* Add more checksummed versions
* Remove all versions not supported by autotools build method
* Add old build system for older versions
* Add suggested changes
* add amgx package
* add amgx variants for mkl and magma support
* fix typo in cmake option
* flake8 fix formatting
* Apply suggestions from code review - use mkl virtual provider
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Apply suggestions from code review - fix copypasta
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
* Add additional variants to netcdf-fortran
* Fix duplicate variant
* Clean up variants based on review feedback
* Addtional variant changes
* Convert jna variant to single line
* Fix proper version constraints for jna variant
* cp2k: prettify arch-file, call pkg-config directly
this allows to re-use the arch-file without having to load the complete
Spack environment, for example after a dev-build
* cp2k: use consistency check instead of blas lib enum
this makes using other BLAS/LAPACK implementations possible without
explicitly adding support for them
* cp2k: add basic support for Cray and XL Compilers, correct Intel fp mode
* cp2k: add myself as maintainer
* cp2k: use "master" to denote the git version
* cp2k: use spack_cc/fc/cxx when possible, set CXX explicitly
* cp2k: set __MKL when using the MKL, not just the Intel compiler
* cp2k: drop self. when referencing spec where possible
* cp2k: add forgotten elpa+openmp dep
* cp2k: set C++14 for recent versions
* Add new version: "Abseil LTS branch, Feb 2020, Patch 1"
* Build shared libraries by default with new version
* Older versions do not support building shared libraries
* Update bat and make the url dynamic
- Now depending on the version it will calculate
the url
- This also fixes a weird issue that was reported
on Darwin, back when I reported that Rust wasn't
linking properly on Darwin #15887 on the comment
by hartzell i was also experiencing this issue
* Remove unnecessary stuff
- Removes the need for LLVM
- Removes the need for version calculation.
* Keep the versions on 1 line
* Pass flake8 tests
* Add new package exa
* Format and fix a silly typo
* Fix SHA256 SUM and make URL calculation dynamic
* Remove unnecessary URL calculation
* Update package.py
* Keep the version on 1 line
* Pass flake8 checks
* Add ShengBTE
Adding a new package; ShengBTE. I tried adding it a MakefilePackage, and use build_directory = 'Src', but it was as if the build_directory gets ignored and make complains about target not found. and using the make funtion here instead of os.system, I get errors that config.f90 is not found which is already available under Src as well.
* more enhancmenets
fix lint; use mkl spec; use build_directory variable
* one more fix
* Use Makefile template
* Update var/spack/repos/builtin/packages/shengbte/package.py
use mkl instead of intel-mkl as a dependency
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update var/spack/repos/builtin/packages/shengbte/package.py
update recipe as suggested by reviewer
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* enhance recipe
remove white space; changes as suggested by reviewer
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
* Fixed SSL pathing for older versions of Python (i.e. @:3.6.999).
* Fixed an issue where the 'python~ssl' variant wasn't properly being respected.
* Improved the '~ssl' patch by making it functional instead of diff-based (enables 3.X.Y patches).
* Fixed comment formatting to satisfy 'flake8' style requirements.
On Power architectures cuDNN will install in a target directory. This
sets cuDNN_ROOT to point to the subdirectory to help dependents use
this install.
The singularity info should actually suggest where you might find the info about the post-install steps.
Co-authored-by: george.hartzell <george.hartzell@sana.com>
Building the `py-jupyter` stack on macOS with AppleClang breaks on
the `py-qtconsole` -> `py-qtconsole` -> `qt +opengl` package build
environment setup with:
```
==> Error: AttributeError: Query of package 'mesa' for 'libs' failed
...
==> Error: Failed to install qt due to ChildError: AttributeError: Query of package 'mesa' for 'libs' failed
```
This tries to add more library targets build by `mesa` to avoid this.
* LLVM: Python Dependency
Effort to expose the linked python library when building LLVM.
This might fix the forward propagation of libintl that comes
with the static python library build on darwin.
* LLDB Py: Remove Ignored Old Flags
Changed in LLVM 10.0+
This allows horovod to be built with frameworks=pytorch,tensorflow.
I tracked down the crash I observed in #15719, where loading torch
before tensorflow would cause a crash in:
google::protobuf::internal::(anonymous
namespace)::InitSCC_DFS(google::protobuf::internal::SCCInfoBase*)
The solution is to make tensorflow compile against the protobuf
version Spack provides, instead of allowing it to use it's own.
It's likely we'll want to go after some of the others
that are listed in third_party/systemlibs/syslibs_configure.bzl
in the future.
