Post by arya on Jul 19, 2024 12:20:59 GMT
Hey Team,
Our datasets are growing fast, and while that's awesome for discovering new regions of interest, it's also slowing down our processing times.
First up, I've been dreaming of a GPU implementation of GCTA. Surprisingly, I couldn't find any threads on the forum discussing this. Given that many of us now have GPUs with 32GB or even 80GB of memory (lucky us), loading genetic data wouldn't be a memory issue anymore. Plus, the repetitive nature of the tests would be perfect for parallel processing.
Speaking of cool features, how about a batching system? We can already split analyses into chromosomes and SNP lists, but imagine having a feature like PLINK where calculations are further subdivided into batches... Just a thought.
Lastly, is there any way to add a "resume" feature? Our computer cluster and file management system are incredibly unstable, with frequent random maintenance causing shutdowns. This means we often lose our progress when the system goes down. There's nothing more frustrating than coming in on a Monday morning to find all your jobs failed after running for five days...
Voilà... if you have suggestions i'm all for it !!
A tired PhD student in his last year.
Our datasets are growing fast, and while that's awesome for discovering new regions of interest, it's also slowing down our processing times.
First up, I've been dreaming of a GPU implementation of GCTA. Surprisingly, I couldn't find any threads on the forum discussing this. Given that many of us now have GPUs with 32GB or even 80GB of memory (lucky us), loading genetic data wouldn't be a memory issue anymore. Plus, the repetitive nature of the tests would be perfect for parallel processing.
Speaking of cool features, how about a batching system? We can already split analyses into chromosomes and SNP lists, but imagine having a feature like PLINK where calculations are further subdivided into batches... Just a thought.
Lastly, is there any way to add a "resume" feature? Our computer cluster and file management system are incredibly unstable, with frequent random maintenance causing shutdowns. This means we often lose our progress when the system goes down. There's nothing more frustrating than coming in on a Monday morning to find all your jobs failed after running for five days...
Voilà... if you have suggestions i'm all for it !!
A tired PhD student in his last year.