-
Notifications
You must be signed in to change notification settings - Fork 15
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Quasimapping with multiple threads odd behaviour #127
Comments
This process eventually finished after 25 hours k7 (8 thread) results Timer report: For comparison, k5 results (8 threads) Count all reads: 67115296 Timer report: Different number of mapped reads surprised me a bit |
I've restarted a new run, to see if it is reproducible |
Also seen here #117 |
With k11, using the standard Plasmodium PRG
Quasimap 4.7 million 150bp reads with 1 thread takes 187,000 seconds (40 hours). Quasimapping with 8 threads takes 45,000 seconds (12.5 hours). Maybe this can be closed |
@iqbal-lab Was the number of mapped reads consistent between runs? |
Precisely the same! |
@iqbal-lab In a previous comment you showed that the number of mapped reads was erroneously inconsistent between kmer sizes (when using multiple threads?). Do we know if that issue persists? |
I am now running at k13, will confirm |
@iqbal-lab Where can I find the quasimap output directory for the above please? |
|
Tested multi-threading of quasimap by mapping 250,000 reads (so 500,000 with the reverse complements) from (yoda) /nfs/leia/research/iqbal/bletcher/Pf_benchmark/all_reads/original/PG0496-C.bam to (big) pf3k prg. So here multi-threading gives no speedup. All runs produced consistent results for reads mapped:
|
Can you give exact command line and lsf command? |
|
Hello! Nice! Thanks for this! If it is not too much to ask, could you also make an additional plot please? I think one showing the true speed-up and the theoretical best speed-up (see for e.g. https://stackoverflow.com/questions/26514264/plot-speed-up-curve-vs-number-of-openmp-threads-scalability ) can be useful for us to know if we should try to improve multithreading or not! Thanks! Cheers! |
I'm running on a single dedicated server (not shared)
Running with this commit
With k=5, and 8 threads, quasimapping of a fixed fastq to a fixed PRG takes 1 hr 45 mins.
Output showed 67 million reads processed
With k=7 and 8 threads, it is still running 15 hours after starting, output shows it has processed 37 million (last print out was 2 hours ago)
Machine: ebi7-017
Command
gramtools quasimap --gram-directory results/gramk7 --reads fastq/out.fq.gz --max-threads=8 2>> error_qmapk7_thread8 1>> output_qmapk7_thread8
pwd
/tmp/benchmarking
The text was updated successfully, but these errors were encountered: