[OpenMP] Sort topology after adding processor group layer. (#83943)
Various behavior around creating affinity masks and detecting uniform
topology depends on the topology being sorted.
resort topology after adding processor group layer to ensure that the
updated topology reflects the newly added processor group info.
Observed that the topology was not sorted correctly on high core count
AMD Epyc Genoa (2 sockets, 96 cores, 2 threads) using NUMA (NPS 2+).
GitOrigin-RevId: ea848d0a6d5c17af3eb1a4e39dc712606ac684f6
diff --git a/runtime/src/kmp_affinity.cpp b/runtime/src/kmp_affinity.cpp
index c3ee4de..048bd17 100644
--- a/runtime/src/kmp_affinity.cpp
+++ b/runtime/src/kmp_affinity.cpp
@@ -327,6 +327,9 @@
KMP_CPU_FREE(mask);
_insert_layer(KMP_HW_PROC_GROUP, ids);
__kmp_free(ids);
+
+ // sort topology after adding proc groups
+ __kmp_topology->sort_ids();
}
#endif
