On Oct 23, 2011, at 8:24 PM, Edward Harned wrote: > opinali: > > Good point about not participating in JSR166. I wasn't aware of just what DL > was doing. From the Brian Goetz article "Stick a fork in it" the framework > was supposed to be for compute only on very large arrays. More of the > scientific/academic stuff. I do business application so I wasn't concerned. > Once I found out last summer exactly what it included, yes it was too late. > > Benchmarks can be manipulated in any direction which is why I said you need > to consider the internal workings. > > jCilk is a separate application framework; exactly where parallelism belongs. > > Kirk: > > Unsafe direct memory modification bypasses all the managed runtime safety > guarantees and security features of the Java language. No bounds checks. No > memory overwrite checks, etc. And it adds unnecessary complexity. As far as I > can see there is not other reason for using it in the F/J framework other > then to get faster execution. There is a j.u.c.LinkedBlockingDeque so there > was no need to invent another.
If you get the unsafe stuff wrong it will throw reasonable exceptions at reasonable times. I know, I've done unreasonable things and was on the receiving end of some interesting exceptions ;-). Every VM I've worked with (prior to Java) came with a backdoor (set of unsafe operations). The more famous one in SmallTalk was the method become:. Completely unsafe!! So the trend continues with Unsafe. Not to say that the needed operations should stay in Unsafe. I'd rather see them exposed in a safer way but to take away access to CAS (or the equivalent in IBM land) closes the door on some interesting solutions to concurreny. Look at Cliff Clicks coding style to higher concurrency. It relies on CAS and a number of other hardware/JMM features in a way that average programmers given todays tools should probably shy away from. That said, who are we to mother other developers away from experimenting with these techniques? Regards, Kirk -- You received this message because you are subscribed to the Google Groups "The Java Posse" group. To post to this group, send email to javaposse@googlegroups.com. To unsubscribe from this group, send email to javaposse+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/javaposse?hl=en.
