Skip to content

Firmware-exercise matrix

Where VALIDATION_STATUS.md asks "is the model right vs silicon?", this matrix asks the complementary question: does a real booting firmware actually RUN against the model? A model is only counted when a booting firmware drives it — a banner-only boot or a register-poke unit test does not.

Rubric coverage — authoritative, from the Tier-1 fixtures

Each chip's Tier-1 fixture is real firmware that drives these peripheral classes and self-reports PASS over UART; the grid below is generated from docs/coverage/tier1-matrix.json (ratchet-gated, so it cannot drift). 🟢 driven-by-firmware · 🟠 modeled but fixture does not exercise it · · not modeled on this chip (n/a).

chip clock gpio uart timer dma irq i2c spi adc pwm wdt rtc can rmt proven
esp32 🟢 🟢 🟢 🟢 · 🟢 🟢 🟢 🟢 🟢 · 🟠 9
esp32c3 🟠 🟢 🟢 🟢 🟠 🟢 🟢 🟢 🟢 🟢 · · 8
esp32c6 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 12
esp32s3 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟠 🟠 🟢 · · 🟢 9
nrf52832 🟢 🟢 🟢 🟢 · · 🟢 🟢 🟢 🟢 🟢 🟢 10
nrf52833 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 12
nrf52840 🟢 🟢 🟢 🟢 · · 🟢 🟢 🟢 🟢 🟢 🟢 10
rp2040 🟢 🟢 🟢 🟢 🟠 🟢 🟢 🟢 🟢 🟢 🟢 🟢 11
stm32f103 🟢 🟢 🟢 🟢 🟢 · 🟢 🟢 🟢 · 🟢 🟢 10
stm32f401 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 12
stm32f407 🟢 🟢 🟢 🟢 🟠 🟠 🟢 🟢 🟢 🟠 🟢 🟢 9
stm32f411 🟢 🟢 🟢 🟢 🟠 · 🟢 🟢 🟢 · 🟢 🟢 9
stm32g071 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🔴 🟢 🟢 🟢 🟢 11
stm32g431rb 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 12
stm32g474re 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 12
stm32h563 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 13
stm32h735 🟢 🟢 🟢 🟢 · 🟢 🟢 🟢 🟠 🟢 🟢 · 9
stm32l073 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 · 🟢 🟢 11
stm32l476 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 · 🟢 🟢 11
stm32wb55 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 12
stm32wba52 🟢 🟢 🟢 🟢 🟢 🟢 🟢 🟢 · 🟢 🟢 🟢 11

223 rubric cells are driven by real firmware across 21 chips; 11 are modeled-but-unrecorded (see below). Most chips drive 9–13 of the common peripheral classes — the STM32 family included.

Unrecorded rubric cells (modeled, fixture does not drive them): esp32/rtc, esp32c3/clock, esp32c3/dma, esp32s3/spi, esp32s3/adc, rp2040/dma, stm32f407/dma, stm32f407/irq, stm32f407/pwm, stm32f411/dma, stm32h735/adc.

Beyond the rubric — functional reads, advanced peripherals, shims

What the rubric grid does not cover: real sensor/protocol behavior over a bus (ideally through an unmodified driver decoding sensor values), register-level firmware checks, chip-specific advanced peripherals, and shims (hardcoded stubs or declarative register files with no engine behind them). From validation/firmware_exercise.yaml.

stm32l476

Richest ARM coverage: drives 11 rubric classes via real firmware, plus the gated IO-Link station (real AHT20/BMP280 + shift-register devices over USART3/I2C). Beyond the rubric, PR-gated survival firmware also drives PWR (cubeMX VOS/VOSF handshake), FLASH (ACR 4WS latency dance), DBGMCU (IDCODE=10076415), LPTIM1 (ENABLE-gated ARR/CMP writes, ARROK/CMPOK), SDMMC1 (silicon-diffed CMD/CTIMEOUT handshake) and COMP/TSC (silicon-diffed EN→VALUE and START→EOAF/MCEF). RNG/CRC/FMC/SPI2/SPI3 are unit-tested only — RNG's only coverage is the shared L073 kernel-clock tests over the L073 bus (rcc_kernel_clock_gate.rs:184-254, no L476 test) — and DAC/QUADSPI/SAI1/USB OTG FS are modeled but appear only as reset-value reads in gated transcripts, while LPTIM2/SAI2 are touched by no firmware and no test; SYSCFG and WWDG are inert shims.

Functional device/protocol reads (real driver, decoded value): - IO-Link master station — AHT20 + BMP280 over I2C, SN74HC165 shift register — iolink-station · examples/iolink-station ci/test.sh (firmware-gate) (PR gate)

Advanced peripherals — unit-tested only (no firmware drives them): SPI2, SPI3, CRC, RNG, FMC

Dead (6 modeled, never exercised): DAC, LPTIM2, QUADSPI, SAI1, SAI2, USB OTG FS

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - SYSCFG (configs/chips/stm32l476.yaml:421 (type: stub; configs/peripherals/stm32l476/syscfg.yaml; docs/coverage/silent-path-census.md:403)) — CLOSABLE: author-declared stub — read 0, writes dropped, so EXTI source-select (the block's main job) is not modelled and nothing reads the window back: r12 gates its clock because COMP shares the window (crates/firmware-l476-demo/src/r12.rs:130), and the silent-path census records it as the live stub of examples/ci/l476-bldc-stall.yaml. - WWDG (crates/core/src/peripherals/wwdg.rs:63 (configs/chips/stm32l476.yaml:396; configs/peripherals/stm32l476/wwdg.yaml)) — CLOSABLE: inert register bank — tick() is the trait-default no-op, so an armed watchdog never fires; the PR-gated nucleo_l476rg_misc fingerprint reads CR/CFR/SR off it (crates/core/tests/gates/firmware_survival.rs:623), but that exercises storage only, not watchdog behaviour.

esp32c2

No beyond-rubric content is claimed. The ESP8684-DevKitM-1-H4X smoke fast-loads firmware that checks UART0 output and the GPIO0 red LED latch; this profile does not exercise mask-ROM boot, electrical pad routing, or radio behavior.

esp32c3

Leo air-quality is the sensor-depth proof: the nightly example e2e asserts it explicitly, and the advisory PR workspace shards also exercise it. clock/dma stay unrecorded by design — esp32c3 wires system/rtc_cntl/dma as declarative register files with no behavioural engine, so a PASS would overclaim. The non-rubric declarative wall (crypto, USB device, I2S, UHCI, radio) is not yet swept into this entry — tracked for a later ledger pass.

Functional device/protocol reads (real driver, decoded value): - Leo air-quality — real Sensirion SCD41/SGP41/SPS30 + Melexis MLX90614 + Vishay VEML7700 decode CO2 / humidity / surface condensation to plain-language verdicts — leo-airquality · e2e_leo_airquality.rs:121-143 (nightly CI)

Advanced peripherals — unit-tested only (no firmware drives them): ana_i2c, sha, virtual_wifi (real shared-802.11 medium model two C3 firmwares associate over, 2 tests), wifi_mac (RE'd MAC<->SimNet bridge, RX descriptor ring, 5 tests)

Dead (2 modeled, never exercised): cache, rng

esp32s3

Broad model surface (~35); the rubric grid proves ~9, plus a TMP102 functional read in the nightly fixtures lane. GDMA is a real 65-test mem-to-mem engine (proven via the dma rubric cell), not a stub. spi_mem_flash/extmem stay unit-only: mask-ROM bring-up reads are not firmware exercise.

Functional device/protocol reads (real driver, decoded value): - TMP102 temperature read over I2C (real driver, decoded value + GPIO threshold) — esp32s3-i2c-tmp102 · e2e_i2c_tmp102.rs:147-188 (nightly CI)

Advanced peripherals — unit-tested only (no firmware drives them): aes, ds, hmac, rsa, rng, sha, i2s, lcd_cam, sdmmc, usb_otg, gpspi, pcnt, io_mux, extmem, spi_mem_flash, sens, system, core1_control, crosscore_ipi

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - efuse_stub (crates/core/src/system/xtensa/esp32s3.rs:647 (esp_xtensa_common/system_stub.rs:454)) — CLOSABLE: canned MAC + chip-rev only; no eFuse state, no burn path. The ROM boot and esp-hal accept the canned values. - rtc_cntl_stub (crates/core/src/system/xtensa/esp32s3.rs:640 (esp_xtensa_common/system_stub.rs:203)) — CLOSABLE: register round-trip plus three canned behaviours — PLL_LOCK seeded, TIME_UPDATE snapshot handshake, APP_CPU un-stall. The RTC clock tree, SWD_CONF super-watchdog and sleep domains are not modelled. - wifi_thunks (esp32s3/wifi_thunks.rs:8 (CHEAT(THUNK-LIB))) — IRREDUCIBLE: the ESP32 WiFi MAC/PHY is a closed RF-coprocessor blob with NO executable image — there is no firmware to run, so it can never be firmware-exercised. The lwIP/socket layer above is routed to a real SimNet.

esp32

tier1 fixture drives 8 rubric classes incl. a BMP280 chip-id read over the I2C command-list engine and a real LEDC duty-latch. The PR-gated ESP32-DevKitM-1 profile adds a BME280 chip-id read on classic I2C0.

Functional device/protocol reads (real driver, decoded value): - ESP32-DevKitM-1 + BME280 — CHIP_ID 0x60 over I2C0 GPIO21/22, GPIO23 high on match — esp32-devkitm-1-bme280 · examples/esp32-devkitm-1/evidence/bme280 (local smoke 5/5; fast-boot fixture) (PR gate)

Advanced peripherals — unit-tested only (no firmware drives them): efuse, sha, syscon, twai, mcpwm, rtc_cntl

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - sdio_stub (esp32/sdio_stub.rs:5) — STRUCTURAL: hardcoded FSM_DONE so the boot ROM's SDIO bring-up poll terminates. Removing it breaks boot; the ROM blob is the only 'firmware' that touches it.

esp32-devkitc-32e-gpio-smoke

Narrow ESP32-DevKitC-32E onboarding profile (GPIO23, UART0, I2C0). docs/receipts/2026-11-09-esp32-deep-sleep records stock Arduino-ESP32 timer-wake deep sleep printing LW_DS_OK (bootCount 1→2→3 across DEEPSLEEP_RESET, RTC slow memory kept) and light sleep printing LW_LS_OK. Firmware is PlatformIO-built, not a committed ELF; those receipts are not a required CI lane. The committed system attaches no external I2C device. No beyond-rubric block is listed for this slice.

esp32-devkitc-32e-uart1

Narrow ESP32-DevKitC-32E onboarding profile that routes UART1 TX to J3 pin 12 / GPIO16. Committed evidence at examples/esp32-devkitc-32e-uart1/evidence/run shows examples/esp32-devkitc-32e-uart1/uart1-smoke.yaml passed 3/3: UART0 printed ESP32_DEVKITC32E_UART1_TX_OK, GPIO16 (UART1 TX) saw at least 32 edges from Serial1 printing UUUUUUUU, and the run stopped at max_steps. Firmware is PlatformIO-built via scripts/build-esp32-devkitc-32e-uart1-smoke.sh, not a committed ELF; no required CI lane runs this script. No beyond-rubric block is listed for this slice.

nrf52840

TWIM/PWM are firmware-proven via the rubric fixture; the PR-gated OBD2 scanner e2e drives the RADIO TXEN/READY/START/END handshake (asserts ble_tx > 0), and the #[ignore]d proximity e2e adds an on-demand radio path. The Zephyr BME280 temperature read runs in an external west build.

Functional device/protocol reads (real driver, decoded value): - BME280 temperature read (~25 °C) via an unmodified Zephyr sensor driver — zephyr-bme280-ztest · external west build (not in core cargo test) (on-demand) - Stock ArduinoCore-mbed sketch on the Arduino Nano 33 BLE Rev2: the scripted USB host enumerates the native port (nrf52840_usbd) and opens its CDC serial port; the Adafruit BME280 library reads the sensor over Wire (TWI0 master, byte registers) and prints 25.00 C, 1013.25 hPa, 50.00 %RH on Serial (USB CDC) — arduino-nano-33-ble-bme280 · crates/core/tests/nrf/arduino_nano_33_ble_bme280.rs (tests/fixtures/nrf52/arduino-nano-33-ble-bme280.elf) (PR gate)

Advanced peripherals — unit-tested only (no firmware drives them): TWIS, SPIS, CCM, RNG, TEMP, ECB, GPIOTE, PPI, NVMC, FICR, UICR, NFCT, PDM, EGU, QSPI

Dead (8 modeled, never exercised): QDEC, AAR, ACL, COMP, CRYPTOCELL, I2S, LPCOMP, MWU

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - USBREGULATOR (nrf52/usbregulator.rs (70 lines, 0 tests)) — thin VBUS-ready register stub

nrf52832

Shares the nRF52 model; tier1 fixture drives TWIM/SPIM/SAADC/WDT/RTC/PWM. (The separate firmware_survival demo ELF is RAM-broken — built for 256KB on a 64KB part — but the tier1 fixture is correct.)

Advanced peripherals — unit-tested only (no firmware drives them): TWIS, SPIS, CCM, RADIO, RNG, TEMP, ECB, GPIOTE, PPI, NVMC, FICR, UICR, NFCT, PDM, EGU0-5

Dead (7 modeled, never exercised): AAR, BPROT, COMP, I2S, LPCOMP, MWU, QDEC

nrf52833

Shares the nRF52 model with nrf52840/nrf52832; tier-1 drives 12 rubric classes (incl. EasyDMA and a real TIMER0 IRQ) and the micro:bit-v2 smoke proves UARTE0. Nothing beyond the rubric is firmware-driven: TWIS/SPIS/CCM/RADIO/NFCT/GPIOTE/TEMP/RNG/ECB/EGU/PDM/NVMC/PPI/USBD/FICR/UICR are unit-tested models, and QDEC/AAR/ACL/COMP/LPCOMP/MWU/I2S have no tests at all.

Advanced peripherals — unit-tested only (no firmware drives them): TWIS, SPIS, CCM, RADIO, GPIOTE, NFCT, TEMP, RNG, ECB, EGU0-5, PDM, NVMC, PPI, USBD, FICR, UICR

Dead (7 modeled, never exercised): QDEC, AAR, ACL, COMP, LPCOMP, MWU, I2S

⚠ ACL is on the part (nRF52833 PS feature list + ch. 6.3; Zephyr nrf52833.dtsi acl@4001e000) but shares the NVMC window the descriptor declares, so no ACL instance is wired.

rp2040

tier1 fixture drives clock/timer/gpio/spi/i2c (PL022/DW_apb); boots unmodified Zephyr. PIO0 is exercised by the required nightly PIO onboarding lane — not by a tier-1 fixture: firmware-rp2040-pio-onboarding boots through this descriptor and must reach PIO_OK (crates/core/tests/rp2040_pio_onboarding.rs:82-135), built and required by the core-full job's LABWIRED_REQUIRE_FIRMWARE (core-ci.yml:2014, :2034, run at :2062; cross-build-excluded from the PR shards).

stm32h563

Drives all 13 rubric classes (incl. FDCAN→can) via real firmware — the broadest rubric coverage. UDS bootloader/ECU examples exist but are not yet gated as automated tests.

Advanced peripherals — unit-tested only (no firmware drives them): UDS bootloader/ECU (0x10/0x27/0x34/0x36/0x37/0x85/0x28 — example firmware, not gated)

stm32h533re

No beyond-rubric peripheral behavior is claimed. The NUCLEO-H533RE profile exercises only USART2 output and GPIOA LED control in its committed smoke.

ch32l103c8t6

No tier-1 fixture or hosted model-verification gate. The PR-gated vendor test runs a committed RV32IMAC ELF that blinks PC13 through GPIOC_CFGHR (negative control: PC0 stays idle), and a second ELF that checks MUL/DIVU/AMOADD.W results and then prints an exact string on USART1 (PA9); a negative control moves the USART1 clock gate and requires silence. A datasheet product-table test pins memory, clock and the 37 LQFP48 GPIOs. This local simulator evidence does not cover WCH-Link, PFIC, SysTick, timers, LPTIM, ADC, I2C, SPI, CAN FD, USB, USB PD or low-power modes, and is not hosted model_verified or hardware confirmation.

ch32v003f4p6

No tier-1 fixture or hosted model-verification gate. The WCH-CH32V003EVT board profile maps its P4 LED1 through a configured jumper to PC0; committed RV32E/ILP32E bare-metal C firmware executes from flash and the PR-gated board smoke requires two GPIO edges. A second RV32EC (compressed) fixture clocks USART1 and prints a line through polled TX on PD5; the board smoke asserts the exact string. A negative control asserts silence when the USART1 clock gate is not set. A third, stock WCH noneos-sdk ELF (PlatformIO ch32v, -march=rv32ecxw, unmodified) runs the vendor startup and SystemInit to 48 MHz, polls the STK SysTick through four Delay_Ms(10) calls and prints exact printf lines on USART1; a negative control without the STK block stops after the boot line. This local simulator evidence does not cover WCH-Link, PFIC, ADC, I2C, SPI, timers, or USB, and is not hosted model_verified or hardware confirmation.

ch32v005f6p6

No tier-1 fixture or hosted model-verification gate. The PR-gated vendor test runs a committed RV32E ELF that blinks PC0 (TSSOP20 pin 10; negative control: PD0 stays idle), and a second RV32EC+Zmmul ELF that checks a MUL result and then prints an exact string on USART1 (PD5); a negative control moves the USART1 clock gate and requires silence. A DS0 test pins memory, clock and the 18 TSSOP20 GPIOs. This local simulator evidence does not cover WCH-Link, PFIC, SysTick, timers, ADC, OPA, I2C, SPI, DMA or USART2, and is not hosted model_verified or hardware confirmation.

ch32x035c8t6

No tier-1 fixture or hosted model-verification gate. The PR-gated vendor test runs a committed RV32IMAC ELF that blinks PB12 through GPIOB_CFGHR (negative control: PB0 stays idle), and a second ELF that checks MUL/DIVU/AMOADD.W results and then prints an exact string on USART1 (PB10); a negative control moves the USART1 clock gate and requires silence. A datasheet test pins memory, clock and the 34 modelled LQFP48 GPIOs (pins 16..23 are not modelled). This local simulator evidence does not cover WCH-Link, PFIC, SysTick, timers, ADC, OPA, CMP, PIOC, I2C, SPI, DMA, USB FS or USB PD, and is not hosted model_verified or hardware confirmation.

ch32v203c8t6

No tier-1 fixture or hosted model-verification gate. The PR-gated vendor test runs a committed RV32IMAC ELF that blinks PC13 through GPIOC_CFGHR (negative control: PC0 stays idle), and a second ELF that checks MUL/DIVU/AMOADD.W results and then prints an exact string on USART1 (PA9); a negative control moves the USART1 clock gate and requires silence. A DS0 Table 2-1 test pins memory, clock and the 37 LQFP48 GPIOs. This local simulator evidence does not cover WCH-Link, PFIC, SysTick, timers, ADC, I2C, SPI, CAN or USB, and is not hosted model_verified or hardware confirmation.

ch32v003f4u6

QFN20 package of the ch32v003f4p6 die. No tier-1 fixture or hosted model-verification gate. The PR-gated vendor test runs the committed F4P6 RV32EC USART1 fixture unchanged and asserts the exact PD5 TX string, and runs the F4P6 blink fixture and requires GPIOC PC0 edges. A DS0 bonding test asserts the descriptor matches F4P6 except for package, pin count and bonded pins. This local simulator evidence does not cover WCH-Link, PFIC, ADC, I2C, SPI, timers, or USB, and is not hosted model_verified or hardware confirmation.

ch32v006k8u6

No tier-1 fixture or hosted model-verification gate. The PR-gated vendor test runs a committed RV32E ELF that blinks PB0 on GPIOB, the port the CH32V003 lacks (negative control: PC0 stays idle), and a second RV32EC+Zmmul ELF that checks a MUL result and then prints an exact string on USART1 (PD5); a negative control moves the USART1 clock gate and requires silence. A DS0 test pins memory, clock and the 31 QFN32 GPIOs. This local simulator evidence does not cover WCH-Link, PFIC, SysTick, timers, ADC, OPA, I2C, SPI, DMA or USART2, and is not hosted model_verified or hardware confirmation.

ch32v003a4m6

SOP16 package of the ch32v003f4p6 die. No tier-1 fixture or hosted model-verification gate. The PR-gated vendor test runs the committed F4P6 RV32EC USART1 fixture unchanged and asserts the exact PD5 TX string, and runs the F4P6 blink fixture and requires GPIOC PC0 edges. A DS0 bonding test asserts the descriptor matches F4P6 except for package, pin count and bonded pins. This local simulator evidence does not cover WCH-Link, PFIC, ADC, I2C, SPI, timers, or USB, and is not hosted model_verified or hardware confirmation.

ch32v003j4m6

SOP8 package of the ch32v003f4p6 die. No tier-1 fixture or hosted model-verification gate. The PR-gated vendor test runs the committed F4P6 RV32EC USART1 fixture unchanged and asserts the exact PD5 TX string. SOP8 has no PC0 pad, so a separate committed fixture blinks PC4 (package pin 7); the test requires PC4 edges and a negative control requires zero PC0 edges. This local simulator evidence does not cover WCH-Link, PFIC, ADC, I2C, SPI, timers, or USB, and is not hosted model_verified or hardware confirmation.

nrf5340

No tier1 fixture. Unmodified Zephyr v3.7 hello_world boots end-to-end through CLOCK/UARTE0/RTC1. TWIM1 is a behavioural EasyDMA model; the PR-gated nrf5340_clock_boot test attaches the declarative BMI270 model and reads CHIP_ID 0x24 with a repeated-START transaction. Nine peripheral blocks remain benign boot stubs.

Functional device/protocol reads (real driver, decoded value): - Unmodified Zephyr v3.7 hello_world boots + prints over UARTE0 EasyDMA — nrf5340-zephyr · firmware_survival::test_nrf5340_zephyr_survival (PR gate)

Advanced peripherals — unit-tested only (no firmware drives them): GPIO, TIMER0-2, RTC0, TWIM1 + BMI270 CHIP_ID EasyDMA transaction

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - DCNF/FPU/CACHE/SPU/OSC_REG/CTRLAP/GPIOTE0/DPPIC/FICR (nrf5340_*_stub) — STRUCTURAL: SystemInit pokes them once at boot and never polls them again, so a register stub is byte-faithful for the hello_world boot path.

nrf54l05

Minimal nRF54L05 application-core smoke using real RRAM and SRAM geometry, UARTE20 EasyDMA and GPIO P2.09. Only a synthetic board is tested; no beyond-rubric firmware function exists. Full Zephyr/SDK boot, BLE radio, TrustZone and silicon parity are untested.

nrf54l10

Minimal nRF54L10 application-core smoke using real RRAM and SRAM geometry, UARTE20 EasyDMA and GPIO P2.09. Only a synthetic board is tested; no beyond-rubric firmware function exists. Full Zephyr/SDK boot, BLE radio, TrustZone and silicon parity are untested.

nrf54l15

No tier1 fixture. Unmodified Zephyr v4.4 hello_world boots end-to-end (nRF54L CLOCK/LFCLK, GRTC kernel tick, TAMPC approtect gate, RRAM-mapped UARTE20 EasyDMA) and the PR-gated smart-ring probe reads four I2C sensor IDs. Beyond the rubric TEMP/EGU/GPIOTE are unit-tested models no firmware drives; TAMPC/RRAMC/regulators/FICR/UICR/DPPIC are zero-filled stub windows the boot path pokes but never polls.

Advanced peripherals — unit-tested only (no firmware drives them): TEMP, EGU10, EGU20, GPIOTE20, GPIOTE30

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - TAMPC (configs/chips/nrf54l15.yaml:363 (nrf54l_tampc_stub)) — STRUCTURAL: SystemInit READS the protect-domain signal registers, and the MDK deliberately hangs the part on locked+high — zero is the only boot-safe answer. Read once, never polled. - RRAMC (configs/chips/nrf54l15.yaml:390 (nrf54l_rramc_stub)) — STRUCTURAL: firmware executes from RRAM and no boot path polls the controller for a settled value. Closable only by firmware that writes NVM at runtime — none does. - REGULATORS (configs/chips/nrf54l15.yaml:398 (nrf54l_regulators_stub)) — STRUCTURAL: DCDC/LDO selection is write-only on the boot path; nothing reads it back. - FICR (configs/chips/nrf54l15.yaml:423 (nrf54l_ficr_stub)) — STRUCTURAL: factory info read once at boot; zero is the honest answer for an unprovisioned part. - UICR (configs/chips/nrf54l15.yaml:429 (nrf54l_uicr_stub)) — STRUCTURAL: unprovisioned UICR reads zero, which the boot path accepts; APPROTECT provisioning is not modelled. - DPPIC20/DPPIC30 (configs/chips/nrf54l15.yaml:407,413 (nrf54l_dppic_stub)) — CLOSABLE: event-routing fabric only — poked during peripheral init, never polled for a settled value, and no firmware routes events through DPPI.

nrf54lm20a

No tier1 fixture. The strict-onboarding snake lab is the only firmware: SPIM22 drives an RM67162 AMOLED and the gate asserts rendered frame ink (a rubric class, so unrecorded here). Beyond the rubric TEMP/EGU/GPIOTE are unit-tested models no firmware drives; TAMPC/RRAMC/regulators/DPPIC20/DPPIC30 are zero-filled stub windows.

Advanced peripherals — unit-tested only (no firmware drives them): TEMP, EGU10, EGU20, GPIOTE20, GPIOTE30

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - TAMPC (configs/chips/nrf54lm20a.yaml:329 (nrf54l_tampc_stub)) — STRUCTURAL: SystemInit READS the protect-domain signal registers, and the MDK deliberately hangs the part on locked+high — zero is the only boot-safe answer. Read once, never polled. - RRAMC (configs/chips/nrf54lm20a.yaml:351 (nrf54l_rramc_stub)) — STRUCTURAL: firmware executes from RRAM and no boot path polls the controller for a settled value. Closable only by firmware that writes NVM at runtime — none does. - REGULATORS (configs/chips/nrf54lm20a.yaml:358 (nrf54l_regulators_stub)) — STRUCTURAL: DCDC/LDO selection is write-only on the boot path; nothing reads it back. - DPPIC20/DPPIC30 (configs/chips/nrf54lm20a.yaml:365,369 (nrf54l_dppic_stub)) — CLOSABLE: event-routing fabric only — poked during peripheral init, never polled for a settled value, and no firmware routes events through DPPI.

esp32c6

Richest Espressif rubric coverage: the tier-1 fixture drives all 12 of its declared classes. clock and irq are real engines, not register stand-ins — the native PCR enforces CLK_EN (gated UART0 reads back 0, then recovers) and interrupt_core0+intpri route both the software doorbell and UART0's peripheral source into real mcause traps. Beyond the rubric only io_mux and hp_sys remain — SVD-derived declarative register files with no behavioural engine.

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - io_mux (configs/chips/esp32c6.yaml:491 (declarative; configs/peripherals/esp32c6/io_mux.yaml)) — CLOSABLE: reset-value register storage; pad function writes are recorded but never read back or electrically enforced — the DevKitC demo writes U0TXD's MCU_SEL and the tier-1 console both pass regardless of routing. The S3's real io_mux model is the port path. - hp_sys (configs/chips/esp32c6.yaml:516 (declarative; configs/peripherals/esp32c6/hp_sys.yaml)) — CLOSABLE: declarative register file with no engine; nothing drives it today, so the false-success path is latent — timeout monitor, SDIO control, ROM-table lock and memory test are not modelled.

esp32h2

No beyond-rubric H2 peripheral is exercised: the board smoke covers only the UART0 FIFO/console path. This narrow profile does not model mask ROM, the USB bridge, IO_MUX, WS2812/RMT, button/strap behavior, Bluetooth LE, IEEE 802.15.4, Thread, Zigbee, or the other H2 peripherals.

esp32c5

The C5 DevKitC-1-N8R8 smoke reads a BME280 ID and raw temperature over the C5 I2C0 command-list engine, observes GPIO5 output edges, and reports a UART0 marker. The profile does not model ROM boot, flash/XIP, PSRAM, wireless, the RGB LED, PCR clock gating, or electrical IO_MUX enforcement.

Functional device/protocol reads (real driver, decoded value): - BME280 ID and raw-temperature read over GPIO8/9-routed C5 I2C0 — esp32c5-devkitc-1-bme280-smoke · examples/esp32c5-devkitc-1-bme280-smoke/evidence/logs/run_smoke.log (PR gate)

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - io_mux (configs/chips/esp32c5.yaml (declarative; configs/peripherals/esp32c6/io_mux.yaml)) — CLOSABLE: register words are stored, but pad electrical mux behavior is not enforced; the smoke's GPIO output and I2C route use the modeled GPIO/matrix subset directly.

esp32c61

The DevKitC-1-N8R2 profile exercises the UART0 TX FIFO, GPIO4 output latch, and the C61 GPIO-matrix I2C0 route to a BME280 on GPIO6/GPIO22. Firmware reads the sensor ID and a raw-temperature vector. It does not model mask-ROM boot, the physical USB bridge, electrical IO_MUX, other GPIO-matrix functions, GPIO26..29, GPIO interrupts, the RGB LED, Wi-Fi 6, Bluetooth LE, PSRAM, or sensor conversion timing.

Functional device/protocol reads (real driver, decoded value): - BME280 ID and raw-temperature read over GPIO6/22-routed C61 I2C0 — esp32c61-bme280-smoke · examples/esp32c61-devkitc-1-n8r2-bme280/evidence/run/logs/run_smoke.log (PR gate)

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - io_mux (configs/chips/esp32c61.yaml (declarative; configs/peripherals/esp32c6/io_mux.yaml)) — Register storage only; electrical pad muxing is not enforced.

esp32p4

The P4X-C5 Function EV board smoke checks BME280 chip ID 0x60 and the model's raw-temperature vector 0x082DC0 over HP I2C0 routed through GPIO7/8, then reports over UART0. The profile models the GPIO data/enable/input head, four matrix selectors, I2C0 transaction registers, and IO_MUX setup words without pad electricals. It does not model mask ROM, the second HP core, the LP core, MIPI-CSI/DSI, the H.264/JPEG/PPA media blocks, USB, Ethernet, the companion ESP32-C5 radio or the other P4 peripherals.

Functional device/protocol checks (register-level firmware asserts observable behavior): - BME280 ID and raw-temperature register read through GPIO7/8-routed HP I2C0 (register-level firmware) — firmware-esp32p4-demo · examples/esp32p4x-c5-function-ev-board/evidence/run/logs/run_smoke.log (ungated)

esp32s2-arduino

Arduino matrix profile of the ESP32-S2 on its genuine mask ROM: stock Arduino-ESP32 runs L0-L8 (UART0, SYSTIMER, RMT WS2812, I2C0, GP-SPI2, SENS ADC, LEDC, TWAI). Those are rubric classes, so no beyond-rubric firmware exercise is recorded in this ledger. Wi-Fi, USB OTG, PSRAM and the ULP are not modelled.

esp32s2

No beyond-rubric S2 peripheral is exercised: the DevKitC-1-N8R2 board smoke covers only the UART0 TX FIFO through host_console. This direct-ELF profile declares UART0 only; it does not model mask ROM, flash-cache/MMU boot, PSRAM, the GPIO matrix, the RGB LED, USB OTG, Wi-Fi or the other S2 peripherals.

rp2040-feather

Board variant of the rp2040 die with the Feather's 133 MHz clock and 8 MiB flash; it reuses the rp2040 behavioural models. No beyond-rubric peripheral is exercised under this name: the Feather smoke reads BME280 CHIP_ID over I2C1 and drives D13 through SIO; both are rubric classes, not beyond-rubric content. PIO, USB and the onboard NeoPixel are not exercised on this board.

rp2350-feather

Board variant of the rp2350 die with the Feather's 8 MiB flash; it reuses the rp2350 behavioural models. No beyond-rubric peripheral is exercised under this name: the Feather smoke toggles red D13 (GPIO7) through SIO and writes UART0; both are rubric classes, not beyond-rubric content. PIO, USB, PSRAM and the onboard NeoPixel are not exercised on this board.

esp32c6fh4

Package variant of the esp32c6 die for the Seeed XIAO ESP32-C6: the yaml copies esp32c6.yaml and changes only the flash size to 4 MB, so it reuses the esp32c6 models and shims. No beyond-rubric peripheral is exercised under this name: the XIAO smoke toggles the active-low user LED on GPIO15 through the C6 GPIO block, a rubric class. The radio, RF switch, USB-Serial/JTAG console and LP core are not exercised on this board.

stm32h723zg

Two firmware paths drive this die. The NUCLEO-H723ZG Rust smoke (core-onboarding-smoke) drives RCC enables, GPIOB (LD1 on PB0), GPIOD AF7 and USART3 TX. The stock stm32duino NUCLEO_H723ZG row in the Arduino matrix (core-arduino-matrix-smoke, nightly + push to main; evidence validation/arduino-matrix/evidence/nucleo-h723zg) drives PWR supply/VOS bring-up, the PLL clock tree, CRS (HAL_RCCEx_CRSConfig), USART3, GPIO, I2C1 (INA219 exact registers), SPI1 (MAX31855 exact frame), ADC1 plus the ADC12 common CCR (analogRead of A0 = PA3 held at 1651 mV reads exactly 512), TIM2 PWM on PA0 (1.000 kHz, 50.2 % duty on the pad) and SysTick (micros). FDCAN1/2, CRC, RNG, IWDG, WWDG, DBGMCU, USART1/2/6, UART4, LPUART1, I2C2/3, SPI2/3, TIM1/3/4/8 and ADC2 are declared with the H735 instance list but no H723 firmware drives them; the FDCAN message RAM is not mapped, so the L8 CAN cell is a declared skip.

stm32h753zi

The stock stm32duino NUCLEO_H753ZI row in the Arduino matrix (core-arduino-matrix-smoke, nightly + push to main; evidence validation/arduino-matrix/evidence/nucleo-h753zi) drives PWR LDO supply and VOS0 bring-up, the PLL clock tree (480 MHz), CRS (HAL_RCCEx_CRSConfig), USART3, GPIO, I2C1 (INA219 exact registers), SPI1 (MAX31855 exact frame), ADC1 plus the ADC12 common CCR (analogRead of A0 = PA3 held at 1651 mV reads exactly 512), TIM2 PWM on PA0 (1.000 kHz, 50.2 % duty on the pad) and SysTick (micros). FDCAN1/2, CRC, RNG, IWDG, WWDG, DBGMCU, USART1/2/6, UART4, LPUART1, I2C2/3, SPI2/3, TIM1/3/4/8 and ADC2 are declared but no H753 firmware drives them. CRYP, HASH, Ethernet, USB OTG, QUADSPI, FMC, ADC3 and DMA are not modelled; the FDCAN message RAM is not mapped, so the L8 CAN cell is a declared skip.

esp32s3-zero

Board variant of the esp32s3 die: same configure_xtensa_esp32s3 wiring and the same committed tier-1 ELF, which non-ignored CLI/workspace tests boot under this chip's name and assert through TIER1 ... PASS (SYSTIMER runs the clock check; USB-Serial/JTAG is the console of the TMP102 e2e, so neither is a gap). The yaml's IRAM/ROM-thunk/XIP windows are memory-map declarations, not per-chip models, and its eFuse/RTC_CNTL stubs are the shared family stubs listed on the esp32s3 entry. spi_mem_flash/extmem stay unit-only here too: mask-ROM bring-up reads are not firmware exercise.

Advanced peripherals — unit-tested only (no firmware drives them): aes, ds, hmac, rsa, rng, sha, i2s, lcd_cam, sdmmc, usb_otg, gpspi, pcnt, io_mux, extmem, spi_mem_flash, sens, system, core1_control, crosscore_ipi

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - efuse_stub (crates/core/src/system/xtensa/esp32s3.rs:647 (esp_xtensa_common/system_stub.rs:454; configs/chips/esp32s3-zero.yaml:72)) — CLOSABLE: canned MAC + chip-rev only; no eFuse state, no burn path. The ROM boot and esp-hal accept the canned values. - rtc_cntl_stub (crates/core/src/system/xtensa/esp32s3.rs:640 (esp_xtensa_common/system_stub.rs:203; configs/chips/esp32s3-zero.yaml:66)) — CLOSABLE: register round-trip plus three canned behaviours — PLL_LOCK seeded, TIME_UPDATE snapshot handshake, APP_CPU un-stall. The RTC clock tree, SWD_CONF super-watchdog and sleep domains are not modelled. - wifi_thunks (esp32s3/wifi_thunks.rs:8 (CHEAT(THUNK-LIB))) — IRREDUCIBLE: the ESP32 WiFi MAC/PHY is a closed RF-coprocessor blob with NO executable image — there is no firmware to run, so it can never be firmware-exercised. The lwIP/socket layer above is routed to a real SimNet. (Same shim as the esp32s3 entry.)

esp32s3-devkitc-1-n16r8-adc

Narrow DevKitC-1 v1.1 N16R8 board profile over the esp32s3 die, with no beyond-rubric content of its own: its register-level firmware starts one SENS ADC1 one-shot conversion on GPIO4 / ADC1_CH3 and asserts code 2047 for the centered 1650 mV input (the ADC rubric cell and the family shims stay as listed on the esp32s3 entry). The local analog model is deterministic; no noise, calibration or attenuation curve is modelled.

stm32f103

Drives 10 rubric classes via its tier-1 fixture, and the nightly J1939 monitor drives bxCAN1 end to end (per-SA BAM reassembly + engine-speed decode); AFIO and DBGMCU are exercised by the PR-gated Zephyr hello (pinctrl remap, debug-init) and PWR by the Arduino startup; beyond the rubric, the shared CRC engine's always-run coverage is the H563 bench oracle (h563_conformance.rs:144) plus IDR-width masking (crc.rs:204), and the F103 polynomial is additionally pinned by the sim-only thumb oracle (stm32f1_exec_oracle.rs:300) and the conformance digest, which CI skips without its prebuilt ELF — while WWDG is an inert shim rather than a dead model, and the USB-device and BKP windows also resolve to zero stubs.

Functional device/protocol reads (real driver, decoded value): - J1939 engine-bus monitor — per-SA BAM transport reassembly and engine-speed decode from a replayed CAN capture (ENGINE idle_rpm=600; 9 DM1 source addresses) — f103-j1939-monitor · examples/f103-j1939-monitor/j1939-replay.yaml (nightly coverage-matrix required target, .github/workflows/core-coverage-matrix-smoke.yml:171) (nightly CI)

Advanced peripherals — unit-tested only (no firmware drives them): CRC

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - usb_dev (configs/chips/stm32f103.yaml:216 (type: stub; configs/peripherals/stm32f103/usb.yaml)) — CLOSABLE: author-declared stub — writes are dropped and every read returns 0, so the F1 USB-FS device engine (EPnR/CNTR/ISTR/BTABLE/PMA) is not modelled; no gated firmware opens the window, and the declarative usb.yaml is decode-only. - bkp (configs/chips/stm32f103.yaml:222 (type: stub)) — CLOSABLE: author-declared stub with no schema wired in the descriptor — the backup-domain registers (RTC backup data and calibration) are not modelled; no gated firmware touches the window. - WWDG (crates/core/src/peripherals/wwdg.rs:63 (configs/chips/stm32f103.yaml:185)) — CLOSABLE: inert register bank — tick() is the trait-default no-op, so an armed watchdog never fires. No behavioural test and no gated firmware drives it; reset values are pinned by the bench-F103 probe in stm32f1_mmio_diff.

stm32f103c4

Exact low-density x4/x6 geometry slice for stm32f103c4; it reuses the shared medium-density STM32F103 peripheral map, so firmware-exercise claims stay under stm32f103. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f103c6

Exact low-density x4/x6 geometry and peripheral windows for stm32f103c6 (SPI1, I2C1, USART1/2, TIM1-3, ADC1/2, DMA1; no SPI2, I2C2, USART3 or TIM4). A dedicated smoke boots firmware, writes USART2 and toggles PA5; USB and BKP remain explicit untouched stubs. ES0348's user-mode DBGMCU_IDCODE restriction is reflected as a zero read; stm32_variant_geometry asserts exact geometry. This is model evidence, not a physical C6 capture.

stm32f103r4

Exact low-density x4/x6 geometry slice for stm32f103r4; it reuses the shared medium-density STM32F103 peripheral map, so firmware-exercise claims stay under stm32f103. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f103r4h

Exact low-density x4/x6 geometry slice for stm32f103r4h; it reuses the shared medium-density STM32F103 peripheral map, so firmware-exercise claims stay under stm32f103. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f103r6

Exact low-density x4/x6 geometry for stm32f103r6 (32 KB flash, 10 KB SRAM, LQFP64, 51 GPIOs, 16 ADC channels), with only the low-density SPI1, I2C1, USART1/2, TIM1-3, ADC1/2 and DMA1 instances exposed. Its dedicated smoke runs USART2 output and PA5 edges; USB and backup-register windows remain explicit untouched stubs. Exact geometry is asserted by stm32_variant_geometry. This is local simulator evidence, not a physical R6 capture.

stm32f103r6h

Exact low-density x4/x6 geometry slice for stm32f103r6h; it reuses the shared medium-density STM32F103 peripheral map, so firmware-exercise claims stay under stm32f103. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f103r8

Exact STM32F103R8 LQFP64 model (medium-density x8: 64 KiB flash, 20 KiB SRAM, 51 GPIOs, 16 ADC channels). Same die and peripheral map as the public stm32f103 (C8) engine, so firmware-exercise rubric claims stay on the STM32F103 family row. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry, and the exact part is executed by the stm32-exact-catalog arm-spin cell in core-coverage-matrix-smoke.yml.

stm32f103rb

Exact STM32F103RBT6 LQFP64 model with 128 KiB flash and 20 KiB SRAM. The NUCLEO-F103RB source smoke exercises PA5 GPIO output and USART2; firmware-exercise rubric claims remain on the STM32F103 family row until exact-part rubric coverage is added.

stm32f103rb-nucleo-bme280-smoke

Narrow NUCLEO-F103RB + BME280 onboarding profile (USART2/ST-LINK VCP, LD2 on PA5, I2C1 remapped to PB8/PB9). Committed evidence at examples/nucleo-f103rb-bme280/evidence shows examples/nucleo-f103rb-bme280/onboarding-smoke.yaml passed 3/3: USART2 printed NUCLEO_F103RB_BME280_OK ID=0x60 I2C1=PB8/PB9 after a firmware read of BME280.CHIP_ID (0xD0) at 0x77, gpioa:5 (LD2) saw at least 32 edges, and the run stopped at max_steps. Firmware is built from crates/firmware-nucleo-f103rb-bme280-smoke, not a committed ELF; no required CI lane runs this script. No beyond-rubric block is listed for this slice.

stm32f103t4

Exact low-density x4/x6 geometry slice for stm32f103t4; it reuses the shared medium-density STM32F103 peripheral map, so firmware-exercise claims stay under stm32f103. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f103t6

Exact low-density x4/x6 geometry slice for stm32f103t6; it reuses the shared medium-density STM32F103 peripheral map, so firmware-exercise claims stay under stm32f103. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f401

All 12 declared rubric classes are driven by its tier-1 fixture — including a byte-exact memory-to-memory transfer on DMA2 plus the EXTI→NVIC and TIM1-PWM checks — while PWR_CR.VOS is programmed by the PR-gated Arduino startup and DBGMCU_CR by the Zephyr SoC-debug init; TIM2 input capture is driven by two register-level firmwares (HC-SR04 echo width through the CC2 interrupt, and PWM-input period/duty in slave reset mode); that leaves the other stream controller, DMA1, as the chip's one beyond-rubric gap.

Functional device/protocol reads (real driver, decoded value): - TIM2 input capture — HC-SR04 echo width (CH1 rising + CH2 indirect falling, CC2 interrupt) within one 1 µs tick of distance × 58 µs, 2-400 cm — timer-capture/echo-capture · timer_input_capture.rs:hc_sr04_echo_width_is_measured_by_tim2_input_capture (ungated) - TIM2 PWM-input mode (TS=TI1FP1, SMS=reset) — CCR1 period and CCR2 high time equal to the injected edge cycles exactly — timer-capture/freq-meter · timer_input_capture.rs:pwm_input_mode_measures_period_and_duty_exactly (ungated)

Advanced peripherals — unit-tested only (no firmware drives them): DMA1

stm32f401cb

Exact orderable geometry slice for stm32f401cb on the shared F401 peripheral engine; firmware-exercise claims stay under the family entry. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f401cc

Exact orderable geometry slice for stm32f401cc on the shared F401 peripheral engine; firmware-exercise claims stay under the family entry. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f401cd

Exact orderable geometry slice for stm32f401cd on the shared F401 peripheral engine; firmware-exercise claims stay under the family entry. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f401ce

Exact orderable geometry slice for stm32f401ce on the shared F401 peripheral engine; firmware-exercise claims stay under the family entry. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f401rb

Exact orderable geometry slice for stm32f401rb on the shared F401 peripheral engine; firmware-exercise claims stay under the family entry. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f401rc

Exact orderable geometry slice for stm32f401rc on the shared F401 peripheral engine; firmware-exercise claims stay under the family entry. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f401rd

Exact orderable geometry slice for stm32f401rd on the shared F401 peripheral engine; firmware-exercise claims stay under the family entry. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f401re

Exact orderable geometry slice for stm32f401re on the shared F401 peripheral engine; firmware-exercise claims stay under the family entry. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f401vb

Exact orderable geometry slice for stm32f401vb on the shared F401 peripheral engine; firmware-exercise claims stay under the family entry. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f401vc

Exact orderable geometry slice for stm32f401vc on the shared F401 peripheral engine; firmware-exercise claims stay under the family entry. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f401vd

Exact orderable geometry slice for stm32f401vd on the shared F401 peripheral engine; firmware-exercise claims stay under the family entry. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f401ve

Exact orderable geometry slice for stm32f401ve on the shared F401 peripheral engine; firmware-exercise claims stay under the family entry. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f405

No tier-1 target (absent from tier1-matrix.json), so this entry carries the whole story: the committed part-specific fixture runs nightly via examples/feather-f405/tier1-smoke.yaml and PR-gated firmware_survival::test_stm32f405_tier1_survival, driving clock/gpio/timer/i2c/spi/adc/wdt/rtc over USART2 and explicitly leaving dma/irq/pwm unclaimed; PWR and both DMA controllers are modeled but firmware-untouched, and DBGMCU is never exercised at all.

Advanced peripherals — unit-tested only (no firmware drives them): PWR, DMA1, DMA2

Dead (1 modeled, never exercised): DBGMCU

stm32f407

Its 9-class tier-1 fixture leaves dma/irq/pwm unrecorded, and the DMA gap is real: no firmware drives either stream controller (only the descriptor-level differential does); the PR-gated smoke reads DBGMCU IDCODE, the Arduino startup programs PWR_CR.VOS then reads REV_ID back, and the arduino-matrix-gate L8_can loopback sketch drives bxCAN1 to self-reception (ID 0x123, data 0xA5), so debug, power and CAN are firmware-exercised; the AHT20+BMP280 e2e drives two real I²C device twins but decodes no physical value.

Advanced peripherals — unit-tested only (no firmware drives them): DMA1, DMA2

stm32f407ie

Exact orderable geometry slice for stm32f407ie on the shared F407 peripheral engine; firmware-exercise claims stay under the family entry. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f407ig

Exact orderable geometry slice for stm32f407ig on the shared F407 peripheral engine; firmware-exercise claims stay under the family entry. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f407ve

Exact orderable geometry slice for stm32f407ve on the shared F407 peripheral engine; firmware-exercise claims stay under the family entry. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f407vg

Exact orderable geometry slice for stm32f407vg on the shared F407 peripheral engine; firmware-exercise claims stay under the family entry. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f407ze

Exact orderable geometry slice for stm32f407ze on the shared F407 peripheral engine; firmware-exercise claims stay under the family entry. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f407zg

Exact orderable geometry slice for stm32f407zg on the shared F407 peripheral engine; firmware-exercise claims stay under the family entry. Exact memory, package, GPIO, ADC and order-code geometry is asserted by stm32_variant_geometry.

stm32f401cdu6

384 KiB Black-Pill package variant of the F401. Its PR-gated firmware is the UART-only black-pill demo (firmware_survival::test_stm32f401cdu6_demo_survival, OK over USART2); the install-canary lane runs the sibling stm32f401 tier-1 ELF against this descriptor but asserts only TIER1 clock PASS, because this descriptor declares no RCC clock gates and zero-window stubs for RTC/IWDG/WWDG/DMA1-2/CRC/SDIO/SYSCFG/I2S extensions/OTG FS/DBG — the sibling fixture's gate-proving timer/spi/adc/pwm checks, its wdt/rtc checks and its byte-exact DMA2 transfer all fail here (a dev run prints clock/gpio/i2c/irq PASS and seven FAIL lines, dma-ndtr among them). PWR is a real stm32f4-profile model that no firmware on this descriptor drives.

Advanced peripherals — unit-tested only (no firmware drives them): PWR

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - RTC (configs/chips/stm32f401cdu6.yaml:66 (type: stub; configs/peripherals/stm32f401/rtc.yaml)) — CLOSABLE: zero window — the calendar RTC the sibling stm32f401 descriptor wires for real (configs/chips/stm32f401.yaml:164) is absent here; the black-pill demo never opens it, and the sibling fixture's DR-reset check fails (rtc-dr-reset). - WWDG (configs/chips/stm32f401cdu6.yaml:72 (type: stub; configs/peripherals/stm32f401/wwdg.yaml)) — CLOSABLE: zero window — not even the inert wwdg register bank the F103/L476 descriptors carry; no gated firmware opens it. - IWDG (configs/chips/stm32f401cdu6.yaml:78 (type: stub; configs/peripherals/stm32f401/iwdg.yaml)) — CLOSABLE: zero window — the LSI-clocked engine the sibling descriptor wires (configs/chips/stm32f401.yaml:155) is absent, so the sibling fixture's KR-unlock/RLR round-trip fails (wdt-unprotected). - I2S2EXT/I2S3EXT (configs/chips/stm32f401cdu6.yaml:84,104 (type: stub; configs/peripherals/stm32f401/i2s2ext.yaml, i2s3ext.yaml)) — CLOSABLE: zero windows for the I2S2/I2S3 extension blocks; no STM32 I2S model exists (the engine's I2S models are the nRF52 and ESP32-S3 ones) and no gated firmware opens them. - SDIO (configs/chips/stm32f401cdu6.yaml:174 (type: stub; configs/peripherals/stm32f401/sdio.yaml)) — CLOSABLE: zero window; the engine carries a real STM32 SDMMC host (sdmmc.rs, wired on the L476) but not this earlier F4 SDIO controller, and no firmware opens the window. - SYSCFG (configs/chips/stm32f401cdu6.yaml:195 (type: stub; configs/peripherals/stm32f401/syscfg.yaml)) — CLOSABLE: zero window — EXTI source-select (EXTICR) is not modelled; the fixture triggers its EXTI line through SWIER, so nothing reads it back. - CRC (configs/chips/stm32f401cdu6.yaml:232 (type: stub; configs/peripherals/stm32f401/crc.yaml)) — CLOSABLE: zero window — the engine's real CRC model (crc.rs) is not wired on this descriptor; no firmware drives it. - DMA1/DMA2 (configs/chips/stm32f401cdu6.yaml:248,254 (type: stub; the real engine is stm32f4_dma, used by configs/chips/stm32f401.yaml:216,226)) — CLOSABLE: both stream controllers are zero windows, so the sibling fixture's byte-exact DMA2 memcpy fails here (dma-ndtr) — the stm32f401 row's dma pass belongs to that descriptor's real DMA2, not this window. Port path: declare type: stm32f4_dma with the AHB1ENR gates and stream_irqs as the F401/F405/F407/F767 descriptors do. - OTG_FS global/host/device/pwrclk (configs/chips/stm32f401cdu6.yaml:309,315,321,327 (type: stub; configs/peripherals/stm32f401/otg_fs_.yaml)) — CLOSABLE: four zero windows — the engine's STM32 OTG FS/DWC2 model (usb_otg.rs, wired on the L476) is not wired here (the F103 USB-device window is the same stub class), no gated firmware opens them, and the declarative otg_fs_.yaml schemas are decode-only. - DBG (configs/chips/stm32f401cdu6.yaml:336 (type: stub; configs/peripherals/stm32f401/dbg.yaml)) — CLOSABLE: zero window with no IDCODE wired even though the yaml header records the CDU6's bench-read DBGMCU IDCODE 0x10016433 — the sibling descriptor's real dbgmcu model (configs/chips/stm32f401.yaml:187, idcode 0x10006411) is not reused; no gated firmware reads it.

stm32f446re

Minimal real-register firmware smoke: reset/stack, RCC, GPIO PA5 LED and USART2 UART banner. No beyond-rubric function is exercised; the model deliberately omits F446-exclusive blocks, and no stock HAL/Zephyr or physical silicon run exists.

Dead (2 modeled, never exercised): PWR, FLASH_CTRL

stm32f411re

NUCLEO-F411RE (same die and register set as stm32f411ceu6, LQFP64). PR-gated raw-register smoke (examples/nucleo-f411re/io-smoke.yaml): reset/stack, RCC, GPIOA, USART2 banner and PA5 LED edges. Stock stm32duino also runs the Arduino matrix row L0-L7 (INA219 over I2C1 at the exact tier, MAX31855 over SPI1, ADC1_IN0 at a board-held 1650 mV read at 10 bits = 511, TIM2_CH1 PWM on PA5 at 1.000 kHz), but that lane runs nightly and on main, not on PRs, so nothing is recorded as functional here. No physical silicon run exists.

Advanced peripherals — unit-tested only (no firmware drives them): PWR

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - WWDG/I2S2EXT/I2S3EXT/SDIO/CRC/DMA1/DMA2/OTG_FS/DBG (configs/chips/stm32f411re.yaml (type: stub; same windows as stm32f411ceu6)) — CLOSABLE: the same zero windows as the Black Pill descriptor; see the stm32f411ceu6 row for each one.

stm32l432kc

NUCLEO-L432KC. PR-gated raw-register smoke (examples/nucleo-l432kc/io-smoke.yaml): reset/stack, RCC, GPIOA/GPIOB, USART2 banner and PB3 LED edges. Stock stm32duino also runs the Arduino matrix row L0-L8 (INA219 over I2C1, MAX31855 over SPI1, ADC1_IN5 at a board-held 1650 mV, TIM2_CH2 PWM on PB3, bxCAN1 loopback), but that lane runs nightly and on main, not on PRs, so nothing is recorded as functional here. No physical silicon run exists.

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - SYSCFG (configs/chips/stm32l432kc.yaml (type: stub; configs/peripherals/stm32l476/syscfg.yaml)) — CLOSABLE: author-declared stub shared with the L476 — read 0, writes dropped, so EXTI source-select is not modelled.

stm32f411ceu6

WeAct Black Pill. Its tier-1 fixture (PR-gated firmware_survival::test_stm32f411_tier1_survival; nightly coverage-matrix cell examples/stm32f411ceu6-blackpill/io-smoke.yaml) drives clock/gpio/timer/i2c/spi (SPI1+SPI5)/adc/wdt/rtc over USART2, leaving dma unrecorded (the fixture never attempts it, and both stream controllers are zero stubs) and irq/pwm n/a (the fixture does not attempt either class). PWR is a real stm32f4-profile model no firmware drives; the descriptor's remaining stub windows are listed below.

Advanced peripherals — unit-tested only (no firmware drives them): PWR

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - WWDG (configs/chips/stm32f411ceu6.yaml:112 (type: stub; configs/peripherals/stm32f411/wwdg.yaml)) — CLOSABLE: zero window — not even the inert wwdg register bank the F103/L476 descriptors carry; no gated firmware opens it. - I2S2EXT/I2S3EXT (configs/chips/stm32f411ceu6.yaml:127,147 (type: stub; configs/peripherals/stm32f411/i2s2ext.yaml, i2s3ext.yaml)) — CLOSABLE: zero windows for the I2S2/I2S3 extension blocks; no STM32 I2S model exists (the engine's I2S models are the nRF52 and ESP32-S3 ones) and no gated firmware opens them. - SDIO (configs/chips/stm32f411ceu6.yaml:220 (type: stub; configs/peripherals/stm32f411/sdio.yaml)) — CLOSABLE: zero window; the engine carries a real STM32 SDMMC host (sdmmc.rs, wired on the L476) but not this earlier F4 SDIO controller, and no firmware opens the window. - SYSCFG (configs/chips/stm32f411ceu6.yaml:247 (type: stub; configs/peripherals/stm32f411/syscfg.yaml)) — CLOSABLE: zero window — EXTI source-select (EXTICR) is not modelled, and no gated firmware configures it. - CRC (configs/chips/stm32f411ceu6.yaml:303 (type: stub; configs/peripherals/stm32f411/crc.yaml)) — CLOSABLE: zero window — the engine's real CRC model (crc.rs) is not wired on this descriptor; no firmware drives it. - DMA1/DMA2 (configs/chips/stm32f411ceu6.yaml:319,325 (type: stub; the real engine is stm32f4_dma, wired by the sibling F401/F405/F407/F767 descriptors)) — CLOSABLE: both stream controllers are zero windows, which is why the fixture's dma cell is unrecorded; the yaml comment at :33 still says only the F1/L4 channel layout is modelled, but the stream-controller engine stm32f4_dma now exists — port path is to wire it with the AHB1ENR gates and stream_irqs as the F401 descriptor does. - OTG_FS global/host/device/pwrclk (configs/chips/stm32f411ceu6.yaml:380,386,392,398 (type: stub; configs/peripherals/stm32f411/otg_fs_.yaml)) — CLOSABLE: four zero windows — the engine's STM32 OTG FS/DWC2 model (usb_otg.rs, wired on the L476) is not wired here (the F103 USB-device window is the same stub class), no gated firmware opens them, and the declarative otg_fs_.yaml schemas are decode-only. - DBG (configs/chips/stm32f411ceu6.yaml:410 (type: stub; configs/peripherals/stm32f411/dbg.yaml)) — CLOSABLE: zero window and no IDCODE by design — the descriptor says the F411 DEV_ID/REV_ID has not been read off a part and inventing one would make a probe-identify path silently wrong; no gated firmware reads it.

stm32f746

Sim-derived F7 (no bench part) and no tier-1 target — absent from TIER1_TARGETS and tier1-matrix.json: the only firmware that runs is the bare-metal Discovery smoke, PR-gated as firmware_survival::test_stm32f746_discovery_smoke_survival and rebuilt by the nightly coverage-matrix cell (examples/stm32f7-discovery/uart-smoke.yaml, crate firmware-stm32f746-demo). It un-gates GPIOIEN/USART1EN, configures PI1, prints OK over USART1 and sets PI1 via BSRR; beyond that the descriptor adds only five zero-window stubs to the F4-shared RCC/GPIO/systick set. FMC/SDRAM, SAI, SDMMC, DCMI, RNG and GPIOJ/K are unmapped holes, not stubs.

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - LTDC (configs/chips/stm32f746.yaml:151 (type: stub; no schema)) — CLOSABLE: zero window — the display controller is not modelled and the smoke never opens it. - Ethernet (configs/chips/stm32f746.yaml:155 (type: stub; no schema)) — CLOSABLE: zero window; no Ethernet model exists in the engine and no firmware opens it. - DMA2D (configs/chips/stm32f746.yaml:159 (type: stub; no schema)) — CLOSABLE: zero window; the 2D blitter is not modelled and no firmware opens it. - USB OTG FS (configs/chips/stm32f746.yaml:163 (type: stub; no schema)) — CLOSABLE: zero window for the OTG FS core — the engine's STM32 OTG FS/DWC2 model (usb_otg.rs, wired on the L476) is not wired here (the F103 window is the same stub class) and the smoke never opens it. - QUADSPI (configs/chips/stm32f746.yaml:167 (type: stub; no schema)) — CLOSABLE: zero window; the engine's STM32 QUADSPI model (quadspi.rs, wired on the L476) is not wired here and no firmware drives the window.

stm32f767

Not a tier-1 target (absent from TIER1_TARGETS and tier1-matrix.json), so this entry carries the whole story: its committed part-specific fixture runs PR-gated as firmware_survival::test_stm32f767_tier1_survival and in the nightly coverage-matrix cell examples/nucleo-f767zi/tier1-smoke.yaml, driving clock/gpio/timer/i2c/spi/adc/wdt/rtc over USART2 and leaving dma/irq/pwm unclaimed — the fixture never attempts them even though the descriptor now wires stm32f4_dma and nvic (the fixture's header comment still says no DMA/NVIC id is declared). That leaves PWR (real stm32f4-profile model) and both DMA stream controllers unit-only, and DBGMCU (real model, IDCODE 0x10006451 per RM0410) exercised by nothing at all.

Advanced peripherals — unit-tested only (no firmware drives them): PWR, DMA1, DMA2

Dead (1 modeled, never exercised): DBGMCU

stm32g071

SIM-DERIVED first G0 port: its committed tier-1 fixture drives all 12 declared classes over USART2 (clock/gpio/timer/pwm/dma/irq/i2c/spi/adc/wdt/rtc plus the implicit uart; nightly-ratcheted by the full-CI tier1_matrix_ratchet step (core-ci.yml:2087), fixture drift checked on Sundays), and the PR-gated firmware_survival::test_nucleo_g071rb_smoke_survival (nightly coverage-matrix cell examples/nucleo-g071rb/uart-smoke.yaml, crate firmware-stm32g0-demo) pins the dedicated stm32g0 RCC layout. Beyond the rubric nothing runs: PWR/LPTIM1/LPTIM2/DAC1/DBGMCU are family models with no unit test and no firmware (the port's KNOWN_LIMITATIONS.md records LPTIM/CRC/EXTI-SYSCFG as declared-but-untouched), CRC is driven by no firmware and its always-run coverage is the H563 bench oracle (h563_conformance.rs:144) plus the F103 sim-only thumb oracle (stm32f1_exec_oracle.rs:300) and IDR-width masking (crc.rs:204), with no G0-specific test asserting a computed value, and WWDG/SYSCFG are an inert register bank and a stub window.

Advanced peripherals — unit-tested only (no firmware drives them): CRC

Dead (5 modeled, never exercised): PWR, LPTIM1, LPTIM2, DAC1, DBGMCU

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - WWDG (configs/chips/stm32g071.yaml:349 (type: wwdg; crates/core/src/peripherals/wwdg.rs:63)) — CLOSABLE: inert register bank — tick() is the trait-default no-op, so an armed watchdog never fires; no behavioural test and no gated firmware drives it (unlike the L476, no misc fingerprint reads its CR/CFR/SR). - SYSCFG (configs/chips/stm32g071.yaml:375 (type: syscfg; crates/core/src/peripherals/generic_factory.rs:739)) — CLOSABLE: hardcoded StubPeripheral — reads return 0 (plus a canned CCCSR=READY at 0x20 that is an H7-HAL seed, meaningless on G0) and writes are dropped, so EXTI source-select is not modelled; the tier-1 fixture pends its NVIC line via ISPR0 and nothing reads the window back.

stm32c031c6

No tier-1 target. First STM32C0 part: the stock-STM32duino Arduino matrix row (core-arduino-matrix-smoke, board stm32c031c6) runs USART2 console, SysTick, LD4 GPIO, Wire/I2C1 (INA219 CONFIG/BUS read back exactly), SPI1 (MAX31855 frame) and TIM1_CH1 PWM measured on the PA5 pad. Those are rubric classes. Beyond the rubric nothing runs: CRC, PWR, DBGMCU and the DMAMUX are family models no firmware drives, the ADC is a register file the C0 HAL talks to without a conversion (matrix L5 is skipped, not passed), and SYSCFG is the shared stub.

Advanced peripherals — unit-tested only (no firmware drives them): CRC

Dead (2 modeled, never exercised): PWR, DBGMCU

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - SYSCFG (configs/chips/stm32c031c6.yaml (type: syscfg); crates/core/src/peripherals/generic_factory.rs (syscfg arm)) — CLOSABLE: StubPeripheral returns zero except a canned H7-style CCCSR readiness bit at 0x20, and drops writes. GPIO source selection lives in the G0-style EXTI EXTICR registers, not here.

stm32g0b1re

No tier-1 target. This is the narrow IO-Link customer firmware descriptor: the PR-triggered iolinki-reference workflow builds pinned STM32G0B1RE/TIOL112 sources and requires g0_reference_firmware in both execution modes, asserting TIM2 at 1 MHz, the actual falling-edge WAKE ISR clearing FPR, and USART1 configured for COM2. GPIO mux/polarity/W1C and AF1 UART pad routing have separate regressions. This proves startup and interrupt prerequisites, not electrical C/Q, master exchange or physical hardware. No beyond-rubric functional device read or advanced peripheral is modeled by this slice; SYSCFG is the only declared non-rubric block and remains a stub.

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - SYSCFG (configs/chips/stm32g0b1re.yaml (type: syscfg); crates/core/src/peripherals/generic_factory.rs (syscfg arm)) — CLOSABLE: StubPeripheral returns zero except a canned H7-style CCCSR readiness bit at 0x20, and drops writes. This is not a G0 SYSCFG behavioral model. GPIO source selection is modeled in the separate G0 EXTI-owned EXTICR registers, not through this stub.

stm32g431rb

All 12 declared rubric classes are driven by its committed tier-1 fixture (tests/fixtures/tier1/stm32g431rb.elf, sha256 in tests/fixtures/tier1/MANIFEST.json; fast_boot target, USART2 console), asserted by the stm32g431rb cell in core-coverage-matrix-smoke.yml and nightly-ratcheted by the full-CI tier1_matrix_ratchet step. Same RM0440 G4 peripheral map as stm32g474re, so the beyond-rubric blocks (PWR/CRC/DBGMCU, driven there by vendor firmware) are shared with the G474 row; no G4-specific firmware drives them for the G431 die and no G431 bench capture exists, so REV_ID stays a visible placeholder rather than a claimed value.

stm32g474re

All 12 declared rubric classes are driven by its committed tier-1 fixture (nightly-ratcheted by the full-CI tier1_matrix_ratchet step (core-ci.yml:2087), fast_boot target, USART2 console), and the two PR-gated vendor firmwares cover the non-rubric blocks: the Zephyr hello's SoC init read-modify-writes DBGMCU_CR (the reason the descriptor declares DBGMCU) and the Arduino serial startup calls HAL_PWREx_ControlVoltageScaling (PWR_CR VOS + SR2.VOSF poll) plus HAL_CRC_Init (CRC_POL/INIT/CR — the reason that survival case exists), so PWR/CRC/DBGMCU are all firmware-driven and no non-rubric block is left unexercised. The CRC polynomial engine itself is still driven by no firmware: its always-run coverage is the H563 bench oracle (h563_conformance.rs:144) plus the F103 sim-only thumb oracle (stm32f1_exec_oracle.rs:300) and IDR-width masking (crc.rs:204), and no G4-specific test asserts a computed value.

stm32h735

SIM-DERIVED first Cortex-M7 part. Its committed tier-1 fixture is PR-gated (firmware_survival::test_stm32h735_tier1_survival, Session builder) and drives clock/gpio/timer/pwm/i2c/spi/wdt/irq plus UART on USART3, leaving adc unrecorded and dma/rtc n/a (neither declared); the nightly h735-telematics-lab cell is the only other gated firmware (SPI1 TFT, USART1 BG770A modem, USART3 console), and its +QGPSLOC decode is not value-asserted — a parse failure falls back to a constant, so the transcript proves the stack ran, not the fix. Beyond the rubric, PWR (PwrH7) is driven only by the un-gated stm32h735-hal-demo/embassy-demo bring-up — no CI gate and no PwrH7 unit test — RNG is exercised only by the L073 kernel-clock unit tests over the L073 bus (rcc_kernel_clock_gate.rs:184-254 — no H735 behavioral test drives it, and rng.rs has no unit tests) and no firmware, DBGMCU is driven by nothing at all, FDCAN1/2 are pinned by fdcan.rs's 18 model unit tests plus the estate test's reset-value and CCCR-independence probes — no firmware touches them — CRC has the shared H563 bench oracle (h563_conformance.rs:144) and F103 sim-only thumb oracle (stm32f1_exec_oracle.rs:300) plus the IDR-width unit test, but no H7-specific computed-value check, and SYSCFG/WWDG are a seeded stub window and an inert register bank.

Advanced peripherals — unit-tested only (no firmware drives them): CRC, FDCAN1, FDCAN2, RNG

Dead (2 modeled, never exercised): PWR, DBGMCU

⚠ PwrH7 was written for the H7 HAL's pwr.freeze() (VOSRDY/ACTVOSRDY) and the two community-HAL demos do run it, but neither example is wired into any test or workflow and pwr.rs has no PwrH7 unit test; DBGMCU has no firmware and no test at all.

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - SYSCFG (configs/chips/stm32h735.yaml:450 (type: syscfg; crates/core/src/peripherals/generic_factory.rs:739; docs/coverage/silent-path-census.md:404)) — CLOSABLE: hardcoded StubPeripheral with one canned value — CCCSR (0x20) reads READY=1 so the H7 HAL's rcc.freeze() compensation-cell poll exits (examples/stm32h735-hal-demo/src/main.rs:43); writes are dropped, so the enable never latches and EXTI source-select is not modelled. The census records it as the telematics lab's only live stub. - WWDG (configs/chips/stm32h735.yaml:487 (type: wwdg; crates/core/src/peripherals/wwdg.rs:63)) — CLOSABLE: inert register bank — tick() is the trait-default no-op, so an armed watchdog never fires; no behavioural test and no gated firmware drives it.

stm32l031k6

NUCLEO-L031K6. PR-gated raw-register smoke (examples/nucleo-l031k6/io-smoke.yaml): reset/stack, RCC IOPENR/APB1ENR, GPIOA/GPIOB, USART2 banner STM32L031K6_OK and PB3 LED edges. No Arduino matrix row and no physical silicon run exist, so nothing is recorded as functional here. USB, LCD, DAC and LPTIM are absent on this die and are not declared. SYSCFG is the real STM32L0 model; PWR, FLASH, ADC, DMA, SPI and I2C are also modelled. This smoke does not drive those blocks.

stm32l073

Silicon-diffed L0 part (SWD capture). Its committed tier-1 fixture (nightly-ratcheted by the full-CI tier1_matrix_ratchet step (core-ci.yml:2087), fixture drift checked on Sundays) drives clock/gpio/timer/dma/irq/i2c/spi/adc/wdt/rtc over USART2 — pwm is n/a, there is no advanced timer — and three PR-gated firmwares run on it: the hardware-validated firmware-l073-demo reads DBGMCU_IDCODE (DEV=20086447, asserted byte-for-byte against silicon) and computes CRC-32 by value (CRC=B874177A, also silicon-exact), the Arduino serial startup programs PWR_CR.VOS (and chip_conformance's reset oracle pins PWR's reset words), and the Zephyr hello boots. RNG is firmware-driven: the PR-gated demo enables HSI48, completes the DRDY handshake and reads a word (the drawn value is informational — the survival transcript asserts only DEV/CLK/CRC/DMA), and rcc_kernel_clock_gate.rs additionally pins that gate over the production bus. DAC and LPTIM1 are modeled but never exercised; SYSCFG/USB FS/LCD are stubs and WWDG an inert register bank. SCB is a core block installed by configure_cortex_m for every Cortex-M part (9 scb.rs tests plus the SCB ratchet at nvic_masking_config_path.rs:163; the fixture's IRQ handler reads ICSR), so it is not listed per-chip.

Dead (2 modeled, never exercised): DAC, LPTIM1

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - SYSCFG (configs/chips/stm32l073.yaml:408 (type: stub; configs/peripherals/stm32l073/syscfg_comp.yaml; docs/coverage/silent-path-census.md:381)) — CLOSABLE: author-declared stub — read 0, writes dropped (examples/nucleo-l073rz/VALIDATION.md:284); the PR-gated Arduino startup read-modify-writes SYSCFG+0x20 during HAL_RCC_OscConfig, but the RMW is fed 0 and its write is dropped, and the tier-1 fixture pends its NVIC line via ISPR0, so nothing depends on the window. - USB FS (configs/chips/stm32l073.yaml:416 (type: stub; configs/peripherals/stm32l073/usb_fs.yaml)) — CLOSABLE: author-declared stub — read 0, writes dropped (VALIDATION.md:284); the L0 USB-FS device engine (EPnR/CNTR/ISTR/BTABLE/PMA) is not modelled and no gated firmware opens the window. - LCD (configs/chips/stm32l073.yaml:425 (type: stub; configs/peripherals/stm32l073/lcd.yaml)) — CLOSABLE: author-declared stub — the L073's glass controller (the L072-vs-L073 differentiator) is not modelled and no gated firmware opens the window. - WWDG (configs/chips/stm32l073.yaml:358 (type: wwdg; crates/core/src/peripherals/wwdg.rs:63)) — CLOSABLE: inert register bank — tick() is the trait-default no-op, so an armed watchdog never fires; no behavioural test and no gated firmware drives it (the L073 reset capture does not cover its window).

stm32u575

SIM-DERIVED STM32U575 target with NO tier-1 target (absent from TIER1_TARGETS and tier1-matrix.json), so this entry carries the whole story. Firmware that runs: the two PR-gated firmware_survival boots — the stock Zephyr 3.7.2 hello (U5 RCC PLL1/ready pairs, PWR, DBGMCU_CR, ICACHE, USART1 VCP console) and the Arduino L0 serial (HAL_PWREx_ControlVoltageScaling SVMSR/VOSR poll, HAL_CRC_Init CRC_POL/INIT/CR, HAL_RCCEx_CRSConfig, HAL_ICACHE_Enable) — plus crates/firmware-stm32u575-demo's Rust io-smoke and Star Catcher OLED game: io-smoke is built and run by strict_onboarding (cross-build-excluded from the PR shards, run by the nightly core-full job), while the six-check Star Catcher simulator is included in the core-onboarding-smoke.yml matrix on main pushes, nightly schedule, and manual dispatch (not PR). The local smoke checks PC13 input/B1, nonblank SSD1306 output, and USART1 TX routing; see examples/nucleo-u575zi/VALIDATION.md for current evidence. Also the un-gated CubeU5 HAL smoke (examples/nucleo-u575zi/VALIDATION.md §B: HAL_Init → ICACHE → PLL1/VOS → USART1; no CI gate, and its only beyond-rubric touches are the already-recorded PWR and ICACHE windows). The Arduino matrix L0–L8 runs on its push-to-main/nightly lane (core-arduino-matrix-smoke.yml:15, NOT PR) and drives the rubric classes (INA219 on I2C1, MAX31855 on SPI1, ADC/PWM/timer, FDCAN1 internal loopback). Beyond the rubric, three external 60North reference firmware cases exercise GPDMA1 in the LabWired simulator: hal-spi-dma, hal-uart-dma, and st-dma-flash-to-ram. core-onboarding-smoke runs them on main pushes, nightly schedule, and manual dispatch (not PR); this is firmware-model acceptance, not a silicon capture or Renode differential. The normal firmware_survival ELFs do not access the GPDMA base; 22 gpdma.rs unit tests also cover the model. RNG is exercised only by the L073 kernel-clock unit tests over the L073 bus (rcc_kernel_clock_gate.rs:184-254 — no U575 behavioral test drives it; register_coverage.rs:128,169-179 only pokes the window), and the U5 ICACHE profile uses a limited register-contract model: CACHEINV self-clears to BSYENDF, WAYSEL locks while enabled, FCR clears status, and BSYENDIE gates the IRQ; flat memory means cache contents and line effects are not simulated. PWR/CRC/CRS/DBGMCU are all firmware-driven — the CRC polynomial engine (crc.rs:66) is still driven by no firmware (its always-run coverage is the H563 bench oracle, h563_conformance.rs:144, plus the F103 sim-only thumb oracle, stm32f1_exec_oracle.rs:300, and IDR-width masking, crc.rs:204 — no U5 test asserts a computed value) and CRS has no SYNC source or trim loop, but the HAL never polls one.

Advanced peripherals — unit-tested only (no firmware drives them): RNG

stm32u5a5zj

SIM-DERIVED NUCLEO-U5A5ZJ-Q target with no tier-1 row or prebuilt CLI firmware. The in-process stm32u5_exec_oracle checks SHA-256 of abc, AES-128 ECB, 4^13 mod 497, and negative PKA subtraction sign extension through 4160 bits; it also rejects 4161-bit operands. Separate tests cover SHA-1/SHA-224/MD5 register vectors on the shared HASH model via U575 and an RFC 4231 HMAC unit vector. The nightly board firmware smoke now drives SAES AES-128 ECB with the NIST SP 800-38A known-answer vector; other SAES unit vectors cover CBC/CTR, and the model implements AES-256 but has no AES-256 vector yet. HASH and PKA remain register-oracle-only. The SAES model does not cover GCM/CCM or TrustZone key selection. No physical U5A5 capture or acceptance is claimed; see docs/boards/stm32u5a5zj.md for model limits.

Functional device/protocol reads (real driver, decoded value): - NUCLEO-U5A5ZJ-Q SAES AES-128 ECB — NIST SP 800-38A known-answer block — nucleo-u5a5-saes-smoke · examples/nucleo-u5a5zj/evidence/saes (local smoke 2/2; CI matrix) (nightly CI)

Advanced peripherals — unit-tested only (no firmware drives them): HASH, PKA

stm32u545

SIM-DERIVED STM32U545RE (NUCLEO-U545RE-Q), a U575-family derivative checked by u545_u575_drift_guard, with explicit DBGMCU IDCODE and system-memory flash-size peripheral-field exceptions; the U575 firmware evidence does not automatically transfer to U545. Firmware that runs: the PR-gated firmware_survival blinky (RCC AHB2ENR1/APB2ENR gates, PA9/PA10 AF7, USART1 VCP banner, PA5 loop) and the nucleo_u545re machine-run pin tests (LD2 PA5 edges, B1 PC13 press/release via the VCP, USART1 RX queue-injected echo (PA10 pad input is not routed), 256 KiB-bank flash erase); the ignored-by-default Cube HAL run (MSI->PLL1 160 MHz, SMPS, ICACHE, HAL_GPIO_ReadPin PC13) is evidence in examples/nucleo-u545re/VALIDATION.md, not a PR gate. Its ICACHE peripheral uses the shared limited U5 register-contract model (cache contents are not simulated); GPDMA1 remains unit-tested-only for this U545 profile.

Advanced peripherals — unit-tested only (no firmware drives them): GPDMA1, RNG

stm32wb55

Dual-core WB (Cortex-M4 app + M0+ radio). Its committed tier-1 fixture (nightly-ratcheted by the full-CI tier1_matrix_ratchet step (core-ci.yml:2087)) drives all 12 declared classes including dma through DMA1; on the PR gate both firmware_survival boots cover every beyond-rubric block the descriptor declares: the Zephyr hello exercises the HSEM inter-core lock (the RLR read of Zephyr's z_stm32_hsem_lock is granted to CPU1 — this is why HSEM is NOT dead here, unlike the L476, which does not carry the block: the WB boot firmware actually drives the lock), the classic RCC BDCR LSE path, PWR and DBGMCU_CR; the Arduino serial startup writes CRC_POL at 0x4002_3014 (the reason that survival case exists), takes the same HSEM lock path and PWR backup-access. No beyond-rubric block is left unexercised and there are no shims.

stm32wba52

First WBA (Cortex-M33). Its committed tier-1 fixture (nightly-ratcheted by the full-CI tier1_matrix_ratchet step (core-ci.yml:2087)) drives all 11 of its declared classes — adc is n/a, ADC4 is not declared (configs/chips/stm32wba52.yaml:205) — including dma through GPDMA1, the descriptor's only DMA controller; on the PR gate both firmware_survival boots cover every beyond-rubric block: the Zephyr hello exercises the WBA-specific RCC (CFGR1@0x1C, BDCR1@0xF0, the PLL1CFGR PLL1RCLKPRE → PLL1RCLKPRERDY handshake at 0x28), the PWR VOSR handshake, DBGMCU_CR at 0xE0044004 and ICACHE; the Arduino serial startup calls HAL_PWREx_ControlVoltageScaling (VOSR write + SVMSR ACTVOSRDY poll), writes CRC_POL at 0x4002_3014 (the reason that survival case exists) and HAL_ICACHE_Enable. ICACHE is the descriptor's only shim.

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - ICACHE (configs/chips/stm32wba52.yaml:95 (type: icache; crates/core/src/peripherals/generic_factory.rs:734)) — CLOSABLE: hardcoded StubPeripheral::new(0x00) — read 0, writes dropped. The PR-gated Arduino startup's HAL_ICACHE_Enable and the Zephyr boot write it and neither polls a status bit, so the enable never latches and the boot cannot tell.

atmega328pb

No tier-1 target. The PR-gated exact-part AVR-GCC smoke prints PB_USART1_OK through the PB-only USART1 and drives PB5 high (avr_atmega328pb_usart1::atmega328pb_firmware_uses_usart1_and_drives_pb5; examples/atmega328pb-usart1/io-smoke.yaml). No silicon capture or package-pad map exists yet. Nothing beyond the rubric is firmware-exercised in this slice.

atmega328p

No tier-1 target (absent from TIER1_TARGETS and tier1-matrix.json). Firmware that runs: the PR-gated AVR golden avr_nano_golden_survival::arduino_nano_golden_prints_and_blinks (core-ci.yml:688) boots the committed PlatformIO blink ELF and asserts nano-ok on USART0 plus a PB5 toggle (avr_nano_golden_survival.rs:100-102); strict_onboarding boots the same ELF through examples/arduino-nano-blinky/io-smoke.yaml in the nightly core-full job (cross-build-excluded from the PR shards, workspace-test-shards.json:44-48; discovery documented at strict_onboarding.rs:369), and the nightly Arduino matrix drives L0–L7 on the part — L8 CAN is skipped, the chip has no CAN controller (validation/arduino-matrix/boards.yaml:306-322); its L3/L4 cells are raw register/frame oracles (INA219 config 0x399F + bus 0x19CA, validation/arduino-matrix/sketches/L3_i2c_sensor/src/main.ino:20-21,92; MAX31855 frame 0x01901600, validation/arduino-matrix/sketches/L4_spi_sensor/src/main.ino:66-67), not decoded physical values, so functional stays empty. Every peripheral the descriptor declares is a rubric class or kit-attach plumbing: the spi/i2c instances are parking controllers that exist only to drain attached slaves into the AVR CPU, which owns the real SPCR/SPSR/SPDR and TWI (configs/chips/atmega328p.yaml:28-39; crates/core/src/cpu/avr.rs:91-106), PORTB/C/D are avr_gpio and the ADC input side is avr_adc — so nothing beyond the rubric is declared and the lists below are empty.

atmega2560

No tier-1 target. Firmware that runs: the PR-gated avr_mega2560_io integration test boots the committed Arduino-core ELF (tests/fixtures/avr/arduino-mega-2560-io.elf, built for arduino:avr:mega from examples/arduino-mega-2560-io) and asserts the 12 LED pins on ports E/G/F/B/H, analogRead(A8)/(A9) through ADCSRB.MUX5, millis() against simulated time, USART0 Serial and an EICALL through a function pointer. The descriptor reuses the 328P rubric classes (avr_gpio ports A..L, avr_adc with 16 channels, the parking spi/i2c controllers and the usart0 peer host), so nothing beyond the rubric is declared and the lists below are empty.

atmega32u4

No tier-1 target. The Arduino Leonardo matrix row (validation/arduino-matrix/boards/arduino-leonardo.yaml) builds the stock ArduinoCore-avr sketches and runs L0-L7 on the model (8 pass, 1 skip: no CAN); Serial is native USB CDC on the CPU's USB device model with a scripted host that enumerates it, and the CPU-owned SPI, TWI and ADC (ADCSRB.MUX5) serve the L3/L4/L5 sensor kits. That lane needs PlatformIO and runs on the nightly/main matrix, not on pull requests; cpu::avr::usb unit tests are the PR-gated evidence. Timer3, the high-speed Timer4, INTn/PCINT and EEPROM are not modelled. No beyond-rubric behavior or physical acceptance is claimed and the lists below are empty.

atmega1281

No tier-1 target. The PR-gated avr_atmega1281_firmware integration test boots the AVR-GCC image built from examples/atmega1281-io/src/main.c and asserts 1281 OK on USART0 plus PB5 high; examples/atmega1281-io/io-smoke.yaml runs the same ELF through labwired test. The exact descriptor exposes 54 GPIO pads on ports A..G and eight ADC inputs. EEPROM, USART1, Timer3 and PWM are present on silicon but not modeled or declared. No beyond-rubric behavior or physical acceptance is claimed.

atmega2561

64-pin sibling of the ATmega2560, so no tier-1 target. The PR-gated avr_atmega2561 test boots the committed 3-byte-PC image tests/fixtures/avr/atmega2561-io.elf and asserts 2561 OK on the USART0 sink plus PB5 driven high; examples/atmega2561-io/io-smoke.yaml runs the same fixture through labwired test and is the board entry in configs/ci/boards.yml. The descriptor deliberately narrows the shared public core to this SKU: 54 GPIO on ports A..G (PG0..PG5 only), USART0 only and eight ADC inputs. EEPROM (4 KB on silicon), USART1 and Timer3 are on silicon but are not modelled/declared in this slice; ports H..L, ADC8..ADC15 and Timer4/5 are 100-pin ATmega2560 features absent from the 64-pin part; Table 2-1 lists twelve 16-bit-resolution PWM channels for the 2560 and six for the 2561. There is no beyond-rubric content and the lists below are empty.

atmega4809

No tier-1 target. The Arduino Nano Every matrix row (validation/arduino-matrix/boards/arduino-nano-every.yaml) builds the stock ArduinoCore-megaavr sketches and runs L0, L1, L2, L3, L4, L5 and L7 on the model, and examples/arduino-nano-every/io-smoke.yaml runs the same Nano Every firmware through labwired test; both need PlatformIO and run on the nightly/main matrix, not on pull requests. PR-gated: avr_nano_every_bme280 (crates/core/tests/avr/avr_nano_every_bme280.rs) boots the committed stock-Arduino sketch tests/fixtures/avr/arduino-nano-every-bme280.elf, in which the Adafruit BME280 library reads the sensor over Wire on the TWI0 host model (megaavr_twi) and prints 25.00 C, 1013.25 hPa and 50.00 %RH on USART3; a second case moves the sensor environment and a third removes the sensor. avr_nano_every_max31855 (crates/core/tests/avr/avr_nano_every_max31855.rs) boots tests/fixtures/avr/arduino-nano-every-max31855.elf, in which the Adafruit MAX31855 library reads the converter over SPI on the SPI0 host model (megaavr_spi) and prints 25.00 C and 22.0000 C; a second case moves both junction temperatures, a third wires the chip select to another pad and a fourth removes the converter. The matrix L3 cell is a raw INA219 register oracle and the L4 cell an exact MAX31855 frame. The TCA0/TCB waveform outputs are not modelled, so the PWM matrix cell is skipped. TWI0 client mode, dual mode, smart mode, quick command and the bus time-out are refused, and the TWI0 pads are not routed. SPI0 Buffer mode and client mode are refused; the multi-host SS input and the SPI clock phase / polarity are not modelled. No physical acceptance is claimed.

Functional device/protocol reads (real driver, decoded value): - Adafruit BME280 library over Wire (TWI0) — compensated 25.00 C, 1013.25 hPa, 50.00 %RH on Serial (USART3) — arduino-nano-every-bme280 · crates/core/tests/avr/avr_nano_every_bme280.rs (tests/fixtures/avr/arduino-nano-every-bme280.elf) (PR gate) - Adafruit MAX31855 library over SPI (SPI0) — thermocouple 25.00 C and cold junction 22.0000 C on Serial (USART3) — arduino-nano-every-max31855 · crates/core/tests/avr/avr_nano_every_max31855.rs (tests/fixtures/avr/arduino-nano-every-max31855.elf) (PR gate)

atsamd21

No tier-1 target. The firmware that runs is the PR-gated firmware_survival::test_atsamd21_nano33_smoke_survival (step core-ci.yml:779; case firmware_survival.rs:1323): the committed Nano 33 IoT smoke ELF (tests/fixtures/atsamd21-nano33-smoke.elf, built from crates/firmware-atsamd21-demo) writes PM.APBCMASK + GCLK SERCOM5_CORE and prints OK on Serial1 (crates/firmware-atsamd21-demo/src/main.rs:35-41). The same demo is built and run by strict_onboarding (nightly core-full, through examples/nano-33-iot/io-smoke.yaml) and the nightly coverage-matrix cell nano-33-iot (core-coverage-matrix-smoke.yml:75-81, examples/nano-33-iot/uart-smoke.yaml). Beyond the rubric the descriptor declares one block, the USB window, as a zero stub that no firmware opens.

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - USB (configs/chips/atsamd21.yaml:68 (type: stub; crates/core/src/peripherals/stub.rs:12)) — CLOSABLE: author-declared zero window — reads return 0 and writes are dropped, so the SAM D21 native-USB device (the Nano 33 IoT's USB CDC console) is not modelled; the PR-gated smoke drives SERCOM5 (Serial1) only and no gated firmware opens the window (docs/boards/nano-33-iot.md:32 lists USB as unmodelled).

atsamd21g18a

No tier-1 target. The firmware that runs is the SAM D21 bring-up smoke built from crates/firmware-samd21-demo (committed copy at tests/fixtures/samd21-smoke.elf; the gated lanes build the crate, not that copy): examples/samd21-smoke/io-smoke.yaml asserts OK, samd21 sercom0 up and led on plus three PORT memory_value reads, and is executed by strict_onboarding (nightly core-full, cross-build-excluded from the PR shards, workspace-test-shards.json:44-48) and the nightly coverage-matrix cell samd21-smoke (core-coverage-matrix-smoke.yml:296-303). It drives the rubric classes (clock through PM/GCLK, GPIO through PORT, uart through SERCOM0) and touches both beyond-rubric declarative banks directly — the SYSCTRL.PCLKSR poll and the NVMCTRL.CTRLB wait-state write — which are register storage with no behavioural engine behind them. PM/GCLK are the clock-class banks and WDT the wdt-class bank: same declarative shape, class-owned, not listed here.

Functional device/protocol reads (real driver, decoded value): - Stock ArduinoCore-samd sketch on the Arduino Nano 33 IoT: the scripted USB host enumerates the native port (samd21_usb) and opens its CDC serial port; the Adafruit BME280 library reads the sensor over Wire (SERCOM4) and prints 25.00 C, 1013.25 hPa, 50.00 %RH on Serial (USB CDC) — arduino-nano-33-iot-bme280 · crates/core/tests/samd/arduino_nano_33_iot_bme280.rs (tests/fixtures/samd/arduino-nano-33-iot-bme280.elf) (PR gate)

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - SYSCTRL (configs/chips/atsamd21g18a.yaml:144 (type: declarative; configs/peripherals/atsamd21g18a/sysctrl.yaml)) — CLOSABLE: declarative register file with no engine — the smoke's OSC8MRDY poll (crates/firmware-samd21-demo/src/main.rs:151-155) is bounded (for _ in 0..10_000) and passes even with PCLKSR reset to 0, so it does not depend on the hand-set value; that value (0x1_8ADF, configs/peripherals/atsamd21g18a/sysctrl.yaml:250-272, value at :264) is what unbounded driver spins like the Arduino SAMD core's DFLLRDY wait need, and the pin is the estate test (atsamd21_peripheral_estate.rs:116-139), the chip's chip_conformance behavior gate (chip_conformance.rs:86-93). No oscillator/DFLL state is modelled, and regenerating the file from the SVD drops the value. ⚠️ examples/samd21-smoke/io-smoke.yaml:17-19 and crates/firmware-samd21-demo/src/main.rs:149-150 still claim a PCLKSR=0 model hangs / never returns — those comments are STALE (the poll is bounded); do not 'correct' this note back from them. - NVMCTRL (configs/chips/atsamd21g18a.yaml:172 (type: declarative; configs/peripherals/atsamd21g18a/nvmctrl.yaml)) — CLOSABLE: declarative register file — the smoke's CTRLB.RWS wait-state write (crates/firmware-samd21-demo/src/main.rs:147) is stored, never read back or polled, and no flash-controller behaviour follows (no programming, erase or ready timing).

atsamd21e15a

Geometry-only orderable on the atsamd21g18a engine; no per-orderable firmware exercise or beyond-rubric content is claimed.

atsamd21e16a

Geometry-only orderable on the atsamd21g18a engine; no per-orderable firmware exercise or beyond-rubric content is claimed.

atsamd21e17a

Geometry-only orderable on the atsamd21g18a engine; no per-orderable firmware exercise or beyond-rubric content is claimed.

atsamd21e18a

Geometry-only orderable on the atsamd21g18a engine; no per-orderable firmware exercise or beyond-rubric content is claimed.

atsamd21g15a

Geometry-only orderable on the atsamd21g18a engine; no per-orderable firmware exercise or beyond-rubric content is claimed.

atsamd21g16a

Geometry-only orderable on the atsamd21g18a engine; no per-orderable firmware exercise or beyond-rubric content is claimed.

atsamd21g17a

Geometry-only orderable on the atsamd21g18a engine; no per-orderable firmware exercise or beyond-rubric content is claimed.

atsamd21j15a

Geometry-only orderable on the atsamd21g18a engine; no per-orderable firmware exercise or beyond-rubric content is claimed.

atsamd21j16a

Geometry-only orderable on the atsamd21g18a engine; no per-orderable firmware exercise or beyond-rubric content is claimed.

atsamd21j17a

Geometry-only orderable on the atsamd21g18a engine; no per-orderable firmware exercise or beyond-rubric content is claimed.

atsamd21j18a

Geometry-only orderable on the atsamd21g18a engine; no per-orderable firmware exercise or beyond-rubric content is claimed.

atsamd51

No tier-1 target. The firmware that runs is the PR-gated firmware_survival::test_atsamd51_metro_m4_smoke_survival (step core-ci.yml:779; case firmware_survival.rs:1336): the committed Metro M4 smoke ELF (tests/fixtures/atsamd51-metro-m4-smoke.elf, built from crates/firmware-atsamd51-demo) writes MCLK.APBBMASK + GCLK PCHCTRL[24] and prints OK on Serial1 (crates/firmware-atsamd51-demo/src/main.rs:35-41). The same demo is built and run by strict_onboarding (nightly core-full, through examples/metro-m4/io-smoke.yaml) and the nightly coverage-matrix cell metro-m4 (core-coverage-matrix-smoke.yml:82-88, examples/metro-m4/uart-smoke.yaml). Beyond the rubric the descriptor declares two zero windows, USB and QSPI, that no firmware opens.

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - USB (configs/chips/atsamd51.yaml:69 (type: stub; crates/core/src/peripherals/stub.rs:12)) — CLOSABLE: author-declared zero window — reads return 0 and writes are dropped, so the SAMD51 native-USB device is not modelled; the PR-gated smoke drives SERCOM3 (Serial1) only and no gated firmware opens the window (docs/boards/metro-m4.md:27-28 lists QSPI/USB/NeoPixel as deliberate stub windows). - QSPI (configs/chips/atsamd51.yaml:73 (type: stub; crates/core/src/peripherals/stub.rs:12)) — CLOSABLE: author-declared zero window — the SAMD51 QSPI controller (the board's onboard QSPI flash) is not modelled; firmware executes from internal flash and nothing opens the window (docs/boards/metro-m4.md:27-28).

atsamd51-pybadge

No tier-1 target and no firmware runs on this descriptor. It is the atsamd51 descriptor with the PyBadge application layout (reset_vector_offset 0x4000, UF2 bootloader resident below). Its only gate, crates/core/tests/samd/pybadge_config.rs, builds the PyBadge bus from configs/systems/pybadge.yaml and checks the reset SP/PC fetched from a stub vector table at 0x4000; it steps no firmware. The Metro M4 smoke that exercises atsamd51 loads atsamd51.yaml, not this file. Beyond the rubric it declares the same two zero windows as atsamd51, USB and QSPI.

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - USB (configs/chips/atsamd51-pybadge.yaml:159 (type: stub; crates/core/src/peripherals/stub.rs:12)) — CLOSABLE: author-declared zero window, same as atsamd51: reads return 0 and writes are dropped; the SAMD51 native-USB device is not modelled (configs/systems/pybadge.yaml lists native USB as a stub). - QSPI (configs/chips/atsamd51-pybadge.yaml:163 (type: stub; crates/core/src/peripherals/stub.rs:12)) — CLOSABLE: author-declared zero window, same as atsamd51: the QSPI controller (PyBadge onboard QSPI flash) is not modelled (configs/systems/pybadge.yaml lists QSPI flash as a stub).

efr32mg26

L2 on the conformance board (silicon reset oracle, 243/243 registers — docs/coverage/chip-conformance.md:39) and no tier-1 row, so this entry carries the story. Two firmwares are gated: the bare-metal demo, rebuilt by the nightly coverage-matrix cell examples/brd2709a/uart-smoke.yaml (core-coverage-matrix-smoke.yml:207-213), and the BRD2709A agent deck, named as this chip's behavior_gate in chip_conformance.rs:350 and asserted in process by efr32_deck_behavior::the_deck_firmware_drives_every_part (advisory PR workspace shards; a superset of the nightly CLI cell examples/brd2709a/deck-smoke.yaml, core-coverage-matrix-smoke.yml:221-227 — the CLI cell asserts 100% ink + lit, while the in-process gate adds the uniform top_colour=0x001F check): the committed bare-metal deck drives CMU-gated USART0 as SPI (ST7789 panel — all 170x320 pixels carry the 0x001F the firmware wrote), USART2 as I2S (INMP441 mic — even slots driven, odd tristated), IADC0 (fader code=2048) and five GPIO contacts through the Series-2 port model plus the USART ROUTE registers. That covers the board's spi/i2s/adc/gpio path where no tier-1 fixture exists. What the deck does NOT touch: the analog-bus allocation (the fader converts without allocating — on silicon that bus reads 0), SMU, MSC and the BLE controller — those are the shims and unit-only models below, and no in-tree gated firmware drives MSC or virtual_ble (the XMODEM bootloader is bench-only; firmware-mg26-ble is wired to no lane). I2C/timer are rubric classes, not recorded here.

Functional device/protocol reads (real driver, decoded value): - BRD2709A agent deck — ST7789 panel over USART0/SPI, INMP441 microphone over USART2/I2S, IADC fader and five GPIO contacts, asserted against the published display artifact (lit, every pixel inked, uniform 0x001F), the I2S slot halves (left=4 right=0), the 12-bit conversion (2048) and each contact's idle polarity — firmware-mg26-deck · crates/core/tests/vendor/efr32_deck_behavior.rs:138-246 (in process) + examples/brd2709a/deck-smoke.yaml (coverage-matrix cell, core-coverage-matrix-smoke.yml:221-227) (nightly CI)

Advanced peripherals — unit-tested only (no firmware drives them): MSC (14 msc.rs unit tests plus efr32_msc_persistence.rs, which drives WREN/ADDRB/WDATA/BUSY and reads the word back through ordinary memory; the only firmware that drives it, firmware-mg26-bootloader, is bench-only — docs/boards/brd2709a.md:58-63), virtual_ble (LabWired's DECLARED controller at 0x4F000000, NOT silicon — the EFR32 radio is a closed RAIL blob with no register documentation; 13 virtual_ble.rs unit tests plus efr32mg26_ble.rs two-machine advertising/scanning through this descriptor; firmware-mg26-ble exists but no lane runs it)

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - gpio_busalloc (configs/chips/efr32mg26.yaml:265-269 (type efr32s2_busalloc; crates/core/src/peripherals/efr32/busalloc.rs:46-52)) — CLOSABLE: the three analog-bus words store and read back and adc0_owns() decodes them (6 unit tests), but IADC0 is not gated on the model — a firmware that never allocates still converts the pad, which reads 0 on silicon. Neither gated in-tree firmware allocates (the deck's fader converts without one; firmware-mg26-adc also skips it and its examples/brd2709a/adc-smoke.yaml is wired into no workflow); the external silabs-arduino analogRead is the one that writes it. Port path: gate IADC0 on adc0_owns, as RouteGate did for the USART/I2C routing. - smu (configs/chips/efr32mg26.yaml:529-534 (type efr32s2_smu; crates/core/src/peripherals/efr32/smu.rs:29-49)) — CLOSABLE: register file with header reset values and read-only members (4 unit tests, including SystemInit's exact sequence) but NO protection enforced — PPUFS never latches and the non-secure alias move is not modelled. Declared so a Simplicity-SDK SystemInit (CLKEN1_SET + PPUSATD*_CLR, three instructions in) cannot bus-fault; no in-tree gated firmware drives or polls it.

imxrt1064

No tier-1 target (absent from TIER1_TARGETS and tier1-matrix.json). Firmware that runs: the committed Teensy 4.1 smoke ELF, PR-gated as firmware_survival::test_imxrt1064_teensy41_smoke_survival (case firmware_survival.rs:1359-1372) and rebuilt by the nightly coverage-matrix cell examples/teensy-41/uart-smoke.yaml (core-coverage-matrix-smoke.yml:96-102); it un-gates CCM CCGR0/CCGR3, configures GPIO2_IO03 and prints OK over LPUART6 from a soft-float image linked in DTCM because FlexSPI/XIP is not modelled — the chip yaml deliberately claims neither 4 MiB SiP flash nor FlexSPI (configs/chips/imxrt1064.yaml:7-9). board_config_gates.rs builds the same descriptor. Beyond the rubric there are exactly two blocks, both below; the CCM is the clock class and is firmware-driven.

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - iomuxc (configs/chips/imxrt1064.yaml:37-41 (type imx_iomuxc; crates/core/src/peripherals/imx_iomuxc.rs:7-11)) — CLOSABLE: sticky register stub — every SW_MUX_CTL/SW_PAD_CTL/SELECT_INPUT word is stored and read back (2 unit tests pin only that read-back), but no alternate-function routing follows; the smoke never opens the window and a DTCM-linked image cannot reveal a pin-mux mistake. - flexspi (configs/chips/imxrt1064.yaml:59-62 (type stub; no schema)) — CLOSABLE: zero window at 0x402A8000 — the FlexSPI controller and XIP are not modelled and the descriptor claims no 4 MiB SiP flash; firmware is linked into DTCM because XIP cannot boot here, and nothing opens the window.

mimxrt1062-teensy41

No tier-1 target. Stock Teensyduino (PJRC core, PlatformIO teensy@6.0.0) boots the Teensy 4.1 profile from the FlexSPI NOR IVT (FlexRAM, PLLs, DCDC, SysTick on the 100 kHz reference, USB1 device controller) and prints through USB CDC to a scripted host; the arduino-matrix row validation/arduino-matrix/boards/teensy41.yaml runs L0-L3, L5 and L7 on it (6 pass, 3 skip: no LPSPI, QuadTimer or FlexCAN model). That lane needs PlatformIO and runs on the nightly/main matrix, not on pull requests. LPI2C1 is a real model (an INA219 is read through stock Wire); the declared stubs are listed as shims. No physical acceptance is claimed.

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - iomuxc_gpr (configs/chips/mimxrt1062-teensy41.yaml:150-152 (type stub)) — CLOSABLE: declared stub (reads 0, writes dropped), documented in the chip header; the Teensyduino startup writes it and the model does not act on it. - flexram (configs/chips/mimxrt1062-teensy41.yaml:153 (type stub)) — CLOSABLE: declared stub (reads 0, writes dropped), documented in the chip header; the Teensyduino startup writes it and the model does not act on it. - snvs (configs/chips/mimxrt1062-teensy41.yaml:154 (type stub)) — CLOSABLE: declared stub (reads 0, writes dropped), documented in the chip header; the Teensyduino startup writes it and the model does not act on it. - pit (configs/chips/mimxrt1062-teensy41.yaml:155 (type stub)) — CLOSABLE: declared stub (reads 0, writes dropped), documented in the chip header; the Teensyduino startup writes it and the model does not act on it. - ocotp (configs/chips/mimxrt1062-teensy41.yaml:156 (type stub)) — CLOSABLE: declared stub (reads 0, writes dropped), documented in the chip header; the Teensyduino startup writes it and the model does not act on it. - src (configs/chips/mimxrt1062-teensy41.yaml:157 (type stub)) — CLOSABLE: declared stub (reads 0, writes dropped), documented in the chip header; the Teensyduino startup writes it and the model does not act on it. - gpc (configs/chips/mimxrt1062-teensy41.yaml:158 (type stub)) — CLOSABLE: declared stub (reads 0, writes dropped), documented in the chip header; the Teensyduino startup writes it and the model does not act on it. - tmr1..tmr4 (QuadTimer1-4) (configs/chips/mimxrt1062-teensy41.yaml:159-162 (type stub)) — CLOSABLE: declared stub (reads 0, writes dropped), documented in the chip header; the Teensyduino startup writes it and the model does not act on it. - pwm1..pwm4 (eFlexPWM1-4) (configs/chips/mimxrt1062-teensy41.yaml:163-166 (type stub)) — CLOSABLE: declared stub (reads 0, writes dropped), documented in the chip header; the Teensyduino startup writes it and the model does not act on it. - scs_debug (CoreDebug DHCSR..DEMCR) (configs/chips/mimxrt1062-teensy41.yaml:174 (type stub)) — CLOSABLE: declared stub (reads 0, writes dropped), documented in the chip header; the Teensyduino startup writes it and the model does not act on it.

mkw41z4

No tier-1 target (absent from TIER1_TARGETS and tier1-matrix.json). Firmware that runs: five PR-gated firmware_survival ELFs — the bare-metal LPUART0 smoke, the genuine NXP MCUXpresso SystemInit + BOARD_BootClockRUN boot (MCG FEE + RSIM RF_OSC_READY + LPUART), unmodified Zephyr v3.7 hello_world, the stock Zephyr fxos8700 sample, and the bare-metal cow-activity tag — plus the tag's nightly CLI cell (examples/kw41z-cow-activity/calm.yaml, core-coverage-matrix-smoke.yml:200-202); strict_onboarding skips this chip (strict_onboarding.rs:99-104). The descriptor's beyond-rubric blocks are all engine-less declarative banks: SIM is driven by the gated vendor boot (SIM.COPC written, CLOCK_SetSimConfig programs SIM.CLKDIV1, BOARD_RfOscInit reads SIM.SDID), SMC only by the uncalled BOARD_BootClockVLPR path, and PMC/RCM/TRNG0 by nothing. The radio wall — BTLE_RF, the nine XCVR* blocks, GENFSK, ZLL, ANT and CMT — is present in the vendored SVD (tests/fixtures/real_world/mkw41z4.svd) and deliberately not modelled: those windows are unmapped, so radio firmware faults loudly (configs/chips/mkw41z4.yaml:7-14)._

Functional device/protocol reads (real driver, decoded value): - FXOS8700 accelerometer read over Kinetis I2C1 drives a PCD8544 LCD render (cattle-activity tag): pixel-level cow features are asserted in the calm pose, and a tilt injected on the sensor channels (x=+2.0 g, y=-2.0 g, z=1.0 g) must change >=300 framebuffer pixels — kw41z-lcd-activity · firmware_survival.rs:2111-2187 (render) and :2195-2285 (tilt reaction); nightly CLI cell examples/kw41z-cow-activity/calm.yaml (PR gate)

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - SIM (configs/chips/mkw41z4.yaml:67-72 (type declarative; configs/peripherals/mkw41z4/sim.yaml)) — CLOSABLE: engine-less bank — the gated vendor boot writes SIM.COPC (crates/firmware-kw41z-nxp/vendor/system_MKW41Z4.c:87-88), programs SIM.CLKDIV1 via CLOCK_SetSimConfig (vendor/fsl_clock.c:475-479, called from BOARD_BootClockRUN at vendor/clock_config.c:163), and BOARD_RfOscInit reads SIM.SDID (vendor/clock_config.c:174); the hand-set SDID REVID is what makes that routine skip the rev-1.0 workaround (configs/chips/mkw41z4.yaml:62-66), and no gating or clock-divider behaviour follows the writes. - PMC (configs/chips/mkw41z4.yaml:91-96 (type declarative; configs/peripherals/mkw41z4/pmc.yaml)) — CLOSABLE: zero-behaviour power-mode bank with no engine; no in-tree firmware opens it — the vendor VLPR path that writes SMC is not called by this board's bring-up. - SMC (configs/chips/mkw41z4.yaml:97-102 (type declarative; configs/peripherals/mkw41z4/smc.yaml)) — CLOSABLE: engine-less power-mode controller — BOARD_BootClockVLPR's SMC_SetPowerModeVlpr + PMSTAT poll (vendor/clock_config.c:112-114) is dead code on this board; were it called, the poll would HANG, not pass — PMSTAT is a read-only 0x01 (Run) with no engine behind it to advance it (configs/peripherals/mkw41z4/smc.yaml:75-86), while kSMC_PowerStateVlpr is 0x04 (vendor/fsl_smc.h:63), so SMC_GetPowerModeState (fsl_smc.h:275-277) never equals it and the while loop never exits. - RCM (configs/chips/mkw41z4.yaml:103-108 (type declarative; configs/peripherals/mkw41z4/rcm.yaml)) — CLOSABLE: reset-cause/control bank with no engine; no gated firmware reads it. - TRNG0 (configs/chips/mkw41z4.yaml:203-210 (type declarative; configs/peripherals/mkw41z4/trng0.yaml)) — CLOSABLE: register file only — no entropy source or ready-state engine and IRQ 13 is never pended; no firmware or test drives it (the radio crypto path that would is unmodelled).

ra4m1

No tier-1 target (absent from TIER1_TARGETS and tier1-matrix.json). Firmware that runs: the bare-metal Uno R4 Minima smoke, PR-gated as firmware_survival::test_ra4m1_uno_r4_smoke_survival (case firmware_survival.rs:1346-1358) and rebuilt by the nightly coverage-matrix cell examples/arduino-uno-r4-minima/uart-smoke.yaml (core-coverage-matrix-smoke.yml:89-95); it waits on the ra_sysc HOCO/OSCSF handshake, drives P111 through PORT1 PCNTR1/PCNTR3 and prints OK over SCI2 TDR, and board_config_gates.rs builds the descriptor. Since 2026-11-07 stock ArduinoCore-renesas sketches also run on it through the UNO R4 WiFi row of the Arduino matrix (validation/arduino-matrix/boards.yaml uno-r4-wifi, nightly core-arduino-matrix-smoke.yml, not PR-gated): Serial on SCI9 through the ICU event link, millis()/micros() on AGT0, Wire on IIC1 (INA219 exact), SPI on RSPI SPI0 (MAX31855), analogRead on ADC0 at 14 bits, analogWrite on GPT2 GTIOC2B (P102 pad edges), and CAN0 in internal loopback on ra_can (L8_can, register-level; validation/arduino-matrix/evidence/uno-r4-wifi/L8_can with a stubbed-CAN0 negative control). The CPU-visible stubs it adds are listed below.

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - usbfs (configs/chips/ra4m1.yaml usbfs (type stub; crates/core/src/peripherals/stub.rs)) — CLOSABLE: zero window at 0x40090000 — the R7FA4M1AB USB FS device (the Minima's USB CDC console) is not modelled; the PR-gated smoke drives SCI2 on the D0/D1 header only, and the UNO R4 WiFi matrix row prints on SCI9 (docs/boards/arduino-uno-r4-minima.md:25). - mpu_spmon (configs/chips/ra4m1.yaml mpu_spmon (type stub)) — CLOSABLE: FSP SystemInit arms the stack pointer monitor; its bounds are dropped and no NMI is raised. - adc1_reserved (configs/chips/ra4m1.yaml adc1_reserved (type stub)) — Silicon-shaped: reserved space ArduinoCore-renesas analogReference() opens as a second ADC unit on every boot.

rp2350

No tier-1 target (absent from TIER1_TARGETS and tier1-matrix.json). Firmware that runs: the committed bare-metal M33 UART0 smoke, PR-gated as firmware_survival::test_rp2350_demo_survival (case firmware_survival.rs:1523-1533) and rebuilt by the nightly coverage-matrix cell examples/rp2350-demo/io-smoke.yaml (core-coverage-matrix-smoke.yml:276-283); strict_onboarding builds and runs the same ELF through examples/pico2/io-smoke.yaml in the nightly core-full workspace step (core-ci.yml:2062; cross-build-excluded from the PR shards — scripts/ci/workspace-test-shards.json strict_onboarding entry). It proves the RP2350 base map (UART0 at 0x40070000, not the RP2040's 0x40034000) and that a cortex-m-rt image boots from XIP flash 0x10000000; its only other MMIO is a scratch store into the PIO0 window used as a mocked LED (crates/firmware-rp2350-demo/src/main.rs:9-10), so no PIO program is loaded and no state machine is observed. rp2350_clkrst_profile.rs (3 unit tests) pins the moved CLOCKS/RESETS/XOSC/PLL map. Beyond the rubric SYSINFO/XIP_CTRL are engine-less declarative files and POWERMAN/TBMAN are zero stubs; PIO0/1/2 are the shared RP2040 PIO model (9 pio.rs unit tests; hardware_fidelity.rs:194-268 drives it at the RP2040 base, and the PIO onboarding lane boots the rp2040 descriptor, not this one), so all three are firmware-untouched here.

Advanced peripherals — unit-tested only (no firmware drives them): PIO0, PIO1, PIO2

Shims (hardcoded stubs or engine-less declarative register files — not real fidelity): - SYSINFO (configs/chips/rp2350.yaml:43-48 (type declarative; configs/peripherals/rp2350/sysinfo.yaml)) — CLOSABLE: engine-less register file whose hand-set CHIP_ID (0x30004927) and PLATFORM=ASIC(2) are what a pico-sdk image reads; nothing reads it here — no USB enumeration is modelled, so the values are never exercised. - POWERMAN/TICKS (configs/chips/rp2350.yaml:68-71 (type stub)) — CLOSABLE: 48 KiB zero window — reads 0 and writes are dropped, with no power sequencing and no TICKS time source; no firmware opens it. - xip_ctrl (configs/chips/rp2350.yaml:124-129 (type declarative; configs/peripherals/rp2350/xip_ctrl.yaml)) — CLOSABLE: declarative cache-control storage — no XIP cache behaviour; firmware executes out of the flash region through the flat memory map and never opens the window. - tbman (configs/chips/rp2350.yaml:156-160 (type stub)) — STRUCTURAL: zero window — PLATFORM reads 0, which is the pico-sdk's ASIC path and the correct answer for the twin (configs/chips/rp2350.yaml:156-160); no firmware polls it.

nrf51822

S110 application-region descriptor (micro:bit V1 / Calliope mini 1): flash from 0x18000 only, the MBR/SoftDevice range an empty hole, the SoftDevice API served by crates/nrf-softdevice-hle. Every block is a reused nRF52 model (legacy-UART personality for UART0, 1 KiB NVMC pages) plus an nRF51 FICR. No beyond-rubric content is recorded: the only gated firmware is the bare legacy-UART smoke (PR-gated as firmware_survival::test_nrf51822_microbit_v1_smoke_survival, rebuilt by the nightly coverage-matrix cell examples/microbit-v1/io-smoke.yaml); official MakeCode images (radio, Bluetooth bonding + Nordic UART round trip) run under crates/core/examples/sd_hle_run.rs and renode-spike-prime's e2e_bond_uart.py, which no labwired-core lane runs.

rv1103

No beyond-rubric board firmware content is recorded. The descriptor declares CPU/memory geometry only, with no modeled RV1103 peripherals or boot path. The generic Cortex-A7 CPU smoke does not exercise RV1103 board firmware; Luckfox Pico Mini A/Linux execution remains unsupported.

dspic33fj32mc204

M0 + M1a + M1b + M1c + M2 + M3 bring-up: boot (PC 0 -> reset GOTO) with UART1 TX, timers T1/T2/T3, the interrupt fabric (IVT/AIVT vector fetch, IFS0/IEC0/IPC0-2, IFS1/IEC1/IPC3-4, IFS3/IPC14-15 sources), GPIO ports A/B/C, PPS, input capture 1/2, output compare 1/2, the DSP MAC class (MAC/MPY/MPY.N/MSC/CLR/MOVSAC/LAC/SAC/SAC.R with X/Y prefetches, write-back, CORCON saturation/rounding and OA/OB/SA/SB), Motor Control PWM 1 (edge-aligned/single-event/center-aligned time base, three duty generators with buffering, complementary/independent pairs with dead time, P1OVDCON override, FLTA1 fault states, PWM1IF postscaling), QEI1 (x2/x4 decoding, POS1CNT/MAX1CNT, match/index resets with IMV, UPDN, CNTERR/CEID, digital filter, 16-bit timer), ADC1 (10/12-bit, FORM formats, SAMP/DONE/ASAM, manual/TMR3/PWM1/internal triggers, CHPS, CSCNA scan, BUFM, SMPI, TAD/SAMC), SPI1 (master/slave, 8/16-bit, prescalers, SPIROV, loopback device) and I2C1 (master start/restart/stop, ACKSTAT, BRG, TMP102 attach, slave register/test seams). Four committed XC16 v1.70 firmwares boot on the model: xc16-uart-smoke.c prints HI (crates/core/tests/pic/pic24_xc16_elf_boot.rs), xc16-dsp-mac.c runs the DS70000157G example-1 MAC sequence (crates/core/tests/pic/pic24_xc16_dsp.rs), xc16-m2-pwm-qei.c configures PWM1+QEI1 with injected quadrature edges (crates/core/tests/pic/pic24_xc16_m2.rs) and xc16-m3-adc-spi-i2c.c samples AN4, loops SPI1 and reads the TMP102 (crates/core/tests/pic/pic24_xc16_m3.rs); the hand fixture runs in the PR shards (crates/core/tests/pic/pic24_m0_smoke.rs) and through examples/dspic33fj32mc204/io-smoke.yaml in strict_onboarding. The DSP accumulator arithmetic (ADD/SUB on ACCA/ACCB) and PSV/EDS remain unmodeled (docs/pic24-m0-reference.md), so no advanced blocks exist to list.

dspic33fj32mc202

M0 + M1 + M2 + M3 bring-up on the 28-pin sibling of the dsPIC33FJ32MC204: boot (PC 0 -> reset GOTO) and UART1 TX; the 28-pin GPIO package (RA0-RA4, RB0-RB15, no PORTC; TRIS/LAT/ODC masks derived from the descriptor's pic24_gpio pads); the shared T1 timer, Motor Control PWM 1 (complementary pair with dead time on RB14/RB15) and QEI1 (x4 index reset, QEA/QEB/INDX on RB0/RB1/RB2) on 202 package pins; and the shared ADC1/SPI1/I2C1 models on the 202's AN0-AN5 analog set (AN0=RA0, AN1=RA1, AN2=RB0..AN5=RB3; AN6-AN8 are RC0-RC2 and absent on the 28-pin package), SPI1 loopback and I2C1 TMP102 at 0x48. Four real XC16 v1.70 firmwares: xc16-uart-smoke.c prints HI and toggles LATB0 (crates/core/tests/pic/pic24_xc16_202_smoke.rs); xc16-m1-gpio.c prints G1, makes PORTA/PORTB outputs and toggles RA0/RB0 (crates/core/tests/pic/pic24_xc16_202_m1.rs), including the reset masks and pad levels; xc16-m2-pwm-qei.c waits on T1IF and configures PWM1+QEI1 (crates/core/tests/pic/pic24_xc16_202_m2.rs), with injected quadrature edges, PWM pad checks and the direction-aware index reset; xc16-m3-adc-spi-i2c.c samples AN4 (RB2), shifts 0xA5 through SPI1, reads the TMP102 at 0x48 and prints M3 OK (crates/core/tests/pic/pic24_xc16_202_m3.rs), with the 0xBEEF completion marker. The 202 shares the 204's PIC24/dsPIC33F core, 32 KB flash / 2 KB SRAM, UART1 and instruction set. The 202 track now covers M0-M3 with nothing left in that range; there is no silicon capture, so the results are simulation-derived only.

pic16f1829

L1 smoke, SIM-DERIVED; no beyond-rubric content. First enhanced mid-range PIC16 part, on the PIC16 core's new enhanced mode (49 instructions: MOVLB/MOVLP, BRA/BRW/CALLW, ADDFSR, MOVIW/MOVWI in all forms, LSLF/LSRF/ASRF, ADDWFC/SUBWFB, RESET): exact PIC16F1829 descriptor (DS40001440E) with 8K words of flash, 32 BSR banks with the core registers and 0x70-0x7F common RAM in each, FSR access to banked, linear (0x2000) and program-flash (0x8000) space, a 16-level stack with STKPTR/TOS and STVREN overflow/underflow reset, and hardware context save into the bank-31 shadows. Real SDCC 4.4.0 C firmware boots through its C runtime and prints a banner on the EUSART TXREG (0x19A), sets LATA5, reads RA5 back and toggles it, sums a stack buffer through FSR/MOVIW, takes three software INT interrupts and then TMR0 overflow interrupts that preempt a checked work loop. Its ISR has no software context save, so wrong shadow restore fails the run (crates/core/tests/pic/pic16f1829_smoke.rs, examples/pic16f1829/io-smoke.yaml). EUSART RX and baud timing, TMR1/2/4/6, CCP/ECCP, MSSP, ADC, comparators, DAC, CPS, data EEPROM, WDT, IOC, clock switching and CONFIG decoding beyond STVREN are not modelled.

pic16f877a

L1 smoke, SIM-DERIVED; no beyond-rubric content. First PIC16 part, on a new PIC16 classic mid-range core (35 instructions): exact PIC16F877A descriptor (DS39582B) with 8K words of flash, 368 B SRAM in four banks with the 0x70-0x7F shared block, and an 8-level circular stack. Real SDCC 4.4.0 pic14 C firmware boots through its C runtime: bank switching via STATUS.RP0, shared-RAM locals, and generic-pointer string reads as computed MOVWF PCLATH, MOVWF PCL jumps into RETLW tables. It prints a banner on the USART TXREG (0x19), sets RB0, reads it back and toggles it (crates/core/tests/pic/pic16f877a_smoke.rs, examples/pic16f877a/io-smoke.yaml). USART RX and baud timing, TMR1/2 (TMR0 timer mode is modelled but unused by this smoke), CCP, MSSP, ADC, comparators, data EEPROM, WDT, PSP and CONFIG decoding are not modelled.

pic18f4520

L1 smoke, SIM-DERIVED; no beyond-rubric content. First PIC18 part: exact PIC18F4520 descriptor (DS39631E) with 32 KB flash, a 31-level stack and the 0x80 Access Bank split. Real SDCC 4.4.0 C firmware boots through its C runtime: the FSR1/FSR2 software stack via MOVFF on the indirect aliases, TBLRD string reads and MULLW. It prints a banner on the EUSART TXREG (0xFAD), sets RB0 through LATB, reads it back and toggles it (crates/core/tests/pic/pic18f4520_smoke.rs, examples/pic18f4520/io-smoke.yaml). EUSART RX and baud timing, timers, CCP/ECCP, MSSP, ADC, data EEPROM, WDT, CONFIG decoding and XINST are not modelled.

pic24f04ka200

L1 smoke, SIM-DERIVED; no beyond-rubric content. Exact PIC24F04KA200 descriptor (DS30009937C): 14-pin package, 4 KB flash, 512 B SRAM, PORTA 0x007F / PORTB 0xC310 / no PORTC, UART1 at 0x0220. A real XC16 v1.70 ELF prints HI on UART1 and toggles RA0 (crates/core/tests/pic/pic24_ka_smoke.rs, examples/pic24f04ka200/io-smoke.yaml). Data EEPROM (absent), UART2, timers, ADC, SPI and I2C are not exercised for this part.

pic24f04ka201

L1 smoke, SIM-DERIVED; no beyond-rubric content. Exact PIC24F04KA201 descriptor (DS30009937C): 20-pin package, 4 KB flash, 512 B SRAM, PORTA 0x007F / PORTB 0xF397 / no PORTC, UART1 at 0x0220. A real XC16 v1.70 ELF prints HI on UART1 and toggles RA0 (crates/core/tests/pic/pic24_ka_smoke.rs, examples/pic24f04ka201/io-smoke.yaml). Data EEPROM (absent), UART2, timers, ADC, SPI and I2C are not exercised for this part.

pic24f08ka101

L1 smoke, SIM-DERIVED; no beyond-rubric content. Exact PIC24F08KA101 descriptor (DS39927C): 20-pin package, 8 KB flash, 1536 B SRAM, PORTA 0x007F / PORTB 0xF397 / no PORTC, UART1 at 0x0220. A real XC16 v1.70 ELF prints HI on UART1 and toggles RA0 (crates/core/tests/pic/pic24_ka_smoke.rs, examples/pic24f08ka101/io-smoke.yaml). Data EEPROM (512 B at 0x7FFE00), UART2, timers, ADC, SPI and I2C are not exercised for this part.

pic24f08ka102

L1 smoke, SIM-DERIVED; no beyond-rubric content. Exact PIC24F08KA102 descriptor (DS39927C): 28-pin package, 8 KB flash, 1536 B SRAM, PORTA 0x00FF / PORTB 0xFFFF / no PORTC, UART1 at 0x0220. A real XC16 v1.70 ELF prints HI on UART1 and toggles RA0 (crates/core/tests/pic/pic24_ka_smoke.rs, examples/pic24f08ka102/io-smoke.yaml). Data EEPROM (512 B at 0x7FFE00), UART2, timers, ADC, SPI and I2C are not exercised for this part.

pic24f16ka101

L1 smoke, SIM-DERIVED; no beyond-rubric content. Exact PIC24F16KA101 descriptor (DS39927C): 20-pin package, 16 KB flash, 1536 B SRAM, PORTA 0x007F / PORTB 0xF397 / no PORTC, UART1 at 0x0220. A real XC16 v1.70 ELF prints HI on UART1 and toggles RA0 (crates/core/tests/pic/pic24_ka_smoke.rs, examples/pic24f16ka101/io-smoke.yaml). Data EEPROM (512 B at 0x7FFE00), UART2, timers, ADC, SPI and I2C are not exercised for this part.

pic24f16ka102

L1 smoke, SIM-DERIVED; no beyond-rubric content. Exact PIC24F16KA102 descriptor (DS39927C): 28-pin package, 16 KB flash, 1536 B SRAM, PORTA 0x00FF / PORTB 0xFFFF / no PORTC, UART1 at 0x0220. A real XC16 v1.70 ELF prints HI on UART1 and toggles RA0 (crates/core/tests/pic/pic24_ka_smoke.rs, examples/pic24f16ka102/io-smoke.yaml). Data EEPROM (512 B at 0x7FFE00), UART2, timers, ADC, SPI and I2C are not exercised for this part.

litex-vexriscv

L1 smoke, SIM-DERIVED; no beyond-rubric content. Boot, the LiteX UART console and the countdown-timer one-shot are proven by tests/fixtures/litex-vexriscv-smoke.elf and examples/litex-vexriscv/io-smoke.yaml. SPI flash, Ethernet and MMCM are declared stubs and timer IRQ delivery is not wired (examples/litex-vexriscv/VALIDATION.md), so no advanced blocks exist to list.

stm32l552

L1 smoke, SIM-DERIVED; no beyond-rubric content. Boot from 0x08000000, the new L5 RCC gates (AHB2ENR GPIO, APB1ENR2.LPUART1EN), the LPUART1 console (PG7/PG8, AF8) and the three-LED GPIO walk with ODR readback plus the PC13 button are proven by tests/fixtures/nucleo-l552ze-q-smoke.elf and examples/nucleo-l552ze-q/io-smoke.yaml. TrustZone/SAU/GTZC, MPU/PMSAv8 enforcement and the L5 non-secure FLASH controller are not modelled; FLASH/EXTI/ADC/FDCAN/USB are declared stubs (examples/nucleo-l552ze-q/VALIDATION.md), so no advanced blocks exist to list.

cc2538sf53

L1 smoke plus a stock Contiki-NG boot, SIM-DERIVED (not yet L2: no repeated-CI history); beyond the rubric the flash controller and the AES/SHA-256 cryptoprocessor are firmware-driven. UNMODIFIED upstream Contiki-NG v4.9 examples/hello-world (TARGET=cc2538dk, full RPL Lite / 6LoWPAN / CSMA netstack) boots and prints its banner, the link-layer address from the info page's primary IEEE address and "Hello, world" every 10 s (tests/fixtures/cc2538dk-contiki-hello.elf, examples/cc2538dk/contiki-hello.yaml). It runs through the in-tree clean-room ROM (SWRU333A CCA boot to VTOR 0x0020_2000, the 0x48 utility API memcpy/memset; NOT a TI dump), SYS_CTRL CLOCK_STA.SYNC_32K, the RF core alone in an empty channel (ISRXON, timed FSM, RSSI_VALID/CCA, RFRND, MTCTRL), the sleep-timer count and SysTick; the sleep-timer compare IRQ is not used by this firmware and is covered by cc2538_sleep_timer_irq. The flash controller (FLASH_CTRL page erase and FCTL/FADDR/FWDATA write sequence, NOR AND rule, CCA lock bits) behind the ROM's PageErase/ProgramFlash runs UNMODIFIED Contiki-NG examples/storage/cfs-coffee test-cfs (CFS test 1/2 completed with 0 errors) and example-coffee (examples/cc2538dk/contiki-test-cfs.yaml, contiki-coffee.yaml), and the steps of Renode's cc2538_flash_controller.robot (examples/cc2538dk/flash-controller.yaml). The AES/SHA-256 cryptoprocessor (two-channel DMA, key store, AES ECB/CBC/CTR/CBC-MAC/CCM/GCM at 128/192/256 bits, SHA-256, level IRQ 143) runs the UNMODIFIED Contiki-NG cc2538-common/crypto suite, every vector OK (examples/cc2538dk/crypto-{ecb,cbc,ctr,cbc-mac,gcm,ccm,sha256}.yaml); cbc, cbc-mac, ctr and ccm are a superset of Renode's cc2538_cryptoprocessor.robot. Three nodes on a shared 802.15.4 air run the UNMODIFIED Contiki-NG examples/rpl-udp (one udp-server RPL root, two udp-clients; tests/fixtures/cc2538dk-contiki-udp-{server,client}.elf, crates/core/tests/cc2538_rpl_udp.rs), the topology of Renode's cc2538_rpl-udp.robot: both clients join the DODAG and get three requests echoed, through the RF core's TX/RX FIFOs, frame filtering, AUTOACK and FIFOP IRQ 141 and the uDMA (software transfers, IRQ 46). The L1 smoke (tests/fixtures/cc2538dk-smoke.elf, examples/cc2538dk/io-smoke.yaml) still proves the SYS_CTRL reset clock and RCGCUART gate, the PL011 UART0 console and the LED2-LED4 address-masked GPIO_DATA walk plus the key pull-ups. The air is lossless (no path loss, range or collisions) and the RF error IRQ, CSP and MAC-timer count are not modelled; the uDMA moves only on software requests; SYS_CTRL gating, the IOC, GPIO interrupts and the watchdog are not modelled; SSI0/SSI1 are modelled in master mode only and the L1 smoke runs an SSI0 loopback self-test (TX-FIFO queueing, 8/16-bit frames, RX overrun); I2CM (byte-command master) and GPTimer0-3 (one-shot/periodic, split 16-bit, prescale) are modelled and exercised by the imported-system onboarding smoke, as is a full PKA SRAM fill/readback (plain RAM); the PKA engine, I2CS, USB and ADC are declared stubs (examples/cc2538dk/VALIDATION.md).

r7fa6m5bh

L1 UART+LED smoke, SIM-DERIVED; no beyond-rubric content. Boot from the 0x0 code-flash vector table, the SCI0 console (P411/P410, 9600 from BRR 25), the LED1-LED3 walk (P006-P008) with PCNTR1 PODR readback and the PWPR-gated PmnPFS PCR pull-up that drives the S1/S2 levels are proven by tests/fixtures/ek-ra6m5-smoke.elf and examples/ek-ra6m5/io-smoke.yaml. The exact-system smoke examples/renode-onboarding/renesas-r7fa6m5bh-tier1-smoke.yaml also checks the SVD-derived MSTPCRA reset, MSTPCRB write/readback and MSTPCRE reset. Module-stop gating, clock generation, ICU event linking and the PFS peripheral mux are not modelled; ICU, DMAC/DTC, ELC, RTC, IWDT/WDT, USBFS/USBHS, IIC, CANFD, AGT, ETHERC, SPI, GPT, ADC12 and FACI remain declared stubs, and SCI1-9 and the other blocks are not mapped (examples/ek-ra6m5/VALIDATION.md).

r7fa8m1ah

L1 UART+LED smoke, SIM-DERIVED; no beyond-rubric content. Boot from the 0x02000000 code-flash vector table, the SCICKCR CKSREQ/CKSRDY clock-switch handshake (SCICKCR reads back 01 = MOCO), the PWPRS-gated PmnPFS PCR pull-up that drives the S1/S2 levels, the SCI9 SCI_B console (PA14/PA15, 9600 from CCR2.BRR 25, CCR3 8N1 LSB-first readback, CESR.TIST/RIST after CCR0.TE/RE) and the LED1-LED3 walk (P600/P414/P107) with PCNTR1 PODR readback are proven by tests/fixtures/ek-ra8m1-smoke.elf and examples/ek-ra8m1/io-smoke.yaml. The exact-system smoke examples/renode-onboarding/renesas-r7fa8m1ah-tier1-smoke.yaml also checks the SVD-derived MSTPCRA reset, MSTPCRB write/readback and MSTPCRE reset. Module-stop gating, clock generation (no rate is derived from SCICKCR), ICU event linking, TrustZone and the PFS peripheral mux are not modelled; ICU, DMAC/DTC, FCACHE, FACI, ELC, RTC, IWDT/WDT, AGT, USBFS/USBHS, IIC, GPT, ADC12, ETHERC, SPI and CANFD remain declared stubs, and SCI0-8 and the other blocks are not mapped (examples/ek-ra8m1/VALIDATION.md).

same70q21b

L1 USART+LED smoke, SIM-DERIVED; no beyond-rubric content. Boot from the 0x00400000 flash vector table on the Cortex-M7, the WDT_MR.WDDIS write, the MATRIX CCFG_SYSIO.SYSIO4 PB4 release, the PMC PCER0 gates (PIOA/PIOB/PIOC/USART1) with PCSR0 readback and MCKRDY, the PIO_ABCDSR1/2 + PDR mux of PA21 (peripheral A) and PB4 (peripheral D), the USART1 EDBG console (115384 baud from CD 6 / FP 4 on the 12-MHz main RC) and the LED0 walk (PC8, active low) with ODSR and pad readback plus the SW0 (PA11) pull-up level are proven by tests/fixtures/sam-e70-xplained-smoke.elf and examples/sam-e70-xplained/io-smoke.yaml. Clock generation, the PIO peripheral mux side effects, SYSIO pad gating and the watchdog timeout are not modelled; EEFC, the RSTC/SUPC/RTT/WDT/RTC system controller and the bus matrix are declared stubs and GMAC, TRNG, TC, SPI, QSPI, TWIHS, MCAN, AFEC, DACC, PWM, HSMCI, SSC, USBHS and XDMAC are not mapped (examples/sam-e70-xplained/VALIDATION.md), so no advanced blocks exist to list.

amap42kp

L1 UART+LED smoke plus stock Zephyr v3.7.0 hello_world (tests/fixtures/amap42kp-zephyr-hello.elf, examples/apollo4p-evb/zephyr-hello.yaml: the AmbiqSuite HAL low-power init through PWRCTRL, the CPU block DAXI flush with DAXIREADY, SYNC_READ, and the clean-room boot-ROM br_util_delay_cycles), SIM-DERIVED; no beyond-rubric content. Boot from the post-SBL vector table at 0x00018000 (reset_vector_offset; the secure bootloader is not modelled, the boot ROM only as a clean-room helper table), the PWRCTRL UART0 power-domain handshake (DEVPWREN write, DEVPWRSTATUS polled until PWRSTUART0) and the LP performance mode read from MCUPERFREQ, the PADKEY-gated PINCFG writes, the PL011-layout UART0 console (115200 from IBRD 13 / FBRD 1; UARTCR readback) and the LED0-LED2 walk (GPIO30/90/97, active low) through single-bit WTC/WTS strobes with RD readback and the WT latch check, plus the BTN0/BTN1 PULLCFG pull-up levels, are proven by tests/fixtures/apollo4p-evb-smoke.elf and examples/apollo4p-evb/io-smoke.yaml. The source-built examples/apollo4p-evb/io-smoke.yaml and exact imported-system examples/renode-onboarding/ambiq-apollo4-smoke.yaml inject UART byte R, check PL011 RIS=0x30/MIS=0x10 and NVIC IRQ 15 pending, then read DR and check deassertion. The GPIO60/GPIO47 FNCSEL 4 UART mux is stored but the GPIO model does not route non-GPIO functions; power domains have no effect on their peripherals; CLKGEN, MCUCTRL, RSTGEN, IOM0-7, MSPI0-2, ADC, timers, STIMER, RTC, WDT, security, crypto, GPU/display, SDIO, USB and audio are declared stubs (examples/apollo4p-evb/VALIDATION.md), so no advanced blocks exist to list.

mk64f12

L1 UART-only smoke, SIM-DERIVED; no beyond-rubric content. Boot from 0x00000000, the classic Kinetis UART0 console (PTB16/PTB17 ALT3, 115200 from the 20.97152 MHz FEI reset clock) and the active-low RGB LED walk with PDOR readback are proven by tests/fixtures/frdm-k64f-smoke.elf and examples/frdm-k64f/io-smoke.yaml. SIM clock gating, PORT pad mux/pulls and the MCG PLL/FEE are not modelled; ENET+KSZ8081 (TCP deferred), RNGA and FTFE are declared stubs and eDMA, PIT/FTM, I2C/SPI, ADC/DAC, USB, CAN and SDHC are not mapped (examples/frdm-k64f/VALIDATION.md), so no advanced blocks exist to list.

s32k388

L1 LPUART+LED smoke, SIM-DERIVED; no beyond-rubric content. Boot of CM7_0 from the 0x00401000 vector table named by the boot header's CM7_0 slot (IVT marker 0x5AA55AA5 at 0x00400000; the boot ROM is not modelled and the IVT is not parsed), the LPUART2 console (115384 baud from OSR 16 / SBR 13 on the 24-MHz AIPS_SLOW_CLK) and the SIUL2 GPIO path on the six RGB LED pads (PTF17-19, PTF22-24: MSCR SSS/OBE/IBE, GPDO bytes, PGPDO11 and MPGPDO11 writes with GPDI/PGPDI11 readback) plus the PTA1 pull-up/pull-down levels are proven by tests/fixtures/s32k388-evb-smoke.elf and examples/s32k388-evb/io-smoke.yaml. The SWT_0 unlock/disable and the MC_ME partition-1 clock gating are executed but not checked (both are declared stubs); clocks stay at the 48-MHz FIRC reset values; the input mux (IMCR), INV and the external interrupts are not modelled; MC_CGM, PLL, PMC, MSCM, XRDC, STM_0 and DMAMUX are declared stubs and FlexCAN, LPSPI, LPI2C, eMIOS, eDMA, GMAC, ADC and the other M7 cores are not mapped (examples/s32k388-evb/VALIDATION.md), so no advanced blocks exist to list.

mimxrt595s

L1 USART+LED smoke, SIM-DERIVED; no beyond-rubric content. Boot of the Cortex-M33 from the 0x08001000 vector table behind the FlexSPI configuration block (FCB tag 'FCFB' at 0x08000400; the boot ROM is not modelled and the FCB is not parsed), the FLEXCOMM0 USART console (PSELID = USART, FIFO TX, 115384 baud from OSR 16 / BRG 26 on the 48-MHz FRO_DIV4 function clock) and the high-speed GPIO path on the RGB LED (P0_14, P1_0, P3_17: DIRSET, SET, byte B, word W, NOT, MPIN under MASK and CLR with PIN readback) plus the SW1/SW2 released levels (P0_25, P0_10) are proven by tests/fixtures/mimxrt595-evk-smoke.elf and examples/mimxrt595-evk/io-smoke.yaml. The CLKCTL0/1 clock selects and gates (read back by the smoke), the RSTCTL1 reset releases and the IOPCTL pad functions are stored but have no effect; GPIO pin interrupts, the USART interrupt line and baud timing are not modelled; SYSCTL0/1, PMC, CACHE64, WWDT0, INPUTMUX and the FlexSPI0 controller are declared stubs and FLEXCOMM1-16, CTIMER, OS timer, DMA, USB, display, PowerQuad and the HiFi4 DSP are not mapped (examples/mimxrt595-evk/VALIDATION.md), so no advanced blocks exist to list.

sam4s16c

L1 smoke, SIM-DERIVED; no beyond-rubric content. Boot, the PMC clock enables (PCER0) and master-clock-ready status, the UART0 console and the SAM4S PIO LED/button path with ODSR readback are proven by tests/fixtures/sam4s-xplained-smoke.elf and examples/sam4s-xplained/io-smoke.yaml. TC/SPI/TWI/EEFC and the RSTC/SUPC/RTT/WDT/RTC system controller are declared stubs (examples/sam4s-xplained/VALIDATION.md), so no advanced blocks exist to list.

stm32f072

L1 smoke, SIM-DERIVED; no beyond-rubric content. Boot, the F0 clock gates (AHBENR/APB2ENR), the USART1 console and the four-LED GPIO walk with GPIOC ODR readback are proven by tests/fixtures/stm32f072b-disco-smoke.elf and examples/stm32f072b-disco/io-smoke.yaml. USB, TSC and CRS are declared stubs (examples/stm32f072b-disco/VALIDATION.md), so no advanced blocks exist to list.

arty-litex-vexriscv

L1 smoke, SIM-DERIVED; no beyond-rubric content. Boot, the LiteX UART console, the countdown-timer one-shot and the Control-and-Status LED/switch/button CSR are proven by tests/fixtures/arty-litex-smoke.elf and examples/arty-litex-vexriscv/io-smoke.yaml. SPI flash, Ethernet and MMCM are declared stubs (examples/arty-litex-vexriscv/VALIDATION.md), so no advanced blocks exist to list.

mpfs095t

L1 + stock RTOS, SIM-DERIVED. First RV64 part: one PolarFire SoC U54 hart (RV64IMAC M-mode, CLINT in the core) runs unmodified Zephyr v4.1.0 hello_world, synchronization and tests/kernel/common (12 suites, 63/63) from the DDR window on the MMUART1 16550 console (examples/mpfs-disco-kit). PLIC, MSS GPIO/I2C/SPI/QSPI and the system-controller mailbox are declared stubs, so no advanced blocks exist to list.

am6234

L1 + stock RTOS, SIM-DERIVED. First AArch64 part: one Cortex-A53 PE (A64 Armv8.0, EL0-EL3, stage-1 MMU, GICv3 and generic timer in the core) runs unmodified Zephyr v4.3.0 hello_world, synchronization and tests/kernel/common (12 suites, 63/63, user mode at EL0) for sk_am62/am6234/a53 from EL3, on the polled MAIN_UART0 16550 console with the virtual-timer tick (examples/sk-am62). Pad configuration and the mailbox are declared stubs and UART interrupts are not requested, so no advanced blocks exist to list.

r5f52618bgfp

L1 + stock RTOS, SIM-DERIVED. First Renesas RX part: the RXv3 core (integer, DSP accumulators, single-precision FPU; ICU in the core) runs unmodified Zephyr v4.3.0 hello_world, synchronization, tests/kernel/common (12 suites, 61 pass + 1 skip) and tests/drivers/uart/uart_basic_api (7/7 on the interrupt-driven SCI6 driver: RXI6/TXI6/TEI6 as ICU vectors 227/228/229) for ek_rx261 from the reset vector, with the CMT0 tick (examples/ek-rx261). PORT/MPC, DTC and BSC are declared stubs and SCI6 ERI is not requested, so no advanced blocks exist to list.

pic32mz2048efm144

L1 + stock RTOS, SIM-DERIVED. First MIPS part: the MIPS32r2 M5150 core (integer ISA with delay slots, CP0, eight shadow register sets; the PIC32MZ EVIC in the core, EIC mode) runs unmodified upstream NuttX (apache/nuttx 7956039dfd) for an out-of-tree Curiosity PIC32MZ EF 2.0 board from the reset vector: nsh with typed commands over the interrupt-driven UART2, and the full ostest (34 test groups, 0 errors) on the Timer1 tick with software-interrupt context switches (examples/curiosity-pic32mz-ef). Ports/PPS, OSC and CFG are declared stubs, so no advanced blocks exist to list.

vegaboard-ri5cy

L1 smoke, SIM-DERIVED; no beyond-rubric content. Boot, the RV32IMC core (a real multiply is asserted) and the LPUART0 console are proven by tests/fixtures/vegaboard-ri5cy-smoke.elf and examples/vegaboard-ri5cy/io-smoke.yaml. The SCG/PCC clock blocks, INTMUX and low-power timers are declared stubs (examples/vegaboard-ri5cy/VALIDATION.md), so no advanced blocks exist to list.

xck26-rpu

L1 + stock RTOS, SIM-DERIVED. First Cortex-R part: the Zynq UltraScale+ XCK26 RPU core 0 (Cortex-R5F, PMSAv7 16-region MPU, VFPv3-D16 on the pinned tlib engine) runs unmodified Zephyr v4.1.0 hello_world, synchronization (TTC0 match IRQs through the RPU GIC, WFI idle) and tests/kernel/common (12 suites, 61 pass / 2 skip / 0 fail of 63, userspace syscalls through the MPU) on the PS UART1 console (examples/kv260-r5, native CLI). Only the GIC, Cadence UART0/1 and TTC0 are installed (native_rpu_peripherals), so no advanced blocks exist to list.

at32f421g8u7

Bare AT32F421G8U7 (QFN28), reference-derived, no silicon capture. The committed Artery BSP 2.1.9 SDK smoke (examples/at32f421g8u7/io-smoke.yaml, PR-gated coverage-matrix cell, and the vendor::at32f421_contract Rust gate) switches CRM to the 120 MHz PLL, erases and programs the last flash sector, copies four words by DMA1 memory-to-memory and toggles PA5 from real TMR6 interrupts over USART2. The comparator/EXINT line-21 path, ADC ordinary scan with DMA handshake and the shared IWDG expiry reset are exercised only by register-level unit tests.

Advanced peripherals — unit-tested only (no firmware drives them): {'p': 'CMP', 'ev': 'crates/core/src/peripherals/at32f421.rs; crates/core/tests/vendor/at32f421_contract.rs', 'note': 'Mux/polarity/lock and EXINT line-21 rising/falling routing; no firmware drives it yet.'}

mspm0g3507

Local UART0 smoke; no beyond-rubric content. The LP-MSPM0G3507 profile boots a bare-metal image and captures LP_MSPM0G3507_UART0_OK through the TXDATA register path (examples/lp-mspm0g3507-launchpad/evidence/). This proves core execution and register-side TX only; GPRCM, IOMUX/pad routing, UART timing/FIFO, GPIO and the other LaunchPad peripherals are outside the profile (examples/lp-mspm0g3507-launchpad/VALIDATION.md), so no beyond-rubric blocks exist to list.