No randomized or unverified custom initial RAM state Emulator settings to initialize RAM a specific way are not allowed unless the entire RAM state as a whole is proven to be a possible starting state for the console being emulated. This also applies to fully random RAM state initialization, which is guaranteed to generate an invalid starting state the vast majority of the time for most consoles.
There was some discussion on the irc today about certain SNES JRPGs (like Tales of Phantasia) having different initial RNG configurations based on whether BizHawk or lsnes was used. After doing some digging, it turned out that the source of the discrepancy was that the game seeds its initial RNG state with values of the RAM prior to initialization. lsnes sets its uninitialized RAM to all 0x55s, while BizHawk apparently uses some fixed pseudorandomly generated RAM state. This brings up the question of what counts as a "proper," "valid," or "acceptable" initial WRAM configuration. I could not find any documentation on this topic so I thought that I would make a thread for people to discuss their philosophy with regard to this issue and perhaps even some actual data (if anyone has it)! After some contemplation, I justified the "legitimacy" of my BizHawk run to myself by arguing that since the random number generation only depends on two bytes, the entropy is low enough (65536 states) that it is possible for a snes to power on with the RNG seed that BizHawk generates for it. But that begs the question, why is it the case that adjusting the initial RAM values is not allowed by site rules? Why is the particular set of random values generated by BizHawk (which more likely than not do not correspond in any way to the values that a physical console would produce) preferred over some customized starting configuration? Just wondering what people's thoughts are on this matter.
But what is the "conservative route?" Taking one (or two) specific strings of RAM and declaring them to be the "legitimate ones" (even though we have no idea whether or not they are actually valid)?
Some bits are correlated, some are anti-correlated. This might be true of groups as well. We don't know enough to demonstrate which power on states are valid and which are not. In order to get sufficient statistical power, as much as a few hundred or few thousands cold power-ons can be be required.
If all TASes were unpublished except console verified ones, the site would be very bare indeed!
0x55 is such a fun pattern, multiple titles break without it, despite the fact that SNES WRAM is DRAM-based, and is thus completely unpredictable upon power-up, with serious variations across each run.
