mersenneforum.org Optimising for Azure F-series
 Register FAQ Search Today's Posts Mark Forums Read

 2022-07-25, 10:34 #1 Th3D0kt0r   Jul 2022 102 Posts Optimising for Azure F-series Hi all - I'm quite new to the GIMPS project, but am really enjoying being involved :-) I have access to a few Azure subscriptions, so I thought I'd spin up a couple of VMs and try get mprime working. I've provisioned an Ubuntu VM using an F2s_v2, which has 2 x vCPUS (Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz) and 4GB RAM: https://docs.microsoft.com/en-us/azu...es/fsv2-series The CPUs are Cascade Lake, but the performance of the current work assignment (LL Double-check) is surprisingly slow. The current exponent will take around 13 days to complete, but a similar job on my old 4-core i7 laptop only took around 5 days. I'm guessing that either I haven't optimised things properly, or I just don't get how it's meant to be working. CPU utilisation seems to be capped out at 50%, so there's definitely extra performance to be had. This is the output of lscpu: Code: Architecture: x86_64 CPU op-mode(s): 32-bit, 64-bit Byte Order: Little Endian Address sizes: 46 bits physical, 48 bits virtual CPU(s): 2 On-line CPU(s) list: 0,1 Thread(s) per core: 2 Core(s) per socket: 1 Socket(s): 1 NUMA node(s): 1 Vendor ID: GenuineIntel CPU family: 6 Model: 85 Model name: Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz Stepping: 7 CPU MHz: 2593.907 BogoMIPS: 5187.81 Virtualization: VT-x Hypervisor vendor: Microsoft Virtualization type: full L1d cache: 32 KiB L1i cache: 32 KiB L2 cache: 1 MiB L3 cache: 35.8 MiB NUMA node0 CPU(s): 0,1 Has anyone else had any experience with Cascade Lake processors or thoughts on how to troubleshoot? Any assistance gratefully received :-)
 2022-07-25, 14:30 #2 axn     Jun 2003 22·7·193 Posts Sounds about right. Single core 13 days vs 4 core 5 days sounds pretty good.
2022-07-25, 14:46   #3
kriesel

"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

25·3·71 Posts

Quote:
 Originally Posted by Th3D0kt0r CPU utilisation seems to be capped out at 50%, so there's definitely extra performance to be had. This is the output of lscpu: Code: Architecture: x86_64 CPU op-mode(s): 32-bit, 64-bit Byte Order: Little Endian Address sizes: 46 bits physical, 48 bits virtual CPU(s): 2 On-line CPU(s) list: 0,1 Thread(s) per core: 2 Core(s) per socket: 1 Socket(s): 1
That's one core, both hyperthreads. 50% of 2 hyperthreads is one thread. It's maxed out. That's same as Colab free; 1 core 2 hyperthreads. Mprime typically does not benefit from running dual hyperthreads on a core for fft-based computations such as LL. It is typically memory bound, so much so George trades more instructions to reduce memory bandwidth requirements per fft iteration.

Last fiddled with by kriesel on 2022-07-25 at 14:52

2022-07-25, 21:17   #4
Th3D0kt0r

Jul 2022

210 Posts

Quote:
 Originally Posted by kriesel That's one core, both hyperthreads. 50% of 2 hyperthreads is one thread. It's maxed out. That's same as Colab free; 1 core 2 hyperthreads. Mprime typically does not benefit from running dual hyperthreads on a core for fft-based computations such as LL. It is typically memory bound, so much so George trades more instructions to reduce memory bandwidth requirements per fft iteration.

Thanks for the info. Does that mean that there's not much difference in provisioning a single core VM (F1s_v2) vs a two-core system? I'm just thinking that I could spin up two VMs for the price of one, if that's the case.

Additionally, what's the recommendation with regards to configuration of memory usage? The current VM has 4GB RAM, but mprime is only using about 300MB. There's nothing else running on the system, so it can certainly go higher. I have tried tweaking the memory usage in prime95 on a Windows box, but even with restarting the app it never seems to try and use what I've made available.

 Similar Threads Thread Thread Starter Forum Replies Last Post Godzilla Miscellaneous Math 1 2021-02-18 06:01 Sam Kennedy Factoring 22 2020-06-16 09:05 mackerel Hardware 11 2018-03-23 17:58 GP2 Cloud Computing 8 2017-08-24 01:27 M0CZY GMP-ECM 0 2010-10-27 13:39

All times are UTC. The time now is 23:35.

Sun Sep 25 23:35:48 UTC 2022 up 38 days, 21:04, 0 users, load averages: 0.94, 1.17, 1.21

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.

≠ ± ∓ ÷ × · − √ ‰ ⊗ ⊕ ⊖ ⊘ ⊙ ≤ ≥ ≦ ≧ ≨ ≩ ≺ ≻ ≼ ≽ ⊏ ⊐ ⊑ ⊒ ² ³ °
∠ ∟ ° ≅ ~ ‖ ⟂ ⫛
≡ ≜ ≈ ∝ ∞ ≪ ≫ ⌊⌋ ⌈⌉ ∘ ∏ ∐ ∑ ∧ ∨ ∩ ∪ ⨀ ⊕ ⊗ 𝖕 𝖖 𝖗 ⊲ ⊳
∅ ∖ ∁ ↦ ↣ ∩ ∪ ⊆ ⊂ ⊄ ⊊ ⊇ ⊃ ⊅ ⊋ ⊖ ∈ ∉ ∋ ∌ ℕ ℤ ℚ ℝ ℂ ℵ ℶ ℷ ℸ 𝓟
¬ ∨ ∧ ⊕ → ← ⇒ ⇐ ⇔ ∀ ∃ ∄ ∴ ∵ ⊤ ⊥ ⊢ ⊨ ⫤ ⊣ … ⋯ ⋮ ⋰ ⋱
∫ ∬ ∭ ∮ ∯ ∰ ∇ ∆ δ ∂ ℱ ℒ ℓ
𝛢𝛼 𝛣𝛽 𝛤𝛾 𝛥𝛿 𝛦𝜀𝜖 𝛧𝜁 𝛨𝜂 𝛩𝜃𝜗 𝛪𝜄 𝛫𝜅 𝛬𝜆 𝛭𝜇 𝛮𝜈 𝛯𝜉 𝛰𝜊 𝛱𝜋 𝛲𝜌 𝛴𝜎𝜍 𝛵𝜏 𝛶𝜐 𝛷𝜙𝜑 𝛸𝜒 𝛹𝜓 𝛺𝜔