Message boards :
Number crunching :
x64 Windows CPU build testing on beta
Message board moderation
Author | Message |
---|---|
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Finally we have x64 binary for Windows deployed Windows/x86 running on an AMD x86_64 or Intel EM64T CPU 8.05 (sse3) 1 Feb 2017, 19:41:37 UTC 6 GigaFLOPS Please consider to enable CPU work fetch on your x64 Windows hosts for beta project. SETI apps news We're not gonna fight them. We're gonna transcend them. |
Michel Makhlouta Send message Joined: 21 Dec 03 Posts: 169 Credit: 41,799,743 RAC: 0 |
I've joined Beta and attached my client, also got WU's that are x86 for my 4770K Edit: I am running 1 WU 64 sse3 |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz [Family 6 Model 60 Stepping 3] x64 selected: v_vGetPowerSpectrumUnrolled2 0.000071 0.00000 avx_ChirpData_d 0.001861 0.00000 v_vTranspose4x16ntw 0.000872 0.00000 AK SSE folding 0.000920 0.00000 v_avxGetPowerSpectrum 0.000068 0.00000 avx_ChirpData_c 0.002006 0.00000 v_vTranspose4x16ntw 0.000689 0.00000 AK SSE folding 0.000934 0.00000 v_vGetPowerSpectrumUnrolled 0.000052 0.00000 avx_ChirpData_a 0.001763 0.00000 v_vTranspose4x16ntw 0.000726 0.00000 BH SSE folding 0.000363 0.00000 <<<<<<< almost 3-times faster than prev 2 samples x86 selected: Optimal function choices: -------------------------------------------------------- name timing error -------------------------------------------------------- v_BaseLineSmooth (no other) v_avxGetPowerSpectrum 0.000041 0.00000 avx_ChirpData_d 0.002374 0.00000 v_avxTranspose4x16ntw 0.000971 0.00000 JS AVX_a folding 0.000440 0.00000 v_avxGetPowerSpectrum 0.000044 0.00000 avx_ChirpData_b 0.001798 0.00000 v_avxTranspose4x16ntw 0.001004 0.00000 JS AVX_a folding 0.000270 0.00000 <<<<<<<<<<quite different timing v_vGetPowerSpectrumUnrolled 0.000037 0.00000 avx_ChirpData_d 0.001988 0.00000 v_avxTranspose4x16ntw 0.000849 0.00000 JS AVX_a folding 0.000461 0.00000 Hm... very weird choice for x64. Could you run few more (2-3 each) but with app_config.xml-supplied -verb command line option to both x64 and x86 apps? I would say this small sample set shows that inner benchmarking screwed enough in real-life conditions. Host load too inconsistent for such short bench in the beginning could pick up right function. SETI apps news We're not gonna fight them. We're gonna transcend them. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Why is that? AVX is normally faster on this CPU. Look edited post. Seems embedded benchmark can't cope correctly with long-term (regarding its duration) host load fluctuations. SETI apps news We're not gonna fight them. We're gonna transcend them. |
Michel Makhlouta Send message Joined: 21 Dec 03 Posts: 169 Credit: 41,799,743 RAC: 0 |
do i need to make the same changes? i didn't compare with x86 but i think it is slower... i have 4 cores dedicated to SoG and 4 cores running CPU WU, i7 4770K clocked at 4GHz |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
I suppose smth like this: <app_config> <app> <name>setiathome_v8</name> </app> <app_version> <app_name>setiathome_v8</app_name> <cmdline>-verb</cmdline> </app_version> </app_config> SETI apps news We're not gonna fight them. We're gonna transcend them. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
However, the same should apply for the X86 app. It applies indeed. Its functions choice fluctuating too. Maybe in less degree for that those few tasks I selected randomly, but.. SETI apps news We're not gonna fight them. We're gonna transcend them. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
It's actually SSE2 one. SETI apps news We're not gonna fight them. We're gonna transcend them. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
OK, when you will have some time allow 2-3 tasks with 8.04 in current config then edit -verb to -v I'll make sure next build will accept both variants... EDIT: maybe this not because app don't understand -verb but because x64 deployed under x64 platform so separate entry app_config.xml required for it?... Hope some more skilled "BOINC-admins" will suggest those missed lines. SETI apps news We're not gonna fight them. We're gonna transcend them. |
Juha Send message Joined: 7 Mar 04 Posts: 388 Credit: 1,857,738 RAC: 0 |
This works: <app_config> <app_version> <app_name>setiathome_v8</app_name> <cmdline>--verb</cmdline> </app_version> <app_version> <app_name>setiathome_v8</app_name> <plan_class>sse3</plan_class> <cmdline>--verb</cmdline> </app_version> </app_config> 8.05 comes under sse3 plan class so it's needed in app_config.xml too. Now, what's so special about JS AVX_c folding that it always fails on the x64 build but never on the x32 build? |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
There are enough tasks for this particualr host processed, thanks. Seems the most damaged part is folding. No AVX for it in x64 and one another fold function actualy fails (too big error). Will see if changing compiler could improve situation. Also I'm collecting AMD-related results still. SETI apps news We're not gonna fight them. We're gonna transcend them. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
As soon as I build it and Eric deploy it. That is, nothing definitive. SETI apps news We're not gonna fight them. We're gonna transcend them. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Ah, perhaps that's graphics-enabled one. So what is worth to try is to see if it can show SETI graphics if requested. Nothing else versus prev one. EDIT: or just same binary re-issue with SSE3+ only lifted. SETI apps news We're not gonna fight them. We're gonna transcend them. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Found time to test graphics on x64 build - it's work OK, just as with x86 one. Speed very close to x86 though. SETI apps news We're not gonna fight them. We're gonna transcend them. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Just take some patience and wait while box will be filled with data. On my x64 PC it happened for some reason... SETI apps news We're not gonna fight them. We're gonna transcend them. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Attempted to use clang toolchain from VS2015. Even after dealing with IUnknown issue it can't compile SETI code: 'void __cdecl _mm256_maskstore_ps(float * __ptr64,union __m256i,union __m256)': Intrinsic not yet implemented: So, only VC++ so far. SETI apps news We're not gonna fight them. We're gonna transcend them. |
Graham Thomas Send message Joined: 19 May 99 Posts: 32 Credit: 3,880,875 RAC: 4 |
This thread's been quiet for a while. I looked at it because I saw yesterday that I've got a number of 8.05 units. I'm not involved in beta-testing, so can I assume that the x64 application is now out of beta? I've got an old (bought 2011) 64-bit AMD PC, so it will be interesting to see how well it handles the new app. By the way, I'm really pleased with the 8.22 app for my GPU. With previous versions I had to tweak the relevant mb_cmdline_8.xx... files to stop my display from crashing and units locking up after recovery (thanks again for your help with that, Raistmer). After the mess with 8.23 and erroneous nvidia units I reset the project and then forgot to re-tweak the mb_cmdline_8.22... files - but everything's been running fine. I don't know what changed, but I'm glad it did and hope it continues with future versions. Graham Thomas |
Mike Send message Joined: 17 Feb 01 Posts: 34255 Credit: 79,922,639 RAC: 80 |
Yes, correct. With each crime and every kindness we birth our future. |
Ghia Send message Joined: 7 Feb 17 Posts: 238 Credit: 28,911,438 RAC: 50 |
The new v.8.05 is running fine here, but seems slower than v. 8.00.... ...Grete... Humans may rule the world...but bacteria run it... |
Mike Send message Joined: 17 Feb 01 Posts: 34255 Credit: 79,922,639 RAC: 80 |
The new v.8.05 is running fine here, but seems slower than v. 8.00.... What i have seen its approx 1000 seconds faster than 8.00 on your host. With each crime and every kindness we birth our future. |
©2024 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.