Newer
Older
WAUTELET Philippe
committed
#MNH_LIC Copyright 1994-2023 CNRS, Meteo-France and Universite Paul Sabatier
#MNH_LIC This is part of the Meso-NH software governed by the CeCILL-C licence
#MNH_LIC version 1. See LICENSE, CeCILL-C_V1-en.txt and CeCILL-C_V1-fr.txt
#MNH_LIC for details. version 1.
##########################################################
# #
# Compiler Options #
# #
##########################################################
#OBJDIR_PATH=${WORKDIR}
#
# Processor type
#TP= -tp=sandybridge, haswell, skylake
#all x86 TP= -tp=px
#Pentium4 64-bit: p7-64
#Nuwa n375: Xeon E5-2680 v2: Ivy Bridge => between Sandy bridge and Haswell => sandybridge
#aeropc45-3: Xeon W-2265: Cascade Lake (version amelioree de Skylake) => skylake
#TP= -tp=sandybridge
#TP= -tp=skylake
#TP= -tp=p7-64
TP= -tp=px
#
#Version of CUDA
#(8.0 at least if compute capability >= 6.0)
CUDALEVEL=cuda11.4
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
#
#Compute capability of GPU
#
OPT_CPTCAP=
#Aeropc45: cc50
#Nuwa: cc35
#Ouessant Firestone K80: cc35
#Ouessant Minsky P100: cc60
#Nuwa V100: cc70
#OPT_CPTCAP= cc35,cc50,cc60,cc70
OPT_CPTCAP= cc35
#
#Compiler info level
#OPT_INFO = -Minfo=ftn,accel,inline,ipa,loop,lre,mp,opt,par,unified,vect,ccff
OPT_INFO = -Minfo=accel
#
#Compiler profiling options
# OPT_PROF = -Mprof=dwarf
#
#PW: if -Ktrap=fp: nvprof/pgprof do not work OPT_BASE = -Ktrap=fp ...
#PW: -Ktrap=fp raises false alarms with NVHPC 21.7
#PW: -g: big impact on performance
OPT_BASE = $(TP) -Mbackslash -Mextend -Kieee -Mnofma -Mallocatable=95 -g
#
OPT_PERF0 = -O0
OPT_PERF1 = -O1
OPT_PERF2 = -O2
#
OPT_MANAGED = -Mframe -Mnostack_arrays -acc=host,gpu -gpu=nofma,$(OPT_CPTCAP),$(CUDALEVEL),managed $(OPT_INFO) $(OPT_PROF)
OPT_MULTICORE = -acc=multicore $(OPT_INFO) $(OPT_PROF)
OPT_NOOPENACC = -acc=host $(OPT_INFO) $(OPT_PROF)
OPT_OPENACC = -Mframe -Mnostack_arrays -acc=host,gpu,verystrict -gpu=nofma,lineinfo,$(OPT_CPTCAP),$(CUDALEVEL) $(OPT_INFO) $(OPT_PROF)
OPT_OPENACC_PCAST = -Mframe -Mnostack_arrays -acc=host,gpu,verystrict -gpu=autocompare,nofma,lineinfo,$(OPT_CPTCAP),$(CUDALEVEL) $(OPT_INFO) $(OPT_PROF)
#
OPT_CHECK = -C #-Mchkfpstk -Mchkptr
OPT_I8 = -i8
OPT_R8 = -r8
#
ifeq "$(VER_USER)" "ZSOLVER"
CPPFLAGS += -DCARTESIANGEOMETRY -DPIECEWISELINEAR
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
PROG_LIST += MG_MAIN MG_MAIN_MNH_ALL
endif
#
IGNORE_OBJS += pgprof.o
#
# Real/integer 4/8 option
#
MNH_REAL ?=8
MNH_INT ?=4
#
ifneq "$(MNH_REAL)" "4"
OPT_BASE += $(OPT_R8)
endif
#
OPT_BASE_I4 := $(OPT_BASE)
ifeq "$(MNH_INT)" "8"
OPT_BASE += $(OPT_I8)
LFI_INT ?=8
else
LFI_INT ?=4
endif
#
#
OPT = $(OPT_BASE) $(OPT_PERF2)
OPT0 = $(OPT_BASE) $(OPT_PERF0)
OPT_NOCB = $(OPT_BASE) $(OPT_PERF2)
#
ifeq "$(OPTLEVEL)" "DEBUG"
OPT = $(OPT_BASE) $(OPT_PERF0) $(OPT_CHECK)
OPT0 = $(OPT_BASE) $(OPT_PERF0) $(OPT_CHECK)
OPT_NOCB = $(OPT_BASE) $(OPT_PERF0)
endif
#
ifeq "$(OPTLEVEL)" "MANAGED"
CPPFLAGS += -DMNH_OPENACC -DMNH_GPUDIRECT
# CPPFLAGS += -DMNH_OPENACC -ta=tesla
OPT = $(OPT_BASE) $(OPT_MANAGED) $(OPT_PERF2)
OPT0 = $(OPT_BASE) $(OPT_MANAGED) $(OPT_PERF0)
OPT_NOCB = $(OPT_BASE) $(OPT_MANAGED) $(OPT_PERF2)
CXXFLAGS = -acc -Kieee -Mnofma $(OPT_MANAGED)
endif
#
ifeq "$(OPTLEVEL)" "MULTICORE"
CPPFLAGS += -DMNH_OPENACC
OPT = $(OPT_BASE) $(OPT_MULTICORE)
OPT0 = $(OPT_BASE) $(OPT_MULTICORE) $(OPT_PERF0)
OPT_NOCB = $(OPT_BASE) $(OPT_MULTICORE)
CXXFLAGS = -acc -Kieee -Mnofma $(OPT_MULTICORE)
endif
#
ifeq "$(OPTLEVEL)" "OPENACCO0"
CPPFLAGS += -DMNH_OPENACC -DMNH_GPUDIRECT
# CPPFLAGS += -DMNH_OPENACC -ta=tesla
OPT = $(OPT_BASE) $(OPT_OPENACC) $(OPT_PERF0)
OPT0 = $(OPT_BASE) $(OPT_OPENACC) $(OPT_PERF0)
OPT_NOCB = $(OPT_BASE) $(OPT_OPENACC) $(OPT_PERF0)
CXXFLAGS = -acc -Kieee -Mnofma $(OPT_OPENACC)
OBJS_REPROD= spll_mode_sum_ll.o
$(OBJS_REPROD) : OPT = $(OPT_BASE) $(OPT_OPENACC) $(OPT_PERF0) -Mvect=nosimd -Minfo=all -g
endif
#
ifeq "$(OPTLEVEL)" "OPENACC"
CPPFLAGS += -DMNH_OPENACC -DMNH_GPUDIRECT
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
# CPPFLAGS += -DMNH_OPENACC -ta=tesla
OPT = $(OPT_BASE) $(OPT_OPENACC) $(OPT_PERF2)
OPT0 = $(OPT_BASE) $(OPT_OPENACC) $(OPT_PERF0)
OPT_NOCB = $(OPT_BASE) $(OPT_OPENACC) $(OPT_PERF2)
CXXFLAGS = -acc -Kieee -Mnofma $(OPT_OPENACC)
OBJS_REPROD= spll_mode_sum_ll.o
$(OBJS_REPROD) : OPT = $(OPT_BASE) $(OPT_OPENACC) $(OPT_PERF2) -Mvect=nosimd -Minfo=all -g
OBJS_O1_OPENACC= spll_ice4_tendencies.o spll_turb_ver_thermo_flux.o
$(OBJS_O1_OPENACC) : OPT = $(OPT_BASE) $(OPT_OPENACC) $(OPT_PERF1)
endif
#
ifeq "$(OPTLEVEL)" "OPENACC_PCAST"
CPPFLAGS += -DMNH_OPENACC
# CPPFLAGS += -DMNH_OPENACC -ta=tesla
OPT = $(OPT_BASE) $(OPT_OPENACC_PCAST) $(OPT_PERF2)
OPT0 = $(OPT_BASE) $(OPT_OPENACC_PCAST) $(OPT_PERF0)
OPT_NOCB = $(OPT_BASE) $(OPT_OPENACC_PCAST) $(OPT_PERF2)
CXXFLAGS = -acc -Kieee -Mnofma $(OPT_OPENACC_PCAST)
OBJS_REPROD= spll_mode_sum_ll.o
$(OBJS_REPROD) : OPT = $(OPT_BASE) $(OPT_OPENACC_PCAST) $(OPT_PERF2) -Mvect=nosimd -Minfo=all -g
endif
#
ifeq "$(OPTLEVEL)" "OPENACCDEFONLY"
CPPFLAGS += -DMNH_OPENACC -D_FAKEOPENACC
OPT = $(OPT_BASE) $(OPT_NOOPENACC) $(OPT_PERF2)
OPT0 = $(OPT_BASE) $(OPT_NOOPENACC) $(OPT_PERF0)
OPT_NOCB = $(OPT_BASE) $(OPT_NOOPENACC) $(OPT_PERF2)
CXXFLAGS = -Kieee -Mnofma $(OPT_NOOPENACC)
OBJS_REPROD= spll_mode_sum_ll.o
$(OBJS_REPROD) : OPT = $(OPT_BASE) $(OPT_NOOPENACC) $(OPT_PERF2) -Mvect=nosimd -Minfo=all -g
endif
#
ifeq "$(OPTLEVEL)" "NOOPENACC"
#CPPFLAGS += -D_OPT_LINEARIZED_LOOPS
OPT = $(OPT_BASE) $(OPT_NOOPENACC) $(OPT_PERF2)
OPT0 = $(OPT_BASE) $(OPT_NOOPENACC) $(OPT_PERF0)
OPT_NOCB = $(OPT_BASE) $(OPT_NOOPENACC) $(OPT_PERF2)
CXXFLAGS = -Kieee -Mnofma $(OPT_NOOPENACC)
OBJS_REPROD= spll_mode_sum_ll.o
$(OBJS_REPROD) : OPT = $(OPT_BASE) $(OPT_NOOPENACC) $(OPT_PERF2) -Mvect=nosimd -Minfo=all -g
endif
#
CC = nvc
CXXFLAGS += $(TP)
CFLAGS += $(TP)
FC = nvfortran
ifeq "$(VER_MPI)" "MPIAUTO"
F90 = mpif90
CC = mpicc
CXX = mpicxx
else
F90 = nvfortran
CC = nvc
CXX = nvc++
endif
#
F77FLAGS = $(OPT)
F77 = $(F90)
F90FLAGS = $(OPT)
FX90 = $(F90)
FX90FLAGS = $(OPT)
#
LDFLAGS = -Wl,-warn-once $(OPT)
#
# preprocessing flags
#
#CPP = cpp -P -traditional -Wcomment -D_OPENACC=201711 -imacros MNH_OPENACC_NV_CR.CPP
CPP = mnh_expand -D_OPENACC=201711 -imacros MNH_OPENACC_NV_CR.CPP
WAUTELET Philippe
committed
CPPFLAGS_C = -DLITTLE_endian
CPPFLAGS_SURFEX =
WAUTELET Philippe
committed
CPPFLAGS_SURCOUCHE += -DMNH_COMPILER_NVHPC
CPPFLAGS_RAD =
CPPFLAGS_NEWLFI = -DSWAPIO -DLINUX -DLFI_INT=${LFI_INT}
WAUTELET Philippe
committed
CPPFLAGS_MNH = -DMNH=MNH -DSFX_MNH
CPPFLAGS_MNH += -Uvector -Upixel
CPPFLAGS_MNH += -DMNH_EXPAND -DMNH_EXPAND_OPENACC
#CPPFLAGS_MNH += -imacros MNH_OPENACC_NV_CR.CPP
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
#
# BITREP flags
#
#if MNH_BITREP exists => compile with the BITREP library
MNH_BITREP = YES
ifeq "$(MNH_BITREP)" "YES"
CPPFLAGS_MNH += -DMNH_BITREP
endif
#
# LIBTOOLS flags
#
#if MNH_TOOLS exists => compile the tools
MNH_TOOLS = yes
#
## IOLFI flag
#
#if MNH_IOLFI exists => compile the NEWLFI library (for LFI files)
MNH_IOLFI=yes
#
## COMPRESS flag
#
#if MNH_COMPRESS exists => compile the COMPRESS library (for LFI files)
MNH_COMPRESS=yes
#
## S4PY flag
#
#if MNH_S4PY exists => compile the libs4py library (for epygram)
#MNH_S4PY=no
#
WAUTELET Philippe
committed
## MGSOLVER flag (multigrid solver)
#
#if MNH_MGSOLVER exists => compile the multigrid solver
MNH_MGSOLVER ?= YES
#
## ecCodes or grib_api selection
#MNH_GRIBAPI: if set to no: use ecCodes
# if set to yes: use grib_api (deprecated library)
#
MNH_GRIBAPI=no
#
# Netcdf/HDF5 flags
#
#HDF_CONF= CXXFLAGS=$(TP)
HDF_OPT ?= $(TP)
NETCDF_OPT ?= $(TP)
CDF_CONF="CPP=cpp"
HDF_CONF="CPP=cpp"
#
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
##########################################################
# #
# Source of MESONH PACKAGE Distribution #
# #
##########################################################
#DIR_SURFEX += ARCH_SRC/surfex.MNH-462
OBJS_NOCB += spll_isba.o
#
include Makefile.MESONH.mk
#
##########################################################
# #
# extra VPATH, Compilation flag modification #
# systeme module , etc ... #
# external precompiled module librairie #
# etc ... #
# #
##########################################################
OBJS_O1 += spll_modd_isba_n.o spll_mode_construct_ll.o \
spll_init_surf_atm_n.o spll_mode_scatter_ll.o spll_convert_patch_teb.o \
spll_define_mask_n.o spll_del1dfield_ll.o spll_mode_fm.o spll_mode_gather_ll.o \
spll_convect_updraft.o spll_convect_updraft_shal.o \
spll_mode_dustopt.o spll_mode_saltopt.o \
spll_aeroopt_get.o spll_write_lfifm1_for_diag_supp.o spll_write_lfifm1_for_diag.o spll_write_lfifm_n.o \
#spll_unpack_1d_2d_from2d.o
#spll_pack_isba_patch_n.o
#spll_phys_param_n.o
$(OBJS_O1) : OPT = $(OPT_BASE) $(OPT_PERF1)
OBJS_O0= spll_mode_mppdb.o \
spll_mode_repro_sum.o \
spll_modd_les_n.o \
spll_default_desfm_n.o
#default_desfmn: workaround a nvhpc 21.9 bug (run stall in default_desfmn)
WAUTELET Philippe
committed
# spll_fft55.o spll_fft.o spll_flat_invz.o \
# spll_fast_terms.o
# spll_modd_ch_solver_n.o \
# spll_modd_dummy_gr_field_n.o spll_modd_dyn_n.o
# spll_mode_sum_ll.o
$(OBJS_O0) : OPT = $(OPT_BASE) $(OPT_PERF0)
OBJS_O2= spll_mode_device.o
$(OBJS_O2) : OPT = $(OPT_BASE) $(OPT_OPENACC) $(OPT_PERF2)
ifneq "$(findstring 8,$(LFI_INT))" ""
OBJS_I8=spll_NEWLFI_ALL.o
$(OBJS_I8) : OPT = $(OPT_BASE) $(OPT_PERF2) $(OPT_I8)
endif
ifeq "$(MNH_INT)" "8"
OBJS_I4=spll_modd_netcdf.o
$(OBJS_I4) : OPT = $(OPT_BASE_I4)
endif
SPLL = spll_new