Message boards :
Number crunching :
Linux (ARM processor) app and alternatives
Message board moderation
Previous · 1 . . . 3 · 4 · 5 · 6 · 7 · 8 · Next
Author | Message |
---|---|
sorcrosc Send message Joined: 13 Dec 08 Posts: 3 Credit: 2,374,066 RAC: 0 |
Thanks Raistmer. What about targeting v7 instead of v6? |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Thanks Raistmer. What about targeting v7 instead of v6? Well, 8.04 not my build so "ideally" doesn't mean it's really how it is (cause I just don't know how it is in 8.04) Maybe worth to get 2 binaries instead of one if performance gain for newer chips is great enough. As I understand v7 has different ISA than v6? Then it can be the reason of failure too, not just NEON... SETI apps news We're not gonna fight them. We're gonna transcend them. |
sorcrosc Send message Joined: 13 Dec 08 Posts: 3 Credit: 2,374,066 RAC: 0 |
Ok, I tried 8.02 from seti which reports v7 target and neon too, and it works. Sorry for my useless intrusion :) |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
KWSN-Linux-MBbench v3.0 cache-keeping edition Running on parallella at Sun 12 Feb 2017 04:27:15 PM UTC ---------------------------------------------------------------- Starting benchmark run... ---------------------------------------------------------------- Suspending BOINC Listing wu-file(s) in /testWUs : #WisGen1_v8.wu #WisGen2_v8.wu PG0009_tiny_v8.wu PG0395_tiny_v8.wu PG0444_tiny_v8.wu PG1327_tiny_v8.wu Listing executable(s) in /APPS : setiathome_8.02_arm-unknown-linux-gnueabihf Listing executable in /REF_APPS : setiathome_8.02_arm-unknown-linux-gnueabihf ---------------------------------------------------------------- Current WU: #WisGen1_v8.wu ---------------------------------------------------------------- Skipping default app setiathome_8.02_arm-unknown-linux-gnueabihf, displaying saved result(s) Elapsed Time: ....................... 43 seconds ---------------------------------------------------------------- Running app with command : .......... setiathome_8.02_arm-unknown-linux-gnueabihf -verb ./setiathome_8.02_arm-unknown-linux-gnueabihf -verb 44.24 sec 39.16 sec 2.34 sec Elapsed Time : ...................... 44 seconds Speed compared to default : ......... 97 % ----------------- Comparing results Result : Strongly similar, Q= 100.0% ---------------------------------------------------------------- Done with #WisGen1_v8.wu ==================================================================== Current WU: #WisGen2_v8.wu ---------------------------------------------------------------- Skipping default app setiathome_8.02_arm-unknown-linux-gnueabihf, displaying saved result(s) Elapsed Time: ....................... 41 seconds ---------------------------------------------------------------- Running app with command : .......... setiathome_8.02_arm-unknown-linux-gnueabihf -verb ./setiathome_8.02_arm-unknown-linux-gnueabihf -verb 41.41 sec 36.57 sec 2.78 sec Elapsed Time : ...................... 42 seconds Speed compared to default : ......... 97 % ----------------- Comparing results Result : Strongly similar, Q= 100.0% ---------------------------------------------------------------- Done with #WisGen2_v8.wu ==================================================================== Current WU: PG0009_tiny_v8.wu ---------------------------------------------------------------- Running default app with command :... setiathome_8.02_arm-unknown-linux-gnueabihf -verb ./setiathome_8.02_arm-unknown-linux-gnueabihf -verb 1973.78 sec 1955.63 sec 15.77 sec Elapsed Time: ....................... 1974 seconds ---------------------------------------------------------------- Running app with command : .......... setiathome_8.02_arm-unknown-linux-gnueabihf -verb ./setiathome_8.02_arm-unknown-linux-gnueabihf -verb 1968.85 sec 1949.63 sec 17.09 sec Elapsed Time : ...................... 1969 seconds Speed compared to default : ......... 100 % ----------------- Comparing results Result : Strongly similar, Q= 100.0% ---------------------------------------------------------------- Done with PG0009_tiny_v8.wu ==================================================================== Current WU: PG0395_tiny_v8.wu ---------------------------------------------------------------- Running default app with command :... setiathome_8.02_arm-unknown-linux-gnueabihf -verb ./setiathome_8.02_arm-unknown-linux-gnueabihf -verb 2281.79 sec 2262.67 sec 16.98 sec Elapsed Time: ....................... 2282 seconds ---------------------------------------------------------------- Running app with command : .......... setiathome_8.02_arm-unknown-linux-gnueabihf -verb ./setiathome_8.02_arm-unknown-linux-gnueabihf -verb 2317.32 sec 2298.56 sec 16.55 sec Elapsed Time : ...................... 2318 seconds Speed compared to default : ......... 98 % ----------------- Comparing results Result : Strongly similar, Q= 100.0% ---------------------------------------------------------------- Done with PG0395_tiny_v8.wu ==================================================================== Current WU: PG0444_tiny_v8.wu ---------------------------------------------------------------- Running default app with command :... setiathome_8.02_arm-unknown-linux-gnueabihf -verb ./setiathome_8.02_arm-unknown-linux-gnueabihf -verb 2139.51 sec 2118.51 sec 16.86 sec Elapsed Time: ....................... 2139 seconds ---------------------------------------------------------------- Running app with command : .......... setiathome_8.02_arm-unknown-linux-gnueabihf -verb ./setiathome_8.02_arm-unknown-linux-gnueabihf -verb 2119.59 sec 2098.61 sec 17.04 sec Elapsed Time : ...................... 2119 seconds Speed compared to default : ......... 100 % ----------------- Comparing results Result : Strongly similar, Q= 100.0% ---------------------------------------------------------------- Done with PG0444_tiny_v8.wu ==================================================================== Current WU: PG1327_tiny_v8.wu ---------------------------------------------------------------- Running default app with command :... setiathome_8.02_arm-unknown-linux-gnueabihf -verb ./setiathome_8.02_arm-unknown-linux-gnueabihf -verb 2439.85 sec 2399.87 sec 36.05 sec Elapsed Time: ....................... 2439 seconds ---------------------------------------------------------------- Running app with command : .......... setiathome_8.02_arm-unknown-linux-gnueabihf -verb ./setiathome_8.02_arm-unknown-linux-gnueabihf -verb 2433.24 sec 2394.51 sec 34.89 sec Elapsed Time : ...................... 2433 seconds Speed compared to default : ......... 100 % ----------------- Comparing results Result : Strongly similar, Q= 100.0% ---------------------------------------------------------------- Done with PG1327_tiny_v8.wu ==================================================================== Hosts CPU data ... model name : ARMv7 Processor rev 0 (v7l) Done with Benchmark run! Removing temporary files! Resuming BOINC These "tiny" more manageable for such slow devices. Next will be test with verbose new ones. SETI apps news We're not gonna fight them. We're gonna transcend them. |
Tom Rinehart Send message Joined: 12 Dec 01 Posts: 113 Credit: 13,255,975 RAC: 6 |
The errors on ARMv6 devices (Raspberry Pi 1s and Zeros) might be FFTW 3.3.6-pl1. 8.02 and 8.03 are built against the Debian repository FFTW library which includes automatic testing for if NEON is present. If NEON is present, it uses it. It looks like this is in FFTW 3.3.6-pl1, but maybe it is not. I will look to see if the Debian package has a patch that does this. |
Tom Rinehart Send message Joined: 12 Dec 01 Posts: 113 Credit: 13,255,975 RAC: 6 |
The errors on ARMv6 devices (Raspberry Pi 1s and Zeros) might be FFTW 3.3.6-pl1. 8.02 and 8.03 are built against the Debian repository FFTW library which includes automatic testing for if NEON is present. If NEON is present, it uses it. It looks like this is in FFTW 3.3.6-pl1, but maybe it is not. I will look to see if the Debian package has a patch that does this. The Debian package has a patch for FFTW that does proper NEON run-time detection. I've applied it and will test it to see if it works on my Raspberry Pi 1. |
Tom Rinehart Send message Joined: 12 Dec 01 Posts: 113 Credit: 13,255,975 RAC: 6 |
If you are building this app, this is the Debian patch for fftw-3.3.6-pl1/simd-support/neon.c: Description: Fix runtime detection of NEON support on armhf and arm64 Author: Edmund Grimley Evans <edmund.grimley.evans@gmail.com> Bug-Debian: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=767138 Forwarded: no Reviewed-by: Sébastien Villemot <sebastien@debian.org> Last-Update: 2016-09-20 --- This patch header follows DEP-3: http://dep.debian.net/deps/dep3/ --- a/simd-support/neon.c +++ b/simd-support/neon.c @@ -23,54 +23,45 @@ #if HAVE_NEON -/* check for an environment where signals are known to work */ -#if defined(unix) || defined(linux) - # include <signal.h> - # include <setjmp.h> - - static jmp_buf jb; - - static void sighandler(int x) - { - UNUSED(x); - longjmp(jb, 1); - } - - static int really_have_neon(void) - { - void (*oldsig)(int); - oldsig = signal(SIGILL, sighandler); - if (setjmp(jb)) { - signal(SIGILL, oldsig); - return 0; - } else { - /* paranoia: encode the instruction in binary because the - assembler may not recognize it without -mfpu=neon */ - /*asm volatile ("vand q0, q0, q0");*/ - asm volatile (".long 0xf2000150"); - signal(SIGILL, oldsig); - return 1; - } - } - - int X(have_simd_neon)(void) - { - static int init = 0, res; - - if (!init) { - res = really_have_neon(); - init = 1; - } - return res; - } +#ifdef __linux__ +#ifdef __aarch64__ + +/* HWCAP_ASIMD is defined in <asm/hwcap.h> but not included by <sys/auxv.h>. + Since all current AArch64 implementations have NEON/ASIMD it is probably + better to return 1 than include a header file which is not intended for + use by user programs. */ + +int X(have_simd_neon)(void) +{ + return 1; +} #else -/* don't know how to autodetect NEON; assume it is present */ - int X(have_simd_neon)(void) - { - return 1; - } + +#include <sys/auxv.h> + +int X(have_simd_neon)(void) +{ + static int cached = 2; + int ret; + + /* This should be thread-safe in all reasonable circumstances. */ + ret = cached; + if (ret == 2) + { + ret = !!(getauxval(AT_HWCAP) & HWCAP_ARM_NEON); + cached = ret; + } + return ret; +} + +#endif + +#else + +#error Please implement a run-time test for NEON/ASIMD for your platform. + #endif #endif |
Tom Rinehart Send message Joined: 12 Dec 01 Posts: 113 Credit: 13,255,975 RAC: 6 |
Success! This patch makes it so it runs correctly on a Raspberry Pi 1 (ARMv6 - no NEON). |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
The 'BUG: unsupported FP instruction in kernel mode' Bug hopefully has been fixed on the Original Raspberry Pi, anyone wishing to try it, and running it here on the Main project, can get the latest fixed kernel by doing: sudo apt-get install rpi-update sudo rpi-update If you get 4.4.48+ #964 or 4.9.9+ #965 or later then you'll have a kernel with the fix. 8.02 apps at the Main project should work no problem now, and 8.03 at Beta should too, 8.04 don't work on the Pi full stop, expect 8.05 soon hopefully. Claggy |
Tom Rinehart Send message Joined: 12 Dec 01 Posts: 113 Credit: 13,255,975 RAC: 6 |
Claggy - Thanks. I will update my Pi 1 and try it. I sent 8.05 to Eric this morning. Hopefully he will get it on Beta soon. - Tom |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
Claggy - There's a further posting on Beta about it, once that kernel or a later version comes out as a production kernel, then i'll get the Pi News thread unlocked and post there too. Claggy |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Here are results from PG tiny test: KWSN-Linux-MBbench v3.0 cache-keeping edition stderr.setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_default.PG0444_tiny_v8.wu.txt Optimal function choices: -------------------------------------------------------- name timing error -------------------------------------------------------- v_BaseLineSmooth (no other) v_GetPowerSpectrum 0.005215 0.00000 test vfp_GetPowerSpectrum 0.003225 0.00000 test neon_GetPowerSpectrum 0.003501 0.00000 test vfp_GetPowerSpectrum 0.003225 0.00000 choice v_ChirpData 0.178233 0.00000 test fpu_ChirpData 0.173405 1.51106 test fpu_opt_ChirpData 0.193583 0.00000 test v_ChirpData 0.178233 0.00000 choice v_Transpose 0.149689 0.00000 test v_Transpose2 0.077378 0.00000 test v_Transpose4 0.043121 0.00000 test v_Transpose8 0.058047 0.00000 test fftwf_transpose 0.027199 0.00000 test v_pfTranspose2 0.072181 0.00000 test v_pfTranspose4 0.046788 0.00000 test v_pfTranspose8 0.076512 0.00000 test v_vfpTranspose2 0.083128 0.00000 test fftwf_transpose 0.027199 0.00000 choice FPU opt folding 0.044374 0.00000 test opt VFP folding 0.034257 0.19827 test opt NEON folding 0.027513 0.00000 test opt NEON folding 0.027513 0.00000 choice Test duration 30.50 seconds stderr.setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_NEONchirp.PG0444_tiny_v8.wu.txt Optimal function choices: -------------------------------------------------------- name timing error -------------------------------------------------------- v_BaseLineSmooth (no other) v_GetPowerSpectrum 0.005769 0.00000 test vfp_GetPowerSpectrum 0.003613 0.00000 test neon_GetPowerSpectrum 0.003797 0.00000 test vfp_GetPowerSpectrum 0.003613 0.00000 choice v_ChirpData 0.191753 0.00000 test fpu_ChirpData 0.191053 1.51106 test fpu_opt_ChirpData 0.211642 0.00000 test neon_ChirpData 0.079536 0.00000 test neon_ChirpData 0.079536 0.00000 choice v_Transpose 0.163054 0.00000 test v_Transpose2 0.084624 0.00000 test v_Transpose4 0.047427 0.00000 test v_Transpose8 0.064182 0.00000 test fftwf_transpose 0.029658 0.00000 test v_pfTranspose2 0.079094 0.00000 test v_pfTranspose4 0.046948 0.00000 test v_pfTranspose8 0.076827 0.00000 test v_vfpTranspose2 0.083352 0.00000 test fftwf_transpose 0.029658 0.00000 choice FPU opt folding 0.044367 0.00000 test opt VFP folding 0.034358 0.19827 test opt NEON folding 0.027593 0.00000 test opt NEON folding 0.027593 0.00000 choice Test duration 36.68 seconds stderr.setiathome_8.04_arm-unknown-linux-gnueabihf.PG0444_tiny_v8.wu.txt Optimal function choices: -------------------------------------------------------- name timing error -------------------------------------------------------- v_BaseLineSmooth (no other) v_GetPowerSpectrum 0.005215 0.00000 test vfp_GetPowerSpectrum 0.003225 0.00000 test neon_GetPowerSpectrum 0.003557 0.00000 test vfp_GetPowerSpectrum 0.003225 0.00000 choice v_ChirpData 0.174492 0.00000 test fpu_ChirpData 0.184554 1.51106 test fpu_opt_ChirpData 0.180076 0.00000 test v_ChirpData 0.174492 0.00000 choice v_Transpose 0.149246 0.00000 test v_Transpose2 0.077268 0.00000 test v_Transpose4 0.043231 0.00000 test v_Transpose8 0.058766 0.00000 test fftwf_transpose 0.027126 0.00000 test v_pfTranspose2 0.072034 0.00000 test v_pfTranspose4 0.041720 0.00000 test v_pfTranspose8 0.068238 0.00000 test v_vfpTranspose2 0.076070 0.00000 test fftwf_transpose 0.027126 0.00000 choice FPU opt folding 0.038182 0.00000 test opt VFP folding 0.030074 0.19827 test opt NEON folding 0.023993 0.00000 test opt NEON folding 0.023993 0.00000 choice Test duration 31.46 seconds 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 |
Here is my binary: https://cloud.mail.ru/public/7sVS/sEgfNQqQb Uploaded from Windows host do don't forget to chmod for execution. Has sense to run it on NEON-enabled ARMs only cause linked vs unpatched FFTW + additional speedup comes directly from additional NEON-based function usage (ChirpData) 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 |
Seems have vfp_ChirpData working too, but for Parallella it benched slower than NEON one: before Chirp test: vfp_ChirpData in[0].xy=(1.00000,0.00000)] in[1].xy=(1.00000,0.00000)] in[2].xy=(1.00000,0.00001)] after Chirp test: vfp_ChirpData out[0].xy=(1.00000,0.00000)] out[1].xy=(1.00000,-0.00000)] out[2].xy=(1.00000,-0.00000)] vfp_ChirpData 0.084548 0.00000 test in[0].xy=(1.00000,0.00000)] in[1].xy=(1.00000,0.00000)] in[2].xy=(1.00000,0.00001)] after Chirp test: neon_ChirpData out[0].xy=(1.00000,0.00000)] out[1].xy=(1.00000,-0.00000)] out[2].xy=(1.00000,-0.00000)] neon_ChirpData 0.072051 0.00000 test neon_ChirpData 0.072051 0.00000 choice 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 |
What steps should I do to have FFTW3.3.6 patched and not fail on ARM w/o NEON chips? SETI apps news We're not gonna fight them. We're gonna transcend them. |
Tom Rinehart Send message Joined: 12 Dec 01 Posts: 113 Credit: 13,255,975 RAC: 6 |
What steps should I do to have FFTW3.3.6 patched and not fail on ARM w/o NEON chips? I just cut and paste removing the - lines and adding the + lines. The file was short enough. You can also use the Linux patch command: http://www.thegeekstuff.com/2014/12/patch-command-examples Then make clean make sudo make install |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Thanks! And here is VFP-only chirp testing: KWSN-Linux-MBbench v3.0 cache-keeping edition Running on parallella at Mon 13 Feb 2017 10:02:50 PM UTC ---------------------------------------------------------------- Starting benchmark run... ---------------------------------------------------------------- Suspending BOINC Listing wu-file(s) in /testWUs : #WisGen1_v8.wu #WisGen2_v8.wu PG0009_tiny_v8.wu PG0395_tiny_v8.wu PG0444_tiny_v8.wu PG1327_tiny_v8.wu Listing executable(s) in /APPS : setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_NEONchirp setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_VFPChirp_enabled_NEONchirp_disabled Listing executable in /REF_APPS : setiathome_8.02_arm-unknown-linux-gnueabihf ---------------------------------------------------------------- Current WU: #WisGen1_v8.wu ---------------------------------------------------------------- Skipping default app setiathome_8.02_arm-unknown-linux-gnueabihf, displaying saved result(s) Elapsed Time: ....................... 43 seconds ---------------------------------------------------------------- Running app with command : .......... setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_NEONchirp -verb ./setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_NEONchirp -verb 105.74 sec 100.42 sec 3.11 sec Elapsed Time : ...................... 105 seconds Speed compared to default : ......... 40 % ----------------- Comparing results Result : Strongly similar, Q= 99.99% ---------------------------------------------------------------- Running app with command : .......... setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_VFPChirp_enabled_NEONchirp_disabled ./setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_VFPChirp_enabled_NEONchirp_disabled 106.06 sec 100.62 sec 3.21 sec Elapsed Time : ...................... 106 seconds Speed compared to default : ......... 40 % ----------------- Comparing results Result : Strongly similar, Q= 99.99% ---------------------------------------------------------------- Done with #WisGen1_v8.wu ==================================================================== Current WU: #WisGen2_v8.wu ---------------------------------------------------------------- Skipping default app setiathome_8.02_arm-unknown-linux-gnueabihf, displaying saved result(s) Elapsed Time: ....................... 41 seconds ---------------------------------------------------------------- Running app with command : .......... setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_NEONchirp -verb ./setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_NEONchirp -verb 106.34 sec 100.25 sec 2.61 sec Elapsed Time : ...................... 107 seconds Speed compared to default : ......... 38 % ----------------- Comparing results Result : Strongly similar, Q= 99.99% ---------------------------------------------------------------- Running app with command : .......... setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_VFPChirp_enabled_NEONchirp_disabled ./setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_VFPChirp_enabled_NEONchirp_disabled 101.24 sec 96.28 sec 2.96 sec Elapsed Time : ...................... 101 seconds Speed compared to default : ......... 40 % ----------------- Comparing results Result : Strongly similar, Q= 99.99% ---------------------------------------------------------------- Done with #WisGen2_v8.wu ==================================================================== Current WU: PG0009_tiny_v8.wu ---------------------------------------------------------------- Skipping default app setiathome_8.02_arm-unknown-linux-gnueabihf, displaying saved result(s) Elapsed Time: ....................... 1974 seconds ---------------------------------------------------------------- Running app with command : .......... setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_NEONchirp -verb ./setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_NEONchirp -verb 1594.12 sec 1574.16 sec 16.37 sec Elapsed Time : ...................... 1594 seconds Speed compared to default : ......... 123 % ----------------- Comparing results Result : Strongly similar, Q= 99.99% ---------------------------------------------------------------- Running app with command : .......... setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_VFPChirp_enabled_NEONchirp_disabled ./setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_VFPChirp_enabled_NEONchirp_disabled 1582.39 sec 1564.17 sec 15.64 sec Elapsed Time : ...................... 1582 seconds Speed compared to default : ......... 124 % ----------------- Comparing results Result : Strongly similar, Q= 99.99% ---------------------------------------------------------------- Done with PG0009_tiny_v8.wu ==================================================================== Current WU: PG0395_tiny_v8.wu ---------------------------------------------------------------- Skipping default app setiathome_8.02_arm-unknown-linux-gnueabihf, displaying saved result(s) Elapsed Time: ....................... 2282 seconds ---------------------------------------------------------------- Running app with command : .......... setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_NEONchirp -verb ./setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_NEONchirp -verb 1854.03 sec 1835.67 sec 16.36 sec Elapsed Time : ...................... 1854 seconds Speed compared to default : ......... 123 % ----------------- Comparing results Result : Strongly similar, Q= 99.95% ---------------------------------------------------------------- Running app with command : .......... setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_VFPChirp_enabled_NEONchirp_disabled ./setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_VFPChirp_enabled_NEONchirp_disabled 1855.14 sec 1836.23 sec 16.12 sec Elapsed Time : ...................... 1855 seconds Speed compared to default : ......... 123 % ----------------- Comparing results Result : Strongly similar, Q= 99.95% ---------------------------------------------------------------- Done with PG0395_tiny_v8.wu ==================================================================== Current WU: PG0444_tiny_v8.wu ---------------------------------------------------------------- Skipping default app setiathome_8.02_arm-unknown-linux-gnueabihf, displaying saved result(s) Elapsed Time: ....................... 2139 seconds ---------------------------------------------------------------- Running app with command : .......... setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_NEONchirp -verb ./setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_NEONchirp -verb 1698.12 sec 1676.46 sec 16.17 sec Elapsed Time : ...................... 1698 seconds Speed compared to default : ......... 125 % ----------------- Comparing results Result : Strongly similar, Q= 99.95% ---------------------------------------------------------------- Running app with command : .......... setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_VFPChirp_enabled_NEONchirp_disabled ./setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_VFPChirp_enabled_NEONchirp_disabled 1740.50 sec 1721.98 sec 15.40 sec Elapsed Time : ...................... 1740 seconds Speed compared to default : ......... 122 % ----------------- Comparing results Result : Strongly similar, Q= 99.95% ---------------------------------------------------------------- Done with PG0444_tiny_v8.wu ==================================================================== Current WU: PG1327_tiny_v8.wu ---------------------------------------------------------------- Skipping default app setiathome_8.02_arm-unknown-linux-gnueabihf, displaying saved result(s) Elapsed Time: ....................... 2439 seconds ---------------------------------------------------------------- Running app with command : .......... setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_NEONchirp -verb ./setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_NEONchirp -verb 1496.14 sec 1458.92 sec 34.12 sec Elapsed Time : ...................... 1496 seconds Speed compared to default : ......... 163 % ----------------- Comparing results Result : Strongly similar, Q= 99.96% ---------------------------------------------------------------- Running app with command : .......... setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_VFPChirp_enabled_NEONchirp_disabled ./setiathome-8.0.armv7l-unknown-linux-gnueabihf_R_VFPChirp_enabled_NEONchirp_disabled 1538.38 sec 1500.97 sec 34.63 sec Elapsed Time : ...................... 1539 seconds Speed compared to default : ......... 158 % ----------------- Comparing results Result : Strongly similar, Q= 99.96% ---------------------------------------------------------------- Done with PG1327_tiny_v8.wu ==================================================================== Hosts CPU data ... model name : ARMv7 Processor rev 0 (v7l) Done with Benchmark run! Removing temporary files! Resuming BOINC Optimal function choices: -------------------------------------------------------- name timing error -------------------------------------------------------- v_BaseLineSmooth (no other) vfp_GetPowerSpectrum 0.003558 0.00000 vfp_ChirpData 0.093684 0.00000 fftwf_transpose 0.029237 0.00000 opt NEON folding 0.027707 0.00000 Slower than NEON one but better than generic. So I'll do commit and post corresponding build (after FFTW patching perhaps to allow non-NEON chips). BTW, look as chirp share increases with AR increase. This particular bench is very revealing IMHO. All other params but ChirpData are the same here. SETI apps news We're not gonna fight them. We're gonna transcend them. |
MarkJ Send message Joined: 17 Feb 08 Posts: 1139 Credit: 80,854,192 RAC: 5 |
Is that fixed chirp version going to appear on beta as 8.06 then? Currently still getting 8.04 on beta as stock app on Pi2 and Pi3. BOINC blog |
Tom Rinehart Send message Joined: 12 Dec 01 Posts: 113 Credit: 13,255,975 RAC: 6 |
Raistmer - Thanks for all your work on fixing the ARM app so it uses the VFP and NEON chirp functions. I will try building it and running it on my various ARM devices including my Raspberry Pi 1 with Claggy's patched kernel. - Tom |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Great, it saves me from FFTW patching cause Parallella prevers NEON anyway. SETI apps news We're not gonna fight them. We're gonna transcend them. |
©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.