add msquic

[ktprime]

can add the result of different quic benchmark?

[victorstewart]

funny i just check out your hash table the other day.

msquic can't be added here because it handles its own sockets/event loop and there's no way to bifurcate out that feature-- thus this wouldn't be a true apples<->apples comparison, which as the point... identical event loops and harvesting/feeding code, only differentiation being the library used.

msquic also uses at least 2 threads.

when i first wrote this benchmark to decide which library to use, i did run a static binary of msquic and jailed it to 1 logical core.. and it performed very well, but i don't off hand remember what rank. but it wasn't #1.

i haven't looked at this in a while or updated the library versions it builds against or the performance results, but i should get around to that.

[victorstewart]

i'll speak with the msquic maintainer about roping in a binary build here though

[ktprime]

I only use google quiche in my projcet. from my benchmark it can reach 520MB/s in my pc intel 9700. the benchmark client/server is a simple single-thread ping-pong without using http3/encrytion on wsl2.

-----------------------------------------------------------------------TPS: 69632/sec, QPS: 128119/sec, RBW 500.47, SBW: 500.47 MB/sec Cons: 1
-----------------------------------------------------------------------TPS: 75163/sec, QPS: 134783/sec, RBW 526.50, SBW: 526.50 MB/sec Cons: 1
-----------------------------------------------------------------------TPS: 75554/sec, QPS: 134097/sec, RBW 523.82, SBW: 523.82 MB/sec Cons: 1
-----------------------------------------------------------------------TPS: 75150/sec, QPS: 133255/sec, RBW 520.53, SBW: 520.53 MB/sec Cons: 1