You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@harmony.apache.org by Maksim Ananjev <ma...@gmail.com> on 2007/08/28 08:28:23 UTC

[drvm][jit] Operator Strength Reduction

Hi!

JIT optimizer now lacks "operator strength reduction" pass. It is
well-known optimization that reformulates costly operations (such as
multiplications) in terms of less expensive ones (such as additions)

I'm going to implement it according to this paper: Cooper, Simpson,
Vick "Operator Strength Reduction" (2001)

I hope it will improve Harmony in performance domain: especially for
numerically intensive applications.

-- 
Maksim

Re: [drvm][jit] Operator Strength Reduction

Posted by Maksim Ananjev <ma...@gmail.com>.
Tim,

Thanks for your detailed reply.
I'll complete these procedures as soon as possible.


2007/8/29, Tim Ellison <t....@gmail.com>:
> Sorry for the delay in responding Maksim.
>
> Maksim Ananjev wrote:
> > Tim,
> >
> > Thanks. I'm really not an authorized contributor yet. What is the way
> > of sending you questionnaire and ICLA? Should I print them, fill by
> > hand and send by post office? Or scan them and send by e-mail? Or
> > there is a way to submit them over the internet?  I have search over
> > the ASF site, but found no hints on this point.
>
> The questionnaire [1] is maintained by the Harmony project.  Please
> complete it and send a copy to private@harmony.apache.org.
>
> The ICLA [2] is maintained by the ASF, so please complete that and send
> it directly to the ASF secretary (as detailed on the form itself).
>
> In both cases, we are just looking for statements that show you
> understand that the work you are contributing can form part of an Apache
> licensed distribution.
>
> [1] http://harmony.apache.org/auth_cont_quest.html
> [2] http://www.apache.org/licenses/icla.txt
>
> Regards,
> Tim
>


-- 
Maksim

Re: [drvm][jit] Operator Strength Reduction

Posted by Tim Ellison <t....@gmail.com>.
Sorry for the delay in responding Maksim.

Maksim Ananjev wrote:
> Tim,
> 
> Thanks. I'm really not an authorized contributor yet. What is the way
> of sending you questionnaire and ICLA? Should I print them, fill by
> hand and send by post office? Or scan them and send by e-mail? Or
> there is a way to submit them over the internet?  I have search over
> the ASF site, but found no hints on this point.

The questionnaire [1] is maintained by the Harmony project.  Please
complete it and send a copy to private@harmony.apache.org.

The ICLA [2] is maintained by the ASF, so please complete that and send
it directly to the ASF secretary (as detailed on the form itself).

In both cases, we are just looking for statements that show you
understand that the work you are contributing can form part of an Apache
licensed distribution.

[1] http://harmony.apache.org/auth_cont_quest.html
[2] http://www.apache.org/licenses/icla.txt

Regards,
Tim

Re: [drvm][jit] Operator Strength Reduction

Posted by Maksim Ananjev <ma...@gmail.com>.
Tim,

Thanks. I'm really not an authorized contributor yet. What is the way
of sending you questionnaire and ICLA? Should I print them, fill by
hand and send by post office? Or scan them and send by e-mail? Or
there is a way to submit them over the internet?  I have search over
the ASF site, but found no hints on this point.


2007/8/28, Tim Ellison <t....@gmail.com>:
> Maksim Ananjev wrote:
> > Hi!
> >
> > JIT optimizer now lacks "operator strength reduction" pass. It is
> > well-known optimization that reformulates costly operations (such as
> > multiplications) in terms of less expensive ones (such as additions)
> >
> > I'm going to implement it according to this paper: Cooper, Simpson,
> > Vick "Operator Strength Reduction" (2001)
> >
> > I hope it will improve Harmony in performance domain: especially for
> > numerically intensive applications.
>
> Great!  Please keep us up to date on your progress, and send in patches
> via JIRA.  At first glance, I did not see you name on the list of
> questionnaire [1] responses we have got.  If you have not already done
> so please complete one and send it to us.
>
> [1] http://harmony.apache.org/auth_cont_quest.html
>
> Thanks,
> Tim
>


-- 
Maksim

Re: [drvm][jit] Operator Strength Reduction

Posted by Tim Ellison <t....@gmail.com>.
Maksim Ananjev wrote:
> Hi!
> 
> JIT optimizer now lacks "operator strength reduction" pass. It is
> well-known optimization that reformulates costly operations (such as
> multiplications) in terms of less expensive ones (such as additions)
> 
> I'm going to implement it according to this paper: Cooper, Simpson,
> Vick "Operator Strength Reduction" (2001)
> 
> I hope it will improve Harmony in performance domain: especially for
> numerically intensive applications.

Great!  Please keep us up to date on your progress, and send in patches
via JIRA.  At first glance, I did not see you name on the list of
questionnaire [1] responses we have got.  If you have not already done
so please complete one and send it to us.

[1] http://harmony.apache.org/auth_cont_quest.html

Thanks,
Tim