From: Szabolcs Nagy <szabolcs.nagy@arm.com>
To: Sunil Pandey <skpgkp2@gmail.com>
Cc: Joseph Myers <joseph@codesourcery.com>,
andrey.kolesov@intel.com,
GNU C Library <libc-alpha@sourceware.org>
Subject: Re: [PATCH 5/6] x86-64: Add pow/powf to libmvec microbenchmark
Date: Thu, 11 Nov 2021 10:39:31 +0000 [thread overview]
Message-ID: <20211111103931.GY1982710@arm.com> (raw)
In-Reply-To: <CAMAf5_dmzTV-FpNut29d19Cjx+d9Jbh1qgZQhtuuv3UjMX+2jw@mail.gmail.com>
The 11/10/2021 19:31, Sunil Pandey via Libc-alpha wrote:
> On Wed, Nov 10, 2021 at 12:27 PM Joseph Myers <joseph@codesourcery.com>
> wrote:
>
> > On Tue, 9 Nov 2021, Sunil K Pandey via Libc-alpha wrote:
> >
> > > +# Random inputs in [a=0.01,b=100.00], [c=0.01,d=10.00]
> >
> > I think there should be a reasonable proportion of negative exponents,
> > unless you have evidence that negative exponents are rarely used in
> > practice. (The *first* argument of pow being negative is of use (produces
> > finite results) in much more limited cases, so may well be much rarer.)
> >
>
> Negative exponents not used in the input file to avoid denormal
> processing. Negative
> exponent
> may hit the underflow domain when pow(x, -y) produces denormals as results
> or as intermediate
> values. Both going to touch very slow special pow() path and invokes Intel
> software assist for
> denormals processing. All of this will distort the overall performance
> result of the main path.
> Also the base argument should be small so that it doesn’t produce denormals.
i'd expect typical negative exponents would not
cause subormal computations.
does powf(100, -10) have intermediate underflow
with the x86 libmvec code?
next prev parent reply other threads:[~2021-11-11 10:39 UTC|newest]
Thread overview: 53+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-11-10 3:07 [PATCH 0/6] Implement microbenchmark for libmvec Sunil K Pandey
2021-11-10 3:07 ` [PATCH 1/6] x86-64: Create microbenchmark infrastructure " Sunil K Pandey
2021-11-10 3:53 ` Noah Goldstein
2021-11-11 18:34 ` Sunil Pandey
2021-11-12 19:17 ` [PATCH v2 0/6] Implement microbenchmark " Sunil K Pandey
2021-11-12 19:17 ` [PATCH v2 1/6] x86-64: Create microbenchmark infrastructure " Sunil K Pandey
2021-11-12 21:02 ` Noah Goldstein
2021-11-12 22:49 ` Sunil Pandey
2021-11-13 19:47 ` H.J. Lu
2021-11-14 2:59 ` Sunil Pandey
2021-11-15 21:06 ` [PATCH v3 0/6] Implement microbenchmark " Sunil K Pandey
2021-11-15 21:06 ` [PATCH v3 1/6] x86-64: Create microbenchmark infrastructure " Sunil K Pandey
2021-11-16 17:21 ` H.J. Lu
2021-11-16 18:37 ` [PATCH] " Sunil K Pandey
2021-11-16 18:40 ` H.J. Lu
2021-11-15 21:06 ` [PATCH v3 2/6] x86-64: Add vector cos/cosf to libmvec microbenchmark Sunil K Pandey
2021-11-15 21:06 ` [PATCH v3 3/6] x86-64: Add vector exp/expf " Sunil K Pandey
2021-11-15 21:06 ` [PATCH v3 4/6] x86-64: Add vector log/logf " Sunil K Pandey
2021-11-15 21:06 ` [PATCH v3 5/6] x86-64: Add vector pow/powf " Sunil K Pandey
2021-11-15 21:06 ` [PATCH v3 6/6] x86-64: Add vector sin/sinf " Sunil K Pandey
2021-11-12 19:17 ` [PATCH v2 2/6] x86-64: Add vector cos/cosf " Sunil K Pandey
2021-11-12 19:17 ` [PATCH v2 3/6] x86-64: Add vector exp/expf " Sunil K Pandey
2021-11-12 19:17 ` [PATCH v2 4/6] x86-64: Add vector log/logf " Sunil K Pandey
2021-11-12 23:18 ` Joseph Myers
2021-11-13 1:37 ` Sunil Pandey
2021-11-13 1:44 ` Joseph Myers
2021-11-13 6:14 ` Sunil Pandey
2021-11-16 0:12 ` Joseph Myers
2021-11-23 17:40 ` [PATCH v4 0/5] Add vector math functions to microbenchmark Sunil K Pandey
2021-11-23 17:40 ` [PATCH v4 1/5] x86-64: Add vector cos/cosf to libmvec microbenchmark Sunil K Pandey
2021-11-24 12:22 ` H.J. Lu
2021-11-23 17:40 ` [PATCH v4 2/5] x86-64: Add vector exp/expf " Sunil K Pandey
2021-11-24 12:24 ` H.J. Lu
2021-11-23 17:40 ` [PATCH v4 3/5] x86-64: Add vector log/logf " Sunil K Pandey
2021-11-24 12:25 ` H.J. Lu
2021-11-23 17:40 ` [PATCH v4 4/5] x86-64: Add vector pow/powf " Sunil K Pandey
2021-11-24 12:26 ` H.J. Lu
2021-11-23 17:40 ` [PATCH v4 5/5] x86-64: Add vector sin/sinf " Sunil K Pandey
2021-11-24 12:30 ` H.J. Lu
2021-11-23 18:30 ` [PATCH v2 4/6] x86-64: Add vector log/logf " Sunil Pandey
2021-11-23 22:13 ` Joseph Myers
2021-11-12 19:17 ` [PATCH v2 5/6] x86-64: Add vector pow/powf " Sunil K Pandey
2021-11-12 19:18 ` [PATCH v2 6/6] x86-64: Add vector sin/sinf " Sunil K Pandey
2021-11-10 3:07 ` [PATCH 2/6] x86-64: Add cos/cosf " Sunil K Pandey
2021-11-10 3:07 ` [PATCH 3/6] x86-64: Add exp/expf " Sunil K Pandey
2021-11-10 3:07 ` [PATCH 4/6] x86-64: Add log/logf " Sunil K Pandey
2021-11-10 3:07 ` [PATCH 5/6] x86-64: Add pow/powf " Sunil K Pandey
2021-11-10 20:27 ` Joseph Myers
2021-11-11 3:31 ` Sunil Pandey
2021-11-11 10:39 ` Szabolcs Nagy [this message]
2021-11-11 17:40 ` Sunil Pandey
2021-11-11 18:51 ` Sunil Pandey
2021-11-10 3:07 ` [PATCH 6/6] x86-64: Add sin/sinf " Sunil K Pandey
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=20211111103931.GY1982710@arm.com \
--to=szabolcs.nagy@arm.com \
--cc=andrey.kolesov@intel.com \
--cc=joseph@codesourcery.com \
--cc=libc-alpha@sourceware.org \
--cc=skpgkp2@gmail.com \
/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 read-only IMAP folder(s) and NNTP newsgroup(s).