Difference between revisions of "Out-of-order execution"
m (+cat) |
m (better cat) |
||
| Line 9: | Line 9: | ||
{{semi-stub}} | {{semi-stub}} | ||
| − | [[Category: | + | [[Category: CPU Basics]] |
Revision as of 23:56, 14 December 2018
Out-of-order execution (also called dynamic execution) is a technique for improving the performance of a processor by making use of what would otherwise be unused processor cycles. It does this by executing instructions in a different order than they appear in the object code.
A certain amount of 'book-keeping' is required, to make sure that all the operands of a given instruction (especially including those calculated in previous instructions) are available, and that a location is available to store the results - the register renaming mechanism is used to handle that issue.
Out-of-order execution is common in superscalar machines which want to avoid the 'stalling' technique for dealing with conflicts.
It is a valuable technique because it means that software (both programmers and compilers) does not have to worry about execution order, in order to obtain the maximum performance; this is all dealt with by the hardware.
