hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Colin P. McCabe" <cmcc...@apache.org>
Subject Re: [crypto][chimera] Next steps
Date Mon, 22 Feb 2016 23:14:28 GMT
Hi Jochen,

Many CPUs come with built-in support for certain cryptographic and/or
hash/checksum-related primitives.  For example, modern x86 CPUs have
CRC32C implemented in hardware.  Currently, this must be accessed via
inline assembly expressed in JNI.  It is worth it... at least in the
case of checksumming, you often see 5x or 10x reductions in the amount
of CPU used.  The gains for moving from pure Java to using the openSSL
AES functions are similar.  Perhaps someday Java will gain native
support for these features.  Until that point, though, JNI will be
necessary to get reasonable performance on modern hardware.

best,
Colin

On Mon, Feb 22, 2016 at 10:02 AM, Jochen Wiedmann
<jochen.wiedmann@gmail.com> wrote:
> On Mon, Feb 22, 2016 at 6:28 PM, Colin P. McCabe <cmccabe@apache.org> wrote:
>
>> What is the strategy for handling JNI components?
>
> Wrong question, IMO. Should better be: What are the reasons for using
> JNI components? Couldn't they be replaced? If so, that would very much
> enhance the long term prospects of crypto|chimera|whatever.
>
> Jochen
>
> --
> The next time you hear: "Don't reinvent the wheel!"
>
> http://www.keystonedevelopment.co.uk/wp-content/uploads/2014/10/evolution-of-the-wheel-300x85.jpg

Mime
View raw message