Hi Stephen et al,
On 8/1/06, Stephen J. Turnbull <stephen(a)xemacs.org> wrote:
>>>>> "vin" == Vin Shelton
vin> Does anyone think the below patch will have a noticeable
vin> effect on the speed byte compiler?
I imagine it will be *measurable* (eg, by profiling the
byte-compilation phase of a build) but probably not noticable.
Interestingly enough, when I edited my original message, I changed
"measurable" to "noticeable", so I was aware of the distinction when
asked the question. :-)
vin> Barring any concerns, I will check this in for 21.4.20.
The main concern I have is that Aidan admits that the code hasn't been
audited for memory leaks, which is always a concern when moving from
alloca (which is effectively garbage-collected by the stack
discipline) to malloc. OTOH, byte-compilation is a very rare event;
any leaks probably will not be noticable.
I guess I should have mentioned that I did the analysis and there are
no alternate return paths from the particular routine in question, so
the malloc'ed space is always free'd. My main concern was around
performance, but as you point out, the byte compiler is used somewhat
rarely, so I guess I shouldn't worry.
Whoever you are, no matter how lonely,
the world offers itself to your imagination,
calls to you like the wild geese, harsh and exciting--
over and over announcing your place
in the family of things. Mary Oliver