BenchmarkFibPlaceholderRun calls compileFile(b, "scripts/factorial.ts"). The
repository also has scripts/fib.ts, which nothing benchmarks. The function's
own comment says "(currently placeholder)"; nothing downstream does.
The timed region is right and the workload is real — recursion, fused -- on a
parameter, ~5M calls per iteration, flat in b.N. The numbers are sound; only
the label is wrong.
The label misleads in practice. Anyone reasoning about what a "Fib" delta should
do reasons about the wrong source, which is how I came to attribute #43's +30.9%
to an array-index guard when factorial.ts contains no arrays (see #52).
Renaming the benchmark to what it runs costs a discontinuity in any stored
timeline keyed on the name. Pointing it at fib.ts changes the workload. I would
rename: the recursion workload is worth keeping and nothing else in the suite
covers it.
BenchmarkFibPlaceholderRuncallscompileFile(b, "scripts/factorial.ts"). Therepository also has
scripts/fib.ts, which nothing benchmarks. The function'sown comment says "(currently placeholder)"; nothing downstream does.
The timed region is right and the workload is real — recursion, fused
--on aparameter, ~5M calls per iteration, flat in
b.N. The numbers are sound; onlythe label is wrong.
The label misleads in practice. Anyone reasoning about what a "Fib" delta should
do reasons about the wrong source, which is how I came to attribute #43's +30.9%
to an array-index guard when
factorial.tscontains no arrays (see #52).Renaming the benchmark to what it runs costs a discontinuity in any stored
timeline keyed on the name. Pointing it at
fib.tschanges the workload. I wouldrename: the recursion workload is worth keeping and nothing else in the suite
covers it.