Re: Re: ORCA GPUs

From: Julio Maia (jmaia_at_ks.uiuc.edu)
Date: Fri Nov 02 2018 - 13:13:21 CDT

Hi Kelly.
Yeah. MOZYME is not currently supported on GPUs. I have no idea if Prof. Stewart plans on having that done on MOPAC any time soon.

Thanks,
Julio

> On Nov 2, 2018, at 1:04 PM, McGuire, Kelly <mcg05004_at_byui.edu> wrote:
>
> Yep, that worked Thanks! . So, I read that the GPU won't help if MOZYME is called. Are you able to explain that?
>
> Kelly L. McGuire
> PhD Scholar
> Biophysics
> Department of Physiology and Developmental Biology
> Brigham Young University
> LSB 3050
> Provo, UT 84602
>
> From: Gerard Rowe <GerardR_at_usca.edu <mailto:GerardR_at_usca.edu>>
> Sent: Friday, November 2, 2018 11:56:39 AM
> To: McGuire, Kelly; namd-l_at_ks.uiuc.edu <mailto:namd-l_at_ks.uiuc.edu>
> Subject: Re: ORCA GPUs
>
> The CUDA-enabled version of NAMD is only able to use CUDA acceleration for pure MD simulations, and it doesn't allow for CUDA-MM during a QM/MM calculation. You only want CUDA enabled for the QM portion, which is called independently by MOPAC (as long as you have the GPU enabled binaries installed). When launching (GPU-QM)/MM calculations, you need to call the multicore NAMD binary without CUDA support. It's a bit confusing.
> From: McGuire, Kelly <mcg05004_at_byui.edu <mailto:mcg05004_at_byui.edu>>
> Sent: Friday, November 2, 2018 1:04:48 PM
> To: Gerard Rowe; namd-l_at_ks.uiuc.edu <mailto:namd-l_at_ks.uiuc.edu>
> Subject: Re: ORCA GPUs
>
> Hey Gerard, ok so I downloaded the CPU+GPU MOPAC 2016 and installed it. I also have NAMD 2.13b2 CUDA. I just tried running a MOPAC QM/MM job, and I get the error:
>
> FATAL ERROR: QM forces not compatible with CUDA
>
> I have my paths set to the correct NAMD and MOPAC, what else could be wrong?
>
> Kelly L. McGuire
> PhD Scholar
> Biophysics
> Department of Physiology and Developmental Biology
> Brigham Young University
> LSB 3050
> Provo, UT 84602
>
> From: Gerard Rowe <GerardR_at_usca.edu <mailto:GerardR_at_usca.edu>>
> Sent: Friday, November 2, 2018 8:14:41 AM
> To: namd-l_at_ks.uiuc.edu <mailto:namd-l_at_ks.uiuc.edu>; McGuire, Kelly
> Subject: Re: ORCA GPUs
>
> The NAMD side of things won't benefit from CUDA in a QM/MM situation because the bottleneck is the QM calculation. MOPAC does support GPU acceleration for systems after a certain atom number cutoff (100+, I think). If you want a GPU accelerated QM/MM calculation, you need to use the conventional NAMD build and specify the GPU accelerated MOPAC in the QM setup. Orca doesn't have GPU support at this time.
>
> -Gerard
> From: owner-namd-l_at_ks.uiuc.edu <mailto:owner-namd-l_at_ks.uiuc.edu> <owner-namd-l_at_ks.uiuc.edu <mailto:owner-namd-l_at_ks.uiuc.edu>> on behalf of McGuire, Kelly <mcg05004_at_byui.edu <mailto:mcg05004_at_byui.edu>>
> Sent: Thursday, November 1, 2018 9:48:47 PM
> To: namd-l_at_ks.uiuc.edu <mailto:namd-l_at_ks.uiuc.edu>
> Subject: namd-l: ORCA GPUs
>
> QM/MM with ORCA and MOPAC is working great now with our CPUs. I just tried a QM/MM with ORCA and GPUs, but I get the error:
>
> FATAL ERROR: QM forces are not compatible with CUDA at this time
>
> I am using NAMD 2.13b2 Linux x86 multicore CUDA and ORCA 4.0.1
>
> I thought ORCA/NAMD QM/MM hybrid simulations were compatible with GPUs...
>
>
> Kelly L. McGuire
> PhD Scholar
> Biophysics
> Department of Physiology and Developmental Biology
> Brigham Young University
> LSB 3050
> Provo, UT 84602

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2018 - 23:21:30 CST