new version 3.3 of FFTW. Compiled with gfortran but only for multicore support (single core library is still the old version 3.2.2). Now either POSIX threads or OMP can be used by choosing the corresponding library file

Some further polishing (besides of new description for FFTW) of the makefile

code/include/fftw3.f

@@ -1,4 +1,3 @@
-! original file taken from the FFTW package ver. 3.2.2 see http://fftw.org
       INTEGER FFTW_R2HC
       PARAMETER (FFTW_R2HC=0)
       INTEGER FFTW_HC2R

code/makefile

@@ -7,19 +7,18 @@
 # Uses openmp to parallelise the material subroutines (set number of cores with "export DAMASK_NUM_THREADS=n" to n)
 # Uses linux threads to parallelise fftw3 (should also be possible with openmp)
 # Besides of the f90 files written at MPIE, the two library files of fftw3 "libfftw3_threads.a" "libfftw3.a" are also needed
-# Install fftw3 (v3.2.2 is tested) with "./configure --enable-threads --enable-float"  and "make", "make install" is not needed
+# Install fftw3 (v3.3 is tested) with "./configure --enable-threads --enable-sse2" and "make"; "make install" is not needed
-# as long as the two library files are copied to the source code directory.
+# as long as the two library files "libfftw3_threads.a" "libfftw3.a" are copied to the code/include directory.
 
 # OPTIONS = standard (alternative): meaning
 #-------------------------------------------------------------
 # PRECISION = double (single): floating point precision
 # F90 = ifort (gfortran): compiler, choose Intel or GNU
-# VERSION = 10 (12): version of Intel compiler. More aggressive optimization if VERSION =12
+# PORTABLE = TRUE (FALSE): decision, if executable is optimized for the machine on which it was built. Until now only for ifort
-# PORTABLE = TRUE (FALSE): decision, if executable is optimized for the machine on which it was built
+# OPTIMIZATION = DEFENSIVE (OFF,AGGRESSIVE,ULTRA): Optimization mode, O0, O2, O3
-# OPTIMIZATION = DEFENSIVE (OFF,AGGRESSIVE): Optimization mode, O0, O2, O3
 # OPENMP = TRUE (FALSE): OpenMP multiprocessor support
-# ACML = OFF (ON): link with AMD math core library (v. 4.4 need to be installed)
 # PREFIX: specifie an arbitrary prefix
+# SUFFIX: specife an arbitrary suffix, e.g
 # COMPILERNAME = overwrite name of Compiler, e.g. using mpich-g90 instead of ifort
 
 # Here are some usefull debugging switches. Switch on by uncommenting last line:
@@ -31,57 +30,57 @@ DEBUG1 =-check bounds -g
 DEBUG2 =-check arg_temp_created
 #check from time to time
 DEBUG3 =-fp-stack-check -g -traceback -gen-interfaces -warn interfaces
-
+#should not be done for OpenMP, but set "ulimit -s unlimited" on shell. Problably it helps also to unlimit other limits
-#should not be done for OpenMP, but set "ulimit -s unlimited" on shell. Porblably it helps to also unlimit other limits
 DEBUG4 =-heap-arrays
 
-#SUFFIX =$(DEBUG1) $(DEBUG2) $(DEBUG3) 
+#checks for standard
-#SUFFIX =$(DEBUG1) $(DEBUG3) 
+DEBUG5 =stand std03/std95
 
+#SUFFIX =$(DEBUG1) $(DEBUG2) $(DEBUG3) 
+
+#BLAS for OPENMP=OFF
+BLAS_ifort =-L /opt/acml4.4.0/ifort64/lib -lacml
+BLAS_gfortran =
 
 ifeq ($(F90), )
 F90 =ifort
 endif
 
+
 ifeq ($(OPTIMIZATION), )
 OPTIMIZATION =DEFENSIVE
 endif
 MAXOPTI =$(OPTIMIZATION)
 
-ifeq ($(F90), ifort)
+ifeq ($(OPTIMIZATION),ULTRA)
-
+MAXOPTI=AGGRESSIVE
-ifeq ($(PORTABLE), FALSE)
-PORTABLE_SWITCH =-xHost
-endif
-ifneq ($(VERSION), 12)
-ifeq ($(OPTIMIZATION), AGGRESSIVE)
-MAXOPTI =DEFENSIVE
-endif
 endif
 
+ifeq ($(OPTIMIZATION),AGGRESSIVE)
+MAXOPTI=DEFENSIVE
+endif
+
+ifeq ($(PORTABLE),FALSE)
+PORTABLE_SWITCH =-msse3
 endif
 
 ifneq ($(OPENMP), OFF)
 OPENMP_FLAG_ifort =-openmp -openmp-report0 -parallel
 OPENMP_FLAG_gfortran =-fopenmp
+BLAS_ifort =-L /opt/acml4.4.0/ifort64_mp/lib -lacml_mp
+BLAS_gfortran =
 OPENMP =ON
 endif
 
-ifeq ($(ACML), ON)
-BLAS_ON_ifort =-L /opt/acml4.4.0/ifort64_mp/lib -lacml_mp
-BLAS_OFF_ifort =-L /opt/acml4.4.0/ifort64/lib -lacml
-BLAS_ON_gfortran =
-BLAS_ON_gfortran =
-endif
-
 OPTIMIZATION_OFF_ifort =-O0
 OPTIMIZATION_OFF_gfortran =-O0
 OPTIMIZATION_DEFENSIVE_ifort =-O2
 OPTIMIZATION_DEFENSIVE_gfortran =-O2
-OPTIMIZATION_AGGRESSIVE_ifort =-O3 $(PORTABLE_SWITCH) -ip
+OPTIMIZATION_AGGRESSIVE_ifort =-O3 $(PORTABLE_SWITCH) -ip -static -fp-model fast=2 -no-prec-div
-OPTIMIZATION_AGGRESSIVE_gfortran =-O3
+OPTIMIZATION_ULTRA_ifort =-O3 $(PORTABLE_SWITCH) -ip -static -fp-model fast=2 -no-prec-div
+OPTIMIZATION_AGGRESSIVE_gfortran =-O3 -march=opteron -ffast-math -funroll-loops -ftree-vectorize -ftree-loop-linear $(PORTABLE_SWITCH)
 
-COMPILE_OPTIONS_ifort =-fpp -diag-disable 8291,8290 
+COMPILE_OPTIONS_ifort =-fpp -diag-disable 8291,8290
 COMPILE_OPTIONS_gfortran =-xf95-cpp-input -ffree-line-length-none
 
 COMPILE =${OPENMP_FLAG_${F90}} ${COMPILE_OPTIONS_${F90}} ${OPTIMIZATION_${OPTIMIZATION}_${F90}} -c
@@ -94,13 +93,13 @@ endif
 ifeq ($(PRECISION),single)
 DAMASK_spectral_single.exe: DAMASK_spectral_single.o CPFEM.a
 	$(PREFIX)	$(COMPILERNAME) ${OPENMP_FLAG_${F90}} -o DAMASK_spectral_single.exe DAMASK_spectral_single.o CPFEM.a include/libfftw3f_threads.a include/libfftw3f.a\
-  constitutive.a advanced.a basics.a -lpthread ${BLAS_${OPENMP}_${F90}}
+  constitutive.a advanced.a basics.a -lpthread ${BLAS_${F90}}
 DAMASK_spectral_single.o: DAMASK_spectral_single.f90 CPFEM.o 
 	$(PREFIX)	$(COMPILERNAME) $(COMPILE_MAXOPTI) DAMASK_spectral_single.f90 $(SUFFIX)
 else
 DAMASK_spectral.exe: DAMASK_spectral.o CPFEM.a
 	$(PREFIX)	$(COMPILERNAME) ${OPENMP_FLAG_${F90}} -o DAMASK_spectral.exe DAMASK_spectral.o CPFEM.a include/libfftw3_threads.a include/libfftw3.a\
-  constitutive.a advanced.a basics.a -lpthread ${BLAS_${OPENMP}_${F90}}
+  constitutive.a advanced.a basics.a -lpthread ${BLAS_${F90}}
 DAMASK_spectral.o: DAMASK_spectral.f90 CPFEM.o 
 	$(PREFIX)	$(COMPILERNAME) $(COMPILE_MAXOPTI) DAMASK_spectral.f90 $(SUFFIX)
 endif
@@ -184,6 +183,8 @@ else
 DAMASK_spectral_interface.o: DAMASK_spectral_interface.f90 prec.o
 	$(PREFIX)	$(COMPILERNAME) $(COMPILE) DAMASK_spectral_interface.f90 $(SUFFIX)
 prec.o: prec.f90
+	@echo $(OPTIMIZATION)
+	@echo $(MAXOPTI)
 	$(PREFIX)	$(COMPILERNAME) $(COMPILE) prec.f90 $(SUFFIX)
 endif