1) Am I right in saying that when you install a software package you a installing a binary file of the program?

2) If this is true how can this same binary file run on different CPU architectures i.e. Intel, AMD, Sparc etc?

3) The aim of compiling a kernel from source is to optimise it for your CPU architecture and other hardware. Is this true for programs too? If I compile xmoto from source for example will it run more efficiently, faster etc?

Thanks in advance,

Boler

1) Usually, you are installing a binary file. However, some packages contain programs that are written in non-compiled languages like Python or Java. OK, Java is sort-of compiled, but the upshot is that the Java program can't run without a local JVE, and THAT is compiled.

2) Usually the same binary can't run on different architectures. Packages like this are likely to be for a non-compiled program language like Java.

3)You can see some performance enhancement from compiling yourself, but frequently the difference isn't really noticeable. The best reason for compiling from source is to add some functionality that might not be in the binary distribution. Or simply because you want to.......

1. Yes

2. Not exactly. The binary is good for an arch. Mainstream cpus (intel AMD) come in two basic forms (after the pentium) 32bit and 64 bit. If you have a 32 bit mainstream cpu you run the iX86 (X can vary but usuall 3 as in i386) (32bit) binary. If you have a 64 bit mainstream cpu you can run a 32bit OS or a 64bit OS. You can run a 32bit binary on a 64bit OS but you cannot run a 64bit binary on a 32bit OS. Now Sparc, MAC, or whatever pretty much require a binary that is made just for them.

3. In theory yes, but in practice not really. The difference between a precompiled binary and compiling it yourself is usually in the low single digits difference in speed. When one considers the time and effort it can take to compile some (not all) stuff, one is better off just using the precompiled binary.