Hi Kirk, thanks for your reply.
Unfortunately I'm made a very shameful mistake. The println at the end of the loop always gets called; even if nothing is found. And I made the content so that the query would not find anything since I'm currently not yet able to return a result. So the initial loop never has a println and the second loop always has a println. After I added a guard to the println in the second loop, the performance of both loops is exactly the same. So it wasn't a JIT issue after all, but just plain stupidity. On Wednesday, December 27, 2017 at 11:21:05 AM UTC+2, Kirk Pepperdine wrote: > > Hi, > > > Can you make count a field and rerun? Not sure the for loop is being > unrolled in either case as the index is a long. I’ve not checked unrolling > but using a long can cause the JIT to miss optimizations that it would > normally apply if an int was used instead. You might want to see what > JITWatch can tell you. > > — Kirk > > On Dec 27, 2017, at 10:09 AM, Peter Veentjer <[email protected] > <javascript:>> wrote: > > As part of an experiment, I'm working on querying large volumes of data > which is stored offheap. > > The content of each record is stored in a chunk of offheap memory. So > instead of having an array of object references, it is an array of records > (no pointer chasing). > > My confusion is about some code I'm generating based on the query content. > There are 2 flavors; one flavor where I'm printing if I found something and > the other flavor increments a local long variable and print this at the end > of the loop. > > The strange thing is that the first one (printing when the correct entry > is found), is 15x faster than the one where I'm increasing the local > counter. > > So this is the first: > > import java.util.*; > public class FullTableScan_e872b2bd8f274cc18b37ac2a0e3df2ed extends > com.hazelcast.simplemap.impl.FullTableScan{ > public void run(){ > long offset=slabPointer; > for(long l=0;l<recordIndex;l++){ > if((unsafe.getInt(offset+12)==10000) && > (unsafe.getBoolean(null,offset+16)==true)){ > System.out.println("found"); > } > offset+=recordDataSize; > } > > } > public void init(Map<String, Object> binding){ > } > } > > > And this is the second: > import java.util.*; > public class FullTableScan_e872b2bd8f274cc18b37ac2a0e3df2ed extends > com.hazelcast.simplemap.impl.FullTableScan{ > public void run(){ > long offset=slabPointer; > long count = 0; > for(long l=0;l<recordIndex;l++){ > if((unsafe.getInt(offset+12)==10000) && > (unsafe.getBoolean(null,offset+16)==true)){ > count++; > } > offset+=recordDataSize; > } > System.out.println("count:"+count); > } > public void init(Map<String, Object> binding){ > } > } > > What could be the reason of this huge performance difference? It isn't a > warmup problem since it was running for 5 minutes. Could there be some data > dependency with the second loop that prevents the loop to be unrolled? I > should analyze the assembler; perhaps this will shed some light on the > situation. > > -- > You received this message because you are subscribed to the Google Groups > "mechanical-sympathy" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected] <javascript:>. > For more options, visit https://groups.google.com/d/optout. > > > -- You received this message because you are subscribed to the Google Groups "mechanical-sympathy" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
