From 430c5793c5f0e039e06f856acbc62eef42a1641c Mon Sep 17 00:00:00 2001 From: Thomas Sibley Date: Thu, 21 Aug 2014 15:03:57 -0700 Subject: [PATCH] MarkDuplicates: Comment out optical duplicate tracking On some inputs, MarkDuplicates seems to take orders of magnitude longer with a runtime out of proportion to the input size. Inspecting the output of jstack -F seems to show the Java thread stuck in findOpticalDuplicates(). strace showed almost 100% of syscall time spent in futex(). The process would sometimes succeed after a while and sometimes lead to a JVM crash. Since this tracking is just metrics bookkeeping and doesn't affect the resulting sam/bam, just disable it! This cut runtime from 100-170 minutes to 2-3 minutes for one such "weird" input file containing only ~950k reads. --- src/java/picard/sam/MarkDuplicates.java | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/java/picard/sam/MarkDuplicates.java b/src/java/picard/sam/MarkDuplicates.java index c625905..a9cf0a7 100644 --- a/src/java/picard/sam/MarkDuplicates.java +++ b/src/java/picard/sam/MarkDuplicates.java @@ -689,7 +689,7 @@ public class MarkDuplicates extends AbstractDuplicateFindingAlgorithm { } } - trackOpticalDuplicates(list); + //trackOpticalDuplicates(list); } /** -- 1.8.4