Intel Linux Now Supports AMD_performance_monitor

[schmidtbag]

Point of using parallels, is that both compared things share similarities.

Here only FLOSS is shared...


3D graphic industry is FULL of people who like to DO IT YOUR OWN WAY. Both MS and ARB are means to decide what should be picked up by everybody after single-vendor implementation proved its usefullness.

For that matter good performance measuring extension will be highly implementation dependednt. Different GPU architecture are after all DIFFERENT.


And last but not least. Remind me:

How competitor team embrancing same solution help fragment the world?

Look at Windows/proprietary side of equation. Did he provided us with shiny example of measurement extensions beeing implemented by big 3?

The impact of developers doing their own thing in a FOSS environment is all that matters to make a comparison. Competing teams embracing the same solution is NOT fragmentation, and I personally never said otherwise. I was actually saying Mesa was a bad example of fragmentation since all the main devs for it are overall in agreement with how things should be done, they're just doing things at their own pace.

The thing to keep in mind is fragmentation is only a problem when the new product is (or ever will be) no better than the product it was trying to replace/improve. This is exactly why Mir was so frowned upon - rather than Canonical compromising and finding ways to improve upon wayland, they decided to ditch it and do their own thing. This is detrimental to the community, because everyone (driver devs, toolkit devs, DE devs, distro devs, etc) need to accommodate the change. Anyway, distros too make a good example of devs going their separate ways.

But more to the point of the 3D industry doing things their own way, there are 2 sides to the coin. There's the Mesa approach, where all the devs are overall striving for the same goal (even with competing hardware), which results in a clean, consistent, and highly compatible way to use video hardware. Then there's the proprietary approach, where the hardware (should) operate with micro-optimized performance and all advertised features. The problems with the FOSS approach to video drivers is performance likely won't be micro-optimized, you get limiting tools, and you can't stray too far with special features exclusive to a platform without crippling everyone else. The problem with proprietary is you're dependent on the developer supporting the software you want to use, which can sometimes leave you in the dark. Proprietary can also limit what you're able to run on your system (for example, I'm pretty sure you can't run nvidia blob and catalyst at the same time, but, you can do nouveau and radeon at the same time).

So for video drivers, fragmentation is a good thing because you can either take the route of the community (consistent operation) or you can take the route of the corporation (guaranteed performance). People may hate nvidia to not cooperating with the community but nvidia isn't wrong in their decisions, because their linux drivers are overall better than their windows drivers. That clearly shows nvidia knows what they're doing and therefore has every right to not listen to the community's demands. Why change your strategy when it's currently proving to be the best?

As for Windows, I get the impression the OS's graphics stack and kernel don't change that often, so nvidia, intel, AMD, VIA, etc all have to accommodate the OS. With Linux being open source and relatively unpopular, it's easier for these companies to modify the OS for their drivers than to accommodate the drivers for the OS.

[przemoli]

The impact of developers doing their own thing in a FOSS environment is all that matters to make a comparison. Competing teams embracing the same solution is NOT fragmentation, and I personally never said otherwise. I was actually saying Mesa was a bad example of fragmentation since all the main devs for it are overall in agreement with how things should be done, they're just doing things at their own pace.

The thing to keep in mind is fragmentation is only a problem when the new product is (or ever will be) no better than the product it was trying to replace/improve. This is exactly why Mir was so frowned upon - rather than Canonical compromising and finding ways to improve upon wayland, they decided to ditch it and do their own thing. This is detrimental to the community, because everyone (driver devs, toolkit devs, DE devs, distro devs, etc) need to accommodate the change. Anyway, distros too make a good example of devs going their separate ways.

But more to the point of the 3D industry doing things their own way, there are 2 sides to the coin. There's the Mesa approach, where all the devs are overall striving for the same goal (even with competing hardware), which results in a clean, consistent, and highly compatible way to use video hardware. Then there's the proprietary approach, where the hardware (should) operate with micro-optimized performance and all advertised features. The problems with the FOSS approach to video drivers is performance likely won't be micro-optimized, you get limiting tools, and you can't stray too far with special features exclusive to a platform without crippling everyone else. The problem with proprietary is you're dependent on the developer supporting the software you want to use, which can sometimes leave you in the dark. Proprietary can also limit what you're able to run on your system (for example, I'm pretty sure you can't run nvidia blob and catalyst at the same time, but, you can do nouveau and radeon at the same time).

So for video drivers, fragmentation is a good thing because you can either take the route of the community (consistent operation) or you can take the route of the corporation (guaranteed performance). People may hate nvidia to not cooperating with the community but nvidia isn't wrong in their decisions, because their linux drivers are overall better than their windows drivers. That clearly shows nvidia knows what they're doing and therefore has every right to not listen to the community's demands. Why change your strategy when it's currently proving to be the best?

As for Windows, I get the impression the OS's graphics stack and kernel don't change that often, so nvidia, intel, AMD, VIA, etc all have to accommodate the OS. With Linux being open source and relatively unpopular, it's easier for these companies to modify the OS for their drivers than to accommodate the drivers for the OS.

Well. If I cut out pure BS from that statement, what will I get?

Zero.

Some really strange assumptions there:
1) That Intel/AMD floss teams aren't doing it the corporate way... BIG corporations pay them for what?
2) Mesa strive for consistency over performace... If you hear anything about perf, its that its improving. Always.
3) Windows do not change graphic stack often... Maybe You where fulled by long interwals between Win releases. But anyway, each and every means new graphic stack.
4) Mir reqire everybody to adopt... Like 1% means everybody... (Driver devs + toolkit devs + distro devs + DE devs => 1% or they do it the wrong way. And canonical is ready to do the heavy lifting. That is not the best example for fragmentation out there...)
5) All of that is on-topic. But irrelevant to trolling by mark45.

[schmidtbag]

Well. If I cut out pure BS from that statement, what will I get?

Zero.

Some really strange assumptions there:
1) That Intel/AMD floss teams aren't doing it the corporate way... BIG corporations pay them for what?
2) Mesa strive for consistency over performace... If you hear anything about perf, its that its improving. Always.
3) Windows do not change graphic stack often... Maybe You where fulled by long interwals between Win releases. But anyway, each and every means new graphic stack.
4) Mir reqire everybody to adopt... Like 1% means everybody... (Driver devs + toolkit devs + distro devs + DE devs => 1% or they do it the wrong way. And canonical is ready to do the heavy lifting. That is not the best example for fragmentation out there...)
5) All of that is on-topic. But irrelevant to trolling by mark45.

You clearly are reading everything the way you want to but not the way I actually worded it. In fact, every single one of your points is something I said, just in a lot fewer words. The only thing I'm disagreeing with is mark45 isn't trolling - you are actually being more of a troll right now, saying my post is BS when you've basically summarized what I said.