Request for reviews (XS): 6879902: CTW failure jdk6_18/hotspot/src/cpu/sparc/vm/assembler_sparc.hpp:845 (original) (raw)

Vladimir Kozlov Vladimir.Kozlov at Sun.COM
Mon Oct 5 12:47:43 PDT 2009

Good.

Vladimir

Christian Thalinger wrote:

On Fri, 2009-10-02 at 12:32 +0200, Christian Thalinger wrote:

Just for the record... I have already prepared another webrev but did had time yet to post it:

http://cr.openjdk.java.net/~twisti/6879902/webrev.02/ and Tom had a look at it: I looked at this again and I don't think that stf should be using argslot. c2i only uses argslot/nextargslot for loads which is correct since we should be moving it from interpreter layout to compiled layout. Using it for stores is wrong. After Vladimir's comment I thought that other compiled code was already using argslot for stores so it wasn't any worse but now I can see that it's not true. I think the original fix of just correcting the stf code was the way to go. It can use RegisterOrConstant with ensurers2. We can keep the other changes or not depending on how you feel about them. They generate better code for 64-bit and deals with the simm13 issue in general which I like. If you keep them I'd suggest using ensurers2 in you argslot routines instead of hand coding that. I wouldn't mind renaming ensurers2 since it's such a useless name. Here is an updated webrev: http://cr.openjdk.java.net/~twisti/6879902/webrev.03/ -- Christian

More information about the hotspot-compiler-dev mailing list