This is the mail archive of the
newlib@sourceware.org
mailing list for the newlib project.
Re: Add memcpy support for SPARC
- From: Sebastian Huber <sebastian dot huber at embedded-brains dot de>
- To: Aditya Upadhyay <aadit0402 at gmail dot com>
- Cc: newlib <newlib at sourceware dot org>, Salil Sirotia <salil dot sirotia at gmail dot com>
- Date: Mon, 4 Jun 2018 17:33:48 +0200 (CEST)
- Subject: Re: Add memcpy support for SPARC
- References: <CAMZxAxebN+LExRGG2m+bTBDncyC_-6CNBPH6ESD+ykmViGGqJw@mail.gmail.com>
Hello,
----- Am 24. Mai 2018 um 21:09 schrieb Aditya Upadhyay aadit0402@gmail.com:
> Hi Developers,
>
> Is this ok to add memcpy implementation from Open Solaries?
>
> https://bitbucket.org/a3217055/illumos-gate-import/src/741592a55c4c9746705ce0fd9232f5acded43ed0/usr/src/lib/libc/sparc/gen/memcpy.s?at=default&fileviewer=file-view-default
>
> If yes, then Could you please tell us rules about modifying the
> Licance of third party code?
independent of the license issue. Is this memcpy() implementation actually faster? Which benchmark is used to determine faster? If it is faster, why is it faster? Could the generic code be re-structured to make the GCC generated code faster on this target? Is there a general GCC issue on this target that leads to a slow memcpy()?