Re: Correct way to format spufs file output.

Re: Correct way to format spufs file output.

[Arnd Bergmann]

On Thursday 19 October 2006 05:30, Dwayne Grant McConnell wrote:
> In a recent submission I added the lslr file and used "%llx" for the 
> format string. You mentioned that it should probably be "0x%llx" so it 
> would be clearly parsed as hex so I changed it in the next submission. But 
> I noticed that there seems to be some inconsistent usage of 0x as follows:

Thanks for bringing this up, I guess I screwed up in some way here, so
we should fix it up one way or another:

> signal1_type (%llu)
> signal2_type (%llu)

These are fine, they can only ever be 1 or 0.

> npc (%llx)

I think we used to access this in _very_ old versions of libspe,
before we move to a syscall based interface.

> decr (%llx)
> decr_status (%llx)
> spu_tag_mask (%llx)
> event_mask (%llx)
> event_status (%llx)
> srr0 (%llx)

These are used exclusively for debugging purposes, and no publically
available version of gdb accesses them, so I guess we can still change
them, although it's not nice.

> phys_id (0x%llx)

This one is used in some forks of libspe, we should not change it.

> object_id (0x%llx)

This is used in libspe, gdb and oprofile, but only in fairly recent
versions.

> lslr (0x%llx)

As this is introduced by your own patch, there is no precedent for
it yet.

Current kernels now also have 'cntl' (0x%08lx), which was introduced
in 2.6.19 and is so far unused. I guess we should change that one
to be consistant with the others as well.

> Should all the %llx be changed to 0x%llx or should the 0x be dropped from 
> those that have it or is the inconsistency acceptable?

I'd rather have it consistant. Moreover, I guess the "%llx" format is
actually harmful, because that means you can not use the same format
for read and write. The simple_attr_write function currently uses
the simple_strtol helper to interpret the value written to it, and that
requires the input to be wither decimal, or hexadecimal with a preceding
0x. I'd suggest we change all files to take a 0x%llx format on output.

	Arnd <><

Re: Correct way to format spufs file output.

[Dwayne Grant McConnell]

On Fri, 20 Oct 2006, Arnd Bergmann wrote:

> On Thursday 19 October 2006 05:30, Dwayne Grant McConnell wrote:
> > In a recent submission I added the lslr file and used "%llx" for the 
> > format string. You mentioned that it should probably be "0x%llx" so it 
> > would be clearly parsed as hex so I changed it in the next submission. But 
> > I noticed that there seems to be some inconsistent usage of 0x as follows:
> 
> Thanks for bringing this up, I guess I screwed up in some way here, so
> we should fix it up one way or another:
> 
> > signal1_type (%llu)
> > signal2_type (%llu)
> 
> These are fine, they can only ever be 1 or 0.
> 
> > npc (%llx)
> 
> I think we used to access this in _very_ old versions of libspe,
> before we move to a syscall based interface.
> 
> > decr (%llx)
> > decr_status (%llx)
> > spu_tag_mask (%llx)
> > event_mask (%llx)
> > event_status (%llx)
> > srr0 (%llx)
> 
> These are used exclusively for debugging purposes, and no publically
> available version of gdb accesses them, so I guess we can still change
> them, although it's not nice.
> 
> > phys_id (0x%llx)
> 
> This one is used in some forks of libspe, we should not change it.
> 
> > object_id (0x%llx)
> 
> This is used in libspe, gdb and oprofile, but only in fairly recent
> versions.
> 
> > lslr (0x%llx)
> 
> As this is introduced by your own patch, there is no precedent for
> it yet.
> 
> Current kernels now also have 'cntl' (0x%08lx), which was introduced
> in 2.6.19 and is so far unused. I guess we should change that one
> to be consistant with the others as well.
> 
> > Should all the %llx be changed to 0x%llx or should the 0x be dropped from 
> > those that have it or is the inconsistency acceptable?
> 
> I'd rather have it consistant. Moreover, I guess the "%llx" format is
> actually harmful, because that means you can not use the same format
> for read and write. The simple_attr_write function currently uses
> the simple_strtol helper to interpret the value written to it, and that
> requires the input to be wither decimal, or hexadecimal with a preceding
> 0x. I'd suggest we change all files to take a 0x%llx format on output.

I think %0xllx is the way to go. I would even advocate changing 
signal1_type and signal2_type unless it is actually too dangerous. Is 
there even a case where changing from %llu to %0xllx would break things? 
Perhaps with the combination of a old library with a new kernel?

-- 
Dwayne Grant McConnell <decimal@us.ibm.com>
Lotus Notes Mail: Dwayne McConnell [Mail]/Austin/IBM@IBMUS
Lotus Notes Calendar: Dwayne McConnell [Calendar]/Austin/IBM@IBMUS

Re: Correct way to format spufs file output.

[Arnd Bergmann]

On Friday 20 October 2006 15:54, Dwayne Grant McConnell wrote:
> I think %0xllx is the way to go. I would even advocate changing 
> signal1_type and signal2_type unless it is actually too dangerous.

There is absolutely no reason why these should be hexadecimal, they
are basically implementing a bool.

> Is there even a case where changing from %llu to %0xllx would break things? 
> Perhaps with the combination of a old library with a new kernel?

Right, a library or some script that has been written assuming there
is no leading 0x.

	Arnd <><

Re: Correct way to format spufs file output.

[Dwayne Grant McConnell]

On Fri, 20 Oct 2006, Arnd Bergmann wrote:

> On Friday 20 October 2006 15:54, Dwayne Grant McConnell wrote:
> > I think %0xllx is the way to go. I would even advocate changing 
> > signal1_type and signal2_type unless it is actually too dangerous.
> 
> There is absolutely no reason why these should be hexadecimal, they
> are basically implementing a bool.
> 
> > Is there even a case where changing from %llu to %0xllx would break things? 
> > Perhaps with the combination of a old library with a new kernel?
> 
> Right, a library or some script that has been written assuming there
> is no leading 0x.

Okay. Thanks for considering it. I'll leave the %llu as is.

-- 
Dwayne Grant McConnell <decimal@us.ibm.com>
Lotus Notes Mail: Dwayne McConnell [Mail]/Austin/IBM@IBMUS
Lotus Notes Calendar: Dwayne McConnell [Calendar]/Austin/IBM@IBMUS

Re: [Cbe-oss-dev] Correct way to format spufs file output.

[Benjamin Herrenschmidt]

On Fri, 2006-10-20 at 16:38 +0200, Arnd Bergmann wrote:
> On Friday 20 October 2006 15:54, Dwayne Grant McConnell wrote:
> > I think %0xllx is the way to go. I would even advocate changing 
> > signal1_type and signal2_type unless it is actually too dangerous.
> 
> There is absolutely no reason why these should be hexadecimal, they
> are basically implementing a bool.
> 
> > Is there even a case where changing from %llu to %0xllx would break things? 
> > Perhaps with the combination of a old library with a new kernel?
> 
> Right, a library or some script that has been written assuming there
> is no leading 0x.

We might still want to "broadcast" that we might change that and see how
much gets broken. It's very likely that nothing will.

Ben.