Mesa Mega Driver Patches Published

[phoronix]

Phoronix: Mesa Mega Driver Patches Published

Eric Anholt of Intel has published his initial patches for implementing the "mega drivers" concept within Mesa...

http://www.phoronix.com/vr.php?view=MTQ3Mzc

[not.sure]

Doesn't seem to make too much sense from a technical point of view. Anybody has an idea about the politics side?

[mrugiero]

Doesn't seem to make too much sense from a technical point of view. Anybody has an idea about the politics side?

On politics, if forced (i.e., if it's not just a build time option) means you need to rebuild the whole mesa to use any driver that is not upstream.
On tech, I think it has some security/stability relevance, since they don't need to make all the symbols public.

[smitty3268]

Doesn't seem to make too much sense from a technical point of view. Anybody has an idea about the politics side?

Hiding all the symbols is a huge win on the technical side - although possibly it could be done in another way, i'm not sure about that.

[Daktyl198]

Doesn't seem to make too much sense from a technical point of view. Anybody has an idea about the politics side?

um...

and is noting about a 2.61% performance increase

It reduces some CPU load due to moving some calls to internal instead of across linked libraries. Also, once Gallium gets added to the mix, it will create a single binary that's about 1/6 the size of the current Mesa package (if you include ALL drivers, like some distros do)

[schmidtbag]

Michael, any chances on benchmarking this driver against the ones it merges?

[zoomblab]

I hope this stays optional.

Mega binaries are not a good practice and the performance benefits seem negligible. Modularization is harder but almost always the better approach.

[chrisr]

I really hope this is a configurable option

Eric Anholt of Intel has published his initial patches for implementing the "mega drivers" concept within Mesa...

This is exactly the kind of thing I don't need when trying to perform a "git bisect".

[uid313]

Is this even a good idea?

Big monolithic things are usually bad.

Isn't there any alternative ways do accomplish similar benefits without the drawbacks?

I don't think anyone else (Apple or Microsoft, etc) would use this solution.

[brosis]

I know the much easier and future proof approach.
Gallium.