Merge branch 'v1.1.0-rc.0'

.pre-commit-config.yaml

@@ -1,17 +1,34 @@
+default_language_version:
+    python: python3.6
+
 exclude: '^tools/(prettify/fprettify|build_utils/fypp)'
 fail_fast: false
 repos:
 - repo: https://github.com/pre-commit/mirrors-yapf
   rev: v0.24.0
   hooks:
   - id: yapf
+    language_version: "python3"
 - repo: https://github.com/pre-commit/pre-commit-hooks
   rev: v2.0.0
   hooks:
   - id: flake8
+    language_version: "python2"
+    exclude: '^src/acc/libsmm_acc/libcusmm/(tune|predict)_'
+  - id: flake8
+    language_version: "python3"
   - id: check-ast
     stages: [manual]
     language_version: "python2"
+    exclude: '^src/acc/libsmm_acc/libcusmm/(tune|predict)_'
   - id: check-ast
     stages: [manual]
     language_version: "python3"
+- repo: local
+  hooks:
+  - id: check-header
+    name: check file headers
+    entry: ./.pre-commit/check_header.py --verbose
+    language: script
+    types: [text]
+    exclude: '^tools/'

.pre-commit/check_header.py

@@ -0,0 +1,90 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+####################################################################################################
+# Copyright (C) by the DBCSR developers group - All rights reserved                                #
+# This file is part of the DBCSR library.                                                          #
+#                                                                                                  #
+# For information on the license, see the LICENSE file.                                            #
+# For further information please visit https://dbcsr.cp2k.org                                      #
+# SPDX-License-Identifier: GPL-2.0+                                                                #
+####################################################################################################
+
+import argparse
+import re
+import mmap
+import sys
+from os import path
+from contextlib import contextmanager
+
+TYPES = {
+    'c_cpp': ['.c', 'h', '.cc', '.hh', '.cxx', '.hpp', '.cu'],
+    'python': ['.py'],
+    'fortran': ['.F', '.f', '.f90', '.f03'],
+}
+
+# max number of lines allowed between header and top of file
+ALLOWED_LINES = 5
+
+# some assumed max line length to terminate early for large files
+MAX_LINE_LENGTH = 128
+
+
+@contextmanager
+def mmap_open(name, mode='r'):
+    access = mmap.ACCESS_READ if mode == 'r' else mmap.ACCESS_WRITE
+    with open(name, mode + 'b') as fhandle:
+        fmapped = mmap.mmap(fhandle.fileno(), 0, access=access)
+        yield fmapped
+        fmapped.close()
+
+
+def check_header(header_dir, files, verbose=False):
+    retval = 0
+    header_re = {}
+    header_len = {}
+
+    for headertype in TYPES:
+        with open(path.join(header_dir, headertype), 'rb') as fhandle:
+            header_content = fhandle.read()
+            header_re[headertype] = re.compile(re.escape(header_content))
+            header_len[headertype] = len(header_content)
+
+    ext_map = {e: t for t, exts in TYPES.items() for e in exts}
+
+    for fpath in files:
+        _, fext = path.splitext(fpath)
+
+        if fext not in ext_map:
+            if verbose:
+                print("? {} ... unknown file type, ignoring".format(fpath))
+            continue
+
+        with mmap_open(fpath) as fmapped:
+            header_type = ext_map[fext]
+            match = header_re[header_type].search(fmapped, 0, ALLOWED_LINES * MAX_LINE_LENGTH + header_len[header_type])
+
+            if not match:
+                print("✗ {} ... required header not found".format(fpath))
+                retval = 1
+                continue
+
+            lines_above = fmapped[0:match.start()].splitlines()
+            if len(lines_above) > ALLOWED_LINES:
+                print("✗ {} ... header not within first {} lines".format(fpath, ALLOWED_LINES))
+                retval = 1
+                continue
+
+        if verbose:
+            print("✓ {}".format(fpath))
+
+    sys.exit(retval)
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(description="Check files for header presence")
+    parser.add_argument('files', metavar='FILE', type=str, nargs='+', help="files to check")
+    parser.add_argument('--verbose', '-v', action='store_true', default=False)
+    args = parser.parse_args()
+
+    header_dir = path.join(path.dirname(path.abspath(__file__)), 'headers')
+    check_header(header_dir, args.files, args.verbose)

.pre-commit/headers/c_cpp

@@ -0,0 +1,8 @@
+/*------------------------------------------------------------------------------------------------*
+ * Copyright (C) by the DBCSR developers group - All rights reserved                              *
+ * This file is part of the DBCSR library.                                                        *
+ *                                                                                                *
+ * For information on the license, see the LICENSE file.                                          *
+ * For further information please visit https://dbcsr.cp2k.org                                    *
+ * SPDX-License-Identifier: GPL-2.0+                                                              *
+ *------------------------------------------------------------------------------------------------*/

.pre-commit/headers/fortran

@@ -0,0 +1,8 @@
+!--------------------------------------------------------------------------------------------------!
+! Copyright (C) by the DBCSR developers group - All rights reserved                                !
+! This file is part of the DBCSR library.                                                          !
+!                                                                                                  !
+! For information on the license, see the LICENSE file.                                            !
+! For further information please visit https://dbcsr.cp2k.org                                      !
+! SPDX-License-Identifier: GPL-2.0+                                                                !
+!--------------------------------------------------------------------------------------------------!

.pre-commit/headers/python

@@ -0,0 +1,8 @@
+####################################################################################################
+# Copyright (C) by the DBCSR developers group - All rights reserved                                #
+# This file is part of the DBCSR library.                                                          #
+#                                                                                                  #
+# For information on the license, see the LICENSE file.                                            #
+# For further information please visit https://dbcsr.cp2k.org                                      #
+# SPDX-License-Identifier: GPL-2.0+                                                                #
+####################################################################################################

.travis.yml

@@ -37,6 +37,11 @@ env:
   - USE_MPI=OFF USE_OPENMP=ON USE_SMM=libxsmm
   - USE_MPI=OFF USE_OPENMP=OFF USE_SMM=libxsmm
 
+matrix:
+  include:
+    - os: linux
+      python: 3.6
+
 matrix:
   exclude:
     - os: osx
@@ -87,7 +92,12 @@ jobs:  # the following jobs inherit only the first OS
   - name: "Run pre-commit hook on changed files"
     script:
     - set -o pipefail
-    - git diff --name-only --diff-filter=AM HEAD...$TRAVIS_BRANCH | xargs pre-commit run --files
+      # fetch remotes first to ensure that referenced branches (other than the default) are available
+    - git config remote.origin.fetch "+refs/heads/*:refs/remotes/origin/*"
+    - git fetch
+    # yapf requires python3.6.4+
+    - pyenv global 3.6.7 2.7.15
+    - git diff --name-only --diff-filter=AM HEAD "origin/${TRAVIS_BRANCH}" | xargs pre-commit run --files
   - name: "Run pre-commit hook 'check-ast' on all files"
     script:
     - pre-commit run --hook-stage manual --all-files check-ast
@@ -96,7 +106,9 @@ jobs:  # the following jobs inherit only the first OS
     if: tag IS present
     script:
       # do an in-source config for simplicity since we are not building anything here
-      - cmake .
+      - mkdir -p build
+      - cd build
+      - cmake ..
       - make dist
     deploy:
       provider: releases

CMakeLists.txt

@@ -88,6 +88,14 @@ endif ()
 
 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")

Makefile

@@ -84,6 +84,7 @@ endif
          dirs makedep \
 	 default_target $(LIBRARY) all \
          toolversions \
+         toolflags \
          doxify doxifyclean \
          pretty prettyclean doxygen/clean doxygen \
          install clean realclean help \
@@ -144,7 +145,7 @@ dirs:
 	@mkdir -p $(LIBDIR)
 
 version:
-	@echo "DCBSR Version: "$(MAJOR)"."$(MINOR)"."$(PATCH)" ("$(DATE)")"
+	@echo "DBCSR Version: "$(MAJOR)"."$(MINOR)"."$(PATCH)" ("$(DATE)")"
 OTHER_HELP += "version : Print DBCSR version"
 
 toolversions:
@@ -158,6 +159,11 @@ else
 	$(FC) --version
 endif
 endif
+ifneq ($(CXX),)
+	@echo "========== CXX =========="
+	$(CXX) --version
+	@echo ""
+endif
 ifneq ($(CC),)
 	@echo "=========== CC ==========="
 ifeq (Cray,$(shell $(CC) -V 2>&1 | head -n1 | cut -d' ' -f1))
@@ -168,6 +174,11 @@ else
 	$(CC) --version
 endif
 endif
+ifneq ($(LD),)
+	@echo "========== LD =========="
+	$(LD) --version
+	@echo ""
+endif
 ifneq ($(NVCC),)
 	@echo "========== NVCC =========="
 	$(NVCC) --version
@@ -186,6 +197,40 @@ endif
 
 OTHER_HELP += "toolversions : Print versions of build tools"
 
+toolflags:
+ifneq ($(FCFLAGS),)
+	@echo "========== FCFLAGS =========="
+	@echo $(FCFLAGS)
+	@echo ""
+endif
+ifneq ($(CXXFLAGS),)
+	@echo "========== CXXFLAGS =========="
+	@echo $(CXXFLAGS)
+	@echo ""
+endif
+ifneq ($(CFLAGS),)
+	@echo "========== CFLAGS =========="
+	@echo $(CFLAGS)
+	@echo ""
+endif
+ifneq ($(LDFLAGS),)
+	@echo "========== LDFLAGS =========="
+	@echo $(LDFLAGS)
+	@echo ""
+endif
+ifneq ($(NVFLAGS),)
+	@echo "========== NVFLAGS =========="
+	@echo $(NVFLAGS)
+	@echo ""
+endif
+ifneq ($(GPUVER),)
+	@echo "========== GPUVER =========="
+	@echo $(GPUVER)
+	@echo ""
+endif
+
+OTHER_HELP += "toolflags : Print flags used with build tools"
+
 else
 # stage 2: Include $(OBJDIR)/all.dep, expand target all, and get list of dependencies.
 all: $(foreach e, $(BIN_NAMES), $(e))
@@ -222,6 +267,14 @@ help:
 	@echo "================= Other Targets ================="
 	@printf "%s\n" $(OTHER_HELP) | awk -F ':' '{printf "%-28s%s\n", $$1, $$2}'
 	@echo "help                         Print this help text"
+	@echo "================= Variables ====================="
+	@echo "For convenience, some variables can be set during compilation,"
+	@echo "e.g. make VARIABLE=value (multiple variables are possible):"
+	@echo "MPI=0    : disable MPI compilation"
+	@echo "GNU=0    : disable GNU compiler compilation and enable Intel compiler compilation"
+	@echo "CHECKS=1 : enable GNU compiler checks and DBCSR asserts"
+	@echo "CINT=1   : generate the C interface"
+	@echo "GPU=1    : enable GPU support"
 
 ifeq ($(INCLUDE_DEPS),)
 install: $(LIBRARY)
@@ -239,12 +292,12 @@ install: $(LIBRARY)
 else
 install:
 	@printf "  ... modules ..."
-	@if [[ -n "$(wildcard $(addprefix $(OBJDIR)/, $(PUBLICFILES:.F=.mod)))" ]] ; then \
+	@if [ -n "$(wildcard $(addprefix $(OBJDIR)/, $(PUBLICFILES:.F=.mod)))" ] ; then \
 		cp $(addprefix $(OBJDIR)/, $(PUBLICFILES:.F=.mod)) $(PREFIX)/include ; \
 		echo " done." ; \
 	else echo " no modules were installed!" ; fi
 	@printf "  ... headers ..."
-	@if [[ -n "$(PUBLICHEADERS)" ]] ; then \
+	@if [ -n "$(PUBLICHEADERS)" ] ; then \
 		cp $(PUBLICHEADERS) $(PREFIX)/include ; \
 		echo " done." ; \
 	else echo " no headers were installed!" ; fi
@@ -266,6 +319,8 @@ test:
 OTHER_HELP += "test    : Run the unittests available in tests/"
 
 clean:
+	rm -f $(TESTSDIR)/libcusmm_libcusmm_unittest_multiply.cu
+	rm -f $(TESTSDIR)/libcusmm_timer_multiply.cu
 	rm -rf $(OBJDIR)
 	rm -f $(LIBCUSMM_ABS_DIR)/parameters.h $(LIBCUSMM_ABS_DIR)/cusmm_kernels.h $(LIBCUSMM_ABS_DIR)/*.so
 OTHER_HELP += "clean : Remove intermediate object and mod files, but not the libraries and executables"
@@ -439,7 +494,7 @@ FYPPFLAGS ?= -n
 	$(CXX) -c $(CXXFLAGS) $<
 
 libcusmm.o: libcusmm.cpp parameters.h cusmm_kernels.h
-	$(CXX) -c $(CXXFLAGS) -DARCH_NUMBER=$(ARCH_NUMBER) $<
+	$(CXX) -c $(CXXFLAGS) $(CXXOMPFLAGS) -DARCH_NUMBER=$(ARCH_NUMBER) $<
 
 %.o: %.cu
 	$(NVCC) -c $(NVFLAGS) -I'$(SRCDIR)' $<

Makefile.inc

@@ -34,11 +34,12 @@
 #
 # For convenience, some variables can be set during compilation,
 # e.g. make `VARIABLE=value` (multiple variables are possible):
-# 1) MPI=0    : disable MPI compilation
-# 2) GNU=0    : disable GNU compiler compilation and enable Intel compiler compilation
-# 3) CHECKS=1 : enable GNU compiler checks and DBCSR asserts
-# 4) CINT=1   : generate the C interface
-# 5) GPU=1    : enable GPU support
+# 1) OMP=0    : disable OpenMP compilation
+# 2) MPI=0    : disable MPI compilation
+# 3) GNU=0    : disable GNU compiler compilation and enable Intel compiler compilation
+# 4) CHECKS=1 : enable GNU compiler checks and DBCSR asserts
+# 5) CINT=1   : generate the C interface
+# 6) GPU=1    : enable GPU support
 #
 
 #######################################################################################
@@ -136,15 +137,20 @@ endif
 
 ifneq (0,$(GNU)) # DEFAULT, just edit...
   CXXFLAGS += -std=c++11
-  FCFLAGS  += -fopenmp -std=f2003 -ffree-form -fimplicit-none -ffree-line-length-512
+  FCFLAGS  += -std=f2003 -ffree-form -fimplicit-none -ffree-line-length-512
   OPTFLAGS += -O3 -g -fno-omit-frame-pointer -funroll-loops
 else
  # Intel compiler flags
   CXXFLAGS += -std=c++11
-  FCFLAGS  += -fopenmp -free
+  FCFLAGS  += -free
   OPTFLAGS += -O2 -g
 endif
 
+ifneq (0,$(OMP)) # using OpenMP
+  CXXOMPFLAGS = -fopenmp
+  FCFLAGS  += -fopenmp
+endif
+
 OPTFLAGS    +=
 CFLAGS      += $(OPTFLAGS)
 CXXFLAGS    += $(OPTFLAGS)

README.md

@@ -15,19 +15,19 @@ You absolutely need:
 * 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 BLAS implementation (reference, OpenBLAS-bundled and MKL have been tested)
-* a Python version installed (2.7 or 3.6+ have been tested)
+* a Python version installed (2.7 or 3.6+ have been tested) with Numpy
 
 Optionally:
 
 * [libxsmm](https://github.com/hfp/libxsmm) (1.8.2+ with make-only, 1.10+ with cmake) for Small Matrix Multiplication acceleration
 * [CMake](https://cmake.org/) (3.10+)
 
-To build with CUDA support you further need:
+To build [libcusmm](src/acc/libsmm_acc/libcusmm) (DBCSR's CUDA backend), you further need:
 
 * CUDA Toolkit
 * a C++ compiler which supports at least C++11 standard
 
-We test against GNU and Intel compilers on Linux systems.
+We test against GNU and Intel compilers on Linux systems, GNU compiler on MacOS systems.
 
 ## Getting started