Select it with --save-type sram (SaveDevice::Sram). The array is byte-wide,
starts erased, and repeats every 32 KiB through the 0x0e00xxxx save window.
A CPU byte store is staged and committed when its step completes, so failed
steps roll back and reads later in the step see the byte. Unlike Flash, SRAM
can be read from ROM code. Halfword/word access and DMA stay diagnostic until
the 8-bit bus behavior for those widths is verified.
SRAM shares the save-image API, host save files, and save states. Its device
code is new, so existing save states keep loading.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HhfHfHEuzv2AUTd6erW1gU
Stores to I/O addresses with no register returned Unmapped and stopped the
machine. Hardware ignores them. Emulated writes there now pay their bus cost and
change nothing; host writes stay strict. The unmodeled internal memory control
register at 0x04xx0800 and all unused I/O reads remain diagnostics.
Rollback tests that used unused I/O as their failing word now run the block on
to a register that rejects the stored value: SIOMULTI2 outside multiplayer
mode, an RCNT GPIO serial IRQ, or 64-sample wave playback. Where no reachable
register can fail (JOYSTAT, POSTFLG/HALTCNT, IME, DMA3CNT), the test now checks
that the block commits and the unused word is ignored.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HhfHfHEuzv2AUTd6erW1gU
MUL/MLA with Rd=Rm and long multiplies with RdHi or RdLo equal to Rm returned
UnsupportedInstruction and stopped the machine. The ARM7TDMI reads every
operand before writing the result, so these forms execute normally. R15
operands and RdHi=RdLo remain diagnostics.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HhfHfHEuzv2AUTd6erW1gU
DISPCNT enable bits for BG3 in Mode 1, BG0/BG1 in Mode 2, and every background
but BG2 in Modes 3-5 failed the whole frame with UnsupportedLayers, and the
window turned that into a stopped session. Hardware has no such backgrounds and
ignores the bits, so the renderer now masks them. The error variant is removed.
Prohibited OBJ mode 3 and shape 3 stay diagnostics: their hardware output is
not established here.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HhfHfHEuzv2AUTd6erW1gU
Stores to 0x08-0x0c returned ReadOnly and stopped the machine. Hardware ignores
them, and a second game's boot code stores to a ROM mirror. Emulated writes now
pay their bus cost and change nothing. The 0x0d window stays a diagnostic until
EEPROM is modeled there, and host writes stay strict.
Tests that used a ROM store as their failing write now cross an ignored ROM
word into the 0x0d window.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HhfHfHEuzv2AUTd6erW1gU
A core, DMA, rendering, window, or audio error returned before the save was
written, so every completed in-game save since launch was lost. The window now
writes changed bytes on that path too. The diagnostic stays the session error;
a refused or failed write is reported as output. A failed state restore leaves
save memory undefined and is never written. Terminal runs keep the strict rule.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HhfHfHEuzv2AUTd6erW1gU
Machine::save_state and load_state encode the CPU, every RAM, and every device
through Snapshot impls into a versioned, checksummed format. Rejected loads leave
the machine unchanged. The ROM window saves to and loads from one state file
with F5 and F9.