[tex-live] strange discrepancy in running time of etex between TL2015 and TL2017

jfbu jfbu at free.fr
Thu Jul 27 08:58:43 CEST 2017

Hi Reinhard

Le 27 juil. 2017 at02:15, Reinhard Kotucha  :

> On 2017-07-26 at 22:00:30 +0200, jfbu wrote:
>> 3. run in this work repertory the test file which I will indicate
>> next via
>> "time etex testmachin.tex"
>> - with TL2017
>> - with TL2015
> Dear Jean-Francois,
> I get
>  TL2017:  computed in 9.27 seconds.
>  TL2015:  computed in 9.21 seconds.
> No significant difference, but in both cases I get the error message
>  ! Undefined control sequence.
>  l.31 \ifx\x\Z \message{OK}\else \ERROR
>                                        \fi
> and thus the return value of "time" isn't usable.  What went wrong?

Thanks a lot for trying out, the most likely cause is that the 


line got hard-wrapped somehow so that it occupies multiple lines

(it should be on only one line)

and this means the \x has spaces in it and the \ifx test fails
because the \Z computed by \fdef\Z {\Machin {1000}} contains
the first 1000 digits of Pi with no spaces.

I really should not have left that piece of code in there, my wrong.

It makes sense as an error check, but as I validated it once at the
time of 1.2k upload to CTAN a few months back,
it was superfluous to keep it here

(I try again here xint 1.2k. OK, no issue on my side)

> My system is an AMD 2.4 GHz CPU running Gentoo Linux and TeX binaries
> are from tug.org.
> I must admit that benchmarks are confusing me ATM.  I've written a
> tiny program in several programming languages which does nothing else
> than to multiply numbers in nested for-loops.  On my old AMD machine,
> luajit was as fast as C.  On my brand-new Intel i7-7700K luajit is
> twice as fast as C.  Don't know why.  The CPU clock frequency is
> obviously not all that counts.

I have no idea: perhaps the luajit was compiled with top-notch
machines and software and performs at its best in such contexts?



More information about the tex-live mailing list