Github user DaveBirdsall commented on a diff in the pull request:
https://github.com/apache/incubator-trafodion/pull/250#discussion_r49110123
--- Diff: core/sql/sqlcomp/nadefaults.cpp ---
@@ -4333,59 +4336,8 @@ void NADefaults::updateSystemParameters(NABoolean
reInit)
case MAX_ESPS_PER_CPU_PER_OP:
{
- // set 2 ESPs per node, as a starting point.
- #define DEFAULT_ESPS_PER_NODE 2
-
- Lng32 numESPsPerNode = DEFAULT_ESPS_PER_NODE;
- Lng32 coresPerNode = 1;
- // Make sure the gpClusterInfo points at an NAClusterLinux
object.
- // In osim simulation mode, the pointer can point at a
NAClusterNSK
- // object, for which the method numTSEsForPOS() is not defined.
- NAClusterInfoLinux* gpLinux =
dynamic_cast<NAClusterInfoLinux*>(gpClusterInfo);
-
- // number of POS TSE
- Lng32 numTSEsPerCluster = gpLinux->numTSEsForPOS();
-
- // cluster nodes
- Lng32 nodesdPerCluster = gpClusterInfo->getTotalNumberOfCPUs();
-
- // TSEs per node
- Lng32 TSEsPerNode = numTSEsPerCluster/nodesdPerCluster;
-
- // cores per node
- coresPerNode = gpClusterInfo->numberOfCpusPerSMP();
-
- // For Linux/nt, we conservatively allocate ESPs per node as
follows
- // - 1 ESP per 2 cpu cores if cores are equal or less than TSEs
- // - 1 ESP per TSE if number of cores is more than double the
TSEs
- // - 1 ESP per 2 TSEs if cores are more than TSEs but less than
double the TSEs
- // - 1 ESP per node. Only possible on NT or workstations
- // - number of cores less than TSEs and there are 1 or 2
cpur cores per node
- // - number of TSEs is less than cpu cores and there 1 or 2
TSEs per node.
- // This case is probable if virtual nodes are used
-
- // TSEsPerNode is 0 for arkcmps started by the seapilot
universal comsumers
- // in this case we only consider cpu cores
- if ((coresPerNode <= TSEsPerNode) || (TSEsPerNode == 0))
- {
- if (coresPerNode > 1)
- numESPsPerNode = DEFAULT_ESPS_PER_NODE;
- }
- else if (coresPerNode > (TSEsPerNode*2))
- {
- numESPsPerNode = TSEsPerNode;
- }
- else if (TSEsPerNode > 1)
- {
- numESPsPerNode = TSEsPerNode/2;
- }
- else // not really needed since numESPsPerNode is set to 1 from
above
- {
- numESPsPerNode = DEFAULT_ESPS_PER_NODE;
- }
-
-
- ftoa_((float)(numESPsPerNode)/(float)(coresPerNode), valuestr);
+ float espsPerCore = computeNumESPsPerCore(FALSE);
--- End diff --
Not sure I understand. We pass FALSE here to computeNumESPsPerCore, and
when I read that method, it goes to the else case (that is, !aggressive). So it
appears "not aggressive" is currently hard-coded?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
