1# Copyright (c) 2017-2018, Lawrence Livermore National Security, LLC. 2# Produced at the Lawrence Livermore National Laboratory. LLNL-CODE-734707. 3# All Rights reserved. See files LICENSE and NOTICE for details. 4# 5# This file is part of CEED, a collection of benchmarks, miniapps, software 6# libraries and APIs for efficient high-order finite element and spectral 7# element discretizations for exascale applications. For more information and 8# source code availability see http://github.com/ceed. 9# 10# The CEED research is supported by the Exascale Computing Project 17-SC-20-SC, 11# a collaborative effort of two U.S. Department of Energy organizations (Office 12# of Science and the National Nuclear Security Administration) responsible for 13# the planning and preparation of a capable exascale ecosystem, including 14# software, applications, hardware, advanced system engineering and early 15# testbed platforms, in support of the nation's exascale computing imperative. 16 17ifeq (,$(filter-out undefined default,$(origin CC))) 18 CC = gcc 19endif 20ifeq (,$(filter-out undefined default,$(origin FC))) 21 FC = gfortran 22endif 23NVCC = $(CUDA_DIR)/bin/nvcc 24 25# ASAN must be left empty if you don't want to use it 26ASAN ?= 27NDEBUG ?= 1 28 29LDFLAGS ?= 30UNDERSCORE ?= 1 31 32# OCCA_DIR env variable should point to OCCA master (github.com/libocca/occa) 33OCCA_DIR ?= ../occa 34 35# env variable MAGMA_DIR can be used too 36MAGMA_DIR ?= ../magma 37# If CUDA_DIR is not set, check for nvcc, or resort to /usr/local/cuda 38CUDA_DIR ?= $(or $(patsubst %/,%,$(dir $(patsubst %/,%,$(dir \ 39 $(shell which nvcc 2> /dev/null))))),/usr/local/cuda) 40 41# Warning: SANTIZ options still don't run with /gpu/occa 42# export LSAN_OPTIONS=suppressions=.asanignore 43AFLAGS = -fsanitize=address #-fsanitize=undefined -fno-omit-frame-pointer 44 45CFLAGS = -std=c99 -Wall -Wextra -Wno-unused-parameter -fPIC -MMD -MP 46FFLAGS = -cpp -Wall -Wextra -Wno-unused-parameter -Wno-unused-dummy-argument -fPIC -MMD -MP 47# If using the IBM XL Fortran (xlf) replace FFLAGS appropriately: 48ifneq ($(filter %xlf %xlf_r,$(FC)),) 49 FFLAGS = -qpreprocess -qextname -qpic -MMD 50endif 51 52CFLAGS += $(if $(NDEBUG),-O2 -DNDEBUG=1,-g) 53 54ifeq ($(UNDERSCORE), 1) 55 CFLAGS += -DUNDERSCORE 56endif 57 58FFLAGS += $(if $(NDEBUG),-O2 -DNDEBUG=1,-g) 59 60CFLAGS += $(if $(ASAN),$(AFLAGS)) 61FFLAGS += $(if $(ASAN),$(AFLAGS)) 62LDFLAGS += $(if $(ASAN),$(AFLAGS)) 63CPPFLAGS = -I./include 64LDLIBS = -lm 65OBJDIR := build 66LIBDIR := lib 67 68# Installation variables 69prefix ?= /usr/local 70bindir = $(prefix)/bin 71libdir = $(prefix)/lib 72okldir = $(prefix)/lib/okl 73includedir = $(prefix)/include 74pkgconfigdir = $(libdir)/pkgconfig 75INSTALL = install 76INSTALL_PROGRAM = $(INSTALL) 77INSTALL_DATA = $(INSTALL) -m644 78 79# Get number of processors of the machine 80NPROCS := $(shell getconf _NPROCESSORS_ONLN) 81# prepare make options to run in parallel 82MFLAGS := -j $(NPROCS) --warn-undefined-variables \ 83 --no-print-directory --no-keep-going 84 85PROVE ?= prove 86PROVE_OPTS ?= -j $(NPROCS) 87DARWIN := $(filter Darwin,$(shell uname -s)) 88SO_EXT := $(if $(DARWIN),dylib,so) 89 90ceed.pc := $(LIBDIR)/pkgconfig/ceed.pc 91libceed := $(LIBDIR)/libceed.$(SO_EXT) 92libceed.c := $(wildcard ceed*.c) 93BACKENDS := /cpu/self/ref /cpu/self/tmpl /cpu/self/opt 94 95# Tests 96tests.c := $(sort $(wildcard tests/t[0-9][0-9]-*.c)) 97tests.f := $(sort $(wildcard tests/t[0-9][0-9]-*.f)) 98tests := $(tests.c:tests/%.c=$(OBJDIR)/%) 99ctests := $(tests) 100tests += $(tests.f:tests/%.f=$(OBJDIR)/%) 101#examples 102examples.c := $(sort $(wildcard examples/ceed/*.c)) 103examples.f := $(sort $(wildcard examples/ceed/*.f)) 104examples := $(examples.c:examples/ceed/%.c=$(OBJDIR)/%) 105examples += $(examples.f:examples/ceed/%.f=$(OBJDIR)/%) 106#mfemexamples 107mfemexamples.cpp := $(sort $(wildcard examples/mfem/*.cpp)) 108mfemexamples := $(mfemexamples.cpp:examples/mfem/%.cpp=$(OBJDIR)/mfem-%) 109petscexamples.c := $(sort $(wildcard examples/petsc/*.c)) 110petscexamples := $(petscexamples.c:examples/petsc/%.c=$(OBJDIR)/petsc-%) 111 112# backends/[ref & occa & magma] 113ref.c := $(sort $(wildcard backends/ref/*.c)) 114template.c := $(sort $(wildcard backends/template/*.c)) 115optimized.c:= $(sort $(wildcard backends/optimized/*.c)) 116occa.c := $(sort $(wildcard backends/occa/*.c)) 117magma_preprocessor := python backends/magma/gccm.py 118magma_pre_src := $(filter-out %_tmp.c, $(wildcard backends/magma/ceed-*.c)) 119magma_dsrc := $(wildcard backends/magma/magma_d*.c) 120magma_tmp.c := $(magma_pre_src:%.c=%_tmp.c) 121magma_tmp.cu := $(magma_pre_src:%.c=%_cuda.cu) 122magma_allsrc.c := $(magma_dsrc) $(magma_tmp.c) 123magma_allsrc.cu:= $(magma_tmp.cu) 124 125# Output using the 216-color rules mode 126rule_file = $(notdir $(1)) 127rule_path = $(patsubst %/,%,$(dir $(1))) 128last_path = $(notdir $(patsubst %/,%,$(dir $(1)))) 129ansicolor = $(shell echo $(call last_path,$(1)) | cksum | cut -b1-2 | xargs -IS expr 2 \* S + 17) 130emacs_out = @printf " %10s %s/%s\n" $(1) $(call rule_path,$(2)) $(call rule_file,$(2)) 131color_out = @if [ -t 1 ]; then \ 132 printf " %10s \033[38;5;%d;1m%s\033[m/%s\n" \ 133 $(1) $(call ansicolor,$(2)) \ 134 $(call rule_path,$(2)) $(call rule_file,$(2)); else \ 135 printf " %10s %s\n" $(1) $(2); fi 136# if TERM=dumb, use it, otherwise switch to the term one 137output = $(if $(TERM:dumb=),$(call color_out,$1,$2),$(call emacs_out,$1,$2)) 138 139# if V is set to non-nil, turn the verbose mode 140quiet = $(if $(V),$($(1)),$(call output,$1,$@);$($(1))) 141 142.SUFFIXES: 143.SUFFIXES: .c .o .d 144.SECONDEXPANSION: # to expand $$(@D)/.DIR 145 146.SECONDARY: $(magma_tmp.c) $(magma_tmp.cu) 147 148%/.DIR : 149 @mkdir -p $(@D) 150 @touch $@ 151 152.PRECIOUS: %/.DIR 153 154this: $(libceed) $(ceed.pc) 155# run 'this' target in parallel 156all:;@$(MAKE) $(MFLAGS) V=$(V) this 157 158$(libceed) : LDFLAGS += $(if $(DARWIN), -install_name @rpath/$(notdir $(libceed))) 159 160libceed.c += $(ref.c) 161libceed.c += $(template.c) 162libceed.c += $(optimized.c) 163ifneq ($(wildcard $(OCCA_DIR)/lib/libocca.*),) 164 $(libceed) : LDFLAGS += -L$(OCCA_DIR)/lib -Wl,-rpath,$(abspath $(OCCA_DIR)/lib) 165 $(libceed) : LDLIBS += -locca 166 libceed.c += $(occa.c) 167 $(occa.c:%.c=$(OBJDIR)/%.o) : CFLAGS += -I$(OCCA_DIR)/include 168 BACKENDS += /cpu/occa /gpu/occa /omp/occa /ocl/occa 169endif 170ifneq ($(wildcard $(MAGMA_DIR)/lib/libmagma.*),) 171 CUDA_LIB_DIR := $(wildcard $(foreach d,lib lib64,$(CUDA_DIR)/$d/libcudart.${SO_EXT})) 172 CUDA_LIB_DIR := $(patsubst %/,%,$(dir $(firstword $(CUDA_LIB_DIR)))) 173 ifneq ($(CUDA_LIB_DIR),) 174 cuda_link = -Wl,-rpath,$(CUDA_LIB_DIR) -L$(CUDA_LIB_DIR) -lcublas -lcusparse -lcudart 175 omp_link = -fopenmp 176 magma_link_static = -L$(MAGMA_DIR)/lib -lmagma $(cuda_link) $(omp_link) 177 magma_link_shared = -L$(MAGMA_DIR)/lib -Wl,-rpath,$(abspath $(MAGMA_DIR)/lib) -lmagma 178 magma_link := $(if $(wildcard $(MAGMA_DIR)/lib/libmagma.${SO_EXT}),$(magma_link_shared),$(magma_link_static)) 179 magma_allsrc.o = $(magma_allsrc.c:%.c=$(OBJDIR)/%.o) $(magma_allsrc.cu:%.cu=$(OBJDIR)/%.o) 180 $(libceed) : LDLIBS += $(magma_link) 181 $(tests) $(examples) : LDLIBS += $(magma_link) 182 $(libceed) : $(magma_allsrc.o) 183 libceed.c += $(magma_allsrc.c) 184 libceed.cu += $(magma_allsrc.cu) 185 $(magma_allsrc.c:%.c=$(OBJDIR)/%.o) : CFLAGS += -DADD_ -I$(MAGMA_DIR)/include -I$(CUDA_DIR)/include 186 $(magma_allsrc.cu:%.cu=$(OBJDIR)/%.o) : NVCCFLAGS += --compiler-options=-fPIC -DADD_ -I$(MAGMA_DIR)/include -I$(MAGMA_DIR)/magmablas -I$(MAGMA_DIR)/control -I$(CUDA_DIR)/include 187 BACKENDS += /gpu/magma 188 endif 189endif 190 191export BACKENDS 192 193# generate magma_tmp.c and magma_cuda.cu from magma.c 194$(magma_tmp.c) $(magma_tmp.cu): $(magma_pre_src) | $$(@D)/.DIR 195 $(magma_preprocessor) $< 196 197$(libceed) : $(libceed.c:%.c=$(OBJDIR)/%.o) $(libceed.cu:%.cu=$(OBJDIR)/%.o) | $$(@D)/.DIR 198 $(call quiet,CC) $(LDFLAGS) -shared -o $@ $^ $(LDLIBS) 199 200$(OBJDIR)/%.o : %.c | $$(@D)/.DIR 201 $(call quiet,CC) $(CPPFLAGS) $(CFLAGS) -c -o $@ $(abspath $<) 202 203$(OBJDIR)/%.o : %.cu | $$(@D)/.DIR 204 $(call quiet,NVCC) $(CPPFLAGS) $(NVCCFLAGS) -c -o $@ $(abspath $<) 205 206$(OBJDIR)/% : tests/%.c | $$(@D)/.DIR 207 $(call quiet,CC) $(CPPFLAGS) $(CFLAGS) $(LDFLAGS) -o $@ $(abspath $<) -lceed $(LDLIBS) 208 209$(OBJDIR)/% : tests/%.f | $$(@D)/.DIR 210 $(call quiet,FC) $(CPPFLAGS) $(FFLAGS) $(LDFLAGS) -o $@ $(abspath $<) -lceed $(LDLIBS) 211 212$(OBJDIR)/% : examples/ceed/%.c | $$(@D)/.DIR 213 $(call quiet,CC) $(CPPFLAGS) $(CFLAGS) $(LDFLAGS) -o $@ $(abspath $<) -lceed $(LDLIBS) 214 215$(OBJDIR)/% : examples/ceed/%.f | $$(@D)/.DIR 216 $(call quiet,FC) $(CPPFLAGS) $(FFLAGS) $(LDFLAGS) -o $@ $(abspath $<) -lceed $(LDLIBS) 217 218$(OBJDIR)/mfem-% : examples/mfem/%.cpp $(libceed) | $$(@D)/.DIR 219 $(MAKE) -C examples/mfem CEED_DIR=`pwd` $* 220 mv examples/mfem/$* $@ 221 222$(OBJDIR)/petsc-% : examples/petsc/%.c $(libceed) | $$(@D)/.DIR 223 $(MAKE) -C examples/petsc CEED_DIR=`pwd` $* 224 mv examples/petsc/$* $@ 225 226$(tests) $(examples) : $(libceed) 227$(tests) $(examples) : LDFLAGS += -Wl,-rpath,$(abspath $(LIBDIR)) -L$(LIBDIR) 228 229run-% : $(OBJDIR)/% 230 @tests/tap.sh $(<:build/%=%) 231# Test core libCEED 232test : $(tests:$(OBJDIR)/%=run-%) $(examples:$(OBJDIR)/%=run-%) 233 234# run test target in parallel 235tst : ;@$(MAKE) $(MFLAGS) V=$(V) test 236# CPU C tests only for backend % 237ctc-% : $(ctests);@$(foreach tst,$(ctests),$(tst) /cpu/$*;) 238 239prove : $(tests) $(examples) 240 $(info Testing backends: $(BACKENDS)) 241 $(PROVE) $(PROVE_OPTS) --exec 'tests/tap.sh' $(tests:$(OBJDIR)/%=%) $(examples:$(OBJDIR)/%=%) 242# run prove target in parallel 243prv : ;@$(MAKE) $(MFLAGS) V=$(V) prove 244 245alltests := $(tests) $(examples) $(mfemexamples) $(if $(PETSC_DIR),$(petscexamples)) 246prove-all : $(alltests) 247 $(info Testing backends: $(BACKENDS)) 248 $(PROVE) $(PROVE_OPTS) --exec 'tests/tap.sh' $(alltests:$(OBJDIR)/%=%) 249 250examples : $(examples) 251 252$(ceed.pc) : pkgconfig-prefix = $(abspath .) 253$(OBJDIR)/ceed.pc : pkgconfig-prefix = $(prefix) 254.INTERMEDIATE : $(OBJDIR)/ceed.pc 255%/ceed.pc : ceed.pc.template | $$(@D)/.DIR 256 @sed "s:%prefix%:$(pkgconfig-prefix):" $< > $@ 257 258# The occa executable is not linked with RPATH by default, so we need it to find its libocca.so 259OCCA := $(if $(DARWIN),DYLD_LIBRARY_PATH,LD_LIBRARY_PATH)=$(OCCA_DIR)/lib $(OCCA_DIR)/bin/occa 260OKL_KERNELS := $(wildcard backends/occa/*.okl) 261 262okl-cache : 263 $(OCCA) cache ceed $(OKL_KERNELS) 264 265okl-clear: 266 $(OCCA) clear -y -l ceed 267 268install : $(libceed) $(OBJDIR)/ceed.pc 269 $(INSTALL) -d "$(DESTDIR)$(includedir)" "$(DESTDIR)$(libdir)" "$(DESTDIR)$(okldir)" "$(DESTDIR)$(pkgconfigdir)" 270 $(INSTALL_DATA) include/ceed.h "$(DESTDIR)$(includedir)/" 271 $(INSTALL_DATA) include/ceedf.h "$(DESTDIR)$(includedir)/" 272 $(INSTALL_DATA) $(libceed) "$(DESTDIR)$(libdir)/" 273 $(INSTALL_DATA) $(OBJDIR)/ceed.pc "$(DESTDIR)$(pkgconfigdir)/" 274 $(INSTALL_DATA) $(OKL_KERNELS) "$(DESTDIR)$(okldir)/" 275 276.PHONY : all cln clean print test tst prove prv examples style install doc okl-cache okl-clear 277 278cln clean : 279 $(RM) *.o *.d $(libceed) 280 $(RM) -r *.dSYM $(OBJDIR) $(LIBDIR)/pkgconfig 281 $(MAKE) -C examples/ceed clean 282 $(MAKE) -C examples/mfem clean 283 $(MAKE) -C examples/petsc clean 284 (cd examples/nek5000 && bash make-nek-examples.sh clean) 285 $(RM) $(magma_tmp.c) $(magma_tmp.cu) backends/magma/*~ backends/magma/*.o 286 287distclean : clean 288 rm -rf doc/html 289 290doc : 291 doxygen Doxyfile 292 293style : 294 astyle --style=google --indent=spaces=2 --max-code-length=80 \ 295 --keep-one-line-statements --keep-one-line-blocks --lineend=linux \ 296 --suffix=none --preserve-date --formatted \ 297 *.[ch] tests/*.[ch] backends/*/*.[ch] examples/*/*.[ch] examples/*/*.[ch]pp 298 299print : 300 @echo $(VAR)=$($(VAR)) 301 302print-% : 303 $(info [ variable name]: $*) 304 $(info [ origin]: $(origin $*)) 305 $(info [ value]: $(value $*)) 306 $(info [expanded value]: $($*)) 307 $(info ) 308 @true 309 310-include $(libceed.c:%.c=build/%.d) $(tests.c:tests/%.c=build/%.d) 311