Question

Basically I am interested to know how far you go, to ensure your users get the best application.

1. Which one to prefer dynamic/static memory allocation, when both are possible.
2. Scraping int and using short or int8_t. Will it really help? Or it would be more of a headache?
3. Allocating a “proportionally” large amount of memory dynamically (due to lack of info on actual memory required). And then reallocating it to smaller memory (compact) and deleting the former. Eg. I allocate 1000 units of memory. Then after calculating, I determine that only 400 units are required. So I allocate new 400 units, copy the 400 items and then delete 1000 units. In this process i end up allocating even more memory i.e 1400 units at some time. So should I actually take pains to allocate new 400 units and let 600 units waste? [main question]
4. Any effective use of union to help lower memory usage?

I am sure there are more important points missing. Any other known or indigenous way to increase memory and time performance are welcome.

Answer 1

2. Static. Don't allocate memory when you don't need to.

3. If the system is a 32-bit system, then the actual instructions to the processor will actually prefer uint32_t or int32_t...

5. If I need a generic data type that would need to be allocated several times in a loop, then a union can come in handy... to avoid allocating memory.

6. From my experience, allocating memory can be a slow process if it is done in a loop... You will see a performance increase if you allocate all the memory ahead of time, and then use it from a pool. When the system has go find huge blocks of memory, it can be a bottleneck.

7. In real-time situations with large chunks of memory... yes... otherwise... eh... not really

Hope this helps… If you follow this style, you will notice improved performance.