dzbarsky wrote: > TBH, I haven't looked at this check before. Could you please describe how we > store data before and now.
Previously, the generator emitted 6,355 fixed-size entries containing one source UTF32 value and 19 replacement UTF32 values. Every entry reserved enough space for the longest replacement, and skeleton() converted the replacement to UTF-8 on each lookup. Now each unique replacement is stored once in a shared UTF-8 array. Each sorted entry contains the source code point and a 16-bit offset and length into that array. skeleton() appends those bytes directly. This reduces the table from 508,400 to 63,327 bytes. https://github.com/llvm/llvm-project/pull/202626 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
