Thanks for the hints,
So, I found valuable material from IBM research
(http://www.research.ibm.com/journal/sj39-1.html)
In particular, http://www.research.ibm.com/journal/sj/391/suganuma.html
is very detailed, yet does not mention prefetch issues.
There it says that for IBMJDK1.1.7 Pentium and PentiumPro family
(Pentium Pro, Pentium II, and Pentium III) specific instructions are
generated. They say code scheduling in 1.1.7 is not yet available for
the PentiumPro family, but they promise for 1.1.8. So maybe, something
is already there.
SHUDO Kazuyuki wrote:
> Java Grande Forum (*1) Numerics Working Group (*2)
> (JGNWG) has been trying to allow JVM and JIT developers
> to use PowerPC's fmac (fused multiply and accumulate)
> instruction. In May 1998, They proposed an
> `associativefp' modifier, but the proposal is refused by
> Sun. But a `strictfp' modifier which is introduced into
> JDK 1.2 is an important result of the discussion.
fma is also an issue with Titanium, since it also uses fma's for fast
IEEE standard conforming floating point operations...
> JIT compiler may utilize even MMX instructions of
> x86. Of course, current JITs cannot utilize them
> efficiently, but I have a paper written by Intel
> researchers. The paper describes the method for JIT to
> exploit MMX instructions.
Could only find an announcement for two Intel tutorials at SC and
JavaGrande, but nothing written. Do you have something printable?
Regards,
Wolfgang.
----------------------------------------------------------------------
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
