Re: JavaMemoryModel: a naive proposal

From: Bill Pugh (pugh@cs.umd.edu)
Date: Thu Jul 12 2001 - 17:13:01 EDT


At 11:26 PM +0200 7/12/01, Assaf Schuster wrote:
> >
>> Plus, SC would substantially hurt performance of JVM's on many
>> uniprocessors, and cripple it on SMP's. I don't think our JSR could
>> ever be approved if it crippled the performance of Java.
>>
>
>I recall a paper by Marc Hill proposing SC as a hardware memory model.
>He basically points out that there are many optimizations that can help SC
>gain performance.

At the hardware level. And no major, successful processor
incorporates those ideas.

You also need to enforce SC at the software level. Which requires
disabling all compiler transformations that reorder heap memory
accesses, including elimination of redundant reads. Although there
has some work on reducing the compiler cost of enforcing SC, that
work has all been in the context of much more structured parallelism.

Looking at SC might be a interesting idea for a future programming
language, to be implemented on future hardware. Might be.

But on existing hardware with existing compiler technology, it isn't feasible.

        Bill
-------------------------------
JavaMemoryModel mailing list - http://www.cs.umd.edu/~pugh/java/memoryModel



This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:33 EDT