<div>Dear Viktor,</div><div>In principal, that should be possible. Other people should be able to help you with it. But I would not recommend to run CP2K on RTX 4080 or consumer cards in general. Their floating point performance with 64-bit floating points (that's what we use in CP2K) is usually not better than on CPU. Due to data transfers, CP2K is practically slower on this kind of hardware and requires GPGPU cards such as the NVIDIA L series if you want to stick to the same microarchitecture.</div><div>HTH,</div><div>Frederick</div><br /><div class="gmail_quote"><div dir="auto" class="gmail_attr">Victor Nazarychev schrieb am Samstag, 10. Mai 2025 um 15:40:34 UTC+2:<br/></div><blockquote class="gmail_quote" style="margin: 0 0 0 0.8ex; border-left: 1px solid rgb(204, 204, 204); padding-left: 1ex;">Dear Dobromir and Cp2k users,<br><br>I am writing to ask you about the following question: I have two RTX 4080 GPUs in my computer, and one processor with 96 CPU cores. Can I run two separate calculations using 48 cores on each GPU? Will the calculations be as fast as if I were to run a single task on the processor using 48 CPU cores?<br><br>Thank you for your attention.<br><br>Best regards, <div>Victor<br><div><br></div></div><div class="gmail_quote"><div dir="auto" class="gmail_attr">пятница, 17 января 2025 г. в 14:39:47 UTC+3, Dobromir A Kalchevski: <br></div></div><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Hello,<div><br></div><div>I figured the errors out: I had not included the CUDA directories before the rest in PATH and LD_LIBRARY_PATH, but after them.</div><div><br></div><div>Best Regards,</div><div>Dobromir</div><div><br></div><div> <br><br></div><div class="gmail_quote"><div dir="auto" class="gmail_attr">On Friday, January 17, 2025 at 2:25:24 AM UTC+2 Dobromir A Kalchevski wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Hello,<div><br></div><div>I noticed that the package SpFFT gives the same kind of errors during compilation.</div><div><br></div><div>It is just SpFFT and SIRIUS.<br><div><br></div><div>Best Regards,</div><div>Dobromir</div><div><br></div><div> <br><br></div></div><div class="gmail_quote"><div dir="auto" class="gmail_attr">On Friday, January 17, 2025 at 12:27:50 AM UTC+2 Dobromir A Kalchevski wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Hello,<div><br></div><div>I am trying to compile cp2k 2025.1 for GPU and during the sirius build I get the errors, attached in the log file. The sirius version is the one that cp2k downloads - 7.6.1.</div><div><br></div><div>I use the command</div><div>./install_cp2k_toolchain.sh --mpi-mode=mpich --with-mpich --with-gcc --enable-cuda --gpu-ver=H100 --with-cusolvermp</div><div><br></div><div>My uname -a is: Linux samara 6.12.8-1530.native #1 SMP PREEMPT Fri Jan 3 14:35:55 PST 2025 x86_64 GNU/Linux</div><div><br></div><div>My GCC version is 14.1.0</div><div>My nvidia driver version is 560</div><div>My CUDA version is 12.6.3</div><div>My video card is RTX 4080 Super</div><div><br></div><div>cp2k compiles fine for CPU only.</div><div><br></div><div>All previous dependencies compile fine.</div><div><br></div><div>For the purpose of searching the errors are of type:</div><div><br></div><div>Installing from scratch into /home/lsd/Downloads/cp2k-2025.1_mpich_gcc14_cuda/tools/toolchain/install/sirius-7.6.1<br>patching file src/potential/xc_functional_base.hpp<br>| ^~~~~~~~~<br>| float3<br><br>/usr/include/bits/mathcalls-narrow.h:39:8: error: ‘_Float32x’ does not<br>name a type; did you mean ‘float3’?<br>...<br>/usr/include/bits/iscanonical.h:51:24: error: ‘int iscanonical’<br>redeclared as different kind of entity<br><br> 51 | inline int iscanonical (_Float128 __val) { return __iscanonicalf128 (__val); }<br> | ^~~~~~~~~<br>...<br>/usr/include/bits/iscanonical.h:49:12: note: previous declaration ‘int<br>iscanonical(long double)’<br><br> 49 | inline int iscanonical (long double __val) { return __iscanonicall (__val); }<br> | ^~~~~~~~~~~<br>...<br><br>--error 0x1 --</div><div><br></div><div>Best Regards,</div><div>Dobromir</div><div><br></div><div> </div></blockquote></div></blockquote></div></blockquote></div></blockquote></div>
<p></p>
-- <br />
You received this message because you are subscribed to the Google Groups "cp2k" group.<br />
To unsubscribe from this group and stop receiving emails from it, send an email to <a href="mailto:cp2k+unsubscribe@googlegroups.com">cp2k+unsubscribe@googlegroups.com</a>.<br />
To view this discussion visit <a href="https://groups.google.com/d/msgid/cp2k/7bbb073b-b2e9-4942-9189-52ed6e3dcfc4n%40googlegroups.com?utm_medium=email&utm_source=footer">https://groups.google.com/d/msgid/cp2k/7bbb073b-b2e9-4942-9189-52ed6e3dcfc4n%40googlegroups.com</a>.<br />