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 = $(CUDA_DIR)/bin/nvcc 24 25# ASAN must be left empty if you don't want to use it 26ASAN ?= 27 28LDFLAGS ?= 29UNDERSCORE ?= 1 30 31# MFEM_DIR env variable should point to sibling directory 32ifneq ($(wildcard ../mfem/.*),) 33 MFEM_DIR?=../mfem 34endif 35 36# OCCA_DIR env variable should point to OCCA master (github.com/libocca/occa) 37OCCA_DIR ?= ../occa 38 39# env variable MAGMA_DIR can be used too 40MAGMA_DIR ?= ../magma 41# If CUDA_DIR is not set, check for nvcc, or resort to /usr/local/cuda 42CUDA_DIR ?= $(or $(patsubst %/,%,$(dir $(patsubst %/,%,$(dir \ 43 $(shell which nvcc 2> /dev/null))))),/usr/local/cuda) 44 45# Warning: SANTIZ options still don't run with /gpu/occa 46# export LSAN_OPTIONS=suppressions=.asanignore 47AFLAGS = -fsanitize=address #-fsanitize=undefined -fno-omit-frame-pointer 48 49OPT = -O -g 50CFLAGS = -std=c99 $(OPT) -Wall -Wextra -Wno-unused-parameter -fPIC -MMD -MP 51NVCCFLAGS = $(OPT) 52# If using the IBM XL Fortran (xlf) replace FFLAGS appropriately: 53ifneq ($(filter %xlf %xlf_r,$(FC)),) 54 FFLAGS = $(OPT) -qpreprocess -qextname -qpic -MMD 55else # gfortran/Intel-style options 56 FFLAGS = -cpp $(OPT) -Wall -Wextra -Wno-unused-parameter -Wno-unused-dummy-argument -fPIC -MMD -MP 57endif 58 59ifeq ($(UNDERSCORE), 1) 60 CFLAGS += -DUNDERSCORE 61endif 62 63ifeq ($(COVERAGE), 1) 64 CFLAGS += --coverage 65 LDFLAGS += --coverage 66endif 67 68CFLAGS += $(if $(ASAN),$(AFLAGS)) 69FFLAGS += $(if $(ASAN),$(AFLAGS)) 70LDFLAGS += $(if $(ASAN),$(AFLAGS)) 71CPPFLAGS = -I./include 72LDLIBS = -lm 73OBJDIR := build 74LIBDIR := lib 75 76# Installation variables 77prefix ?= /usr/local 78bindir = $(prefix)/bin 79libdir = $(prefix)/lib 80okldir = $(prefix)/lib/okl 81includedir = $(prefix)/include 82pkgconfigdir = $(libdir)/pkgconfig 83INSTALL = install 84INSTALL_PROGRAM = $(INSTALL) 85INSTALL_DATA = $(INSTALL) -m644 86 87# Get number of processors of the machine 88NPROCS := $(shell getconf _NPROCESSORS_ONLN) 89# prepare make options to run in parallel 90MFLAGS := -j $(NPROCS) --warn-undefined-variables \ 91 --no-print-directory --no-keep-going 92 93PROVE ?= prove 94PROVE_OPTS ?= -j $(NPROCS) 95DARWIN := $(filter Darwin,$(shell uname -s)) 96SO_EXT := $(if $(DARWIN),dylib,so) 97 98ceed.pc := $(LIBDIR)/pkgconfig/ceed.pc 99libceed := $(LIBDIR)/libceed.$(SO_EXT) 100libceed.c := $(wildcard ceed*.c) 101BACKENDS := /cpu/self/ref /cpu/self/tmpl /cpu/self/opt 102 103# Tests 104tests.c := $(sort $(wildcard tests/t[0-9][0-9]-*.c)) 105tests.f := $(sort $(wildcard tests/t[0-9][0-9]-*.f)) 106tests := $(tests.c:tests/%.c=$(OBJDIR)/%) 107ctests := $(tests) 108tests += $(tests.f:tests/%.f=$(OBJDIR)/%) 109#examples 110examples.c := $(sort $(wildcard examples/ceed/*.c)) 111examples.f := $(sort $(wildcard examples/ceed/*.f)) 112examples := $(examples.c:examples/ceed/%.c=$(OBJDIR)/%) 113examples += $(examples.f:examples/ceed/%.f=$(OBJDIR)/%) 114#mfemexamples 115mfemexamples.cpp := $(sort $(wildcard examples/mfem/*.cpp)) 116mfemexamples := $(mfemexamples.cpp:examples/mfem/%.cpp=$(OBJDIR)/mfem-%) 117petscexamples.c := $(sort $(wildcard examples/petsc/*.c)) 118petscexamples := $(petscexamples.c:examples/petsc/%.c=$(OBJDIR)/petsc-%) 119 120# backends/[ref & occa & magma] 121ref.c := $(sort $(wildcard backends/ref/*.c)) 122template.c := $(sort $(wildcard backends/template/*.c)) 123optimized.c:= $(sort $(wildcard backends/optimized/*.c)) 124occa.c := $(sort $(wildcard backends/occa/*.c)) 125magma_preprocessor := python backends/magma/gccm.py 126magma_pre_src := $(filter-out %_tmp.c, $(wildcard backends/magma/ceed-*.c)) 127magma_dsrc := $(wildcard backends/magma/magma_d*.c) 128magma_tmp.c := $(magma_pre_src:%.c=%_tmp.c) 129magma_tmp.cu := $(magma_pre_src:%.c=%_cuda.cu) 130magma_allsrc.c := $(magma_dsrc) $(magma_tmp.c) 131magma_allsrc.cu:= $(magma_tmp.cu) 132 133# Output using the 216-color rules mode 134rule_file = $(notdir $(1)) 135rule_path = $(patsubst %/,%,$(dir $(1))) 136last_path = $(notdir $(patsubst %/,%,$(dir $(1)))) 137ansicolor = $(shell echo $(call last_path,$(1)) | cksum | cut -b1-2 | xargs -IS expr 2 \* S + 17) 138emacs_out = @printf " %10s %s/%s\n" $(1) $(call rule_path,$(2)) $(call rule_file,$(2)) 139color_out = @if [ -t 1 ]; then \ 140 printf " %10s \033[38;5;%d;1m%s\033[m/%s\n" \ 141 $(1) $(call ansicolor,$(2)) \ 142 $(call rule_path,$(2)) $(call rule_file,$(2)); else \ 143 printf " %10s %s\n" $(1) $(2); fi 144# if TERM=dumb, use it, otherwise switch to the term one 145output = $(if $(TERM:dumb=),$(call color_out,$1,$2),$(call emacs_out,$1,$2)) 146 147# if V is set to non-nil, turn the verbose mode 148quiet = $(if $(V),$($(1)),$(call output,$1,$@);$($(1))) 149 150.SUFFIXES: 151.SUFFIXES: .c .o .d 152.SECONDEXPANSION: # to expand $$(@D)/.DIR 153 154.SECONDARY: $(magma_tmp.c) $(magma_tmp.cu) 155 156%/.DIR : 157 @mkdir -p $(@D) 158 @touch $@ 159 160.PRECIOUS: %/.DIR 161 162this: $(libceed) $(ceed.pc) 163# run 'this' target in parallel 164all:;@$(MAKE) $(MFLAGS) V=$(V) this 165 166$(libceed) : LDFLAGS += $(if $(DARWIN), -install_name @rpath/$(notdir $(libceed))) 167 168libceed.c += $(ref.c) 169libceed.c += $(template.c) 170libceed.c += $(optimized.c) 171ifneq ($(wildcard $(OCCA_DIR)/lib/libocca.*),) 172 $(libceed) : LDFLAGS += -L$(OCCA_DIR)/lib -Wl,-rpath,$(abspath $(OCCA_DIR)/lib) 173 $(libceed) : LDLIBS += -locca 174 libceed.c += $(occa.c) 175 $(occa.c:%.c=$(OBJDIR)/%.o) : CFLAGS += -I$(OCCA_DIR)/include 176 BACKENDS += /cpu/occa /gpu/occa /omp/occa /ocl/occa 177endif 178ifneq ($(wildcard $(MAGMA_DIR)/lib/libmagma.*),) 179 CUDA_LIB_DIR := $(wildcard $(foreach d,lib lib64,$(CUDA_DIR)/$d/libcudart.${SO_EXT})) 180 CUDA_LIB_DIR := $(patsubst %/,%,$(dir $(firstword $(CUDA_LIB_DIR)))) 181 ifneq ($(CUDA_LIB_DIR),) 182 cuda_link = -Wl,-rpath,$(CUDA_LIB_DIR) -L$(CUDA_LIB_DIR) -lcublas -lcusparse -lcudart 183 omp_link = -fopenmp 184 magma_link_static = -L$(MAGMA_DIR)/lib -lmagma $(cuda_link) $(omp_link) 185 magma_link_shared = -L$(MAGMA_DIR)/lib -Wl,-rpath,$(abspath $(MAGMA_DIR)/lib) -lmagma 186 magma_link := $(if $(wildcard $(MAGMA_DIR)/lib/libmagma.${SO_EXT}),$(magma_link_shared),$(magma_link_static)) 187 magma_allsrc.o = $(magma_allsrc.c:%.c=$(OBJDIR)/%.o) $(magma_allsrc.cu:%.cu=$(OBJDIR)/%.o) 188 $(libceed) : LDLIBS += $(magma_link) 189 $(tests) $(examples) : LDLIBS += $(magma_link) 190 $(libceed) : $(magma_allsrc.o) 191 libceed.c += $(magma_allsrc.c) 192 libceed.cu += $(magma_allsrc.cu) 193 $(magma_allsrc.c:%.c=$(OBJDIR)/%.o) : CFLAGS += -DADD_ -I$(MAGMA_DIR)/include -I$(CUDA_DIR)/include 194 $(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 195 BACKENDS += /gpu/magma 196 endif 197endif 198 199export BACKENDS 200 201# generate magma_tmp.c and magma_cuda.cu from magma.c 202$(magma_tmp.c) $(magma_tmp.cu): $(magma_pre_src) | $$(@D)/.DIR 203 $(magma_preprocessor) $< 204 205$(libceed) : $(libceed.c:%.c=$(OBJDIR)/%.o) $(libceed.cu:%.cu=$(OBJDIR)/%.o) | $$(@D)/.DIR 206 $(call quiet,CC) $(LDFLAGS) -shared -o $@ $^ $(LDLIBS) 207 208$(OBJDIR)/%.o : %.c | $$(@D)/.DIR 209 $(call quiet,CC) $(CPPFLAGS) $(CFLAGS) -c -o $@ $(abspath $<) 210 211$(OBJDIR)/%.o : %.cu | $$(@D)/.DIR 212 $(call quiet,NVCC) $(CPPFLAGS) $(NVCCFLAGS) -c -o $@ $(abspath $<) 213 214$(OBJDIR)/% : tests/%.c | $$(@D)/.DIR 215 $(call quiet,LINK.c) -o $@ $(abspath $<) -lceed $(LDLIBS) 216 217$(OBJDIR)/% : tests/%.f | $$(@D)/.DIR 218 $(call quiet,LINK.F) -o $@ $(abspath $<) -lceed $(LDLIBS) 219 220$(OBJDIR)/% : examples/ceed/%.c | $$(@D)/.DIR 221 $(call quiet,LINK.c) -o $@ $(abspath $<) -lceed $(LDLIBS) 222 223$(OBJDIR)/% : examples/ceed/%.f | $$(@D)/.DIR 224 $(call quiet,LINK.F) -o $@ $(abspath $<) -lceed $(LDLIBS) 225 226$(OBJDIR)/mfem-% : examples/mfem/%.cpp $(libceed) | $$(@D)/.DIR 227 $(MAKE) -C examples/mfem CEED_DIR=`pwd` $* 228 mv examples/mfem/$* $@ 229 230$(OBJDIR)/petsc-% : examples/petsc/%.c $(libceed) $(ceed.pc) | $$(@D)/.DIR 231 $(MAKE) -C examples/petsc CEED_DIR=`pwd` $* 232 mv examples/petsc/$* $@ 233 234$(tests) $(examples) : $(libceed) 235$(tests) $(examples) : LDFLAGS += -Wl,-rpath,$(abspath $(LIBDIR)) -L$(LIBDIR) 236 237run-% : $(OBJDIR)/% 238 @tests/tap.sh $(<:build/%=%) 239# Test core libCEED 240test : $(tests:$(OBJDIR)/%=run-%) $(examples:$(OBJDIR)/%=run-%) 241 242# run test target in parallel 243tst : ;@$(MAKE) $(MFLAGS) V=$(V) test 244# CPU C tests only for backend % 245ctc-% : $(ctests);@$(foreach tst,$(ctests),$(tst) /cpu/$*;) 246 247prove : $(tests) $(examples) 248 $(info Testing backends: $(BACKENDS)) 249 $(PROVE) $(PROVE_OPTS) --exec 'tests/tap.sh' $(tests:$(OBJDIR)/%=%) $(examples:$(OBJDIR)/%=%) 250# run prove target in parallel 251prv : ;@$(MAKE) $(MFLAGS) V=$(V) prove 252 253alltests := $(tests) $(examples) $(if $(MFEM_DIR),$(mfemexamples)) $(if $(PETSC_DIR),$(petscexamples)) 254prove-all : $(alltests) 255 $(info Testing backends: $(BACKENDS)) 256 $(PROVE) $(PROVE_OPTS) --exec 'tests/tap.sh' $(alltests:$(OBJDIR)/%=%) 257 258examples : $(examples) 259 260$(ceed.pc) : pkgconfig-prefix = $(abspath .) 261$(OBJDIR)/ceed.pc : pkgconfig-prefix = $(prefix) 262.INTERMEDIATE : $(OBJDIR)/ceed.pc 263%/ceed.pc : ceed.pc.template | $$(@D)/.DIR 264 @sed "s:%prefix%:$(pkgconfig-prefix):" $< > $@ 265 266# The occa executable is not linked with RPATH by default, so we need it to find its libocca.so 267OCCA := $(if $(DARWIN),DYLD_LIBRARY_PATH,LD_LIBRARY_PATH)=$(OCCA_DIR)/lib $(OCCA_DIR)/bin/occa 268OKL_KERNELS := $(wildcard backends/occa/*.okl) 269 270okl-cache : 271 $(OCCA) cache ceed $(OKL_KERNELS) 272 273okl-clear: 274 $(OCCA) clear -y -l ceed 275 276install : $(libceed) $(OBJDIR)/ceed.pc 277 $(INSTALL) -d "$(DESTDIR)$(includedir)" "$(DESTDIR)$(libdir)" "$(DESTDIR)$(okldir)" "$(DESTDIR)$(pkgconfigdir)" 278 $(INSTALL_DATA) include/ceed.h "$(DESTDIR)$(includedir)/" 279 $(INSTALL_DATA) include/ceedf.h "$(DESTDIR)$(includedir)/" 280 $(INSTALL_DATA) $(libceed) "$(DESTDIR)$(libdir)/" 281 $(INSTALL_DATA) $(OBJDIR)/ceed.pc "$(DESTDIR)$(pkgconfigdir)/" 282 $(INSTALL_DATA) $(OKL_KERNELS) "$(DESTDIR)$(okldir)/" 283 284.PHONY : all cln clean print test tst prove prv examples style install doc okl-cache okl-clear 285 286cln clean : 287 $(RM) *.o *.d $(libceed) 288 $(RM) -r *.dSYM $(OBJDIR) $(LIBDIR)/pkgconfig 289 $(MAKE) -C examples/ceed clean 290 $(MAKE) -C examples/mfem clean 291 $(MAKE) -C examples/petsc clean 292 (cd examples/nek5000 && bash make-nek-examples.sh clean) 293 $(RM) $(magma_tmp.c) $(magma_tmp.cu) backends/magma/*~ backends/magma/*.o 294 295distclean : clean 296 $(RM) -r doc/html 297 298doc : 299 doxygen Doxyfile 300 301style : 302 astyle --style=google --indent=spaces=2 --max-code-length=80 \ 303 --keep-one-line-statements --keep-one-line-blocks --lineend=linux \ 304 --suffix=none --preserve-date --formatted \ 305 *.[ch] tests/*.[ch] backends/*/*.[ch] examples/*/*.[ch] examples/*/*.[ch]pp 306 307print : 308 @echo $(VAR)=$($(VAR)) 309 310print-% : 311 $(info [ variable name]: $*) 312 $(info [ origin]: $(origin $*)) 313 $(info [ value]: $(value $*)) 314 $(info [expanded value]: $($*)) 315 $(info ) 316 @true 317 318-include $(libceed.c:%.c=build/%.d) $(tests.c:tests/%.c=build/%.d) 319