Aarch64: Split up QLF_NIL meanings, and other fixes
Splits AARCH64_OPND_QLF_NIL into NIL/ERR/UNKNOWN/UNUSED qualifiers, adds sequence-length checks, and improves code quality.
This series splits the overloaded AARCH64_OPND_QLF_NIL operand qualifier into four distinct qualifiers: NIL, ERR, UNKNOWN, and UNUSED. This eliminates the need for the F_STRICT opcode flag to differentiate between NIL and UNKNOWN. The series also introduces checks to enforce correct qualifier sequence lengths, which would have caught issues addressed in earlier patches, and includes miscellaneous code quality improvements. The proposer intends to commit the changes next week if no issues are raised.
- proposer
Proposes splitting AARCH64_OPND_QLF_NIL into four qualifiers and adding sequence-length checks.
“The main goal of this series is to split up the separate meanings of AARCH64_OPND_QLF_NIL into four operand qualifiers - the existing NIL and ERR qualifiers, and new UNKNOWN and UNUSED qualifiers.”
In Details
This series refactors the AARCH64 instruction decoder, specifically the handling of operand qualifiers. The existing AARCH64_OPND_QLF_NIL was overloaded, requiring the F_STRICT opcode flag to distinguish meanings. This change clarifies the semantics and improves maintainability of the instruction decoding tables. The new qualifiers and sequence-length checks aim to prevent future errors in instruction definition.
For Context
When a CPU executes an instruction, the instruction needs to be decoded to determine what operation to perform and on what data. This involves parsing the instruction's binary representation and identifying its different parts, such as the opcode (the operation to perform) and the operands (the data to operate on). Operand qualifiers provide additional information about the operands. This series cleans up how these qualifiers are handled in the AArch64 architecture, making the code easier to understand and maintain.