Re: Re: Re: 3.0alpha9 CUDA version

From: 张驭洲 (
Date: Fri Mar 12 2021 - 00:09:48 CST

Hi Julio,

That's great! Waiting for the release of HIP version, and thanks for your reply!


发件人:"De Carvalho Maia, Julio Daniel" <>
发送时间:2021-03-12 13:45:00 (星期五)
收件人: "" <>, "Vermaas, Josh" <>, "张驭洲" <>
主题: Re: Re: namd-l: 3.0alpha9 CUDA version

Hi Zhang,

We're still working on the NAMD3.0 support for HIP (But it's coming soon!).

Best regards,

Research Programmer
Beckman Institute for Advanced Science and Technology
Vice Chancellor Research Institutes
University of Illinois at Urbana-Champaign
405 N. Mathews Avenue | M/C 251
Urbana, IL 61801
217-244-1928 |

Under the Illinois Freedom of Information Act any written communication to or from university employees regarding university business is a public record and may be subject to public disclosure.

From: <> on behalf of 张驭洲 <>
Sent: Thursday, March 11, 2021 10:29 PM
To: <>; Vermaas, Josh <>
Subject: Re: Re: namd-l: 3.0alpha9 CUDA version

I wonder if there is any release of HIP version of the namd 3 ?


&gt; -----原始邮件-----
&gt; 发件人: "Vermaas, Josh" <>
&gt; 发送时间: 2021-03-12 09:27:53 (星期五)
&gt; 收件人: "" <>, "Michael Von Domaros" <>
&gt; 抄送:
&gt; 主题: Re: namd-l: 3.0alpha9 CUDA version
&gt; Hi Michael,
&gt; I have NAMD3 running with CUDA 10 on my cluster. I got a copy of the
&gt; source from gitlab.
&gt; It is in the devel branch, and it tagged as 3.0a9. I have compilation
&gt; instructions that are what I used (or at least what I think I used from
&gt; memory) on github.
&gt; Compilation with NAMD3 from there is basically the same as it would be
&gt; anywhere else. You need charm++ compiled, followed by NAMD itself. The
&gt; hardest part for me was actually sorting out which of the many openMPI
&gt; implementations installed on the cluster had the PMI backend so I could
&gt; just call srun if I wanted to do replica exchange stuff.
&gt; -Josh
&gt; On 3/11/21 7:42 PM, Michael Von Domaros wrote:
&gt; &gt; Hi everyone,
&gt; &gt;
&gt; &gt; It appears that the namd 3 alpha 9 builds now require CUDA 11.0 (or
&gt; &gt; something even newer? not sure.) Is this absolutely necessary? I'm
&gt; &gt; asking because our compute cluster is stuck with CUDA 10 and CUDA 10
&gt; &gt; compatible device drivers, because a pretty substantial software stack
&gt; &gt; has been compiled against this version. The admins are somewhat
&gt; &gt; hesitant to update the device drivers, because they don't think that
&gt; &gt; running CUDA 10 apps with newer device drivers will be flawless.
&gt; &gt;
&gt; &gt; So, long story short: Could NAMD3 be compiled against CUDA 10? If so,
&gt; &gt; is there any way to access the source code? As far as I can tell, only
&gt; &gt; source codes for the 2.x series and nightly snapshots are available.
&gt; &gt;
&gt; &gt; Thanks a bunch,
&gt; &gt; Michael
&gt; --
&gt; Josh Vermaas
&gt; Assistant Professor, MSU-DOE Plant Research Lab and Department of Biochemisty and Molecular Biology

This archive was generated by hypermail 2.1.6 : Fri Dec 31 2021 - 23:17:10 CST