Skip to question
asic.fyi

Hardware interview practice

Parse a bounded Roman board revision

MediumFirmware AlgorithmsSystemVerilog

Parse an explicitly sized electrically erasable programmable read-only memory (EEPROM) field as a canonical Roman numeral from I through MMMCMXCIX without allocation or out-of-range reads.

Starting point

Question code

typedef enum { ROMAN_OK, ROMAN_EINVAL, ROMAN_ERANGE } roman_rc_t;
roman_rc_t roman_to_u16(const uint8_t *buf,
                        size_t len,
                        uint16_t *value);
// len is 1..15; a successful value is 1..3999.
Reviewed example

Work through one case

Input
bytes="MCMXCIV", length=7
Expected output
success; revision=1994

The subtractive pairs CM, XC, and IV are legal, and re-encoding 1994 reproduces the exact input, proving canonical spelling.

What to cover

Requirements

  1. Accept only uppercase I, V, X, L, C, D, and M and the six standard subtractive pairs.
  2. Return invalid-input errors before range errors, except that a fully parsed value above 3999 returns the range error.
  3. Reject noncanonical spellings by re-encoding the parsed value and comparing every input byte.
  4. Do not call strlen or modify the caller's output until the complete parse succeeds.
Continue practicing
Firmware AlgorithmsFind device-tree depth without recursionFirmware AlgorithmsAdd decimal lists from a fixed poolFirmware AlgorithmsEnumerate calibration sums with fixed memory