Am 05.08.2016 um 13:41 schrieb Mauro Carvalho Chehab <mche...@s-opensource.com>:
> Em Fri, 5 Aug 2016 11:56:44 +0200 > Markus Heiser <markus.hei...@darmarit.de> escreveu: > >> Am 20.07.2016 um 16:04 schrieb Mauro Carvalho Chehab >> <mche...@s-opensource.com>: >> >>> >>> A completely unrelated question: it seems that Sphinx is using just >>> one CPU to do its builds: >>> >>> %Cpu0 : 3,0 us, 7,6 sy, 0,0 ni, 89,4 id, 0,0 wa, 0,0 hi, 0,0 si, >>> 0,0 st >>> %Cpu1 :100,0 us, 0,0 sy, 0,0 ni, 0,0 id, 0,0 wa, 0,0 hi, 0,0 si, >>> 0,0 st >>> %Cpu2 : 1,3 us, 2,7 sy, 0,0 ni, 95,7 id, 0,3 wa, 0,0 hi, 0,0 si, >>> 0,0 st >>> %Cpu3 : 1,0 us, 3,3 sy, 0,0 ni, 95,7 id, 0,0 wa, 0,0 hi, 0,0 si, >>> 0,0 st >>> KiB Mem : 15861876 total, 5809820 free, 1750528 used, 8301528 buff/cache >>> KiB Swap: 8200188 total, 8200188 free, 0 used. 13382964 avail Mem >>> >>> PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND >>> >>> 5660 mchehab 20 0 325256 89776 8300 R 99,7 0,6 0:22.25 >>> sphinx-bui+ >>> >>> Are there any way to speed it up and make it use all available CPUs? >> >> Hi Mauro, >> >> sorry for the late reply. There is a sphinx-build option "-j N" [1]. >> It is in a *experimental* state in Sphinx v1.2 and has been improved >> in v1.3. Set e.g. "-j2" to the SPHINXOPTS to use two cores. >> >> make SPHINXOPTS=-j2 htmldocs >> >> But take into account what the documentation says: """not all parts and >> not all builders of Sphinx can be parallelized.""". >> >> [1] >> http://www.sphinx-doc.org/en/stable/invocation.html#cmdoption-sphinx-build-j > > Good, thanks! > > Did some tests here on a machine with 32 CPU threads using a PCIe SSD disk, > using Sphinx 1.4.5. > > Using -j32, those are the timings: > > real 0m59.522s > user 1m29.968s > sys 0m4.975s > > not using it, I got: > > real 1m27.814s > user 1m26.465s > sys 0m1.842s > > Not much gain :( > > Regards, > Mauro I think, that only some reading / writing parts of the sphinx implementation are parallel and most of the stuff is sequential. I haven't looked through but I think, generating env and traversing trees etc. are always steps which has to be consecutively. I guess you will get same results with 2 or 4 threads :( --Markus-- -- To unsubscribe from this list: send the line "unsubscribe linux-doc" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
