Possible slowdown in is-prime

[MasterDuke17]

On my system, at HEAD, (^10_000).grep(*.is-prime).tail.say; say now - INIT now reports ~2.2s, but 2018.12 reports ~1.3s.

[lizmat]

$ perl6 --profile -e '(^10_000).grep(*.is-prime).tail.say'

produces a profile for me that spends ~ 99% of the time in is-prime. And I don't think that has been changed at all since then. Could you reproduce this performance difference with something other than is-prime ?

[MasterDuke17]

I haven’t had time to try and reproduce with something different, but I I thought maybe some of the bigint optimizations @jnthn recently landed in MoarVM might have caused this.

…

Sent from my iPhone

On Jun 2, 2019, at 4:53 AM, Elizabeth Mattijsen ***@***.***> wrote: $ perl6 --profile -e '(^10_000).grep(*.is-prime).tail.say' produces a profile for me that spends ~ 99% of the time in is-prime. And I don't think that has been changed at all since then. Could you reproduce this performance difference with something other than is-prime ? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.

[AlexDaniel]

So there's this (GCC-related): #2945

And then this:

¦7d7aaf0c1d2^: «29989␤5.1725937␤»
¦7d7aaf0     : «29989␤6.4280064␤»

7d7aaf0 and this commit specifically: MoarVM/MoarVM@304f568
Haven't checked if it's a compiler difference again, but I guess it's not.

[AlexDaniel]

Then there's also this:

¦787596d00^: «29989␤5.76303753␤»
¦4459364:    «29989␤8.2163708␤»

And the commits in between are:

• c1c108f
• db5becf
• 44e0386
• 5bc6269
• 787596d
• bafa523
• 4459364

Not sure what that is about, could it be that some compiler flags are not being passed correctly? Ping @PatZim