LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
From: Peter Rosin <peda@axentia.se>
To: Liam Beguin <liambeguin@gmail.com>,
	jic23@kernel.org, lars@metafoo.de, pmeerw@pmeerw.net
Cc: linux-kernel@vger.kernel.org, linux-iio@vger.kernel.org,
	devicetree@vger.kernel.org, robh+dt@kernel.org
Subject: Re: [PATCH v6 05/13] iio: afe: rescale: add INT_PLUS_{MICRO,NANO} support
Date: Fri, 30 Jul 2021 09:01:52 +0200	[thread overview]
Message-ID: <8448cecb-16b3-96ff-dfa8-3933325ef946@axentia.se> (raw)
In-Reply-To: <a77e7a18-67dc-5bc5-427b-89d6b1e82b85@axentia.se>

On 2021-07-30 08:49, Peter Rosin wrote:
> On 2021-07-29 17:56, Liam Beguin wrote:
>> On Wed Jul 28, 2021 at 3:19 AM EDT, Peter Rosin wrote:
>>> On 2021-07-28 02:21, Liam Beguin wrote:
>>>> On Fri Jul 23, 2021 at 5:16 PM EDT, Peter Rosin wrote:
>>>>> On 2021-07-21 05:06, Liam Beguin wrote:
>>>>>> From: Liam Beguin <lvb@xiphos.com>
>>>>>>
>>>>>> Some ADCs use IIO_VAL_INT_PLUS_{NANO,MICRO} scale types.
>>>>>> Add support for these to allow using the iio-rescaler with them.
>>>>>>
>>>>>> Signed-off-by: Liam Beguin <lvb@xiphos.com>
>>>>>> ---
>>>>>>  drivers/iio/afe/iio-rescale.c | 14 ++++++++++++++
>>>>>>  1 file changed, 14 insertions(+)
>>>>>>
>>>>>> diff --git a/drivers/iio/afe/iio-rescale.c b/drivers/iio/afe/iio-rescale.c
>>>>>> index d0669fd8eac5..2b73047365cc 100644
>>>>>> --- a/drivers/iio/afe/iio-rescale.c
>>>>>> +++ b/drivers/iio/afe/iio-rescale.c
>>>>>> @@ -41,6 +41,20 @@ int rescale_process_scale(struct rescale *rescale, int scale_type,
>>>>>>  		do_div(tmp, 1000000000LL);
>>>>>>  		*val = tmp;
>>>>>>  		return scale_type;
>>>>>> +	case IIO_VAL_INT_PLUS_NANO:
>>>>>> +		tmp = ((s64)*val * 1000000000LL + *val2) * rescale->numerator;
>>>>>> +		tmp = div_s64(tmp, rescale->denominator);
>>>>>> +
>>>>>> +		*val = div_s64(tmp, 1000000000LL);
>>>>>> +		*val2 = tmp - *val * 1000000000LL;
>>>>>> +		return scale_type;
>>>>
>>>> Hi Peter,
>>>>
>>>>>
>>>>> Hi!
>>>>>
>>>>> My objection from v5 still stands. Did you forget or did you simply send
>>>>> the
>>>>> wrong patch?
>>>>
>>>> Apologies, again I didn't mean to make it seem like I ignored your comments.
>>>> I tried your suggestion, but had issues when *val2 would overflow into
>>>> the integer part.
>>
>> Hi Peter,
>>
>>>
>>> Not saying anything about it not working does indeed make it seem like
>>> you
>>> ignored it :-) Or did I just miss where you said this? Anyway, no
>>> problem,
>>> it can be a mess dealing with a string of commits when there are
>>> numerous
>>> things to take care of between each iteration. And it's very easy to
>>> burn
>>> out and just back away. Please don't do that!
>>
>> It was my mistake. Thanks for the encouragement :-)
>>
>>>
>>>> Even though what I has was more prone to integer overflow with the first
>>>> multiplication, I thought it was still a valid solution as it passed the
>>>> tests.
>>>
>>> I did state that you'd need to add overflow handling from the fraction
>>> calculation and handling for negative values, so it was no surprise that
>>> my original sketchy suggestion didn't work as-is.
>>>
>>>>
>>>>>
>>>>> Untested suggestion, this time handling negative values and
>>>>> canonicalizing any
>>>>> overflow from the fraction calculation.
>>>>>
>>>>> neg = *val < 0 || *val2 < 0;
>>>>> tmp = (s64)abs(*val) * rescale->numerator;
>>>>> rem = do_div(tmp, rescale->denominator);
>>>>> *val = tmp;
>>>>> tmp = rem * 1000000000LL + (s64)abs(*val2) * rescale->numerator;
>>>>> do_div(tmp, rescale->denominator);
>>>>> *val2 = do_div(tmp, 1000000000LL);
>>>>> *val += tmp;
>>>>> if (neg) {
>>>>> if (*val < 0)
>>>>> *val = -*val;
>>>>> else
>>>>> *val2 = -*val;
>>>
>>> This last line should of course be *val2 = -*val2;
>>> Sorry.
>>>
>>>>
>>>> I'll look into this suggestion.
>>>
>>> Thanks!
>>>
>>
>> Starting from what you suggested, here's what I came up with.
>> I also added a few test cases to cover corner cases.
>>
>> 	if (scale_type == IIO_VAL_INT_PLUS_NANO)
>> 		mult = 1000000000LL;
>> 	else
>> 		mult = 1000000LL;
>> 	/*
>> 	 * For IIO_VAL_INT_PLUS_{MICRO,NANO} scale types if *val OR
>> 	 * *val2 is negative the schan scale is negative
>> 	 */
>> 	neg = *val < 0 || *val2 < 0;
>>
>> 	tmp = (s64)abs(*val) * (s32)abs(rescale->numerator);
> 
> Small nit, but I think abs() returns a signed type compatible
> with the argument type. I.e. (s32)abs(rescale->...) where both
> numerator and denominator are already s32 could just as well
> be written without the cast as plain old abs(rescale->...)
> 
> 
>> 	*val = div_s64_rem(tmp, (s32)abs(rescale->denominator), &rem);
>>
>> 	tmp = (s64)rem * mult +
>> 		(s64)abs(*val2) * (s32)abs(rescale->numerator);
>> 	tmp = div_s64(tmp, (s32)abs(rescale->denominator));
>>
>> 	*val += div_s64_rem(tmp, mult, val2);
>>
>> 	/*
>> 	 * If the schan scale or only one of the rescaler elements is
>> 	 * negative, the combined scale is negative.
>> 	 */
>> 	if (neg || ((rescale->numerator < 0) ^ (rescale->denominator < 0)))

Hang on, that's not right. If the value and only one of the rescaler
elements is negative, the result is positive. || is not the correct
logical operation.

>> 		*val = -*val;
> 
> Unconditionally negating *val doesn't negate the combined value when
> *val is zero and *val2 isn't. My test "if (*val < 0)" above attempting
> to take care of this case is clearly not right. It should of course be
> "if (*val > 0)" since *val is not yet negated. Duh!
> 
> In fact, I think a few tests scaling to/from the [-1,1] interval
> would be benefitial for this exact reason.

So, with both these issues taken care of:

 	if (neg ^ ((rescale->numerator < 0) ^ (rescale->denominator < 0))) {
		if (*val > 0)
			*val = -*val;
		else
			*val2 = -*val2;
	}

(bitwise ^ is safe since all operands come from logical operations, i.e.
they are either zero or one and nothing else)

Cheers,
Peter


  reply	other threads:[~2021-07-30  7:02 UTC|newest]

Thread overview: 33+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-07-21  3:06 [PATCH v6 00/13] iio: afe: add temperature rescaling support Liam Beguin
2021-07-21  3:06 ` [PATCH v6 01/13] iio: inkern: apply consumer scale on IIO_VAL_INT cases Liam Beguin
2021-07-21  3:06 ` [PATCH v6 02/13] iio: inkern: apply consumer scale when no channel scale is available Liam Beguin
2021-07-21  3:06 ` [PATCH v6 03/13] iio: inkern: make a best effort on offset calculation Liam Beguin
2021-07-21  3:06 ` [PATCH v6 04/13] iio: afe: rescale: expose scale processing function Liam Beguin
2021-07-21  3:06 ` [PATCH v6 05/13] iio: afe: rescale: add INT_PLUS_{MICRO,NANO} support Liam Beguin
2021-07-23 21:16   ` Peter Rosin
2021-07-28  0:21     ` Liam Beguin
2021-07-28  7:19       ` Peter Rosin
2021-07-29 15:56         ` Liam Beguin
2021-07-30  6:49           ` Peter Rosin
2021-07-30  7:01             ` Peter Rosin [this message]
2021-07-30 20:01               ` Liam Beguin
2021-07-30 19:57             ` Liam Beguin
2021-07-31 17:47         ` Jonathan Cameron
2021-07-21  3:06 ` [PATCH v6 06/13] iio: afe: rescale: add offset support Liam Beguin
2021-07-21  3:06 ` [PATCH v6 07/13] iio: test: add basic tests for the iio-rescale driver Liam Beguin
2021-07-24  8:40   ` kernel test robot
2021-07-21  3:06 ` [PATCH v6 08/13] iio: afe: rescale: reduce risk of integer overflow Liam Beguin
2021-07-23 21:17   ` Peter Rosin
2021-07-28  0:07     ` Liam Beguin
2021-07-28  7:47       ` Peter Rosin
2021-07-29 16:02         ` Liam Beguin
2021-07-21  3:06 ` [PATCH v6 09/13] iio: afe: rescale: fix precision on fractional log scale Liam Beguin
2021-07-23 21:20   ` Peter Rosin
2021-07-28  0:26     ` Liam Beguin
2021-07-28  7:58       ` Peter Rosin
2021-07-29 16:19         ` Liam Beguin
2021-07-21  3:06 ` [PATCH v6 10/13] iio: afe: rescale: add RTD temperature sensor support Liam Beguin
2021-07-21  3:06 ` [PATCH v6 11/13] iio: afe: rescale: add temperature transducers Liam Beguin
2021-07-21  3:06 ` [PATCH v6 12/13] dt-bindings: iio: afe: add bindings for temperature-sense-rtd Liam Beguin
2021-07-21  3:06 ` [PATCH v6 13/13] dt-bindings: iio: afe: add bindings for temperature transducers Liam Beguin
2021-07-23 22:59   ` Rob Herring

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=8448cecb-16b3-96ff-dfa8-3933325ef946@axentia.se \
    --to=peda@axentia.se \
    --cc=devicetree@vger.kernel.org \
    --cc=jic23@kernel.org \
    --cc=lars@metafoo.de \
    --cc=liambeguin@gmail.com \
    --cc=linux-iio@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=pmeerw@pmeerw.net \
    --cc=robh+dt@kernel.org \
    --subject='Re: [PATCH v6 05/13] iio: afe: rescale: add INT_PLUS_{MICRO,NANO} support' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).