Imported ATmega128 onboarding smoke
This profile measures AVR8 execution, GPIOA and polled USART0 TX against the
imported 128 KiB flash / 4 KiB SRAM geometry (SRAM from 0x0100). The
ATmega128 keeps the classic I/O file (PORTA at 0x3B, read-only PINx) and adds
extended I/O at 0x60..0xFF; the AVR core decodes that whole 0x20..0xFF range
with the classic layout rather than the ATmega328P map (where 0x3B is PCIFR).
The chip maps only porta (PINA/DDRA/PORTA at 0x39..0x3B) and the CPU's USART
host window, onto which the core translates UBRR0L/UCSR0B/UCSR0A/UDR0 at
0x29..0x2C. The fixture is
tests/fixtures/avr/atmega128-io-smoke.c
built with -mmcu=atmega128. It runs seven checks: UBRR0L and UCSR0B readback,
UDRE0 set, SBI on PORTA keeps PA7 and PINA0 follows the latch, CBI clears PA7
only, a PINA write does not toggle (PINx is read-only on this part), and a
160-byte SRAM block at 0x122..0x1C1 reads back. It stores the check bits at
0x01C2, prints M128 IO bits=7F and toggles PA0 in a loop. The smoke asserts
the UART text, the result byte 0x7F, at least two PA0 edges and the bounded
max_steps stop after 100,000 instructions.
Negative control: the same firmware and script against the previous CPU-only
chip descriptor (peripherals: []) fails 0/4 (memory violation at 0x10039).
Not claimed: ports B-G (the core mirrors ports A-E, but only porta is mapped
and asserted), UBRR0H/UCSR0C, USART1, timers, ADC, SPI, TWI, interrupts,
EEPROM and the ATmega103 compatibility fuse. The run assumes an ideal power
rail. Result JSON, JUnit assertions, simulator logs, and the snapshot are under
docs/coverage/renode-onboarding/metrics/simavr-atmega128/.
This is local CLI smoke evidence against the imported manifest, not full
ATmega128 coverage or hardware confirmation.