/proc/cpuinfo bad cache info

/proc/cpuinfo bad cache info

[David Balazic]

Hi!

In recent 2.4.x kernels the "Cache: " line in /proc/cpuinfo
reports the amount of the L1 cache or L2 cache or L3 cache or
some combination of them, depending on what code is executed
for this ( different for different CPU types ).

Somebody should decide what information should be reported in that
line and then fix the code.

Party on !

-- 
David Balazic
--------------
"Be excellent to each other." - Bill S. Preston, Esq., & "Ted" Theodore Logan
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

Re: /proc/cpuinfo bad cache info

[H. Peter Anvin]

Followup to:  <3B98DD93.4BDD367C@uni-mb.si>
By author:    David Balazic <david.balazic@uni-mb.si>
In newsgroup: linux.dev.kernel
> 
> In recent 2.4.x kernels the "Cache: " line in /proc/cpuinfo
> reports the amount of the L1 cache or L2 cache or L3 cache or
> some combination of them, depending on what code is executed
> for this ( different for different CPU types ).
> 
> Somebody should decide what information should be reported in that
> line and then fix the code.
> 

We already DO report the information we care about -- the SMP
weighting value -- and thus the code is correct.  The value indicates
how much data is localized to that CPU and therefore how expensive it
is to reschedule a process elsewhere.

Anything that reports anything else is buggy.  This includes things
like adding in the L1 cache in an inclusive cache design, or reporting
the L3 cache when it is a shared cache in the chipset.

	-hpa
-- 
<hpa@transmeta.com> at work, <hpa@zytor.com> in private!
"Unix gives you enough rope to shoot yourself in the foot."
http://www.zytor.com/~hpa/puzzle.txt	<amsp@zytor.com>

Re: /proc/cpuinfo bad cache info

[Alan Cox]

> We already DO report the information we care about -- the SMP
> weighting value -- and thus the code is correct.  The value indicates
> how much data is localized to that CPU and therefore how expensive it
> is to reschedule a process elsewhere.

There are plenty of other reasons for wanting to know more, eg picking which
video encode tables to use - but I still agree they belong in user space
since cpuid is available from user space too