Emuladores de Ajedrez de MESS

[CBF]

Hi Berger,

you can also remove the "Update5.zip" from this site, because 2 days ago (20.6.) I've uploaded
a new complete package with the latest changes ("CPU Divider" and new Prestige artwork).

CBF

[berger]

Done! Thank you for reporting this.

Berger

[Luis a]

Hola,

Tienes un montón de trabajo hecho pero muy disperso. Me pregunto si podrías agruparlo todo en un archivo para reducir el número de enlaces de descarga.

Saludos,
Berger

Vale. Voy a revisarlos con la última versión de CB-Emu y luego los agruparé.

Saludos,
Luis

[berger]

Vale. Voy a revisarlos con la última versión de CB-Emu y luego los agruparé.

¡Gracias!

[tibono]

Hi,
thanks CBF, as usual.

For your information, I noticed a unusually high CPU load using the EAG2100 with CPU divider activated (which is the default setup). Too high on my laptop's core, stuck at 100% load, to avoid clock shifting.
Disabling it using configuration setting gets back to normal load.
I had no similar issue with EAS Budapest or Glasgow, other ones using this new accurate CPU clocking.
Kind regards,
Tibono

[tibono]

Hi,
I performed additional testing. The CPU divider feature seems really costy, whatever the emulated system:

on my laptop (dual core T4500 @2,3Ghz)
feasbu 3,2Mhz previous version (May) uses 16% CPU of 1 core --> new version uses 64% (x4)
feasgla 3,6Mhz ======================= 18% ============================= 72% (x4)
feag2100 6Mhz ======================= 22% ============================= 100% (> x4)

So no real specific with Elite A/G 2100 except it reaches too high a CPU peak on my configuration assuming CPU divider feature.
Which acts as a multiplier as far as CPU consumption is concerned!

KR, Tibono

[CBF]

Hi,
I performed additional testing. The CPU divider feature seems really costy, whatever the emulated system:

Yes Tibono, unfortunately you're right -
this CPU divider feature is definitely not working correctly (at least in MESS version 0.175 which I'm using)!

I can also confirm these problems with all modules using such a CPU divider, and since 'hap' certainly won't help me
(and I don't have enough knowledge of hardware and the MAME internals) the only way for me is to completely
remove the code for these CPU dividers in all devices and go back to older driver versions.

I'm already working on it, but it's a lot of work - the current CB-Emu version is seriously broken by the last MAME updates ...

CBF

[hap]

I agree that the 4*more taxing resources is unfortunate. But it is as expected. The reason is that before each read/write, MAME inserts a break in the emulation flow to check the address bus to apply the cpu divider(this is up to a couple of million times per second).

[CBF]

I agree that the 4*more taxing resources is unfortunate. But it is as expected.

Hmm? Are you serious?

How useful is it to add this CPU divider code just to make the device timing a bit more accurate,
when now even a modern computer (Core i5/3GHz) is not able to emulate a 4-8 MHz 6502 CPU with full (100%) speed???

I've now removed all this CPU divider ('trampoline') code again in CB-Emu and reverted it back to a previous state.
Now there's still one thing in the code, which I'm not sure about:

In older versions there was this code (as example) in MACHINE_CONFIG:

Código: Seleccionar todo

MCFG_CPU_PERIODIC_INT_DRIVER(fidel6502_state, irq0_line_hold, 600) // guessed

This has been changed recently to:

Código: Seleccionar todo

MCFG_TIMER_DRIVER_ADD_PERIODIC("irq_on", fidel6502_state, irq_on, attotime::from_hz(38400/64))
MCFG_TIMER_START_DELAY(attotime::from_hz(38400/64) - attotime::from_hz(38400*2)) // edge!
MCFG_TIMER_DRIVER_ADD_PERIODIC("irq_off", fidel6502_state, irq_off, attotime::from_hz(38400/64))

MCFG_TIMER_DRIVER_ADD_PERIODIC("dummy_timer", fidel6502_state, dummy, attotime::from_hz(XTAL_4MHz))

Have these changes anything to do with the CPU divider, or are these changed lines just a 'better' code?
In other words:
Since I've now removed the CPU divider code again, should I rather use the 1st (older) or the 2nd (newer) version above?
Is this "dummy_timer" necessary at all without the CPU divider code?

CBF

[hap]

dummy_timer is for the CPU divider. If you remove the CPU divider, you won't need this.
IRQ timer is unrelated to it.