret != MPI_SUCCESS) + retval = ret; } return retval; } @@ -625,22 +550,18 @@ int Win::unlock(int rank){ MPI_Win target_win = connected_wins_[rank]; int target_mode = target_win->mode_; target_win->mode_= 0; - target_win->lockers_.remove(comm_->rank()); + target_win->lockers_.remove(rank_); if (target_mode==MPI_LOCK_EXCLUSIVE){ target_win->lock_mut_->unlock(); } - int finished = finish_comms(rank); - XBT_DEBUG("Win_unlock %d - Finished %d RMA calls", rank, finished); - finished = target_win->finish_comms(rank_); - XBT_DEBUG("Win_unlock target %d - Finished %d RMA calls", rank, finished); + flush(rank); return MPI_SUCCESS; } int Win::unlock_all(){ - int i=0; int retval = MPI_SUCCESS; - for (i=0; isize();i++){ + for (int i = 0; i < comm_->size(); i++) { int ret = this->unlock(i); if (ret != MPI_SUCCESS) retval = ret; @@ -649,33 +570,36 @@ int Win::unlock_all(){ } int Win::flush(int rank){ - MPI_Win target_win = connected_wins_[rank]; - int finished = finish_comms(rank_); - XBT_DEBUG("Win_flush on local %d - Finished %d RMA calls", rank_, finished); - finished = target_win->finish_comms(rank); - XBT_DEBUG("Win_flush on remote %d - Finished %d RMA calls", rank, finished); + int finished = finish_comms(rank); + XBT_DEBUG("Win_flush on local %d for remote %d - Finished %d RMA calls", rank_, rank, finished); + if (rank != rank_) { + finished = connected_wins_[rank]->finish_comms(rank_); + XBT_DEBUG("Win_flush on remote %d for local %d - Finished %d RMA calls", rank, rank_, finished); + } return MPI_SUCCESS; } int Win::flush_local(int rank){ int finished = finish_comms(rank); - XBT_DEBUG("Win_flush_local for rank %d - Finished %d RMA calls", rank, finished); + XBT_DEBUG("Win_flush_local on local %d for remote %d - Finished %d RMA calls", rank_, rank, finished); return MPI_SUCCESS; } int Win::flush_all(){ int finished = finish_comms(); - XBT_DEBUG("Win_flush_all on local - Finished %d RMA calls", finished); + XBT_DEBUG("Win_flush_all on local %d - Finished %d RMA calls", rank_, finished); for (int i = 0; i < comm_->size(); i++) { - finished = connected_wins_[i]->finish_comms(rank_); - XBT_DEBUG("Win_flush_all on %d - Finished %d RMA calls", i, finished); + if (i != rank_) { + finished = connected_wins_[i]->finish_comms(rank_); + XBT_DEBUG("Win_flush_all on remote %d for local %d - Finished %d RMA calls", i, rank_, finished); + } } return MPI_SUCCESS; } int Win::flush_local_all(){ int finished = finish_comms(); - XBT_DEBUG("Win_flush_local_all - Finished %d RMA calls", finished); + XBT_DEBUG("Win_flush_local_all on local %d - Finished %d RMA calls", rank_, finished); return MPI_SUCCESS; } @@ -684,53 +608,45 @@ Win* Win::f2c(int id){ } int Win::finish_comms(){ - mut_->lock(); + // This (simulated) mutex ensures that no process pushes to the vector of requests during the waitall. + // Without this, the vector could get redimensioned when another process pushes. + // This would result in the array used by Request::waitall() to be invalidated. + // Another solution would be to copy the data and cleanup the vector *before* Request::waitall + const std::scoped_lock lock(*mut_); //Finish own requests - std::vector *reqqs = requests_; - int size = static_cast(reqqs->size()); + int size = static_cast(requests_.size()); if (size > 0) { - MPI_Request* treqs = &(*reqqs)[0]; + MPI_Request* treqs = requests_.data(); Request::waitall(size, treqs, MPI_STATUSES_IGNORE); - reqqs->clear(); + requests_.clear(); } - mut_->unlock(); return size; } int Win::finish_comms(int rank){ - mut_->lock(); - //Finish own requests - std::vector *reqqs = requests_; - int size = static_cast(reqqs->size()); + // See comment about the mutex in finish_comms() above + const std::scoped_lock lock(*mut_); + // Finish own requests + // Let's see if we're either the destination or the sender of this request + // because we only wait for requests that we are responsible for. + // Also use the process id here since the request itself returns from src() + // and dst() the process id, NOT the rank (which only exists in the context of a communicator). + aid_t proc_id = comm_->group()->actor(rank); + auto it = std::stable_partition(begin(requests_), end(requests_), [proc_id](const MPI_Request& req) { + return (req == MPI_REQUEST_NULL || (req->src() != proc_id && req->dst() != proc_id)); + }); + std::vector myreqqs(it, end(requests_)); + requests_.erase(it, end(requests_)); + int size = static_cast(myreqqs.size()); if (size > 0) { - size = 0; - std::vector myreqqs; - std::vector::iterator iter = reqqs->begin(); - int proc_id = comm_->group()->actor(rank)->get_pid(); - while (iter != reqqs->end()){ - // Let's see if we're either the destination or the sender of this request - // because we only wait for requests that we are responsible for. - // Also use the process id here since the request itself returns from src() - // and dst() the process id, NOT the rank (which only exists in the context of a communicator). - if (((*iter) != MPI_REQUEST_NULL) && (((*iter)->src() == proc_id) || ((*iter)->dst() == proc_id))) { - myreqqs.push_back(*iter); - iter = reqqs->erase(iter); - size++; - } else { - ++iter; - } - } - if(size >0){ - MPI_Request* treqs = &myreqqs[0]; - Request::waitall(size, treqs, MPI_STATUSES_IGNORE); - myreqqs.clear(); - } + MPI_Request* treqs = myreqqs.data(); + Request::waitall(size, treqs, MPI_STATUSES_IGNORE); + myreqqs.clear(); } - mut_->unlock(); return size; } -int Win::shared_query(int rank, MPI_Aint* size, int* disp_unit, void* baseptr) +int Win::shared_query(int rank, MPI_Aint* size, int* disp_unit, void* baseptr) const { const Win* target_win = rank != MPI_PROC_NULL ? connected_wins_[rank] : nullptr; for (int i = 0; not target_win && i < comm_->size(); i++) { @@ -763,5 +679,4 @@ void Win::set_errhandler(MPI_Errhandler errhandler) if (errhandler_ != MPI_ERRHANDLER_NULL) errhandler_->ref(); } -} // namespace smpi -} // namespace simgrid +} // namespace simgrid::smpi