Re: Hardware-based vs software-based emulation

From: smf <smf_at_null.net>
Date: Fri, 17 Feb 2017 12:52:45 +0000
Message-ID: <5c88063e-ea97-d4df-e616-ea719a7cb567@null.net>
On 17/02/2017 12:05, groepaz@gmx.net wrote:

> actually.... iirc beros micro64 can poll the inputs once per 
> rasterline at
> least :) (which is probably already faster than the OS can do it....)
>

I haven't tried it and it's closed source, is it any better in practise? 
If it's running full speed and then waiting at the end of the frame, 
then even if the OS polls 10 times per frame, the emulator could still 
have finished it's entire frame before the OS polls for the second time.

http://www.micro64.de/

"Hyper64 is a working proof-of-concept of an experimental C64 emulator 
implementation without resorting to a conventional CPU emulation, but 
instead utilizing a realtime dynamic recompilation engine – which 
disassembles, analyzes and re-assembles the 6510 machine code as native 
32-bit x86 code. It further uses a fast translation code cache.

The A, X and Y registers of the 6510 are mapped to the x86 AL, BL and CL 
registers and the 6510 flags are mapped via the x86 SAHF/LAHF opcodes.

Although this yields to a lot of speed emulation-wise, it also has the 
side-effect of not being cycle exact, but instead being accurate on an 
instruction basis only.

The VIC II emulation is – as far as I'm sophisticated – complete but 
also not cycle exact. The same thing applies to the CIA emulation, whose 
accuracy is based on a VIC II line basis."



       Message was sent through the cbm-hackers mailing list
Received on 2017-02-17 13:02:58

Archive generated by hypermail 2.2.0.