try multicore

[anuragsoni]

This was done just out of curiosity as I wanted to try the multicore compiler. I'm leaving this as a draft here in-case you want to try this on the multicore branch and/or use it for tests as the multicore project progresses :)

[anuragsoni]

Some rough numbers from my naive wrk based testing

Test command: ./_build/default/src/examples/echo.exe -j 128

Numbers from the thread based implementation on the current release

➜  tiny_httpd git:(master) ✗ wrk -t4 -c96 -d15s --latency "http://127.0.0.1:8080/hello"
Running 15s test @ http://127.0.0.1:8080/hello
  4 threads and 96 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency     2.33ms    3.01ms 116.84ms   94.27%
    Req/Sec    11.82k     1.49k   43.51k    98.17%
  Latency Distribution
     50%    1.97ms
     75%    3.04ms
     90%    4.50ms
     99%    8.14ms
  706580 requests in 15.10s, 1.38GB read
Requests/sec:  46794.26
Transfer/sec:     93.45MB

Numbers from this branch with domainslib:

➜  tiny_httpd git:(try-multicore) wrk -t4 -c96 -d15s --latency "http://127.0.0.1:8080/hello"
Running 15s test @ http://127.0.0.1:8080/hello
  4 threads and 96 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    97.24us  689.25us  41.86ms   99.35%
    Req/Sec    56.87k     7.97k   70.87k    62.33%
  Latency Distribution
     50%   54.00us
     75%   65.00us
     90%   74.00us
     99%  321.00us
  1696983 requests in 15.05s, 3.31GB read
Requests/sec: 112752.17
Transfer/sec:    225.17MB

Both tests are run on a 4 core intel 8th gen 8550u cpu.

[craff]

This is impressive speed up. Did you try with two distant machine?

[c-cube]

You can use moonpool for that now, I have a bench here :)