Middle-end: Optimize reversed CRC table-based implementation
GCC now generates faster table-based CRC implementations by pre-reversing the lookup table.
GCC’s middle-end now optimizes reversed CRC calculations by generating a reversed polynomial lookup table directly. This eliminates the need for explicit bit reflection before and after the CRC computation, improving code generation for all targets using table-based reversed CRCs. The change removes the overhead of reflecting input data and CRC values, resulting in faster CRC calculations.
In Details
This commit introduces calculate_reversed_crc, assemble_reversed_crc_table, and generate_reversed_crc_table in expr.cc. It modifies expand_reversed_crc_table_based to use these new functions, removing the gen_reflecting_code parameter. The change avoids generating explicit bit reflection code. The performance gain arises from simplifying the IR exposed to the back end.
For Context
Cyclic Redundancy Checks (CRCs) are algorithms used to detect accidental changes to raw data. They are often implemented using lookup tables for speed. Different CRC variants exist, including reversed CRCs. This commit optimizes the implementation of reversed CRC calculations in GCC by pre-calculating a reversed lookup table, which avoids costly bit reflection operations during runtime. This results in faster CRC calculations.