[PATCH] GPU Exploitation Infrastructure (original) (raw)
Ben Walsh ben_walsh at uk.ibm.com
Wed Jan 31 13:16:21 UTC 2018
- Previous message: [PATCH] GPU Exploitation Infrastructure
- Next message: Possible addition of a default 'getOne' method on Iterable?
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Thank you Alan and David.
In light of your responses, I'll take this to the Valhalla project.
Thanks, Ben
From: David Holmes <david.holmes at oracle.com> To: Alan Bateman <Alan.Bateman at oracle.com>, Ben Walsh <ben_walsh at uk.ibm.com>, core-libs-dev at openjdk.java.net Date: 31/01/2018 01:33 Subject: Re: [PATCH] GPU Exploitation Infrastructure
Project Sumatra was looking at GPU use:
It's inactive though.
David
On 31/01/2018 7:46 AM, David Holmes wrote:
On 31/01/2018 1:35 AM, Alan Bateman wrote:
On 30/01/2018 13:55, Ben Walsh wrote:
This patch provides the infrastructure to enable the exploitation of a GPU by the compiler to accelerate certain suitable Java constructs.
When enabled, a suitable compiler can attempt to accelerate the following Java constructs by launching the corresponding lambda expression on the GPU: IntStream.range().parallel().forEach() IntStream.rangeClosed().parallel().forEach() where: defines upper and lower bounds is a correctly defined lambda expression As it stands, with the HotSpot compiler, this patch performs a "no op" in the newly added in-built native library method. This can be extended so that the HotSpot compiler attempts the acceleration detailed above instead. I would like to pair with a sponsor to contribute this patch ... The function prototypes in jvm.h are JVM* functions rather than JNI native method names. You may want to look at adding a JNI function to libjava with a wrapper or else use RegisterNatives to map the native method name to a JVM* function. BTW: Is this something for one the project repos (Valhalla or Panama) rather the main line (jdk/jdk)? This should be looked at in the context of Panama I think, though IIRC there was a distinct GPU related project as well? Either way GPU support
in mainline doesn't seem like something to just be dropped in like this without much more higher-level discussion taking place.
Plus hotspot changes - even stubs - must be reviewed on the appropriate hotspot mailing list. David
-Alan
Unless stated otherwise above: IBM United Kingdom Limited - Registered in England and Wales with number 741598. Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU
- Previous message: [PATCH] GPU Exploitation Infrastructure
- Next message: Possible addition of a default 'getOne' method on Iterable?
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]