[Openmp-commits] [openmp] r360890 - Fix hwloc topology traversal code unable to handle situation where L2 cache is common for the packages

[Paul Osmialowski via Openmp-commits]

Author: pawosm01
Date: Thu May 16 06:16:24 2019
New Revision: 360890

URL: http://llvm.org/viewvc/llvm-project?rev=360890&view=rev
Log:
Fix hwloc topology traversal code unable to handle situation where L2 cache is common for the packages

Currently cores within package that share the same L2 cache are grouped together.
The current logic behind this assumes that the L2 cache is always at deeper
(or the same) level than the package itself. In case when L2 cache is common
for all packages (and the packages are at deeper level than L2 cache) the whole of
the further topology discovery fails to find any computational units resulting in
following assertion:

Assertion failure at kmp_affinity.cpp(715): nActiveThreads == __kmp_avail_proc.
OMP: Error #13: Assertion failure at kmp_affinity.cpp(715).

This patch adds a bit of a logic that prevents such situation from occurring.

Differential Revision: https://reviews.llvm.org/D61796


Modified:
    openmp/trunk/runtime/src/kmp_affinity.cpp

Modified: openmp/trunk/runtime/src/kmp_affinity.cpp
URL: http://llvm.org/viewvc/llvm-project/openmp/trunk/runtime/src/kmp_affinity.cpp?rev=360890&r1=360889&r2=360890&view=diff
==============================================================================
--- openmp/trunk/runtime/src/kmp_affinity.cpp (original)
+++ openmp/trunk/runtime/src/kmp_affinity.cpp Thu May 16 06:16:24 2019
@@ -530,7 +530,7 @@ static int __kmp_hwloc_process_obj_core_
 static int __kmp_hwloc_check_numa() {
   hwloc_topology_t &tp = __kmp_hwloc_topology;
   hwloc_obj_t hT, hC, hL, hN, hS; // hwloc objects (pointers to)
-  int depth;
+  int depth, l2cache_depth, package_depth;
 
   // Get some PU
   hT = hwloc_get_obj_by_type(tp, HWLOC_OBJ_PU, 0);
@@ -548,8 +548,10 @@ static int __kmp_hwloc_check_numa() {
     }
   }
 
+  package_depth = hwloc_get_type_depth(tp, HWLOC_OBJ_PACKAGE);
+  l2cache_depth = hwloc_get_cache_type_depth(tp, 2, HWLOC_OBJ_CACHE_UNIFIED);
   // check tile, get object by depth because of multiple caches possible
-  depth = hwloc_get_cache_type_depth(tp, 2, HWLOC_OBJ_CACHE_UNIFIED);
+  depth = (l2cache_depth < package_depth) ? package_depth : l2cache_depth;
   hL = hwloc_get_ancestor_obj_by_depth(tp, depth, hT);
   hC = NULL; // not used, but reset it here just in case
   if (hL != NULL &&