Merge branch 'release/2.0.0-alpha2'

cp2k · Jun 28, 2019 · 5fbd363 · 5fbd363
2 parents 6de25b4 + fa126ce
commit 5fbd363
Show file tree

Hide file tree

Showing 96 changed files with 2,930 additions and 1,376 deletions.
diff --git a/.ci/daint.cscs.ch/Jenkinsfile b/.ci/daint.cscs.ch/Jenkinsfile
@@ -7,35 +7,46 @@ pipeline {
                 checkout scm
             }
         }
-        stage('build') {
-            steps {
-                run_batch("0:15:00")
-            }
-        }
-        stage('test') {
-            steps {
-                run_batch("1:00:00")
+        stage("build and test") {
+            parallel {
+                stage("GNU") {
+                    stages {
+                        stage('build') {
+                            steps {
+                                run_batch("0:15:00", "gnu", "build")
+                            }
+                        }
+                        stage('test') {
+                            steps {
+                                run_batch("1:00:00", "gnu", "test")
+                            }
+                        }
+                    }
+                }
             }
         }
     }
 }
 
-def run_batch(timelimit) {
+def run_batch(timelimit, environment, task) {
     def (account, basename) = env.JOB_NAME.split('/')
+    def sbatch_script = ".ci/daint.cscs.ch/${environment}.${task}.sh"
+    def sbatch_out = "sbatch.${env.BUILD_TAG}.${environment}.${task}.out"
+
     // avoid using the shell for variable expansion to
     // get the final command displayed in Jenkins
     try {
         sh """
         sbatch --wait \
             --time="${timelimit}" \
             --account="${account}" \
-            --job-name="${basename}" \
-            --output="sbatch.${env.BUILD_TAG}.${env.STAGE_NAME}.out" \
-            .ci/daint.cscs.ch/${env.STAGE_NAME}.sh
+            --job-name="${basename}.${environment}.${task}" \
+            --output="${sbatch_out}" \
+            ${sbatch_script}
         """
     }
     finally {
-        echo readFile("sbatch.${env.BUILD_TAG}.${env.STAGE_NAME}.out")
+        echo readFile("${sbatch_out}")
     }
 }
 

diff --git a/.ci/daint.cscs.ch/build.sh → .ci/daint.cscs.ch/gnu.build.sh b/.ci/daint.cscs.ch/build.sh → .ci/daint.cscs.ch/gnu.build.sh
@@ -16,18 +16,22 @@ set -o pipefail
 module swap PrgEnv-cray PrgEnv-gnu
 module load daint-gpu cudatoolkit CMake/3.12.0
 module unload cray-libsci_acc
+module list
 
 set -o xtrace  # do not set earlier to avoid noise from module
 
 umask 0002  # make sure group members can access the data
 
-mkdir --mode=0775 -p "${SCRATCH}/${BUILD_TAG}"
-cd "${SCRATCH}/${BUILD_TAG}"
+mkdir --mode=0775 -p "${SCRATCH}/${BUILD_TAG}.gnu"
+cd "${SCRATCH}/${BUILD_TAG}.gnu"
 
 cmake \
+    -DCMAKE_SYSTEM_NAME=CrayLinuxEnvironment \
     -DUSE_CUDA=ON \
     -DUSE_CUBLAS=ON \
     -DWITH_GPU=P100 \
+    -DBLAS_FOUND=ON -DBLAS_LIBRARIES="-lsci_gnu_mpi_mp" \
+    -DLAPACK_FOUND=ON -DLAPACK_LIBRARIES="-lsci_gnu_mpi_mp" \
     -DMPIEXEC_EXECUTABLE="$(command -v srun)" \
     -DTEST_MPI_RANKS=${SLURM_NTASKS} \
     -DTEST_OMP_THREADS=${SLURM_CPUS_PER_TASK} \

diff --git a/.ci/daint.cscs.ch/test.sh → .ci/daint.cscs.ch/gnu.test.sh b/.ci/daint.cscs.ch/test.sh → .ci/daint.cscs.ch/gnu.test.sh
@@ -16,13 +16,14 @@ set -o pipefail
 module swap PrgEnv-cray PrgEnv-gnu
 module load daint-gpu cudatoolkit CMake/3.12.0
 module unload cray-libsci_acc
+module list
 
 set -o xtrace  # do not set earlier to avoid noise from module
 
 umask 0002  # make sure group members can access the data
 
-mkdir --mode=0775 -p "${SCRATCH}/${BUILD_TAG}"
-cd "${SCRATCH}/${BUILD_TAG}"
+mkdir --mode=0775 -p "${SCRATCH}/${BUILD_TAG}.gnu"
+cd "${SCRATCH}/${BUILD_TAG}.gnu"
 
 export CRAY_CUDA_MPS=1 # enable the CUDA proxy for MPI+CUDA
 export OMP_PROC_BIND=TRUE # set thread affinity

diff --git a/.travis.yml b/.travis.yml
@@ -19,11 +19,13 @@ addons:
     - gcc
     - cmake
     - openmpi
+    update: true
 
 cache:
   pip: true
   directories:
   - $HOME/deps
+  - $HOME/Library/Caches/Homebrew
 
 env:
   global:

diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -21,7 +21,7 @@ set_property(CACHE USE_SMM PROPERTY STRINGS blas libxsmm)
 option(USE_CUDA "Build with CUDA support" OFF)
 cmake_dependent_option(USE_CUBLAS "Build with CUBLAS support" OFF "USE_CUDA" OFF)
 set(WITH_GPU "P100" CACHE STRING "Set the CUDA GPU architecture if CUDA is enabled (default: P100)")
-set_property(CACHE WITH_GPU PROPERTY STRINGS K20X K40 K80 P100)
+set_property(CACHE WITH_GPU PROPERTY STRINGS K20X K40 K80 P100 V100)
 
 enable_language(Fortran)
 enable_testing() # enables the `make test` target
@@ -37,6 +37,7 @@ if (USE_CUDA)
   set(CUDA_ARCH_NUMBER_K40  35)
   set(CUDA_ARCH_NUMBER_K80  37)
   set(CUDA_ARCH_NUMBER_P100 60)
+  set(CUDA_ARCH_NUMBER_V100 70)
   set(CUDA_ARCH_NUMBER ${CUDA_ARCH_NUMBER_${WITH_GPU}})
 
   # assume that the backend compiler for nvcc understands the -std=c++11
@@ -85,62 +86,8 @@ if (NOT CMAKE_BUILD_TYPE)
     FORCE)
 endif ()
 
-# COMPILER CONFIGURATION:
-
-get_filename_component(Fortran_COMPILER_NAME ${CMAKE_Fortran_COMPILER} NAME)
-
-# if the fortran compiler command is a CRAY wrapper, find out what the underlying compiler is
-if (Fortran_COMPILER_NAME MATCHES "ftn")
-  execute_process(COMMAND ftn -craype-verbose OUTPUT_VARIABLE ftn_verbose_OUT ERROR_QUIET)
-  separate_arguments(ftn_verbose_OUT NATIVE_COMMAND "${ftn_verbose_OUT}")
-  list(GET ftn_verbose_OUT 0 Fortran_COMPILER_NAME)
-endif ()
-
-# Depending on the fortran compiler, set the appropriate compiler flags
-if (Fortran_COMPILER_NAME MATCHES "^gfortran.*")
-  set(CMAKE_CXX_FLAGS              "-std=c++11")
-  set(CMAKE_CXX_FLAGS_RELEASE      "-O3 -funroll-loops")
-  set(CMAKE_CXX_FLAGS_COVERAGE     "-O0 -fprofile-arcs -ftest-coverage")
-  set(CMAKE_CXX_FLAGS_DEBUG        "-O0 -ggdb")
-  set(CMAKE_Fortran_FLAGS          "-ffree-form -ffree-line-length-none -std=f2003")
-  set(CMAKE_Fortran_FLAGS_RELEASE  "-O3 -funroll-loops")
-  set(CMAKE_Fortran_FLAGS_DEBUG    "-O0 -ggdb")
-  set(CMAKE_Fortran_FLAGS_COVERAGE "-O0 -fprofile-arcs -ftest-coverage")
-  set(F2008_COMPILER_FLAGS         "-std=f2008ts")
-
-  # on Apple we may compile with GCC Fortran, but build/link C/C++ code with Apple's clang, requiring special treatment
-  if (APPLE AND CMAKE_CXX_COMPILER_ID MATCHES "AppleClang")
-    if (CMAKE_BUILD_TYPE MATCHES Coverage)
-      # when building with coverage suppport, shared libs/executables must be explicitly linked to avoid undefined symbols
-      SET(CMAKE_EXE_LINKER_FLAGS  "${CMAKE_EXE_LINKER_FLAGS} -lgcov --coverage")
-    endif ()
-  endif ()
-elseif (Fortran_COMPILER_NAME MATCHES "^ifort.*")
-  set(CMAKE_CXX_FLAGS_RELEASE      "-O3")
-  set(CMAKE_CXX_FLAGS_DEBUG        "-O0 -debug")
-  set(CMAKE_Fortran_FLAGS          "-free -stand f03 -fpp")
-  # Disable the line-length-extension warning #5268
-  set(CMAKE_Fortran_FLAGS_RELEASE  "-O3 -diag-disable=5268")
-  set(CMAKE_Fortran_FLAGS_DEBUG    "-O0 -debug")
-  set(F2008_COMPILER_FLAGS         "-stand f08")
-elseif (Fortran_COMPILER_NAME MATCHES "^pgf.*")
-  set(CMAKE_CXX_FLAGS_RELEASE      "-fast")
-  set(CMAKE_CXX_FLAGS_DEBUG        "-g")
-  set(CMAKE_Fortran_FLAGS          "-Mfreeform -Mextend -Mallocatable=03")
-  set(CMAKE_Fortran_FLAGS_RELEASE  "-fast")
-  set(CMAKE_Fortran_FLAGS_DEBUG    "-g")
-  set(F2008_COMPILER_FLAGS         "")
-else ()
-  message(WARNING "\
-Unknown compiler, trying with just '-O2'/'-O0 -g'.\n\
-You will most likely have to pass additonal options for free-form Fortran 2003/2008 for your compiler!\n\
-Please open an issue at https://github.com/cp2k/dbcsr/issues with the reported compiler name and the required flags.")
-  message("-- Fortran_COMPILER_NAME: " ${Fortran_COMPILER_NAME})
-  message("-- CMAKE_Fortran_COMPILER full path: " ${CMAKE_Fortran_COMPILER})
-
-  set(CMAKE_Fortran_FLAGS_RELEASE  "-O2")
-  set(CMAKE_Fortran_FLAGS_DEBUG    "-O0 -g")
-endif ()
+include(CompilerConfiguration)
+include(CheckCompilerSupport)
 
 file(STRINGS VERSION VERSION_INFO)
 foreach(line ${VERSION_INFO})

diff --git a/Makefile b/Makefile
@@ -25,11 +25,6 @@ LIBNAME      := dbcsr
 LIBRARY      := lib$(LIBNAME)
 default_target: $(LIBRARY)
 
-# Check if FYPP is available  ===============================================
-ifeq (, $(shell which $(FYPPEXE) 2>/dev/null ))
-$(error "No FYPP submodule available, please read README.md on how to properly download DBCSR")
-endif
-
 # Read the configuration ====================================================
 MODDEPS = "lower"
 include $(INCLUDEMAKE)
@@ -50,7 +45,9 @@ else ifeq ($(GPUVER),K80)
  ARCH_NUMBER = 37
 else ifeq ($(GPUVER),P100)
  ARCH_NUMBER = 60
-else ifeq ($(GPUVER),) # Default to the newest GPU
+else ifeq ($(GPUVER),V100)
+ ARCH_NUMBER = 70
+else ifeq ($(GPUVER),) # Default to the P100
  ARCH_NUMBER = 60
 else
  $(error GPUVER not recognized)
@@ -233,6 +230,12 @@ OTHER_HELP += "toolflags : Print flags used with build tools"
 
 else
 # stage 2: Include $(OBJDIR)/all.dep, expand target all, and get list of dependencies.
+
+# Check if FYPP is available  ===============================================
+ifeq (, $(shell which $(FYPPEXE) 2>/dev/null ))
+$(error "No FYPP submodule available, please read README.md on how to properly download DBCSR")
+endif
+
 all: $(foreach e, $(BIN_NAMES), $(e))
 
 ifeq ($(BIN_NAME),)
@@ -494,7 +497,7 @@ FYPPFLAGS ?= -n
 	$(CXX) -c $(CXXFLAGS) $<
 
 libcusmm.o: libcusmm.cpp parameters.h cusmm_kernels.h
-	$(CXX) -c $(CXXFLAGS) $(CXXOMPFLAGS) -DARCH_NUMBER=$(ARCH_NUMBER) $<
+	$(CXX) -c $(CXXFLAGS) -DARCH_NUMBER=$(ARCH_NUMBER) $<
 
 %.o: %.cu
 	$(NVCC) -c $(NVFLAGS) -I'$(SRCDIR)' $<
@@ -505,7 +508,7 @@ libcusmm_benchmark.o: libcusmm_benchmark.cu parameters.h
 $(LIBDIR)/%:
 ifneq ($(LD_SHARED),)
 	@echo "Creating shared library $@"
-	@$(LD_SHARED) -o $(@:.a=.so) $^
+	@$(LD_SHARED) $(LDFLAGS) -o $(@:.a=.so) $^ $(LIBS)
 else
 	@echo "Updating archive $@"
 	@$(AR) $@ $?

diff --git a/Makefile.inc b/Makefile.inc
@@ -17,7 +17,7 @@
 # a) set the NVCC variable, e.g. NVCC = nvcc
 # b) specify -D__DBCSR_ACC in FCFLAGS variable
 # c) set the GPUVER variable, e.g. GPUVER = P100 for P100 card
-#    (possible values are K20X, K40, K80, P100)
+#    (possible values are K20X, K40, K80, P100, V100)
 # d) set the NVFLAGS variable,
 #    e.g. NVFLAGS = -O3 -g -w --std=c++11
 #    in the Makefile, the -arch will be appended with the correct compute version
@@ -137,7 +137,7 @@ endif
 
 ifneq (0,$(GNU)) # DEFAULT, just edit...
   CXXFLAGS += -std=c++11
-  FCFLAGS  += -std=f2003 -ffree-form -fimplicit-none -ffree-line-length-512
+  FCFLAGS  += -std=f2008ts -ffree-form -fimplicit-none -ffree-line-length-512
   OPTFLAGS += -O3 -g -fno-omit-frame-pointer -funroll-loops
 else
  # Intel compiler flags
@@ -147,7 +147,7 @@ else
 endif
 
 ifneq (0,$(OMP)) # using OpenMP
-  CXXOMPFLAGS = -fopenmp
+  CXXFLAGS += -fopenmp
   FCFLAGS  += -fopenmp
 endif
 
@@ -157,14 +157,6 @@ CXXFLAGS    += $(OPTFLAGS)
 FCFLAGS     += $(OPTFLAGS)
 LDFLAGS     += $(FCFLAGS)
 
-#######################################################################################
-#
-# C interface requires new F2008ts standard
-
-ifeq ($(CINT),1)
-FCFLAGS    := $(subst -std=f2003,-std=f2008ts,$(FCFLAGS))
-endif
-
 #######################################################################################
 #
 # Macro for MPI parallelization.
@@ -230,7 +222,5 @@ endif
 #######################################################################################
 #
 # Enable for MacOS compilation (-D__ACCELERATE is for Apple Accelerate library)
-
 #FCFLAGS += -D__NO_STATM_ACCESS -D__ACCELERATE
-
 #######################################################################################
diff --git a/README.md b/README.md
@@ -1,19 +1,23 @@
 # DBCSR: Distributed Block Compressed Sparse Row matrix library 
 
 [![Build Status](https://travis-ci.org/cp2k/dbcsr.svg?branch=develop)](https://travis-ci.org/cp2k/dbcsr) [![codecov](https://codecov.io/gh/cp2k/dbcsr/branch/develop/graph/badge.svg)](https://codecov.io/gh/cp2k/dbcsr)
+[![Licence](https://img.shields.io/badge/license-GPL%20v2.0-blue.svg)](./LICENSE)
+[![GitHub Releases](https://img.shields.io/github/release-pre/cp2k/dbcsr.svg)](https://github.com/cp2k/dbcsr/releases)
 
 DBCSR is a library designed to efficiently perform sparse matrix matrix multiplication, among other operations.
 It is MPI and OpenMP parallel and can exploit GPUs via CUDA.
 
-![DBCSR logo](tools/logo/logo.png)
+<p align="center">
+<img src="docs/logo/logo.png" width="500">
+</p>
 
 ## Prerequisites
 
 You absolutely need:
 
 * GNU make
-* a Fortran compiler which supports at least Fortran 2003 (respectively 2008+TS when using the C-bindings)
-* a LAPACK implementation (reference, OpenBLAS-bundled and MKL have been tested)
+* a Fortran compiler which supports at least Fortran 2008 (including the TS 29113 when using the C-bindings)
+* a LAPACK implementation (reference, OpenBLAS-bundled and MKL have been tested. Note: DBCSR linked to OpenBLAS 0.3.6 gives wrong results on Power9 architectures.)
 * a BLAS implementation (reference, OpenBLAS-bundled and MKL have been tested)
 * a Python version installed (2.7 or 3.6+ have been tested) with Numpy
 
@@ -71,10 +75,19 @@ The configuration flags are (default first):
     -DUSE_OPENMP=<ON|OFF>
     -DUSE_SMM=<blas|libxsmm>
     -DUSE_CUDA=<OFF|ON>
+    -DUSE_CUBLAS=<OFF|ON>
     -DWITH_C_API=<ON|OFF>
     -DWITH_EXAMPLES=<ON|OFF>
-    -DWITH_GPU=<P100|K20X|K40|K80>
+    -DWITH_GPU=<P100|K20X|K40|K80|V100>
     -DTEST_MPI_RANKS=<auto,N>
+    -DTEST_OMP_THREADS=<2,N>
+    -DCMAKE_BUILD_TYPE=<Release|Debug|Coverage>
+
+
+Building on Cray requires an additional flag to make sure CMake is able to properly
+detect the compiler behind the compiler wrappers:
+
+    -DCMAKE_SYSTEM_NAME=CrayLinuxEnvironment
 
 ## Contributing to DBCSR
 

diff --git a/VERSION b/VERSION
@@ -1,6 +1,6 @@
 MAJOR = 2
 MINOR = 0
-PATCH = 0-alpha1
+PATCH = 0-alpha2
 # A specific DATE (YYYY-MM-DD) fixes an official release, otherwise
 # it is considered Development version.
-DATE  = 2019-04-16
+DATE  = 2019-06-28
diff --git a/cmake/CheckCompilerSupport.cmake b/cmake/CheckCompilerSupport.cmake
@@ -0,0 +1,20 @@
+
+include(CheckFortranSourceCompiles)
+
+set(CHECK_PROGRAMS
+  f2008-norm2.f90
+  f2008-block_construct.f90
+  f2008-contiguous.f90
+  )
+
+foreach (prog ${CHECK_PROGRAMS})
+  get_filename_component(prog_ext ${prog} EXT)  # get the src extension to pass along
+  get_filename_component(prog_name ${prog} NAME_WE)
+
+  file(READ "${CMAKE_CURRENT_LIST_DIR}/compiler-tests/${prog}" prog_src)
+  check_fortran_source_compiles("${prog_src}" "${prog_name}" SRC_EXT "${prog_ext}")
+
+  if (NOT ${prog_name})
+    message(FATAL_ERROR "Your compiler does not support all required F2008 features")
+  endif ()
+endforeach ()