From: Eric Blake <eblake@redhat.com>
To: Markus Armbruster <armbru@redhat.com>
Cc: Alexey Kardashevskiy <aik@ozlabs.ru>,
Peter Maydell <peter.maydell@linaro.org>,
qemu-devel@nongnu.org
Subject: Re: [Qemu-devel] [PATCH v2] utils: Add pow2ceil()
Date: Mon, 23 Feb 2015 14:20:18 -0700 [thread overview]
Message-ID: <54EB9992.20601@redhat.com> (raw)
In-Reply-To: <87sidwo7a7.fsf@blackfin.pond.sub.org>
[-- Attachment #1: Type: text/plain, Size: 1942 bytes --]
On 02/23/2015 10:40 AM, Markus Armbruster wrote:
>>> int64_t pow2floor(int64_t value)
>>> {
>>> assert(value > 0);
>>> return 0x8000000000000000u >> clz64(value);
>>> }
>>
>> Needs to be 0x8000000000000000ull for 32-bit machines to compile correctly.
>
> Why?
Because 0x8000000000000000u is only required to be a 'long', and on
32-bit machines, your constant would overflow long. See, for example,
commit 5cb6be2ca. You NEED the 'll' suffix to ensure that the compiler
doesn't reject the constant as an overflow.
>
>> Why is the parameter int64_t? Wouldn't it be more useful to have:
>>
>> uint64_t pow2floor(uint64_t value)
>
> Crossed my mind, too. However, the existing callers pass *signed*
> arguments.
I guess it means auditing callers either way.
>
>>> int64_t pow2ceil(int64_t value)
>>> {
>>
>> Again, why allow signed inputs?
>>
>>> assert(value <= 0x4000000000000000)
>>> if (value <= 1)
>>> return 1;
>>
>> In particular, this slams all negative values to a result of 1, which
>> doesn't necessarily make sense.
>
> It implements a straightforward contract: return the smallest power of
> two greater or equal to the argument. The function's domain is the set
> of int64_t arguments where this value can be represented in int64_t,
> i.e. [-2^63..2^62].
>
> Feel free to suggest a more sensible contract.
But why should I claim that the nearest power of 2 greater than -3 is
positive 1, when I could argue that it should instead be -2 (nearest
positive or negative power of 2 rounding towards +inf) or -4 (nearest
positive or negative power of 2 rounding away from 0)? Since there are
multiple potential contracts once negative values are involved, I find
it easier to just make the contract require a positive input to begin with.
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 604 bytes --]
next prev parent reply other threads:[~2015-02-23 21:20 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-02-23 12:23 [Qemu-devel] [PATCH v2] utils: Add pow2ceil() Alexey Kardashevskiy
2015-02-23 13:59 ` Markus Armbruster
2015-02-23 16:17 ` Eric Blake
2015-02-23 17:40 ` Markus Armbruster
2015-02-23 21:20 ` Eric Blake [this message]
2015-02-24 9:39 ` Markus Armbruster
2015-02-24 11:39 ` Peter Maydell
2015-02-24 13:09 ` Markus Armbruster
2015-02-25 0:40 ` Alexey Kardashevskiy
2015-02-25 10:45 ` Markus Armbruster
2015-03-12 15:29 ` Richard Henderson
2015-03-12 16:45 ` Eric Blake
2015-03-13 19:04 ` Richard Henderson
2015-03-13 7:33 ` Markus Armbruster
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=54EB9992.20601@redhat.com \
--to=eblake@redhat.com \
--cc=aik@ozlabs.ru \
--cc=armbru@redhat.com \
--cc=peter.maydell@linaro.org \
--cc=qemu-devel@nongnu.org \
/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
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).