More AMD PowerPlay Code Is Being Refactored

[phoronix]

Phoronix: More AMD PowerPlay Code Is Being Refactored

Rex Zhu of AMD published his latest set of 16 patches for refactoring PowerPlay code within the open-source DRM driver space. This is a big refactoring: nearly 19,000 new lines of code and 34,000 deleted lines of code...

http://www.phoronix.com/scan.php?page=news_item&px=PowerPlay-Code-Refactoring

[Azpegath]

Oooh my god, deleting 34000 lines of code is always such a nice rush! Just the bare thought of de-duplicating code gives me the nice kind of shudders...

[libv]

I haven't bothered to read up properly, but i guess that this is with DAL and/or ATI style code. There's going to be tons of similar bullshit about, and i still wonder how one could burn 19kloc still on this topic for just 1 or two chipset generations. 19.0/34.0 = 56% which suddenly does not sound that impressive anymore when you realize that this is 19k left. There's some pretty nasty things happening there still if it needs _that_ much code, and ATI style solutions are the reason.

[pal666]

I haven't bothered to read up properly

no need to say more

[libv]

I have now read up properly and my assumptions (Which i clearly marked as such) are now confirmed. ATI thinking. This is hopefully not the last refactor. Also, how can this AMD driver be 32MB in source code? Wasn't AtomBIOS supposed to hide everything?

[schmidtbag]

Oooh my god, deleting 34000 lines of code is always such a nice rush! Just the bare thought of de-duplicating code gives me the nice kind of shudders...

Hahaha I know what you mean. Nothing feels better than to take a heaping pile of filth and watch it all just go away in front of your eyes with no effort. Like vacuuming up a big pile of dust or using a pressure washer to clean up a grimy patio:

https://media0.giphy.com/media/LD7TOKhaP8Kqc/200.gif

[schmidtbag]

Hahaha I know what you mean. Nothing feels better than to take a heaping pile of filth and watch it all just go away in front of your eyes with no effort. Like vacuuming up a big pile

Well, I shouldn't say no effort. I don't know if it was as simple and just highlighting a bunch of rows and pressing delete. (BTW, what happened to the edit button?)

[planetguy]

Spammers were posting something clean, then editing it to be their spam. Michael made it available to premium users only to stop that.

[srkelley5]

Why did AMD have so much duplicated code?

[bridgman]

Until the products have reached production you don't know for sure how different the code for each chip will end up being, and making abstraction decisions too early brings a different set of problems including risk of destabilizing already-released parts.

In "sensitive" areas like power management there's an argument for waiting until all the parts for a generation are in production before optimizing for code sharing. In other areas (eg userspace drivers) you see more aggressive code sharing at the start.