<html><head><meta http-equiv="content-type" content="text/html; charset=utf-8"></head><body dir="auto">Hi Yuzhi (&amp;Junhan)!<div><br></div><div>First, thanks for getting involved! We love to have more people involved, and will do our best to welcome you to the extended team. free energy is definitely an area we want to improve further, and contributions like the ones below sound great.</div><div><br></div><div>Why don't you joint our next open developer call every second Wednesday? Or, if the timezone makes things inconvenient we could schedule a separate meeting.</div><div><br></div><div>Some of the things we will like discuss would be:</div><div><br></div><div>1. We are gradually moving to support both CUDA, SYCL and Hip so we run well on all accelerators - but this also means we need to avoid hacking things just for CUDA, and rather have a clean hierarchical implementation where different platforms might initially support different accelerated features. Much of this stuff has already started to happen in gmx 2021.</div><div><br></div><div>2. For new features and functional forms we tend to be restrictive, partly because it means a collective undertaking for the entire team the next 5-10 years to ensure that feature works with all other features. We have not quite settled things on free energy yet, and we might consider modifying the interactions, but we don't want 4-5 different implementations from different teams - instead we should try to settle on which one will serve us all best.</div><div><br></div><div>3. Documentation and Unit tests are critical to ensure new code keeps working ;-)</div><div><br></div><div>Cheers,</div><div><br></div><div>Erik</div><div><br><div dir="ltr">--<div>Erik Lindahl &lt;erik.lindahl@scilifelab.se&gt;</div><div>Professor of Biophysics</div><div>Science for Life Laboratory</div><div>Stockholm University &amp; KTH</div><div>Office (SciLifeLab): +46 8 524 81567</div><div>Cell (Sweden): +46 73 4618050&nbsp;</div><div>Cell (US): 1 267 307 8746</div><div><br></div></div><div dir="ltr"><br><blockquote type="cite">On Mar 15, 2021, at 12:13, Yuzhi Zhang &lt;352@pku.edu.cn&gt; wrote:<br><br></blockquote></div><blockquote type="cite"><div dir="ltr"><div class=" __aliyun_node_has_color" style="margin:0px;padding:0px;border:0px;outline:0px;line-height:1.7;font-family:Tahoma, Arial, STHeiti, SimSun;font-size:14px;">
        <div style="margin:0px;padding:0px;border:0px;outline:0px;clear:both;">
                Dear&nbsp;Gromacs&nbsp;developers,&nbsp;
                <div style="margin:0px;padding:0px;border:0px;outline:0px;clear:both;">
                        &nbsp;&nbsp;&nbsp;&nbsp;Based&nbsp;on&nbsp;Gromacs-2020.2,&nbsp;we've&nbsp;done&nbsp;a&nbsp;series&nbsp;of&nbsp;development&nbsp;related&nbsp;to&nbsp;FEP&nbsp;in&nbsp;following&nbsp;aspects&nbsp;.&nbsp;We&nbsp;would&nbsp;like&nbsp;to&nbsp;contribute&nbsp;our&nbsp;codes&nbsp;to&nbsp;Gromacs&nbsp;and&nbsp;we're&nbsp;looking&nbsp;forward&nbsp;to&nbsp;hearing your&nbsp;opinions&nbsp;first.
                </div>
                <div style="margin:0px;padding:0px;border:0px;outline:0px;clear:both;">
                        &nbsp;&nbsp;&nbsp;&nbsp;High-performance&nbsp;implementations:&nbsp;
                </div>
                <div style="margin:0px;padding:0px;border:0px;outline:0px;clear:both;">
                        &nbsp;&nbsp;&nbsp;&nbsp;1.&nbsp;&nbsp;Offload&nbsp;FEP-PME&nbsp;to&nbsp;GPU.&nbsp;&nbsp;Our&nbsp;status:&nbsp;finished. We've&nbsp;also&nbsp;noticed&nbsp;that&nbsp;this&nbsp;function&nbsp;has&nbsp;been&nbsp;implemented&nbsp;in&nbsp;commit&nbsp;f7be07e3cc901eb03700d93248fc09b573370282&nbsp;by&nbsp;M.&nbsp;Lundborg&nbsp;et.&nbsp;al.&nbsp;and&nbsp;has&nbsp;been&nbsp;merged&nbsp;in&nbsp;Gromacs-2021.&nbsp;So&nbsp;we&nbsp;won't&nbsp;commit&nbsp;this.&nbsp;
                </div>
                <div style="margin:0px;padding:0px;border:0px;outline:0px;clear:both;">
                        &nbsp;&nbsp;&nbsp;&nbsp;2.&nbsp;&nbsp;Offload non-bonded free energy kernel&nbsp;to&nbsp;GPU.&nbsp;Our&nbsp;status:&nbsp;finished.&nbsp;
                </div>
                <div style="margin:0px;padding:0px;border:0px;outline:0px;clear:both;">
                        &nbsp;&nbsp;&nbsp;&nbsp;3.&nbsp;&nbsp;Offload&nbsp;bonded free energy kernel&nbsp;to&nbsp;GPU.&nbsp;Our&nbsp;status:&nbsp;finished.
                </div>
                <div style="margin:0px;padding:0px;border:0px;outline:0px;clear:both;">
                        &nbsp;&nbsp;&nbsp;&nbsp;4.&nbsp;&nbsp;Support&nbsp;GPU&nbsp;updates for FEP.&nbsp;Our&nbsp;status:&nbsp;finished&nbsp;md-integrator. We've&nbsp;noticed&nbsp;that&nbsp;this&nbsp;has&nbsp;also&nbsp;been&nbsp;implemented&nbsp;in&nbsp;Gromacs-2021,&nbsp;but&nbsp;we'are&nbsp;still&nbsp;interested&nbsp;in&nbsp;supporting&nbsp;GPU&nbsp;updates&nbsp;of&nbsp;sd-integrator.&nbsp;&nbsp;Similar&nbsp;discussion&nbsp;can&nbsp;be&nbsp;found&nbsp;in&nbsp;issue&nbsp;&nbsp;#3258&nbsp;on&nbsp;gitlab.&nbsp;
                </div>
                <div style="margin:0px;padding:0px;border:0px;outline:0px;clear:both;">
                        &nbsp; &nbsp; 5.&nbsp; Support GPU replica exchange. Our status: finished.
                </div>
                <div style="margin:0px;padding:0px;border:0px;outline:0px;clear:both;">
                        <br>
                </div>
                <div style="margin:0px;padding:0px;border:0px;outline:0px;clear:both;">
                        &nbsp;&nbsp;&nbsp;&nbsp;New&nbsp;features:&nbsp;
                </div>
                <div style="margin:0px;padding:0px;border:0px;outline:0px;clear:both;">
                        &nbsp;&nbsp;&nbsp;&nbsp;1.&nbsp;&nbsp;Support&nbsp;"Soft&nbsp;Bond&nbsp;Potential"&nbsp;in&nbsp;FEP,&nbsp;which&nbsp;is&nbsp;proposed&nbsp;in&nbsp;&nbsp;DOI:&nbsp;10.1021/acs.jctc.6b00991&nbsp;and&nbsp;allows&nbsp;a&nbsp;more&nbsp;smooth&nbsp;change&nbsp;of&nbsp;bond&nbsp;topology&nbsp;when&nbsp;there&nbsp;is&nbsp;bond&nbsp;formation&nbsp;or&nbsp;breaking&nbsp;in&nbsp;a&nbsp;FEP&nbsp;transformation,&nbsp;such&nbsp;as&nbsp;5-member&nbsp;ring&nbsp;to&nbsp;6-member&nbsp;ring.&nbsp;&nbsp;Our&nbsp;status:&nbsp;finished.&nbsp;
                </div>
                <div style="margin:0px;padding:0px;border:0px;outline:0px;clear:both;">
                        &nbsp;&nbsp;&nbsp;&nbsp;2.&nbsp;&nbsp;Split&nbsp;softcore&nbsp;parameters&nbsp;for&nbsp;VDW&nbsp;and&nbsp;Coulomb&nbsp;interactions.&nbsp;&nbsp;We&nbsp;noticed&nbsp;that&nbsp;currently&nbsp;Gromacs&nbsp;supports&nbsp;only&nbsp;one&nbsp;parameter&nbsp;"sc-alpha"&nbsp;to&nbsp;modify&nbsp;soft-core&nbsp;interactions.&nbsp;&nbsp;After&nbsp;our&nbsp;tests,&nbsp;this&nbsp;will&nbsp;cause&nbsp;a&nbsp;phase-transition-like&nbsp;phenomenon&nbsp;at&nbsp;middle&nbsp;lambdas&nbsp;in&nbsp;some&nbsp;cases.&nbsp;&nbsp;Our&nbsp;status:&nbsp;finished.&nbsp;&nbsp;We&nbsp;also&nbsp;noticed&nbsp;recently&nbsp;there have&nbsp;been&nbsp;some&nbsp;new&nbsp;forms&nbsp;of&nbsp;softcores&nbsp;&nbsp;like&nbsp;gapsys&nbsp;softcore&nbsp;(DOI:&nbsp;&nbsp;dx.doi.org/10.1021/ct300220p&nbsp;and&nbsp;discussed&nbsp;in&nbsp;<a href="https://gitlab.com/gromacs/gromacs/-/merge_requests/882" target="_blank">https://gitlab.com/gromacs/gromacs/-/merge_requests/882</a>&nbsp;)&nbsp;and&nbsp;Amber&nbsp;SSC2&nbsp;softcore&nbsp;(DOI:&nbsp;10.1021/acs.jctc.0c00237.)&nbsp;&nbsp;And&nbsp;we&nbsp;would&nbsp;like&nbsp;to&nbsp;ask&nbsp;if&nbsp;you&nbsp;consider&nbsp;it's&nbsp;necessary&nbsp;to&nbsp;support&nbsp;such&nbsp;new&nbsp;softcore&nbsp;potentials.&nbsp;
                </div>
                <div style="margin:0px;padding:0px;border:0px;outline:0px;clear:both;">
                        &nbsp;&nbsp;&nbsp;&nbsp;These&nbsp;terms&nbsp;are&nbsp;all&nbsp;that&nbsp;we&nbsp;want&nbsp;to&nbsp;contribute&nbsp;to&nbsp;Gromacs.&nbsp;But&nbsp;we&nbsp;have&nbsp;to&nbsp;admit&nbsp;that&nbsp;there&nbsp;is&nbsp;still&nbsp;a&nbsp;lot&nbsp;of&nbsp;work&nbsp;to&nbsp;test&nbsp;and&nbsp;standardize&nbsp;our&nbsp;codes&nbsp;to&nbsp;meet&nbsp;all&nbsp;requirements&nbsp;of&nbsp;Gromacs.&nbsp;&nbsp;Is&nbsp;there&nbsp;any&nbsp;suggestion?&nbsp;
                </div>
                <div style="margin:0px;padding:0px;border:0px;outline:0px;clear:both;">
                        <br>
                </div>
                <div style="margin:0px;padding:0px;border:0px;outline:0px;clear:both;">
                        &nbsp;&nbsp;&nbsp;&nbsp;Best&nbsp;wishes!
                </div>
                <div style="margin:0px;padding:0px;border:0px;outline:0px;clear:both;">
                        &nbsp;&nbsp;&nbsp;&nbsp;Happy Simulating!<br>
                </div>
                <div style="margin:0px;padding:0px;border:0px;outline:0px;clear:both;">
                        &nbsp;&nbsp;&nbsp;&nbsp;Yuzhi &amp; Junhan&nbsp;
                </div>
        </div>
</div>
<div style="margin:0px;padding:0px;border:0px;outline:0px;clear:both;font-size:0px;height:1px;overflow:hidden;">
</div>
<div style="margin:0px;padding:0px;border:0px;outline:0px;line-height:20px;clear:both;">
        <br style="font-family:Helvetica, Tahoma, Arial;font-size:14px;white-space:normal;">
</div><span>-- </span><br><span>Gromacs Developers mailing list</span><br><span></span><br><span>* Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List before posting!</span><br><span></span><br><span>* Can't post? Read http://www.gromacs.org/Support/Mailing_Lists</span><br><span></span><br><span>* For (un)subscribe requests visit</span><br><span>https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers or send a mail to gmx-developers-request@gromacs.org.</span></div></blockquote></div></body></html>