Skip to contentSkip to conversation

General CHRP/MPC5K2 Platform and drivers support - to comment

PATCH In review Nicolas DET 13 participants 34 messages Oct 19, 2006 – Oct 25, 2006

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 General CHRP/MPC5K2 Platform and drivers support - to comment

This 'big' patch adds support for CHRP/MPC52xx based platform. Here, this is the bPlan's Efika computer (http://www.bplan-gmbh.de/efika_spec_en.html)

We know this patch is not totaly compliant with Documentation/CodingStyle/.

We would like people to comment and review it. This way we would provide a new patch with the changes required (if any) for an upcomming merge in the kernel.

This patch has been applied on the kernel 2.6.18.1.

begin:vcard
fn:Nicolas DET ( bplan GmbH )
n:DET;Nicolas
org:bplan GmbH
adr:;;;;;;Germany
email;internet:[email protected]
title:Software Entwicklung
tel;work:+49 6171 9187 - 31
x-mozilla-html:FALSE
url:http://www.bplan-gmbh.de
version:2.1
end:vcard

#1 lore raw

So, like, the other day Nicolas DET mumbled:

Quoted 4 lines

This 'big' patch adds support for CHRP/MPC52xx based platform. Here, this is the bPlan's Efika computer
(http://www.bplan-gmbh.de/efika_spec_en.html)

In mail like this with buried patches, could we at least get the full "git diff --stat HEAD" or "git apply --stat /path/to/patch" included in the base mail too please?

Thanks,
jdl

#2 lore raw
In reply to Nicolas DET
Quoted 7 lines · Nicolas DET

This 'big' patch adds support for CHRP/MPC52xx based platform. Here, this is the bPlan's Efika computer
(http://www.bplan-gmbh.de/efika_spec_en.html)

We know this patch is not totaly compliant with Documentation/CodingStyle/.

We would like people to comment and review it.

Can you send the patch unzipped, and not as an attachement? This would make reviewing and commenting much easier.

--linas

#3 lore raw
Quoted 11 lines · Linas Vepstas
Quoted 7 lines · Nicolas DET

This 'big' patch adds support for CHRP/MPC52xx based platform. Here, this is the bPlan's Efika computer
(http://www.bplan-gmbh.de/efika_spec_en.html)

We know this patch is not totaly compliant with Documentation/CodingStyle/.

We would like people to comment and review it.

Can you send the patch unzipped, and not as an attachement? This would make reviewing and commenting much easier.

Yes, and please break it up into more digestible chunks.

cheers,
g.

Signature · 4 lines
Grant Likely, B.Sc. P.Eng.
Secret Lab Technologies Ltd.
[email protected]
(403) 399-0195
#4 lore raw
In reply to Linas Vepstas
Quoted 11 lines · Linas Vepstas
Quoted 7 lines · Nicolas DET

This 'big' patch adds support for CHRP/MPC52xx based platform. Here, this is the bPlan's Efika computer
(http://www.bplan-gmbh.de/efika_spec_en.html)

We know this patch is not totaly compliant with Documentation/CodingStyle/.

We would like people to comment and review it.

Can you send the patch unzipped, and not as an attachement? This would make reviewing and commenting much easier.

It's too big for the list.

Ben.

#5 lore raw
Quoted 6 lines · Benjamin Herrenschmidt
Quoted 5 lines · Linas Vepstas
Quoted 4 lines · Nicolas DET

This 'big' patch adds support for CHRP/MPC52xx based platform. Here, this is the bPlan's Efika computer
(http://www.bplan-gmbh.de/efika_spec_en.html)

Quoted 8 lines
Quoted 6 lines
Quoted 3 lines

We know this patch is not totaly compliant with Documentation/CodingStyle/.

We would like people to comment and review it.

Can you send the patch unzipped, and not as an attachement? This would make reviewing and commenting much easier.

It's too big for the list.

I think the reasons its to big is it includes patches for platform, uart, and enet drivers on 52xx.

- k

#6 lore raw
In reply to Nicolas DET
Quoted 5 lines · On Thu, 19 Oct 2006 14:39:01 +0200 Nicolas DET <[email protected]> wrote:

We know this patch is not totaly compliant with Documentation/CodingStyle/.

We would like people to comment and review it. This way we would provide a new patch with the changes required (if any) for an upcomming merge in the kernel.

Hi,

Like others have said, a monolithic patch like this is really hard to review. Please split it up, there are several logical ways it could be cut, drivers vs base platform support, etc.

First reaction from just browsing through it:

* Lots of SillyCaps, i.e. codingstyle issues, especially in the bestcomm code
* No need to have filenames at top of files (they're even wrong in some cases)
* You reinvented pr_debug() (everyone loves doing this, myself included)
* Some hardcoded baudrate stuff in the serial driver. Should probably be passed in from somewhere -- device tree?
* Network drivers should be posted to netdev, not the arch list

More thorough review would be easier to do with split-up patches.

-Olof

#7 lore raw
In reply to Nicolas DET
Quoted 10 lines · Nicolas DET

This 'big' patch adds support for CHRP/MPC52xx based platform. Here, this is the bPlan's Efika computer (http://www.bplan-gmbh.de/efika_spec_en.html)

We know this patch is not totaly compliant with Documentation/CodingStyle/.

We would like people to comment and review it. This way we would provide a new patch with the changes required (if any) for an upcomming merge in the kernel.

This patch has been applied on the kernel 2.6.18.1.

I think it would be extremely useful to get a dump of what the device-tree looks like. We have to be very careful to device nodes for 52xx components are competed consistent between this 'CHRP' style and a more embedded 52xx boards.

Some high level comments:
1. lets stick with the 52xx naming, instead of 5k2
2. PIC code needs to be updated for new interrupt model (as well as remove of pt_regs)
3. use standard kernel debug macros
4. look at replacing sram_allocator w/rheap

if you repost the patch broken up into driver and platform bits it will be easier to provide more detailed comments.

- kumar

#8 lore raw
Quoted 15 lines · Kumar Gala
Quoted 13 lines · Nicolas DET

This 'big' patch adds support for CHRP/MPC52xx based platform. Here, this is the bPlan's Efika computer (http://www.bplan-gmbh.de/
efika_spec_en.html)

We know this patch is not totaly compliant with Documentation/ CodingStyle/.

We would like people to comment and review it. This way we would provide a new patch with the changes required (if any) for an upcomming merge in the kernel.

This patch has been applied on the kernel 2.6.18.1.

Quoted 7 lines
Quoted 2 lines

I think it would be extremely useful to get a dump of what the device- tree looks like. We have to be very careful to device nodes for 52xx components are competed consistent between this 'CHRP' style and a more embedded 52xx boards.

They should use either the /soc bus or the of_platform stuff (I'm coming up with some reworked of_platform that can be useful to them).

Ben.

#9 lore raw
In reply to Kumar Gala
Quoted 16 lines · Kumar Gala
Quoted 13 lines · Nicolas DET

This 'big' patch adds support for CHRP/MPC52xx based platform. Here, this is the bPlan's Efika computer (http://www.bplan-gmbh.de/
efika_spec_en.html)

We know this patch is not totaly compliant with Documentation/ CodingStyle/.

We would like people to comment and review it. This way we would provide a new patch with the changes required (if any) for an upcomming merge in the kernel.

This patch has been applied on the kernel 2.6.18.1.

Quoted 7 lines
Quoted 2 lines

I think it would be extremely useful to get a dump of what the device- tree looks like. We have to be very careful to device nodes for 52xx components are competed consistent between this 'CHRP' style and a more embedded 52xx boards.

Well, the efika is a CHRP board, and thus more akin to the pegasos or IBM boxes than the embedded 52xx boards with regard to the device trees. Nico will provide you an lsprop output.

Quoted 3 lines

Some high level comments:
2. PIC code needs to be updated for new interrupt model (as well as remove of pt_regs)

Notice that the efika does not have the interrupt tree bindings in the firmware. I implemented those on pegasos last month, but linux refused to boot, and i am currently waiting for comments from benh on the lsprop output.

Friendly,

Sven Luther

#10 lore raw
In reply to Kumar Gala
Quoted 8 lines · Kumar Gala
Quoted 1 line · Nicolas DET

This 'big' patch adds support for CHRP/MPC52xx based platform. Here, this is the bPlan's Efika computer (http://www.bplan-gmbh.de/efika_spec_en.html)

Some high level comments:
1. lets stick with the 52xx naming, instead of 5k2

Ok. Will be done

Quoted 1 line

2. PIC code needs to be updated for new interrupt model (as well as remove of pt_regs)

Ok.

Quoted 1 line

3. use standard kernel debug macros

Loads of changes in the pipe line ;-)

Quoted 2 lines

4. look at replacing sram_allocator w/rheap

This SRAM allocator is the exact same from the original Linux one. In fact, it is the original one. Would it be possible to accept this code as it is and schedule rheap integration later ?

Quoted 1 line

if you repost the patch broken up into driver and platform bits it will be easier to provide more detailed comments.

Ok. I'm alraedy working on it since I post the first one :-) . Thank you very much for your comments. I hope to re submit a much better patch today.

Regads

begin:vcard
fn:Nicolas DET ( bplan GmbH )
n:DET;Nicolas
org:bplan GmbH
adr:;;;;;;Germany
email;internet:[email protected]
title:Software Entwicklung
tel;work:+49 6171 9187 - 31
x-mozilla-html:FALSE
url:http://www.bplan-gmbh.de
version:2.1
end:vcard

#11 lore raw
Quoted 18 lines · Nicolas DET
Quoted 4 lines · Kumar Gala
Quoted 1 line · Nicolas DET

This 'big' patch adds support for CHRP/MPC52xx based platform.

Here, this is the bPlan's Efika computer (http://www.bplan-gmbh.de/efika_spec_en.html)

Quoted 4 lines

Some high level comments:
1. lets stick with the 52xx naming, instead of 5k2

Ok. Will be done

Quoted 1 line

2. PIC code needs to be updated for new interrupt model (as well

as remove of pt_regs)

Ok.

Quoted 8 lines
Quoted 1 line

3. use standard kernel debug macros

Loads of changes in the pipe line ;-)

Quoted 2 lines

4. look at replacing sram_allocator w/rheap

This SRAM allocator is the exact same from the original Linux one. In fact, it is the original one. Would it be possible to accept this code as it is and schedule rheap integration later ?

I dont follow, what do you mean by 'original Linux one' ?

Quoted 7 lines
Quoted 1 line

if you repost the patch broken up into driver and platform bits

it will be easier to provide more detailed comments.

Ok. I'm alraedy working on it since I post the first one :-) . Thank you very much for your comments. I hope to re submit a much better patch today.

Regads
<nd.vcf>

#12 lore raw
Quoted 9 lines · Kumar Gala
Quoted 7 lines
Quoted 3 lines

4. look at replacing sram_allocator w/rheap

This SRAM allocator is the exact same from the original Linux one. In fact, it is the original one. Would it be possible to accept this code as it is and schedule rheap integration later ?

I dont follow, what do you mean by 'original Linux one' ?

There are already some kind of Bestcomm DMA engine supoprt for some embbeded system (ARCH=ppc). In order to speed up the developement, the API is (almost) compatible so we can use the drivers wrote for this API. For example, the Ethernet driver enterily rely on this API. However, the 'soon to be release' ATA driver has been rewritten.

I did not at the time (yet) to look at rheap, but I know the current allocator is not really good (it can not free!).

I was just thinking, as this code is alraedy in some Linux tree, we could still use it there, and plan to replace it by a much better one (keeping the API compatible) later on.

Regards,

begin:vcard
fn:Nicolas DET ( bplan GmbH )
n:DET;Nicolas
org:bplan GmbH
adr:;;;;;;Germany
email;internet:[email protected]
title:Software Entwicklung
tel;work:+49 6171 9187 - 31
x-mozilla-html:FALSE
url:http://www.bplan-gmbh.de
version:2.1
end:vcard

#13 lore raw
Quoted 12 lines · Nicolas DET
Quoted 5 lines · Kumar Gala
Quoted 5 lines
Quoted 3 lines

4. look at replacing sram_allocator w/rheap

This SRAM allocator is the exact same from the original Linux

one. In

Quoted 1 line
Quoted 1 line

fact, it is the original one. Would it be possible to accept

this code

Quoted 2 lines
Quoted 1 line

as it is and schedule rheap integration later ?

Quoted 7 lines
Quoted 1 line

I dont follow, what do you mean by 'original Linux one' ?

There are already some kind of Bestcomm DMA engine supoprt for some embbeded system (ARCH=ppc). In order to speed up the developement, the API is (almost) compatible so we can use the drivers wrote for this API. For example, the Ethernet driver enterily rely on this API. However, the 'soon to be release' ATA driver has been rewritten.

I did not at the time (yet) to look at rheap, but I know the current allocator is not really good (it can not free!).

I was just thinking, as this code is alraedy in some Linux tree, we could still use it there, and plan to replace it by a much better one (keeping the API compatible) later on.

Is this in the kernel tree already, I am not coming across it.

- k

#14 lore raw
Quoted 6 lines
Quoted 4 lines

I was just thinking, as this code is alraedy in some Linux tree, we could still use it there, and plan to replace it by a much better one (keeping the API compatible) later on.

Is this in the kernel tree already, I am not coming across it.

Probably only in the WD tree and not even attempted to be merged like most of the gunk in there...

Ben

#15 lore raw
In reply to Kumar Gala
Quoted 26 lines · Kumar Gala
Quoted 23 lines · Nicolas DET
Quoted 4 lines · Kumar Gala
Quoted 1 line · Nicolas DET

This 'big' patch adds support for CHRP/MPC52xx based platform.

Here, this is the bPlan's Efika computer
(http://www.bplan-gmbh.de/efika_spec_en.html)

Quoted 4 lines

Some high level comments:
1. lets stick with the 52xx naming, instead of 5k2

Ok. Will be done

Quoted 1 line

2. PIC code needs to be updated for new interrupt model (as well as

remove of pt_regs)

Ok.

Quoted 1 line

3. use standard kernel debug macros

Loads of changes in the pipe line ;-)

Quoted 9 lines
Quoted 6 lines
Quoted 2 lines

4. look at replacing sram_allocator w/rheap

This SRAM allocator is the exact same from the original Linux one. In fact, it is the original one. Would it be possible to accept this code as it is and schedule rheap integration later ?

I dont follow, what do you mean by 'original Linux one' ?

There is a driver for Linux (not in the official kernel.org tree) for this chipset. This is the one I'm speaking about You can find it into various embbeded kernel like MontaVista tree, etc...

By the way, I uploaded the current patch on a private URL:
http://powernico.free.fr/chrpmpc52x_bigone.patch.bz2

If people wish to have a look.

begin:vcard
fn:Nicolas DET ( bplan GmbH )
n:DET;Nicolas
org:bplan GmbH
adr:;;;;;;Germany
email;internet:[email protected]
title:Software Entwicklung
tel;work:+49 6171 9187 - 31
x-mozilla-html:FALSE
url:http://www.bplan-gmbh.de
version:2.1
end:vcard

#16 lore raw

Ok, here's a batch of comments. They go on top of the comments I sent about your device-tree, which you might want to make public.

Also note that the size of the patch is partially due to stale ## and .orig files :)

In general, one big thing is: Don't bother with re-using the arch/ppc code, API, .h etc... There is not enough re-use there to justify it (FEC driver is new, serial can easily be changed, as for USB).

* /proc/ppc64/rtas changes: we need to continue that discussion separately and ask for paulus point of view. I prefer /proc/powerpc personally with a /proc/ppc64->/proc/powerpc symlink on 64 bits machines only.

* mpc5k2_bestcomm.c (& helpers):

- name changes (mpc5200_bestcomm.c would be better)
- location change -> arch/powerpc/sysdev
- look into using rheap instead of your current sram bits (and I don't
care about being compatible with arch/ppc code).
- the whole sdma_io_va/pa/... looks like bullcrap to me. The bestcomm
interface should use the normal dma mapping APIs
- I don't like having those helpers like fec or ata helpers in that
file. I think we need to rethink the API between the bestcomm and the drivers. I understand that part of this is related to the fact that the tasks themselves in the bestcomm are tailored for various devices and with different descriptor formats. That sucks. I suppose we can live with that for now (with a bit more cleanup) but we should have a healthy discussion on how to rework the whole thing properly.

In general, I think there's a lot more code in those files that would be necessary ;)

* mpc5k2.c :

Well, to start with, I don't think we need that file at all :) A lot of the stuff in there is completely unnecessary provided that you make better use of the device-tree.

The main thing is: use of_platform devices, not platform devices, so that you have a linkage with the device-tree. In fact, you could register of platform devices for every direct children of your "buildin" node (which is to be renamed, I hope, according to my comments about your device-tree).

You'll even be able to make good use of some work I've been doing lately for Cell by adding a call to a single function to the main chrp setup code that will register for you all those devices. You might want to keep PCI at the root of the tree for now though that won't even be necessary.

The core of platform matching code will take care of matching devices to drivers based on standard OF properties (name,device_type,compatible). Drivers will use standard prom_parse.c functions to obtain resources like addresses (instead of directly accessing "reg" property from a useless helper) etc...

But first, let's get your device-tree in shape :)

* mpc5k2_pic.c :

This should go in arch/powerpc/sysdev (and be renamed as for other files, something like mpc5200_pic.c would be fine).

There are several issues here, though some of them are definitely related to the chip design being completely alien to common sense:

- enable/disable/end shouldn't be used that way anymore. use
mask/unmask and set the appropriate IRQ flow handler for your IRQs. That is, basically, adapt your driver to genirq. That will simplify.

- You need to add an irq host, probably set it as default, with the
appropriate xlate, map, etc... routines as done by other controllers in 2.6.18.

- the if/else bits in enable/disable (to become mask/unmask) should
probably be changed into something a bit more sane. You have a large freedom of chosing what kind of IRQ "hardware" numbers you use. Since 2.6.18 new irq code, there is no direct relationship anymore between a linux irq number (virtual irq number) and a HW number manipulated by a PIC. Thus you can do what you want with those HW numbers, like, for example, using some top bits to represent the "base" (IRQ0...3, SDMA, PERP, ...) and use bottom bits for the actual bit in the mask. That will simplify your mask/unmask implementation.

* chrp/pci.c

I dislike the whole "is_bplan" thingy with ugly magic numbers. The is_pegasos with 3 values was already on the limit .... In fact, you probably don't even need it. The whole pcibios_fixup change is useless, let the fixup code run, it will do no harm, and in fact will be necessary as soon as you have a correct interrupt-tree anyway. Which means basically that you shoudn't even need to patch this file...

* chrp/setup.c

Did you do any change to chrp_init2() appart from moving it around ? Please don't move it for no reason or explain why you did it, it makes the patch bigger than necessary and more difficult to read.

Why did you add back bits that I removed, like:

+ if (_chrp_type == _CHRP_Pegasos)
+ ppc_md.get_irq = i8259_irq;

This should be handled by the chrp code just fine now (basically 8259 takes over when no mpic was found and sets that callback).

You may have noticed that I changed the code a bit too regaring how irq controllers are handled in chrp in 2.6.18. You should follow that instead of bringing back the old scheme. Basically add a chrp_find_mpc5200_pic or something like that and have it change ppc_md.get_irq itself if it finds it.

This all init2 vs init_alt business is very ugly. What is your reason for doing so ? What is chrp_init2 doesn't work for you ? (you should fix that instead).

* fec driver

(That driver should ultimately be submited to the netdev list & Jeff Garzik, but here's a fist pass at comments on things that need to be fixed before you get there).

I'm not sure you have enough files here to justify a subdir. Especially since you should probably get rid of fec_phy.c and replace it with the generic PHY code that Andy Fleming wrote in drivers/net/phy/. That old state machine PHY code from montavista has outlived its usefulness

The driver should be in drivers/net, in fact, it should just be one or two files: drivers/net/mpc5200_fec.{c,h}

Do not keep #ifdefs around for chrp/non-chrp. I don't want the mpc5200 stuff to have any difference between platforms. The driver wasn't in the tree before so I don't care about whatevber old API was used by an out of tree driver in arch/ppc. We are defining an API for use by any mpc5200 based device under arch/powerpc here, nothing specific to EFIKA or CHRP.

+ if ( _chrp_type == _CHRP_E5K2 )
+ return (132*1000*1000);
+

Gack ! Make that a property in the device-tree instead. By converting the driver to be an of_platform device, you'll get your device-node reference for free anyway.

kfree_skb() -> dev_kfree_skb() afaik. Also, you might want to make sure you are getting the right irq/non-irq/any version of it depending on the context you are calling it in.

You are never calling netif_carrier_on/off... might be useful to do so to inform the kernel about the state of your link... though if you use the generic PHY layer, it will do it for you.

+ if (status & 0x370000) {

What about using symbolic constants instead ?

virt_to_phys() is deprecated. See my comments in the bestcomm code too, use the dma mapping API instead.

Your rx code doesn't seem to do the fairly common optimisation of copying the data to a new skb when it's small enough rather than allocating a new one and replacing...

fec_interrupt() has some serious coding style issues.

I'm a bit dubious about the implementation of the re-init code... it looks racy vs. interrupts, and I don't like at all you calling back into open(). I'd rather have a separate common routine called by both open and reinit. You may also want to defer that to a work queue.

ethtool ioctls: You are using a deprecated interface. grep for ethtool_ops for the right way to do it.

+ while(!priv->sequence_done)
+ schedule();
+

This is evil. Don't do that. In fact, you should not need to do it, look at what other drivers do.

* serial stuff

Pretty much all of the same comments as the fec driver regarding things like:

+static unsigned long get_ipbfreq(void)
+{
+#ifdef CONFIG_PPC_CHRP
+ if ( _chrp_type == _CHRP_E5K2 )
+ return 132*1000*1000;
+
+ return 0;
+#else
+ return __res.bi_ipbfreq
+#endif
+}
+

Just get rid of the bi case and remove the ifdef. Now howevr that I see at least 2 drivers having the same function to get this ibp freq I'm starting to wonder why you don't just put the value in the device-tree possibly in the parent node ?

Same for the default baudrate. Remove the ifdef gunk. Use the baudrate that's passed in by the system and possibly implement something akin to check_legacy_serial_console() in arch/powerpc/kernel/legacy_serial.c to get proper auto-detect. (I would even accept adding a special case to detect the mpc5200 serial in there)

Make it an of_platform device anyway, that's better. Just ifdef to be compatible with the stuff in arch/ppc (I didn't even know we ever merged that... oh well).

* USB

Do an ohci-ppc-of.c that use an OF platform device for probing

Ben.

#17 lore raw

Curious question.

What are you guys going to do when the PowerPC name is defunct?

Like, last month :)

Power Architecture is where it is at. The trademark is even going to lapse. It's a bit too late for the ppc->powerpc tree breakout now, but wouldn't it just confuse people to be using a "Power Architecture" processor or SoC of some type, using collections of definitions from the Power ISA 2.03 and have this "powerpc" thing pop up?

It confused me even before, because ppc and ppc64 have also been used to support real POWER (with a capital P, O, W, E and R) processors, and now these are lumped in with powerpc which is no better than ppc64 in these terms?

Just flexing my marketing exec muscles, see if they work, never done it before. Oh... *crack*.. that wasn't a good noise :]

Signature · 2 lines
Matt Sealey <[email protected]>
Genesi, Manager, Developer Relations
Quoted 15 lines · Benjamin Herrenschmidt

Ok, here's a batch of comments. They go on top of the comments I sent about your device-tree, which you might want to make public.

Also note that the size of the patch is partially due to stale ## and .orig files :)

In general, one big thing is: Don't bother with re-using the arch/ppc code, API, .h etc... There is not enough re-use there to justify it (FEC driver is new, serial can easily be changed, as for USB).

* /proc/ppc64/rtas changes: we need to continue that discussion separately and ask for paulus point of view. I prefer /proc/powerpc personally with a /proc/ppc64->/proc/powerpc symlink on 64 bits machines only.

[snip]

#18 lore raw
Quoted 11 lines · Matt Sealey

Curious question.

What are you guys going to do when the PowerPC name is defunct?

Like, last month :)

Power Architecture is where it is at. The trademark is even going to lapse. It's a bit too late for the ppc->powerpc tree breakout now, but wouldn't it just confuse people to be using a "Power Architecture" processor or SoC of some type, using collections of definitions from the Power ISA 2.03 and have this "powerpc" thing pop up?

Who the f***k cares. We've always given a chip what the newspeak party line of the days is, and we will here aswell. If the IBM/Mot/PowerYodda overlords are unhappy with that it's their problem, not ours.

#19 lore raw

Very eloquently put.

Signature · 2 lines
Matt Sealey <[email protected]>
Genesi, Manager, Developer Relations
Quoted 17 lines · Christoph Hellwig
Quoted 11 lines · Matt Sealey

Curious question.

What are you guys going to do when the PowerPC name is defunct?

Like, last month :)

Power Architecture is where it is at. The trademark is even going to lapse. It's a bit too late for the ppc->powerpc tree breakout now, but wouldn't it just confuse people to be using a "Power Architecture" processor or SoC of some type, using collections of definitions from the Power ISA 2.03 and have this "powerpc" thing pop up?

Who the f***k cares. We've always given a chip what the newspeak party line of the days is, and we will here aswell. If the IBM/Mot/PowerYodda overlords are unhappy with that it's their problem, not ours.

#20 lore raw
In reply to Matt Sealey
Quoted 3 lines

What are you guys going to do when the PowerPC name is defunct?

Like, last month :)

If some marketeers want to rewrite history, let them try; but we cannot, simply because we use Git. Keeps people honest ;-)

Segher

#21 lore raw
In reply to Matt Sealey
Quoted 3 lines · Matt Sealey

Curious question.

What are you guys going to do when the PowerPC name is defunct?

It's not really, is it ? I don't care anyway, it will stay PowerPC in linux of course :)

Quoted 4 lines

Like, last month :)

Power Architecture is where it is at. The trademark is even going to lapse

Who cares ? Besides, while PAPR says "Power architecture", the processor architecture specification says PowerPC :)

Quoted 4 lines

. It's a bit too late for the ppc->powerpc tree breakout now, but wouldn't it just confuse people to be using a "Power Architecture" processor or SoC of some type, using collections of definitions from the Power ISA 2.03 and have this "powerpc" thing pop up?

It's PowerPC ISA :)

Quoted 4 lines

It confused me even before, because ppc and ppc64 have also been used to support real POWER (with a capital P, O, W, E and R) processors, and now these are lumped in with powerpc which is no better than ppc64 in these terms?

Historically, POWER means something else ... then with POWER3, POWER processors became compatible with the PowerPC architecture, then IBM played name changing game a couple of times and nobody knows what's up anymore :)

Quoted 2 lines

Just flexing my marketing exec muscles, see if they work, never done it before. Oh... *crack*.. that wasn't a good noise :]

AFAIK, the processor instruction set architecture is PowerPC an that will not change.

Ben.

#22 lore raw
In reply to Matt Sealey
Quoted 4 lines · Matt Sealey

It confused me even before, because ppc and ppc64 have also been used to support real POWER (with a capital P, O, W, E and R) processors, and now these are lumped in with powerpc which is no better than ppc64 in these terms?

"PowerPC" is the name of the instruction set architecture, "POWER" is the name of some specific implementations of the PowerPC architecture made by a certain large multinational computer company. Unless you're talking about the really really old ISA that was the predecessor of PowerPC, that is. :)

Paul.

#23 lore raw
Quoted 7 lines · Paul Mackerras

Matt Sealey writes:

"PowerPC" is the name of the instruction set architecture, "POWER" is the name of some specific implementations of the PowerPC architecture made by a certain large multinational computer company. Unless you're talking about the really really old ISA that was the predecessor of PowerPC, that is. :)

http://www.power.org/news/articles/new_brand/#isa

I'm talking about the one they consolidated from all the different specs Freescale, AMCC and IBM had about what makes a PowerPC processor. Since POWER is part of it and it's all Power.org now.. they seem to have changed the name.

I don't think they have actually added anything new, that wasn't there before, just consolidated it. You won't seen an IBM chip with SPE or VLE unless they license the gaggle of patents Freescale have on those units, maybe.. but there is always a possibility. All new chips are being called Power (not POWER).

Signature · 2 lines
Matt Sealey <[email protected]>
Genesi, Manager, Developer Relations
#24 lore raw
Quoted 17 lines · Matt Sealey
Quoted 7 lines · Paul Mackerras

Matt Sealey writes:

"PowerPC" is the name of the instruction set architecture, "POWER" is the name of some specific implementations of the PowerPC architecture made by a certain large multinational computer company. Unless you're talking about the really really old ISA that was the predecessor of PowerPC, that is. :)

http://www.power.org/news/articles/new_brand/#isa

I'm talking about the one they consolidated from all the different specs Freescale, AMCC and IBM had about what makes a PowerPC processor. Since POWER is part of it and it's all Power.org now.. they seem to have changed the name.

Quoted 6 lines

I don't think they have actually added anything new, that wasn't there before, just consolidated it. You won't seen an IBM chip with SPE or VLE unless they license the gaggle of patents Freescale have on those units, maybe.. but there is always a possibility. All new chips are being called Power (not POWER).

The ISA version 2.03 that you are talking about does, in fact, rename it to "Power ISA" (big "P" little "ower") from "PowerPC ISA". This spec is mostly a big glom of all of FSL and IBM's architecure documentation, although there were some minor changes made and some things were added. If you read the spec carefully (which, btw, I do *not* recommend, as it is about 900 pages of spaghetti documentation), you will find things in there that do not exist in any implemented processor to date.

Cheers,
B

#25 lore raw
In reply to Matt Sealey
Quoted 6 lines · Matt Sealey

http://www.power.org/news/articles/new_brand/#isa

I'm talking about the one they consolidated from all the different specs Freescale, AMCC and IBM had about what makes a PowerPC processor. Since POWER is part of it and it's all Power.org now.. they seem to have changed the name.

So, are you proposing we drop support for all the PowerPC processors that don't conform to the recently-published V2.03 ISA? You don't want 6xx, 7xx, 7xxx, 52xx, etc. supported any more? :) (joke :)

Seriously, the v2.03 ISA covers only IBM's latest POWER processors (i.e. not POWER3, POWER4, or even POWER5), and Book E processors. We support a lot more than that.

Paul.

#26 lore raw

It supports the original processors, G3, G4 etc. as well. Maybe not the 601? I'm not sure. All the PowerPC ISA Books are in there and information only seems to have been ADDED (or moved to another book or been given a new book).

If a processor is not supported it was probably designed and end-of-lined well before they made the original "Book" specs :D

Have you even looked at it? Even the dumb 'evolution' picture on Power.org's site makes this quite, quite clear. Most of the spec is identical, to the letter, over vast amounts of pages. There are change bars in front of everything that changed and those that have; it seems they are just lifted and reworked from Freescale's AltiVec/SPE documentation, reworded in Book E so you can have a Book E with AltiVec, or has it's own new book (VLE, which is an IBM-ified copy of the e200 Core Reference Manual Addendum)

Signature · 2 lines
Matt Sealey <[email protected]>
Genesi, Manager, Developer Relations
Quoted 18 lines · Paul Mackerras
Quoted 6 lines · Matt Sealey

http://www.power.org/news/articles/new_brand/#isa

I'm talking about the one they consolidated from all the different specs Freescale, AMCC and IBM had about what makes a PowerPC processor. Since POWER is part of it and it's all Power.org now.. they seem to have changed the name.

So, are you proposing we drop support for all the PowerPC processors that don't conform to the recently-published V2.03 ISA? You don't want 6xx, 7xx, 7xxx, 52xx, etc. supported any more? :) (joke :)

Seriously, the v2.03 ISA covers only IBM's latest POWER processors (i.e. not POWER3, POWER4, or even POWER5), and Book E processors. We support a lot more than that.

Paul.

#27 lore raw
Quoted 8 lines · Matt Sealey

It supports the original processors, G3, G4 etc. as well. Maybe not the 601? I'm not sure. All the PowerPC ISA Books are in there and information only seems to have been ADDED (or moved to another book or been given a new book).

If a processor is not supported it was probably designed and end-of-lined well before they made the original "Book" specs :D

Quoted 9 lines

Have you even looked at it? Even the dumb 'evolution' picture on Power.org's site makes this quite, quite clear. Most of the spec is identical, to the letter, over vast amounts of pages. There are change bars in front of everything that changed and those that have; it seems they are just lifted and reworked from Freescale's AltiVec/SPE documentation, reworded in Book E so you can have a Book E with AltiVec, or has it's own new book (VLE, which is an IBM-ified copy of the e200 Core Reference Manual Addendum)

There is far more new in the 2.03 spec than just merging of various existing specs as you are stating. Please be more careful with general statements like this. The Book III-E has some significant changes from Book-E 1.0.

- k

#28 lore raw
Quoted 9 lines · Kumar Gala
Quoted 4 lines

so you can have a Book E with AltiVec, or has it's own new book (VLE, which is an IBM-ified copy of the e200 Core Reference Manual Addendum)

There is far more new in the 2.03 spec than just merging of various existing specs as you are stating. Please be more careful with general statements like this. The Book III-E has some significant changes from Book-E 1.0.

I don't think any of it suddenly makes a billion existing processors completely incompatible.

Signature · 2 lines
Matt Sealey <[email protected]>
Genesi, Manager, Developer Relations
#29 lore raw
In reply to Matt Sealey
Quoted 4 lines · Matt Sealey

It supports the original processors, G3, G4 etc. as well. Maybe not the 601? I'm not sure. All the PowerPC ISA Books are in there and information only seems to have been ADDED (or moved to another book or been given a new book).

Why are we still talking about this? None of it matters.

Who cares if the new ISA spec incorporates all of the old ones? The old specs cannot be unpublished. powerpc is a sufficient name for kernel source purposes and it will continue to have meaning in the minds of developers. Marketing glossies can (and do) use the name de jour, but we don't have to.

As a side note; The move from ppc/ppc64->powerpc was a *technical* decision, not a marketing one. It was done so there would be only one code base. The choice of arch/powerpc was almost arbitrary.

Cheers,
g.

Signature · 4 lines
Grant Likely, B.Sc. P.Eng.
Secret Lab Technologies Ltd.
[email protected]
(403) 399-0195
#30 lore raw

Arbitrary naming is fine.

Thinking up a bunch of excuses like "we refuse to bow to our corporate marketing overlords" and "but it's definitely called PowerPC ISA, and anyway, that new one which has a new name, I didn't read anyway, but it undefines every processor we support" is just.. well I dunno. Making excuses I guess.

I'm fine with the name I was just curious (see first line of the original email) about how these things ARE named. It lead to a query I am still wondering about the answer to; Genesi/bplan could have a new G5 platform and whatever support code is required submitted to the tree soon. Since it will have a fairly-PAPR-compliant firmware, this makes it logical to put it in the CHRP platform for us. But CHRP doesn't include LPAR, this is a pSeries platform feature. Segher suggested we use the Maple tree (!!) as this runs SLOF.

What platform do you think a 970MP box with a CHRP-compliant (well, let us say, Pegasos-compatible) firmware go, if we eventually want to support LPAR, especially without code duplication and MBs of "pull this out of pSeries and drop it into CHRP, even though CHRP technically doesn't support any of this" stuff? Especially since Ben is being quite strict about what goes into CHRP nowadays.

Is it time for a PAPR platform? Or is that against the Linux anti-marketing-name lobby we have in here?

Signature · 2 lines
Matt Sealey <[email protected]>
Genesi, Manager, Developer Relations
Quoted 21 lines · Grant Likely
Quoted 4 lines · Matt Sealey

It supports the original processors, G3, G4 etc. as well. Maybe not the 601? I'm not sure. All the PowerPC ISA Books are in there and information only seems to have been ADDED (or moved to another book or been given a new book).

Why are we still talking about this? None of it matters.

Who cares if the new ISA spec incorporates all of the old ones? The old specs cannot be unpublished. powerpc is a sufficient name for kernel source purposes and it will continue to have meaning in the minds of developers. Marketing glossies can (and do) use the name de jour, but we don't have to.

As a side note; The move from ppc/ppc64->powerpc was a *technical* decision, not a marketing one. It was done so there would be only one code base. The choice of arch/powerpc was almost arbitrary.

Cheers,
g.

#31 lore raw
Quoted 1 line · On Wed, 25 Oct 2006 18:18:43 +0200 Matt Sealey <[email protected]> wrote:

Is it time for a PAPR platform?

There is a PAPR platform in the tree already, it's named pseries.

-Olof

#32 lore raw
In reply to Matt Sealey
Quoted 2 lines · Matt Sealey

Is it time for a PAPR platform? Or is that against the Linux anti-marketing-name lobby we have in here?

:) All I would say is that if your platform really is PAPR compliant, then sure, we could have a papr platform, but don't underestimate the amount of work it will take to be PAPR compliant. There is a ton of stuff in there.

Paul.

#33 lore raw
In reply to Matt Sealey
Quoted 4 lines · Matt Sealey

It supports the original processors, G3, G4 etc. as well. Maybe not the 601? I'm not sure. All the PowerPC ISA Books are in there and information only seems to have been ADDED (or moved to another book or been given a new book).

The only 32-bit processors it supports are Book E processors. IBM removed all the material relating to 32-bit implementations from the earlier V2.0x architectures, because IBM was no longer interested in 32-bit server processors.

Paul.