Update for ncrc-intel.mk needed for C5 #42
Comments
|
More details from @nikizadehgfdl:
Do we want this option, and if so, should it be removed from the
Do we want this additional verbosity all the time? There are some additional verbose flags that are not used normally in the templates: https://github.com/NOAA-GFDL/mkmf/blob/master/templates/ncrc-intel.mk#L169-L173 If users want the additional verbosity all the time, these can be added to the CFLAGS and FFLAGS but probably should be removed from the If users want the additional verbosity rarely, they can be added to CFLAGS_VERBOSE and FLAGS_VERBOSE. |
|
Hi @bensonr, could you provide some feedback on the proposed changes to Specifically, whether And then for the two others that increase verbosity, should they be used all the time, or only if the |
|
I just started compiling on t5 login node. With intel-oneapi/2022.0.2 I get the following warnings So, -qno-opt-dynamic-align is not even supported and there is no point putting it in the c5 template, neither is -sox. |
|
@nikizadehgfdl - this template is for intel classic ifort and not for use with ifx. The -qno-opt-dynamic-align is necessary for run-to-run reproducibility when using vector ISAs. Without this option, the runtime environment is free to change data alignment, which changes what data can be pipelined. |
|
I should also add that any C/C++ options will probably need to change with icx/icpcx due to it being a Clang-based front-end. I think that will make it hard to maintain a single intel.mk file in favor of a classic version and a new oneapi.mk version. |
|
Since What about the options that increase verbosity ( |
- rename ncrc-intel.mk to ncrc-intel-classic.mk for clarity, and create new ncrc-intel-oneapi.mk - For both, change -xsse2 to -march=core-avx2. This changes the ISA from xsse2 to avx2, and provides generic hardware support. It's believed the avx2 ISA can now support run-to-run reproducibility. - For intel-classic, add -qno-opt-dynamic-align, needed for run-to-run reproducibility when using vector ISAs. - For both, add -qopt-report-phase=vec and -qopt-report=2 to CFLAGS and FLAGS when VERBOSE is used - For intel-oneapi, remove -sox from FFLAGS as ifx does not support it
- rename ncrc-intel.mk to ncrc-intel-classic.mk for clarity, and create new ncrc-intel-oneapi.mk - For both, change -xsse2 to -march=core-avx2. This changes the ISA from xsse2 to avx2, and provides generic hardware support. It's believed the avx2 ISA can now support run-to-run reproducibility. - For intel-classic, add -qno-opt-dynamic-align, needed for run-to-run reproducibility when using vector ISAs. - For both, add -qopt-report-phase=vec and -qopt-report=2 to CFLAGS and FLAGS when VERBOSE is used - For intel-oneapi, remove -sox from FFLAGS as ifx does not support it
as C5's environment contains its own gettid Linux system call
C5 will use different Intel templates, but the C3/C4 Intel template should remain the same to avoid answer changes.
… classic, and "production" or ifort/ifx - For all 3, change -xsse2 to -march=core-avx2. This changes the ISA from xsse2 to avx2, and provides generic hardware support. It's believed the avx2 ISA can now support run-to-run reproducibility. - For all 3, add -qopt-report-phase=vec and -qopt-report=2 to CFLAGS and FLAGS when VERBOSE is used - For intel-classic and intel, add -qno-opt-dynamic-align, needed for run-to-run reproducibility when using vector ISAs. - Also for intel-classic and intel, remove -ftrapuv from CFLAGS_DEBUG as icx (clang) does not support it - For intel-oneapi, remove -sox from FFLAGS as ifx does not support it
ISA to FFLAGS as it is not a supported icx option
as C5's environment contains its own gettid Linux system call available in the GNU standard library. The intel-classic environment does not have the gettid system function defined. More details: NOAA-GFDL/FMS#276 The error one sees is: src/FMS/affinity/affinity.c:55:14: error: static declaration of 'gettid' follows non-static declaration static pid_t gettid(void) ^ /usr/include/bits/unistd_ext.h:34:16: note: previous declaration is here extern __pid_t gettid (void) __THROW; ^ 1 error generated.
…repro/debug Leave prod unchanged for now
based on the number of cores available. This setting is undesired and potentially dangerous for batch jobs, as it assumes the job can use all cores available. Instead, for batch jobs the paralellism should be the number of cores allocated to the job. (Updates to FRE do this.) For interactive jobs, users can set MAKEFLAGS themselves (e.g. in their shell init files) or as the make -j option.
1. GNU Compiler Collection, not GNU 2. nvhpc instead of pgi
…and remove unsupported "-h nosecond_underscore" option
|
done, in #44 |
C5's environment contains its own gettid Linux system call available in the GNU standard library. More details: NOAA-GFDL/FMS#276 The error one sees is: src/FMS/affinity/affinity.c:55:14: error: static declaration of 'gettid' follows non-static declaration static pid_t gettid(void) ^ /usr/include/bits/unistd_ext.h:34:16: note: previous declaration is here extern __pid_t gettid (void) __THROW; ^ 1 error generated.
C5's environment contains its own gettid Linux system call available in the GNU standard library More details: NOAA-GFDL/FMS#276
#42 For gaea C5 gcc make template, set HAVE_GETTID macro
…AVE_GETTID be added to the CPPDEFS.
From @nikizadehgfdl, we are planning to have a single Intel compiler template that works on both Intel and non-Intel hardware.
In the past, we had switched to the xsse2 instruction set to support run-to-run reproduciblity, and it is believed that this is no longer needed for recent Intel compilers.
The update needed is the ISA. Existing:
ISA = -xsse2Proposed:
ISA = -march=core-avx2 -qopt-report-phase=vec -qopt-report=2 -qno-opt-dynamic-alignNote: this is two changes: 1) to avx2, and 2) generic hardware support. The first change only would be:
ISA = -xCORE-AVX2 -qopt-report-phase=vec -qopt-report=2 -qno-opt-dynamic-alignThe text was updated successfully, but these errors were encountered: