[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [DotGNU]Will unrolling an instruction slow down execution ?
From: |
Gopal V |
Subject: |
Re: [DotGNU]Will unrolling an instruction slow down execution ? |
Date: |
Wed, 14 May 2003 19:30:38 +0530 |
User-agent: |
Mutt/1.2.5i |
If memory serves me right, Rhys Weatherley wrote:
> That's a pretty fair assumption, modulo register assignments to stack
> positions. The main tricky part is with exceptions - you need to "ReExecute"
> in the interpreter to make the exception work.
I'm hoping that the offsets of data structures like System_String->length
etc will remain constant ?... Those -8(%ebp) stuff sounds too freaky to
hand code ...
I see that you have store-fetch forwarding in registers which is cool ...!
So
stloc.0
ldloc.0
is only one memory write and there is not even a need for a register move
there !.
> The hard stuff will eventually turn into helper functions, called from the
> JIT. For now, the CVM interpreter does the hard stuff.
I mean
x86_call(&unroll,_StringEquals);
> Parts of the generic unroller are now checked in, although they haven't been
> enabled just yet because it is incomplete.
I'll look into this in detail later .. Right now I'm re-writing the
shared copyrighted parts of the Java compiler ...
Java is more important than a 2% speed of the JIT , right ? :-)
Gopal
--
The difference between insanity and genius is measured by success