For a fair comparison of KJT benchmark, future %. need to be excluded from blas.
Sent from my iPhone On 28 Aug, 2017, at 7:57 PM, Roger Hui <[email protected]> wrote: > See also http://code.jsoftware.com/wiki/Essays/JKT_Benchmark > > On Aug 28, 2017 03:17, "Joey K Tuttle" <[email protected]> wrote: > >> I have installed the new beta on my MacBook and on a Ubuntu server. No >> particular problems, but some fiddling around was required to get jQt >> working on the Mac. I only use ssh access to the Linux server, so that >> removes the fiddling. >> >> I decided to run my favorite benchmark (inverting a matrix). I have been >> doing this for about 47 years - starting with APLSV running on a 360 Model >> 50 with 32Kbyte (woo hoo!) workspaces... As I recall, the inverse of a 20 >> 20 matrix took several seconds and maxed out the available workspace. Kind >> of puts into perspective where processing (and memory) has gone. Anyway, >> here are some results - >> >> jkt@set1:~$ ja >> JVERSION >> Engine: j805/j64/linux >> Release: commercial/2016-12-11T08:02:52 >> Library: 8.05.14 >> Platform: Linux 64 >> Installer: J805 install >> InstallPath: /usr/local/lib/j64-805 >> Contact: www.jsoftware.com >> >> 5 (6!:2) 'mi=. %. 1000 1000 ?.@$ 0' >> 2.71839 >> >> NB. Not too shabby >> >> (3!:1 mi) fwrite 'v805' NB. store the result for later comparison >> 8000048 >> >> exit 0 >> jkt@set1:~$ >> jkt@set1:~$ jb >> JVERSION >> Engine: j806/j64/linux >> Beta-5: commercial/2017-08-23T10:33:49 >> Library: 8.06.06 >> Platform: Linux 64 >> Installer: J806 install >> InstallPath: /usr/local/lib/j64-806 >> Contact: www.jsoftware.com >> >> 5 (6!:2) 'mi=. %. 1000 1000 ?.@$ 0' >> 0.97644 >> >> NB. Whoa! J806 more than twice as fast as J805 on my favorite benchmark. >> >> (3!:1 mi) fwrite 'v806' >> 8000048 >> v5=. fread 'v805' >> v6=. fread 'v806' >> v5 -: v6 >> 0 >> NB. Well, the results aren't identical - but they are probably "very >> close" ... More research required. >> >> exit 0 >> jkt@set1:~$ cat /proc/cpuinfo >> processor : 0 >> vendor_id : GenuineIntel >> cpu family : 6 >> model : 30 >> model name : Intel(R) Xeon(R) CPU X3430 @ 2.40GHz >> stepping : 5 >> microcode : 0x3 >> cpu MHz : 1200.000 >> cache size : 8192 KB >> physical id : 0 >> siblings : 4 >> core id : 0 >> cpu cores : 4 >> apicid : 0 >> initial apicid : 0 >> fpu : yes >> fpu_exception : yes >> cpuid level : 11 >> wp : yes >> flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca >> cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx >> rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology >> nonstop_tsc aperfmperf pni dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 >> xtpr pdcm sse4_1 sse4_2 popcnt lahf_lm ida dtherm tpr_shadow vnmi >> flexpriority ept vpid >> bogomips : 4800.60 >> clflush size : 64 >> cache_alignment : 64 >> address sizes : 36 bits physical, 48 bits virtual >> power management: >> .... >> >> The other 3 cores are the same, but irrelevant because j is only using 1 >> anyway... >> >> Since matrix inverse is implemented essentially in j (I presume that is >> still true), I'm guessing that the speed up comes from better copy and >> memory management and that will have a nice impact on a lot of systems! >> >> Congrats on the improvements. >> >> - joey >> >>> On 2017Aug 24, at 05:59, Eric Iverson <[email protected]> wrote: >>> >>> New zip installers are available for win 32/64, linux 32/64, and mac 64 >>> built with the latest source. >> >> ---------------------------------------------------------------------- >> For information about J forums see http://www.jsoftware.com/forums.htm > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
