From 9843799e7fc5c8f53f4260713884b2a413382183 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Tiziano=20M=C3=BCller?= Date: Thu, 7 Feb 2019 16:59:12 +0100 Subject: [PATCH] Drop -ffast-math everywhere Check the discussion and latest failures with LAPACK using gcc-8.2.0 and `-ffast-math` here: https://github.com/cp2k/cp2k/pull/193 Just. Not. Worth. It. --- arch/CRAY-XC30-gfortran-cuda.psmp | 2 +- arch/CRAY-XC30-gfortran-mkl.psmp | 2 +- arch/CRAY-XC30-gfortran.psmp | 2 +- arch/CRAY-XK7-gfortran-cuda.psmp | 2 +- arch/Darwin-IntelMacintosh-gfortran.popt | 2 +- arch/Darwin-IntelMacintosh-gfortran.psmp | 2 +- arch/Darwin-IntelMacintosh-gfortran.sopt | 2 +- arch/Darwin-IntelMacintosh-gfortran.ssmp | 2 +- arch/FreeBSD-gfortran.popt | 2 +- arch/FreeBSD-gfortran.psmp | 2 +- arch/FreeBSD-gfortran.sopt | 2 +- arch/FreeBSD-gfortran.ssmp | 2 +- arch/Linux-x86-64-cuda-ma.psmp | 2 +- arch/Linux-x86-64-dbcsr-cuda.popt | 2 +- arch/Linux-x86-64-gfortran.warn | 2 +- arch/Linux-x86-64-gfortran_mkl_elpa.popt | 2 +- arch/Linux-x86-64-gfortran_opencl_nvidia.sopt | 2 +- arch/Linux-x86-64-ma.sopt | 2 +- arch/Linux-x86-64-mingw64-minimal.sopt | 4 ++-- arch/PR59336.sdbg | 2 +- arch/test_tsan.sdbg | 2 +- tools/autotune_grid/config.in | 2 +- tools/toolchain/install_cp2k_toolchain.sh | 16 ++++++++-------- tools/toolchain/scripts/install_reflapack.sh | 10 +++++----- tools/toolchain/scripts/install_scalapack.sh | 2 +- 25 files changed, 37 insertions(+), 37 deletions(-) diff --git a/arch/CRAY-XC30-gfortran-cuda.psmp b/arch/CRAY-XC30-gfortran-cuda.psmp index afa8dd574a..501bb1696e 100644 --- a/arch/CRAY-XC30-gfortran-cuda.psmp +++ b/arch/CRAY-XC30-gfortran-cuda.psmp @@ -15,7 +15,7 @@ CPPFLAGS = CXXFLAGS = -O3 -I${CUDA_PATH}/include -std=c++11 DFLAGS = -D__FFTW3 -D__parallel -D__SCALAPACK -D__HAS_smm_dnn -D__ACC -D__DBCSR_ACC CFLAGS = $(DFLAGS) -FCFLAGS = $(DFLAGS) -O3 -fopenmp -mavx -funroll-loops -ffast-math -ftree-vectorize \ +FCFLAGS = $(DFLAGS) -O3 -fopenmp -mavx -funroll-loops -ftree-vectorize \ -ffree-form -ffree-line-length-512 LDFLAGS = $(FCFLAGS) NVFLAGS = $(DFLAGS) -O3 -arch sm_35 -Xcompiler='-fopenmp' --std=c++11 diff --git a/arch/CRAY-XC30-gfortran-mkl.psmp b/arch/CRAY-XC30-gfortran-mkl.psmp index 39402dc622..1807f33658 100644 --- a/arch/CRAY-XC30-gfortran-mkl.psmp +++ b/arch/CRAY-XC30-gfortran-mkl.psmp @@ -12,7 +12,7 @@ AR = ar -r CPPFLAGS = DFLAGS = -D__MKL -D__FFTW3 -D__parallel -D__SCALAPACK -D__HAS_smm_dnn CFLAGS = $(DFLAGS) -FCFLAGS = $(DFLAGS) -O3 -fopenmp -mavx -funroll-loops -ffast-math -ftree-vectorize \ +FCFLAGS = $(DFLAGS) -O3 -fopenmp -mavx -funroll-loops -ftree-vectorize \ -ffree-form -ffree-line-length-512 \ -I${MKLROOT}/include -I${MKLROOT}/include/fftw LDFLAGS = $(FCFLAGS) diff --git a/arch/CRAY-XC30-gfortran.psmp b/arch/CRAY-XC30-gfortran.psmp index 7ec2e28307..e29827f42a 100644 --- a/arch/CRAY-XC30-gfortran.psmp +++ b/arch/CRAY-XC30-gfortran.psmp @@ -10,7 +10,7 @@ AR = ar -r CPPFLAGS = DFLAGS = -D__FFTW3 -D__parallel -D__SCALAPACK -D__HAS_smm_dnn CFLAGS = $(DFLAGS) -FCFLAGS = $(DFLAGS) -O3 -fopenmp -mavx -funroll-loops -ffast-math -ftree-vectorize \ +FCFLAGS = $(DFLAGS) -O3 -fopenmp -mavx -funroll-loops -ftree-vectorize \ -ffree-form -ffree-line-length-512 LDFLAGS = $(FCFLAGS) LIBS = -lfftw3 -lfftw3_threads diff --git a/arch/CRAY-XK7-gfortran-cuda.psmp b/arch/CRAY-XK7-gfortran-cuda.psmp index eba48e547f..e39f533d59 100644 --- a/arch/CRAY-XK7-gfortran-cuda.psmp +++ b/arch/CRAY-XK7-gfortran-cuda.psmp @@ -16,7 +16,7 @@ CPPFLAGS = CXXFLAGS = -O3 -I${CUDA_PATH}/include -std=c++11 DFLAGS = -D__GEMINI -D__FFTW3 -D__parallel -D__SCALAPACK -D__HAS_smm_dnn -D__ACC -D__DBCSR_ACC CFLAGS = $(DFLAGS) -FCFLAGS = $(DFLAGS) -O3 -fopenmp -march=bdver1 -funroll-loops -ffast-math -ftree-vectorize \ +FCFLAGS = $(DFLAGS) -O3 -fopenmp -march=bdver1 -funroll-loops -ftree-vectorize \ -ffree-form -ffree-line-length-512 LDFLAGS = $(FCFLAGS) NVFLAGS = $(DFLAGS) -O3 -arch sm_35 -Xcompiler='-fopenmp' --std=c++11 diff --git a/arch/Darwin-IntelMacintosh-gfortran.popt b/arch/Darwin-IntelMacintosh-gfortran.popt index b55e916ecc..ee46b42d75 100644 --- a/arch/Darwin-IntelMacintosh-gfortran.popt +++ b/arch/Darwin-IntelMacintosh-gfortran.popt @@ -10,7 +10,7 @@ AR = ar -r RANLIB = ranlib DFLAGS = -D__MPI_VERSION=3 -D__parallel -D__SCALAPACK \ -D__NO_STATM_ACCESS -D__ACCELERATE -D__LIBXC -D__LIBINT -FCFLAGS = -I $(LIBXC_INCLUDE_DIR) -O2 -ffast-math -funroll-loops \ +FCFLAGS = -I $(LIBXC_INCLUDE_DIR) -O2 -funroll-loops \ -ftree-vectorize -ffree-form $(DFLAGS) LDFLAGS = $(FCFLAGS) LIBS = -framework Accelerate -lscalapack \ diff --git a/arch/Darwin-IntelMacintosh-gfortran.psmp b/arch/Darwin-IntelMacintosh-gfortran.psmp index 8cac0a46f2..cfb559dbe9 100644 --- a/arch/Darwin-IntelMacintosh-gfortran.psmp +++ b/arch/Darwin-IntelMacintosh-gfortran.psmp @@ -10,7 +10,7 @@ AR = ar -r RANLIB = ranlib DFLAGS = -D__MPI_VERSION=3 -D__parallel -D__SCALAPACK \ -D__NO_STATM_ACCESS -D__ACCELERATE -D__LIBXC -D__LIBINT -FCFLAGS = -I $(LIBXC_INCLUDE_DIR) -fopenmp -O2 -ffast-math \ +FCFLAGS = -I $(LIBXC_INCLUDE_DIR) -fopenmp -O2 \ -funroll-loops -ftree-vectorize -ffree-form $(DFLAGS) LDFLAGS = $(FCFLAGS) LIBS = -framework Accelerate -lscalapack \ diff --git a/arch/Darwin-IntelMacintosh-gfortran.sopt b/arch/Darwin-IntelMacintosh-gfortran.sopt index e149cde3ec..ecfe46e916 100644 --- a/arch/Darwin-IntelMacintosh-gfortran.sopt +++ b/arch/Darwin-IntelMacintosh-gfortran.sopt @@ -8,7 +8,7 @@ LD = gfortran AR = ar -r RANLIB = ranlib DFLAGS = -D__NO_STATM_ACCESS -D__ACCELERATE -D__LIBXC -D__LIBINT -FCFLAGS = -I $(LIBXC_INCLUDE_DIR) -O2 -ffast-math -funroll-loops \ +FCFLAGS = -I $(LIBXC_INCLUDE_DIR) -O2 -funroll-loops \ -ftree-vectorize -ffree-form $(DFLAGS) LDFLAGS = $(FCFLAGS) LIBS = -framework Accelerate \ diff --git a/arch/Darwin-IntelMacintosh-gfortran.ssmp b/arch/Darwin-IntelMacintosh-gfortran.ssmp index e58c5c97c2..9c9a53bd77 100644 --- a/arch/Darwin-IntelMacintosh-gfortran.ssmp +++ b/arch/Darwin-IntelMacintosh-gfortran.ssmp @@ -8,7 +8,7 @@ LD = gfortran AR = ar -r RANLIB = ranlib DFLAGS = -D__NO_STATM_ACCESS -D__ACCELERATE -D__LIBXC -D__LIBINT -FCFLAGS = -I $(LIBXC_INCLUDE_DIR) -fopenmp -ffast-math -funroll-loops \ +FCFLAGS = -I $(LIBXC_INCLUDE_DIR) -fopenmp -funroll-loops \ -ftree-vectorize -ffree-form -O2 $(DFLAGS) LDFLAGS = $(FCFLAGS) LIBS = -framework Accelerate \ diff --git a/arch/FreeBSD-gfortran.popt b/arch/FreeBSD-gfortran.popt index c82aa2f4a7..0611ece72e 100644 --- a/arch/FreeBSD-gfortran.popt +++ b/arch/FreeBSD-gfortran.popt @@ -5,6 +5,6 @@ LD = mpif90 AR = ar -r CPPFLAGS = DFLAGS = -D__FFTW3 -D__NO_STATM_ACCESS -D__MPI_VERSION=2 -D__parallel -D__SCALAPACK -FCFLAGS = -O2 -ffast-math -funroll-loops -ftree-vectorize -march=native -ffree-form $(DFLAGS) +FCFLAGS = -O2 -funroll-loops -ftree-vectorize -march=native -ffree-form $(DFLAGS) LDFLAGS = $(FCFLAGS) LIBS = -llapack -lblas -lscalapack -lfftw3 -lmpich diff --git a/arch/FreeBSD-gfortran.psmp b/arch/FreeBSD-gfortran.psmp index 1e99fe0b49..34c6ebd1e4 100644 --- a/arch/FreeBSD-gfortran.psmp +++ b/arch/FreeBSD-gfortran.psmp @@ -5,6 +5,6 @@ LD = mpif90 AR = ar -r CPPFLAGS = DFLAGS = -D__FFTW3 -D__NO_STATM_ACCESS -D__MPI_VERSION=2 -D__parallel -D__SCALAPACK -FCFLAGS = -O2 -fopenmp -ffast-math -funroll-loops -ftree-vectorize -march=native -ffree-form $(DFLAGS) +FCFLAGS = -O2 -fopenmp -funroll-loops -ftree-vectorize -march=native -ffree-form $(DFLAGS) LDFLAGS = $(FCFLAGS) LIBS = -llapack -lblas -lscalapack -lfftw3 -lfftw3_omp -lmpich diff --git a/arch/FreeBSD-gfortran.sopt b/arch/FreeBSD-gfortran.sopt index e2f951095d..e6bf71aaa8 100644 --- a/arch/FreeBSD-gfortran.sopt +++ b/arch/FreeBSD-gfortran.sopt @@ -5,7 +5,7 @@ LD = gfortran AR = ar -r CPPFLAGS = DFLAGS = -D__FFTW3 -D__NO_STATM_ACCESS -FCFLAGS = -O2 -ffast-math -funroll-loops -ftree-vectorize -march=native -ffree-form $(DFLAGS) +FCFLAGS = -O2 -funroll-loops -ftree-vectorize -march=native -ffree-form $(DFLAGS) LDFLAGS = $(FCFLAGS) LIBS = -llapack -lblas -lfftw3 diff --git a/arch/FreeBSD-gfortran.ssmp b/arch/FreeBSD-gfortran.ssmp index 3c9f742091..9a75842c31 100644 --- a/arch/FreeBSD-gfortran.ssmp +++ b/arch/FreeBSD-gfortran.ssmp @@ -5,7 +5,7 @@ LD = gfortran AR = ar -r CPPFLAGS = DFLAGS = -D__FFTW3 -D__NO_STATM_ACCESS -FCFLAGS = -O2 -fopenmp -ffast-math -funroll-loops -ftree-vectorize -march=native -ffree-form $(DFLAGS) +FCFLAGS = -O2 -fopenmp -funroll-loops -ftree-vectorize -march=native -ffree-form $(DFLAGS) LDFLAGS = $(FCFLAGS) LIBS = -llapack -lblas -lfftw3 -lfftw3_omp diff --git a/arch/Linux-x86-64-cuda-ma.psmp b/arch/Linux-x86-64-cuda-ma.psmp index a328e801a7..b93a802fa9 100644 --- a/arch/Linux-x86-64-cuda-ma.psmp +++ b/arch/Linux-x86-64-cuda-ma.psmp @@ -20,7 +20,7 @@ GPUVER = K20X CPPFLAGS = CXXFLAGS = -O3 -I${CUDA_PATH}/include -std=c++11 DFLAGS = -D__GFORTRAN -D__parallel -D__SCALAPACK -D__BLACS -D__FFTSG -D__LIBINT -D__ACC -D__DBCSR_ACC -D__HAS_smm_dnn -D__HWLOC -GFLAGS = -g -fopenmp -O3 -ffast-math $(DFLAGS) +GFLAGS = -g -fopenmp -O3 $(DFLAGS) FCFLAGS = -ffree-form -fcray-pointer $(GFLAGS) CFLAGS = $(GFLAGS) -I$(HWLOC_BASE)/include LDFLAGS = $(FCFLAGS) -mkl=sequential diff --git a/arch/Linux-x86-64-dbcsr-cuda.popt b/arch/Linux-x86-64-dbcsr-cuda.popt index 01a7be7823..4ca7e38df3 100644 --- a/arch/Linux-x86-64-dbcsr-cuda.popt +++ b/arch/Linux-x86-64-dbcsr-cuda.popt @@ -12,7 +12,7 @@ GPUVER = K20X CPPFLAGS = CXXFLAGS = -O3 -I${CUDA_PATH}/include -std=c++11 DFLAGS = -D__parallel -D__SCALAPACK -D__ACC -D__DBCSR_ACC -FCFLAGS = -g -O3 -ffast-math -ffree-form $(DFLAGS) -I$(GFORTRAN_INC) +FCFLAGS = -g -O3 -ffree-form $(DFLAGS) -I$(GFORTRAN_INC) LDFLAGS = $(FCFLAGS) NVFLAGS = $(DFLAGS) -g -O3 -arch sm_35 -Xcompiler='-fopenmp' --std=c++11 diff --git a/arch/Linux-x86-64-gfortran.warn b/arch/Linux-x86-64-gfortran.warn index 9fe72c4e6b..f209b1092c 100644 --- a/arch/Linux-x86-64-gfortran.warn +++ b/arch/Linux-x86-64-gfortran.warn @@ -46,7 +46,7 @@ WFLAGS = -pedantic\ DFLAGS = -D__LIBINT -D__FFTW3 -D__LIBXC2 -D__LIBINT_MAX_AM=6 -D__LIBDERIV_MAX_AM1=5 -D__ACC -D__DBCSR_ACC -D__PW_CUDA -D__parallel -D__SCALAPACK -D__LIBPEXSI -D__MPI_VERSION=3 -FCFLAGS = -fopenmp -I$(CP2KINSTALLDIR)/include -std=f2008 -fimplicit-none -ffree-form -fno-omit-frame-pointer -g -O1 -O3 -march=native -ffast-math $(PROFOPT) $(DFLAGS) $(WFLAGS) +FCFLAGS = -fopenmp -I$(CP2KINSTALLDIR)/include -std=f2008 -fimplicit-none -ffree-form -fno-omit-frame-pointer -g -O1 -O3 -march=native $(PROFOPT) $(DFLAGS) $(WFLAGS) LDFLAGS = -L$(CP2KINSTALLDIR)/lib/ -L/usr/local/cuda/lib64 $(FCFLAGS) NVFLAGS = $(DFLAGS) -g -O2 -arch sm_35 -arch sm_35 -Xcompiler='-fopenmp' --std=c++11 CFLAGS = $(DFLAGS) -I$(CP2KINSTALLDIR)/include -fno-omit-frame-pointer -g -O1 diff --git a/arch/Linux-x86-64-gfortran_mkl_elpa.popt b/arch/Linux-x86-64-gfortran_mkl_elpa.popt index fcffd33da8..3b6b8e4c25 100644 --- a/arch/Linux-x86-64-gfortran_mkl_elpa.popt +++ b/arch/Linux-x86-64-gfortran_mkl_elpa.popt @@ -10,7 +10,7 @@ LD = mpif90 AR = ar -r CPPFLAGS = DFLAGS = -D__parallel -D__SCALAPACK -D__LIBINT -D__LIBXC2 -D__FFTW3 -D__ELPA=201308 -FCFLAGS = -O3 -ffast-math -funroll-loops -ftree-vectorize -march=native -ffree-form $(DFLAGS) -g -I${FFTW3_PATH}/include -I${ELPA_INC}/modules -I${ELPA_INC}/elpa +FCFLAGS = -O3 -funroll-loops -ftree-vectorize -march=native -ffree-form $(DFLAGS) -g -I${FFTW3_PATH}/include -I${ELPA_INC}/modules -I${ELPA_INC}/elpa LDFLAGS = $(FCFLAGS) -L$(LIBINT_PATH) -L${FFTW3_PATH}/lib64 -L$(INTEL_MKL_LIB) -L$(LIBXC_PATH)/lib64 LIBS = \ -lmkl_scalapack_lp64 -lmkl_blacs_openmpi_lp64 -lmkl_gf_lp64 -lmkl_sequential -lmkl_core \ diff --git a/arch/Linux-x86-64-gfortran_opencl_nvidia.sopt b/arch/Linux-x86-64-gfortran_opencl_nvidia.sopt index d71f2629e0..22544c99c6 100644 --- a/arch/Linux-x86-64-gfortran_opencl_nvidia.sopt +++ b/arch/Linux-x86-64-gfortran_opencl_nvidia.sopt @@ -42,7 +42,7 @@ DFLAGS = -D__FFTW3 \ -D__HAS_ISO_C_BINDING -AFLAGS = $(DFLAGS) -O3 -mtune=native -ffast-math -funroll-loops -ftree-vectorize +AFLAGS = $(DFLAGS) -O3 -mtune=native -funroll-loops -ftree-vectorize CPPFLAGS = $(DFLAGS) -traditional -C CFLAGS = $(AFLAGS) FCFLAGS = $(AFLAGS) -ffree-form -ffree-line-length-none diff --git a/arch/Linux-x86-64-ma.sopt b/arch/Linux-x86-64-ma.sopt index edab494634..d3d5ba51d6 100644 --- a/arch/Linux-x86-64-ma.sopt +++ b/arch/Linux-x86-64-ma.sopt @@ -17,7 +17,7 @@ LD = $(FC) AR = ar -r CPPFLAGS = DFLAGS = -D__LIBINT -D__HWLOC -GFLAGS = -g -O3 -ffast-math $(DFLAGS) +GFLAGS = -g -O3 $(DFLAGS) FCFLAGS = -ffree-form -fcray-pointer $(GFLAGS) CFLAGS = $(GFLAGS) -I$(HWLOC_BASE)/include LDFLAGS = $(FCFLAGS) -mkl=sequential diff --git a/arch/Linux-x86-64-mingw64-minimal.sopt b/arch/Linux-x86-64-mingw64-minimal.sopt index e58719c295..29ed0dbbe5 100644 --- a/arch/Linux-x86-64-mingw64-minimal.sopt +++ b/arch/Linux-x86-64-mingw64-minimal.sopt @@ -17,8 +17,8 @@ LD = x86_64-w64-mingw32-gfortran AR = x86_64-w64-mingw32-ar -r DFLAGS = -D__NO_STATM_ACCESS -D__NO_IPI_DRIVER -D__MINGW CPPFLAGS = -P -traditional $(DFLAGS) -CFLAGS = $(DFLAGS) -O2 -ffast-math -FCFLAGS = $(DFLAGS) -O2 -ffast-math -ffree-form -ffree-line-length-none \ +CFLAGS = $(DFLAGS) -O2 +FCFLAGS = $(DFLAGS) -O2 -ffree-form -ffree-line-length-none \ -ftree-vectorize -funroll-loops -std=f2008 LDFLAGS = $(FCFLAGS) -static LIBS = $(OPENBLAS_LIBPATH)/libopenblas.a diff --git a/arch/PR59336.sdbg b/arch/PR59336.sdbg index 9412689d75..18777a0c2f 100644 --- a/arch/PR59336.sdbg +++ b/arch/PR59336.sdbg @@ -9,7 +9,7 @@ AR = gcc-ar -r CPPFLAGS = DFLAGS = -D__FFTW3 -FCFLAGS = -fno-prefetch-loop-arrays -O3 -march=native -funroll-loops -ffast-math -ffree-form $(DFLAGS) +FCFLAGS = -fno-prefetch-loop-arrays -O3 -march=native -funroll-loops -ffree-form $(DFLAGS) LDFLAGS = $(FCFLAGS) LIBS = -llapack -lblas diff --git a/arch/test_tsan.sdbg b/arch/test_tsan.sdbg index 56636c8d02..012aff3355 100644 --- a/arch/test_tsan.sdbg +++ b/arch/test_tsan.sdbg @@ -12,7 +12,7 @@ AR = ar -r CPPFLAGS = DFLAGS = -D__LIBINT -FCFLAGS = -gdwarf-3 -fPIE -pie -fsanitize=thread -fno-omit-frame-pointer -fopenmp -O0 -march=native -ffast-math -ffree-form $(DFLAGS) +FCFLAGS = -gdwarf-3 -fPIE -pie -fsanitize=thread -fno-omit-frame-pointer -fopenmp -O0 -march=native -ffree-form $(DFLAGS) LDFLAGS = $(FCFLAGS) -L/data/vjoost/libint/sanitize/install/lib/ LIBS = -llapack -lblas -lderiv -lint -lstdc++ diff --git a/tools/autotune_grid/config.in b/tools/autotune_grid/config.in index 1774d5ac90..7f8ffbbda0 100644 --- a/tools/autotune_grid/config.in +++ b/tools/autotune_grid/config.in @@ -7,7 +7,7 @@ Nrun=3 # #### Gfortran flags FC_comp="ftn -ffree-form -pipe" -FCFLAGS_OPT=" -O3 -ffast-math -funroll-loops -ftree-vectorize -fno-omit-frame-pointer -g -march=ivybridge -fgcse-sm -fgcse-las -fmerge-all-constants " +FCFLAGS_OPT=" -O3 -funroll-loops -ftree-vectorize -fno-omit-frame-pointer -g -march=ivybridge -fgcse-sm -fgcse-las -fmerge-all-constants " FCFLAGS_NATIVE="-march=native" # Cray flags diff --git a/tools/toolchain/install_cp2k_toolchain.sh b/tools/toolchain/install_cp2k_toolchain.sh index 3b87290162..e3edd90b37 100755 --- a/tools/toolchain/install_cp2k_toolchain.sh +++ b/tools/toolchain/install_cp2k_toolchain.sh @@ -922,12 +922,12 @@ done # setup compiler flags, leading to nice stack traces on crashes but # still optimised -CFLAGS="-O2 -ftree-vectorize -g -fno-omit-frame-pointer -march=native -ffast-math $TSANFLAGS" -FFLAGS="-O2 -ftree-vectorize -g -fno-omit-frame-pointer -march=native -ffast-math $TSANFLAGS" -F77FLAGS="-O2 -ftree-vectorize -g -fno-omit-frame-pointer -march=native -ffast-math $TSANFLAGS" -F90FLAGS="-O2 -ftree-vectorize -g -fno-omit-frame-pointer -march=native -ffast-math $TSANFLAGS" -FCFLAGS="-O2 -ftree-vectorize -g -fno-omit-frame-pointer -march=native -ffast-math $TSANFLAGS" -CXXFLAGS="-O2 -ftree-vectorize -g -fno-omit-frame-pointer -march=native -ffast-math $TSANFLAGS" +CFLAGS="-O2 -ftree-vectorize -g -fno-omit-frame-pointer -march=native $TSANFLAGS" +FFLAGS="-O2 -ftree-vectorize -g -fno-omit-frame-pointer -march=native $TSANFLAGS" +F77FLAGS="-O2 -ftree-vectorize -g -fno-omit-frame-pointer -march=native $TSANFLAGS" +F90FLAGS="-O2 -ftree-vectorize -g -fno-omit-frame-pointer -march=native $TSANFLAGS" +FCFLAGS="-O2 -ftree-vectorize -g -fno-omit-frame-pointer -march=native $TSANFLAGS" +CXXFLAGS="-O2 -ftree-vectorize -g -fno-omit-frame-pointer -march=native $TSANFLAGS" export CFLAGS=$(allowed_gcc_flags $CFLAGS) export FFLAGS=$(allowed_gfortran_flags $FFLAGS) @@ -1035,12 +1035,12 @@ LD_arch="IF_MPI(${MPIFC}|${FC})" # we always want good line information and backtraces BASEFLAGS="-march=native -fno-omit-frame-pointer -g ${TSANFLAGS}" -OPT_FLAGS="-O3 -funroll-loops -ffast-math" +OPT_FLAGS="-O3 -funroll-loops" NOOPT_FLAGS="-O1" # those flags that do not influence code generation are used always, the others if debug FCDEB_FLAGS="-ffree-form -std=f2008 -fimplicit-none" -FCDEB_FLAGS_DEBUG="-fsanitize=leak -fcheck=all -ffpe-trap=invalid,zero,overflow -finit-derived -finit-real=snan -finit-integer=-42 -fno-fast-math -Werror=realloc-lhs-all -finline-matmul-limit=0" +FCDEB_FLAGS_DEBUG="-fsanitize=leak -fcheck=all -ffpe-trap=invalid,zero,overflow -finit-derived -finit-real=snan -finit-integer=-42 -Werror=realloc-lhs-all -finline-matmul-limit=0" # code coverage generation flags COVERAGE_FLAGS="-O1 -coverage -fkeep-static-functions" diff --git a/tools/toolchain/scripts/install_reflapack.sh b/tools/toolchain/scripts/install_reflapack.sh index 75c0b97e3e..cb4eb6a075 100755 --- a/tools/toolchain/scripts/install_reflapack.sh +++ b/tools/toolchain/scripts/install_reflapack.sh @@ -37,14 +37,14 @@ case "$with_reflapack" in cat < make.inc SHELL = /bin/sh FORTRAN = $FC -OPTS = $FFLAGS -frecursive -fno-fast-math -DRVOPTS = $FFLAGS -frecursive -fno-fast-math -NOOPT = $FFLAGS -O0 -frecursive -fno-fast-math +OPTS = $FFLAGS -frecursive +DRVOPTS = $FFLAGS -frecursive +NOOPT = $FFLAGS -O0 -frecursive LOADER = $FC -LOADOPTS = $FFLAGS -Wl,--enable-new-dtags -fno-fast-math +LOADOPTS = $FFLAGS -Wl,--enable-new-dtags TIMER = INT_ETIME CC = $CC -CFLAGS = $CFLAGS -fno-fast-math +CFLAGS = $CFLAGS ARCH = ar ARCHFLAGS= cr RANLIB = ranlib diff --git a/tools/toolchain/scripts/install_scalapack.sh b/tools/toolchain/scripts/install_scalapack.sh index 60ae27979b..f429494f0d 100755 --- a/tools/toolchain/scripts/install_scalapack.sh +++ b/tools/toolchain/scripts/install_scalapack.sh @@ -39,7 +39,7 @@ case "$with_scalapack" in CDEFS = -DAdd_ FC = ${MPIFC} CC = ${MPICC} -NOOPT = ${FFLAGS} -O0 -fno-fast-math +NOOPT = ${FFLAGS} -O0 FCFLAGS = ${FFLAGS} ${MATH_CFLAGS} CCFLAGS = ${CFLAGS} ${MATH_CFLAGS} FCLOADER = \$(FC)