Mesa Considers Raising CPU Support Baseline

[skeevy420]

Oddly enough I remember when Linux was for enthusiasts with weird, wonderful and "ancient" hardware.

Since when did it change to only become about consumers and gamers who should probably be running Windows anyway?

Sure, they can regress the "baseline" all they want. Other projects that do happen to want to keep the old PPC MacBook or SGI machines will simply pick up the slack and fix the brokenness (probably just a build time flag anyway).

For me and x86 it's about drawing a line between obviously old hardware that is holding back the full potential of modern hardware. We're clearly at a point in the architecture where so many advancements have happened over so many years that we need a new baseline for what is modern x86. Michael's benchmarks show us over and over again that we have better performing systems when we're running better than generically optimized code...like with those AOCC benchmarks the other day.

I'm sure the 7 people running PII and PIII systems are totally upset about this.

But, seriously though, the small handful of people this will effect would be better off pooling their resources together and coming up with Legacy Linux & LegacyFreeBSD. I think that's the best option because all the rest involve lots wasted space storing code a CPU will never run; targeted binaries and whatnot. Seems like bloating up the binaries to make them both optimized and compatible everywhere defeats the purpose of trimming the fat for an optimized modern system or legacy system. Especially the legacy systems. Legacy people should be wanting to trim the fat more than us modern folks since PII's and PIII's are starving for cycles a hell of a lot more than a Zen 1 or Zen 2.

It isn't about running anyone off or making Linux less accessible; it's the opposite, it's about letting the weird, wonderful, and modern hardware shine just was well as the weird, wonderful, and ancient hardware already does.

[edge-case]

Your reaction to people using ancient hardware is, "F you, Nvidia!"? That doesn't make any sense...

No - stop taking the photo so literally / pedantically - he's obviously giving the middle finger to users of "ancient hardware".. and using a Linux / Linus related photo to make his point. Quite clever, really.

Was it really not obvious to you, or are you just trolling?

[skeevy420]

Hardware that people can't use Chrome or Firefox on anyway for 7-5 years and has been required since Windows 8 (and a Windows 7 update).
It's basically garbage. I'd have more use as sand rather than silicon.

No reason they shouldn't just use an older version, they aren't going to get any benefit from upgrading their Mesa.
Not like someone is going to be using a Pentium 4 with a RX 6800XT. :V

Hold my beer


Well.....that is if you know someone I can rob to get an RX 6800XT. I'm not getting one legally, that's for sure.

[skeevy420]

My tumble dryer is over 20 years old and has only ever needed a replacement drum drive belt which I did myself, we won't be giving you a membership card to the miser, thrift & recycling club.

In fairness, a 20 year old processor is the basis for a Spacebar Heater. Someone bragging about the recycling club should also be for pushing more energy efficient hardware.

I had to replace my dryer last year. After the third belt repair the blower and heater started going bad. Luckily right when that happened my aunt got her stimulus check (first one) and replaced her washer and dryer so I got a free dryer. Talk about timing.

[Nille_kungen]

I can't see this as anything negative, for those that needs other defaults they will have to pass a argument when compiling.

[ermo]

Seems to me that requiring the same base level as x86_64 (which includes SSE2) should be fine at this point (I run a couple of old PhII boxes which are at this level).

I don't really get why there isn't a subscription-based (binary) package manager where you can subscribe to your base architecture level and then just get the packages tagged as meeting your subscription requirements. I guess the old, incumbent distros out there haven't yet arrived at a point where having this capability is par for the course.

This would also help distributors set the base level -- over time, systems will likely start requesting newer architecture levels more often (the statistics of which can be published in simple graphs) and since it could be purely based on server-side download numbers (which are logged anyway), you aren't really compromising people's privacy in any way, while still getting a good idea of what's happening on the ground.

[skeevy420]

ermo
To wager an educated guess: base x86_64 levels are just now a thing. Before now it would have had to have been purely architecture based, a bunch of -march=xxx repos, or based on more exotic compiler flags like Clear Linux and essentially making up their own base level standard as they go forward. Now that we have base levels that opens up options for distributions and package managers so now they just need the time to adapt.

[kpedersen]

It isn't about running anyone off or making Linux less accessible; it's the opposite, it's about letting the weird, wonderful, and modern hardware shine just was well as the weird, wonderful, and ancient hardware already does.

Yeah, in all fairness I do get that. In particular for my older machines I simply do not run the latest version of Linux. Part of the fun of old shite is also running old shite software.

I also understand that the Linux desktop is tiny and insignificant (this isn't intended for offense. I am a big fan of the BSD's and those are remarkably smaller still!) so there is an issue of manpower. I guess I just don't want to see ~10 year old machines artificially dropped because companies like Intel are slowly wriggling their way into governing the direction of Linux. Blind consumerism of macOS is the reason that people escape to Linux for after all!

I think the only modern distro that actually cares about older hardware these days is Raspberry Pi Desktop (https://www.raspberrypi.org/software...ry-pi-desktop/) and I am pretty sure that Debian, being quite conservative will make sure that the more compatible build flags are used for the i386 arch.

[ermo]

ermo
To wager an educated guess: base x86_64 levels are just now a thing. Before now it would have had to have been purely architecture based, a bunch of -march=xxx repos, or based on more exotic compiler flags like Clear Linux and essentially making up their own base level standard as they go forward. Now that we have base levels that opens up options for distributions and package managers so now they just need the time to adapt.

My impression is that -march=haswell has been seen as the "new" x86_64 target for a while now (AVX2, FMA etc.), while basic x86_64 (16 int + 16 float/simd registers w/SSE2) has been the natural baseline for, quelle surprise, x86_64 since it landed on the market 10+ years ago?

The only "new" thing here is the levels in between I think?

Happy to learn something new as always, so don't be afraid to correct my perspective if you feel I'm barking up the wrong tree. =)

[angrypie]

What the fuck? AMD made SSE2 architectural in AMD64, meaning you can't have a 64-bit x86 implementation without it. Only 32-bit only CPUs would be affected.