diff --git a/.ycm_extra_conf.py b/.ycm_extra_conf.py
index a48a781b9af0e36d56e9b5330aaa93518696fec6..f79487e8578b15e5fc71177dceec4f4a36fb419d 100644
--- a/.ycm_extra_conf.py
+++ b/.ycm_extra_conf.py
@@ -36,7 +36,7 @@ import ycm_core
# CHANGE THIS LIST OF FLAGS. YES, THIS IS THE DROID YOU HAVE BEEN LOOKING FOR.
flags = [
'-DNDEBUG',
- '-DNMC_HAVE_CTHREAD',
+ '-DARB_HAVE_CTHREAD',
'-std=c++11',
'-x',
'c++',
@@ -54,7 +54,7 @@ flags = [
'modcc',
'-I',
'/cm/shared/apps/cuda/8.0.44/include',
- '-DNMC_HAVE_GPU'
+ '-DARB_HAVE_GPU'
]
# Set this to the absolute path to the folder (NOT the file!) containing the
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 58d126e6b6624c0b301fb6ecaf482664744efad6..11efe3df7d5417689a6d7e84ca52d6a466a9ccbc 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -29,23 +29,23 @@ set(CMAKE_ARCHIVE_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/lib)
#----------------------------------------------------------
# Option to enable assertions
#----------------------------------------------------------
-option(NMC_WITH_ASSERTIONS "enable EXPECTS() assertions in code" OFF)
-if(NMC_WITH_ASSERTIONS)
- add_definitions("-DNMC_HAVE_ASSERTIONS")
+option(ARB_WITH_ASSERTIONS "enable EXPECTS() assertions in code" OFF)
+if(ARB_WITH_ASSERTIONS)
+ add_definitions("-DARB_HAVE_ASSERTIONS")
endif()
#----------------------------------------------------------
# Option to enable traces
#----------------------------------------------------------
-option(NMC_WITH_TRACE "enable TRACE() macros in code" OFF)
-if(NMC_WITH_TRACE)
- add_definitions("-DNMC_HAVE_TRACE")
+option(ARB_WITH_TRACE "enable TRACE() macros in code" OFF)
+if(ARB_WITH_TRACE)
+ add_definitions("-DARB_HAVE_TRACE")
endif()
#----------------------------------------------------------
# Option to disable auto running of modcc compiler
#----------------------------------------------------------
-option(NMC_AUTO_RUN_MODCC_ON_CHANGES
+option(ARB_AUTO_RUN_MODCC_ON_CHANGES
"Rerun modcc compiler whenever *.mod file or modcc compiler change" ON)
#----------------------------------------------------------
@@ -55,17 +55,17 @@ set(EXTERNAL_LIBRARIES "")
set(EXTERNAL_INCLUDES "")
#----------------------------------------------------------
-# list of libraries built inside NestMC
+# list of libraries built inside Arbor
#----------------------------------------------------------
-set(NESTMC_LIBRARIES "")
+set(ARB_LIBRARIES "")
#----------------------------------------------------------
# Threading model selection
#----------------------------------------------------------
-set(NMC_THREADING_MODEL "cthread" CACHE STRING "set the threading model, one of cthread/tbb/serial")
-set_property(CACHE NMC_THREADING_MODEL PROPERTY STRINGS cthread tbb serial )
+set(ARB_THREADING_MODEL "cthread" CACHE STRING "set the threading model, one of cthread/tbb/serial")
+set_property(CACHE ARB_THREADING_MODEL PROPERTY STRINGS cthread tbb serial )
-if(NMC_THREADING_MODEL MATCHES "tbb")
+if(ARB_THREADING_MODEL MATCHES "tbb")
# download and build TBB
include(${CMAKE_SOURCE_DIR}/cmake/tbb/TBBGet.cmake)
include(${CMAKE_SOURCE_DIR}/cmake/tbb/TBBBuild.cmake)
@@ -75,8 +75,8 @@ if(NMC_THREADING_MODEL MATCHES "tbb")
# configure TBB
find_package(TBB REQUIRED tbb tbbmalloc)
- add_definitions(-DNMC_HAVE_TBB)
- set(NMC_WITH_TBB TRUE)
+ add_definitions(-DARB_HAVE_TBB)
+ set(ARB_WITH_TBB TRUE)
# The TBB CMake package creates the TBB::tbb and TBB:tbbmalloc targets.
# The following should be sufficient, however on some systems (e.g. Cray PE),
@@ -91,11 +91,11 @@ if(NMC_THREADING_MODEL MATCHES "tbb")
# The TBB headers are in the downloaded TBB source
list(APPEND EXTERNAL_INCLUDES ${tbb_root}/include)
-elseif(NMC_THREADING_MODEL MATCHES "cthread")
+elseif(ARB_THREADING_MODEL MATCHES "cthread")
find_package(Threads REQUIRED)
- add_definitions(-DNMC_HAVE_CTHREAD)
- set(NMC_WITH_CTHREAD TRUE)
+ add_definitions(-DARB_HAVE_CTHREAD)
+ set(ARB_WITH_CTHREAD TRUE)
list(APPEND EXTERNAL_LIBRARIES ${CMAKE_THREAD_LIBS_INIT})
if(CMAKE_USE_PTHREADS_INIT)
@@ -116,8 +116,8 @@ endif()
#----------------------------------------------------------
# CUDA support
#----------------------------------------------------------
-option(NMC_WITH_CUDA "use CUDA for GPU offload" OFF)
-if(NMC_WITH_CUDA)
+option(ARB_WITH_CUDA "use CUDA for GPU offload" OFF)
+if(ARB_WITH_CUDA)
find_package(CUDA REQUIRED)
# Turn off annoying and incorrect warnings generated in the JSON file.
@@ -125,14 +125,14 @@ if(NMC_WITH_CUDA)
set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS};-Xcudafe \"--diag_suppress=not_used_in_template_function_params\";-Xcudafe \"--diag_suppress=cast_to_qualified_type\")
# set the CUDA target specfic flags
- # code regions protected by NMC_HAVE_CUDA should only be available to the CUDA
- # compiler, which regions protected by NMC_HAVE_GPU are visible to both host
+ # code regions protected by ARB_HAVE_CUDA should only be available to the CUDA
+ # compiler, which regions protected by ARB_HAVE_GPU are visible to both host
# and device compiler when targetting GPU.
- set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS};-DNMC_HAVE_CUDA)
- set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS};-DNMC_HAVE_GPU)
+ set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS};-DARB_HAVE_CUDA)
+ set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS};-DARB_HAVE_GPU)
set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS};-arch=sm_60)
- add_definitions(-DNMC_HAVE_GPU)
+ add_definitions(-DARB_HAVE_GPU)
include_directories(SYSTEM ${CUDA_INCLUDE_DIRS})
list(APPEND EXTERNAL_LIBRARIES ${CUDA_LIBRARIES})
endif()
@@ -140,28 +140,28 @@ endif()
#----------------------------------------------------------
# Cray/BGQ/Generic Linux/other flag?
#----------------------------------------------------------
-set(NMC_SYSTEM_TYPE "Generic" CACHE STRING
+set(ARB_SYSTEM_TYPE "Generic" CACHE STRING
"Choose a system type to customize flags")
-set_property(CACHE NMC_SYSTEM_TYPE PROPERTY STRINGS Generic Cray BGQ )
+set_property(CACHE ARB_SYSTEM_TYPE PROPERTY STRINGS Generic Cray BGQ )
# Cray specific flags
-if(${NMC_SYSTEM_TYPE} MATCHES "Cray")
+if(${ARB_SYSTEM_TYPE} MATCHES "Cray")
# we no longer set the -dynamic flag for the compilere here, instead dynamic
# linking should be enabled by the user by setting the environment variable:
# CRAYPE_LINK_TYPE=dynamic
#set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -dynamic")
- add_definitions(-DNMC_HAVE_CRAY)
+ add_definitions(-DARB_HAVE_CRAY)
endif()
#----------------------------------------------------------
# MPI support
#----------------------------------------------------------
-set(NMC_DISTRIBUTED_MODEL "serial" CACHE STRING "set the global communication model, one of serial/mpi/dryrun")
-set_property(CACHE NMC_DISTRIBUTED_MODEL PROPERTY STRINGS serial mpi dryrun)
+set(ARB_DISTRIBUTED_MODEL "serial" CACHE STRING "set the global communication model, one of serial/mpi/dryrun")
+set_property(CACHE ARB_DISTRIBUTED_MODEL PROPERTY STRINGS serial mpi dryrun)
-if(NMC_DISTRIBUTED_MODEL MATCHES "mpi")
+if(ARB_DISTRIBUTED_MODEL MATCHES "mpi")
# BGQ specific flags
- if(${NMC_SYSTEM_TYPE} MATCHES "BGQ" )
+ if(${ARB_SYSTEM_TYPE} MATCHES "BGQ" )
# On BGQ, set CXX to the mpi wrapper, and pass it a static
add_definitions(-DMPICH2_CONST=const)
set(MPI_FOUND TRUE)
@@ -171,46 +171,46 @@ if(NMC_DISTRIBUTED_MODEL MATCHES "mpi")
find_package(MPI REQUIRED)
endif()
include_directories(SYSTEM ${MPI_C_INCLUDE_PATH})
- add_definitions(-DNMC_HAVE_MPI)
+ add_definitions(-DARB_HAVE_MPI)
# unfortunate workaround for C++ detection in system mpi.h
add_definitions(-DMPICH_SKIP_MPICXX=1 -DOMPI_SKIP_MPICXX=1)
set_property(DIRECTORY APPEND_STRING PROPERTY COMPILE_OPTIONS "${MPI_C_COMPILE_FLAGS}")
- set(NMC_WITH_MPI TRUE)
+ set(ARB_WITH_MPI TRUE)
-elseif(NMC_DISTRIBUTED_MODEL MATCHES "dryrun")
- add_definitions(-DNMC_HAVE_DRYRUN)
- set(NMC_WITH_DRYRUN TRUE)
+elseif(ARB_DISTRIBUTED_MODEL MATCHES "dryrun")
+ add_definitions(-DARB_HAVE_DRYRUN)
+ set(ARB_WITH_DRYRUN TRUE)
-elseif(NMC_DISTRIBUTED_MODEL MATCHES "serial")
+elseif(ARB_DISTRIBUTED_MODEL MATCHES "serial")
# no additional set up needed
else()
- message( FATAL_ERROR "-- Distributed communication model '${NMC_DISTRIBUTED_MODEL}' not supported, use one of serial/mpi/dryrun")
+ message( FATAL_ERROR "-- Distributed communication model '${ARB_DISTRIBUTED_MODEL}' not supported, use one of serial/mpi/dryrun")
endif()
#----------------------------------------------------------
# Built-in profiler
#----------------------------------------------------------
-option(NMC_WITH_PROFILING "use built-in profiling of miniapp" OFF)
-if(NMC_WITH_PROFILING)
- add_definitions(-DNMC_HAVE_PROFILING)
+option(ARB_WITH_PROFILING "use built-in profiling of miniapp" OFF)
+if(ARB_WITH_PROFILING)
+ add_definitions(-DARB_HAVE_PROFILING)
endif()
#----------------------------------------------------------
# vectorization target
#----------------------------------------------------------
-set(NMC_VECTORIZE_TARGET "none" CACHE STRING "CPU target for vectorization {KNL,AVX,AVX2,AVX512}")
-set_property(CACHE NMC_VECTORIZE_TARGET PROPERTY STRINGS none KNL AVX AVX2 AVX512)
+set(ARB_VECTORIZE_TARGET "none" CACHE STRING "CPU target for vectorization {KNL,AVX,AVX2,AVX512}")
+set_property(CACHE ARB_VECTORIZE_TARGET PROPERTY STRINGS none KNL AVX AVX2 AVX512)
-if(NMC_VECTORIZE_TARGET STREQUAL "KNL")
+if(ARB_VECTORIZE_TARGET STREQUAL "KNL")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${CXXOPT_KNL}")
-elseif(NMC_VECTORIZE_TARGET STREQUAL "AVX")
+elseif(ARB_VECTORIZE_TARGET STREQUAL "AVX")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${CXXOPT_AVX}")
-elseif(NMC_VECTORIZE_TARGET STREQUAL "AVX2")
+elseif(ARB_VECTORIZE_TARGET STREQUAL "AVX2")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${CXXOPT_AVX2}")
-elseif(NMC_VECTORIZE_TARGET STREQUAL "AVX512")
+elseif(ARB_VECTORIZE_TARGET STREQUAL "AVX512")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${CXXOPT_AVX512}")
endif()
@@ -232,22 +232,22 @@ endif()
# Validation data generation
#----------------------------------------------------------
# destination directory for generated data
-set(NMC_VALIDATION_DATA_DIR "${PROJECT_SOURCE_DIR}/validation/data" CACHE PATH
+set(ARB_VALIDATION_DATA_DIR "${PROJECT_SOURCE_DIR}/validation/data" CACHE PATH
"location of generated validation data")
#----------------------------------------------------------
# Whether to build validation data
#----------------------------------------------------------
# turn off by default
-option(NMC_BUILD_VALIDATION_DATA "generate validation data" OFF)
-if (NMC_BUILD_VALIDATION_DATA)
+option(ARB_BUILD_VALIDATION_DATA "generate validation data" OFF)
+if (ARB_BUILD_VALIDATION_DATA)
# Whether to attempt to use julia to build validation data
find_program(JULIA_BIN julia)
if(JULIA_BIN STREQUAL "JULIA_BIN-NOTFOUND")
message(STATUS "julia not found; will not automatically build validation data sets from julia scripts")
- set(NMC_BUILD_JULIA_VALIDATION_DATA FALSE)
+ set(ARB_BUILD_JULIA_VALIDATION_DATA FALSE)
else()
- set(NMC_BUILD_JULIA_VALIDATION_DATA TRUE)
+ set(ARB_BUILD_JULIA_VALIDATION_DATA TRUE)
endif()
# Whether to attempt to use nrniv to build validation data
@@ -255,9 +255,9 @@ if (NMC_BUILD_VALIDATION_DATA)
find_program(NRNIV_BIN nrniv)
if(NRNIV_BIN STREQUAL "NRNIV_BIN-NOTFOUND")
message(STATUS "nrniv not found; will not automatically build NEURON validation data sets")
- set(NMC_BUILD_NRN_VALIDATION_DATA FALSE)
+ set(ARB_BUILD_NRN_VALIDATION_DATA FALSE)
else()
- set(NMC_BUILD_NRN_VALIDATION_DATA TRUE)
+ set(ARB_BUILD_NRN_VALIDATION_DATA TRUE)
endif()
endif()
@@ -279,7 +279,7 @@ endif()
# Setup subdirs
#----------------------------------------------------------
# only include validation data if flag is set
-if(NMC_BUILD_VALIDATION_DATA)
+if(ARB_BUILD_VALIDATION_DATA)
add_subdirectory(validation)
endif()
diff --git a/README.md b/README.md
index aedf199a3d8d245272a3c8127c88daae6471bd56..728641df9e46a5cf3e5c19e1f113524bc6f284ca 100644
--- a/README.md
+++ b/README.md
@@ -41,13 +41,13 @@ cd tests
## MPI
-Set the `NMC_WITH_MPI` option either via the ccmake interface, or via the command line as shown below.
+Set the `ARB_WITH_MPI` option either via the ccmake interface, or via the command line as shown below.
To ensure that CMake detects MPI correctly, you should specify the MPI wrapper for the compiler by setting the `CXX` and `CC` environment variables.
```
export CXX=mpicxx
export CC=mpicc
-cmake <path to CMakeLists.txt> -DNMC_WITH_MPI=ON
+cmake <path to CMakeLists.txt> -DARB_WITH_MPI=ON
```
## TBB
@@ -57,7 +57,7 @@ When TBB is installed, it comes with some scripts that can be run to set up the
The scripts set the `TBB_ROOT` environment variable, which is used by the CMake configuration to find TBB.
```
-cmake <path to CMakeLists.txt> -DNMC_THREADING_MODEL=tbb
+cmake <path to CMakeLists.txt> -DARB_THREADING_MODEL=tbb
```
### TBB on Cray systems
@@ -71,11 +71,11 @@ Note, the CMake package provided by TBB is very fragile, and won't work if CMake
```
export CRAYPE_LINK_TYPE=dynamic
-cmake <path-to-arbor-source> -DNMC_THREADING_MODEL=tbb
+cmake <path-to-arbor-source> -DARB_THREADING_MODEL=tbb
# NOTE: specifying CMAKE_SYSTEM_NAME won't work, instead let CMake automatically
# detect the build environment as above.
-cmake <path-to-arbor-source> -DNMC_THREADING_MODEL=tbb -DCMAKE_SYSTEM_NAME=CrayLinuxEnvironment
+cmake <path-to-arbor-source> -DARB_THREADING_MODEL=tbb -DCMAKE_SYSTEM_NAME=CrayLinuxEnvironment
```
```
@@ -137,15 +137,15 @@ cd build_knl
# run cmake with all the magic flags
export CC=`which icc`
export CXX=`which icpc`
-cmake <path to CMakeLists.txt> -DCMAKE_BUILD_TYPE=release -DNMC_THREADING_MODEL=tbb -DNMC_WITH_PROFILING=ON -DNMC_VECTORIZE_TARGET=KNL
+cmake <path to CMakeLists.txt> -DCMAKE_BUILD_TYPE=release -DARB_THREADING_MODEL=tbb -DARB_WITH_PROFILING=ON -DARB_VECTORIZE_TARGET=KNL
make -j
```
The flags passed into cmake are described:
- `-DCMAKE_BUILD_TYPE=release` : build in release mode with `-O3`.
- - `-DNMC_THREADING_MODEL=tbb` : use TBB for threading on multi-core
- - `-DNMC_WITH_PROFILING=ON` : use internal profilers that print profiling report at end
- - `-DNMC_VECTORIZE_TARGET=KNL` : generate AVX512 instructions, alternatively you can use:
+ - `-DARB_THREADING_MODEL=tbb` : use TBB for threading on multi-core
+ - `-DARB_WITH_PROFILING=ON` : use internal profilers that print profiling report at end
+ - `-DARB_VECTORIZE_TARGET=KNL` : generate AVX512 instructions, alternatively you can use:
- `AVX2` for Haswell & Broadwell
- `AVX` for Sandy Bridge and Ivy Bridge
diff --git a/doc/install.rst b/doc/install.rst
index bbd632526c46bfc93b543e98e1fb8ffbab967bc3..cf860e8ad230364f48a1739040c3aaaac6897c0c 100644
--- a/doc/install.rst
+++ b/doc/install.rst
@@ -31,9 +31,9 @@ The easiest way to acquire the latest version of Arbor is to check the code out
.. code-block:: bash
- git clone https://github.com/eth-cscs/nestmc-proto.git
+ git clone https://github.com/eth-cscs/arbor.git
-You can also point your browser to our `Github page <https://github.com/eth-cscs/nestmc-proto>`_ and download a zip file.
+You can also point your browser to our `Github page <https://github.com/eth-cscs/arbor>`_ and download a zip file.
.. _install_desktop:
diff --git a/doc/math/model/symbols.tex b/doc/math/model/symbols.tex
index 7f2bf3a352fe1527aa8647a9bc148bb53bffaddc..a771dca599c28aead66cd1f2e92f461d5444200d 100644
--- a/doc/math/model/symbols.tex
+++ b/doc/math/model/symbols.tex
@@ -35,7 +35,7 @@ We try to use the same units as NEURON, except for the specific membrane capacit
$I_e$ & $nA$ & $10^{-9} \cdot A$ & yes \\
\hline
\end{tabular}
-\caption{The units chosen for parameters and variables in NEST MC. The NEURON column indicates whether the same units have been used as NEURON.}
+\caption{The units chosen for parameters and variables in Arbor. The NEURON column indicates whether the same units have been used as NEURON.}
\label{tbl:units}
\end{table}
@@ -146,7 +146,7 @@ which can be expressed more compactly as
\end{equation}
where $G\in\mathbb{R}^{n\times n}$ is the conductance matrix, and $v, i \in \mathbb{R}^{n}$ are voltage and current vectors respectively.
-In NestMC the weights are chosen such that the conductance has units $\mu S$, voltage has units $mV$ and current has units $nA$.
+In Arbor the weights are chosen such that the conductance has units $\mu S$, voltage has units $mV$ and current has units $nA$.
\begin{center}
diff --git a/lmorpho/CMakeLists.txt b/lmorpho/CMakeLists.txt
index 126b93f8f280d888ddb4575a48dcc7a37d7ea316..67271afaf156482e1d7685efbaf61c898a416ede 100644
--- a/lmorpho/CMakeLists.txt
+++ b/lmorpho/CMakeLists.txt
@@ -1,5 +1,5 @@
add_executable(lmorpho lmorpho.cpp lsystem.cpp lsys_models.cpp morphio.cpp)
-target_link_libraries(lmorpho LINK_PUBLIC nestmc)
+target_link_libraries(lmorpho LINK_PUBLIC arbor)
target_link_libraries(lmorpho LINK_PUBLIC ${EXTERNAL_LIBRARIES})
diff --git a/lmorpho/morphio.cpp b/lmorpho/morphio.cpp
index dd9c36675c72ba6a6b310c1ebaf8954d2889db9b..9c3547296a864c7868bebb6731f47e957f2fef28 100644
--- a/lmorpho/morphio.cpp
+++ b/lmorpho/morphio.cpp
@@ -62,7 +62,7 @@ void multi_file::open(unsigned n) {
using arb::io::swc_record;
-// TODO: Move this functionality to nestmc library.
+// TODO: Move this functionality to arbor library.
std::vector<swc_record> as_swc(const arb::morphology& morph) {
using kind = swc_record::kind;
std::map<int, int> parent_end_id;
diff --git a/mechanisms/CMakeLists.txt b/mechanisms/CMakeLists.txt
index dbd1e1a9d7ebb084842c52d703a42af85d344d61..ba6e30623285036e735ceb5519a4dc05cc3c97b0 100644
--- a/mechanisms/CMakeLists.txt
+++ b/mechanisms/CMakeLists.txt
@@ -10,14 +10,14 @@ set(mod_srcdir "${CMAKE_CURRENT_SOURCE_DIR}/mod")
set(modcc_opt)
set(mech_dir "${CMAKE_CURRENT_SOURCE_DIR}/multicore")
file(MAKE_DIRECTORY "${mech_dir}")
-if(NMC_VECTORIZE_TARGET STREQUAL "KNL")
+if(ARB_VECTORIZE_TARGET STREQUAL "KNL")
set(modcc_target "avx512")
-elseif(NMC_VECTORIZE_TARGET STREQUAL "AVX")
+elseif(ARB_VECTORIZE_TARGET STREQUAL "AVX")
set(modcc_opt "-O")
set(modcc_target "cpu")
-elseif(NMC_VECTORIZE_TARGET STREQUAL "AVX2")
+elseif(ARB_VECTORIZE_TARGET STREQUAL "AVX2")
set(modcc_target "avx2")
-elseif(NMC_VECTORIZE_TARGET STREQUAL "AVX512")
+elseif(ARB_VECTORIZE_TARGET STREQUAL "AVX512")
set(modcc_target "avx512")
else()
set(modcc_target "cpu")
@@ -46,20 +46,20 @@ build_modules(
# Make a library with the implementations of the mechanism kernels
-if(NMC_WITH_CUDA)
+if(ARB_WITH_CUDA)
# make list of the .cu files that implement the mechanism kernels
foreach(mech ${mechanisms})
list(APPEND cuda_mech_sources ${mech_dir}/${mech}_impl.cu)
endforeach()
# compile the .cu files into a library
- cuda_add_library(gpu_mechanisms ${cuda_mech_sources})
+ cuda_add_library(arbormechcu ${cuda_mech_sources})
# force recompilation on changes to modcc or the underlying .mod files
- if (NMC_AUTO_RUN_MODCC_ON_CHANGES)
- add_dependencies(gpu_mechanisms build_all_gpu_mods)
+ if (ARB_AUTO_RUN_MODCC_ON_CHANGES)
+ add_dependencies(arbormechcu build_all_gpu_mods)
endif()
- list(APPEND NESTMC_LIBRARIES gpu_mechanisms)
- set(NESTMC_LIBRARIES "${NESTMC_LIBRARIES}" PARENT_SCOPE)
+ list(APPEND ARB_LIBRARIES arbormechcu)
+ set(ARB_LIBRARIES "${ARB_LIBRARIES}" PARENT_SCOPE)
endif()
diff --git a/miniapp/CMakeLists.txt b/miniapp/CMakeLists.txt
index c309edd910cb9ce2ae897ae26da7e966d9fd2c89..86d5f906c4c4ef38308c7205c47b780472a10967 100644
--- a/miniapp/CMakeLists.txt
+++ b/miniapp/CMakeLists.txt
@@ -8,10 +8,10 @@ set(MINIAPP_SOURCES
add_executable(miniapp.exe ${MINIAPP_SOURCES})
-target_link_libraries(miniapp.exe LINK_PUBLIC ${NESTMC_LIBRARIES})
+target_link_libraries(miniapp.exe LINK_PUBLIC ${ARB_LIBRARIES})
target_link_libraries(miniapp.exe LINK_PUBLIC ${EXTERNAL_LIBRARIES})
-if(NMC_WITH_MPI)
+if(ARB_WITH_MPI)
target_link_libraries(miniapp.exe LINK_PUBLIC ${MPI_C_LIBRARIES})
set_property(TARGET miniapp.exe APPEND_STRING PROPERTY LINK_FLAGS "${MPI_C_LINK_FLAGS}")
endif()
diff --git a/miniapp/README.md b/miniapp/README.md
index 9f8bd3822fdd8cca08442412100e5676564bdf17..066daf12b37bbc1dba23471ce7b700916133a56b 100644
--- a/miniapp/README.md
+++ b/miniapp/README.md
@@ -1,6 +1,6 @@
-# NEST MC miniapp
+# Arbor miniapp
-The miniapp is a simple benchmark of the NEST MC framework.
+The miniapp is a simple benchmark for the Arbor library.
## the model
diff --git a/miniapp/io.cpp b/miniapp/io.cpp
index 2fcee0ca3ca526ddae4553be073b9331d38195f0..b90d56bbae72f7722adff758e777e857744dc251 100644
--- a/miniapp/io.cpp
+++ b/miniapp/io.cpp
@@ -122,7 +122,7 @@ cl_options read_options(int argc, char** argv, bool allow_write) {
try {
cl_options defopts;
- CustomCmdLine cmd("nest mc miniapp harness", "0.1");
+ CustomCmdLine cmd("arbor miniapp harness", "0.1");
TCLAP::ValueArg<std::string> ifile_arg(
"i", "ifile",
diff --git a/miniapp/miniapp.cpp b/miniapp/miniapp.cpp
index 08a759d18f17561913bb6b75c789ccb2e5c66bda..b79b456b879ccf6680e63ba8bfa7bcfbbf571b20 100644
--- a/miniapp/miniapp.cpp
+++ b/miniapp/miniapp.cpp
@@ -193,7 +193,7 @@ int main(int argc, char** argv) {
void banner(hw::node_info nd) {
std::cout << "==========================================\n";
- std::cout << " NestMC miniapp\n";
+ std::cout << " Arbor miniapp\n";
std::cout << " - distributed : " << global_policy::size()
<< " (" << std::to_string(global_policy::kind()) << ")\n";
std::cout << " - threads : " << nd.num_cpu_cores
diff --git a/modcc/backends/avx2.hpp b/modcc/backends/avx2.hpp
index 081b3fe23263963a02655a923314bce06691f15a..ff8709cc9f5294a10b58eec1dc4660e0b9365f2a 100644
--- a/modcc/backends/avx2.hpp
+++ b/modcc/backends/avx2.hpp
@@ -77,7 +77,7 @@ struct simd_intrinsics<targetKind::avx2> {
tb << "_mm256_exp_pd(";
}
else {
- tb << "arb::multicore::nmc_mm256_exp_pd(";
+ tb << "arb::multicore::arb_mm256_exp_pd(";
}
break;
case tok::log:
@@ -85,7 +85,7 @@ struct simd_intrinsics<targetKind::avx2> {
tb << "_mm256_log_pd(";
}
else {
- tb << "arb::multicore::nmc_mm256_log_pd(";
+ tb << "arb::multicore::arb_mm256_log_pd(";
}
break;
default:
@@ -102,7 +102,7 @@ struct simd_intrinsics<targetKind::avx2> {
tb << "_mm256_pow_pd(";
}
else {
- tb << "arb::multicore::nmc_mm256_pow_pd(";
+ tb << "arb::multicore::arb_mm256_pow_pd(";
}
emit_operands(tb, arg_emitter(base), arg_emitter(exp));
diff --git a/scripts/travis/build.sh b/scripts/travis/build.sh
index 25620686923375c57894939e44ac0bbde2f3ed81..4ac2df17496c2a87a2756c89613fa553a9c6244b 100755
--- a/scripts/travis/build.sh
+++ b/scripts/travis/build.sh
@@ -44,7 +44,7 @@ cd $build_path
#
progress "Configuring with cmake"
-cmake_flags="-DNMC_WITH_ASSERTIONS=on -DNMC_THREADING_MODEL=${WITH_THREAD} -DNMC_DISTRIBUTED_MODEL=${WITH_DISTRIBUTED} ${CXX_FLAGS}"
+cmake_flags="-DARB_WITH_ASSERTIONS=on -DARB_THREADING_MODEL=${WITH_THREAD} -DARB_DISTRIBUTED_MODEL=${WITH_DISTRIBUTED} ${CXX_FLAGS}"
echo "cmake flags: ${cmake_flags}"
cmake .. ${cmake_flags} || error "unable to configure cmake"
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 2e06e3ab4e66e7d12a83af99334749942821a4c3..44488b94712c94b7b5e0fd99edfd1859f23086b2 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -38,30 +38,30 @@ set(CUDA_SOURCES
backends/gpu/kernels/time_ops.cu
)
-if(NMC_WITH_MPI)
+if(ARB_WITH_MPI)
set(BASE_SOURCES ${BASE_SOURCES} communication/mpi.cpp)
-elseif(NMC_WITH_DRYRUN)
+elseif(ARB_WITH_DRYRUN)
set(BASE_SOURCES ${BASE_SOURCES} communication/dryrun_global_policy.cpp)
endif()
-if(NMC_WITH_CTHREAD)
+if(ARB_WITH_CTHREAD)
set(BASE_SOURCES ${BASE_SOURCES} threading/cthread.cpp)
endif()
-add_library(nestmc ${BASE_SOURCES})
-list(APPEND NESTMC_LIBRARIES nestmc)
+add_library(arbor ${BASE_SOURCES})
+list(APPEND ARB_LIBRARIES arbor)
-if(NMC_WITH_CUDA)
- cuda_add_library(gpu ${CUDA_SOURCES})
- list(APPEND NESTMC_LIBRARIES gpu)
+if(ARB_WITH_CUDA)
+ cuda_add_library(arborcu ${CUDA_SOURCES})
+ list(APPEND ARB_LIBRARIES arborcu)
endif()
-if (NMC_AUTO_RUN_MODCC_ON_CHANGES)
- add_dependencies(nestmc build_all_mods)
- if (NMC_WITH_CUDA)
- add_dependencies(gpu build_all_gpu_mods)
+if (ARB_AUTO_RUN_MODCC_ON_CHANGES)
+ add_dependencies(arbor build_all_mods)
+ if (ARB_WITH_CUDA)
+ add_dependencies(arborcu build_all_gpu_mods)
endif()
endif()
-# Make changes to the NMC_LIBRARIES visible to the parent scope.
-set(NESTMC_LIBRARIES "${NESTMC_LIBRARIES}" PARENT_SCOPE)
+# Make changes to the ARB_LIBRARIES visible to the parent scope.
+set(ARB_LIBRARIES "${ARB_LIBRARIES}" PARENT_SCOPE)
diff --git a/src/backends/fvm.hpp b/src/backends/fvm.hpp
index d85f69c417462e3ca77d35e5f0c726aaa0f847f9..be7a11538c47b30595ea8c8a584f40d79c16bb95 100644
--- a/src/backends/fvm.hpp
+++ b/src/backends/fvm.hpp
@@ -34,7 +34,7 @@ struct null_backend: public multicore::backend {
} // namespace arb
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
#include <backends/gpu/fvm.hpp>
#else
namespace arb { namespace gpu {
diff --git a/src/backends/matrix_storage.md b/src/backends/matrix_storage.md
index d6ed40c2cf9121dc480f1a8604906e2e96b18166..f87f277c15f59ac8ef262685def7a19b5e50b015 100644
--- a/src/backends/matrix_storage.md
+++ b/src/backends/matrix_storage.md
@@ -8,9 +8,13 @@ An NxN Hines matrix can be stored compactly with 3 vectors of length N:
* `p`: the parent index
Additionally, we often store N*1 vectors that have one value per compartment, e.g. voltage, solution or rhs vectors.
-In NestMC multicompartment a single cell has a matrix structure associated with in, that is derived directly from the connections between its constituent compartments. NestMC groups these cells into groups of cells, called `cell_groups`. The matrices for all the cells in a group are packed together
+In Arbor, a single multicompartment cell has an associated matrix structure
+that is derived directly from the connections between its constituent
+compartments. Arbor groups these cells into groups of cells, called
+`cell_groups`. The matrices for all the cells in a group are packed together
-The matrix packing applies the same packing operation to each vector associated with a matrix, i.e. the `u`, `d`, `p` and solution, voltage vectors.
+The matrix packing applies the same packing operation to each vector associated
+with a matrix, i.e. the `u`, `d`, `p` and solution, voltage vectors.
In this discussion we use a simple example group of matrices to illustrate the storage methods, because an example is more illustrative than a formal description:
* 7 vectors labeled `{a, b, c, d, e, f, g}`
diff --git a/src/backends/multicore/intrin.hpp b/src/backends/multicore/intrin.hpp
index 05f5e2b5701e256e69c6a5321c591552cd3b6976..23d7b05cec1f5632cf634540dc76559a3198594a 100644
--- a/src/backends/multicore/intrin.hpp
+++ b/src/backends/multicore/intrin.hpp
@@ -55,24 +55,24 @@ constexpr int exp_bias = 1023;
constexpr double dsqrth = 0.70710678118654752440;
// Useful constants in vector registers
-const __m256d nmc_m256d_zero = _mm256_set1_pd(0.0);
-const __m256d nmc_m256d_one = _mm256_set1_pd(1.0);
-const __m256d nmc_m256d_two = _mm256_set1_pd(2.0);
-const __m256d nmc_m256d_nan = _mm256_set1_pd(std::numeric_limits<double>::quiet_NaN());
-const __m256d nmc_m256d_inf = _mm256_set1_pd(std::numeric_limits<double>::infinity());
-const __m256d nmc_m256d_ninf = _mm256_set1_pd(-std::numeric_limits<double>::infinity());
+const __m256d arb_m256d_zero = _mm256_set1_pd(0.0);
+const __m256d arb_m256d_one = _mm256_set1_pd(1.0);
+const __m256d arb_m256d_two = _mm256_set1_pd(2.0);
+const __m256d arb_m256d_nan = _mm256_set1_pd(std::numeric_limits<double>::quiet_NaN());
+const __m256d arb_m256d_inf = _mm256_set1_pd(std::numeric_limits<double>::infinity());
+const __m256d arb_m256d_ninf = _mm256_set1_pd(-std::numeric_limits<double>::infinity());
}
-static void nmc_mm256_print_pd(__m256d x, const char *name) __attribute__ ((unused));
-static void nmc_mm256_print_epi32(__m128i x, const char *name) __attribute__ ((unused));
-static void nmc_mm256_print_epi64x(__m256i x, const char *name) __attribute__ ((unused));
-static __m256d nmc_mm256_exp_pd(__m256d x) __attribute__ ((unused));
-static __m256d nmc_mm256_subnormal_pd(__m256d x) __attribute__ ((unused));
-static __m256d nmc_mm256_frexp_pd(__m256d x, __m128i *e) __attribute__ ((unused));
-static __m256d nmc_mm256_log_pd(__m256d x) __attribute__ ((unused));
-static __m256d nmc_mm256_pow_pd(__m256d x, __m256d y) __attribute__ ((unused));
+static void arb_mm256_print_pd(__m256d x, const char *name) __attribute__ ((unused));
+static void arb_mm256_print_epi32(__m128i x, const char *name) __attribute__ ((unused));
+static void arb_mm256_print_epi64x(__m256i x, const char *name) __attribute__ ((unused));
+static __m256d arb_mm256_exp_pd(__m256d x) __attribute__ ((unused));
+static __m256d arb_mm256_subnormal_pd(__m256d x) __attribute__ ((unused));
+static __m256d arb_mm256_frexp_pd(__m256d x, __m128i *e) __attribute__ ((unused));
+static __m256d arb_mm256_log_pd(__m256d x) __attribute__ ((unused));
+static __m256d arb_mm256_pow_pd(__m256d x, __m256d y) __attribute__ ((unused));
-void nmc_mm256_print_pd(__m256d x, const char *name) {
+void arb_mm256_print_pd(__m256d x, const char *name) {
double *val = (double *) &x;
std::cout << name << " = { ";
for (size_t i = 0; i < 4; ++i) {
@@ -82,7 +82,7 @@ void nmc_mm256_print_pd(__m256d x, const char *name) {
std::cout << "}\n";
}
-void nmc_mm256_print_epi32(__m128i x, const char *name) {
+void arb_mm256_print_epi32(__m128i x, const char *name) {
int *val = (int *) &x;
std::cout << name << " = { ";
for (size_t i = 0; i < 4; ++i) {
@@ -92,7 +92,7 @@ void nmc_mm256_print_epi32(__m128i x, const char *name) {
std::cout << "}\n";
}
-void nmc_mm256_print_epi64x(__m256i x, const char *name) {
+void arb_mm256_print_epi64x(__m256i x, const char *name) {
uint64_t *val = (uint64_t *) &x;
std::cout << name << " = { ";
for (size_t i = 0; i < 4; ++i) {
@@ -148,7 +148,7 @@ void nmc_mm256_print_epi64x(__m256i x, const char *name) {
// the standard library seems not to do that, so we are getting differences
// compared to std::exp() for large exponents.
//
-__m256d nmc_mm256_exp_pd(__m256d x) {
+__m256d arb_mm256_exp_pd(__m256d x) {
__m256d x_orig = x;
__m256d px = _mm256_floor_pd(
@@ -193,8 +193,8 @@ __m256d nmc_mm256_exp_pd(__m256d x) {
// Compute 1 + 2*P(x**2) / (Q(x**2)-P(x**2))
x = _mm256_div_pd(px, _mm256_sub_pd(qx, px));
- x = _mm256_add_pd(detail::nmc_m256d_one,
- _mm256_mul_pd(detail::nmc_m256d_two, x));
+ x = _mm256_add_pd(detail::arb_m256d_one,
+ _mm256_mul_pd(detail::arb_m256d_two, x));
// Finally, compute x *= 2**n
__m256i n64 = _mm256_cvtepi32_epi64(n);
@@ -211,23 +211,23 @@ __m256d nmc_mm256_exp_pd(__m256d x) {
);
__m256d is_nan = _mm256_cmp_pd(x_orig, x_orig, 3 /* _CMP_UNORD_Q */ );
- x = _mm256_blendv_pd(x, detail::nmc_m256d_inf, is_large);
- x = _mm256_blendv_pd(x, detail::nmc_m256d_zero, is_small);
- x = _mm256_blendv_pd(x, detail::nmc_m256d_nan, is_nan);
+ x = _mm256_blendv_pd(x, detail::arb_m256d_inf, is_large);
+ x = _mm256_blendv_pd(x, detail::arb_m256d_zero, is_small);
+ x = _mm256_blendv_pd(x, detail::arb_m256d_nan, is_nan);
return x;
}
-__m256d nmc_mm256_subnormal_pd(__m256d x) {
+__m256d arb_mm256_subnormal_pd(__m256d x) {
__m256i x_raw = _mm256_castpd_si256(x);
__m256i exp_mask = _mm256_set1_epi64x(detail::dexp_mask);
__m256d x_exp = _mm256_castsi256_pd(_mm256_and_si256(x_raw, exp_mask));
// Subnormals have a zero exponent
- return _mm256_cmp_pd(x_exp, detail::nmc_m256d_zero, 0 /* _CMP_EQ_OQ */);
+ return _mm256_cmp_pd(x_exp, detail::arb_m256d_zero, 0 /* _CMP_EQ_OQ */);
}
-__m256d nmc_mm256_frexp_pd(__m256d x, __m128i *e) {
+__m256d arb_mm256_frexp_pd(__m256d x, __m128i *e) {
__m256i exp_mask = _mm256_set1_epi64x(detail::dexp_mask);
__m256i mant_mask = _mm256_set1_epi64x(detail::dmant_mask);
@@ -252,13 +252,13 @@ __m256d nmc_mm256_frexp_pd(__m256d x, __m128i *e) {
// Treat special cases
__m256d is_zero = _mm256_cmp_pd(
- x_orig, detail::nmc_m256d_zero, 0 /* _CMP_EQ_OQ */
+ x_orig, detail::arb_m256d_zero, 0 /* _CMP_EQ_OQ */
);
__m256d is_inf = _mm256_cmp_pd(
- x_orig, detail::nmc_m256d_inf, 0 /* _CMP_EQ_OQ */
+ x_orig, detail::arb_m256d_inf, 0 /* _CMP_EQ_OQ */
);
__m256d is_ninf = _mm256_cmp_pd(
- x_orig, detail::nmc_m256d_ninf, 0 /* _CMP_EQ_OQ */
+ x_orig, detail::arb_m256d_ninf, 0 /* _CMP_EQ_OQ */
);
__m256d is_nan = _mm256_cmp_pd(x_orig, x_orig, 3 /* _CMP_UNORD_Q */ );
@@ -266,13 +266,13 @@ __m256d nmc_mm256_frexp_pd(__m256d x, __m128i *e) {
// have already prepared it as a power of 2
__m256i is_denorm = _mm256_cmpeq_epi64(x_exp, _mm256_set1_epi64x(-1022));
- x = _mm256_blendv_pd(x, detail::nmc_m256d_zero, is_zero);
- x = _mm256_blendv_pd(x, detail::nmc_m256d_inf, is_inf);
- x = _mm256_blendv_pd(x, detail::nmc_m256d_ninf, is_ninf);
- x = _mm256_blendv_pd(x, detail::nmc_m256d_nan, is_nan);
+ x = _mm256_blendv_pd(x, detail::arb_m256d_zero, is_zero);
+ x = _mm256_blendv_pd(x, detail::arb_m256d_inf, is_inf);
+ x = _mm256_blendv_pd(x, detail::arb_m256d_ninf, is_ninf);
+ x = _mm256_blendv_pd(x, detail::arb_m256d_nan, is_nan);
// FIXME: We treat denormalized numbers as zero here
- x = _mm256_blendv_pd(x, detail::nmc_m256d_zero,
+ x = _mm256_blendv_pd(x, detail::arb_m256d_zero,
_mm256_castsi256_pd(is_denorm));
x_exp = _mm256_blendv_epi8(x_exp, _mm256_set1_epi64x(0), is_denorm);
@@ -299,12 +299,12 @@ __m256d nmc_mm256_frexp_pd(__m256d x, __m128i *e) {
//
// ln(1+x) = x - 0.5*x^2 + x^3*P(x)/Q(x)
//
-__m256d nmc_mm256_log_pd(__m256d x) {
+__m256d arb_mm256_log_pd(__m256d x) {
__m256d x_orig = x;
__m128i x_exp;
// x := x', x_exp := g
- x = nmc_mm256_frexp_pd(x, &x_exp);
+ x = arb_mm256_frexp_pd(x, &x_exp);
// convert x_exp to packed double
__m256d dx_exp = _mm256_cvtepi32_pd(x_exp);
@@ -320,13 +320,13 @@ __m256d nmc_mm256_log_pd(__m256d x) {
// Precompute both branches
// 2*x - 1
- __m256d x2m1 = _mm256_sub_pd(_mm256_add_pd(x, x), detail::nmc_m256d_one);
+ __m256d x2m1 = _mm256_sub_pd(_mm256_add_pd(x, x), detail::arb_m256d_one);
// x - 1
- __m256d xm1 = _mm256_sub_pd(x, detail::nmc_m256d_one);
+ __m256d xm1 = _mm256_sub_pd(x, detail::arb_m256d_one);
// dx_exp - 1
- __m256d dx_exp_m1 = _mm256_sub_pd(dx_exp, detail::nmc_m256d_one);
+ __m256d dx_exp_m1 = _mm256_sub_pd(dx_exp, detail::arb_m256d_one);
x = _mm256_blendv_pd(xm1, x2m1, lt_sqrth);
dx_exp = _mm256_blendv_pd(dx_exp, dx_exp_m1, lt_sqrth);
@@ -381,25 +381,25 @@ __m256d nmc_mm256_log_pd(__m256d x) {
// Treat exceptional cases
__m256d is_inf = _mm256_cmp_pd(
- x_orig, detail::nmc_m256d_inf, 0 /* _CMP_EQ_OQ */);
+ x_orig, detail::arb_m256d_inf, 0 /* _CMP_EQ_OQ */);
__m256d is_zero = _mm256_cmp_pd(
- x_orig, detail::nmc_m256d_zero, 0 /* _CMP_EQ_OQ */);
+ x_orig, detail::arb_m256d_zero, 0 /* _CMP_EQ_OQ */);
__m256d is_neg = _mm256_cmp_pd(
- x_orig, detail::nmc_m256d_zero, 17 /* _CMP_LT_OQ */);
- __m256d is_denorm = nmc_mm256_subnormal_pd(x_orig);
+ x_orig, detail::arb_m256d_zero, 17 /* _CMP_LT_OQ */);
+ __m256d is_denorm = arb_mm256_subnormal_pd(x_orig);
- ret = _mm256_blendv_pd(ret, detail::nmc_m256d_inf, is_inf);
- ret = _mm256_blendv_pd(ret, detail::nmc_m256d_ninf, is_zero);
+ ret = _mm256_blendv_pd(ret, detail::arb_m256d_inf, is_inf);
+ ret = _mm256_blendv_pd(ret, detail::arb_m256d_ninf, is_zero);
// We treat denormalized cases as zeros
- ret = _mm256_blendv_pd(ret, detail::nmc_m256d_ninf, is_denorm);
- ret = _mm256_blendv_pd(ret, detail::nmc_m256d_nan, is_neg);
+ ret = _mm256_blendv_pd(ret, detail::arb_m256d_ninf, is_denorm);
+ ret = _mm256_blendv_pd(ret, detail::arb_m256d_nan, is_neg);
return ret;
}
// Equivalent to exp(y*log(x))
-__m256d nmc_mm256_pow_pd(__m256d x, __m256d y) {
- return nmc_mm256_exp_pd(_mm256_mul_pd(y, nmc_mm256_log_pd(x)));
+__m256d arb_mm256_pow_pd(__m256d x, __m256d y) {
+ return arb_mm256_exp_pd(_mm256_mul_pd(y, arb_mm256_log_pd(x)));
}
} // end namespace multicore
diff --git a/src/communication/global_policy.hpp b/src/communication/global_policy.hpp
index 20af00fc999ee843f546b729a61659f6c02c88fa..e26406beee393a796a8de485e371922aee1b0d8f 100644
--- a/src/communication/global_policy.hpp
+++ b/src/communication/global_policy.hpp
@@ -20,9 +20,9 @@ namespace std {
}
}
-#if defined(NMC_HAVE_MPI)
+#if defined(ARB_HAVE_MPI)
#include "mpi_global_policy.hpp"
-#elif defined(NMC_HAVE_DRYRUN)
+#elif defined(ARB_HAVE_DRYRUN)
#include "dryrun_global_policy.hpp"
#else
#include "serial_global_policy.hpp"
diff --git a/src/communication/mpi_global_policy.hpp b/src/communication/mpi_global_policy.hpp
index 32f5beedd81b60de4a41432673e1f7c25949fe65..49ff2cb611040956195debe3351c5df1ddf75462 100644
--- a/src/communication/mpi_global_policy.hpp
+++ b/src/communication/mpi_global_policy.hpp
@@ -1,7 +1,7 @@
#pragma once
-#ifndef NMC_HAVE_MPI
-#error "mpi_global_policy.hpp should only be compiled in a NMC_HAVE_MPI build"
+#ifndef ARB_HAVE_MPI
+#error "mpi_global_policy.hpp should only be compiled in a ARB_HAVE_MPI build"
#endif
#include <cstdint>
diff --git a/src/hardware/gpu.cpp b/src/hardware/gpu.cpp
index 6e8ac4cb6c45ce8522696272916822165f0b21ba..bd5f2584acf898a1b93dc4ba2396fbe1581a6ce5 100644
--- a/src/hardware/gpu.cpp
+++ b/src/hardware/gpu.cpp
@@ -1,11 +1,11 @@
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
#include <cuda_runtime.h>
#endif
namespace arb {
namespace hw {
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
unsigned num_gpus() {
int n;
cudaGetDeviceCount(&n);
diff --git a/src/hardware/memory.cpp b/src/hardware/memory.cpp
index 6cce80dac82b2eb3ebe1464bee74d3509942d34f..65c085a530f8899187ac5e427004a4d5f029d49b 100644
--- a/src/hardware/memory.cpp
+++ b/src/hardware/memory.cpp
@@ -6,7 +6,7 @@ extern "C" {
}
#endif
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
#include <cuda_runtime.h>
#endif
@@ -24,7 +24,7 @@ memory_size_type allocated_memory() {
}
#endif
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
memory_size_type gpu_allocated_memory() {
std::size_t free;
std::size_t total;
diff --git a/src/hardware/power.cpp b/src/hardware/power.cpp
index 05c5303f6ddb7c3af6c843635f2b0cff9e4dca80..4819b6307a481dd8e59623bf6c79b787d3a4a22b 100644
--- a/src/hardware/power.cpp
+++ b/src/hardware/power.cpp
@@ -5,7 +5,7 @@
namespace arb {
namespace hw {
-#ifdef NMC_HAVE_CRAY
+#ifdef ARB_HAVE_CRAY
energy_size_type energy() {
energy_size_type result = energy_size_type(-1);
diff --git a/src/memory/allocator.hpp b/src/memory/allocator.hpp
index 74b8c921e7f32d26947ef033a31e9d11eb2c6bc2..3774cd30fd5a794a84676e354d5445122203e95a 100644
--- a/src/memory/allocator.hpp
+++ b/src/memory/allocator.hpp
@@ -2,7 +2,7 @@
#include <limits>
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
#include <cuda.h>
#include <cuda_runtime.h>
#endif
@@ -137,7 +137,7 @@ namespace impl {
}
#endif
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
namespace cuda {
template <size_type Alignment>
class pinned_policy {
@@ -247,7 +247,7 @@ namespace impl {
}
};
} // namespace cuda
-#endif // #ifdef NMC_HAVE_GPU
+#endif // #ifdef ARB_HAVE_GPU
} // namespace impl
template<typename T, typename Policy >
@@ -325,7 +325,7 @@ namespace util {
}
};
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
template <size_t Alignment>
struct type_printer<impl::cuda::pinned_policy<Alignment>>{
static std::string print() {
@@ -371,7 +371,7 @@ template <class T, size_t alignment=(512/8)>
using hbw_allocator = allocator<T, impl::knl::hbw_policy<alignment>>;
#endif
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
// For pinned and allocation set the default alignment to correspond to
// the alignment of 1024 bytes, because pinned memory is allocated at
// page boundaries. It is allocated at page boundaries (typically 4k),
diff --git a/src/memory/gpu.hpp b/src/memory/gpu.hpp
index bb187e6bf25dfa4cf61a408fda00025e6fe31206..6ba8699b96918985975db120c84844c3bcf5ac6b 100644
--- a/src/memory/gpu.hpp
+++ b/src/memory/gpu.hpp
@@ -1,6 +1,6 @@
#pragma once
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
#include <string>
#include <cstdint>
diff --git a/src/memory/host_coordinator.hpp b/src/memory/host_coordinator.hpp
index 329521a50cca9b21618c912656c383d55c152349..607fb954b55f75613e32d2e0a238554d42cff0e2 100644
--- a/src/memory/host_coordinator.hpp
+++ b/src/memory/host_coordinator.hpp
@@ -11,7 +11,7 @@
#include "allocator.hpp"
#include "util.hpp"
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
#include "gpu.hpp"
#endif
@@ -22,7 +22,7 @@ namespace memory {
template <typename T, class Allocator>
class host_coordinator;
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
template <typename T, class Allocator>
class device_coordinator;
#endif
@@ -123,7 +123,7 @@ public:
std::copy(from.begin(), from.end(), to.begin());
}
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
// copy memory from device to host
template <class Alloc>
void copy(
diff --git a/src/memory/memory.hpp b/src/memory/memory.hpp
index c0b660d8dd7bf8c4bd2c960a2a8a22aaa8c56e91..fe5f9b7090ecdbff609d73c9403ba0271a3113e8 100644
--- a/src/memory/memory.hpp
+++ b/src/memory/memory.hpp
@@ -6,7 +6,7 @@
#include "definitions.hpp"
#include "host_coordinator.hpp"
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
#include "device_coordinator.hpp"
#endif
@@ -28,7 +28,7 @@ std::ostream& operator<< (std::ostream& o, host_view<T> const& v) {
return o;
}
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
// specialization for pinned vectors. Use a host_coordinator, because memory is
// in the host memory space, and all of the helpers (copy, set, etc) are the
// same with and without page locked memory
diff --git a/src/memory/wrappers.hpp b/src/memory/wrappers.hpp
index 50a9092399ecabcb979840bd1b327c3f31977911..0371370927c05b03cae5450e4c18ab134124dc02 100644
--- a/src/memory/wrappers.hpp
+++ b/src/memory/wrappers.hpp
@@ -5,7 +5,7 @@
#include <memory/memory.hpp>
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
#include <cuda.h>
#include <cuda_runtime.h>
#endif
@@ -95,7 +95,7 @@ namespace util {
return is_on_host<typename std::decay<T>::type>::value;
}
- #ifdef NMC_HAVE_GPU
+ #ifdef ARB_HAVE_GPU
template <typename T>
struct is_on_gpu : std::false_type {};
@@ -131,7 +131,7 @@ auto on_host(const C& c) -> decltype(make_const_view(c)) {
return make_const_view(c);
}
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
template <
typename C,
typename = typename std::enable_if<util::is_on_gpu_v<C>()>::type
diff --git a/src/profiling/profiler.cpp b/src/profiling/profiler.cpp
index b260e2faf477b80f942dab7d8778c1bce4ab4673..703766963340c950fe2fb1c057e4d841df82177f 100644
--- a/src/profiling/profiler.cpp
+++ b/src/profiling/profiler.cpp
@@ -1,6 +1,6 @@
#include <numeric>
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
#include <cuda_profiler_api.h>
#endif
@@ -18,7 +18,7 @@ namespace util {
// a program control which parts of the program are to be profiled. It is a
// simple wrapper around the API calls with a mutex to ensure correct behaviour
// when multiple threads attempt to start or stop the profiler.
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
namespace gpu {
bool is_running_nvprof = false;
std::mutex gpu_profiler_mutex;
@@ -298,7 +298,7 @@ profiler_node profiler::performance_tree() {
}
-#ifdef NMC_HAVE_PROFILING
+#ifdef ARB_HAVE_PROFILING
namespace data {
profiler_wrapper profilers_(profiler("root"));
}
diff --git a/src/profiling/profiler.hpp b/src/profiling/profiler.hpp
index fb5c09aa7ba51aea0c41c6f5174b95544b2b17db..75281e3e846632b0ad7371abd69a6c592ee553ba 100644
--- a/src/profiling/profiler.hpp
+++ b/src/profiling/profiler.hpp
@@ -202,7 +202,7 @@ private:
region_type* current_region_ = &root_region_;
};
-#ifdef NMC_HAVE_PROFILING
+#ifdef ARB_HAVE_PROFILING
namespace data {
using profiler_wrapper = arb::threading::enumerable_thread_specific<profiler>;
extern profiler_wrapper profilers_;
@@ -225,7 +225,7 @@ void profiler_enter(const char* n);
/// enter nested profiler regions in a single call
template <class...Args>
void profiler_enter(const char* n, Args... args) {
-#ifdef NMC_HAVE_PROFILING
+#ifdef ARB_HAVE_PROFILING
get_profiler().enter(n);
profiler_enter(args...);
#endif
diff --git a/src/threading/cthread.hpp b/src/threading/cthread.hpp
index 57bdbc32b6e75bdc4e7b0ec928b96c620cfaec84..da70a2b54d6490eb8e74fd179de7eee3f2b4dc34 100644
--- a/src/threading/cthread.hpp
+++ b/src/threading/cthread.hpp
@@ -1,7 +1,7 @@
#pragma once
-#if !defined(NMC_HAVE_CTHREAD)
- #error "this header can only be loaded if NMC_HAVE_CTHREAD is set"
+#if !defined(ARB_HAVE_CTHREAD)
+ #error "this header can only be loaded if ARB_HAVE_CTHREAD is set"
#endif
// task_group definition
diff --git a/src/threading/serial.hpp b/src/threading/serial.hpp
index 0a182eafa1c2785064e8118095bb6c7067921aee..8f54823b39f3633bb0e67c2864e775cd63932d22 100644
--- a/src/threading/serial.hpp
+++ b/src/threading/serial.hpp
@@ -1,7 +1,7 @@
#pragma once
-#if !defined(NMC_HAVE_SERIAL)
- #error "this header can only be loaded if NMC_HAVE_SERIAL is set"
+#if !defined(ARB_HAVE_SERIAL)
+ #error "this header can only be loaded if ARB_HAVE_SERIAL is set"
#endif
#include <algorithm>
diff --git a/src/threading/tbb.hpp b/src/threading/tbb.hpp
index 59dc8ebf27e3d22872ab875d3af185fbb464523b..e9760c21ca3e40cb03814fa5095f3a851e2e96dc 100644
--- a/src/threading/tbb.hpp
+++ b/src/threading/tbb.hpp
@@ -1,7 +1,7 @@
#pragma once
-#if !defined(NMC_HAVE_TBB)
- #error this header can only be loaded if NMC_HAVE_TBB is set
+#if !defined(ARB_HAVE_TBB)
+ #error this header can only be loaded if ARB_HAVE_TBB is set
#endif
#include <string>
diff --git a/src/threading/threading.cpp b/src/threading/threading.cpp
index 1316949b5c9e977d2edf14c1447954237fefe0f4..35ebf467872f29ae8b362770a6c7c47179ff37f2 100644
--- a/src/threading/threading.cpp
+++ b/src/threading/threading.cpp
@@ -13,27 +13,27 @@ namespace threading {
// Test environment variables for user-specified count of threads.
//
-// NMC_NUM_THREADS is used if set, otherwise OMP_NUM_THREADS is used.
+// ARB_NUM_THREADS is used if set, otherwise OMP_NUM_THREADS is used.
//
// If neither variable is set, returns no value.
//
// Valid values for the environment variable are:
-// 0 : NestMC is responsible for picking the number of threads.
+// 0 : Arbor is responsible for picking the number of threads.
// >0: The number of threads to use.
//
// Throws std::runtime_error:
-// NMC_NUM_THREADS or OMP_NUM_THREADS is set with invalid value.
+// ARB_NUM_THREADS or OMP_NUM_THREADS is set with invalid value.
util::optional<size_t> get_env_num_threads() {
const char* str;
// select variable to use:
- // If NMC_NUM_THREADS_VAR is set, use $NMC_NUM_THREADS_VAR
- // else if NMC_NUM_THREAD set, use it
+ // If ARB_NUM_THREADS_VAR is set, use $ARB_NUM_THREADS_VAR
+ // else if ARB_NUM_THREAD set, use it
// else if OMP_NUM_THREADS set, use it
- if (auto nthreads_var_name = std::getenv("NMC_NUM_THREADS_VAR")) {
+ if (auto nthreads_var_name = std::getenv("ARB_NUM_THREADS_VAR")) {
str = std::getenv(nthreads_var_name);
}
- else if (! (str = std::getenv("NMC_NUM_THREADS"))) {
+ else if (! (str = std::getenv("ARB_NUM_THREADS"))) {
str = std::getenv("OMP_NUM_THREADS");
}
@@ -71,7 +71,7 @@ size_t num_threads_init() {
// number of threads.
size_t num_threads() {
// TODO: this is a bit of a hack until we have user-configurable threading.
-#if defined(NMC_HAVE_SERIAL)
+#if defined(ARB_HAVE_SERIAL)
return 1;
#else
static size_t num_threads_cached = num_threads_init();
diff --git a/src/threading/threading.hpp b/src/threading/threading.hpp
index 9049d0a9d3a0425b8b383cd1c2ea35fd4eb5d5c4..2e90ac36219eab4465e32639d5b7a08e6d485249 100644
--- a/src/threading/threading.hpp
+++ b/src/threading/threading.hpp
@@ -7,13 +7,13 @@ namespace threading {
// Test environment variables for user-specified count of threads.
// Potential environment variables are tested in this order:
-// 1. use the environment variable specified by NMC_NUM_THREADS_VAR
-// 2. use NMC_NUM_THREADS
+// 1. use the environment variable specified by ARB_NUM_THREADS_VAR
+// 2. use ARB_NUM_THREADS
// 3. use OMP_NUM_THREADS
// 4. If no variable is set, returns no value.
//
// Valid values for the environment variable are:
-// 0 : NestMC is responsible for picking the number of threads.
+// 0 : Arbor is responsible for picking the number of threads.
// >0 : The number of threads to use.
//
// Throws std::runtime_error:
@@ -25,11 +25,11 @@ size_t num_threads();
} // namespace threading
} // namespace arb
-#if defined(NMC_HAVE_TBB)
+#if defined(ARB_HAVE_TBB)
#include "tbb.hpp"
-#elif defined(NMC_HAVE_CTHREAD)
+#elif defined(ARB_HAVE_CTHREAD)
#include "cthread.hpp"
#else
- #define NMC_HAVE_SERIAL
+ #define ARB_HAVE_SERIAL
#include "serial.hpp"
#endif
diff --git a/src/util/config.hpp b/src/util/config.hpp
index 5103bcf3e599a4f5425a5553d9016bb5413b51c5..dd8f867151cc96cfa357067d97e1fc6d2b2be4ea 100644
--- a/src/util/config.hpp
+++ b/src/util/config.hpp
@@ -23,13 +23,13 @@ constexpr bool has_memory_measurement = true;
constexpr bool has_memory_measurement = false;
#endif
-#ifdef NMC_HAVE_CRAY
+#ifdef ARB_HAVE_CRAY
constexpr bool has_power_measurement = true;
#else
constexpr bool has_power_measurement = false;
#endif
-#ifdef NMC_HAVE_GPU
+#ifdef ARB_HAVE_GPU
constexpr bool has_cuda = true;
#else
constexpr bool has_cuda = false;
diff --git a/src/util/debug.hpp b/src/util/debug.hpp
index 331ca6af8888e96d8e4b51fbb62b6482835547cf..214e2269f0602c2cd437e78190b03b629f9c7f5f 100644
--- a/src/util/debug.hpp
+++ b/src/util/debug.hpp
@@ -97,13 +97,13 @@ impl::sepval<Seq, Separator> sepval(const Seq& seq, Separator sep) {
} // namespace util
} // namespace arb
-#ifdef NMC_HAVE_TRACE
+#ifdef ARB_HAVE_TRACE
#define TRACE(vars...) arb::util::debug_emit_trace(__FILE__, __LINE__, #vars, ##vars)
#else
#define TRACE(...)
#endif
-#ifdef NMC_HAVE_ASSERTIONS
+#ifdef ARB_HAVE_ASSERTIONS
#ifdef __GNUC__
#define DEBUG_FUNCTION_NAME __PRETTY_FUNCTION__
#else
@@ -116,4 +116,4 @@ impl::sepval<Seq, Separator> sepval(const Seq& seq, Separator sep) {
#else
#define EXPECTS(condition) \
(void)(false && (condition))
-#endif // def NMC_HAVE_ASSERTIONS
+#endif // def ARB_HAVE_ASSERTIONS
diff --git a/src/util/range.hpp b/src/util/range.hpp
index c177f80b328faa0c6822e6e9297ef2eec64250ad..75e2b5791062a33f2ea283ae40a63d16aec695d5 100644
--- a/src/util/range.hpp
+++ b/src/util/range.hpp
@@ -26,7 +26,7 @@
#include <type_traits>
#include <utility>
-#ifdef NMC_HAVE_TBB
+#ifdef ARB_HAVE_TBB
#include <tbb/tbb_stddef.h>
#endif
@@ -108,7 +108,7 @@ struct range {
return (*this)[n];
}
-#ifdef NMC_HAVE_TBB
+#ifdef ARB_HAVE_TBB
template <
typename V = iterator,
typename = enable_if_t<is_forward_iterator<V>::value>
diff --git a/tests/global_communication/CMakeLists.txt b/tests/global_communication/CMakeLists.txt
index b6e38f99c4c2a4f3b571907f6bb4c1a9b5109b64..88061eec44842ffe1f19ab8e1303989d8a793bcd 100644
--- a/tests/global_communication/CMakeLists.txt
+++ b/tests/global_communication/CMakeLists.txt
@@ -16,10 +16,10 @@ add_executable(global_communication.exe ${COMMUNICATION_SOURCES} ${HEADERS})
set(TARGETS global_communication.exe)
foreach(target ${TARGETS})
- target_link_libraries(${target} LINK_PUBLIC nestmc gtest)
+ target_link_libraries(${target} LINK_PUBLIC arbor gtest)
target_link_libraries(${target} LINK_PUBLIC ${EXTERNAL_LIBRARIES})
- if(NMC_WITH_MPI)
+ if(ARB_WITH_MPI)
target_link_libraries(${target} LINK_PUBLIC ${MPI_C_LIBRARIES})
set_property(TARGET ${target} APPEND_STRING PROPERTY LINK_FLAGS "${MPI_C_LINK_FLAGS}")
endif()
diff --git a/tests/global_communication/test_communicator.cpp b/tests/global_communication/test_communicator.cpp
index 532e7444ff34e3b2c5140e62fdd713abd74bea18..147daead5124b2c7acc34022505f724be40efb38 100644
--- a/tests/global_communication/test_communicator.cpp
+++ b/tests/global_communication/test_communicator.cpp
@@ -38,7 +38,7 @@ TEST(communicator, policy_basics) {
//
// Here we defined proxy types for testing the gather_spikes functionality.
// These are a little bit simpler than the spike and source types used inside
-// NestMC, to simplify the testing.
+// Arbor, to simplify the testing.
// Proxy for a spike source, which represents gid as an integer.
struct source_proxy {
diff --git a/tests/global_communication/test_mpi.cpp b/tests/global_communication/test_mpi.cpp
index d122ff3660313f3a01a934678f11c9a482396a12..94c1559de19df0e1c26b3b1ce4330def6c4a71aa 100644
--- a/tests/global_communication/test_mpi.cpp
+++ b/tests/global_communication/test_mpi.cpp
@@ -1,4 +1,4 @@
-#ifdef NMC_HAVE_MPI
+#ifdef ARB_HAVE_MPI
#include "../gtest.h"
@@ -145,4 +145,4 @@ TEST(mpi, gather) {
}
}
-#endif // NMC_HAVE_MPI
+#endif // ARB_HAVE_MPI
diff --git a/tests/performance/io/CMakeLists.txt b/tests/performance/io/CMakeLists.txt
index 0d9b1b1e6a1a26123d6d43f7f58a97b9eaa6211d..d6043b491301507e82e76ce1a643aaec6c9c1bec 100644
--- a/tests/performance/io/CMakeLists.txt
+++ b/tests/performance/io/CMakeLists.txt
@@ -7,10 +7,10 @@ set(DISK_IO_SOURCES
add_executable(disk_io.exe ${DISK_IO_SOURCES} ${HEADERS})
-target_link_libraries(disk_io.exe LINK_PUBLIC nestmc)
+target_link_libraries(disk_io.exe LINK_PUBLIC arbor)
target_link_libraries(disk_io.exe LINK_PUBLIC ${EXTERNAL_LIBRARIES})
-if(NMC_WITH_MPI)
+if(ARB_WITH_MPI)
target_link_libraries(disk_io.exe LINK_PUBLIC ${MPI_C_LIBRARIES})
set_property(TARGET disk_io.exe APPEND_STRING PROPERTY LINK_FLAGS "${MPI_C_LINK_FLAGS}")
endif()
diff --git a/tests/ubench/CMakeLists.txt b/tests/ubench/CMakeLists.txt
index 94bd3979c223dbd880e0d89625450354943e9fdc..258f646a7f75030133f50052d0dbbab39b9b1def 100644
--- a/tests/ubench/CMakeLists.txt
+++ b/tests/ubench/CMakeLists.txt
@@ -69,14 +69,14 @@ foreach(bench_src ${bench_sources})
endforeach()
-if(NMC_WITH_CUDA)
+if(ARB_WITH_CUDA)
cuda_include_directories("${gbench_install_dir}/include")
foreach(bench_src ${bench_sources_cuda})
string(REGEX REPLACE "\\.[^.]*$" "" bench_exe "${bench_src}")
cuda_add_executable("${bench_exe}" EXCLUDE_FROM_ALL "${bench_src}")
add_dependencies("${bench_exe}" gbench)
target_link_libraries("${bench_exe}" "${gbench_install_dir}/lib/libbenchmark.a")
- target_link_libraries("${bench_exe}" LINK_PUBLIC ${NESTMC_LIBRARIES})
+ target_link_libraries("${bench_exe}" LINK_PUBLIC ${ARB_LIBRARIES})
list(APPEND bench_exe_list ${bench_exe})
endforeach()
diff --git a/tests/ubench/README.md b/tests/ubench/README.md
index 80a7c538a18862ea65a5b30d86c84b01471a09d9..6ab4186760c1b84246319f7db5685ef649eddc6e 100644
--- a/tests/ubench/README.md
+++ b/tests/ubench/README.md
@@ -126,11 +126,15 @@ Platform:
#### Motivation
-The reduction by key pattern with repeated keys is used when "point process" mechanism contributions to currents are collected.
-More than one point process, typically synapses, can be attached to a compartment,
-and when their contributions are computed and added to the per-compartment current in parallel, care must be taken to avoid race conditions.
-Early versions of NestMC used cuda atomic operations to perform the accumulation, which works quite well up to a certain point.
-However, performance with atomics decreases as the number of synapses per compartment increases, i.e. as the number of threads performing simultatneous atomic updates on the same location increases.
+The reduction by key pattern with repeated keys is used when "point process"
+mechanism contributions to currents are collected. More than one point process,
+typically synapses, can be attached to a compartment, and when their
+contributions are computed and added to the per-compartment current in
+parallel, care must be taken to avoid race conditions. Early versions of Arbor
+used cuda atomic operations to perform the accumulation, which works quite well
+up to a certain point. However, performance with atomics decreases as the
+number of synapses per compartment increases, i.e. as the number of threads
+performing simultatneous atomic updates on the same location increases.
#### Implementations
@@ -153,11 +157,15 @@ Platform:
* gcc version 5.2.0
* nvcc version 8.0.61
-Results are presented as speedup for warp intrinsics vs atomics, for both single and double precision.
-Note that the P100 GPU has hardware support for double precision atomics, and we expect much larger speedup for double precision on Keplar GPUs that emulate double precision atomics with CAS.
-The benchmark updates `n` locations, each with an average density of `d` keys per location.
-This is equivalent to `n` compartments with `d` synapses per compartment.
-Atomics are faster for the case where both `n` and `d` are small, however the gpu is backend is for throughput simulations, with large cell groups with at least 10k compartments in total.
+Results are presented as speedup for warp intrinsics vs atomics, for both
+single and double precision. Note that the P100 GPU has hardware support for
+double precision atomics, and we expect much larger speedup for double
+precision on Keplar GPUs that emulate double precision atomics with CAS. The
+benchmark updates `n` locations, each with an average density of `d` keys per
+location. This is equivalent to `n` compartments with `d` synapses per
+compartment. Atomics are faster for the case where both `n` and `d` are small,
+however the gpu is backend is for throughput simulations, with large cell
+groups with at least 10k compartments in total.
*float*
diff --git a/tests/unit/CMakeLists.txt b/tests/unit/CMakeLists.txt
index 5d58b80ad140ca30a98e960805a510eaed5a6751..11808a1371202ae41efc558c8203aa018661cdcf 100644
--- a/tests/unit/CMakeLists.txt
+++ b/tests/unit/CMakeLists.txt
@@ -87,7 +87,7 @@ set(TEST_SOURCES
stats.cpp
)
-if(NMC_VECTORIZE_TARGET STREQUAL "AVX2")
+if(ARB_VECTORIZE_TARGET STREQUAL "AVX2")
list(APPEND TEST_SOURCES test_intrin.cpp)
endif()
@@ -96,23 +96,23 @@ set(TARGETS test.exe)
add_executable(test.exe ${TEST_SOURCES})
target_compile_definitions(test.exe PUBLIC "-DDATADIR=\"${PROJECT_SOURCE_DIR}/data\"")
-if (NMC_AUTO_RUN_MODCC_ON_CHANGES)
+if (ARB_AUTO_RUN_MODCC_ON_CHANGES)
add_dependencies(test.exe build_test_mods)
endif()
target_include_directories(test.exe PRIVATE "${mech_proto_dir}/..")
-if(NMC_WITH_CUDA)
+if(ARB_WITH_CUDA)
set(TARGETS ${TARGETS} test_cuda.exe)
cuda_add_executable(test_cuda.exe ${TEST_CUDA_SOURCES})
endif()
foreach(target ${TARGETS})
target_link_libraries(${target} LINK_PUBLIC gtest)
- target_link_libraries(${target} LINK_PUBLIC ${NESTMC_LIBRARIES})
+ target_link_libraries(${target} LINK_PUBLIC ${ARB_LIBRARIES})
target_link_libraries(${target} LINK_PUBLIC ${EXTERNAL_LIBRARIES})
- if(NMC_WITH_MPI)
+ if(ARB_WITH_MPI)
target_link_libraries(${target} LINK_PUBLIC ${MPI_C_LIBRARIES})
set_property(TARGET ${target} APPEND_STRING PROPERTY LINK_FLAGS "${MPI_C_LINK_FLAGS}")
endif()
diff --git a/tests/unit/test_intrin.cpp b/tests/unit/test_intrin.cpp
index e5c83a8206f6fc51f8811b381aed38a5597da467..5cb3310c648c84fc75321292cf8c35f2c5d83289 100644
--- a/tests/unit/test_intrin.cpp
+++ b/tests/unit/test_intrin.cpp
@@ -32,7 +32,7 @@ TEST(intrin, exp256) {
for (size_t i = 0; i < 16/simd_len; ++i) {
- __m256d vv = nmc_mm256_exp_pd(vvalues[i]);
+ __m256d vv = arb_mm256_exp_pd(vvalues[i]);
double *intrin = (double *) &vv;
for (size_t j = 0; j < simd_len; ++j) {
double v = values[i*simd_len + j];
@@ -59,7 +59,7 @@ TEST(intrin, frexp256) {
for (size_t i = 0; i < 16/simd_len; ++i) {
__m128i vexp;
- __m256d vbase = nmc_mm256_frexp_pd(vvalues[i], &vexp);
+ __m256d vbase = arb_mm256_frexp_pd(vvalues[i], &vexp);
double *vbase_ = (double *) &vbase;
int *vexp_ = (int *) &vexp;
for (size_t j = 0; j < simd_len; ++j) {
@@ -99,7 +99,7 @@ TEST(intrin, log256) {
for (size_t i = 0; i < 16/simd_len; ++i) {
- __m256d vv = nmc_mm256_log_pd(vvalues[i]);
+ __m256d vv = arb_mm256_log_pd(vvalues[i]);
double *intrin = (double *) &vv;
for (size_t j = 0; j < simd_len; ++j) {
double v = values[i*simd_len + j];
diff --git a/tests/unit/test_range.cpp b/tests/unit/test_range.cpp
index 90b9f9faa67455972774f8dddff3815f0064a0ba..d666e45ffac614779e3a2187b4d8a942a81352ec 100644
--- a/tests/unit/test_range.cpp
+++ b/tests/unit/test_range.cpp
@@ -9,7 +9,7 @@
#include <type_traits>
#include <unordered_map>
-#ifdef NMC_HAVE_TBB
+#ifdef ARB_HAVE_TBB
#include <tbb/tbb_stddef.h>
#endif
@@ -600,7 +600,7 @@ TEST(range, is_sorted_by) {
EXPECT_TRUE(util::is_sorted_by(seq, [](int x) { return x+2; }, std::greater<int>{}));
}
-#ifdef NMC_HAVE_TBB
+#ifdef ARB_HAVE_TBB
TEST(range, tbb_split) {
constexpr std::size_t N = 20;
diff --git a/tests/validation/CMakeLists.txt b/tests/validation/CMakeLists.txt
index 45773d26e33109530aab96c8fd6d91ad568ca7e2..399cfe84478d6ff03a6b5b7982498e2077dd708c 100644
--- a/tests/validation/CMakeLists.txt
+++ b/tests/validation/CMakeLists.txt
@@ -14,21 +14,21 @@ set(VALIDATION_SOURCES
validate.cpp
)
-if(NMC_VALIDATION_DATA_DIR)
+if(ARB_VALIDATION_DATA_DIR)
if ("${CMAKE_VERSION}" MATCHES "^3.[789].")
- message(WARNING "CMake ${CMAKE_VERSION} has broken FindCUDA; omitting NMC_DATADIR define.")
+ message(WARNING "CMake ${CMAKE_VERSION} has broken FindCUDA; omitting ARB_DATADIR define.")
else()
- add_definitions("-DNMC_DATADIR=\"${NMC_VALIDATION_DATA_DIR}\"")
+ add_definitions("-DARB_DATADIR=\"${ARB_VALIDATION_DATA_DIR}\"")
endif()
endif()
add_executable(validate.exe ${VALIDATION_SOURCES})
target_link_libraries(validate.exe LINK_PUBLIC gtest)
-target_link_libraries(validate.exe LINK_PUBLIC ${NESTMC_LIBRARIES})
+target_link_libraries(validate.exe LINK_PUBLIC ${ARB_LIBRARIES})
target_link_libraries(validate.exe LINK_PUBLIC ${EXTERNAL_LIBRARIES})
-if(NMC_WITH_MPI)
+if(ARB_WITH_MPI)
target_link_libraries(validate.exe LINK_PUBLIC ${MPI_C_LIBRARIES})
set_property(TARGET validate.exe APPEND_STRING PROPERTY LINK_FLAGS "${MPI_C_LINK_FLAGS}")
endif()
@@ -39,6 +39,6 @@ set_target_properties(
RUNTIME_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/tests"
)
-if(NMC_BUILD_VALIDATION_DATA)
+if(ARB_BUILD_VALIDATION_DATA)
add_dependencies(validate.exe validation_data)
endif()
diff --git a/tests/validation/validate.cpp b/tests/validation/validate.cpp
index 6431a4977456627ed7860a730c6734a0e155a98d..2a8eec47366d71e6f7c785fb3d229f8d52666656 100644
--- a/tests/validation/validate.cpp
+++ b/tests/validation/validate.cpp
@@ -26,10 +26,10 @@ const char* usage_str =
" -h, --help Display usage information and exit\n"
"\n"
"Validation data is searched by default in the directory specified by\n"
-"NMC_DATADIR at compile time. If NMC_DATADIR does not correspond to a\n"
+"ARB_DATADIR at compile time. If ARB_DATADIR does not correspond to a\n"
"directory, the tests will try the paths './validation/data' and\n"
"'../validation/data'. This default path can be overridden with the\n"
-"NMC_DATADIR environment variable, or with the -p command-line option.\n";
+"ARB_DATADIR environment variable, or with the -p command-line option.\n";
int main(int argc, char **argv) {
using to::parse_opt;
diff --git a/tests/validation/validate_ball_and_stick.cpp b/tests/validation/validate_ball_and_stick.cpp
index 62b5e2b557a08c52f90df538906c91e22e65b898..b5fd22902d8a4a4ccdf00c828af912ff369fae69 100644
--- a/tests/validation/validate_ball_and_stick.cpp
+++ b/tests/validation/validate_ball_and_stick.cpp
@@ -48,7 +48,7 @@ void run_ncomp_convergence_test(
{"name", "membrane voltage"},
{"model", model_name},
{"dt", dt},
- {"sim", "nestmc"},
+ {"sim", "arbor"},
{"units", "mV"},
{"backend_kind", to_string(backend)}
};
diff --git a/tests/validation/validate_kinetic.cpp b/tests/validation/validate_kinetic.cpp
index 749e0b3b0c5869575f7d5b7323f67fbc232ecae0..986d8cbc54c3c8e124b2de5ede5e1ae9a1305a7c 100644
--- a/tests/validation/validate_kinetic.cpp
+++ b/tests/validation/validate_kinetic.cpp
@@ -35,7 +35,7 @@ void run_kinetic_dt(
rec.add_probe(0, 0, cell_probe_address{{0, 0.5}, cell_probe_address::membrane_voltage});
probe_label plabels[1] = {"soma.mid", {0u, 0u}};
- meta["sim"] = "nestmc";
+ meta["sim"] = "arbor";
meta["backend_kind"] = to_string(backend);
convergence_test_runner<float> runner("dt", plabels, meta);
diff --git a/tests/validation/validate_soma.cpp b/tests/validation/validate_soma.cpp
index 8b0f35ba54800fbb1f11933119496c6cf8723984..86f8dfd1388dfb8404bb5f69d3177e177f9e06aa 100644
--- a/tests/validation/validate_soma.cpp
+++ b/tests/validation/validate_soma.cpp
@@ -38,7 +38,7 @@ void validate_soma(backend_kind backend) {
nlohmann::json meta = {
{"name", "membrane voltage"},
{"model", "soma"},
- {"sim", "nestmc"},
+ {"sim", "arbor"},
{"units", "mV"},
{"backend_kind", to_string(backend)}
};
diff --git a/tests/validation/validate_synapses.cpp b/tests/validation/validate_synapses.cpp
index f4b3b7b03672e82b5c69815f1c9aeafe2c71785d..03c2ba703e2378dc9e40ec321e1011cef9a33459 100644
--- a/tests/validation/validate_synapses.cpp
+++ b/tests/validation/validate_synapses.cpp
@@ -32,7 +32,7 @@ void run_synapse_test(
nlohmann::json meta = {
{"name", "membrane voltage"},
{"model", syn_type},
- {"sim", "nestmc"},
+ {"sim", "arbor"},
{"units", "mV"},
{"backend_kind", to_string(backend)}
};
diff --git a/tests/validation/validation_data.cpp b/tests/validation/validation_data.cpp
index 2ce470f3063a79ed0599c0754cfcc3d539a9d66e..8e9250fe3a81e77b3361d0e1c8a63c0811ec708b 100644
--- a/tests/validation/validation_data.cpp
+++ b/tests/validation/validation_data.cpp
@@ -15,23 +15,23 @@ namespace arb {
trace_io g_trace_io;
-#ifndef NMC_DATADIR
-#define NMC_DATADIR ""
+#ifndef ARB_DATADIR
+#define ARB_DATADIR ""
#endif
util::path trace_io::find_datadir() {
// If environment variable is set, use that in preference.
- if (const char* env_path = std::getenv("NMC_DATADIR")) {
+ if (const char* env_path = std::getenv("ARB_DATADIR")) {
return util::path(env_path);
}
- // Otherwise try compile-time path NMC_DATADIR and hard-coded
+ // Otherwise try compile-time path ARB_DATADIR and hard-coded
// relative paths below in turn, returning the first that
// corresponds to an existing directory.
const char* paths[] = {
- NMC_DATADIR,
+ ARB_DATADIR,
"./validation/data",
"../validation/data"
};
diff --git a/tests/validation/validation_data.hpp b/tests/validation/validation_data.hpp
index c1944eb3ba144a05fc75a4016cc06f667a9a3ece..ed2cb09ff76bd3c5d8829e6b99bd746f8d87f477 100644
--- a/tests/validation/validation_data.hpp
+++ b/tests/validation/validation_data.hpp
@@ -88,9 +88,9 @@ private:
float min_dt_ = 0.001f;
float sample_dt_ = 0.005f;
- // Returns value of NMC_DATADIR environment variable if set,
+ // Returns value of ARB_DATADIR environment variable if set,
// otherwise make a 'best-effort' search for the data directory,
- // starting with NMC_DATADIR preprocessor define if defined and
+ // starting with ARB_DATADIR preprocessor define if defined and
// if the directory exists, or else try './validation/data'
// and '../validation/data'.
static util::path find_datadir();
diff --git a/validation/CMakeLists.txt b/validation/CMakeLists.txt
index b65c4fcf0030099558381c66b6635dc6a3dc74df..252d099a0f70fddffd39bd4e435da0044c2c0a85 100644
--- a/validation/CMakeLists.txt
+++ b/validation/CMakeLists.txt
@@ -28,7 +28,7 @@ endfunction()
include(CMakeParseArguments)
function(add_validation_data)
cmake_parse_arguments(ADD_VALIDATION_DATA "" "OUTPUT" "DEPENDS;COMMAND" ${ARGN})
- set(out "${NMC_VALIDATION_DATA_DIR}/${ADD_VALIDATION_DATA_OUTPUT}")
+ set(out "${ARB_VALIDATION_DATA_DIR}/${ADD_VALIDATION_DATA_OUTPUT}")
string(REGEX REPLACE "([^;]+)" "${CMAKE_CURRENT_SOURCE_DIR}/\\1" deps "${ADD_VALIDATION_DATA_DEPENDS}")
add_custom_command(
OUTPUT "${out}"
@@ -43,7 +43,7 @@ function(add_validation_data)
endfunction()
-if(NMC_BUILD_NRN_VALIDATION_DATA)
+if(ARB_BUILD_NRN_VALIDATION_DATA)
add_subdirectory(ref/neuron)
endif()
diff --git a/validation/ref/numeric/CMakeLists.txt b/validation/ref/numeric/CMakeLists.txt
index 510f58ee6db5d7b77ba01dcecc00c94dbf3817ea..ed5761ec7c08e636588b0e5e60485893d7baea55 100644
--- a/validation/ref/numeric/CMakeLists.txt
+++ b/validation/ref/numeric/CMakeLists.txt
@@ -1,6 +1,6 @@
# note: function add_validation_data defined in validation/CMakeLists.txt
-if(NMC_BUILD_JULIA_VALIDATION_DATA)
+if(ARB_BUILD_JULIA_VALIDATION_DATA)
add_validation_data(
OUTPUT numeric_kin1.json
DEPENDS numeric_kin1.jl
diff --git a/validation/ref/numeric/numeric_rallpack1.jl b/validation/ref/numeric/numeric_rallpack1.jl
index 8b249760ef5d1818bcbdf8327d19632c7775877c..b13a5effe60c3a7d2daa154c3bc1a125c26683ce 100644
--- a/validation/ref/numeric/numeric_rallpack1.jl
+++ b/validation/ref/numeric/numeric_rallpack1.jl
@@ -14,7 +14,7 @@ function run_cable(x_prop, ts)
# Physical properties:
# f is a fudge factor. rM needs to be the same
- # the same as in nestmc, where we cannot yet set
+ # the same as in arbor, where we cannot yet set
# the membrane conductance parameter. Scaling
# other parameters proportionally, however,
# gives the same dynamics.