Faster CopyFrom(+Ascii)#585
Conversation
benaadams
force-pushed
the
copyfrom-perf-finalized
branch
from
d1f69eb to
3cd7793
Compare
Fixed finalizer block return; by reusing Create rather than .ctor
benaadams
force-pushed
the
copyfrom-perf-finalized
branch
from
3cd7793 to
a3540d5
Compare
|
I was kind of kicking myself for merging the earlier version of this PR last week, so I'm a wary of merging it again. When I tested this before, I remember seeing a little over a 1% RPS improvement in the plaintext benchmark. Your own numbers show about a 2.5% improvement. Either way, I don't think this improvement is dramatic enough for taking a perf related change at this point. I'm worried that there might be yet another undiscovered bug. I feel bad about this because I've pretty much already reviewed and accepted this PR. It's definitely something I will likely merge post 1.0. |
|
Np, completely understand :) |
|
On the plus side, merging help identify and issue and showed the finalizer was doing critical work - so it ended up being a good thing 😜 |
3 participants
Fixed finalizer block return issue in #527 by making the finalizer using MemoryPoolBlock2.Create rather than .ctor directly
Resolves #579