The branch HPSJAVA-494 is now updated for sequential read. I have tested it with several evio files. Both the EvioToLcio and the monitoring app work fine with these changes.

LcioToEvio can be forced back to Memory Mapped mode using the -M switch.

The two modes of reading are equally fast, give or take a second, however, the memory foot print of the Memory Mapped version is significantly larger, requiring > 3GB to run the code, 3.44 GB versus 1.51 GB sequentially. On the farm machines at JLab, it will have a big advantage to run the smaller memory footprint.

Benchmark: LcioToEvio running "DataQualityRecon" (which has a memory leak :-)

(time java -cp distribution/target/hps-distribution-3.3.1-SNAPSHOT-bin.jar  org.hps.evio.EvioToLcio  -d HPS-EngRun2015-2mm-v1  -L SEVERE -x DataQualityRecon.lcsim -DoutputFile=tmp   /data/HPS/data/hps_005184.evio.0 )

Sequential:        time = 2m20.002s user, 0m4.167s system  -- Memory:  1.51 GB

Memory Map:    time = 2m15.889s user, 0m8.047s system  -- Memory = 3.44 GB

I tested this with a *local* compiled copy of jevio. When I remove this copy from the .m2 directory and try to down load a jevio-4.4.6-SNAPSHOT, Maven seems to not be able to retrieve the jar. Perhaps no surprise if this this jar isn’t there?