Recipe changes to combine coreneuron into NEURON#1777
Conversation
|
Thank you for your pull request! Should you want to clear the PR build directory after failures, please use this pipeline. Before running the cleanup pipeline, please ensure that any PR building pipelines have been cancelled or finished running. |
|
The following build jobs of the deployment for this PR failed: |
* list of additional mod files are read from file and those
contain trailing newline char
* comparison done between two list was returning undesired
list of mod files due to leading newline char
|
The following build jobs of the deployment for this PR failed: |
|
The following build jobs of the deployment for this PR failed: |
iomaganaris
left a comment
There was a problem hiding this comment.
Some initial things noted.. I'm wondering also whether we still need to define a libs function in the neuron recipe like we do in the coreneuron recipe to allow dependent libraries to find the proper coreneuron libraries. I haven't looked yet how you handled these. I suppose neuron creates the same libraries' names as before?
I am thinking in oppose way - as neuron doesn't have |
|
To test your PR, use the following on BlueBrain5: unset MODULEPATH
. /gpfs/bbp.cscs.ch/ssd/apps/bsd/pulls/1777/config/modules.sh
module load unstablePlease test the following updated modules:
|
2 similar comments
|
To test your PR, use the following on BlueBrain5: unset MODULEPATH
. /gpfs/bbp.cscs.ch/ssd/apps/bsd/pulls/1777/config/modules.sh
module load unstablePlease test the following updated modules:
|
|
To test your PR, use the following on BlueBrain5: unset MODULEPATH
. /gpfs/bbp.cscs.ch/ssd/apps/bsd/pulls/1777/config/modules.sh
module load unstablePlease test the following updated modules:
|
is used. This will simplify logic in blueconfig CI where we don't have to explicitly specify coreneuron dependency (especially as we have merged neuron and coreneuron)
31d07d1 to
4699d08
Compare
|
To test your PR, use the following on BlueBrain5: unset MODULEPATH
. /gpfs/bbp.cscs.ch/ssd/apps/bsd/pulls/1777/config/modules.sh
module load unstablePlease test the following updated modules:
The following module tests failed: |
…rent name than libcorenrnmech.so as it's now part of NEURON installation"" This reverts commit adc43f4.
|
To test your PR, use the following on BlueBrain5: unset MODULEPATH
. /gpfs/bbp.cscs.ch/ssd/apps/bsd/pulls/1777/config/modules.sh
module load unstablePlease test the following updated modules:
|
|
To test your PR, use the following on BlueBrain5: unset MODULEPATH
. /gpfs/bbp.cscs.ch/ssd/apps/bsd/pulls/1777/config/modules.sh
module load unstablePlease test the following updated modules:
|
1 similar comment
|
To test your PR, use the following on BlueBrain5: unset MODULEPATH
. /gpfs/bbp.cscs.ch/ssd/apps/bsd/pulls/1777/config/modules.sh
module load unstablePlease test the following updated modules:
|
* In #1777 we have made coreneuron optional for upcoming neuron > 8 releases: ``` depends_on("coreneuron+legacy-unit~caliper", when="@:8.99+coreneuron+legacy-unit~caliper") ... ``` * Because of this change, spack is now preferring develop version of neuron when we mix gcc with intel: ``` $ spack spec -I py-bluepyopt%gcc ^neuron%intel Input spec -------------------------------- - py-bluepyopt%gcc - ^neuron%intel Concretized -------------------------------- - py-bluepyopt@1.10.38%gcc@11.2.0+neuron patches=dc37639246b3a3320c840f1c4f635940b8510a896a667d4707c3d355d9c6fb81 arch=linux-rhel7-skylake - ^neuron@develop%intel@2021.4.0+binary~caliper~codegenopt+coreneuron~debug~gpu~interviews~ipo+legacy-fr+legacy-unit+memacs+mod-compatibility+mpi~nmodl~openmp+python+report+rx3d+shared~sympy~sympyopt~tests~unified build_type=RelWithDebInfo model_tests=None patches=708cb04826b394a858069d93e8c08e1e81e914c23e1ef3da0486e8233834ff6c sanitizers=None arch=linux-rhel7-skylake ``` but with single compiler, it uses last release correctly ``` $ spack spec -I py-bluepyopt ^neuron%intel Input spec -------------------------------- - py-bluepyopt - ^neuron%intel Concretized -------------------------------- - py-bluepyopt@1.10.38%intel@2021.4.0+neuron patches=dc37639246b3a3320c840f1c4f635940b8510a896a667d4707c3d355d9c6fb81 arch=linux-rhel7-skylake - ^neuron@8.2.1%intel@2021.4.0+binary~caliper~codegenopt+coreneuron~debug~gpu~interviews~ipo+legacy-fr+legacy-unit+memacs+mod-compatibility+mpi~nmodl~openmp+python+report+rx3d+shared~sympy~sympyopt~tests~unified build_type=RelWithDebInfo model_tests=None patches=708cb04826b394a858069d93e8c08e1e81e914c23e1ef3da0486e8233834ff6c sanitizers=None arch=linux-rhel7-skylake - ^bison@3.8.2%intel@2021.4.0 arch=linux-rhel7-skylake ``` * I tried various spec changes in neuron and coreneuron but don't see any satisfactory change to force spack to use last release than develop. So as part of this commit, pin the last release whenever we are mixing compilers. ANYWAY, this will be removed with #1781.
* In #1777 we have made coreneuron optional for upcoming neuron > 8 releases: ``` depends_on("coreneuron+legacy-unit~caliper", when="@:8.99+coreneuron+legacy-unit~caliper") ... ``` * Because of this change, spack is now preferring develop version of neuron when we mix gcc with intel: ``` $ spack spec -I py-bluepyopt%gcc ^neuron%intel Input spec -------------------------------- - py-bluepyopt%gcc - ^neuron%intel Concretized -------------------------------- - py-bluepyopt@1.10.38%gcc@11.2.0+neuron patches=dc37639246b3a3320c840f1c4f635940b8510a896a667d4707c3d355d9c6fb81 arch=linux-rhel7-skylake - ^neuron@develop%intel@2021.4.0+binary~caliper~codegenopt+coreneuron~debug~gpu~interviews~ipo+legacy-fr+legacy-unit+memacs+mod-compatibility+mpi~nmodl~openmp+python+report+rx3d+shared~sympy~sympyopt~tests~unified build_type=RelWithDebInfo model_tests=None patches=708cb04826b394a858069d93e8c08e1e81e914c23e1ef3da0486e8233834ff6c sanitizers=None arch=linux-rhel7-skylake ``` but with single compiler, it uses last release correctly ``` $ spack spec -I py-bluepyopt ^neuron%intel Input spec -------------------------------- - py-bluepyopt - ^neuron%intel Concretized -------------------------------- - py-bluepyopt@1.10.38%intel@2021.4.0+neuron patches=dc37639246b3a3320c840f1c4f635940b8510a896a667d4707c3d355d9c6fb81 arch=linux-rhel7-skylake - ^neuron@8.2.1%intel@2021.4.0+binary~caliper~codegenopt+coreneuron~debug~gpu~interviews~ipo+legacy-fr+legacy-unit+memacs+mod-compatibility+mpi~nmodl~openmp+python+report+rx3d+shared~sympy~sympyopt~tests~unified build_type=RelWithDebInfo model_tests=None patches=708cb04826b394a858069d93e8c08e1e81e914c23e1ef3da0486e8233834ff6c sanitizers=None arch=linux-rhel7-skylake - ^bison@3.8.2%intel@2021.4.0 arch=linux-rhel7-skylake ``` * I tried various spec changes in neuron and coreneuron but don't see any satisfactory change to force spack to use last release than develop. So as part of this commit, pin the last release whenever we are mixing compilers. ANYWAY, this will be removed with #1781.
* Bring coreneuron recipe options into neuron for coreneuron merge * remove dependency with coreneuron when neuron>=9.0 or develop * coreneuron variant is on by default in neuron * coreneuron mechanism library should always have different name than libcorenrnmech.so as it's now part of NEURON installation * do not add arch specific opt flags with +debug * Automatically bring coreneuron build_type=Debug when neuron+debug is used. This will simplify logic in blueconfig CI where we don't have to explicitly specify coreneuron dependency (especially as we have merged neuron and coreneuron) * Find "right" nrnivmodl-core binary for neurodamus building See relevant PR: neuronsimulator/nrn#2055
CoreNEURON: reduce memory usage in report initialization (#1776) Recipe changes to combine coreneuron into NEURON (#1777) * Bring coreneuron recipe options into neuron for coreneuron merge * remove dependency with coreneuron when neuron>=9.0 or develop * coreneuron variant is on by default in neuron * coreneuron mechanism library should always have different name than libcorenrnmech.so as it's now part of NEURON installation * do not add arch specific opt flags with +debug * Automatically bring coreneuron build_type=Debug when neuron+debug is used. This will simplify logic in blueconfig CI where we don't have to explicitly specify coreneuron dependency (especially as we have merged neuron and coreneuron) * Find "right" nrnivmodl-core binary for neurodamus building See relevant PR: neuronsimulator/nrn#2055 Deploy NEURON's upcoming 9.0a release (#1781) * Updated model-neocortex packages with latest 1.10 tag * add nmodl 0.5.a * remove coreneuron packages Co-authored-by: Ioannis Magkanaris <iomagkanaris@gmail.com> Co-authored-by: Matthias Wolf <matthias.wolf@epfl.ch> Update NEURON version (#1792) - Create NEURON version 9.0.a2 that includes bug fixes [Core]NEURON 8.2.2 (#1802) coreneuron: so long, and thanks for the fish. Coreneuron has been folded into neuRon and should not be referred to as a standalone package any more. Add py-sympy dependency at neuron package level newer spack doesn't seem to propogate python package dependencies (as PYTHONPATH) during build phase. neuron: omitted link flag for Caliper
CoreNEURON: reduce memory usage in report initialization (#1776) Recipe changes to combine coreneuron into NEURON (#1777) * Bring coreneuron recipe options into neuron for coreneuron merge * remove dependency with coreneuron when neuron>=9.0 or develop * coreneuron variant is on by default in neuron * coreneuron mechanism library should always have different name than libcorenrnmech.so as it's now part of NEURON installation * do not add arch specific opt flags with +debug * Automatically bring coreneuron build_type=Debug when neuron+debug is used. This will simplify logic in blueconfig CI where we don't have to explicitly specify coreneuron dependency (especially as we have merged neuron and coreneuron) * Find "right" nrnivmodl-core binary for neurodamus building See relevant PR: neuronsimulator/nrn#2055 Deploy NEURON's upcoming 9.0a release (#1781) * Updated model-neocortex packages with latest 1.10 tag * add nmodl 0.5.a * remove coreneuron packages Co-authored-by: Ioannis Magkanaris <iomagkanaris@gmail.com> Co-authored-by: Matthias Wolf <matthias.wolf@epfl.ch> Update NEURON version (#1792) - Create NEURON version 9.0.a2 that includes bug fixes [Core]NEURON 8.2.2 (#1802) coreneuron: so long, and thanks for the fish. Coreneuron has been folded into neuRon and should not be referred to as a standalone package any more. Add py-sympy dependency at neuron package level newer spack doesn't seem to propogate python package dependencies (as PYTHONPATH) during build phase. neuron: omitted link flag for Caliper
CoreNEURON: reduce memory usage in report initialization (#1776) Recipe changes to combine coreneuron into NEURON (#1777) * Bring coreneuron recipe options into neuron for coreneuron merge * remove dependency with coreneuron when neuron>=9.0 or develop * coreneuron variant is on by default in neuron * coreneuron mechanism library should always have different name than libcorenrnmech.so as it's now part of NEURON installation * do not add arch specific opt flags with +debug * Automatically bring coreneuron build_type=Debug when neuron+debug is used. This will simplify logic in blueconfig CI where we don't have to explicitly specify coreneuron dependency (especially as we have merged neuron and coreneuron) * Find "right" nrnivmodl-core binary for neurodamus building See relevant PR: neuronsimulator/nrn#2055 Deploy NEURON's upcoming 9.0a release (#1781) * Updated model-neocortex packages with latest 1.10 tag * add nmodl 0.5.a * remove coreneuron packages Co-authored-by: Ioannis Magkanaris <iomagkanaris@gmail.com> Co-authored-by: Matthias Wolf <matthias.wolf@epfl.ch> Update NEURON version (#1792) - Create NEURON version 9.0.a2 that includes bug fixes [Core]NEURON 8.2.2 (#1802) coreneuron: so long, and thanks for the fish. Coreneuron has been folded into neuRon and should not be referred to as a standalone package any more. Add py-sympy dependency at neuron package level newer spack doesn't seem to propogate python package dependencies (as PYTHONPATH) during build phase. neuron: omitted link flag for Caliper
Uh oh!
There was an error while loading. Please reload this page.