This is the mail archive of the libstdc++@gcc.gnu.org mailing list for the libstdc++ project.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]
Other format:	[Raw text]

Re: [PATCH, libstdc++, complex] complex multiplication algorithm improved

From: Marc Glisse <marc dot glisse at inria dot fr>
To: Paolo Carlini <paolo dot carlini at oracle dot com>
Cc: Wang Feng <wanng dot fenng at gmail dot com>, libstdc++ at gcc dot gnu dot org, Gabriel Dos Reis <gdr at integrable-solutions dot net>
Date: Wed, 15 Sep 2010 11:21:39 +0200 (CEST)
Subject: Re: [PATCH, libstdc++, complex] complex multiplication algorithm improved
References: <AANLkTikfU1+E9K8B8-6XxvxsCLdu01TcznOTURKaE0vF@mail.gmail.com> <AANLkTikN2zvMaUJL0KPfdfi8ab5iJbtBs5iDeRybnFPD@mail.gmail.com> <4C90866F.5070406@oracle.com> <4C908BCC.9010700@oracle.com>
Reply-to: libstdc++ at gcc dot gnu dot org

On Wed, 15 Sep 2010, Paolo Carlini wrote:

... note that we are talking about the code *in the primary template*,
not in the specializations for float / double / long double and that the
behavior when instantiating with anything else outside float / double /
long double is *unspecified* in the Standard.

For float/double/long double, the code with 3 multiplications is actually significantly slower on many architectures (including x86 and x86_64) because of all the extra additions. So it only makes sense if you expect complex<T> to be used mostly with T that has costly multiplications (double on some archs, multi-precision, etc).

--
Marc Glisse

References:
- [PATCH, libstdc++, complex] complex multiplication algorithm improved
  - From: Wang Feng
- Re: [PATCH, libstdc++, complex] complex multiplication algorithm improved
  - From: Paolo Carlini
- Re: [PATCH, libstdc++, complex] complex multiplication algorithm improved
  - From: Paolo Carlini

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]