What problem does it solve? It prevents benchmark methodology errors when measuring stock upstream SDCC (not z88dk's patched zsdcc), ensuring published numbers use the correct binary, version banner, size rule, and wiki row placement. ## Core Features & Use Cases - Correct toolchain separation: Distinguishes /usr/local/bin/sdcc (vanilla, __sdcccall(1) ABI) from bin/z88dk-zsdcc (ABI 0 via z88dk patch), so numbers never land in the wrong RESULT block or wiki row. - Pinned, reproducible builds: Checks out svn r16639 (SDCC 4.6.0), publishes the #16608 banner, and lists configure flags such as --disable-f8l-port to avoid sdasf8 build failures. - Standardized compile, size, and tick measurement: Uses sdcc -mz80 -DSTATIC -DTIMER, makebin -p, map-section sum sizing, and z88dk-ticks with start/end addresses resolved from .sym/.map files. - Use Case: Re-measuring the SDCC rows in support/benchmarks/*/sdcc/readme.txt for a 4.6.0 release and updating the Benchmarks wiki without contaminating Z88DK/SDCC rows. ## Quick Start Ask the assistant to re-run the vanilla SDCC benchmark for a given support/benchmarks directory and update its sdcc/readme.txt RESULT block with the SDCC 4.6.0 #16608 tag.