Is using `int` more preferable than using `unsigned int`?

I have a code where I am using data-types such as int, unsignedint, and unsigned char. How beneficial/futile (in terms of memory/speed) will the activity of changing all the data-types to int be?

• The code has several structures defined having bytes,int, char inside them. Will it bring any performance improvements in terms of memory/speed considering structure padding and other factors?

• Answer:

  Please note that you can only save memory by using packed structures, but when you use packed structure you actually lose speed, because depending on the architecture, the CPU can read only 4-8 bytes aligned to multiple of 4/8 addresses. Again, this is dependent on the architecture. You save some memory, so if your work requires huge cache of data then you may want to use the benefits of this structure packing. You may save some speed when copying (via memcpy) large chunks of data (you copy less bytes). But otherwise, in real 'work' accessing the data will become slightly slower, because, for example, the CPU will bring and int into memory and isolate the char from the rest of the 'garbage'. Also with packed structures you might have race conditions when writing the same structure, but different fields. Otherwise, by default, the compiler decides on how to pack your data structures. Usually it makes sensible choices. A very interesting reading is Ulrich Drepper's paper: What every programmer should know about memory. http://www.akkadia.org/drepper/cpumemory.pdf