harmony-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Alex Astapchuk (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HARMONY-3077) add the Encoder Library description to the site
Date Tue, 30 Jan 2007 10:21:33 GMT

    [ https://issues.apache.org/jira/browse/HARMONY-3077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12468558

Alex Astapchuk commented on HARMONY-3077:

Hi Nadya,

Thanks for your comments! 
As long as I have some relations to the Encoder and its docs, I'll try to 
address some of the questions (though some of them are too special for me, esp. the language-related

> - Intel64 - do we really only support Intel64? what about amd64?
They are binary compatible, so the Intel64 also implies amd64. 

> - section 4: "The most popular instructions are in place already" - you mean not all
of these 
> 150 functions are implemented? if so, please make this key statement more apparent; otherwise,
> seems the whole 150 are in place and ready for use.

Yes, 150+ instructions are ready. 
The modern IA-32 and Intel64 instruction set contains much more instructions. 
We start filling from the most popular ones, and now we have these 150.

> - section 5: "common generic interface" isn't this tautology?
Hmmm.. I was thinking about something like "common" as "the same for everyone" as it exposes
a single method. And about "generic" as "wide enough to cover most of its need into a single

> - section 5.1: does the section mention all the adapters currently available? if so,
please make them 
Yes. Currently, we have 3 of them. Two of them are very special, and strictly targeted to
clients (these are adapter in Jitrino.JET and adapter in Jitrino.OPT).
The latter one is what most of programmers need, and its declarations are in encoder.h.

This adapter is often informally referred ad 'encoder', for example 'a stub was written using
encoder'. :-)

> a list and not examples + some text. this way, a reader will have the chance to easily
find the adapter 
> they need to read about most of all.

> - section 5.2.2: i don't quite understand why it's necessary to explain how the encode
*does not* work 
> ("This algorithm helps to avoid the following loop on the hot execution path, as shown
> perhaps, we could abridge the section to only show how the encoder works (not to confuse

In fact, this IS how it works. :-)
The "hot execution path" normally refers to a program section that spends a lot of time.
This is the subject for various optimizations. Some of them may contain
	- reducing number of instructions in the path
	- reducing number of memory accesses
	- (sic!)removing loops from the hot path 
	- and there are more...
Often, the loops are the root of the evil, that's why this section exactly states why 
we do such things - we optimize (and speed up) execution by *removing* the loop.

> add the Encoder Library description to the site
> -----------------------------------------------
>                 Key: HARMONY-3077
>                 URL: https://issues.apache.org/jira/browse/HARMONY-3077
>             Project: Harmony
>          Issue Type: Task
>          Components: Website/Documentation
>            Reporter: Svetlana Konovalova
>         Assigned To: Nadya Morozova
>         Attachments: encoder_library.html, encoder_library.xml
> add the Encoder Library document to the site [to the subcomponents/drlvm directory ]

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message