sstsimulator / sst-macro Goto Github PK

SST Macro Element Library

License: Other

Python 2.35% Shell 0.79% C++ 80.60% C 10.28% Objective-C 0.74% Makefile 1.05% M4 1.49% Cuda 0.01% Assembly 2.60% Jupyter Notebook 0.07% CMake 0.01%

simulation simulator snl-other snl-performance-workflow

sst-macro's People

Contributors

Stargazers

Watchers

Forkers

nlslatt calccrypto schnorr sknigh summonersrift amollah adarshyoga edwinaye deanchester beuth-diglab jjwilke shayan-taheri rhenelis pfaizian vadimkutovoi grantmackey-wdc mperrinel nikela allevin calewis enshao jpkenny sunsirui pastens anet18 jmlapre stanso nab880 babyworm berquist lchj yzhf1995 lpsmodsimteam harvv abubekermohammed burlen ko9ma7 venkata-hlrs gvoskuilen cappadokes

sst-macro's Issues

Document clang build process

Cover the building of clang libtooling and sst/macro

Compiling a project with auto global variable removal

node_id indexing doesn't initialize properly

also need to re-add node_id indexing test since that seems to have disappeared

nic traffic matrix does not match byte counts on the switches

Document otf2 build of sst-macro

This should follow the model in the manual for building with sst-core. It should show the basic installation steps for otf2 and then an example configure script for otf2 in the build section.

Running otf2 example input files should also be added similar to the dumpi section.

Weird behavior with topology unit test functions disabled

When all topology unit tests are disabled except topology/test_topology in tests/unit/tests/Makefile.am and all tests in tests//unit_tests/topology/test_topology.h and tests//unit_tests/topology/test_topology.c are commented out so that the first line that runs is unit.validate();, make check prints
FAILED: unit_test_topology.chk-out: Line SUCCESS: fat tree coord check top_fattree.cc:58 missing from CHK file, which makes no sense.

some skeletons in sst-macro/skeletons fail to build

harden OTF2

collect more traces, run them

network_switch::queue_length() returns 0

packet_flow_switch::queue_length

int
packet_flow_switch::queue_length(int port) const
{
  packet_flow_buffer* buf = static_cast<packet_flow_buffer*>(out_buffers_[port]);
  return buf->queue_length();
}

calls packet_flow_buffer::queue_length, which returns 0.

  virtual int
  queue_length() const {
    return 0;
  }

This invalidates all routers that use queue_lengths

-Wno-pragmas needs to be auto-added for gcc

Way too many warnings printed for gcc. It really shouldn't be warning about unknown pragmas.

need to clean up hindexed/indexed for MPI types

confusion between byte displacements and type count displacements

Memory leak running mpi stats

Memory leak results in std::bad_alloc at the following backtrace when running large jobs on linux.

#0 new_error_handler () at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/main/main.cc:70
#1 0x00007fab8eef90eb in operator new(unsigned long) () from /lib64/libstdc++.so.6
#2 0x00007fab8fd9b2de in __gnu_cxx::new_allocator<std::_Rb_tree_nodesstmac::event_queue_entry* >::allocate (
this=0x138d770, __n=1) at /usr/include/c++/4.8.2/ext/new_allocator.h:104
#3 0x00007fab8fd9b25a in std::_Rb_tree<sstmac::event_queue_entry*, sstmac::event_queue_entry*, std::_Identitysstmac::event_queue_entry*, sstmac::native::event_map::event_compare, std::allocatorsstmac::event_queue_entry* >::_M_get_node (this=0x138d770) at /usr/include/c++/4.8.2/bits/stl_tree.h:370
#4 0x00007fab8fd9b177 in std::_Rb_tree<sstmac::event_queue_entry*, sstmac::event_queue_entry*, std::_Identitysstmac::event_queue_entry*, sstmac::native::event_map::event_compare, std::allocatorsstmac::event_queue_entry* >::_M_create_node<sstmac::event_queue_entry* const&> (this=0x138d770) at /usr/include/c++/4.8.2/bits/stl_tree.h:403
#5 0x00007fab8fd9af29 in std::_Rb_tree<sstmac::event_queue_entry*, sstmac::event_queue_entry*, std::_Identitysstmac::event_queue_entry*, sstmac::native::event_map::event_compare, std::allocatorsstmac::event_queue_entry* >::M_insert<sstmac::event_queue_entry* const&> (this=0x138d770, __x=0x0, __p=0x643713090, __v=@0x7fab393ba710: 0x7faaf9bb3010)
at /usr/include/c++/4.8.2/bits/stl_tree.h:1023
#6 0x00007fab8fd9a9b6 in std::_Rb_tree<sstmac::event_queue_entry*, sstmac::event_queue_entry*, std::_Identitysstmac::event_queue_entry*, sstmac::native::event_map::event_compare, std::allocatorsstmac::event_queue_entry* >::_M_insert_unique<sstmac::event_queue_entry* const&> (this=0x138d770, __v=@0x7fab393ba710: 0x7faaf9bb3010)
at /usr/include/c++/4.8.2/bits/stl_tree.h:1382
#7 0x00007fab8fd9a665 in std::set<sstmac::event_queue_entry*, sstmac::native::event_map::event_compare, std::allocatorsstmac::event_queue_entry* >::insert (this=0x138d770, __x=@0x7fab393ba710: 0x7faaf9bb3010)
at /usr/include/c++/4.8.2/bits/stl_set.h:463
#8 0x00007fab8fd99ea3 in sstmac::native::event_map::add_event (this=0x138d700, ev=0x7faaf9bb3010)
at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/backends/native/event_map.cc:54
#9 0x00007fab8fd9baa5 in sstmac::native::event_container::schedule (this=0x138d700, start_time=..., seqnum=1679363,
ev=0x7faaf9bb3010) at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/backends/native/event_container.cc:118
#10 0x00007fab8fda803f in sstmac::event_scheduler::schedule (this=0x15a5ec0, t=..., ev=0x7faaf9bb3010)
at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/common/event_scheduler.cc:315
#11 0x00007fab8fda7ec3 in sstmac::event_scheduler::send_self_event_queue (this=0x15a5ec0, arrival=...,
ev=0x7faaf9bb3010) at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/common/event_scheduler.cc:269
#12 0x00007fab8fe2aa74 in sstmac::hw::pisces_memory_packetizer::handle_payload (this=0x15a5ec0, vn=77, pkt=
0x7faaf9bb4010) at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/hardware/pisces/pisces_memory_model.cc:187
#13 0x00007fab8fe2a7fa in sstmac::hw::pisces_memory_packetizer::inject (this=0x15a5ec0, vn=77, bytes=4096,
byte_offset=757760, msg=0x3d76d7af0)
at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/hardware/pisces/pisces_memory_model.cc:169
#14 0x00007fab8fde73ee in sstmac::hw::packetizer::sendWhatYouCan (this=0x15a5ec0, vn=77)
at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/hardware/common/packetizer.cc:72
#15 0x00007fab8fde708e in sstmac::hw::packetizer::start (this=0x15a5ec0, vn=77, msg=0x3d76d7af0)
at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/hardware/common/packetizer.cc:45
#16 0x00007fab8fe2a333 in sstmac::hw::pisces_memory_model::access (this=0x15a5d60, bytes=1619712, max_bw=500000000,
cb=0x709b5ac00) at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/hardware/pisces/pisces_memory_model.cc:112
#17 0x00007fab8fded698 in sstmac::hw::nic::intranode_send (this=0x15a5b80, payload=0x2b29f4200)
at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/hardware/nic/nic.cc:200
#18 0x00007fab8fdf219c in sstmac::hw::node::send_to_nic (this=0x15a5ae0, netmsg=0x2b29f4200)
at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/hardware/node/node.cc:285
#19 0x00007fab8fead282 in sstmac::sw::operating_system::execute_kernel (this=0x15a61e0, func=sstmac::ami::COMM_SEND,
data=0x2b29f4200) at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/software/process/operating_system.cc:389
#20 0x00007fab8fe6003c in sstmac::sw::operating_system::execute (this=0x15a61e0, func=sstmac::ami::COMM_SEND,
data=0x2b29f4200) at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/software/process/operating_system.h:136
#21 0x00007fab8fe5e3c5 in sstmac::sumi_transport::send (this=0xfa01f60, byte_length=1619712, dst_task=9388,
dst_node=391, dst_app=1, msg=..., needs_ack=false, ty=1)
---Type to continue, or q to quit---
at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/libraries/sumi/sumi_transport.cc:308
#22 0x00007fab8fe5e2a7 in sstmac::sumi_transport::send (this=0xfa01f60, byte_length=1619712, msg=..., sendType=1,
dst_rank=9388, needs_ack=false)
at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/libraries/sumi/sumi_transport.cc:286
#23 0x00007fab8fe5e795 in sstmac::sumi_transport::do_rdma_get (this=0xfa01f60, dst=9388, msg=...)
at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/libraries/sumi/sumi_transport.cc:364
#24 0x00007fab8fd47725 in sumi::transport::rdma_get (this=0xfa01fe0, src=9388, msg=..., needs_send_ack=false,
needs_recv_ack=true) at /home/jpkenny/src/sst-macro-mpi_stats/sumi/sumi/transport.cc:1154
#25 0x00007fab8ff07138 in sumi::mpi_queue::post_rdma (this=0xfa02940, msg=..., needs_send_ack=false,
needs_recv_ack=true) at /home/jpkenny/src/sst-macro-mpi_stats/sumi-mpi/mpi_queue/mpi_queue.cc:742
#26 0x00007fab8ff104c7 in sumi::rendezvous_get::incoming_header (this=0x2cbfcb0, queue=0xfa02940, msg=...,
req=0x48ececfe0) at /home/jpkenny/src/sst-macro-mpi_stats/sumi-mpi/mpi_protocol/rendezvous_rdma.cc:71
#27 0x00007fab8ff035df in sumi::mpi_queue::recv (this=0xfa02940, key=0x34f7857e0, count=202464, type=7, source=172,
tag=172, comm=0x3387000, buffer=0x0) at /home/jpkenny/src/sst-macro-mpi_stats/sumi-mpi/mpi_queue/mpi_queue.cc:289
#28 0x00007fab8ff2e748 in sumi::mpi_api::do_irecv (this=0xfa01f60, buf=0x0, count=202464, datatype=7, source=172,
tag=172, comm=5, request=0x7fab393bb368)
at /home/jpkenny/src/sst-macro-mpi_stats/sumi-mpi/mpi_api_send_recv.cc:224
#29 0x00007fab8ff2e847 in sumi::mpi_api::irecv (this=0xfa01f60, buf=0x0, count=202464, datatype=7, source=172,
tag=172, comm=5, request=0x7fab393bb368)
at /home/jpkenny/src/sst-macro-mpi_stats/sumi-mpi/mpi_api_send_recv.cc:233
#30 0x00007fab8fe8212f in sumi::parsedumpi_callbacks::on_MPI_Irecv (prm=0x7fab393be420, thread=0,
cpu=0x7fab393be400, wall=0x7fab393be3f0, perf=0x7fab393bb3e0, uarg=0x7fab393bf9d0)
at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/skeletons/sumi_undumpi/parsedumpi_callbacks.cc:987
#31 0x00007fab8f5dac06 in libundumpi_grab_irecv (profile=0xfa02fb0,
callout=0x7fab8fe81f4e <sumi::parsedumpi_callbacks::on_MPI_Irecv(dumpi_irecv const*, unsigned short, dumpi_time const*, dumpi_time const*, dumpi_perfinfo const*, void*)>, uarg=0x7fab393bf9d0)
at /home/jpkenny/src/sst-macro-mpi_stats/sst-dumpi/dumpi/libundumpi/bindings.c:239
#32 0x00007fab8f5d8a8a in undumpi_read_single_call (profile=0xfa02fb0, callarr=0x7fab393be4e0, uarg=0x7fab393bf9d0,
mpi_finalized=0x7fab393bf734) at /home/jpkenny/src/sst-macro-mpi_stats/sst-dumpi/dumpi/libundumpi/libundumpi.c:83
#33 0x00007fab8f5d8dfb in undumpi_read_stream_full (
metaname=0xfa02d28 "/srv/tmp/jk/gtc_16384/dumpi-2015.08.19.08.30.41-9389.bin", profile=0xfa02fb0,
callback=0x2d7ee10, uarg=0x7fab393bf9d0, print_progress=false, percent_terminate=-1)
at /home/jpkenny/src/sst-macro-mpi_stats/sst-dumpi/dumpi/libundumpi/libundumpi.c:150
#34 0x00007fab8fe7de35 in sumi::parsedumpi_callbacks::parse_stream (this=0x7fab393bf9d0,
fname="/srv/tmp/jk/gtc_16384/dumpi-2015.08.19.08.30.41-9389.bin", print_progress=false, percent_terminate=-1)
at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/skeletons/sumi_undumpi/parsedumpi_callbacks.cc:126
#35 0x00007fab8fe7cb5d in sumi::parsedumpi::skeleton_main (this=0xfa009f0)
at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/skeletons/sumi_undumpi/parsedumpi.cc:99
#36 0x00007fab8fec0a18 in sstmac::sw::app::run (this=0xfa009f0)
at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/software/process/app.cc:216
#37 0x00007fab8feb9831 in sstmac::sw::thread::run_routine (threadptr=0x2fcc0a0)
at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/software/process/thread.cc:205
#38 0x00007fab8fed024d in sstmac::sw::funcptr::call (this=0x7fab393bffd0, arg=...)
at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/software/threading/context_util.h:80
#39 0x00007fab8fed0090 in sstmac::sw::context_springboard (func_ptr_a=-1880385598, func_ptr_b=32683,
arg_ptr_a=50118816, arg_ptr_b=0)
at /home/jpkenny/src/sst-macro-mpi_stats/sstmac/software/threading/context_util.cc:23
#40 0x00007fab8e607cf0 in ?? () from /lib64/libc.so.6
#41 0x0000000000000000 in ?? ()New Issue for sst-macro

parameter checking (regex enabled) is broken

parameter checking (regex enabled) is broken on master branch (possible others)

verify make dist

Compile from tarballs

Pulling data from .git/refs/heads broken

Line 2171 of https://travis-ci.org/calccrypto/sst-macro/builds/144228551

make[2]: *** No rule to make target /home/travis/build/calccrypto/sst-macro/.git/refs/heads/HEAD', needed by../../sstmac_repo.h'. Stop.

git/refs/heads/HEAD should be git/refs/heads/<branch name>. HEAD is in .git/.

Is it necessary to print the git commit for versioning?

Performance optimization of multithreading

Limit number of thread barriers
Template barrier functions?
General profiling to identify hotspots

regex off by default

regex proofreading should be off by default
very slow implementations are causing timeout tests, performance debugging needed - particularly on Ubuntu

MPI parallel optimization: get async progress threads to play nice with multithreading

Openmpi has progress threads. We pin threads and spinlock. Get these two to cooperate.

another test

New Issue for sst-macro

Skeletons provided do not compile (none of them)

Skeletons provided under skeletons/ directory do not compile, I've tried all of them and I cannot get them compiled. These are my system properties:

Ubuntu 16.04 LTS
g++ (Ubuntu 5.4.0-6ubuntu1~16.04.4) 5.4.0 20160609
Zoltan 1.83
Boost 1.58
Package openmpi-bin 1.10.2 from Ubuntu repositories
sst-macro downloaded from this repository (as I'm writting this)

Also tried with a different system:

g++ (GCC) 4.8.5 20150623 (Red Hat 4.8.5-4)
Zoltan 1.83
Boost 1.63.0
OpenMPI 2.0.2
sst-macro downloaded from this repository (as I'm writting this)

This is the error I'm getting:

sst-macro/skeletons/sendrecv$ make
sst++ -I.  -c main.cc -o main.o
<command-line>:0:10: error: ‘dontallow’ does not name a type
<command-line>:0:10: error: ‘dontallow’ does not name a type
In file included from /home/juanjosegcc/local/sstmacro/6.1.0/include/sstmac/replacements/mutex:12:0,
                 from /usr/include/c++/5/condition_variable:39,
                 from /home/juanjosegcc/local/sstmacro/6.1.0/include/sstmac/replacements/condition_variable:6,
                 from /home/juanjosegcc/local/sstcore/6.1.0/include/sst/core/threadsafe.h:19,
                 from /home/juanjosegcc/local/sstcore/6.1.0/include/sst/core/mempool.h:25,
                 from /home/juanjosegcc/local/sstcore/6.1.0/include/sst/core/activity.h:21,
                 from /home/juanjosegcc/local/sstcore/6.1.0/include/sst/core/event.h:21,
                 from /home/juanjosegcc/local/sstcore/6.1.0/include/sst/core/link.h:17,
                 from /home/juanjosegcc/local/sstmacro/6.1.0/include/sstmac/common/event_handler.h:22,
                 from /home/juanjosegcc/local/sstmacro/6.1.0/include/sstmac/common/sst_event.h:16,
                 from /home/juanjosegcc/local/sstmacro/6.1.0/include/sstmac/common/messages/sst_message.h:19,
                 from /home/juanjosegcc/local/sstmacro/6.1.0/include/sstmac/hardware/network/network_message.h:4,
                 from /home/juanjosegcc/local/sstmacro/6.1.0/include/sumi-mpi/mpi_message.h:16,
                 from /home/juanjosegcc/local/sstmacro/6.1.0/include/sumi-mpi/mpi_request.h:18,
                 from /home/juanjosegcc/local/sstmacro/6.1.0/include/sumi-mpi/mpi_api.h:22,
                 from /home/juanjosegcc/local/sstmacro/6.1.0/include/sumi-mpi/mpi_wrapper.h:8,
                 from /home/juanjosegcc/local/sstmacro/6.1.0/include/sstmac/replacements/mpi.h:2,
                 from main.cc:1:
/usr/include/c++/5/mutex: In function ‘void std::__once_call_impl()’:
/usr/include/c++/5/mutex:706:21: error: ‘__once_callable’ was not declared in this scope
       (*(_Callable*)__once_callable)();
                     ^
/usr/include/c++/5/mutex: In function ‘void std::call_once(std::once_flag&, _Callable&&, _Args&& ...)’:
/usr/include/c++/5/mutex:728:7: error: ‘__once_callable’ was not declared in this scope
       __once_callable = std::__addressof(__bound_functor);
       ^
/usr/include/c++/5/mutex:729:7: error: ‘__once_call’ was not declared in this scope
       __once_call = &__once_call_impl<decltype(__bound_functor)>;
       ^
Makefile:24: recipe for target 'main.o' failed
make: *** [main.o] Error 1

MPI parallel optimization: message aggregation

Do not send single events. Aggregate messages to a given size

Using router parameter breaks with topology=simple_fattree

Should ft be cast to a abstract_fat_tree instead of fat_tree? Also, shouldn't the check be done first, and then the variable be set?

sstmac/hardware/router/fat_tree.cc:54

void
fat_tree_router::set_topology(topology *top)
{
  structured_router::set_topology(top);

  fat_tree* ft = safe_cast(fat_tree, top);
  if (ft->k() != k_ || ft->l() != l_){
    spkt_throw_printf(sprockit::value_error,
                      "fat tree router configuration (k=%d,l=%d) does not match"
                      " topology configuration (k=%d,l=%d)",
                      k_, l_, ft->k(), ft->l());
  }
}

fat tree initialization

fat tree should probably be initialized with k=N, l=M instead of confusing topology geometry input

no warning about inconsistent override for clang

Too many errors coming in from SST core. Just stop the warnings.

all topology unit tests call abort()

this should not happen

round-off errors

epoch->truncate_after can be more robust with respect to round-off - causes issues in certain use cases

mpirun -n 2 skeletons/sendrecv/runsendrecv

Skeletons simulations cannot be executed with MPI

I've got sst-macro built in standalone mode (which allows me to build the skeletons), otherwise they will not build.

When executing the simulations as is, skeletons/sendrecv/runsendrecv, it works.

$ ./runsendrecv
./runsendrecv
  --debug="" \
  --configfile="parameters.ini" \

Rank 0 finished at t=  0.0020 ms
Rank 0 finished at t=  0.0080 ms
Estimated total runtime of           0.00001206 seconds
SSTMAC   repo:   1307bb4f31e033dbbd2d9b82c54d3753446d31a4
SST/macro ran for       0.0009 seconds

But when trying to execute using MPI to execute the simulation (title command), the dynamic linker tells me this:

$ mpirun -n 2 ./runsendrecv
./runsendrecv
  --debug="" \
  --configfile="parameters.ini" \

[qarthos:05771] *** Process received signal ***
[qarthos:05771] Signal: Segmentation fault (11)
[qarthos:05771] Signal code: Address not mapped (1)
[qarthos:05771] Failing at address: 0x60
[qarthos:05771] [ 0] /lib/x86_64-linux-gnu/libpthread.so.0(+0x11390)[0x7f8b9cc35390]
[qarthos:05771] [ 1] /home/juanjosegcc/local/sstmacro/6.1.0/lib/libsstmac.so.6(_ZN6sstmac2hw12interconnect17connect_endpointsEPN8sprockit14sim_parametersES4_+0x128)[0x7f8b9dd9f4d8]
[qarthos:05771] [ 2] /home/juanjosegcc/local/sstmacro/6.1.0/lib/libsstmac.so.6(_ZN6sstmac2hw12interconnectC1EPN8sprockit14sim_parametersEPNS_13event_managerEPNS_9partitionEPNS_16parallel_runtimeE+0xa10)[0x7f8b9dda2550]
[qarthos:05771] [ 3] /home/juanjosegcc/local/sstmacro/6.1.0/lib/libsstmac.so.6(_ZN6sstmac2hw12interconnect19static_interconnectEPN8sprockit14sim_parametersEPNS_13event_managerE+0x43b)[0x7f8b9dda2c0b]
[qarthos:05771] [ 4] /home/juanjosegcc/local/sstmacro/6.1.0/lib/libsstmac.so.6(_ZN6sstmac6native7managerC1EPN8sprockit14sim_parametersEPNS_16parallel_runtimeE+0x29a)[0x7f8b9dd3c91a]
[qarthos:05771] [ 5] /home/juanjosegcc/local/sstmacro/6.1.0/lib/libsstmac.so.6(_ZN6sstmac10run_paramsEPNS_16parallel_runtimeEPN8sprockit14sim_parametersER9sim_stats+0xcb)[0x7f8b9ddc9aab]
[qarthos:05771] [ 6] /home/juanjosegcc/local/sstmacro/6.1.0/lib/libsstmac.so.6(_ZN6sstmac8try_mainEPN8sprockit14sim_parametersEiPPcb+0x291)[0x7f8b9ddca341]
[qarthos:05771] [ 7] /home/juanjosegcc/local/sstmacro/6.1.0/lib/libsstmac_main.so.0(main+0x41)[0x7f8b9e5d6cd1]
[qarthos:05771] [ 8] /lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xf0)[0x7f8b9c87b830]
[qarthos:05771] [ 9] ./runsendrecv[0x403e29]
[qarthos:05771] *** End of error message ***
[qarthos:05770] *** Process received signal ***
[qarthos:05770] Signal: Segmentation fault (11)
[qarthos:05770] Signal code: Address not mapped (1)
[qarthos:05770] Failing at address: 0x60
[qarthos:05770] [ 0] /lib/x86_64-linux-gnu/libpthread.so.0(+0x11390)[0x7fecc79ac390]
[qarthos:05770] [ 1] /home/juanjosegcc/local/sstmacro/6.1.0/lib/libsstmac.so.6(_ZN6sstmac2hw12interconnect17connect_endpointsEPN8sprockit14sim_parametersES4_+0x128)[0x7fecc8b164d8]
[qarthos:05770] [ 2] /home/juanjosegcc/local/sstmacro/6.1.0/lib/libsstmac.so.6(_ZN6sstmac2hw12interconnectC1EPN8sprockit14sim_parametersEPNS_13event_managerEPNS_9partitionEPNS_16parallel_runtimeE+0xa10)[0x7fecc8b19550]
[qarthos:05770] [ 3] /home/juanjosegcc/local/sstmacro/6.1.0/lib/libsstmac.so.6(_ZN6sstmac2hw12interconnect19static_interconnectEPN8sprockit14sim_parametersEPNS_13event_managerE+0x43b)[0x7fecc8b19c0b]
[qarthos:05770] [ 4] /home/juanjosegcc/local/sstmacro/6.1.0/lib/libsstmac.so.6(_ZN6sstmac6native7managerC1EPN8sprockit14sim_parametersEPNS_16parallel_runtimeE+0x29a)[0x7fecc8ab391a]
[qarthos:05770] [ 5] /home/juanjosegcc/local/sstmacro/6.1.0/lib/libsstmac.so.6(_ZN6sstmac10run_paramsEPNS_16parallel_runtimeEPN8sprockit14sim_parametersER9sim_stats+0xcb)[0x7fecc8b40aab]
[qarthos:05770] [ 6] /home/juanjosegcc/local/sstmacro/6.1.0/lib/libsstmac.so.6(_ZN6sstmac8try_mainEPN8sprockit14sim_parametersEiPPcb+0x291)[0x7fecc8b41341]
[qarthos:05770] [ 7] /home/juanjosegcc/local/sstmacro/6.1.0/lib/libsstmac_main.so.0(main+0x41)[0x7fecc934dcd1]
[qarthos:05770] [ 8] /lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xf0)[0x7fecc75f2830]
[qarthos:05770] [ 9] ./runsendrecv[0x403e29]
[qarthos:05770] *** End of error message ***
--------------------------------------------------------------------------
mpirun noticed that process rank 0 with PID 5770 on node qarthos exited on signal 11 (Segmentation fault).
--------------------------------------------------------------------------

I've got sst-macro compiled in standalone mode with CC=mpicc and CXX=mpic++, using gcc 5.4 in Ubuntu 16.04LTS 64bits.

For more information, this is the output for the command ldd:

$ ldd runsendrecv
	linux-vdso.so.1 =>  (0x00007fffa030f000)
	libsstmac_main.so.0 => /home/juanjosegcc/local/sstmacro/6.1.0/lib/libsstmac_main.so.0 (0x00007fb03fe17000)
	libsprockit.so.0 => /home/juanjosegcc/local/sstmacro/6.1.0/lib/libsprockit.so.0 (0x00007fb03fbbc000)
	libundumpi.so.3 => /home/juanjosegcc/local/sstmacro/6.1.0/lib/libundumpi.so.3 (0x00007fb03f91a000)
	libsstmac.so.6 => /home/juanjosegcc/local/sstmacro/6.1.0/lib/libsstmac.so.6 (0x00007fb03f413000)
	libmpi_cxx.so.1 => /usr/lib/libmpi_cxx.so.1 (0x00007fb03f1dc000)
	libmpi.so.12 => /usr/lib/libmpi.so.12 (0x00007fb03ef06000)
	libstdc++.so.6 => /usr/lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007fb03eb84000)
	libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007fb03e87a000)
	libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007fb03e664000)
	libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007fb03e447000)
	libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007fb03e07d000)
	libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007fb03de79000)
	libibverbs.so.1 => /usr/lib/libibverbs.so.1 (0x00007fb03dc69000)
	libopen-rte.so.12 => /usr/lib/libopen-rte.so.12 (0x00007fb03d9ef000)
	libopen-pal.so.13 => /usr/lib/libopen-pal.so.13 (0x00007fb03d752000)
	/lib64/ld-linux-x86-64.so.2 (0x000055ea9f00b000)
	libhwloc.so.5 => /usr/lib/x86_64-linux-gnu/libhwloc.so.5 (0x00007fb03d517000)
	librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1 (0x00007fb03d30f000)
	libutil.so.1 => /lib/x86_64-linux-gnu/libutil.so.1 (0x00007fb03d10c000)
	libnuma.so.1 => /usr/lib/x86_64-linux-gnu/libnuma.so.1 (0x00007fb03cf00000)
	libltdl.so.7 => /usr/lib/x86_64-linux-gnu/libltdl.so.7 (0x00007fb03ccf6000)

topology and routing unit tests are disabled

they should be enabled

python 3 compatibility

Convert all print statements to sys.stdout.write for both 2 and 3 compat

crossbar arbitration policy

crossbar pulls off a single queue - could probably improve arbitration policy to not bias so much to single source

deprecated parameters do not really play nice with param expanders

performance debugging of regex

Regex runs much too slowly for large runs

OTF2 added to test suite

Small OTF2 trace should be conditionally added to test suite when otf is activated

otf2 trace replay with openmp events

collect OTF2 traces for debugging

Lulesh
AMG
HPCG
MiniDFT

~50-100 MPI ranks

This should be plenty

Is minimal_adaptive intentionally left out of routing::to_str?

https://github.com/sstsimulator/sst-macro/blob/master/sstmac/hardware/router/routable.cc#L26

const char*
tostr(routing::algorithm_t algo)
{
  switch(algo) {
      enumcase(minimal);
      enumcase(valiant);
      enumcase(deflt);
      enumcase(ugal);
    default:
      spkt_throw_printf(sprockit::value_error,
                       "invalud routing algorithm enum %d",
                       algo);
  }
  return 0;
}

clang libTooling better configure check

The libtooling check needs to be stricter. Right now, it just checks for header files. There should be a way to ensure abi compatibility between compiler used for SST and compiler used for libtooling so entire configure/compile process doesn't complete before linking error occurs.

parameter namespace issues

has_param() function does not properly check parent + global namespaces for parameter name. This causes get_int_param() and get_optional_int_param() to have inconsistent behaviors.

cpu-affinity should be disabled on OSX

cpu-affinity doesn't work on OSX, configure should ensure that user doesn't enable it there

MPI parallel optimization: speculative Irecvs

Have Irecvs posted speculatively (for a particular size - related to message aggregation)

Allow async progress thread to make progress before reaching sync points

man variable not defined

in devel branch sstmac/software/process/operating_system.cc

#if SSTMAC_USE_MULTITHREAD
  if (os_thread_contexts_.size() == 1){
    os_thread_contexts_.resize(man->nthread());
    os_thread_context& main_ctxt = os_thread_contexts_[0];
    for (int i=1; i < man->nthread(); ++i){
      os_thread_context& ctxt = os_thread_contexts_[i];
      ctxt.stackalloc.init(
        main_ctxt.stackalloc.stacksize(),
        main_ctxt.stackalloc.chunksize(),
        main_ctxt.stackalloc.use_mprot());
    }
  }
#endif

~/install/sstmacro-debug/bin/sst++ -I. -DSSTMAC_HAVE_BOOST=1 -I~/src/boost_1_59_0 -g -c boxml.cc -o boxml.o
make[3]: ~/install/sstmacro-debug/bin/sst++: Command not found
~/install/sstmacro-debug/bin/sst++ -I. -DSSTMAC_HAVE_BOOST=1 -I~/src/boost_1_59_0 -g -c boxml-data.cc -o boxml-data.o
Makefile:27: recipe for target 'boxml.o' failed
make[3]: ~/install/sstmacro-debug/bin/sst++: Command not found
make[3]: *** [boxml.o] Error 127
make[3]: *** Waiting for unfinished jobs....
Makefile:27: recipe for target 'boxml-data.o' failed
make[3]: *** [boxml-data.o] Error 127
make[3]: Leaving directory '~/sst-macro/tests/skeletons/boxml'
Makefile:1534: recipe for target 'boxml' failed
make[2]: *** [boxml] Error 2
make[2]: *** Waiting for unfinished jobs....
CXX: ~/.local/sstmacro/bin/sst++

if sst core was activated
if threading was activate and whether spinlocks/affinity
missing stuff on MPI pdes - should say yes or no
whether mpi sync stats are enable
whether sanity checks are enabled
whether keyword proofreading (regex) is one
whether clang source-to-source is enable

Request: make fat_tree::nearest_common_ancestor_level work for internal nodes

Right now, it seems that fat_tree::nearest_common_ancestor_level assumes that the input coordinates are leaf switches. What about when they are not?