118 lines
5.8 KiB
CMake
118 lines
5.8 KiB
CMake
###################################################################################################
|
|
# IntelLLVM Compiler
|
|
###################################################################################################
|
|
if (OPENMP)
|
|
set (OPENMP_FLAGS "-fiopenmp")
|
|
endif ()
|
|
|
|
if (OPTIMIZATION STREQUAL "OFF" OR OPTIMIZATION STREQUAL "DEBUG")
|
|
set (OPTIMIZATION_FLAGS "-O0")
|
|
elseif (OPTIMIZATION STREQUAL "DEFENSIVE")
|
|
set (OPTIMIZATION_FLAGS "-O2")
|
|
elseif (OPTIMIZATION STREQUAL "AGGRESSIVE")
|
|
#set (OPTIMIZATION_FLAGS "-ipo -O3 -fp-model fast=2 -xHost") # ifx 2022.0 has problems with YAML types and IPO
|
|
set (OPTIMIZATION_FLAGS "-O3 -fp-model fast=2 -xHost")
|
|
endif ()
|
|
|
|
# -assume std_mod_proc_name (included in -standard-semantics) causes problems if other modules
|
|
# (PETSc, HDF5) are not compiled with this option (https://software.intel.com/en-us/forums/intel-fortran-compiler-for-linux-and-mac-os-x/topic/62172)
|
|
set (STANDARD_CHECK "-stand f18 -assume nostd_mod_proc_name")
|
|
set (LINKER_FLAGS "${LINKER_FLAGS} -shared-intel")
|
|
# Link against shared Intel libraries instead of static ones
|
|
set (LINKER_FLAGS "${LINKER_FLAGS} -shared-intel -fc=ifx")
|
|
# enforce use of ifx for MPI wrapper
|
|
|
|
#------------------------------------------------------------------------------------------------
|
|
# Fine tuning compilation options
|
|
set (COMPILE_FLAGS "${COMPILE_FLAGS} -fpp") # preprocessor, needed for CMake < 3.18
|
|
|
|
set (COMPILE_FLAGS "${COMPILE_FLAGS} -no-ftz")
|
|
# disable flush underflow to zero, will be set if -O[1,2,3]
|
|
|
|
set (COMPILE_FLAGS "${COMPILE_FLAGS} -diag-disable")
|
|
# disables warnings ...
|
|
set (COMPILE_FLAGS "${COMPILE_FLAGS} 5268")
|
|
# ... the text exceeds right hand column allowed on the line (enforced by pre-receive hook)
|
|
set (COMPILE_FLAGS "${COMPILE_FLAGS},7624")
|
|
# ... about deprecated forall (has nice syntax and most likely a performance advantage)
|
|
|
|
set (COMPILE_FLAGS "${COMPILE_FLAGS} -warn")
|
|
# enables warnings ...
|
|
set (COMPILE_FLAGS "${COMPILE_FLAGS} declarations")
|
|
# ... any undeclared names (alternative name: -implicitnone)
|
|
set (COMPILE_FLAGS "${COMPILE_FLAGS},general")
|
|
# ... warning messages and informational messages are issued by the compiler
|
|
set (COMPILE_FLAGS "${COMPILE_FLAGS},usage")
|
|
# ... questionable programming practices
|
|
set (COMPILE_FLAGS "${COMPILE_FLAGS},interfaces")
|
|
# ... checks the interfaces of all SUBROUTINEs called and FUNCTIONs invoked in your compilation against an external set of interface blocks
|
|
set (COMPILE_FLAGS "${COMPILE_FLAGS},ignore_loc")
|
|
# ... %LOC is stripped from an actual argument
|
|
set (COMPILE_FLAGS "${COMPILE_FLAGS},alignments")
|
|
# ... data that is not naturally aligned
|
|
set (COMPILE_FLAGS "${COMPILE_FLAGS},unused")
|
|
# ... declared variables that are never used
|
|
|
|
# Additional options
|
|
# -warn: enables warnings, where
|
|
# truncated_source: Determines whether warnings occur when source exceeds the maximum column width in fixed-format files.
|
|
# (too many warnings because we have comments beyond character 132)
|
|
# uncalled: Determines whether warnings occur when a statement function is never called
|
|
# all:
|
|
# -name as_is: case sensitive Fortran!
|
|
|
|
#------------------------------------------------------------------------------------------------
|
|
# Runtime debugging
|
|
set (DEBUG_FLAGS "${DEBUG_FLAGS} -g")
|
|
# Generate symbolic debugging information in the object file
|
|
|
|
set (DEBUG_FLAGS "${DEBUG_FLAGS} -traceback")
|
|
# Generate extra information in the object file to provide source file traceback information when a severe error occurs at run time
|
|
|
|
set (DEBUG_FLAGS "${DEBUG_FLAGS} -gen-interfaces")
|
|
# Generate an interface block for each routine. http://software.intel.com/en-us/blogs/2012/01/05/doctor-fortran-gets-explicit-again/
|
|
|
|
set (DEBUG_FLAGS "${DEBUG_FLAGS} -fp-stack-check")
|
|
# Generate extra code after every function call to ensure that the floating-point (FP) stack is in the expected state
|
|
|
|
set (DEBUG_FLAGS "${DEBUG_FLAGS} -fp-model strict")
|
|
# Trap uninitalized variables
|
|
|
|
set (DEBUG_FLAGS "${DEBUG_FLAGS} -check" )
|
|
# Checks at runtime ...
|
|
set (DEBUG_FLAGS "${DEBUG_FLAGS} bounds")
|
|
# ... if an array index is too small (<1) or too large!
|
|
set (DEBUG_FLAGS "${DEBUG_FLAGS},format")
|
|
# ... for the data type of an item being formatted for output.
|
|
set (DEBUG_FLAGS "${DEBUG_FLAGS},output_conversion")
|
|
# ... for the fit of data items within a designated format descriptor field.
|
|
set (DEBUG_FLAGS "${DEBUG_FLAGS},pointers")
|
|
# ... for certain disassociated or uninitialized pointers or unallocated allocatable objects.
|
|
set (DEBUG_FLAGS "${DEBUG_FLAGS},uninit")
|
|
# ... for uninitialized variables.
|
|
set (DEBUG_FLAGS "${DEBUG_FLAGS} -fpe-all=0 -ftz")
|
|
# ... capture all floating-point exceptions, need to overwrite -no-ftz
|
|
|
|
set (DEBUG_FLAGS "${DEBUG_FLAGS} -init=arrays,zero,minus_huge,snan")
|
|
# ... initialize logical to false, integer to -huge, float+complex to signaling NaN
|
|
|
|
# disable due to compiler bug https://community.intel.com/t5/Intel-Fortran-Compiler/false-positive-stand-f18-and-IEEE-SELECTED-REAL-KIND/m-p/1227336
|
|
#set (DEBUG_FLAGS "${DEBUG_FLAGS} -warn")
|
|
# enables warnings ...
|
|
#set (DEBUG_FLAGS "${DEBUG_FLAGS} errors")
|
|
# ... warnings are changed to errors
|
|
#set (DEBUG_FLAGS "${DEBUG_FLAGS},stderrors")
|
|
# ... warnings about Fortran standard violations are changed to errors
|
|
|
|
set (DEBUG_FLAGS "${DEBUG_FLAGS} -debug-parameters all")
|
|
# generate debug information for parameters
|
|
|
|
set (DEBUG_FLAGS "${DEBUG_FLAGS} -debug all")
|
|
# generate complete debugging information
|
|
|
|
# Additional options
|
|
# -heap-arrays: Should not be done for OpenMP, but set "ulimit -s unlimited" on shell. Probably it helps also to unlimit other limits
|
|
# -check: Checks at runtime, where
|
|
# arg_temp_created: will cause a lot of warnings because we create a bunch of temporary arrays (performance?)
|
|
# stack:
|