ruspopla.blogg.se

1. #WHY DOES MACROMEDIA FLASH MX 2004 NOT RUN FLA FILES SOFTWARE#
2. #WHY DOES MACROMEDIA FLASH MX 2004 NOT RUN FLA FILES CODE#

In a nutshell, if you use a byte, short, or int, you're slowing down the machine. Also, except when memory is statically allocated and you've forced the compiler to pack memory allocated to your variables, that carefully considered allocation of a single byte is going to take up 32 bits of actual memory anyway, because once again, the CPU can access it that way significantly faster. Speaking as a programmer whose first programming experience was Z80 assembly with 4K of RAM, I had to seriously revamp my thinking in recent years, but basically 32-bit processors work significantly better with 32-bit values. And it can't totally be explained by a bunch of bad algorithm choices.

#WHY DOES MACROMEDIA FLASH MX 2004 NOT RUN FLA FILES CODE#

Sure, the code does more today, but the factors still don't add up.

#WHY DOES MACROMEDIA FLASH MX 2004 NOT RUN FLA FILES SOFTWARE#

Hardware has gotten inconceivably faster, but software has hardly picked up the pace (in most domains). You can try to improve the locality of your data structures and your code, but those efforts are often at odds with a lot of modern practices such as object-orientation, enhancement by derivation, and letting the VM manager and the garbage collector worry about it. We have few tools to fight this type of performance problem. The processor waits for the VM manager to deliver the next instruction to RAM. A single button click to bring up a dialog can trigger tens of thousands of lines of code scattered all across virtual memory. Number crunching is still this way.īut today we have interactive, event-driven programs implemented with deep object hierarchies that mask how much code is actually being called. When programs were simpler, smaller and more procedural, performance problems appeared inside loops.

The structure of code, with long chains of hooks in message passing and method invocation, can result in skipping through dozens of pages. Code bloat in general increases working sets. (I don't intend that as a blanket condemnation of STL, it's probably just a funky implementation.)īut it's not just data structures. More than once I've had to replace an STL map with a hash table in order to get decent performance. Powerful, dynamic containers often have high-overhead per node and make little if any effort to improve locality of references. If nodes are scattered far and wide, then the page file can get a workout. In other words: virtual memory thrashing.ĭata structures are often suspect when thrashing is a problem. It is certainly true that a bad algorithm can result in slow code, but most of the difficult performance problems I've had to tackle in the last several years are more closely related with bloat. Lowness comes from incorrectly chosen data structures and algorithms.