Clear Linux Exploring "libSuperX11" As Newest Optimization Effort

[phoronix]

Phoronix: Clear Linux Exploring "libSuperX11" As Newest Optimization Effort

As another optimization for Intel's Clear Linux distribution, a "libSuperX11" library is being considered that fundamentally changes how the X.Org libraries are handled...

http://www.phoronix.com/scan.php?pag...ux-libSuperX11

[King InuYasha]

Why is this being discussed in the Clear Linux mailing list rather than the Xorg development one? This seems like something that should be considered upstream...

[arjan_intel]

Why is this being discussed in the Clear Linux mailing list rather than the Xorg development one? This seems like something that should be considered upstream...

because it's a packaging change not a source code change?

[King InuYasha]

because it's a packaging change not a source code change?

You're literally creating a brand new library in which all applications will link to instead of the component libraries. It sounds an awful lot like a rather significant code change across the board.

[arjan_intel]

You're literally creating a brand new library in which all applications will link to instead of the component libraries. It sounds an awful lot like a rather significant code change across the board.

eh no? applications link to GTK/QT ;-)

and even those who want to link to libXau (random example, pick any)... they still do so.
The packaging is just such that libXau.so has become a 1 line linker script that pulls in the main library, instead of a symlink to another file.

[vegabook]

Intel has already shown itself to be possibly the single most important contributor to Linux, even ahead of Red Hat. I get that Intel gets a lot of uphill from the Ryzen-pushers (including me) but nobody can fault this company for its absolutely massive contribution to Linux. Long may it last and, Thank You Intel.

[mlau]

eh no? applications link to GTK/QT ;-)

and even those who want to link to libXau (random example, pick any)... they still do so.
The packaging is just such that libXau.so has become a 1 line linker script that pulls in the main library, instead of a symlink to another file.

Still, the idea is great IMO, and the worst the X devs can tell you is "no", no?

[AdamOne]

good, let intel spend more cash on developing Linux, and let the community reap the rewards.
Win-win, I say!

[milkylainen]

It's an interesting thing to try either way. A porker like this might get some relevant benefits.
I always tried to adhere to "stupid and slow code is still stupid and slow, no matter how much magic sauce you throw at it."

I am unsure about such fixes vs speeding up more common operations in general.
How much does a typical running X consume in running code vs moving data by common functions for example?
I always come back to optimizing regular glibc functions for a lot of stupid platforms and cpu implementations (powerpc).
But for modern x86 CPUs I guess there is nothing much to benefit from doing that?
x86 CPUs already have a brilliant memory backend and prefetchers...

[salsadoom]

Why is this being discussed in the Clear Linux mailing list rather than the Xorg development one? This seems like something that should be considered upstream...

My guess is just that its not "there" yet. Like its an idea they are throwing around and maybe will begin experimenting with. I think, quite reasonably, they'd like to have something to show the xorg guys and maybe even some benchmarks to demonstrate that its actually a practical benefit. As it is, it sounds like just throwing a lot of armchair ideas around that would be a bunch of work for possibly no benefit. You show up with code however, it becomes a lot more interesting.