DSPHLE: Move CDD1 constants to UCodes.h

[Pokechu22]

These are used by all uCodes, though not all uCodes support MAIL_RESUME or MAIL_CONTINUE.

Extracted from my WIP card ucode HLE stuff, since this is also applicable to the homebrew asndlib/aesndlib ucode.

[iwubcode]

You can leave this as is but this is only used in one place. Seems strange to me to have it in an enum called MailType all by itself. I wonder if we could kill this enum with a function or something:

bool IsCmdList(u32 val)
{
   const u32 mask = 0xFFFF0000;
   const u32 expected_val = 0xBABE0000;
   return (val & mask) == expected_val;
}

or maybe there is something more appropriate?

[Pokechu22]

The way it's actually used is like this:

  else if ((mail & MAIL_CMDLIST_MASK) == MAIL_CMDLIST)
  {
    // A command list address is going to be sent next.
    set_next_is_cmdlist = true;
    cmdlist_size = (u16)(mail & ~MAIL_CMDLIST_MASK);
  }

Probably the best choice is to just use constexpr u32 instead of enum (and it could also be put in the cpp file instead of the header, but I have less of an opinion on that front).

[AdmiralCurtiss]

Technically a behavior change here: If a mail that triggers the above panic alert comes in, it could previously trigger a case in this switch, and now it can't. I assume that's fine?

[Pokechu22]

That wasn't an intentional change; I'll try to verify which is correct now.

[Pokechu22]

The old behavior was correct; the zelda uCode doesn't check that the prefix is CDD1 (but it usually is in practice).

• 56D36052 (Super Mario Sunshine) does this at inst:05db (with a jump table at inst:05d7). The byte sequence CDD1 doesn't appear in it at all (though DCD1 does appear).
• 24B22038 (NTSC GameCube IPL) does not check it at all; neither CDD1 nor DCD1 appear. The comment on SUPPORTS_GBA_CRYPTO about "This was used before the GBA UCode and UCode switching was available." makes this seem a bit reasonable; the CDD1 and DCD1 mails are (I think) part of the powerpc uCode switching logic (which is also why they're standard between different uCode). Note that the IPL doesn't support GBA uCode, but it does (seem to) include a copy of the CARD uCode in it, which Dolphin doesn't correctly handle right now (but we have a workaround that means the CARD uCode is never needed). This version also uses the "light" protocol, while the CDD1 handling is in the default "protocol" only, so we're handling this correctly.
• D643001F (Super Mario Galaxy 2) does this at 0734 (with the jump table at 0730). CDD1 doesn't appear in it at all (though DCD1 does appear).

For what it's worth, the AX uCode is the same:

• 07F88145 (not sure which game I used, but it's apparently a common one) has it at 0f34 and the jump table at 0f30; CDD1 doesn't appear in it at all (though DCD1 does appear). I'm not 100% sure how this is reached, but it does send 0xdcd10002 (DSP_YIELD) beforehand. I think that this is one of the few places where they process mail, though, so it's not like there's a separate state where it won't accept the mail.

  (I also see some other outgoing mail, but it all seems to be in special cases: 0xfbad8080 at 0079 as the instruction after a JMPR (should be unreachable), 0xbaadXXXX at 007e if XXXX, the command to use, is out of bounds (negative (which is impossible based on how it's loaded) or greater than 0x13), 0xecc0XXXX at 0eb3 if a stack overflow (exception 1) or reset (exception 0) occurred where XXXX is the value of $st0 (i.e. the PC before the exception occured), and 0xfeedXXXX at 0ec2 if exception 2 (currently unknown) occurred (again where XXXX is $st0). (That last one is different from the 0x8071feed used by the ROM uCode).

• 4CC52064 (Wii Sports Resort) has it at 0f99 and the jump table at 0f95. CDD1 doesn't appear in it at all (though DCD1 does appear).

So, the old behavior was more accurate here. (The CARD and GBA uCode, however, do check for the CDD1 prefix.)

[Pokechu22]

I've fixed this and also added comments detailing how the various mails are handled.

I'll handle the AX version in #10768 because it turns out AX also has global state (but not enough global state to handle the masking accurately).

[AdmiralCurtiss]

Did you mean treat it as DSP_DONE instead? Not sure how to parse this otherwise.

[Pokechu22]

[it is] treated [by the CPU] as [if it were] DSP_DONE instead is what I had in mind. I'll expand it.