Skip to content

also try collecting AMD GPU info (via rocm-smi) for --show-system-info#3978

Merged
bartoldeman merged 3 commits intoeasybuilders:developfrom
SebastianAchilles:gpuinfo
Mar 23, 2022
Merged

also try collecting AMD GPU info (via rocm-smi) for --show-system-info#3978
bartoldeman merged 3 commits intoeasybuilders:developfrom
SebastianAchilles:gpuinfo

Conversation

@SebastianAchilles
Copy link
Copy Markdown
Member

@SebastianAchilles SebastianAchilles commented Mar 10, 2022

Extends the GPU info to also report info for AMD GPUs

@easybuilders easybuilders deleted a comment from boegelbot Mar 16, 2022
@boegel boegel modified the milestones: 5.0, 4.x Mar 16, 2022
bartoldeman
bartoldeman approved these changes Mar 23, 2022
View reviewed changes
Copy link
Copy Markdown
Contributor

@bartoldeman bartoldeman left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

It's hard to add a test case and there isn't one for Nvidia GPUs either, so this should be ok.

@bartoldeman bartoldeman merged commit 74c9726 into easybuilders:develop Mar 23, 2022
@boegel boegel modified the milestones: 4.x, next release (4.5.4) Mar 25, 2022
@boegel
Copy link
Copy Markdown
Member

boegel commented Mar 26, 2022

Works like a charm on the MI100 system I have access to @ RUG:

$ eb --show-system-info
System information (pg-lab02.hpc.rug.nl):

* OS:
  -> name: RHEL
  -> type: Linux
  -> version: 8.4
  -> platform name: x86_64-unknown-linux

* CPU:
  -> vendor: AMD
  -> architecture: x86_64
  -> family: AMD
  -> arch name: UNKNOWN (archspec is not installed?)
  -> model: AMD EPYC 7542 32-Core Processor
  -> speed: 2894.608
  -> cores: 64
  -> features: 3dnowprefetch,abm,adx,aes,amd_ppin,aperfmperf,apic,arat,avic,avx,avx2,bmi1,bmi2,bpext,cat_l3,cdp_l3,clflush,clflushopt,clwb,clzero,cmov,cmp_legacy,constant_tsc,cpb,cpuid,cqm,cqm_llc,cqm_mbm_local,cqm_mbm_total,cqm_occup_llc,cr8_legacy,cx16,cx8,de,decodeassists,extapic,extd_apicid,f16c,flushbyasid,fma,fpu,fsgsbase,fxsr,fxsr_opt,ht,hw_pstate,ibpb,ibrs,ibs,irperf,lahf_lm,lbrv,lm,mba,mca,mce,misalignsse,mmx,mmxext,monitor,movbe,msr,mtrr,mwaitx,nonstop_tsc,nopl,npt,nrip_save,nx,osvw,overflow_recov,pae,pat,pausefilter,pclmulqdq,pdpe1gb,perfctr_core,perfctr_llc,perfctr_nb,pfthreshold,pge,pni,popcnt,pse,pse36,rdpid,rdrand,rdseed,rdt_a,rdtscp,rep_good,sep,sev,sev_es,sha_ni,skinit,smap,smca,sme,smep,ssbd,sse,sse2,sse4_1,sse4_2,sse4a,ssse3,stibp,succor,svm,svm_lock,syscall,tce,topoext,tsc,tsc_scale,umip,v_spec_ctrl,v_vmsave_vmload,vgif,vmcb_clean,vme,vmmcall,wbnoinvd,wdt,x2apic,xgetbv1,xsave,xsavec,xsaveerptr,xsaveopt,xsaves

* GPU:
  -> AMD
    -> 2x 0x0c34, 4.18.0-348.12.2.el8_5.x86_64

* software:
  -> glibc version: 2.28
  -> Python binary: /usr/bin/python3
  -> Python version: 3.6.8

@branfosj
Copy link
Copy Markdown
Member

branfosj commented Mar 26, 2022 

* GPU:
  -> AMD
    -> 2x 0x0c34, 4.18.0-348.12.2.el8_5.x86_64

Is that the right part of the GPU info output?

@boegel
Copy link
Copy Markdown
Member

boegel commented Mar 26, 2022

Yeah, I just realised that could be better ^_^

Here's the raw output:

$ rocm-smi --showdriverversion --csv
device,Driver version
cardsystem,4.18.0-348.12.2.el8_5.x86_64

$ rocm-smi --showproductname --csv
device,Card series,Card model,Card vendor,Card SKU
card0,Arcturus GL-XL [AMD Instinct MI100],0x0c34,Advanced Micro Devices Inc. [AMD/ATI],D34316
card1,Arcturus GL-XL [AMD Instinct MI100],0x0c34,Advanced Micro Devices Inc. [AMD/ATI],D34316

@SebastianAchilles So from this output, it should be taking the 2nd field ([1]), not the 3rd ([2]) as it is now?

@SebastianAchilles
Copy link
Copy Markdown
Member Author

SebastianAchilles commented Mar 26, 2022

The idea was to print the card model which is the 3rd field ([2]):

$ rocm-smi --showproductname


======================= ROCm System Management Interface =======================
================================= Product Info =================================
GPU[0]          : Card series:          Fiji [Radeon R9 FURY / NANO Series]
GPU[0]          : Card model:           Radeon R9 FURY X
GPU[0]          : Card vendor:          Advanced Micro Devices, Inc. [AMD/ATI]
GPU[0]          : Card SKU:             C88001
================================================================================
============================= End of ROCm SMI Log ==============================

$ rocm-smi --showproductname --csv
device,Card series,Card model,Card vendor,Card SKU
card0,Fiji [Radeon R9 FURY / NANO Series],Radeon R9 FURY X,Advanced Micro Devices Inc. [AMD/ATI],C88001

If Card model does not produce informative output on some cards, we could consider to change it to the Card series the 2nd field ([1]) instead.

@boegel Could you share the output of rocm-smi --showproductname?

@SebastianAchilles
Copy link
Copy Markdown
Member Author

SebastianAchilles commented Mar 26, 2022

The output above is for a newer driver version:

$ rocm-smi --showdriverversion --csv
device,Driver version
cardsystem,5.11.14

@SebastianAchilles
Copy link
Copy Markdown
Member Author

SebastianAchilles commented Mar 26, 2022

I opened a follow-up PR to also add the AMD card series: #3982

@SebastianAchilles SebastianAchilles changed the title add support for collecting GPU info (via rocm-smi) add support for collecting AMD GPU info (via rocm-smi) Mar 28, 2022
@boegel boegel changed the title add support for collecting AMD GPU info (via rocm-smi) also try collecting AMD GPU info (via rocm-smi) for --show-system-info Mar 29, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@bartoldeman bartoldeman bartoldeman approved these changes

Assignees

No one assigned

Labels

enhancement

Projects

None yet

Milestone

4.5.4

Development

Successfully merging this pull request may close these issues.

4 participants

@SebastianAchilles @boegel @branfosj @bartoldeman