Hi there,
> I was wondering if there was a way to estimate topup processing time? I am topup'ing two acquisition sets b=0+256 directions A->P and b=0+256 directions P->A. I was working on it on my laptop and set it running on Thursday and it is still processing. I understand it is doing quite a bit since the resolution is 2 isotropic with 75 slices and many directions that all have an acquisition with 180 phase blips.
>
> It looks like it is using about 9.5 GB of virtual memory on my Macbook Pro. I figure that I can kill this process and move it to my Mac Pro that has 16 GB of RAM. I am guessing this will still take a bit of time to process, but I was wondering if someone might have an estimate based on their experience.
>
> I didn't see that topup utilizes Grid Engine. Is that correct? If it doesn't, would it be possible to split the job and run, say, 7 or 8 instances of topup since each of the volumes, b=0 and all 256 directions, have their own reverse phase encoded volume and then recombine them when complete? Or, would one instance of topup use all the available RAM and I'd be shooting myself in the foot, writing to virtual memory again.
we don't recommend doing what you are doing here. Topup is used to estimate the distortions caused by the object (head) and those are expected to be the same for all directions. So we recommend using just a handful of scans to estimate those. There is another set of distortions that are caused by the diffusion gradients (eddy currents) and for those you obviously need all directions. Those we recommend you use our new tool called eddy to estimate. Eddy will take as one of its inputs the field you estimated with topup, so it will be applied to all the scans. In addition eddy will calculate the unique eddy current distortions and movements for all scans.
eddy is also quite processing intense, but at least less so that topup. In addition eddy is written using OpenMP so it is able to take advantage of multi-processor machines. We regularly run it on data sets with several hundred volumes.
Jesper
|