[8u] RFR 8073108: Use x86 and SPARC CPU instructions for GHASH acceleration

[Yuri Gaevsky]

Hi Martin,

> I didn't know you have already proposed a backport for this bug. I
> assumed nobody did because the ticket was not flagged with
> "jdk8u-fix-request". Sorry.

There is no need to apologise, it's great to finally see the review for JDK-8073108
is progressing. :-)

IIUC, the marking of the appropriate JBS issue with 'jdk8u-fix-request' is the *next step*
in the backporting process. 

In summary, my reading  of  the related OpenJDK documents (see below) is that:
  (a) the formal review should be done *before* setting 'jdk8u-fix-request' label in JBS;
  (b) 'review request for backport' and 'push approval for backport' are different things.

[1] https://wiki.openjdk.java.net/display/jdk8u/Main
  --- cut ---
    If the backport requires more than just cosmetic changes (file location changes, copyright
  header updates) to apply to the 8u tree, *it should first be submitted for review*.
  Push approval for a fix is *then requested by setting the jdk8u-fix-request label* on
  the original JBS bug.
  --- cut ---

[2] https://wiki.openjdk.java.net/display/jdk8u/Guidelines+for+working+on+jdk8u
  --- cut ---
  In general we'll use the process in https://openjdk.java.net/projects/jdk-updates/approval.html.
  ...
  To request maintainer *approval* for a backport, tag your entry in the bug
  database with "jdk8u-fix-request".
  --- cut ---

[3] https://openjdk.java.net/projects/jdk-updates/approval.html
  --- cut ---
  JDK Updates: Requesting *push* approval for fixes
  Rule 0
  ...
  An OpenJDK JDK Updates Author, Committer or Reviewer requesting their fix in a JDK Updates
  repo MUST label the corresponding master/parent issue with jdk<release>u-fix-request
  (where <release> denotes the feature release version, e.g. jdk10u-fix-request) in the JDK Bug System.
  --- cut ---

[4] https://wiki.openjdk.java.net/display/JDKUpdates/How+to+contribute+a+fix
  --- cut ---
  If (and only if) the original patch was modified, *get the change reviewed*
    Note: *the change review is not the approval, which you would get at the next step*
  --- cut ---


> I've not found significant differences between your backport and mine.
> The only minor comment I have is that you missed the word "left" in "//
> Shift left 128 bit value in xmm register by number of bytes." line
> (src/cpu/sparc/vm/vm_version_sparc.cpp).

Great, thank you for re-checking.

You are right, during addition of 'pslldq' instruction into 'assembler_x86.cpp' I haven't
included it just because the corresponding 'psrldq' above in the sources doesn't have it.
I should have explicitly stated that difference in my request.

> What's the difference between [1] and [2]?

There is no difference, the last one is just a "resending" copy.

Best regards,
-Yuri

-----Original Message-----
From: Martin Balao [mailto:mbalao at redhat.com] 
Sent: Thursday, November 21, 2019 07:11 PM
To: Yuri Gaevsky <ygaevsky at azul.com>; jdk8u-dev at openjdk.java.net
Subject: Re: [8u] RFR 8073108: Use x86 and SPARC CPU instructions for GHASH acceleration

Hi Yuri,

I didn't know you have already proposed a backport for this bug. I
assumed nobody did because the ticket was not flagged with
"jdk8u-fix-request". Sorry.

I've not found significant differences between your backport and mine.
The only minor comment I have is that you missed the word "left" in "//
Shift left 128 bit value in xmm register by number of bytes." line
(src/cpu/sparc/vm/vm_version_sparc.cpp).

On 11/21/19 7:38 AM, Yuri Gaevsky wrote:
> Could you please clarify how is your patch different from the similar patch for JDK-8073108 [1] (or [2])?
> 

What's the difference between [1] and [2]?

> NB:  another follow-up patch [3] is required to make that acceleration fully functional.
> 

This one was on my list too. I'll not abstain from duplicating the work.
Thanks for letting me know.

Regards,
Martin.-

> Thanks,
> -Yuri
> 
> [1] http://mail.openjdk.java.net/pipermail/jdk8u-dev/2019-September/010493.html
> [2] http://mail.openjdk.java.net/pipermail/jdk8u-dev/2019-October/010409.html
> [3] http://mail.openjdk.java.net/pipermail/jdk8u-dev/2019-October/010437.html
>