release/1.9: Bug fixes for scotch and crtm-fix, update Ursa site config #1660
There are no files selected for viewing
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -24,3 +24,6 @@ packages: | |
| qt: | ||
| require: | ||
| - '%gcc' | ||
| scotch: | ||
| require: | ||
| - '%gcc' | ||
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -48,7 +48,7 @@ | |
| w3emc@2.10.0, | ||
| nco, | ||
| esmf@8.8.0, | ||
| mapl@2.53.4, | ||
| zlib-ng, | ||
| zstd, | ||
| odc@1.5.2, | ||
|
|
||
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,6 +1,6 @@ | ||
| packages: | ||
| all: | ||
| compiler:: [oneapi@2024.2.1,gcc@11.4.1] | ||
| providers: | ||
| mpi:: [intel-oneapi-mpi@2021.13] | ||
| # Change as appropriate to switch to intel-oneapi-mkl | ||
|
|
@@ -16,22 +16,20 @@ packages: | |
| modules: | ||
| - intel-oneapi-mpi/2021.13.1 | ||
| prefix: /apps/spack-2024-12/linux-rocky9-x86_64/oneapi-2024.2.1/intel-oneapi-mpi-2021.13.1-ss72gbndvat3oz22sa6lhmlbjkeabrn4 | ||
| intel-oneapi-mkl: | ||
| externals: | ||
| - spec: intel-oneapi-mkl@2024.2.1%oneapi@2024.2.1 | ||
| modules: | ||
| - intel-oneapi-mkl/2024.2.1 | ||
| prefix: /apps/spack-2024-12/linux-rocky9-x86_64/gcc-11.4.1/intel-oneapi-mkl-2024.2.1-srqwrbzwo2k7hxuuhlrxttburs5jvlat | ||
| intel-oneapi-runtime: | ||
| externals: | ||
| - spec: intel-oneapi-runtime@2024.2.1%oneapi@2024.2.1 | ||
| prefix: /apps/spack-2024-12/linux-rocky9-x86_64/gcc-11.4.1/intel-oneapi-compilers-2024.2.1-oqhstbmawnrsdw472p4pjsopj547o6xs/compiler/2024.2 | ||
| modules: | ||
| - compiler-rt/2024.2.1 | ||
| gcc-runtime: | ||
| externals: | ||
| - spec: gcc-runtime@11.4.1%gcc@11.4.1 | ||
| prefix: /usr | ||
|
|
||
|
Collaborator
Author
There was a problem hiding this comment. Since this is a new platform, let's just move straight back to Intel MKL instead of OpenBLAS+FFTW with Intel. The GCC build still uses OpenBLAS and FFTW, as intended.
Collaborator
Author
There was a problem hiding this comment. This is what Jessica Meixner tested in ufs-community/ufs-weather-model#2650 (comment)
Collaborator
There was a problem hiding this comment. @climbfuji will going "straight back to Intel MKL instead of OpenBLAS+FFTW with Intel" on ursa but not on other problems cause any inconsistencies? trying to understand if we are going to need to rebuild 1.9.1 entirely on all platforms or if we can hopefully just update scotch...
Collaborator
Author
There was a problem hiding this comment. No, it won't. It's entirely transparent to the users or downstream application. I've asked for this move in general for a very long time. Before we had OpenBLAS with Intel, hpc-stack and jedi-stack used MKL with Intel by default. When spack-stack started, we had problems with getting MKL to work and out of necessity switched to OpenBLAS. In fact, we already switched to MKL on all NRL systems with Intel (and use OpenBLAS with GNU), and nobody noticed.
Collaborator
Author
There was a problem hiding this comment. The spack-stack CI runners are like that, too, by the way. |
||
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Does this mean we are using scotch 7.0.4 or are we moving to 7.0.7? (sorry this is my lack of spack-stack knowledge question)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is still scotch 7.0.4. If it works, then there is no need to update it. The purpose of spack-stack release 1.9.2 is to provide a bug fix / hot fix for the existing 1.9.0 and 1.9.1 releases. Bug fix releases only contain the necessary changes to fix bugs, no changes that aren't necessary.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Do we know if scotch 7.0.4 works on other machines? Or just Ursa... I'm assuming it's going to work on other machines but don't know. Either way WW3 is ready to handle a mix of the two if necessary.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I am pretty sure it will work. We'll install the 1.9.2 release candidate on two machines of your choice with both GNU and oneAPI, and hopefully that will confirm my guess. Ursa + ???