Posts by Eric Driver

1) Message boards : News : Future of the Project (Message 3429)
Posted 6 days ago by Profile Eric Driver
Post:
It interesing, which search are you going to calculate after DS15x12


The plan is to do DS16x10 and DS16x11, and then back to finish off sf3.
2) Message boards : News : Support for Intel GPUs (Message 3423)
Posted 13 days ago by Profile Eric Driver
Post:
I seem to be having problems in getting GPU work. I have an Intel HD 630 GPU as far as I am aware the driver is up to date, but I keep getting the message no GPU work available for some reason. I can run Einstein WU's no problem so BOINC knows the GPU is there.

Thanks


It could be several things. You need to have beta apps selected and the card must support openCL 1.1
3) Message boards : News : Support for Intel GPUs (Message 3420)
Posted 13 days ago by Profile Eric Driver
Post:
I was comparing to a default Windows Get Decic Fields v4.00 (default)windows_x86_64

You may have a point about Driver or implementation

Bill F

The GPU apps should be faster than the default CPU app. However, some of the older cards can be slower, if they work at all.
4) Message boards : News : Support for Intel GPUs (Message 3418)
Posted 14 days ago by Profile Eric Driver
Post:
bunch of WU processed :)

https://numberfields.asu.edu/NumberFields/result.php?resultid=161210884
https://numberfields.asu.edu/NumberFields/result.php?resultid=161211661

and plenty others


Thx! Good to see it's working, and that run times are reasonable on Intel's newest GPUs.


Run times are longer for the Intel GPU tasks. Are they doing larger amounts of science and should the credit amount be higher ?

Bill F


Run times are longer compared to what? Compared to the previous dataset, run times are longer and credits have gone up accordingly.

If comparing to a different gpu then average times will probably be different. Each WU is a fixed amount of work and therefore has a fixed credit. Just because they run longer doesn't mean they deserve more credit. If it's running longer than expected then it could be an inefficient driver or bad OpenCL implementation.
5) Message boards : News : Support for Intel GPUs (Message 3416)
Posted 16 days ago by Profile Eric Driver
Post:
bunch of WU processed :)

https://numberfields.asu.edu/NumberFields/result.php?resultid=161210884
https://numberfields.asu.edu/NumberFields/result.php?resultid=161211661

and plenty others


Thx! Good to see it's working, and that run times are reasonable on Intel's newest GPUs.
6) Message boards : News : Support for Intel GPUs (Message 3410)
Posted 21 days ago by Profile Eric Driver
Post:
Adding an Intel ARC A750 to the pool, watch these spaces => https://numberfields.asu.edu/NumberFields/show_host_detail.php?hostid=2835341 and https://numberfields.asu.edu/NumberFields/results.php?hostid=2835341


It's been several hours and nothing returned yet, I hope that's not a bad sign.

I would expect the Arc A750 to do well, assuming Intel puts out a decent openCL driver.


The machine is busy with Einstein@home tasks :D
Arc A750 doesn't have FP64 though

I'll pause Einstein and let some tasks run


I saw some cpu tasks go through. Any problems with tasks on the Arc A750?
7) Message boards : News : Support for Intel GPUs (Message 3408)
Posted 22 days ago by Profile Eric Driver
Post:
Adding an Intel ARC A750 to the pool, watch these spaces => https://numberfields.asu.edu/NumberFields/show_host_detail.php?hostid=2835341 and https://numberfields.asu.edu/NumberFields/results.php?hostid=2835341


It's been several hours and nothing returned yet, I hope that's not a bad sign.

I would expect the Arc A750 to do well, assuming Intel puts out a decent openCL driver.
8) Message boards : News : Future of the Project (Message 3402)
Posted 27 days ago by Profile Eric Driver
Post:
I am an alumnus of ASU. I think maybe we can try to reach out to the university foundation to see if they can setup a donation page for this project. Another option would be cloud computing. Some alternative cloud providers like Linode may have a cheaper price than AWS or Azure. I (will be come soon) am a software engineer, and I am familiar with cloud computing technologies. Can you list the hardware/network requirements of this project so we can start calculating how much would be required to maintain it?


Hardware is not a problem, I could even use one of my older systems, maybe with an upgraded network card. But to answer your question, one would need the equivalent of a Xeon class processor, an SSD, and a fast network connection.

The bottleneck is the network connection. During the pentathlon earlier this year, the server was processing 400k tasks per day (>16000 file transfers per hour), and the network bandwidth was starting to hit its limit (not sure if it was the connection speed or the ethernet card). So that would be my biggest concern hosting the project somewhere else.
9) Questions and Answers : Windows : Aborted by Project? (Message 3395)
Posted 23 Oct 2022 by Profile Eric Driver
Post:
Apologies A typo with my ID#. That should be 90801.

Thank you,
Wayne Solomon


That ID is the Science United user. I'm not sure how Science United works, but perhaps you originally connected via an account with them? You might want to ask around on their message boards.
10) Message boards : Number crunching : similar issue of server side scripts including at avatar uploading in community preferences (Message 3390)
Posted 20 Oct 2022 by Profile Eric Driver
Post:
Should be working now (see other post).
11) Message boards : Number crunching : cannot upload profile avatar image file (Message 3389)
Posted 20 Oct 2022 by Profile Eric Driver
Post:
I think this is working now. At least I was able to delete my own profile image and load it back with no errors.

It turned out the php-gd package needed to be installed on the server. Thanks for bringing this problem to my attention.
12) Message boards : Number crunching : cannot upload profile avatar image file (Message 3386)
Posted 19 Oct 2022 by Profile Eric Driver
Post:
when creating profiles avatar image by uploading file, server side script has troubles to convert and insert into database

error messages like this


Fatal error: Uncaught Error: Call to undefined function imageCreateFromJPEG() in /home/boincadm/projects/NumberFields/html/user/create_profile.php:129 Stack trace: #0 /home/boincadm/projects/NumberFields/html/user/create_profile.php(261): getImages() #1 /home/boincadm/projects/NumberFields/html/user/create_profile.php(350): process_create_profile() #2 {main} thrown in /home/boincadm/projects/NumberFields/html/user/create_profile.php on line 129


Try it again. I found a typo in the create_profile.php script that I am hoping is the problem.
13) Message boards : Number crunching : More than one task per GPU? (Message 3382)
Posted 18 Oct 2022 by Profile Eric Driver
Post:
Perfect! Thanks!!

ETA: Wow, SEARCH is wierd.

I tried searching for app_config.xml with and without quotes and got nothing. After I got your reply I searched for "app config" and found what I was looking for. Sorry I didn't try that before.


I've never actually tried the search function, but that does sound weird.
14) Message boards : Number crunching : More than one task per GPU? (Message 3380)
Posted 18 Oct 2022 by Profile Eric Driver
Post:
Is it possible to run more than one GPU task on a GPU (assuming CPU is not limiting)?

I searched the forums and found nothing about app_config.xml files.

Thanks.


Yes. Here is my app config for running 2 simultaneously. Each instance uses about 20% of a CPU, but yours might differ.

<app_config>
  <app>
    <name>GetDecics</name>
    <gpu_versions>
      <gpu_usage>0.49</gpu_usage>
      <cpu_usage>0.20</cpu_usage>
    </gpu_versions>
  </app>
</app_config>
15) Message boards : News : Future of the Project (Message 3371)
Posted 8 Oct 2022 by Profile Eric Driver
Post:
I would really like to see sf3 finished, is it possible to make it GPU only and start to feed it to us?
we could calculate sf7 and sf3 side by side, other on CPU and other on GPU.


It is not possible for the app to distinguish between subfields. The only way would be to create a different app for one of the subfields and then only enable the GPU platforms for it. A lot of work for little benefit. Have patience... we will be back to sf3 early next year.
16) Message boards : Number crunching : Issues with GPU's (2 issues) (Message 3348)
Posted 13 Sep 2022 by Profile Eric Driver
Post:
Issue 1: neither of my intel GPU's are getting tasks (HD 520, HD 4600).
standard event log of requesting tasks for intel GPU, got 0 new tasks, delay of (-) seconds.
Issue 2: Project is not using one of my AMD GPU's correctly (AMD R5 240), Its a relatively slow GPU but it never goes past 7% usage with number Feilds tasks ultimately causing the task to eventually fail, while my AMD Rx 460 gets used at 98% consistently. My guess is that this specific GPU is not on the GPU table.


I'm not sure how well the intel GPUs will do, but it should give them work if they support openCL. Do you have beta apps enabled?

I would expect the older AMD card to not work well due to the computational load, which is what you are seeing. The GPU table is only for fine tuning some parameter settings; if your card is not in the table then it uses default settings.
17) Message boards : Science : Availability of source code (Message 3345)
Posted 10 Sep 2022 by Profile Eric Driver
Post:
I am not an expert in computational number theory, but upon reviewing a bit the present code, I think that there is not a lot of improvement we can do without algorithmical advance or changing GMP for the CPU code. For the algorithmic part, it is clear. For GMP, it is because most of the computation is arithmetic using GMP, which means that the speed mostly depends on how GMP is doing. However, GMP is a generic arbitrary precision library, meaning that there are some internal managements that are done to check which algorithm to apply. This in general gives an overhead, and probably also forbids any kind of vectorisation except those already in GMP.

One idea is, from the CUDA code we can see that there is a natural bound on numbers that are used in the computation. If they all fit in 128 bits, then gcc supports __int128. Otherwise, it might (!) help the speed by using or writing an emulated type in which the numbers fit, like the one used in the CUDA code. In this way, at least gcc would have more clues to vectorize (or not!).

fwjmath.


Thanks for the analysis. One thing I will point out is, the numbers grow to almost 1000 bits in size in the later passes through the loop.
18) Message boards : News : New badge type (Message 3335)
Posted 30 Aug 2022 by Profile Eric Driver
Post:
Please add the other sub-field badges. How many sub-fields are involved in this project. I currently have counts on Decic and Q(Sqrt[-10]). I have no counts on Septic.

Thanks

Lorenz H. Menke, Jr.


Septics was an old search from several years ago, so there are no more of those. After the current subfield, there are only 2 more left. I will add badges for those when we get there. Be patient.
19) Message boards : Number crunching : Certificate of Computation (Account) (Message 3323)
Posted 18 Aug 2022 by Profile Eric Driver
Post:
Hello

Information:

The Date in the Certificate of Computation (Account) is - 1 January 1970 - .
It should be the current Date Today (also).

I also see this right above the Certificate of Computation (Account) ..outside the frame:
Warning: time() expects exactly 0 parameters, 1 given in /home/boincadm/projects/NumberFields/html/user/cert1.php on line 27

This should be corrected.

Thank you in advance.


This should be fixed now. Thanks for letting me know.
20) Message boards : News : Future of the Project (Message 3319)
Posted 12 Aug 2022 by Profile Eric Driver
Post:
In worst case is it possible to organize the server at home. I think it should not be very powerful: i7/32GB/10TB.

I think our communitee can help with all expenses (hardware, electricity, new domain etc) if you organize donations.

Similar to Primegrid:
https://www.primegrid.com/donations.php


I don't have enough bandwidth at home to handle the load, which was the main advantage of ASU. Besides, my ISP blocks ports 80 and 443 (and others). I would have to purchase a high speed dedicated connection and last I checked they were over $150 per month.


Next 20


Main page · Your account · Message boards


Copyright © 2022 Arizona State University