[aarch64-port-dev ] ZeroTLAB and block zeroing

[Jasty, Ananth]

Iirc the correctness was due to not firing enough barriers after the zva. Throwing a dsb sy after the zeroes should be conservative until we've found the proper ordering (1 dmb st should really cover everything but we were young and naive back then).

________________________________
From: aarch64-port-dev <aarch64-port-dev-bounces at openjdk.java.net> on behalf of White, Derek <Derek.White at cavium.com>
Sent: Wednesday, October 25, 2017 3:35:43 PM
To: Andrew Haley; Jones, Joel
Cc: aarch64-port-dev at openjdk.java.net
Subject: Re: [aarch64-port-dev ] ZeroTLAB and block zeroing

Hi Andrew,

We also talked about this a year ago. Maybe it's time we fixed it :)

Reported as OpenJDK issue: https://bugs.openjdk.java.net/browse/JDK-8190169

Assigned it to myself.

 - Derek

> -----Original Message-----
> From: aarch64-port-dev [mailto:aarch64-port-dev-
> bounces at openjdk.java.net] On Behalf Of Andrew Haley
> Sent: Tuesday, October 24, 2017 10:24 AM
> To: Jones, Joel <Joel.Jones at cavium.com>
> Cc: aarch64-port-dev at openjdk.java.net
> Subject: Re: [aarch64-port-dev ] ZeroTLAB and block zeroing
>
> On 24/10/17 15:08, Jones, Joel wrote:
> > Certainly worth a shot. We tried this about two years ago, but it caused
> correctness issues if I recall correctly.
>
> zero_words() has been rewritten since then.  There were some nasty
> problems with register usage, but it's all fixable.
>
> --
> Andrew Haley
> Java Platform Lead Engineer
> Red Hat UK Ltd. <https://www.redhat.com>
> EAC8 43EB D3EF DB98 CC77 2FAD A5CD 6035 332F A671