Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.


Messages - ClassicHasClass

Pages: [1] 2 3 ... 16
1
Operating Systems and Porting / Re: [NEWS] Fedora 35 is out folks
« on: November 25, 2021, 10:50:06 pm »
The performance problem is indeed libgraphene again: the fix seems present in Fedora's sources but it doesn't stick. I rebuilt it and the performance regression is largely gone. Here's a binary for those who want it: https://github.com/ebassi/graphene/pull/233#issuecomment-979687134

Still other F34 regressions are still present. It basically works, though. https://www.talospace.com/2021/11/fedora-35-mini-review-on-blackbird-and.html

2
It's pretty much time dependence. My day job is ... different.

3
Most unfortunate.  :(

4
General Discussion / Re: Learning POWER9 assembly
« on: November 13, 2021, 10:18:59 am »
Thanks! Most of what's in Firefox now is just taking better advantage of VMX and VSX. The JIT should blow those doors open, though just remember that the MVP is the first-stage JIT (there are three stages) and Wasm support, so it should not be interpreted as "as good as it gets." And don't forget to do an LTO-PGO build if your distro doesn't; I keep those patches up to date: https://www.talospace.com/2021/11/firefox-94-on-power.html

5
Legacy POWER Hardware / Re: Old stuff survey
« on: November 13, 2021, 10:17:01 am »
Haven't picked up anything new, but I have been testing things as I move them into storage, and my beloved "SR-7100" (Power Mac 7100/66AV with a 400MHz Sonnet G3, max RAM and HPV card) still works just great. That was a nice find.

7
General Discussion / Re: Learning POWER9 assembly
« on: November 11, 2021, 11:51:31 pm »
No, you don't need to be extremely careful, and POWER9 has made this easier by getting rid of dispatch groups (this used to be sensitive to how instructions were packaged up into dispatch groups and made their way through the execution pipeline). Register renaming is also a lot more powerful than it used to be, though if you can make use of more registers manually, that always helps. The biggest sources of slowdowns are inappropriate use of the link register for branching or trampolines, which will foul the return address cache, and avoidable or aliased spills to memory, which is why the FPR<->GPR moves in VSX are so great (to move a GPR to the FPR used to require a memory spill and load, which inevitably aliased if you did this on the stack, so you also needed nops). As a general rule, Power chips are also way better at straightline code than branching, even if the straightline code seems to do more work.

8
Operating Systems and Porting / Re: [NEWS] Fedora 35 is out folks
« on: November 11, 2021, 11:47:08 pm »
On F34, I'm using the juiced graphene, and yes, it was a major improvement. It looks like F35 still didn't pick it up.

9
GPU Compute / Accelerators / Re: Blackbird Radeon RX580
« on: November 10, 2021, 11:13:58 am »
@MPC7500, we have a winner!

Code: [Select]
sudo grubby --update-kernel=ALL --args="modprobe.blacklist=ast"
...achieved the desired result.

Thanks!

This could still make things tricky in the unusual but by no means remote possibility something still goes wrong with the graphics card. It still sounds like a bug that this should be necessary (it wasn't before). I'll investigate this when I do my upgrade.

10
GPU Compute / Accelerators / Re: Blackbird Radeon RX580
« on: November 09, 2021, 01:13:01 pm »
I mainly use Ubuntu 20.04.3 with Xorg and the problem is the same, so I don't think it affects the graphics server.  They changed some damn parameters on the new Kernels and now we have problems and what's worse they still haven't solved it ...!  Classic sorry for Kernel options you mean when configuring the parameters?  Is there this option that you say must be set to 0?  Let me get it right ...

Yes, you want to change the kernel command line that Petitboot uses (which, unless you're doing something really strange, should be the same as Grub's). You don't need to rebuild the kernel for that option.

11
General Discussion / Re: Learning POWER9 assembly
« on: November 09, 2021, 01:09:54 pm »
No, actually, the mscdfr0-type instructions are the ones I hate. eieio is just fun for jokes though it has practical use as a lightweight barrier. But sheer number of instructions aside, RISC really is now just parlance for load-store. In that sense FISC might be more appropriate: https://news.ycombinator.com/item?id=28601455

ARM is just as bad (in fact worse due to their crazy encodings), and I think RISC-V will eventually metastasize. Even MIPS is getting that way. That said, I'll also add as someone handcoding assembly right now for the Firefox JIT that it was so nice to finally have instructions for GPR<->FPR moves in VSX plus a lot more rounding-type instructions rather than having to serialize the FPSCR by twiddling bits. And VSX is way more complete than VMX used to be (I had to tie myself in knots to write good AltiVec routines). ISA 3.0 added a whopping number of instructions to what was already a large instruction set but I find I'm actually using them.

12
GPU Compute / Accelerators / Re: Blackbird Radeon RX580
« on: November 08, 2021, 06:35:56 pm »
Yes Classic I know that I have the Navi, only that I say this because I have noticed that every time there is a problem in the Kernel, concerning the AMD GPUs, this occurs indiscriminately on different types of cards, such as Navi, Nano and others ... I therefore believe that the problem is the same also because coincidentally we have the same problems at the same time ... Maybe you have not read this link sent by the friend MPC, here we are talking about Vega, so another model but problem itself. Here they say they have solved but I did not understand anything about how they did unfortunately. If you can better understand what they did then it's great ... I'll post the link ...

https://gitlab.freedesktop.org/drm/amd/-/issues/1723

It looks like it's the same solution, though I still think the underlying issue is different. You could change this in Petitboot or Grub, but either way you want to add add `amdgpu.aspm=0` to your kernel options.

13
GPU Compute / Accelerators / Re: Blackbird Radeon RX580
« on: November 08, 2021, 06:32:29 pm »
If you disable gdm (easiest way would be something like `systemctl set-default multi-user.target` or add `systemd.unit=multi-user.target`), do you at least get a text boot? You could try messing with additional options from there. You should still be able to install xorg in F35 even if it didn't come with it.

14
GPU Compute / Accelerators / Re: Blackbird Radeon RX580
« on: November 08, 2021, 03:22:32 pm »
@mcarden, that's distressing because that card should be "known working." I wonder if @tle has already updated. My F34 T2 has a WX7100 and it works fine, but I haven't updated it to F35. Do you get at least Petitboot on screen? Can you force fbdev for Xorg?

MauryG5, I think you have a different issue because you're trying to use a Navi-based card, as memory serves. These are Polaris and should already be working just fine.

15
GPU Compute / Accelerators / Re: Blackbird Radeon RX580
« on: November 08, 2021, 12:38:16 pm »
What are your kernel command line options? According to https://bugzilla.kernel.org/show_bug.cgi?id=200695 some people have had trouble with amdgpu.dc=1 (setting it to =0 seemed to fix it), though that problem seems to be somewhat old. There are a lot of comparable WX7100 OpenPOWER systems out there which should be the same generation, so I would expect this to be a more widespread problem (but I haven't upgraded my workstation to 35 yet either, planning to do the BMC-only Blackbird this week).

Pages: [1] 2 3 ... 16