[tex-live] Request for performance check of legacy darwin (luatex)

jfbu jfbu at free.fr
Fri Mar 9 09:30:40 CET 2018


Hi Reinhard,

Le 8 mars 2018 à 23:47, Reinhard Kotucha <reinhard.kotucha at web.de> a écrit :

> On 2018-03-08 at 14:38:13 +0100, jfbu wrote:
>> $ time luatex temptest.tex
>> This is LuaTeX, Version 1.0.4 (TeX Live 2017/MacPorts 2017_1) 
>> restricted system commands enabled.
>> (./temptest.tex . . . . . . . . . .)
>> warning  (pdf backend): no pages of output.
>> Transcript written on temptest.log.
>> 
>> real	0m1.745s
>> user	0m1.716s
>> sys	0m0.025s
>> 
>> $ time luatex temptest.tex
>> This is LuaTeX, Version 1.07.0 (TeX Live 2018) 
>> restricted system commands enabled.
>> (./temptest.tex . . . . . . . . . .)
>> warning  (pdf backend): no pages of output.
>> Transcript written on temptest.log.
>> 
>> real	0m2.618s
>> user	0m2.593s
>> sys	0m0.020s
> 
> Just for the record, these numbers are meaningless if you don't clear
> the filesystem cache before each run.
> 
> When I do
> 
>  time cat somefile > /dev/null
> 
> then the file is read from the disk with abt. 150 MB/s.
> 
> When I run the same command again, the file is read from the
> filesystem cache with 8.5 GB/s.
> 
> In order to get comparable results it's necessary to wipe out the
> cache before running any tests.  The easiest way to accomplish this on
> Linux (at least) is to run the command
> 
>  sync; echo 3 > /proc/sys/vm/drop_caches
> 
> as root.  If you have a system monitor installed on your system
> (xosview, for instance), you can see immediately what happens.
> 

Thanks for advice,
What you say applies when file system operations are significant
for the timed process. In the case at hand I think CPU being 
busy with other things impacts more.

Anyway  I was careful to report
timings which are typical from having done these things
**many many times**, and in **random order**. I did get
some anomalous values, but the overall feeling is that
the timings I reported are clearly significative of some
underlying non-fake truth, in short, I take exception
with the indication that they are meaningless

Best,

Jean-François



More information about the tex-live mailing list