Re: NAMD on AMD GPU

From: jing liang (jingliang2015_at_gmail.com)
Date: Thu Jun 24 2021 - 23:27:11 CDT

Hi,

rocm is installed in "/proc/rocm" but not rocfft. I guess the
rocfft library is this one: https://urldefense.com/v3/__https://github.com/ROCmSoftwarePlatform/rocFFT__;!!DZ3fjg!tEbDn9ILA1YyF1pGVXOjmIWgeAwIEsR8oHkWgTaz-tpsKmUR5E9EK1OlJgqEYELSaw$
?
I can only install that library in my home because rocm was installed by
the sys admin. In this case, changing the argument --rocm-prefix
would change the whole location of rocm right? Is there a specific switch
for changing the location of the rocfft library only?

thanks.

El jue, 24 jun 2021 a las 22:21, Josh Vermaas (<vermaasj_at_msu.edu>) escribió:

> One recommendation I would make is to fetch the source from
> https://urldefense.com/v3/__https://gitlab.com/tcbgUIUC/namd__;!!DZ3fjg!tEbDn9ILA1YyF1pGVXOjmIWgeAwIEsR8oHkWgTaz-tpsKmUR5E9EK1OlJgrFq5K2fw$ . The bugs mentioned about energy
> conservation are fixed in the master branch there. If rocfft.h can't be
> found, either there is no ROCm library installed, or AMD changed the
> directory layout *again*. The NAMD config file looks in /opt/rocm for the
> rocm installation, however that may not be where it was installed on your
> system. If this is the case, you'll need to set the --rocm-prefix argument.
>
> -Josh
> On 6/24/21 3:58 PM, jing liang wrote:
>
> I fetched the NAMD source code from the link:
> http://www.ks.uiuc.edu/Research/namd/alpha/2.15_amdhip/
> <https://urldefense.com/v3/__http://www.ks.uiuc.edu/Research/namd/alpha/2.15_amdhip/__;!!HXCxUKc!mTSBnVcKgph85ZxeUaK1kWcLQlPdo8R4iwA5HC6v3f6pCuqWoCXnyRj9ryJ7c7I$>
>
> and followed the BuildInstructions.md. I could build charm++ single node.
> Then, I tried to build NAMD:
>
> cd Linux-x86_64-g++.hip
> make -j8
>
> but I got the error:
> In file included from src/ComputePmeCUDAMgr.C:12:
> src/CudaPmeSolverUtil.h:11:10: fatal error: rocfft.h: No such file or
> directory
> #include <rocfft.h>
>
> do you have any comments about this?
>
>
>
>
>
> El mié, 23 jun 2021 a las 12:08, jing liang (<jingliang2015_at_gmail.com>)
> escribió:
>
>> Hi,
>>
>> I saw a thread on this mailing list about compiling NAMD on AMD GPUs and
>> it said that there was a problem with energy conservation. Is there an
>> update
>> on this? Could the same alpha version be used?
>>
>> http://www.ks.uiuc.edu/Research/namd/alpha/2.15_amdhip/
>> <https://urldefense.com/v3/__http://www.ks.uiuc.edu/Research/namd/alpha/2.15_amdhip/__;!!HXCxUKc!mTSBnVcKgph85ZxeUaK1kWcLQlPdo8R4iwA5HC6v3f6pCuqWoCXnyRj9ryJ7c7I$>
>>
>> thanks.
>>
>> --
> Josh Vermaas
> vermaasj_at_msu.edu
> Assistant Professor, Plant Research Laboratory and Biochemistry and Molecular Biology
> Michigan State Universityhttps://prl.natsci.msu.edu/people/faculty/josh-vermaas/
>
>

This archive was generated by hypermail 2.1.6 : Fri Dec 31 2021 - 23:17:11 CST