Summary:
Compilation is a mixture of I/O and CPU work, so there's some good optimisation potential even on a single thread by running the promises concurrently. The downside is that the output is a bit garbled, but it's quite a bit faster for me.
To test this, I applied this diff so I could just `time yarn start` for an entire run.
```
diff --git a/xplat/sonar/src/init.js b/xplat/sonar/src/init.js
--- a/xplat/sonar/src/init.js
+++ b/xplat/sonar/src/init.js
@@ -80,4 +80,5 @@
dispatcher(store, logger);
// make init function callable from outside
window.Flipper.init = init;
+ window.close();
});
```
## Before
**Cold Start**
```
696.89user 93.21system 4:21.44elapsed 302%CPU (0avgtext+0avgdata 1111448maxresident)k
1416inputs+372824outputs (3major+28155897minor)pagefaults 0swaps
```
**Hot Start**
```
6.87user 1.62system 0:17.32elapsed 49%CPU (0avgtext+0avgdata 249264maxresident)k
216inputs+182912outputs (0major+251426minor)pagefaults 0swaps
```
## After
**Cold Start**
```
736.86user 95.47system 1:53.13elapsed 735%CPU (0avgtext+0avgdata 970616maxresident)k
42864inputs+356776outputs (42major+25299531minor)pagefaults 0swaps
```
**Hot Start**
```
6.85user 1.60system 0:15.93elapsed 53%CPU (0avgtext+0avgdata 269420maxresident)k
2592inputs+197904outputs (0major+260099minor)pagefaults 0swaps
```
## Summary
It's beneficial in both cases, but the wall clock time for cold start goes from *4:21.44* to *1:53.13* for me (mind that this is a single sample). This is a 2.3x speed-up!
Reviewed By: jknoxville
Differential Revision: D16359215
fbshipit-source-id: 62257ec3bdefbf98356f5ac9418c4906f7cb4b1b
10559a188f
