Hi all,
To better focus our efforts and to help us frame conversations with other groups, we need to define a target set of hardware to support. A big determinant in this is whether we have enough hardware available to represent this target, otherwise we can't really support it. Another big determinant is going to be how many packages won't run on our target. Hardware that is going to require changes to a significant number of packages is probably going to be out of our reach.
I have a couple of Pentium-M-based machines with 1GB of RAM available. Not that old, but I can at least check for AES problems. I'll have to see if I can scrounge up something older from the beige graveyard at work.
Do we have any pre-SSE hardware? Also, I'd like to see if we can support a 512MB minimum, if possible. I'd have to check that in a VM, though.
Thoughts?
jeff
One set of hardware to try and get a hold of is the OLPC hardware. This is widely used in Latin America and Southeast Asia and would be an ideal "we are doing it for the kids" story. The hardware while considered EOL is probably going to be in use until 2020 when I last checked on it a couple of years ago.
On 11 September 2017 at 13:36, Jeff Backus jeff.backus@gmail.com wrote:
Hi all,
To better focus our efforts and to help us frame conversations with other groups, we need to define a target set of hardware to support. A big determinant in this is whether we have enough hardware available to represent this target, otherwise we can't really support it. Another big determinant is going to be how many packages won't run on our target. Hardware that is going to require changes to a significant number of packages is probably going to be out of our reach.
I have a couple of Pentium-M-based machines with 1GB of RAM available. Not that old, but I can at least check for AES problems. I'll have to see if I can scrounge up something older from the beige graveyard at work.
Do we have any pre-SSE hardware? Also, I'd like to see if we can support a 512MB minimum, if possible. I'd have to check that in a VM, though.
Thoughts?
jeff
-- Jeff Backus jeff.backus@gmail.com http://github.com/jsbackus _______________________________________________ X86 mailing list -- x86@lists.fedoraproject.org To unsubscribe send an email to x86-leave@lists.fedoraproject.org
On Mon, Sep 11, 2017 at 01:46:33PM -0400, Stephen John Smoogen wrote:
One set of hardware to try and get a hold of is the OLPC hardware. This is widely used in Latin America and Southeast Asia and would be an ideal "we are doing it for the kids" story. The hardware while considered EOL is probably going to be in use until 2020 when I last checked on it a couple of years ago.
Not a bad idea. I'll see if I can track down someone with one.
jeff
The only machine I can test with is my Atom laptop, but it does have support for SSE2 and 2GB of RAM...
On Mon, Sep 11, 2017 at 05:47:34PM -0000, Ariela Wenner wrote:
The only machine I can test with is my Atom laptop, but it does have support for SSE2 and 2GB of RAM...
Still useful! Integrated Intel graphics?
jeff
I can use my dual Athlon MP (2GB) workstation and my Aspire One (Atom N270, 1.5GB). I just have to make time to get all the files and configuration settings backed up. Somewhere between November and January I'll be moving and I won't be able to test anything, but I will take the AOA with me, so after January I'll still have that.
I will also try to get acquainted with the noc at my new lab when I get there and see if they have any relics lying around.
I've decided to keep my PIII server out of all this and let it live its last couple of months in peace.
On Mon, Sep 11, 2017 at 09:06:25PM +0300, Alexander Ploumistos wrote:
I can use my dual Athlon MP (2GB) workstation and my Aspire One (Atom N270, 1.5GB). I just have to make time to get all the files and configuration settings backed up. Somewhere between November and January I'll be moving and I won't be able to test anything, but I will take the AOA with me, so after January I'll still have that.
Great!
I will also try to get acquainted with the noc at my new lab when I get there and see if they have any relics lying around.
Sounds like a plan!
I've decided to keep my PIII server out of all this and let it live its last couple of months in peace.
Hah! Yeah, I think that's fair. :)
jeff
I have - 2 Atom N270 based netbooks w/ 2GB RAM These have aes, sse, sse2.
- 1 Pentium III w/ 512 MB RAM This one has sse, but doesn't have pae nor sse2. So far, it had seen all Fedora releases having ever been released.
In general, - I'd like to see Fedora to continue supporting all HW it has supported until now, to avoid Fedora loosing users, who already are using Fedora. - Conversely, I do not see much sense in adding older CPUs, because these users are unlikely to have run Fedora. - Though I would feel a little grumpy about it and would consider this to be a strategic management short sightedness, I do have not much of a problem in Fedora dropping support for the Pentium III (i.e. non-aes/non-sse2 CPUs).
Ralf
On 11 September 2017 at 23:06, Ralf Corsepius rc040203@freenet.de wrote:
I have
2 Atom N270 based netbooks w/ 2GB RAM These have aes, sse, sse2.
1 Pentium III w/ 512 MB RAM This one has sse, but doesn't have pae nor sse2. So far, it had seen all Fedora releases having ever been released.
In general,
- I'd like to see Fedora to continue supporting all HW it has supported until now, to avoid Fedora loosing users, who already are using Fedora.
Looking at the current systems hitting the mirror system and using Fedora, the numbers for September have been:
#ofIPs %ofusers release 38159 38.4842% f25 14306 14.4279% f20 8503 8.57546% f26 5605 5.65277% f23 5003 5.04564% f08 4315 4.35177% f22 3884 3.9171% f24 3578 3.60849% f14 3537 3.56714% f21 1745 1.75987% f19 1703 1.71751% f17 1557 1.57027% f18 1272 1.28284% f07 1189 1.19913% f16 1098 1.10736% f10 1082 1.09122% f11 867 0.874389% f09 800 0.806818% f06 358 0.361051% f13 238 0.240028% f12 184 0.185568% rawhide 108 0.10892% f15 23 0.023196% f27 19 0.0191619% f05 15 0.0151278% f04 5 0.00504261% f03 2 0.00201704% f28
The majority of users are on EOL releases but about 15% of rawhide users are on i386... so I would see if there is a way to reach out and advertise to those people to see what they are wanting. The majority of rawhide users are in the US with the next set in India and China.
For users of f25 the majority of users are located in the following countries: 24649 GeoIP Country Edition: IN, India 1068 GeoIP Country Edition: US, United States 564 GeoIP Country Edition: RU, Russian Federation 482 GeoIP Country Edition: DE, Germany 369 GeoIP Country Edition: IT, Italy 345 GeoIP Country Edition: BR, Brazil 329 GeoIP Country Edition: PL, Poland 318 GeoIP Country Edition: CN, China 241 GeoIP Country Edition: FR, France 234 GeoIP Country Edition: GB, United Kingdom
I am not sure if that large number is a data anomaly or not. The networks in India tend to use nat over nat with multiple endpoints so 1 person can show up coming out of multiple ip addresses over a day. In the end though I would say working with the India ambassadors to see what users there are wanting might be a good focus.
- Conversely, I do not see much sense in adding older CPUs, because these users are unlikely to have run Fedora.
- Though I would feel a little grumpy about it and would consider this to be a strategic management short sightedness, I do have not much of a problem in Fedora dropping support for the Pentium III (i.e. non-aes/non-sse2 CPUs).
Ralf
X86 mailing list -- x86@lists.fedoraproject.org To unsubscribe send an email to x86-leave@lists.fedoraproject.org
On Tue, Sep 12, 2017 at 11:51:09AM -0400, Stephen John Smoogen wrote:
For users of f25 the majority of users are located in the following countries: 345 GeoIP Country Edition: BR, Brazil
Not sure if I will get any replies, but I sent an email in the pt-BR mailing list to query about their hardware.
On Sat, Sep 16, 2017 at 09:57:52AM -0300, Athos Ribeiro wrote:
On Tue, Sep 12, 2017 at 11:51:09AM -0400, Stephen John Smoogen wrote:
For users of f25 the majority of users are located in the following countries: 345 GeoIP Country Edition: BR, Brazil
Not sure if I will get any replies, but I sent an email in the pt-BR mailing list to query about their hardware.
Great! Thanks!
jeff
On Tue, Sep 12, 2017 at 11:51:09AM -0400, Stephen John Smoogen wrote:
Looking at the current systems hitting the mirror system and using Fedora, the numbers for September have been:
#ofIPs %ofusers release 38159 38.4842% f25 14306 14.4279% f20 8503 8.57546% f26 5605 5.65277% f23 5003 5.04564% f08 4315 4.35177% f22 3884 3.9171% f24 3578 3.60849% f14 3537 3.56714% f21 1745 1.75987% f19 1703 1.71751% f17 1557 1.57027% f18 1272 1.28284% f07 1189 1.19913% f16 1098 1.10736% f10 1082 1.09122% f11 867 0.874389% f09 800 0.806818% f06 358 0.361051% f13 238 0.240028% f12 184 0.185568% rawhide 108 0.10892% f15 23 0.023196% f27 19 0.0191619% f05 15 0.0151278% f04 5 0.00504261% f03 2 0.00201704% f28
The majority of users are on EOL releases but about 15% of rawhide users are on i386... so I would see if there is a way to reach out and advertise to those people to see what they are wanting. The majority of rawhide users are in the US with the next set in India and China.
For users of f25 the majority of users are located in the following countries: 24649 GeoIP Country Edition: IN, India 1068 GeoIP Country Edition: US, United States 564 GeoIP Country Edition: RU, Russian Federation 482 GeoIP Country Edition: DE, Germany 369 GeoIP Country Edition: IT, Italy 345 GeoIP Country Edition: BR, Brazil 329 GeoIP Country Edition: PL, Poland 318 GeoIP Country Edition: CN, China 241 GeoIP Country Edition: FR, France 234 GeoIP Country Edition: GB, United Kingdom
I am not sure if that large number is a data anomaly or not. The networks in India tend to use nat over nat with multiple endpoints so 1 person can show up coming out of multiple ip addresses over a day. In the end though I would say working with the India ambassadors to see what users there are wanting might be a good focus.
That's a great idea. We might also be able to track down people with hardware, such as the OLPC. I'll put together an e-mail sometime this week.
jeff
On Tue, Sep 12, 2017 at 05:06:43AM +0200, Ralf Corsepius wrote:
I have
- 2 Atom N270 based netbooks w/ 2GB RAM
These have aes, sse, sse2.
- 1 Pentium III w/ 512 MB RAM
This one has sse, but doesn't have pae nor sse2. So far, it had seen all Fedora releases having ever been released.
In general,
- I'd like to see Fedora to continue supporting all HW it has
supported until now, to avoid Fedora loosing users, who already are using Fedora.
Agreed. I see that as our current 'primary goal'.
- Conversely, I do not see much sense in adding older CPUs, because
these users are unlikely to have run Fedora.
- Though I would feel a little grumpy about it and would consider this
to be a strategic management short sightedness, I do have not much of a problem in Fedora dropping support for the Pentium III (i.e. non-aes/non-sse2 CPUs).
We'll see what hurdles we're facing. Since (in theory) SSE2 isn't a requirement for i686 packages, we should be able to (in theory) continue to support it. Particularly since you have hardware we can test on. ;-)
Speaking of which, are you running X on your PIII? If so, what WM are you using and how well does it run?
jeff
On 09/17/2017 06:05 PM, Jeff Backus wrote:
Speaking of which, are you running X on your PIII?
Yes, but I rarely use this machine at all and less so with a WM. ATM, its main use-case is testing SW and it hosting my last pair of floppies ;).
If I am using it, in most cases, I boot it into a login manager (currently lightdm) and ssh into it and am not using a WM.
If so, what WM are you using and how well does it run?
I have been using Xfce/xfwm4, which used to work sufficiently well for "very light" occasional use-cases. However memory demands apparently have gradually grown over the years, which render using Xfce into "constant swapping".
Furthermore, in recent times, the primary problem on this machine has been all Fedora-26 kernels >= 4.10 are exposing "hard lockups", apparently being related to networking[1].
Another set of problems, which is around for quite a while on this machine is apic problems (ATM, I am using pci=noapic).
Ralf
[1] https://bugzilla.redhat.com/show_bug.cgi?id=1482561 For the moment, I resorted to using kernel-4.9.14-200.fc25.i686 with FC26 ontop.
On 11/09/17 18:36, Jeff Backus wrote:
Hi all,
To better focus our efforts and to help us frame conversations with other groups, we need to define a target set of hardware to support. A big determinant in this is whether we have enough hardware available to represent this target, otherwise we can't really support it.
My computer is a Dell Inspiron 1300 laptop
https://uk.hardware.info/product/1727/dell-inspiron-1300/specifications
It has an Intel Celeron M380 (it says here). Though it currently has the maximum 2GB RAM fitted, it is otherwise as per those specs.
/proc/cpuinfo shows processor : 0 vendor_id : GenuineIntel cpu family : 6 model : 13 model name : Intel(R) Celeron(R) M processor 1.60GHz stepping : 8 microcode : 0x20 cpu MHz : 1596.108 cache size : 1024 KB physical id : 0 siblings : 1 core id : 0 cpu cores : 1 apicid : 0 initial apicid : 0 fdiv_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov clflush dts acpi mmx fxsr sse sse2 ss tm pbe nx bts cpuid bugs : bogomips : 3192.21 clflush size : 64 cache_alignment : 64 address sizes : 32 bits physical, 32 bits virtual power management:
Regards, Chris R.
On Tue, Sep 12, 2017 at 03:27:57PM +0100, Christopher Ross wrote:
My computer is a Dell Inspiron 1300 laptop
https://uk.hardware.info/product/1727/dell-inspiron-1300/specifications
It has an Intel Celeron M380 (it says here). Though it currently has the maximum 2GB RAM fitted, it is otherwise as per those specs.
/proc/cpuinfo shows processor : 0 vendor_id : GenuineIntel cpu family : 6 model : 13 model name : Intel(R) Celeron(R) M processor 1.60GHz stepping : 8 microcode : 0x20 cpu MHz : 1596.108 cache size : 1024 KB physical id : 0 siblings : 1 core id : 0 cpu cores : 1 apicid : 0 initial apicid : 0 fdiv_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov clflush dts acpi mmx fxsr sse sse2 ss tm pbe nx bts cpuid bugs : bogomips : 3192.21 clflush size : 64 cache_alignment : 64 address sizes : 32 bits physical, 32 bits virtual power management:
Looks good. Particularly for AES-related issues which seem to be cropping up.
jeff
-
Alexander Ploumistos -
Ariela Wenner -
Athos Ribeiro -
Christopher Ross -
Jeff Backus -
Ralf Corsepius -
Stephen John Smoogen