Re: Best way to store postfix data?

Look into based pointers. You can create a structure where you keep the symbol table as a
based-pointer table (relative offsets) and the executable tree as a based pointer system,
and consequently you can write the data out directly. Even if the data is not originally
contiguous, you can write it out maintaining the relative pointer positions and therefore
you can perfectly well save the relative pointers to the symbol table, then read it back
in. Based pointers allow the pain-free conversion from relative pointers to absolute
pointers.

You can also use based pointers to address an array of proxies to the actual symbol
values.
joe

On Mon, 02 Jul 2007 06:35:03 -0700, Scoots <bssalmons@xxxxxxxxxxxx> wrote:

I agree with you. In fact, I have a tree based logic that is quite
fast (can read, parse, error check, and execute complex statements
with control logic in the same time as the old read one line, execute
one line (only simplest statements) engine.

But unfortunately, my boss wants me to try this approach and then
compare them. I haven't optimized my tree approach (at the moment, it
is a memory hog. 20,000 lines of code was spiking it from ~15mbs usage
to ~60 mbs usage. There's some obvious things I can remove to drop
that to ~40mbs, but there's some memory hog I've missed in there.
Probably my method for building my assign logic. It's fast, but it's
creating objects for each element of an assign statement. Very fast,
but inefficient. As for any memory fragmentation, I could use
CFixedAlloc at the cost of slightly higher memory cost requirements,
but I'm willing to try other solutions.

The reason for the two maps is fairly simple. As I'm building the
logic, it CAN grow. As of right now, my 'instructions' are the same
size if I can move the logic and assignments to virtual 'heap' space.
To me, that seems infinitely preferable to trying to store them inline
and recording the size of the block. Maybe not. Storing references
to variables is no longer an issue, as we're going to be saving the
files so that if the saved preprocessed version and the file share the
same timestamp, we just load the preprocessed. This means I can't
store the variable locations.

I was storing the assignment logic as a string, not the variables.
Those are already in the engine. Instead, I'm trying to come up with
the most efficient way of taking these instructions in, and storing
them. Since I allow control routines, I have to execute at runtime,
and hence, have to build a logic for the program. Post-fix notation
seems like it would help a lot here, as I have both single arguement
functions and double arguement functions. Since I can't guarantee
that the variable addresses will stay the same between runs, I don't
think I can store them (also, erm, they aren't simple variables. I
can't create a simple reference to it. It's an inherited problem from
code written 15 years ago.).

In short, I'm not trying to store the symbol table as strings, just my
assignment statements. Variables are stored entirely separate from
the engine, and I can not change that.

An efficient way of doing this interpretation is just stumping me. I
definately appreciate your help, and I'll be happy to answer any
questions you have or try anything you suggest. Unfortunately, I'm
only an intern, with less than a year of c++ experience (though I've
been coding for much longer than that), so I might come up with some
truly monumentally odd ways of solving things. I'm not opposed to
learning new things. Thanks for the help.
Joseph M. Newcomer [MVP]
email: newcomer@xxxxxxxxxxxx
Web: http://www.flounder.com
MVP Tips: http://www.flounder.com/mvp_tips.htm
.

Relevant Pages

  • Re: Best way to store postfix data?
    ... based-pointer table and the executable tree as a based pointer system, ... Based pointers allow the pain-free conversion from relative pointers to absolute ... with control logic in the same time as the old read one line, execute ... but there's some memory hog I've missed in there. ...
    (microsoft.public.vc.mfc)
  • c++ : a baseclass, derived classes, a list of baseclass pointers, a problem.
    ... virtual void func1 ... Now i created a list of pointers to insances of the baseclass ... How can I execute the right function? ... I use a pointer to the baseclass to loop trough the list. ...
    (comp.programming)
  • Re: Making an ActiveX control out of an executable
    ... Microsoft MVP, MCSD ... >>be coded to execute standalone as well. ... >>message handling differs. ... Thanks for the pointers. ...
    (microsoft.public.vc.atl)
  • Re: IP Level Encryption (kind of long)
    ... >> pointers to executable code like Henrick mentioned. ... >something to execute that file. ... That's one way to fix it. ... it would be to prevent the buffer overflow in the first place ...
    (sci.crypt)
  • Re: Best way to store postfix data?
    ... with control logic in the same time as the old read one line, execute ... but there's some memory hog I've missed in there. ... that seems infinitely preferable to trying to store them inline ... Since I allow control routines, I have to execute at runtime, ...
    (microsoft.public.vc.mfc)