[Gcl-devel] Re: [Axiom-developer] Re: counting allocated bytes

[Stephen Wilson]

Hi Gregory,

Gregory Vanuxem <address@hidden> writes:

> Hello Stephen and Camm,
> 
> I'm particularly interested by this addition. Can it be considered as
> comparable to get-byte-consed in SBCL or CMUCL ?

I do not belive so.  IIRC get-bytes-consed returns an increasing
sequence of integers, effectively counting the number of allocations
performed.

I was actually aiming for somthing similar, but due to my
inexperience with GCL's internals I clearly got something wrong.  I
was using si::allocated as a guide.  The sixth value is the number of
uses of the given type.  Compare the following:

>(si:allocated 'cons)

204399
819
819
512
0
4503032

>(si:total-bytes-allocated)

3296240

>(defvar *c* (make-list 10000))

*C*

>(si:allocated 'cons)

194336
819
819
512
0
4513095

>(si:total-bytes-allocated)

3377336

>(setq *c* nil)

NIL

>(gbc t)

T

>(si:allocated 'cons)

204527
819
819
512
0
4513143

>(si:total-bytes-allocated)

3293360


Notice how si::allocated is returning an increasing integer sequence
in its sixth value,  whereas total-bytes-allocated is, as Camm
mentioned, a report commensurate with (room).


Camm, is it possible that tm_nused is providing the type of
information  necessary to get an accurate picture of how much a
function is consing, predictably increasing even amidst gc's?

Thanks,
Steve