I stumbled across the following article on www.voidnish.com. It compares the differences between using the new / delete pair and the malloc / free pair. The 'Tip for scalar and vector new/delete' in the article helped me realize why my one DLL had a memory leak when handling char arrays.
Obviously a seasoned C++ programmer would know this, but me being a C# / Java programmer and only a C++ programmer when the situation calls for it, I did not realize that mixing the two ways of allocating memory and freeing them again will have such an effect.
Just another reason I love my C# / Java managed code world ;)
Friday, February 24, 2006
Metamodel Programming
I always used meta data to store business rules e.g. tax rates, commission etc. This allowed the system to be modified and even extended after delivery well into the maintenance phase without too much hassle.
Lately I am gravitating more and more to the idea of having all my logic inside a meta model. This meta model will not only contain business logic but also programmatic logic . The meta model must be able to model the interaction between all the various business objects as well as the presentation objects which will in turn allow me to generate code from the meta model.
There are tools available today allowing this, but they are either too expensive or too bulky. Obviously creating a product to do all of this, the product needs to be very generic to allow it to be usable by the mass market.
Because my projects are medium in size, 5 - 10 developer teams, I do not need the bulky huge be everything to everyone packages. The tools needed in my domain will be fewer but more biased towards the domain instead of trying to have a tool generic enough for all domains.
Luckily for me the chances are very good that I will be trying out this new approach on my new upcoming project. We will have to wait and see how many sleepless nights I will have to endure to make this a reality.
Lately I am gravitating more and more to the idea of having all my logic inside a meta model. This meta model will not only contain business logic but also programmatic logic . The meta model must be able to model the interaction between all the various business objects as well as the presentation objects which will in turn allow me to generate code from the meta model.
There are tools available today allowing this, but they are either too expensive or too bulky. Obviously creating a product to do all of this, the product needs to be very generic to allow it to be usable by the mass market.
Because my projects are medium in size, 5 - 10 developer teams, I do not need the bulky huge be everything to everyone packages. The tools needed in my domain will be fewer but more biased towards the domain instead of trying to have a tool generic enough for all domains.
Luckily for me the chances are very good that I will be trying out this new approach on my new upcoming project. We will have to wait and see how many sleepless nights I will have to endure to make this a reality.
Tuesday, February 14, 2006
Intel Pentium M vs. AMD Turion 64
Being an avid laptop fan, I found this article very interesting. I wondered when AMD will come to the party with some real laptop CPUs, the previous attempts were running too hot for my liking. It seems my next laptop will still be a Pentium M, but it is good to see that AMD is making some progress. The mentioned article is available here.
What I do find very interesting is the advantage the Pentium M has over the Turion when the level 2 cache is used. I suspect the level 2 cache plays a rather prominent role in compiling code, which is way my next laptop will again be a Pentium M.
What I do find very interesting is the advantage the Pentium M has over the Turion when the level 2 cache is used. I suspect the level 2 cache plays a rather prominent role in compiling code, which is way my next laptop will again be a Pentium M.
Subscribe to:
Posts (Atom)