Hi Henry, > De: "Henri Tremblay" <henri.tremb...@gmail.com> > À: "mechanical-sympathy" <mechanical-sympathy@googlegroups.com> > Envoyé: Mercredi 27 Décembre 2017 19:47:56 > Objet: Re: Minimal Value Type First, there are actually two specs for value types, the minimal value type which is the one implemented in the early access binaries and "real" value type which will be close (closer) to the value type that will be integrated in Java.
The idea of the minimal value type prototype was to see what we need in the VM to support value types without thinking too much about how reference and value type will co-exist, so you declare a value capable class which is an immutable normal Java class and the VM is able to derive a value type from it, the value capable class is the box and the derived value type is the plain value type. Since October, we are moving to a new specs, where java.lang.Object (and any interfaces) are types that hold either a value type or reference type, but currently, there is no real implementation of this spec. > Thanks. It raises a lot of interesting question. > - Will I finally be able to add a primitive type in a generic (since it's a > value type)? that's the plan, but this is not implemented in the minimal value type prototype. I'm currently developing a prototype of the generics part, but it's not clean enough to share (yet!). > - Why not considering everything an universal type (but then how to box unbox > it)? java.lang.Object (and any interfaces) can play the role of that universal type, and we now know how to box a value type on stack (or in an arena that move like the stack), we call that buffering and not boxing because the value type is seen as a reference but not a reference on the heap. > - And do I tell a value type is a value type? With the current prototype, you put the annotation @jdk.incubator.mvt.ValueCapableClass on an immutable class. In the future, javac will have a language support for that (if you take a look to the source, there is already a proto version of that support for testing). Rémi > On 27 December 2017 at 12:13, Jay Askren < [ mailto:jay.ask...@gmail.com | > jay.ask...@gmail.com ] > wrote: >> [ https://www.infoq.com/news/2017/11/ValueTypesNov10 | >> https://www.infoq.com/news/2017/11/ValueTypesNov10 ] >> On Wednesday, December 27, 2017 at 7:54:11 AM UTC-7, Henri Tremblay wrote: >>> Hi Rémi, >>> Highly interesting. >>> For those like me who are too lazy to read an annotated JVM spec, are you >>> aware >>> of a quick summary of how it works somewhere? (over (how to specify in Java >>> code and API changes) and under (JVM instructions) the hood). >>> Thanks, >>> Henri >>> On 27 December 2017 at 07:30, Remi Forax < fo...@univ-mlv.fr > wrote: >>>> Hi all, >>>> seeing the mail from Peter on how to avoid pointer chasing, instead of >>>> using >>>> off-heap data structure, use value types. >>>> Currently, there is a prototype, named the Minimal Value Type, that allows >>>> people to test value types inside the VM. >>>> The spec is here: >>>> [ http://cr.openjdk.java.net/~dlsmith/values.html | >>>> http://cr.openjdk.java.net/~dlsmith/values.html ] >>>> There are an early access build for Linux and MacOs >>>> [ http://jdk.java.net/valhalla/ | http://jdk.java.net/valhalla/ ] >>>> The version of ASM that let you generate the new bytecodes is here: >>>> [ https://gitlab.ow2.org/asm/asm/tree/MINIMAL_VALUE_TYPE | >>>> https://gitlab.ow2.org/asm/asm/tree/MINIMAL_VALUE_TYPE ] >>>> and you can also get compiled version here: >>>> [ https://github.com/forax/valuetypify/tree/master/deps | >>>> https://github.com/forax/valuetypify/tree/master/deps ] >>>> You can use the valuetypifier, that transform Java 8 code to code with >>>> value >>>> types >>>> [ https://github.com/forax/valuetypify | >>>> https://github.com/forax/valuetypify ] >>>> regards, >>>> Rémi >>>> -- >>>> You received this message because you are subscribed to the Google Groups >>>> "mechanical-sympathy" group. >>>> To unsubscribe from this group and stop receiving emails from it, send an >>>> email >>>> to mechanical-sympathy+unsubscr...@googlegroups.com . >>>> For more options, visit [ https://groups.google.com/d/optout | >>>> https://groups.google.com/d/optout ] . >> -- >> You received this message because you are subscribed to the Google Groups >> "mechanical-sympathy" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email >> to [ mailto:mechanical-sympathy+unsubscr...@googlegroups.com | >> mechanical-sympathy+unsubscr...@googlegroups.com ] . >> For more options, visit [ https://groups.google.com/d/optout | >> https://groups.google.com/d/optout ] . > -- > You received this message because you are subscribed to the Google Groups > "mechanical-sympathy" group. > To unsubscribe from this group and stop receiving emails from it, send an > email > to [ mailto:mechanical-sympathy+unsubscr...@googlegroups.com | > mechanical-sympathy+unsubscr...@googlegroups.com ] . > For more options, visit [ https://groups.google.com/d/optout | > https://groups.google.com/d/optout ] . -- You received this message because you are subscribed to the Google Groups "mechanical-sympathy" group. To unsubscribe from this group and stop receiving emails from it, send an email to mechanical-sympathy+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
