Yes, i did that, there are only few places where #ifdef is used, inside
these functions and inside ReturnRange
Zoran Vasiljevic wrote:
On 14.12.2006, at 04:26, Vlad Seryakov wrote:
I just replaced Tcl_FSStat in fastpath.c with stat and got about 5-9%
increase in serving 26K image file. I am m
On 14.12.2006, at 04:26, Vlad Seryakov wrote:
I just replaced Tcl_FSStat in fastpath.c with stat and got about 5-9%
increase in serving 26K image file. I am making #define which will
make
this compile time configurable, anyway i am in this already.
Before you scatter the code with defines
I do not think that this is the problem, really. I mean, not the
creation or disposal of *that* objects. The problem, if any, is
in the Tcl_FS* calls which do quite a bit of "arithmetic" to
fiddle with the internal path reps and stuff. It is this part that
*may* hit you. Again, if you have a test
I do not think that this is the problem, really. I mean, not the
creation or disposal of *that* objects. The problem, if any, is
in the Tcl_FS* calls which do quite a bit of "arithmetic" to
fiddle with the internal path reps and stuff. It is this part that
*may* hit you. Again, if you have a test
On 13.12.2006, at 17:11, Vlad Seryakov wrote:
Tcl_Obj needs to be created, then deallocated.
which is a fairly quick operation as it mostly just
removes the object from the per-thread list and puts
it back on that list when done.
Tcl_OpenFileChannel and other we use are obsolete in the Tcl
The parts i have doubts are those when Tcl file needs to be open or
something, Tcl_Obj needs to be created, then deallocated.
Tcl_OpenFileChannel and other we use are obsolete in the Tcl core
sources, internally, it creates String object and calls corresponding FS
calls. This is overhead if the
On 13.12.2006, at 16:54, Vlad Seryakov wrote:
Is it possible to have TclVfs code compile time or realtime
configurable?
I started that (configurable) way. Then we talked (I cannot recall
exactly when)
and decided to put it hard-wired all the time.
Instead of using direct open/read/writ
Zoran,
Is it possible to have TclVfs code compile time or realtime configurable?
Zoran Vasiljevic wrote:
On 12.12.2006, at 19:40, Stephen Deasey wrote:
Tcl not being the fastest kid on the block, and folklore saying you
should reboot your AOLserver/NaviServer once/day due to chronic memory
fr
On 12.12.2006, at 21:34, Stephen Deasey wrote:
Suggesting work for other people is easy and fun!
Heh!
But, really, the libumem hint is a good one indeed.
I will try this out ASAP and tell you what I get.
Zoran
On 12/12/06, Zoran Vasiljevic <[EMAIL PROTECTED]> wrote:
On 12.12.2006, at 19:40, Stephen Deasey wrote:
>
> Tcl not being the fastest kid on the block, and folklore saying you
> should reboot your AOLserver/NaviServer once/day due to chronic memory
> fragmentation, it would be nice if these tho
Google tmalloc, Hoard
I am going to try tmalloc
but if do not care about speed, just compile Tcl with default malloc,
glibc malloc even return memory to the system
Zoran Vasiljevic wrote:
On 12.12.2006, at 19:40, Stephen Deasey wrote:
Tcl not being the fastest kid on the block, and folklor
On 12.12.2006, at 19:40, Stephen Deasey wrote:
Tcl not being the fastest kid on the block, and folklore saying you
should reboot your AOLserver/NaviServer once/day due to chronic memory
fragmentation, it would be nice if these thousand node DOMs were
constructed once, at compile time, with one
12 matches
Mail list logo
