G-BLASTN 1.1 Released

[OpenHero]

G-BLASTN v1.1 released.

http://comp.hkbu.edu.hk/~chxw/software/G-BLASTN.html

G-BLASTN is a GPU-accelerated nucleotide alignment tool based on the widely used NCBI-BLAST. G-BLASTN can produce exactly the same results as NCBI-BLAST, and it also has very similar user commands. It also supports a pipeline mode, which can fully use the GPU and CPU resources when handling a batch of medium to large sized queries. Currently, G_BLASTN supports the blastn and megablast modes of NCBI-BLAST. The discontiguous megablast mode is not supported yet.

G-BLASTN is free software and you can browse/download the source code at https://sourceforge.net/projects/gblastn/ or https://github.com/OpenHero/gblastn .

Binary codes of G-BLASTNA are also avialble for the following platforms:
64bit CentOS 6.4: https://github.com/OpenHero/gblastn/...d/v1.1/gblastn
64bit Windows 7: http://sourceforge.net/projects/gbla...n.exe/download

[rhinoceros]

I wonder if the alleged 7X speedup of NVIDIA GTX780 over Intel Core i7-3820 has anything to do with how they set up parallelization for gnblastn vs plain blastn. Based on the supplementary.pdf from http://www.comp.hkbu.edu.hk/~chxw/software/G-BLASTN.htm, I would say yes (if it was set properly, "speedup 8" would be more-or-less exactly twice as fast as "speedup 4"). Anyway, I'm sure there's more to it than that. Looking forward to reading the article. A few GPUs and a good CPU might make a rather decent bioinformatics workstation in the not so distant future

[OpenHero]

I wonder if the alleged 7X speedup of NVIDIA GTX780 over Intel Core i7-3820 has anything to do with how they set up parallelization for gnblastn vs plain blastn. Based on the supplementary.pdf from http://www.comp.hkbu.edu.hk/~chxw/software/G-BLASTN.htm, I would say yes (if it was set properly, "speedup 8" would be more-or-less exactly twice as fast as "speedup 4"). Anyway, I'm sure there's more to it than that. Looking forward to reading the article. A few GPUs and a good CPU might make a rather decent bioinformatics workstation in the not so distant future

Thanks for your interest.
There are 4 main steps in blastn.
1.Prepare the hash table with mask data.
2.Scan the hits in the database and ungapped extention. And the -num_threads command only useful in this step ( the multi-threads only work in this step).
3.Trace back the result in the database.
4.Print the result.

In our work, we have not only speedup the step 2, but also changed some of other parts, you can download the source code in the url.
If there is any question please tell me.

[rhinoceros]

Thanks for your interest.
There are 4 main steps in blastn.
1.Prepare the hash table with mask data.
2.Scan the hits in the database and ungapped extention. And the -num_threads command only useful in this step ( the multi-threads only work in this step).
3.Trace back the result in the database.
4.Print the result.

In our work, we have not only speedup the step 2, but also changed some of other parts, you can download the source code in the url.
If there is any question please tell me.

Hello,

I'm just saying that (assuming 4 cores and sufficient RAM):

blastn -query seqs.fasta -num_threads 4

is a lot slower than:

cat seqs.fasta | parallel --block 100k --recstart '>' --pipe blastn -num_threads 1

Because, as you pointed out, num_threads only parallelizes blast partially. So, perhaps some (how much) of the speed gain you're seeing is due to this..

[OpenHero]

Hello,

I'm just saying that (assuming 4 cores and sufficient RAM):

blastn -query seqs.fasta -num_threads 4

is a lot slower than:

cat seqs.fasta | parallel --block 100k --recstart '>' --pipe blastn -num_threads 1

Because, as you pointed out, num_threads only parallelizes blast partially. So, perhaps some (how much) of the speed gain you're seeing is due to this..

You can try our application and compare them. Our application also support pipeline directly.