Trevor Grant created MAHOUT-1934:
------------------------------------
Summary: OpenMP jars aren't being picked up in distributed Spark
Key: MAHOUT-1934
URL: https://issues.apache.org/jira/browse/MAHOUT-1934
Project: Mahout
Issue Type: Improvement
Affects Versions: 0.13.0
Reporter: Trevor Grant
Assignee: Andrew Palumbo
Fix For: 0.13.0
When executing the following:
```
import org.apache.mahout.math.decompositions._
drmX.checkpoint()
val (U, V, s) = dspca(drmX, k = 30, q = 1)
```
I see the following:
```
import org.apache.mahout.math.decompositions._
res1: org.apache.mahout.math.drm.CheckpointedDrm[Int] =
org.apache.mahout.sparkbindings.drm.CheckpointedDrmSpark@2e8dd198
[INFO] Creating org.apache.mahout.viennacl.opencl.GPUMMul solver
[WARN] Unable to create class GPUMMul: attempting OpenMP version
[INFO] Creating org.apache.mahout.viennacl.openmp.OMPMMul solver
org.apache.mahout.viennacl.openmp.OMPMMul$
[INFO] Unable to create class OMPMMul: falling back to java version
drmReduxX: org.apache.mahout.math.drm.DrmLike[Int] =
OpMapBlock(OpTimesRightMatrix(org.apache.mahout.s...
```
The code executes fine, but it seems the OpenMP is not getting picked up.
To be clear I build these jars with -Pviennacl-omp
I the issue persists through other operations as well. I am running on a 3
node Yarn cluster/ spark 1.6.3
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
