[QE-users] QE-GPU Installation on i9/RTX4070 setup

[R T Phillips]

I have been trying to compile Quantum Espresso for use on this system:
Intel i9-14900 (24 core)
Linux Mint 22.1 Cinnamon version 6.4.6
Nvidia GeForce RTX 4070

with:
nvhpc_2024_2411_Linux_x86_64_cuda_12.6

The prerequisite gcc has been tested before installing nvhpc as 
specified in $1.1 of the nvidia hpc-sdk installation guide 
(https://docs.nvidia.com/hpc-sdk/archive/20.9/pdf/hpc209install.pdf).

The hpc-sdk specification is now said to include cuda and openmpi, so I 
have used only the one installation step at the top level of the nvhpc 
distribution (adding independent cuda or openmpi installations now seems 
to create havoc, though these may have been required in earlier versions 
of the hpc-sdk).

After installation of the nvhpc, nvaccelinfo produces the following 
output. Does this demonstrate that a "proper parallel environment" can 
be detected, as mentioned in the QE Guide?

"CUDA Driver Version:           12040
NVRM version:                  NVIDIA UNIX x86_64 Kernel Module 550.120  
Fri Sep 13 10:10:01 UTC 2024

Device Number:                 0
Device Name:                   NVIDIA GeForce RTX 4070
Device Revision Number:        8.9
Global Memory Size:            12584878080
Number of Multiprocessors:     46
Concurrent Copy and Execution: Yes
Total Constant Memory:         65536
Total Shared Memory per Block: 49152
Registers per Block:           65536
Warp Size:                     32
Maximum Threads per Block:     1024
Maximum Block Dimensions:      1024, 1024, 64
Maximum Grid Dimensions:       2147483647 x 65535 x 65535
Maximum Memory Pitch:          2147483647B
Texture Alignment:             512B
Clock Rate:                    2520 MHz
Execution Timeout:             Yes
Integrated Device:             No
Can Map Host Memory:           Yes
Compute Mode:                  default
Concurrent Kernels:            Yes
ECC Enabled:                   No
Memory Clock Rate:             10501 MHz
Memory Bus Width:              192 bits
L2 Cache Size:                 37748736 bytes
Max Threads Per SMP:           1536
Async Engines:                 2
Unified Addressing:            Yes
Managed Memory:                Yes
Concurrent Managed Memory:     Yes
Preemption Supported:          Yes
Cooperative Launch:            Yes
Unified Memory:                No
Memory Models Flags:           -gpu=mem:separate, -gpu=mem:managed
Default Target:                cc89  "

After the installation of nvhpc, I have tried both the default 
.configure, and also a bash script of QE-7.3.1 (see attached), setting 
the environment variables specified in $2.4 of the QE Users' Guide, and 
nvhpc.

Both produce config.log files which reveal problems in finding various 
components - the Guide warns that this might occur. I have trawled many 
websites for a solution but have not resolved this problem.  I have 
tried some variants of the above, which have led to apparently 
successful configuration runs, but 'make' subsequently fails.

It is pretty obvious that I am completely inexpert at this, and would 
greatly appreciate advice if anyone can spot what I am doing incorrectly.

Richard Phillips, Cavendish Lab. UK


-------------- next part --------------
A non-text attachment was scrubbed...
Name: config.log
Type: text/x-log
Size: 54221 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20250205/73d0d84d/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OC2.sh
Type: application/x-shellscript
Size: 2148 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20250205/73d0d84d/attachment-0001.bin>