I detected a huge performance regression in Arraymancer on my benchmark: https://github.com/mratsim/Arraymancer/blob/master/benchmarks/ex01_xor.nim
It used to run in 10sec but now takes about 30 seconds to run (see https://github.com/mratsim/Arraymancer/issues/221 for screenshots).
I've narrowed down the issue to this commit: https://github.com/nim-lang/Nim/commit/358709e9cbc4ed3e084dc6d1db313a862b8356e4, the previous commit worked OK: https://github.com/nim-lang/Nim/commit/2c1f1f21bf93293940624898cfc37f2c3d42711b.
I have several {.noInit.} in the library to avoid initialization where not needed but it seems like that is now completely ignored.
I confirm the fix. Actually now devel is about 30% faster than before, some perf improvement were hidden by this regression 馃巻 .
Most helpful comment
I confirm the fix. Actually now devel is about 30% faster than before, some perf improvement were hidden by this regression 馃巻 .