Every Monday, unattended, Raku++ is measured
against what the Raku world published that week — Weekly Challenge solutions, new ecosystem
releases, and the raku-corpus golden battery —
with Rakudo as both control and speed reference.
Updated 2026-10-05 at rakupp 06f9384b5.
93.6%
Weekly Challenge — byte-identical of counted
1696 / 1813
raku-corpus programs matching reference
22
standing PWC mismatches (the open set)
2 green of 29 new
fresh ecosystem releases
What moved this week
3 regressions this week. A program that was byte-identical to the Rakudo reference on 2026-09-28 and is not today.
what differs: line 4 — reference False, rakupp True
7 fixed since 2026-09-28. A program that differed from the Rakudo reference on 2026-09-28 and is byte-identical today — raku-corpus is +6 on the week, at 1696 of 1813 matching.
New solutions since the last run, the open set re-run, and a rotating 10% of past passes.
Hover a chart for that week's numbers; click a name in the legend to switch its line off.
raku-corpus
1,800-odd curated programs against committed Rakudo reference outputs. This line should hold or climb; any dip names the file that broke, above.
Benchmarks — Rakudo time ÷ rakupp time
Above 1× means rakupp is faster. Measured back-to-back on the same runner each week; dashed lines mark a Rakudo release or toolchain change. Ratios are the series — absolute times from shared CI hardware are stored in the ledger but not headlined.
interpreter
native (--exe)
kernel
interp
native
rakudo
speedup-interp
speedup-native
flags
startup
4.3ms
4.1ms
156.9ms
36.5×
38.3×
ok
loopsum
6.9ms
4.1ms
395.4ms
57.3×
96.4×
ok
fib
19.2ms
8.8ms
555.2ms
28.9×
63.1×
ok
strcat
5.1ms
4.4ms
198.7ms
39.0×
45.2×
ok
arrayops
80.6ms
77.8ms
345.1ms
4.3×
4.4×
ok
arraypush
51.6ms
155.2ms
588.0ms
11.4×
3.8×
ok
sortnums
53.0ms
30.2ms
316.5ms
6.0×
10.5×
ok
sortby
67.6ms
42.5ms
281.7ms
4.2×
6.6×
ok
regex
84.4ms
54.5ms
361.4ms
4.3×
6.6×
ok
hash
16.7ms
14.8ms
270.2ms
16.2×
18.3×
ok
hashfill
274.9ms
83.5ms
656.5ms
2.4×
7.9×
ok
bigint
11.3ms
8.8ms
255.0ms
22.6×
29.0×
ok
streq
28.4ms
11.7ms
314.1ms
11.1×
26.8×
ok
textsplit
157.2ms
94.7ms
401.3ms
2.6×
4.2×
ok
rats
29.3ms
164.2ms
445.9ms
15.2×
2.7×
ok
multiwhere
678.8ms
677.4ms
431.5ms
0.6×
0.6×
ok
objects
563.4ms
513.3ms
448.8ms
0.8×
0.9×
ok
mainwhen
9.6ms
4.2ms
583.0ms
60.7×
138.8×
ok
Ecosystem — this week's releases
Every distribution the ecosystem published since the last run, installed and run against its own test suite by rakupp test — with Rakudo as the control for anything that failed.
new
pass
self-fail
dep-fail
timeout
other
upstream-broken
ours
29
2
7
7
13
0
2
18
new
distributions that appeared in the REA index since the previous run. This leg sweeps exactly that set: the ecosystem is measured as it is published, week by week, not re-swept whole.
pass
installed under rakupp test, and the distribution’s own test suite came out green.
self-fail
it installed, but its own suite fails.
dep-fail
a dependency broke first, so the distribution itself never got as far as being tested. The dependency is named in the table below.
timeout
still running after 180 seconds, and killed.
other
it failed for a reason the classifier does not recognise; the raw log line is in the table below.
upstream-broken
of the failures rakupp could be answerable for, the ones Rakudo cannot install either. The run re-tries each of them with zef install under Rakudo: if that fails too, the distribution is broken on its own account and is out of scope here.
ours
…and the ones where Rakudo installs and tests it fine. These are rakupp’s to fix — the single actionable number in the row.
The control runs over self-fail, build-fail, timeout and other only — a dep-fail is charged to the dependency, and shows up under its own name in a later week.