[8u] RFR for backport of JDK-8188868: PPC64: Support AES intrinsics on Big Endian

Wed Nov 21 16:19:29 UTC 2018

Hi Martin and Ogata,

Sorry for not replying earlier.

I'm just back from holidays and catching up with the threads.

On 11/21/2018 07:53 AM, Kazunori Ogata wrote:
> Hi Martin,
> 
> I noticed that the changeset JDK-8188868 [1] also needs load_perm() and
> vec_prem(), which are added in another changeset "8185979: PPC64:
> Implement SHA2 intrinsic" [2].  I'm sorry, but I forgot about these two
> functions when I posted my RFR.
> 
> The issue is that applying [2] causes a long chain of dependency, as shown
> below in depending order, and it eventually reached two changesets [5][6]
> that seem to be difficult to backport to JDK8.  [5] is for the compact
> string feature, which is not supported in JDK8, and [6] is a big bundle of
> changes.
> 
> Is it acceptable as a back port process to cherry pick a few methods from
> a dependent changeset, instead of applying whole changeset?
> 
> Possible candidates of cherry picking are:
> 
> 1. Pick load_perm() and vec_perm() from [2]
> 2. Pick vst() and vsix() from [3], and apply [2]
> 3. Pick has_vsx() and vsr[abst]() from [4], and apply [3] and above
> 4. Pick has_mfdscr() from [5] and config_dscr() from [6], and apply [4]
> and above.
> 
> I appreciate your suggestion.

Martin, I know Ogata already said that he will pick up also the changes related
to the SHA2 acceleration following your suggestion, but I'm still curious if it
would be acceptable/reasonable  (assuming one would like to focus only on the
BE AES [1] backport) to cherry pick only load_perm() and vec_perm() from [3] in
order to backport "PPC64: Support AES intrinsics on Big Endian" [1], since both
load_perm() and vec_perm() look rather general and simple (to maintain, handle
in case of a merge conflict with a future backport of [3], etc).

In that case BE AES [1] backport would be simpler since JDK-8185975 [2] (needed
by [1] as you pointed out) applies cleanly except for some offsets. vec_perm()
is also overloaded in [1], so a slight variation of vec_perm() from [2] will be
added anyway by [1].

Thank you.

Best regards,
Gustavo

[1] http://hg.openjdk.java.net/jdk/jdk/rev/9d337e48b178 (8188868: PPC64: Support AES intrinsics on Big Endian)
[2] http://hg.openjdk.java.net/jdk/jdk/rev/1bf8c1e8b79a (8185975: PPC64: Fix vsldoi interface according to the ISA)
[3] http://hg.openjdk.java.net/jdk/jdk/rev/f4962ab855b6 (8185979: PPC64: Implement SHA2 intrinsic)

> 
> [1] 8188868: PPC64: Support AES intrinsics on Big Endian
>      http://hg.openjdk.java.net/jdk/jdk/rev/9d337e48b178
> [2] 8185979: PPC64: Implement SHA2 intrinsic
>      http://hg.openjdk.java.net/jdk/jdk/rev/f4962ab855b6
> [3] 8185969: PPC64: Improve VSR support to use up to 64 registers
>      http://hg.openjdk.java.net/jdk/jdk/rev/057f21a10f5f
> [4] 8154156: PPC64: improve array copy stubs by using vector instructions
>      http://hg.openjdk.java.net/jdk/jdk/rev/c9d756fa846e
> [5] 8149655: PPC64: Implement CompactString intrinsics
>      http://hg.openjdk.java.net/jdk/jdk/rev/6241574f5982
> [6] 8077838: Recent developments for ppc.
>      http://hg.openjdk.java.net/jdk/jdk/rev/c703c89fddbf
> 
> 
> Regards,
> Ogata
> 
> 
> "hotspot-compiler-dev" <hotspot-compiler-dev-bounces at openjdk.java.net>
> wrote on 2018/11/19 18:12:45:
> 
>> From: "Kazunori Ogata" <OGATAK at jp.ibm.com>
>> To: "Doerr, Martin" <martin.doerr at sap.com>
>> Cc: "hotspot-compiler-dev at openjdk.java.net" <hotspot-compiler-
>> dev at openjdk.java.net>, "ppc-aix-port-dev at openjdk.java.net"
> <ppc-aix-port-
>> dev at openjdk.java.net>
>> Date: 2018/11/19 18:16
>> Subject: RE: [8u] RFR for backport of JDK-8188868: PPC64: Support AES
>> intrinsics on Big Endian
>> Sent by: "hotspot-compiler-dev"
> <hotspot-compiler-dev-bounces at openjdk.java.net>
>>
>> Hi Martin,
>>
>> Thank you for your comment.  I see.  I'll backport JDK-8185975 first.
>>
>>> Please don't forget to test the debug build.
>> Yes, I'll test both debug and release builds.
>>
>>
>> Regards,
>> Ogata
>>
>>
>>
>> From:   "Doerr, Martin" <martin.doerr at sap.com>
>> To:     Kazunori Ogata <OGATAK at jp.ibm.com>,
>> "hotspot-compiler-dev at openjdk.java.net"
>> <hotspot-compiler-dev at openjdk.java.net>,
>> "ppc-aix-port-dev at openjdk.java.net" <ppc-aix-port-dev at openjdk.java.net>
>> Date:   2018/11/19 17:56
>> Subject:        RE: [8u] RFR for backport of JDK-8188868: PPC64: Support
> 
>> AES intrinsics on Big Endian
>>
>>
>>
>> Hi Ogata,
>>
>> you need to backport JDK-8185975 before you can use +8 instead of -8.
>> Please don't forget to test the debug build.
>>
>> Best regards,
>> Martin
>>
>>
>> -----Original Message-----
>> From: hotspot-compiler-dev
> <hotspot-compiler-dev-bounces at openjdk.java.net>
>> On Behalf Of Kazunori Ogata
>> Sent: Montag, 19. November 2018 09:28
>> To: hotspot-compiler-dev at openjdk.java.net;
>> ppc-aix-port-dev at openjdk.java.net
>> Subject: [8u] RFR for backport of JDK-8188868: PPC64: Support AES
>> intrinsics on Big Endian
>>
>> Hi,
>>
>> May I get reviews for enhancement backport of JDK-8188868: PPC64:
> Support
>> AES intrinsics on Big Endian?
>>
>> INVALID URI REMOVED
>> u=http-3A__cr.openjdk.java.net_-7Ehorii_jdk8u-5Faes-5Fbe_8188868_webrev.
>> 00_&d=DwIFAg&c=jf_iaSHvJObTbx-siA1ZOg&r=p-
>>
> FJcrbNvnCOLkbIdmQ2tigCrcpdU77tlI2EIdaEcJw&m=BqxW_dQ87Y0TOxyOjkH6hFZIIrhwH2ycEWJdxSDcAeQ&s=ulzXDyQq40EaguNM-
>> svdR9rWGZnClO3c3d_41XGFuHg&e=
>>
>>
>> There is no code change except for an immediate parameter of vsldoi (-8
> in
>>
>> jdk8u was changed to 8 to match with the latest code), besides the
>> difference of directory tree structure.
>>
>> Since not a small number of customers are still using jdk8, especially
> in
>> AIX, JDK-8188868 is important for them to improve performance of secure
>> network transportation.
>>
>>
>> Regards,
>> Ogata
>>
>>
>>
>>
>>
>>
> 
> 