The INTEL architecture supports two implementations of breakpoints for debugging purposes:

The software generated breakpoint using the INT 3 instruction;

The hardware generated breakpoint using the Debugging Registers.

The use of software breakpoints require code modification, whereas the use of debugging registers does not. However, the number of predefined software breakpoints is potentially unlimited whereas there are only 4 breakpoints specifiable using Debugging Registers.

A further distinction between the two types is that software breakpoints only intercept the execution of a particular instruction path, whereas Debugging Registers may be used, in addition, to intercept data fetches and stores from a particular location in virtual memory.

The Kernel Debugger supports both implementations of breakpoints through the use of the:

BR command,which uses Debugging Registers, and

BP command, which uses INT 3 instructions.

The Kernel Debugger limits the predefinition of BP breakpoints to 10, however the programmer may code as may additional INT 3 instructions into his/her program as desired.

The Kernel Debugger refers to breakpoints explicitly set by the BP and BR commands as sticky (implying a certain permanence about them). The G command may have one or more temporary breakpoints established when one or more stop addresses are specified. These are referred to as go breakpoints. Once the Kernel Debugger breaks in go breakpoints are removed. The internal operation of the Kernel Debugger may also necessitate the use of the occasional temporary breakpoints when instruction tracing (see the T and P commands). These are set implicitly and discarded without the user being aware of their existence. Go and temporary breakpoints are created using the INT 3 instruction. Go and sticky BP breakpoints count towards the Kernel Debugger imposed limit of 10, but temporary breakpoints only ever exist singly so do not.