Low-Level Code Optimization in Phoenix
(Page 1 of 4 )
Compiler writing has been reduced to the job of a few people in the past. The developer should stop worrying about the phase at which the code snippet he entered in becomes code that the machine can execute. The appearance of the Phoenix Compiler Framework may make things easier for those who wish to take advantage of it.
I already presented the framework in a nutshell, so this week I will not cover this section again. If you missed my article titled "The Phoenix Compiler Framework," you are free to read it, just search the ASP Free site, or even better, under my profile. Everything I presented there will come in handy for a better understanding of this article, although you do not have to read it first.
Writing code has always been built around the developer. There always has been a direction to make the life of the developer easier. For this purpose, languages appeared that we call "managed," and for this reason, tools exist that help us develop, such as Microsoft Visual Studio or Netbeans.
Of course you can write a great program with just a text editor; however, at some point, that can become so complicated that the developer will "break his legs" in the code, and he will have enough of everything related to coding for a lifetime.
No one wants this to happen. Therefore, specific tools were created to compensate for algorithmic problems with ease of implementation. However, at the end of the day, the code written by a developer will have to be transformed into something that the machine can understand and run.
Nice, readable code will have little importance for the machine. It executes anything without checking its appearance. Moreover, the saddest thing is that the most optimal and fastest codes still are the ones that are hard to read and write. To make the best trade with the developer, the ideal case scenario is to transform your beautiful code into fast code.
By the end of the previous article about Phoenix, you realized the importance of the linear flow of the IR (Intermediate Representation). This is what a back=end programmer of Phoenix will have, and this is what he has to generate.
The front-end is responsible for checking that what you wrote is valid according to the language's semantics and enforce those rules on you. This includes jobs like making sure you do make every cast correct, or even that you made it at all, or that you have not missed a ";" somewhere.
So for this part, we are going to the back end of the compiler to see how your code snippet is going to become executable machine code. One of the people responsible for this job for the Phoenix Compiler is Russell Hadley. The code that he gets from the front-end looks just like your programming teacher at the university told you not to write it.
It is full of labels and "goto"s. It may include a sequence of instructions that says add this with this, flip this bit, check if this variable is zero and if so go to there. Russell is responsible for filtering the code in a manner that, in the end, will make it both efficient and understandable to the platform.
Next: Resource allocation >>
More BrainDump Articles
More By Gabor Bernat
