Would micro64 consume less cpu and would it work smoother if it is
installed on a minimal os?
By the way, you are doing a great work Bero, thank you.
I think that any current x86/x64 linux distribution should do for it,
and as graphic card I would suggest one of the cheapest from the
current or pre-current NVidia serie, and as processor I would suggest
any current cheap tricore/quadcore/hexacore processor (for example AMD
A6-3500 / A6-3600) or any current fast dualcore processor with the
near same single-core performance of a current quadcore/hexacore
processor (for example any current Intel i3, i5 or i7 CPU), but I
think, that you should check the CPU rang lists from the web for
this. :)
I'm developing Micro64 at least on a Notebook with a Intel i7 2630QM
processor and NVIDIA Geforce 555m 3GB graphic chip.
BeRo
Best regards.
On 5 dic, 06:36, Ed Spittles <ed.spitt...@gmail.com> wrote:
> It should at least boot faster, and be in a more predictable environment
> (which makes it more like a console, or an original ROM-based micro, which
> is what it's emulating)
>
> I use tinycore linux for a thin client: have a look athttp://www.linuxjournal.com/article/11023?page=0,0
>
> (It can boot from the network and run from RAM, so an old diskless laptop
> becomes useful. If you boot from CD or USB stick and run from RAM you get
> advantages.)
>
> Cheers
> Ed
>
Why micro64 is so demanding of CPU? is this related to SDL? My first
core sometimes reach to 100% with micro64, but the second core is less
demanded.
Best regards.
It's not SDL-related. It's because Micro64 emulates all components at
least cycle-exact, so:
- C64 CPU: cycle-exact (or as optional dawn slow transistor-
switchlevel simulation halfcycle-exact)
- C64 VIC II: selectively single-pixel-dot-clock-exact (single-pixel
for single-pixel) together with main-halfcycle-exact-logic or dynamic-
pixel-blockwise-exact (but never more than 4 pixels at once) together
with main-halfcycle-exact-logic
- C64 SID: cycle-exact + sample-exact (in 6581 mode, it emulates also
the analog parts (ADCs/DACs,filters,etc.,) as possible exact, here the
CPU time usage killer no. 1 after the transistor-switchlevel CPU
simulation)
- C64 CIAs: cycle-exact
- Cartridge logic: cycle-exact (supports the most .CRT's from the web)
- REU: cycle-exact and complete
- GeoRAM/NeoRAM: cycle-exact and complete
- 1541 CPU: cycle-exact
- 1541 VIAs: cycle-exact (and complete incl. original-buggy shift
register and so on)
- 1541 motor physics: physically time-exact (see http://vserver.rosseaux.net/stuff/m64b.png
for the option setting screen screenshot for it )
- 1541 head logic: cycle-exact and complete including GCR electronics
confusion random bit emulation on invalid raw bit sequences.
- IEC bus: also as optional option time-exact incl. latency delays
- and so on
This all do require simply a lot of CPU time usage.
For example VICE uses here a very dirty trick, VICE emulates the CPU
instruction-for-instruction (not really cycle-for-cycle) together with
in-instruction-clock-increment-macros and dirty cycle difference
compensation tricks, so that VICE is thereby faster than Micro64,
Hoxs64 and so on, only together with loss of some accuracy, but anyway
still a good emulator and competitor to Micro64, without regard to the
weakness in the CPU emulation of VICE.
And Hoxs64 is a very good competitor to Micro64, but it does not all
the components, which Micro64 emulates (optional IEC bus latency,
cartridges, REU, NeoRAM/GeoRAM, PAL emulation, etc.), so that Hoxs64
is thereby also a bit faster than Micro64, but not much. And Hoxs64 is
currently only for Win32/Win64.
BeRo
Anyway, About the cpu use, it is an innovative and interesting aproach
to develop an emulator that doesn't constraint itself for work with
slow old CPUs.
On 5 dic, 16:42, Benjamin Rosseaux <rosse...@googlemail.com> wrote:
> Hi juga,
>
> It's not SDL-related. It's because Micro64 emulates all components at
> least cycle-exact, so:
>
> - C64 CPU: cycle-exact (or as optional dawn slow transistor-
> switchlevel simulation halfcycle-exact)
> - C64 VIC II: selectively single-pixel-dot-clock-exact (single-pixel
> for single-pixel) together with main-halfcycle-exact-logic or dynamic-
> pixel-blockwise-exact (but never more than 4 pixels at once) together
> with main-halfcycle-exact-logic
> - C64 SID: cycle-exact + sample-exact (in 6581 mode, it emulates also
> the analog parts (ADCs/DACs,filters,etc.,) as possible exact, here the
> CPU time usage killer no. 1 after the transistor-switchlevel CPU
> simulation)
> - C64 CIAs: cycle-exact
> - Cartridge logic: cycle-exact (supports the most .CRT's from the web)
> - REU: cycle-exact and complete
> - GeoRAM/NeoRAM: cycle-exact and complete
> - 1541 CPU: cycle-exact
> - 1541 VIAs: cycle-exact (and complete incl. original-buggy shift
> register and so on)
> - 1541 motor physics: physically time-exact (seehttp://vserver.rosseaux.net/stuff/m64b.png
On 5 dic, 14:36, Ed Spittles <ed.spitt...@gmail.com> wrote:
> (Wow - tce was fast to boot! micro64 didn't run on my existing tce machine
> - out of memory - had to boot a different laptop.)
>
> Yes, it does run! I had to load the 'SDL' package. ('ab' is a text-mode
> package manager. Seehttp://wiki.tinycorelinux.net/wiki:package_management_cheat_sheet
> )
>
> It's reporting 46% 24 FPS, just running the built in BASIC.
> This is on a 1.66GHz Atom N455 laptop (with hyperthreading - I don't think
> it's a genuine dual core, but it thinks it is) with 2G ram. It's
> saturating one core. The graphics is integrated. So this is a low-end
> machine.
>
> Having also loaded the 'alsa' package, I have some semblance of sound
> working too.
>
> I typed in these POKEs, in separate lines:
> POKE 54296,15:POKE 54278,240:POKE 54273,100:POKE 54276,33
> (max volume, full sustain ADSR, medium pitched note, sawtooth waveform)
> and got a rough 'boing' sound. I think I'm supposed to get something
> continuous.
>
> What's the quickest and easiest route to running something more multimedia?
>
> Cheers
> Ed
>