Re: [B2-devel] Improved SheepShaver/JIT

[Robert S. <Ro...@xb...>]

I have noticed an approx 13% boost in speed for real world bench marks.
Unstuffing files , loading large images in Graphics converter and
preforming fliters on those images. By using ICC over GCC using the -O3
arugement.  This was on a P4 Northwood.  I will have to see what kind of
boost i get now with your updated code when my p4 machine is among the
living .  My P4 died my Athlon died my G4s 250 gig drive die all in the
weekend.  I will be able to get the athlon machine up first tas all it
needs is a new bios chip.  But I am going to download sheep shaver from
cvs again and try it out on this old machine just to see the speed it
pulling off of an old 600mhz celey.



On Tue, 2004-01-27 at 09:03, Gwenole Beauchesne wrote:
> Hi,
> 
> I have just committed a few changes to the SheepShaver JIT. Basically,
> gprof told me that 64% of the time was spent in block_cache::find().  
> Besides, some extra counters revealed that more than 95% of the blocks are
> available at the head of the blocks list for a hash table entry.
> 
> i.e. I inlined that fast block lookup case in generated code. On a
> Pentium4, Speedometer 4 pure "CPU" benchtest improved by a factor of 3.  
> More testing/tuning has yet to be done but overall the speed increase is
> interesting. ;-)
> 
> Please give it a try.
> 
> BTW, there are also fixes to enable MacBench 5.0 to work.
> 
> Bye,
> Gwenole.
> 
> 
> -------------------------------------------------------
> The SF.Net email is sponsored by EclipseCon 2004
> Premiere Conference on Open Tools Development and Integration
> See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
> http://www.eclipsecon.org/osdn
> _______________________________________________
> basilisk-devel mailing list
> bas...@li...
> https://lists.sourceforge.net/lists/listinfo/basilisk-devel