Re: net-libertas: Better exception handling in if_spi_host_to_card_worker()
On 4 January 2016 at 11:05, Arend van Spriel wrote:
> On 03-01-16 16:18, Rafał Miłecki wrote:
>> On 3 January 2016 at 10:36, Arend van Spriel wrote:
>>> On 02-01-16 12:21, SF Markus Elfring wrote:
> Did you look at the resulting assembly code for different target
> architectures?
Not yet. - Which execution system variants would you recommend for
further comparisons?
>>>
>>> Guess x86{,_64} and arm would be good candidates, ie. CISC vs. RISC.
>>
>> Oh, don't forget about MIPS with its fancy branches handling. You know
>> about it, don't you?
>
> You are asking me, right ;-) ? I have come across my share of MIPS
> platforms here at Broadcom, but I still try to avoid them as much as
> possible.
I was more thinking about author on this patch. But it's indeed an
interesting thing, just to know, how MIPS CPU handles branches ;)
--
Rafał
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Re: net-libertas: Better exception handling in if_spi_host_to_card_worker()
On 03-01-16 16:18, Rafał Miłecki wrote:
> On 3 January 2016 at 10:36, Arend van Spriel wrote:
>> On 02-01-16 12:21, SF Markus Elfring wrote:
Did you look at the resulting assembly code for different target
architectures?
>>>
>>> Not yet. - Which execution system variants would you recommend for
>>> further comparisons?
>>
>> Guess x86{,_64} and arm would be good candidates, ie. CISC vs. RISC.
>
> Oh, don't forget about MIPS with its fancy branches handling. You know
> about it, don't you?
You are asking me, right ;-) ? I have come across my share of MIPS
platforms here at Broadcom, but I still try to avoid them as much as
possible.
> I'm against this patch as well.
and counting... :-p
Regards,
Arend
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Re: net-libertas: Better exception handling in if_spi_host_to_card_worker()
On 3 January 2016 at 10:36, Arend van Spriel wrote:
> On 02-01-16 12:21, SF Markus Elfring wrote:
>>> Did you look at the resulting assembly code for different target
>>> architectures?
>>
>> Not yet. - Which execution system variants would you recommend for
>> further comparisons?
>
> Guess x86{,_64} and arm would be good candidates, ie. CISC vs. RISC.
Oh, don't forget about MIPS with its fancy branches handling. You know
about it, don't you?
I'm against this patch as well.
--
Rafał
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Re: net-libertas: Better exception handling in if_spi_host_to_card_worker()
>>> What the patch tries to do is avoid the extra 'if (err)'. >> >> Yes. - I propose to look at related consequences together with the usage >> of a popular short jump label once more. > > When I read a subject saying "Better exception handling" it sounds like > a functional improvement. Your change does not change anything > functionally and may or may not save a bit of execution time depending > on how smart the compiler is. Can it eventually matter to skip another condition check in three cases? > What you change does is confuse people reading the code. A few software developers might find this proposal unusual. > So please explain why your update improves exception handling here. > I don't see it. How does this feedback fit to the mentioned check avoidance? > The code is not making the driver more robust against failures That's true for this update suggestion. > in this function, which is what I think of reading "better exception > handling". Other implementation details are affected by the shown fine-tuning. Regards, Markus -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: net-libertas: Better exception handling in if_spi_host_to_card_worker()
On 02-01-16 12:21, SF Markus Elfring wrote:
>> I have never seen much evolution going on in this area.
>
> I can get an other impression from a specific document for example.
> https://git.kernel.org/cgit/linux/kernel/git/next/linux-next.git/log/Documentation/CodingStyle
>
>
>> What the patch tries to do is avoid the extra 'if (err)'.
>
> Yes. - I propose to look at related consequences together with the usage
> of a popular short jump label once more.
When I read a subject saying "Better exception handling" it sounds like
a functional improvement. Your change does not change anything
functionally and may or may not save a bit of execution time depending
on how smart the compiler is. What you change does is confuse people
reading the code. So please explain why your update improves exception
handling here. I don't see it. The code is not making the driver more
robust against failures in this function, which is what I think of
reading "better exception handling".
>> Setting coding style aside, the question is whether there is
>> a good metric for the patch.
>
> A software development challenge is to accept changes also around a topic
> like "error handling". My update suggestion for the source file
> "drivers/net/wireless/marvell/libertas/if_spi.c" should only improve
> exception handling. (I came along other source files where more improvements
> will eventually be possible.)
>
> When will the run-time behaviour matter also for exceptional situations?
>
>
>> Did you look at the resulting assembly code for different target
>> architectures?
>
> Not yet. - Which execution system variants would you recommend for
> further comparisons?
Guess x86{,_64} and arm would be good candidates, ie. CISC vs. RISC.
Regards,
Arend
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Re: net-libertas: Better exception handling in if_spi_host_to_card_worker()
> I have never seen much evolution going on in this area. I can get an other impression from a specific document for example. https://git.kernel.org/cgit/linux/kernel/git/next/linux-next.git/log/Documentation/CodingStyle > What the patch tries to do is avoid the extra 'if (err)'. Yes. - I propose to look at related consequences together with the usage of a popular short jump label once more. > Setting coding style aside, the question is whether there is > a good metric for the patch. A software development challenge is to accept changes also around a topic like "error handling". My update suggestion for the source file "drivers/net/wireless/marvell/libertas/if_spi.c" should only improve exception handling. (I came along other source files where more improvements will eventually be possible.) When will the run-time behaviour matter also for exceptional situations? > Did you look at the resulting assembly code for different target > architectures? Not yet. - Which execution system variants would you recommend for further comparisons? Regards, Markus -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: net-libertas: Better exception handling in if_spi_host_to_card_worker()
On 02-01-16 10:08, SF Markus Elfring wrote: >>> I assume that a software development taste can evolve, can't it? >> >> So far, you have gotten several down votes for this kind of change, > > I am curious when more contributors will share corresponding opinions. Let's burn some cycles on this while the holidays give me time to do so. "software development taste" is another term for "coding style". In every project battles are fought over this between friends and foes. I have never seen much evolution going on in this area. >> and no enthusiasm. > > How many software designers and developers can become enthusiastic > about better exception handling to some degree? I had to take a look at this particular patch and I have to say that I don't see, using your favorite term, evolution at work. It looks more like the result of inbred. What the patch tries to do is avoid the extra 'if (err)'. Setting coding style aside, the question is whether there is a good metric for the patch. So does it really safe processing time? Did you look at the resulting assembly code for different target architectures? You got pushed back on the change so you have to come up with solid arguments for your change instead of spewing ideas about evolution in software development. Running Coccinelle is one thing, but understanding the results and what you are ultimately proposing to be changed is more important. Regards, Arend >> The code that is performance critical, you should probably not touch, ever. > > I imagine that technical evolution will result in further considerations > so that "unchangeable" components can be adjusted once more. > > >> The people who wrote it knew what was important and what was not. > > I might come along at some places where the affected knowledge will also > evolve. > > Regards, > Markus > -- > To unsubscribe from this list: send the line "unsubscribe linux-wireless" in > the body of a message to majord...@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html > -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: net-libertas: Better exception handling in if_spi_host_to_card_worker()
>> I assume that a software development taste can evolve, can't it? > > So far, you have gotten several down votes for this kind of change, I am curious when more contributors will share corresponding opinions. > and no enthusiasm. How many software designers and developers can become enthusiastic about better exception handling to some degree? > The code that is performance critical, you should probably not touch, ever. I imagine that technical evolution will result in further considerations so that "unchangeable" components can be adjusted once more. > The people who wrote it knew what was important and what was not. I might come along at some places where the affected knowledge will also evolve. Regards, Markus -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
