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 17CONFIG ?= config.mk 18-include $(CONFIG) 19COMMON ?= common.mk 20-include $(COMMON) 21 22ifeq (,$(filter-out undefined default,$(origin CC))) 23 CC = gcc 24endif 25ifeq (,$(filter-out undefined default,$(origin CXX))) 26 CXX = g++ 27endif 28ifeq (,$(filter-out undefined default,$(origin FC))) 29 FC = gfortran 30endif 31ifeq (,$(filter-out undefined default,$(origin LINK))) 32 LINK = $(CC) 33endif 34NVCC ?= $(CUDA_DIR)/bin/nvcc 35NVCC_CXX ?= $(CXX) 36HIPCC ?= $(HIP_DIR)/bin/hipcc 37 38# ASAN must be left empty if you don't want to use it 39ASAN ?= 40 41LDFLAGS ?= 42UNDERSCORE ?= 1 43 44# MFEM_DIR env variable should point to sibling directory 45ifneq ($(wildcard ../mfem/libmfem.*),) 46 MFEM_DIR ?= ../mfem 47endif 48 49# NEK5K_DIR env variable should point to sibling directory 50ifneq ($(wildcard ../Nek5000/*),) 51 NEK5K_DIR ?= $(abspath ../Nek5000) 52endif 53export NEK5K_DIR 54MPI ?= 1 55 56# CEED_DIR env for NEK5K testing 57export CEED_DIR = $(abspath .) 58 59# XSMM_DIR env variable should point to XSMM master (github.com/hfp/libxsmm) 60XSMM_DIR ?= ../libxsmm 61 62# OCCA_DIR env variable should point to OCCA master (github.com/libocca/occa) 63OCCA_DIR ?= ../occa 64 65# env variable MAGMA_DIR can be used too 66MAGMA_DIR ?= ../magma 67# If CUDA_DIR is not set, check for nvcc, or resort to /usr/local/cuda 68CUDA_DIR ?= $(or $(patsubst %/,%,$(dir $(patsubst %/,%,$(dir \ 69 $(shell which nvcc 2> /dev/null))))),/usr/local/cuda) 70HIP_DIR ?= /opt/rocm 71 72# Check for PETSc in ../petsc 73ifneq ($(wildcard ../petsc/lib/libpetsc.*),) 74 PETSC_DIR ?= ../petsc 75endif 76 77# Warning: SANTIZ options still don't run with /gpu/occa 78# export LSAN_OPTIONS=suppressions=.asanignore 79AFLAGS = -fsanitize=address #-fsanitize=undefined -fno-omit-frame-pointer 80 81# Note: Intel oneAPI C/C++ compiler is now icx/icpx 82CC_VENDOR := $(subst oneAPI,icc,$(patsubst gcc%,gcc,$(firstword $(filter gcc% clang icc oneAPI XL,$(shell $(CC) --version))))) 83FC_VENDOR := $(firstword $(filter GNU ifort XL,$(shell $(FC) --version 2>&1 || $(FC) -qversion))) 84 85# Default extra flags by vendor 86MARCHFLAG.gcc := -march=native 87MARCHFLAG.clang := $(MARCHFLAG.gcc) 88MARCHFLAG.icc := 89OMP_SIMD_FLAG.gcc := -fopenmp-simd 90OMP_SIMD_FLAG.clang := $(OMP_SIMD_FLAG.gcc) 91OMP_SIMD_FLAG.icc := -qopenmp-simd 92OPT.gcc := -ffp-contract=fast 93OPT.clang := $(OPT.gcc) 94CFLAGS.gcc := -fPIC -std=c99 -Wall -Wextra -Wno-unused-parameter -MMD -MP 95CFLAGS.clang := $(CFLAGS.gcc) 96CFLAGS.icc := $(CFLAGS.gcc) 97CFLAGS.XL := -qpic -MMD 98CXXFLAGS.gcc := -fPIC -std=c++11 -Wall -Wextra -Wno-unused-parameter -MMD -MP 99CXXFLAGS.clang := $(CXXFLAGS.gcc) 100CXXFLAGS.icc := $(CXXFLAGS.gcc) 101CXXFLAGS.XL := -qpic -MMD 102FFLAGS.GNU := -fPIC -cpp -Wall -Wextra -Wno-unused-parameter -Wno-unused-dummy-argument -MMD -MP 103FFLAGS.ifort := -fPIC -cpp 104FFLAGS.XL := -qpic -ffree-form -qpreprocess -qextname -MMD 105 106# This check works with compilers that use gcc and clang. It fails with some 107# compilers; e.g., xlc apparently ignores all options when -E is passed, thus 108# succeeds with any flags. Users can pass MARCHFLAG=... if desired. 109cc_check_flag = $(shell $(CC) -E -Werror $(1) -x c /dev/null > /dev/null 2>&1 && echo 1) 110MARCHFLAG := $(MARCHFLAG.$(CC_VENDOR)) 111MARCHFLAG := $(if $(call cc_check_flag,$(MARCHFLAG)),$(MARCHFLAG),-mcpu=native) 112MARCHFLAG := $(if $(call cc_check_flag,$(MARCHFLAG)),$(MARCHFLAG)) 113 114OMP_SIMD_FLAG := $(OMP_SIMD_FLAG.$(CC_VENDOR)) 115OMP_SIMD_FLAG := $(if $(call cc_check_flag,$(OMP_SIMD_FLAG)),$(OMP_SIMD_FLAG)) 116 117OPT ?= -O -g $(MARCHFLAG) $(OPT.$(CC_VENDOR)) $(OMP_SIMD_FLAG) 118CFLAGS ?= $(OPT) $(CFLAGS.$(CC_VENDOR)) 119CXXFLAGS ?= $(OPT) $(CXXFLAGS.$(CC_VENDOR)) 120NVCCFLAGS ?= -ccbin $(CXX) -Xcompiler "$(OPT)" -Xcompiler -fPIC 121HIPCCFLAGS ?= $(filter-out $(OMP_SIMD_FLAG),$(OPT)) -fPIC 122FFLAGS ?= $(OPT) $(FFLAGS.$(FC_VENDOR)) 123 124ifeq ($(COVERAGE), 1) 125 CFLAGS += --coverage 126 CXXFLAGS += --coverage 127 LDFLAGS += --coverage 128endif 129 130CFLAGS += $(if $(ASAN),$(AFLAGS)) 131FFLAGS += $(if $(ASAN),$(AFLAGS)) 132LDFLAGS += $(if $(ASAN),$(AFLAGS)) 133CPPFLAGS += -I./include 134LDLIBS = -lm 135OBJDIR := build 136LIBDIR := lib 137 138# Installation variables 139prefix ?= /usr/local 140bindir = $(prefix)/bin 141libdir = $(prefix)/lib 142includedir = $(prefix)/include 143pkgconfigdir = $(libdir)/pkgconfig 144INSTALL = install 145INSTALL_PROGRAM = $(INSTALL) 146INSTALL_DATA = $(INSTALL) -m644 147 148# Get number of processors of the machine 149NPROCS := $(shell getconf _NPROCESSORS_ONLN) 150# prepare make options to run in parallel 151MFLAGS := -j $(NPROCS) --warn-undefined-variables \ 152 --no-print-directory --no-keep-going 153 154PYTHON ?= python3 155PROVE ?= prove 156PROVE_OPTS ?= -j $(NPROCS) 157DARWIN := $(filter Darwin,$(shell uname -s)) 158SO_EXT := $(if $(DARWIN),dylib,so) 159 160ceed.pc := $(LIBDIR)/pkgconfig/ceed.pc 161libceed := $(LIBDIR)/libceed.$(SO_EXT) 162CEED_LIBS = -lceed 163libceed.c := $(filter-out interface/ceed-cuda.c interface/ceed-hip.c, $(wildcard interface/ceed*.c)) 164gallery.c := $(wildcard gallery/*/ceed*.c) 165libceed.c += $(gallery.c) 166libceed_test := $(LIBDIR)/libceed_test.$(SO_EXT) 167libceeds = $(libceed) $(libceed_test) 168BACKENDS_BUILTIN := /cpu/self/ref/serial /cpu/self/ref/blocked /cpu/self/opt/serial /cpu/self/opt/blocked 169BACKENDS := $(BACKENDS_BUILTIN) 170 171# Tests 172tests.c := $(sort $(wildcard tests/t[0-9][0-9][0-9]-*.c)) 173tests.f := $(sort $(wildcard tests/t[0-9][0-9][0-9]-*.f90)) 174tests := $(tests.c:tests/%.c=$(OBJDIR)/%) 175ctests := $(tests) 176tests += $(tests.f:tests/%.f90=$(OBJDIR)/%) 177# Examples 178examples.c := $(sort $(wildcard examples/ceed/*.c)) 179examples.f := $(sort $(wildcard examples/ceed/*.f)) 180examples := $(examples.c:examples/ceed/%.c=$(OBJDIR)/%) 181examples += $(examples.f:examples/ceed/%.f=$(OBJDIR)/%) 182# MFEM Examples 183mfemexamples.cpp := $(sort $(wildcard examples/mfem/*.cpp)) 184mfemexamples := $(mfemexamples.cpp:examples/mfem/%.cpp=$(OBJDIR)/mfem-%) 185# Nek5K Examples 186nekexamples := $(OBJDIR)/nek-bps 187# PETSc Examples 188petscexamples.c := $(wildcard examples/petsc/*.c) 189petscexamples := $(petscexamples.c:examples/petsc/%.c=$(OBJDIR)/petsc-%) 190# Fluid Dynamics Examples 191fluidsexamples.c := $(sort $(wildcard examples/fluids/*.c)) 192fluidsexamples := $(fluidsexamples.c:examples/fluids/%.c=$(OBJDIR)/fluids-%) 193# Solid Mechanics Examples 194solidsexamples.c := $(sort $(wildcard examples/solids/*.c)) 195solidsexamples := $(solidsexamples.c:examples/solids/%.c=$(OBJDIR)/solids-%) 196 197# Backends/[ref, blocked, template, memcheck, opt, avx, occa, magma] 198ref.c := $(sort $(wildcard backends/ref/*.c)) 199blocked.c := $(sort $(wildcard backends/blocked/*.c)) 200template.c := $(sort $(wildcard backends/template/*.c)) 201ceedmemcheck.c := $(sort $(wildcard backends/memcheck/*.c)) 202opt.c := $(sort $(wildcard backends/opt/*.c)) 203avx.c := $(sort $(wildcard backends/avx/*.c)) 204xsmm.c := $(sort $(wildcard backends/xsmm/*.c)) 205cuda.c := $(sort $(wildcard backends/cuda/*.c)) 206cuda.cpp := $(sort $(wildcard backends/cuda/*.cpp)) 207cuda.cu := $(sort $(wildcard backends/cuda/kernels/*.cu)) 208cuda-shared.c := $(sort $(wildcard backends/cuda-shared/*.c)) 209cuda-shared.cu := $(sort $(wildcard backends/cuda-shared/kernels/*.cu)) 210cuda-gen.c := $(sort $(wildcard backends/cuda-gen/*.c)) 211cuda-gen.cpp := $(sort $(wildcard backends/cuda-gen/*.cpp)) 212cuda-gen.cu := $(sort $(wildcard backends/cuda-gen/kernels/*.cu)) 213occa.cpp := $(sort $(shell find backends/occa -type f -name *.cpp)) 214magma.c := $(sort $(wildcard backends/magma/*.c)) 215magma.cu := $(sort $(wildcard backends/magma/kernels/cuda/*.cu)) 216magma.hip := $(sort $(wildcard backends/magma/kernels/hip/*.hip.cpp)) 217hip.c := $(sort $(wildcard backends/hip/*.c)) 218hip.cpp := $(sort $(wildcard backends/hip/*.cpp)) 219hip.hip := $(sort $(wildcard backends/hip/kernels/*.hip.cpp)) 220 221# Quiet, color output 222quiet ?= $($(1)) 223 224# Cancel built-in and old-fashioned implicit rules which we don't use 225.SUFFIXES: 226 227.SECONDEXPANSION: # to expand $$(@D)/.DIR 228 229%/.DIR : 230 @mkdir -p $(@D) 231 @touch $@ 232 233.PRECIOUS: %/.DIR 234 235lib: $(libceed) $(ceed.pc) 236# run 'lib' target in parallel 237par:;@$(MAKE) $(MFLAGS) V=$(V) lib 238backend_status = $(if $(filter $1,$(BACKENDS)), [backends: $1], [not found]) 239info: 240 $(info ------------------------------------) 241 $(info CC = $(CC)) 242 $(info CXX = $(CXX)) 243 $(info FC = $(FC)) 244 $(info CPPFLAGS = $(CPPFLAGS)) 245 $(info CFLAGS = $(value CFLAGS)) 246 $(info CXXFLAGS = $(value CXXFLAGS)) 247 $(info FFLAGS = $(value FFLAGS)) 248 $(info NVCCFLAGS = $(value NVCCFLAGS)) 249 $(info HIPCCFLAGS = $(value HIPCCFLAGS)) 250 $(info LDFLAGS = $(value LDFLAGS)) 251 $(info LDLIBS = $(LDLIBS)) 252 $(info OPT = $(OPT)) 253 $(info AFLAGS = $(AFLAGS)) 254 $(info ASAN = $(or $(ASAN),(empty))) 255 $(info V = $(or $(V),(empty)) [verbose=$(if $(V),on,off)]) 256 $(info ------------------------------------) 257 $(info MEMCHK_STATUS = $(MEMCHK_STATUS)$(call backend_status,$(MEMCHK_BACKENDS))) 258 $(info AVX_STATUS = $(AVX_STATUS)$(call backend_status,$(AVX_BACKENDS))) 259 $(info XSMM_DIR = $(XSMM_DIR)$(call backend_status,$(XSMM_BACKENDS))) 260 $(info OCCA_DIR = $(OCCA_DIR)$(call backend_status,$(OCCA_BACKENDS))) 261 $(info MAGMA_DIR = $(MAGMA_DIR)$(call backend_status,$(MAGMA_BACKENDS))) 262 $(info CUDA_DIR = $(CUDA_DIR)$(call backend_status,$(CUDA_BACKENDS))) 263 $(info HIP_DIR = $(HIP_DIR)$(call backend_status,$(HIP_BACKENDS))) 264 $(info ------------------------------------) 265 $(info MFEM_DIR = $(MFEM_DIR)) 266 $(info NEK5K_DIR = $(NEK5K_DIR)) 267 $(info PETSC_DIR = $(PETSC_DIR)) 268 $(info ------------------------------------) 269 $(info prefix = $(prefix)) 270 $(info includedir = $(value includedir)) 271 $(info libdir = $(value libdir)) 272 $(info pkgconfigdir = $(value pkgconfigdir)) 273 $(info ------------------------------------) 274 @true 275info-backends: 276 $(info make: 'lib' with optional backends: $(filter-out $(BACKENDS_BUILTIN),$(BACKENDS))) 277.PHONY: lib all par info info-backends 278 279$(libceed) : LDFLAGS += $(if $(DARWIN), -install_name @rpath/$(notdir $(libceed))) 280$(libceed_test) : LDFLAGS += $(if $(DARWIN), -install_name @rpath/$(notdir $(libceed_test))) 281 282# Standard Backends 283libceed.c += $(ref.c) 284libceed.c += $(blocked.c) 285libceed.c += $(opt.c) 286 287# Testing Backends 288test_backends.c := $(template.c) 289TEST_BACKENDS := /cpu/self/tmpl /cpu/self/tmpl/sub 290 291# Memcheck Backend 292MEMCHK_STATUS = Disabled 293MEMCHK := $(shell echo "\#include <valgrind/memcheck.h>" | $(CC) $(CPPFLAGS) -E - >/dev/null 2>&1 && echo 1) 294MEMCHK_BACKENDS = /cpu/self/memcheck/serial /cpu/self/memcheck/blocked 295ifeq ($(MEMCHK),1) 296 MEMCHK_STATUS = Enabled 297 libceed.c += $(ceedmemcheck.c) 298 BACKENDS += $(MEMCHK_BACKENDS) 299endif 300 301# AVX Backed 302AVX_STATUS = Disabled 303AVX_FLAG := $(if $(filter clang,$(CC_VENDOR)),+avx,-mavx) 304AVX := $(filter $(AVX_FLAG),$(shell $(CC) $(OPT) -v -E -x c /dev/null 2>&1)) 305AVX_BACKENDS = /cpu/self/avx/serial /cpu/self/avx/blocked 306ifneq ($(AVX),) 307 AVX_STATUS = Enabled 308 libceed.c += $(avx.c) 309 BACKENDS += $(AVX_BACKENDS) 310endif 311 312# libXSMM Backends 313XSMM_BACKENDS = /cpu/self/xsmm/serial /cpu/self/xsmm/blocked 314ifneq ($(wildcard $(XSMM_DIR)/lib/libxsmm.*),) 315 $(libceeds) : LDFLAGS += -L$(XSMM_DIR)/lib -Wl,-rpath,$(abspath $(XSMM_DIR)/lib) 316 $(libceeds) : LDLIBS += -lxsmm -ldl 317 MKL ?= 318 ifeq (,$(MKL)$(MKLROOT)) 319 BLAS_LIB = -lblas 320 else 321 ifneq ($(MKLROOT),) 322 # Some installs put everything inside an intel64 subdirectory, others not 323 MKL_LIBDIR = $(dir $(firstword $(wildcard $(MKLROOT)/lib/intel64/libmkl_sequential.* $(MKLROOT)/lib/libmkl_sequential.*))) 324 MKL_LINK = -L$(MKL_LIBDIR) -Wl,-rpath,$(MKL_LIBDIR) 325 endif 326 BLAS_LIB = $(MKL_LINK) -Wl,--no-as-needed -lmkl_intel_lp64 -lmkl_sequential -lmkl_core -lpthread -lm -ldl 327 endif 328 $(libceeds) : LDLIBS += $(BLAS_LIB) 329 libceed.c += $(xsmm.c) 330 $(xsmm.c:%.c=$(OBJDIR)/%.o) $(xsmm.c:%=%.tidy) : CPPFLAGS += -I$(XSMM_DIR)/include 331 BACKENDS += $(XSMM_BACKENDS) 332endif 333 334# OCCA Backends 335OCCA_BACKENDS = /cpu/self/occa 336ifneq ($(wildcard $(OCCA_DIR)/lib/libocca.*),) 337 OCCA_MODES := $(shell $(OCCA_DIR)/bin/occa modes) 338 OCCA_BACKENDS += $(if $(filter OpenMP,$(OCCA_MODES)),/cpu/openmp/occa) 339# OCCA_BACKENDS += $(if $(filter OpenCL,$(OCCA_MODES)),/gpu/opencl/occa) 340 OCCA_BACKENDS += $(if $(filter HIP,$(OCCA_MODES)),/gpu/hip/occa) 341 OCCA_BACKENDS += $(if $(filter CUDA,$(OCCA_MODES)),/gpu/cuda/occa) 342 343 $(libceeds) : CPPFLAGS += -I$(OCCA_DIR)/include 344 $(libceeds) : LDFLAGS += -L$(OCCA_DIR)/lib -Wl,-rpath,$(abspath $(OCCA_DIR)/lib) 345 $(libceeds) : LDLIBS += -locca 346 $(libceeds) : LINK = $(CXX) 347 libceed.cpp += $(occa.cpp) 348 BACKENDS += $(OCCA_BACKENDS) 349endif 350 351# CUDA Backends 352CUDA_LIB_DIR := $(wildcard $(foreach d,lib lib64,$(CUDA_DIR)/$d/libcudart.${SO_EXT})) 353CUDA_LIB_DIR := $(patsubst %/,%,$(dir $(firstword $(CUDA_LIB_DIR)))) 354CUDA_LIB_DIR_STUBS := $(CUDA_LIB_DIR)/stubs 355CUDA_BACKENDS = /gpu/cuda/ref /gpu/cuda/shared /gpu/cuda/gen 356ifneq ($(CUDA_LIB_DIR),) 357 $(libceeds) : CPPFLAGS += -I$(CUDA_DIR)/include 358 $(libceeds) : LDFLAGS += -L$(CUDA_LIB_DIR) -Wl,-rpath,$(abspath $(CUDA_LIB_DIR)) 359 $(libceeds) : LDLIBS += -lcudart -lnvrtc -lcuda -lcublas 360 $(libceeds) : LINK = $(CXX) 361 libceed.c += interface/ceed-cuda.c 362 libceed.c += $(cuda.c) $(cuda-shared.c) $(cuda-gen.c) 363 libceed.cpp += $(cuda.cpp) $(cuda-gen.cpp) 364 libceed.cu += $(cuda.cu) $(cuda-shared.cu) $(cuda-gen.cu) 365 BACKENDS += $(CUDA_BACKENDS) 366endif 367 368# HIP Backends 369HIP_LIB_DIR := $(wildcard $(foreach d,lib lib64,$(HIP_DIR)/$d/libamdhip64.${SO_EXT})) 370HIP_LIB_DIR := $(patsubst %/,%,$(dir $(firstword $(HIP_LIB_DIR)))) 371HIP_BACKENDS = /gpu/hip/ref 372ifneq ($(HIP_LIB_DIR),) 373 $(libceeds) : HIPCCFLAGS += -I./include 374 ifneq ($(CXX), $(HIPCC)) 375 CPPFLAGS += $(subst =,,$(shell $(HIP_DIR)/bin/hipconfig -C)) 376 endif 377 $(libceeds) : CPPFLAGS += -I$(HIP_DIR)/include -Wno-unused-function 378 $(libceeds) : LDFLAGS += -L$(HIP_LIB_DIR) -Wl,-rpath,$(abspath $(HIP_LIB_DIR)) 379 $(libceeds) : LDLIBS += -lamdhip64 -lhipblas 380 $(libceeds) : LINK = $(CXX) 381 libceed.c += interface/ceed-hip.c 382 libceed.c += $(hip.c) 383 libceed.cpp += $(hip.cpp) 384 libceed.hip += $(hip.hip) 385 BACKENDS += $(HIP_BACKENDS) 386endif 387 388# MAGMA Backend 389ifneq ($(wildcard $(MAGMA_DIR)/lib/libmagma.*),) 390 MAGMA_ARCH=$(shell nm -g $(MAGMA_DIR)/lib/libmagma.* | grep -c "hipblas") 391 ifeq ($(MAGMA_ARCH), 0) #CUDA MAGMA 392 ifneq ($(CUDA_LIB_DIR),) 393 cuda_link = -Wl,-rpath,$(CUDA_LIB_DIR) -L$(CUDA_LIB_DIR) -lcublas -lcusparse -lcudart 394 omp_link = -fopenmp 395 magma_link_static = -L$(MAGMA_DIR)/lib -lmagma $(cuda_link) $(omp_link) 396 magma_link_shared = -L$(MAGMA_DIR)/lib -Wl,-rpath,$(abspath $(MAGMA_DIR)/lib) -lmagma 397 magma_link := $(if $(wildcard $(MAGMA_DIR)/lib/libmagma.${SO_EXT}),$(magma_link_shared),$(magma_link_static)) 398 $(libceeds) : LDLIBS += $(magma_link) 399 $(tests) $(examples) : LDLIBS += $(magma_link) 400 libceed.c += $(magma.c) 401 libceed.cu += $(magma.cu) 402 $(magma.c:%.c=$(OBJDIR)/%.o) $(magma.c:%=%.tidy) : CPPFLAGS += -DADD_ -I$(MAGMA_DIR)/include -I$(CUDA_DIR)/include 403 $(magma.cu:%.cu=$(OBJDIR)/%.o) : CPPFLAGS += --compiler-options=-fPIC -DADD_ -I$(MAGMA_DIR)/include -I$(MAGMA_DIR)/magmablas -I$(MAGMA_DIR)/control -I$(CUDA_DIR)/include 404 MAGMA_BACKENDS = /gpu/cuda/magma /gpu/cuda/magma/det 405 endif 406 else # HIP MAGMA 407 ifneq ($(HIP_LIB_DIR),) 408 hip_link = -Wl,-rpath,$(HIP_LIB_DIR) -L$(HIP_LIB_DIR) -lhipblas -lhipsparse -lamdhip64 409 omp_link = -fopenmp 410 magma_link_static = -L$(MAGMA_DIR)/lib -lmagma $(hip_link) $(omp_link) 411 magma_link_shared = -L$(MAGMA_DIR)/lib -Wl,-rpath,$(abspath $(MAGMA_DIR)/lib) -lmagma 412 magma_link := $(if $(wildcard $(MAGMA_DIR)/lib/libmagma.${SO_EXT}),$(magma_link_shared),$(magma_link_static)) 413 $(libceeds) : LDLIBS += $(magma_link) 414 $(tests) $(examples) : LDLIBS += $(magma_link) 415 libceed.c += $(magma.c) 416 libceed.hip += $(magma.hip) 417 ifneq ($(CXX), $(HIPCC)) 418 $(magma.c:%.c=$(OBJDIR)/%.o) $(magma.c:%=%.tidy) : CPPFLAGS += -I$(MAGMA_DIR)/include -I$(HIP_DIR)/include -DHAVE_HIP -DADD_ 419 else 420 $(magma.c:%.c=$(OBJDIR)/%.o) $(magma.c:%=%.tidy) : HIPCCFLAGS += -I$(MAGMA_DIR)/include -I$(HIP_DIR)/include -DHAVE_HIP -DADD_ 421 endif 422 $(magma.hip:%.hip.cpp=$(OBJDIR)/%.o) : HIPCCFLAGS += -I$(MAGMA_DIR)/include -I$(MAGMA_DIR)/magmablas -I$(MAGMA_DIR)/control -I$(HIP_DIR)/include -DHAVE_HIP -DADD_ 423 MAGMA_BACKENDS = /gpu/hip/magma /gpu/hip/magma/det 424 endif 425 endif 426 BACKENDS += $(MAGMA_BACKENDS) 427endif 428 429export BACKENDS 430 431libceed.o = $(libceed.c:%.c=$(OBJDIR)/%.o) $(libceed.cpp:%.cpp=$(OBJDIR)/%.o) $(libceed.cu:%.cu=$(OBJDIR)/%.o) $(libceed.hip:%.hip.cpp=$(OBJDIR)/%.o) 432$(filter %fortran.o,$(libceed.o)) : CPPFLAGS += $(if $(filter 1,$(UNDERSCORE)),-DUNDERSCORE) 433$(libceed.o): | info-backends 434$(libceed) : $(libceed.o) | $$(@D)/.DIR 435 $(call quiet,LINK) $(LDFLAGS) -shared -o $@ $^ $(LDLIBS) 436 437$(OBJDIR)/%.o : $(CURDIR)/%.c | $$(@D)/.DIR 438 $(call quiet,CC) $(CPPFLAGS) $(CFLAGS) -c -o $@ $(abspath $<) 439 440$(OBJDIR)/%.o : $(CURDIR)/%.cpp | $$(@D)/.DIR 441 $(call quiet,CXX) $(CPPFLAGS) $(CXXFLAGS) -c -o $@ $(abspath $<) 442 443$(OBJDIR)/%.o : $(CURDIR)/%.cu | $$(@D)/.DIR 444 $(call quiet,NVCC) $(filter-out -Wno-unused-function, $(CPPFLAGS)) $(NVCCFLAGS) -c -o $@ $(abspath $<) 445 446$(OBJDIR)/%.o : $(CURDIR)/%.hip.cpp | $$(@D)/.DIR 447 $(call quiet,HIPCC) $(HIPCCFLAGS) -c -o $@ $(abspath $<) 448 449$(OBJDIR)/% : tests/%.c | $$(@D)/.DIR 450 $(call quiet,LINK.c) $(CEED_LDFLAGS) -o $@ $(abspath $<) $(CEED_LIBS) $(LDLIBS) 451 452$(OBJDIR)/% : tests/%.f90 | $$(@D)/.DIR 453 $(call quiet,LINK.F) -DSOURCE_DIR='"$(abspath $(<D))/"' $(CEED_LDFLAGS) -o $@ $(abspath $<) $(CEED_LIBS) $(LDLIBS) 454 455$(OBJDIR)/% : examples/ceed/%.c | $$(@D)/.DIR 456 $(call quiet,LINK.c) $(CEED_LDFLAGS) -o $@ $(abspath $<) $(CEED_LIBS) $(LDLIBS) 457 458$(OBJDIR)/% : examples/ceed/%.f | $$(@D)/.DIR 459 $(call quiet,LINK.F) -DSOURCE_DIR='"$(abspath $(<D))/"' $(CEED_LDFLAGS) -o $@ $(abspath $<) $(CEED_LIBS) $(LDLIBS) 460 461$(OBJDIR)/mfem-% : examples/mfem/%.cpp $(libceed) | $$(@D)/.DIR 462 +$(MAKE) -C examples/mfem CEED_DIR=`pwd` \ 463 MFEM_DIR="$(abspath $(MFEM_DIR))" CXX=$(CXX) $* 464 mv examples/mfem/$* $@ 465 466# Note: Multiple Nek files cannot be built in parallel. The '+' here enables 467# this single Nek bps file to be built in parallel with other examples, 468# such as when calling `make prove-all -j2`. 469$(OBJDIR)/nek-bps : examples/nek/bps/bps.usr examples/nek/nek-examples.sh $(libceed) | $$(@D)/.DIR 470 +$(MAKE) -C examples MPI=$(MPI) CEED_DIR=`pwd` NEK5K_DIR="$(abspath $(NEK5K_DIR))" nek 471 mv examples/nek/build/bps $(OBJDIR)/bps 472 cp examples/nek/nek-examples.sh $(OBJDIR)/nek-bps 473 474$(OBJDIR)/petsc-% : examples/petsc/%.c $(libceed) $(ceed.pc) | $$(@D)/.DIR 475 +$(MAKE) -C examples/petsc CEED_DIR=`pwd` \ 476 PETSC_DIR="$(abspath $(PETSC_DIR))" OPT="$(OPT)" $* 477 mv examples/petsc/$* $@ 478 479$(OBJDIR)/fluids-% : examples/fluids/%.c $(libceed) $(ceed.pc) | $$(@D)/.DIR 480 +$(MAKE) -C examples/fluids CEED_DIR=`pwd` \ 481 PETSC_DIR="$(abspath $(PETSC_DIR))" OPT="$(OPT)" $* 482 mv examples/fluids/$* $@ 483 484$(OBJDIR)/solids-% : examples/solids/%.c $(libceed) $(ceed.pc) | $$(@D)/.DIR 485 +$(MAKE) -C examples/solids CEED_DIR=`pwd` \ 486 PETSC_DIR="$(abspath $(PETSC_DIR))" OPT="$(OPT)" $* 487 mv examples/solids/$* $@ 488 489libceed_test.o = $(test_backends.c:%.c=$(OBJDIR)/%.o) 490$(libceed_test) : $(libceed.o) $(libceed_test.o) | $$(@D)/.DIR 491 $(call quiet,LINK) $(LDFLAGS) -shared -o $@ $^ $(LDLIBS) 492 493$(examples) : $(libceed) 494$(tests) : $(libceed_test) 495$(tests) : CEED_LIBS = -lceed_test 496$(tests) $(examples) : LDFLAGS += -Wl,-rpath,$(abspath $(LIBDIR)) -L$(LIBDIR) 497 498run-t% : BACKENDS += $(TEST_BACKENDS) 499run-% : $(OBJDIR)/% 500 @tests/tap.sh $(<:$(OBJDIR)/%=%) 501 502external_examples := \ 503 $(if $(MFEM_DIR),$(mfemexamples)) \ 504 $(if $(PETSC_DIR),$(petscexamples)) \ 505 $(if $(NEK5K_DIR),$(nekexamples)) \ 506 $(if $(PETSC_DIR),$(fluidsexamples)) \ 507 $(if $(PETSC_DIR),$(solidsexamples)) 508 509allexamples = $(examples) $(external_examples) 510 511# The test and prove targets can be controlled via pattern searches. The 512# default is to run tests and those examples that have no external dependencies. 513# Examples of finer grained control: 514# 515# make test search='petsc mfem' # PETSc and MFEM examples 516# make prove search='t3' # t3xx series tests 517# make junit search='ex petsc' # core ex* and PETSc tests 518search ?= t ex 519realsearch = $(search:%=%%) 520matched = $(foreach pattern,$(realsearch),$(filter $(OBJDIR)/$(pattern),$(tests) $(allexamples))) 521 522# Test core libCEED 523test : $(matched:$(OBJDIR)/%=run-%) 524 525# Run test target in parallel 526tst : ;@$(MAKE) $(MFLAGS) V=$(V) test 527# CPU C tests only for backend % 528ctc-% : $(ctests);@$(foreach tst,$(ctests),$(tst) /cpu/$*;) 529 530prove : BACKENDS += $(TEST_BACKENDS) 531prove : $(matched) 532 $(info Testing backends: $(BACKENDS)) 533 $(PROVE) $(PROVE_OPTS) --exec 'tests/tap.sh' $(matched:$(OBJDIR)/%=%) 534# Run prove target in parallel 535prv : ;@$(MAKE) $(MFLAGS) V=$(V) prove 536 537prove-all : 538 +$(MAKE) prove realsearch=% 539 540junit-t% : BACKENDS += $(TEST_BACKENDS) 541junit-% : $(OBJDIR)/% 542 @printf " %10s %s\n" TEST $(<:$(OBJDIR)/%=%); $(PYTHON) tests/junit.py $(<:$(OBJDIR)/%=%) 543 544junit : $(matched:$(OBJDIR)/%=junit-%) 545 546all: $(alltests) 547 548examples : $(allexamples) 549ceedexamples : $(examples) 550nekexamples : $(nekexamples) 551mfemexamples : $(mfemexamples) 552petscexamples : $(petscexamples) 553 554# Benchmarks 555allbenchmarks = petsc-bps 556bench_targets = $(addprefix bench-,$(allbenchmarks)) 557.PHONY: $(bench_targets) benchmarks 558$(bench_targets): bench-%: $(OBJDIR)/% 559 cd benchmarks && ./benchmark.sh --ceed "$(BACKENDS)" -r $(*).sh 560benchmarks: $(bench_targets) 561 562$(ceed.pc) : pkgconfig-prefix = $(abspath .) 563$(OBJDIR)/ceed.pc : pkgconfig-prefix = $(prefix) 564.INTERMEDIATE : $(OBJDIR)/ceed.pc 565%/ceed.pc : ceed.pc.template | $$(@D)/.DIR 566 @sed "s:%prefix%:$(pkgconfig-prefix):" $< > $@ 567 568install : $(libceed) $(OBJDIR)/ceed.pc 569 $(INSTALL) -d $(addprefix $(if $(DESTDIR),"$(DESTDIR)"),"$(includedir)"\ 570 "$(libdir)" "$(pkgconfigdir)") 571 $(INSTALL_DATA) include/ceed.h "$(DESTDIR)$(includedir)/" 572 $(INSTALL_DATA) include/ceedf.h "$(DESTDIR)$(includedir)/" 573 $(INSTALL_DATA) include/ceed-hash.h "$(DESTDIR)$(includedir)/" 574 $(INSTALL_DATA) include/ceed-khash.h "$(DESTDIR)$(includedir)/" 575 $(INSTALL_DATA) $(libceed) "$(DESTDIR)$(libdir)/" 576 $(INSTALL_DATA) $(OBJDIR)/ceed.pc "$(DESTDIR)$(pkgconfigdir)/" 577 578.PHONY : cln clean doxygen doc lib install all print test tst prove prv prove-all junit examples style style-c style-py tidy info info-backends 579 580cln clean : 581 $(RM) -r $(OBJDIR) $(LIBDIR) dist *egg* .pytest_cache *cffi* 582 $(MAKE) -C examples clean NEK5K_DIR="$(abspath $(NEK5K_DIR))" 583 $(MAKE) -C tests/python clean 584 $(RM) benchmarks/*output.txt 585 586distclean : clean 587 $(RM) -r doc/html doc/sphinx/build $(CONFIG) 588 589DOXYGEN ?= doxygen 590doxygen : 591 $(DOXYGEN) Doxyfile 592 593doc-html doc-latexpdf doc-epub : doc-% : doxygen 594 make -C doc/sphinx $* 595 596doc : doc-html 597 598style-c : 599 @astyle --options=.astylerc \ 600 $(filter-out include/ceedf.h $(wildcard tests/t*-f.h), \ 601 $(wildcard include/*.h interface/*.[ch] tests/*.[ch] backends/*/*.[ch] \ 602 examples/*/*/*.[ch] examples/*/*.[ch] examples/*/*.[ch]pp gallery/*/*.[ch])) 603 604AUTOPEP8 = autopep8 605style-py : AUTOPEP8_ARGS = --in-place --aggressive 606style-py : 607 @$(AUTOPEP8) $(AUTOPEP8_ARGS) $(wildcard *.py python**/*.py tests/python**/*.py examples**/*.py doc/sphinx/source**/*.py benchmarks/*.py) 608 609style : style-c style-py 610 611CLANG_TIDY ?= clang-tidy 612 613%.c.tidy : %.c 614 $(CLANG_TIDY) $(TIDY_OPTS) $^ -- $(CPPFLAGS) --std=c99 -I$(CUDA_DIR)/include 615 616%.cpp.tidy : %.cpp 617 $(CLANG_TIDY) $(TIDY_OPTS) $^ -- $(CPPFLAGS) --std=c++11 -I$(CUDA_DIR)/include -I$(OCCA_DIR)/include 618 619tidy_c : $(libceed.c:%=%.tidy) 620tidy_cpp : $(libceed.cpp:%=%.tidy) 621 622tidy : tidy_c tidy_cpp 623 624print : 625 @echo $(VAR)=$($(VAR)) 626 627print-% : 628 $(info [ variable name]: $*) 629 $(info [ origin]: $(origin $*)) 630 $(info [ flavor]: $(flavor $*)) 631 $(info [ value]: $(value $*)) 632 $(info [expanded value]: $($*)) 633 $(info ) 634 @true 635 636# "make configure" detects any variables passed on the command line or 637# previously set in config.mk, caching them in config.mk as simple 638# (:=) variables. Variables set in config.mk or on the command line 639# take precedence over the defaults provided in the file. Typical 640# usage: 641# 642# make configure CC=/path/to/my/cc CUDA_DIR=/opt/cuda 643# make 644# make prove 645# 646# The values in the file can be updated by passing them on the command 647# line, e.g., 648# 649# make configure CC=/path/to/other/clang 650 651# All variables to consider for caching 652CONFIG_VARS = CC CXX FC NVCC NVCC_CXX HIPCC \ 653 OPT CFLAGS CPPFLAGS CXXFLAGS FFLAGS NVCCFLAGS HIPCCFLAGS \ 654 LDFLAGS LDLIBS \ 655 MAGMA_DIR XSMM_DIR CUDA_DIR MFEM_DIR PETSC_DIR NEK5K_DIR HIP_DIR 656 657# $(call needs_save,CFLAGS) returns true (a nonempty string) if CFLAGS 658# was set on the command line or in config.mk (where it will appear as 659# a simple variable). 660needs_save = $(or $(filter command line,$(origin $(1))),$(filter simple,$(flavor $(1)))) 661 662configure : 663 $(file > $(CONFIG)) 664 $(foreach v,$(CONFIG_VARS),$(if $(call needs_save,$(v)),$(file >> $(CONFIG),$(v) := $($(v))))) 665 @echo "Configuration cached in $(CONFIG):" 666 @cat $(CONFIG) 667 668wheel : export MARCHFLAG = -march=generic 669wheel : export WHEEL_PLAT = manylinux2010_x86_64 670wheel : 671 docker run -it --user $(shell id -u):$(shell id -g) --rm -v $(PWD):/io -w /io \ 672 -e MARCHFLAG -e WHEEL_PLAT \ 673 quay.io/pypa/$(WHEEL_PLAT) python/make-wheels.sh 674 675.PHONY : configure wheel 676 677# Include *.d deps when not -B = --always-make: useful if the paths are wonky in a container 678-include $(if $(filter B,$(MAKEFLAGS)),,$(libceed.c:%.c=$(OBJDIR)/%.d) $(tests.c:tests/%.c=$(OBJDIR)/%.d)) 679