> On Feb 25, 2020, at 7:05 AM, Matthew Flatt <[email protected]> wrote:
>
> * CS has a single heap with a single-threaded, stop-the-world GC ---
> so allocation and GC easily become a bottleneck.
>
> If GHC's experience is any guide, making the GC itself multithreaded
> may address much of this problem.
>
> Locks on shared system data structures may also be a significant
> obstacle to CPU utilization with places in CS, but I'm not sure.
FWIW some quick timings on a Pollen render of practicaltypography.com. Though
extra cores have diminishing net returns under Racket BC, the returns are still
positive. Under Racket CS, by contrast, net performance degrades with more than
4 cores.
Racket BC
single core
real 4m21.191s
user 3m37.940s
sys 0m42.388s
parallel @ 2 cores
real 2m46.235s
user 4m22.160s
sys 0m56.270s
parallel @ 3 cores
real 1m54.134s
user 4m10.330s
sys 0m54.533s
parallel @ 4 cores
real 1m43.055s
user 4m46.933s
sys 1m5.948s
parallel @ 6 cores
real 1m34.783s
user 6m8.522s
sys 1m32.125s
parallel @ 8 cores
real 1m18.137s
user 6m24.778s
sys 1m38.617s
parallel @ 12 cores
real 1m14.924s
user 8m30.239s
sys 2m14.671s
Racket CS
single core
real 5m1.422s
user 4m16.300s
sys 0m44.253s
parallel @ 2 cores
real 3m25.016s
user 4m45.385s
sys 0m54.634s
parallel @ 3 cores
real 2m52.780s
user 4m57.951s
sys 1m3.184s
parallel @ 4 cores
real 2m42.471s
user 5m22.796s
sys 1m17.889s
parallel @ 6 cores
real 2m44.513s
user 6m26.700s
sys 1m54.549s
parallel @ 8 cores
real 2m56.782s
user 8m4.029s
sys 2m58.554s
parallel @ 12 cores
real 3m2.116s
user 9m34.846s
sys 5m5.443s
--
You received this message because you are subscribed to the Google Groups
"Racket Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/racket-users/3D11BBCF-B0FE-473B-8997-09B7CB60D761%40mbtype.com.
