Jaotc input standard for compile commands
jean-philippe.halimi at intel.com
Fri Nov 16 19:59:06 UTC 2018
Thanks for chiming in despite the wrong mailinglist.
I'll look into this. I think it can bring significant value to the AOT compiler and should not require a lot of code.
From: Vladimir Kozlov [mailto:vladimir.kozlov at oracle.com]
Sent: Friday, November 16, 2018 8:25 AM
To: Doug Simon <doug.simon at oracle.com>; Halimi, Jean-Philippe <jean-philippe.halimi at intel.com>
Cc: hotspot compiler <hotspot-compiler-dev at openjdk.java.net>; graal-dev at openjdk.java.net
Subject: Re: Jaotc input standard for compile commands
We have RFE exactly for that:
If you are interesting you can take it :)
Currently for us AOT is lower priority than libgraal.
On 11/16/18 12:31 AM, Doug Simon wrote:
> Hi Jp,
> That's a question for the HotSpot compiler team but I suspect the answer is no.
>> On 16 Nov 2018, at 01:50, Halimi, Jean-Philippe <jean-philippe.halimi at intel.com> wrote:
>> Dear all,
>> I am currently looking at jaotc and its input capabilities. It looks like it can take a list of methods to compile from a list of bytecode inputs. I have not found a tool that allows to automatically detect hot methods at runtime. Namely, it looks like CDS can dump a log of touched methods, but it requires some extra conversion to be read by jaotc. Also, it may be nicer to AOT-compile only hot methods, not necessarily ones that have been touched once.
>> I was wondering if there was an ongoing effort on that?
More information about the hotspot-compiler-dev