overclock question

[negativeExponent]

the overclock logic is incorrect.

nestopia/source/core/NstApu.cpp

Lines 2382 to 2390 in c24ffe8

	NST_SINGLE_CALL void Apu::Dmc::WriteReg2(const uint data)
	{
	regs.address = 0xC000 | (data << 6);
	if (regs.address != 0)
	{
	overclockingIsSafe = true;
	}
	}

overclockingIsSafe will always be true because regs.address is guaranteed to be at least have a value of 0xC000

nestopia/source/core/NstApu.cpp

Lines 2392 to 2400 in c24ffe8

	NST_SINGLE_CALL void Apu::Dmc::WriteReg3(const uint data)
	{
	regs.lengthCounter = (data << 4) + 1;
	if (regs.lengthCounter != 0)
	{
	overclockingIsSafe = true;
	}
	}

same here. regs.lengthCounter will always be at least 1.

[hizzlekizzle]

What does "safe" actually mean in this context?

[negativeExponent]

when DMC is latched, thats considered unsafe. It causes PCM audio channel to run faster than it should when overclocking. If this was following FCEUx ppu overclock logic, Reg2 and Reg3 writes should only check value of incoming data

https://github.com/libretro/libretro-fceumm/blob/59771792b0709f10c9bc7b408e41ae6df55d6dab/src/sound.c#L289-L302

[keithbowes]

Well, you should probably forward these concerns to the upstream code. I don't think anyone here does anything for the Nestopia backend but only for the libretro interface, so if this were fixed upstream, it would get merged into this core later.

[negativeExponent]

Upstream (if it's still even active) does not have overclock option, not has the libretro port. either case the condition is still incorrect and will always be true.

[keithbowes]

You're right. I didn't think the core had anything that the upstream backend didn't. I guess I was wrong.

Judging from the git logs, it seems that upstream did once have overclocking but it was considered buggy and was removed, but the project's originator readded the code specifically for the libretro core.