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

[jfbu]

Le 29 juil. 2017 à 02:07, Reinhard Kotucha <reinhard.kotucha at web.de> a écrit :

> On 2017-07-28 at 09:15:32 +0200, jfbu wrote:
> 
>> [...]
>> minutes: for very lengthy things (5mns+) my laptop *slows down* in
>> comparison to the desktop or the Linux machine
> 
> Hi Jean-Francois,
> the CPU fan is obviously not able to keep the CPU temperature below a
> certain limit.  In order to prevent the CPU from being destroyed by
> heat, the clock frequency is reduced.  This is intended and expected
> behavior.
> 
> Power consumption increases linearly with clock frequency.  If your
> CPU gets too hot, your system has only two choices: either reduce the
> clock frequency or shutdown the system completely.


Hi Reinhard

somehow, I sort of did not quite think it out this way because
I observed the phenomenon with computations lasting of the order
of one minute, but my fan gets started for longer computations
of at least seven or eight minutes. But I think you must
be right. (anyway there must be _some_ explanation ;-) I was
thinking perhaps some virtual memory thing, but I have an SSD
only on my laptop)


> 
> You can check yourself:  Before starting your program, open another
> terminal window and run
> 
>  while true; do clear; grep 'cpu MHz' /proc/cpuinfo; sleep 1; done
> 
> therein.  This works at least on Linux.  On OS X you probably have to
> run "sysctl -a" and check whether the current CPU frequencies are
> reported.  I don't have OS X.

yes sysctl -a seems report quite a bunch of cpu information

I see things like

hw.cpufamily: 526772277
hw.cpufrequency: 2000000000
hw.cpufrequency_min: 2000000000
hw.cpufrequency_max: 2000000000
hw.cputhreadtype: 1

machdep.cpu.brand_string: Intel(R) Core(TM) i7-3667U CPU @ 2.00GHz

machdep.cpu.thermal.sensor: 1
machdep.cpu.thermal.dynamic_acceleration: 1
machdep.cpu.thermal.invariant_APIC_timer: 1
machdep.cpu.thermal.thresholds: 2
machdep.cpu.thermal.ACNT_MCNT: 1
machdep.cpu.thermal.core_power_limits: 1
machdep.cpu.thermal.fine_grain_clock_mod: 1
machdep.cpu.thermal.package_thermal_intr: 1
machdep.cpu.thermal.hardware_feedback: 0

I have not tried yet to use dynamically

> 
> I suppose that the CPU frequencies are quite low at the beginnig, when
> no other processes are running.  When you start your program, at least
> one CPU will run at full speed.  After a while, when your system gets
> slower, you'll see that CPU frequencies are decreased.  You'll also
> notice that the air coming from the CPU fan is extremely hot.


I will try to investigate if CPU frequency is adapted

> 
> The problem is that laptops are almost as powerful as desktop PCs but
> there is not enough space for a decent cooling system.  This is quite
> unfortunate because lifetime of electronic components heavily depend
> on temperature.


> 
> BTW, I encountered a peculiarity with my Linux machine a few years
> ago.  I could do everything with this machine -- except compiling gcc.
> Whenever I invoked "./configure && make -j3", the power supply turned
> off after a few minutes.  In this case the power supply was too weak
> and replacing it by a bigger one solved the problem.

Lucky you not with Mac: you can replace things.

Here my batteries are 5 years old and changing them is way overdue
but I must bring the thing to an Apple Store, for this I need
to make an appointment, then they want to charge me something
like 200 euros (don't recall the details; but definitely it
is something like at least 50 euros for the physical replacement
not the price of batteries; and if I wait too long possibly
the battery model will not exist anymore).

Best

Jean-Francois