Merge pull request #153 from casparvl/fix_mem_config

Added / updated memory for various systems in MiB units
EESSI · Jun 13, 2024 · 8cadf28 · 8cadf28
2 parents 46db269 + 0a886c1
commit 8cadf28
Show file tree

Hide file tree

Showing 4 changed files with 12 additions and 5 deletions.
diff --git a/config/aws_mc.py b/config/aws_mc.py
@@ -105,6 +105,11 @@
         # steps inherit environment. It doesn't hurt to define this even if srun is not used
         'export SLURM_EXPORT_ENV=ALL'
     ],
+    'extras': {
+        # Node types have somewhat varying amounts of memory, but we'll make it easy on ourselves
+        # All should _at least_ have this amount (30GB * 1E9 / (1024*1024) = 28610 MiB)
+        'mem_per_node': 28610
+    },
 }
 for system in site_configuration['systems']:
     for partition in system['partitions']:

diff --git a/config/it4i_karolina.py b/config/it4i_karolina.py
@@ -62,7 +62,7 @@
                     'extras': {
                         # Make sure to round down, otherwise a job might ask for more mem than is available
                         # per node
-                        'mem_per_node': 219.345 * 1024  # in MiB
+                        'mem_per_node': 235520  # in MiB
                     },
                     'descr': 'CPU Universal Compute Nodes, see https://docs.it4i.cz/karolina/hardware-overview/'
                 },

diff --git a/config/izum_vega.py b/config/izum_vega.py
@@ -62,7 +62,9 @@
                     'extras': {
                         # Make sure to round down, otherwise a job might ask for more mem than is available
                         # per node
-                        'mem_per_node': 238.418 * 1024  # in MiB
+                        # NB: Vega's MaxMemPerNode is set to 256000, but this MUST be a MB/MiB units mistake
+                        # Most likely, it is 256 GB, so 256*1E9/(1024*1024) MiB
+                        'mem_per_node': 244140  # in MiB
                     },
                     'descr': 'CPU partition Standard, see https://en-doc.vega.izum.si/architecture/'
                 },

diff --git a/config/surf_snellius.py b/config/surf_snellius.py
@@ -56,7 +56,7 @@
                     'extras': {
                         # Make sure to round down, otherwise a job might ask for more mem than is available
                         # per node
-                        'mem_per_node': 213.623 * 1024  # in MiB
+                        'mem_per_node': 229376  # in MiB
                     },
                     'descr': 'AMD Rome CPU partition with native EESSI stack'
                 },
@@ -80,7 +80,7 @@
                     'extras': {
                         # Make sure to round down, otherwise a job might ask for more mem than is available
                         # per node
-                        'mem_per_node': 320.434 * 1024  # in MiB
+                        'mem_per_node': 344064  # in MiB
                     },
                     'descr': 'AMD Genoa CPU partition with native EESSI stack'
                 },
@@ -117,7 +117,7 @@
                         GPU_VENDOR: GPU_VENDORS[NVIDIA],
                         # Make sure to round down, otherwise a job might ask for more mem than is available
                         # per node
-                        'mem_per_node': 457.763 * 1024  # in MiB
+                        'mem_per_node': 491520  # in MiB
                     },
                     'descr': 'Nvidia A100 GPU partition with native EESSI stack'
                 },