[Rpcemu] RPCEmu 0.8.11

george greenfield george.greenfield at tiscali.co.uk
Thu Oct 24 05:18:24 PDT 2013


In message <a0f19f9f53.pittdj+ at iyonix.home>
          David Pitt <pittdj at pittdj.co.uk> wrote:

> In message <6d9d959f53.George at tiscali..co.uk>
>   george greenfield <george.greenfield at tiscali.co.uk> wrote:
> 
>> In message <mpro.mv61td00381ke03n2.pittdj at pittdj.co.uk>
>>           David Pitt <pittdj at pittdj.co.uk> wrote:
> 
>>> Peter Howkins, on 23 Oct, wrote:
>>> 
[Snip]
> 
> I have just tried a brief test or two but they don't really tell us
> much. The tests are !Desk_Bogo and Dhrystones.
> 
>                         bogomips  dhrystones
> 
> Ubuntu 12.04 on VMware on an iMac, rpcemu 0.8.11
> Phoebe OS3.80 Interpreter    41    98,396
> SA110  OS5.21 Recompiler    117   481,463
> 
> This is not like for like, no surprises, the Recompiler is a lot
> faster.
> 
> Mac OS X Mavericks, rpcemu 0.8.9
> SA110  OS5.21 Recompiler     159   627,746

ROMark scores c.610k dhrystone MIPs on my 0810/520 Recompiler 
installation here, running on a Dell XPS 3.4GHz i7 Win7/64 platform.
> 
> This is more interesting there is that much difference between the
> Virtual Machine and native installations.
> 
> Mac OS X Mavericks, VRPC
>        OS4.39                320   484,730
> 
> There is the oft repeated story that !Desk_Bogo favours VRPC, but
> Dhrystones indicate rpcemu is faster.

Indeed: that suspicion is supported by Chris Hall's benchmarks here
https://www.riscosopen.org/forum/forums/5/topics/466?page=8
where VRPC on a 2668MHz PC gives only 175% compared to a bog-standard 
S/ARM RPC, i.e. 1524MHz = 100%; the corresponding figure for RPCEmu 
here is approx. 1000MHz = 100%. Obviously that assumes a similar 
processor type in the 2668MHz machine, which may not be the case.
> 
> Raspberry Pi
>        OS5.21                401   995,024

That's a very high dhrystone figure for a Pi: I got 450,080 when I 
tested one recently.
> 
> Emulators not too dusty then!

No indeed.
> 
> Phoebe in hardware might have been a step forward but that might not
> apply to an emulation.
> 
As Peter H has just explained.

-- 
george greenfield



More information about the Rpcemu mailing list