Posts by Raistmer

21) Message boards : Number crunching : Next Lunatics? (Message 1948567)
Posted 9 Aug 2018 by Profile Raistmer
Post:
But based on comments alluded by Stephen and what I remember, our Windows app developer Raistmer said it would be almost impossible to use the exiting CUDA9 Linux app code source for any Windows app simply because the mechanisms the Linux app uses isn't available in the Windows environment.

Ups, hardly I said so :)
What I could say it's no Windows build until sync bug would be resolved (is it?).
CUDA is quite OS-independent AFAIK though I dealt only with older versions.
22) Message boards : Number crunching : SETI@Home Wow!-Event 2018 (Message 1948566)
Posted 9 Aug 2018 by Profile Raistmer
Post:
This time I'm in :)
23) Message boards : Number crunching : To Hyperthread or not to Hyperthread, that is the question (Message 1947598)
Posted 2 Aug 2018 by Profile Raistmer
Post:
Windows scheduler has affinity property that allows to pin process to particular logical CPU. So it's possible to group CPU processes on one real core while GPU processes allocate to anotehr real CPU core.
24) Message boards : Number crunching : Please point me to a discussion of 8.00 vs. 8.05 vs. 8.08 (alt)? (Message 1947597)
Posted 2 Aug 2018 by Profile Raistmer
Post:
what was named (alt) is stock codebase PGOed with VS2017 or VS2015 (don't recall what latest was at that time). prev x64 was buil with older VS.
So, same codebase, different compilers.
25) Message boards : Number crunching : I've Built a Couple OSX CUDA Apps... (Message 1943626)
Posted 10 Jul 2018 by Profile Raistmer
Post:
If the code streams from CUDA and OpenCL apps could decide on the ordering of overflow counts

Little misleading statement. There is CPU apps too. And both OpenCL and CUDA should agree with their order not just align vs each other.
26) Message boards : Number crunching : How to get new host ID most easely? (Message 1939028)
Posted 10 Jun 2018 by Profile Raistmer
Post:
Have an empty cache.

Fake the <rpc_seqno> in client_state - set it low. (BOINC will think you're cheating by running two machines with the same ID, and split them).

New cross-project ID was generated.
But computers page doesn't show new host. Old one with its statistics...
27) Message boards : Number crunching : I've Built a Couple OSX CUDA Apps... (Message 1938675)
Posted 8 Jun 2018 by Profile Raistmer
Post:
Ref: xjf
While bad one: jf

To decrease number of variables I would rebuild exactly with same defines set (x absence means that USE_I386_XEON define missing)
Try to build AVX instead of SSE4.2
28) Message boards : Number crunching : I've Built a Couple OSX CUDA Apps... (Message 1938665)
Posted 8 Jun 2018 by Profile Raistmer
Post:
Would be good to see result.sah files (good vs bad) too.
29) Message boards : Number crunching : I've Built a Couple OSX CUDA Apps... (Message 1938662)
Posted 8 Jun 2018 by Profile Raistmer
Post:
Bads of all signal types usually mean smth wrong with chirping. Chirp is quite fragile part that include trigonometry.
Check if "fast math" ON.
30) Message boards : Number crunching : I've Built a Couple OSX CUDA Apps... (Message 1938633)
Posted 8 Jun 2018 by Profile Raistmer
Post:
Try and tell us :)

Good that you point direct places to fix in our sources but not good that BOINC sources modifications required also :/
31) Message boards : Number crunching : How to get new host ID most easely? (Message 1938473)
Posted 6 Jun 2018 by Profile Raistmer
Post:
I replaced hardware keeping sam eOS installation. So old BOINC app statistics not relevant to new host. Would like to get new host ID w/o complete BOINC reinstall (preferably w/o tasks trashing).

How?
32) Message boards : Number crunching : I've Built a Couple OSX CUDA Apps... (Message 1938470)
Posted 6 Jun 2018 by Profile Raistmer
Post:
Could this help?

Try adding -DCMAKE_CXX_FLAGS=-D__GLIBCXX_USE_CXX11_ABI=0 to your cmake call


https://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html
33) Message boards : Number crunching : I've Built a Couple OSX CUDA Apps... (Message 1938393)
Posted 6 Jun 2018 by Profile Raistmer
Post:
Could be useful in transition https://www.codeplay.com/portal/07-14-17-codeplay-release-clspv-an-opencl-tool-for-vulkan-enabled-devices
34) Message boards : Number crunching : SETIv8 for Linux skylake-avx512 available (Message 1938390)
Posted 6 Jun 2018 by Profile Raistmer
Post:

All that work was done by Raistmer - he'll probably discover this thread in the morning. I'm on UK time, so I'll sign off for tonight.

:D ;D
35) Message boards : Number crunching : SETIv8 for Linux skylake-avx512 available (Message 1938389)
Posted 6 Jun 2018 by Profile Raistmer
Post:
Thanks, I've tried compiling that version with no success. There are multiple errors on Linux; it appears to be meant for Windows. Any other advice would be greatly appreciated.


It's cross-platform sources. Look for configure lines in repo. It known to be buildable on Linux (for x86), checked on quite recent revs.
36) Message boards : Number crunching : SETIv8 for Linux skylake-avx512 available (Message 1938388)
Posted 6 Jun 2018 by Profile Raistmer
Post:
Despite the name, "branches/sah_v7_opt/AKv8/client" is probably the one, but take advice.

Yep, for opt CPU MBv8 build it's right choice.
37) Message boards : Number crunching : SETIv8 for Linux skylake-avx512 available (Message 1938386)
Posted 6 Jun 2018 by Profile Raistmer
Post:
Thanks for that info. I've been told that there's an optimized code base available somewhere to compile from, which would give me better results. I've done more than 11 tasks so the estimates should be reliable in BOINC Manager. The run time figures on the website are inaccurate AFAIK, as it appears to be including suspended time.


Much better to do more formal and stable tests with package availble here: http://lunatics.kwsn.info/index.php?action=downloads;cat=5

Compare your binary with those: http://lunatics.kwsn.info/index.php?action=downloads;cat=48

And as concluding remark: one should understand that w/o direct AVX2 instructions addition to code one will completely depending from compiler own autovectorising abilities.
And if I could believe that optimizing compiler can autoinject SIMD instead of high-level code I hardly believe it will replace already existing SIMD with another one. So, rebuild could touch only small part of existing codebase anyway.
And regarding AVX2 per se - what particular instructions SETI could use from that IA?
38) Message boards : Number crunching : SETIv8 for Linux skylake-avx512 available (Message 1938385)
Posted 6 Jun 2018 by Profile Raistmer
Post:
Thank you, that's very helpful. I did indeed compile from the stock code base. After a google search for optimized SETI@home, I've found no optimized sources. Could you please point me in the right direction?

I would expect more from Google's AI ;)

Anyway, here you go: https://setisvn.ssl.berkeley.edu/svn/branches/sah_v7_opt
39) Message boards : Number crunching : Apple deprecates OpenCl (and OpenGl) (Message 1938383)
Posted 6 Jun 2018 by Profile Raistmer
Post:
Apple

Well, to be completely proprietary in Apple's traditions (I would say it's their basic instinct ). Users of Apple production should be quite familiar with that.
Perhaps they happy enough with such tradition for Apple to hold its market share still :P

For SETI it's another obstacle of course. With possible "no SETI GPU compute on macOS" addition to "No BOINC on iPhones"...
40) Message boards : Number crunching : I've Built a Couple OSX CUDA Apps... (Message 1938382)
Posted 6 Jun 2018 by Profile Raistmer
Post:
When are you going to support Vulkan in your SETI Apps?

After making sure that Vulkan is compute API, not just graphics API.


Previous 20 · Next 20


 
©2018 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.