RFR(S): 8223249: [s390] Cleanup TemplateInterpreterGenerator::generate_fixed_frame
Doerr, Martin
martin.doerr at sap.com
Thu May 2 16:41:05 UTC 2019
Hi Coleen,
thanks for reminding me of JDK-8168795.
I agree. It makes sense to implement it like on PPC.
New webrev:
http://cr.openjdk.java.net/~mdoerr/8223249_s390_generate_fixed_frame/webrev.01/
I think it makes sense to improve other platforms, too (even though I wouldn’t expect a measurable performance improvement).
I don't really like the lengthy load chain.
Best regards,
Martin
-----Original Message-----
From: hotspot-runtime-dev <hotspot-runtime-dev-bounces at openjdk.java.net> On Behalf Of coleen.phillimore at oracle.com
Sent: Donnerstag, 2. Mai 2019 17:49
To: hotspot-runtime-dev at openjdk.java.net
Subject: Re: RFR(S): 8223249: [s390] Cleanup TemplateInterpreterGenerator::generate_fixed_frame
This looks familiar. ppc added load_mirror_from_const_method(). I
think it makes sense to add that to s390 also, and eventually we should
add it and use it on the other platforms.
What do you think?
thanks,
Coleen
On 5/2/19 11:40 AM, Doerr, Martin wrote:
> Hi,
>
> I'd like to simplify and improve the s390 implementation of generate_fixed_frame.
> The current implementation of generate_fixed_frame unnecessarily reloads fields from memory. Most prominent is the reload of ConstMethod* (5 times).
> In addition, unneeded monitor space is allocated in the top frame. Space for monitors gets added on demand (lock_method uses add_monitor_to_stack).
> Also, implementation can get simplified by using MacroAssembler::Clear_Array.
>
> Please review:
> http://cr.openjdk.java.net/~mdoerr/8223249_s390_generate_fixed_frame/webrev.00/
>
> Best regards,
> Martin
>
More information about the hotspot-runtime-dev
mailing list