* The Class F problem has been added to seven of the benchmarks
(BT, SP, LU, CG, MG, FT, and EP).
* The Class E problem has been added to the IS benchmark.
* In version 3.4.1, 'the number of processes' option does not apply.
* MPIFC and FC flags were added.
These versions change the install location of CMake files used
by dependents, but most dependents don't seem to look in this
new location.
Co-authored-by: Harmen Stoppels <harmenstoppels@gmail.com>
Co-authored-by: Chris White <white238@llnl.gov>
This command pokes the environment, Python interpreter
and bootstrap store to check if dependencies needed by
Spack are available.
If any are missing, it shows a comprehensible message.
* locks: allow locks to work under high contention
This is a bug found by Harshitha Menon.
The `lock=None` line shouldn't be a release but should be
```
return (lock_type, None)
```
to inform the caller it couldn't get the lock type requested without
disturbing the existing lock object in the database. There were also a
couple of bugs due to taking write locks at the beginning without any
checking or release, and not releasing read locks before requeueing.
This version no longer gives me read upgrade to write errors, even
running 200 instances on one box.
* Change lock in check_deps_status to read, release if not installed,
not sure why this was ever write, but read definitely is more
appropriate here, and the read lock is only held out of the scope if
the package is installed.
* Release read lock before requeueing to reduce chance of livelock, the
timeout that caused the original issue now happens in roughly 3 of 200
workers instead of 199 on average.
With this commit:
```
$ spack env activate --temp
$ spack install zlib
==> All of the packages are already installed
==> Updating view at /tmp/spack-faiirgmt/.spack-env/view
$ spack install zlib
==> All of the packages are already installed
```
Before this PR:
```
$ spack env activate --temp
$ spack install zlib
==> All of the packages are already installed
$ spack install zlib
==> All of the packages are already installed
```
No view was generated
This commit introduces the command
spack module tcl setdefault <package>
similar to the one already available for lmod
Co-authored-by: Massimiliano Culpo <massimiliano.culpo@gmail.com>
* upcxx: Update the UPC++ package to 2021.9.0
* Add the new release, and a missing older one.
* Remove the spack package cruft for supporting the obsolete build system that
was present in older versions that are no longer supported.
* General cleanups.
Support for library versions older than 2020.3.0 is officially retired,
for two reasons:
1. Releases prior to 2020.3.0 had a required dependency on Python 2,
which is [officially EOL](https://www.python.org/doc/sunset-python-2/)
as of Jan 1 2020, and is no longer considered secure.
2. (Most importantly) The UPC++ development team is unable/unwilling to
support releases more than two years old. UPC++ provides robust
backwards-compatibility to earlier releases of UPC++ v1.0, with very
rare well-documented/well-motivated exceptions. Users are strongly
encouraged to update to a current version of UPC++.
NOTE: Most of the lines changed in this commit are simply re-indentation,
and thus might be best reviewed in a diff that ignores whitespace.
* upcxx: Detect Cray XC more explicitly
This change is necessary to prevent false matches occuring on new Cray Shasta
systems, which do not use the aries network but were incorrectly being treated
as a Cray XC + aries platform.
UPC++ has not yet deployed official native support for Cray Shasta, but this
change is sufficient to allow building the portable backends there.
When running `spack install --log-format junit|cdash ...`, install
errors were ignored. This made spack continue building dependents of
failed install, ignoring `--fail-fast`, and exit 0 at the end.
* locks: allow locks to work under high contention
This is a bug found by Harshitha Menon.
The `lock=None` line shouldn't be a release but should be
```
return (lock_type, None)
```
to inform the caller it couldn't get the lock type requested without
disturbing the existing lock object in the database. There were also a
couple of bugs due to taking write locks at the beginning without any
checking or release, and not releasing read locks before requeueing.
This version no longer gives me read upgrade to write errors, even
running 200 instances on one box.
* Change lock in check_deps_status to read, release if not installed,
not sure why this was ever write, but read definitely is more
appropriate here, and the read lock is only held out of the scope if
the package is installed.
* Release read lock before requeueing to reduce chance of livelock, the
timeout that caused the original issue now happens in roughly 3 of 200
workers instead of 199 on average.
