These are simply some results from my dev machine, just to give a rough idea of what the impact is certain tweaks:
| Kotlin compiler version | kotlinx-html version | hacked kotlinx-html-jvm | reduced nesting | K2 compiler | Kotlin compiler args | realCPU | userCPU | peakMEM | |
|---|---|---|---|---|---|---|---|---|---|
| 1.6.20 | 0.7.3 | no | no | no | none | 1:14.58 | 1.38 | 102368 | |
| 1.7.0 | 0.7.3 | no | no | no | none | 0:48.03 | 1.27 | 101316 | |
| 1.7.22 | 0.7.3 | no | no | no | none | 0:48.97 | 1.27 | 101932 | |
| 1.7.22 | 0.8.0 | no | no | no | none | 0:48.32 | 1.37 | 102016 | |
| 1.8.0 | 0.8.0 | no | no | no | none | 0:48.45 | 1.34 | 100440 | |
| 1.8.10 | 0.8.0 | no | no | no | none | 0:50.92 | 1.41 | 103060 | |
| 1.8.20-Beta | 0.8.0 | no | no | no | none | 0:50.87 | 1.33 | 101252 | |
| 1.8.20-RC | 0.8.0 | no | no | no | none | 0:49.83 | 1.32 | 102588 | |
| 1.8.20-RC | 0.8.0 | yes | no | no | none | 0:30.80 | 1.23 | 102904 | |
| 1.8.20-RC | 0.8.0 | no | no | yes | "-Xuse-k2", "-Xbackend-threads=0", "-Xuse-fast-jar-file-system", "-Xlambdas=indy" | 2:03.16 | 1.93 | 104540 | |
| 1.8.20-RC | 0.8.0 | yes | no | yes | "-Xuse-k2", "-Xbackend-threads=0", "-Xuse-fast-jar-file-system", "-Xlambdas=indy" | 0:24.53 | 1.31 | 100968 | |
| 1.8.20-RC | 0.8.0 | yes | no | yes | "-Xuse-k2" | 0:26.85 | 1.27 | 104620 | |
| 1.8.20-RC | 0.8.0 | yes | yes | yes | "-Xuse-k2", "-Xbackend-threads=0", "-Xuse-fast-jar-file-system", "-Xlambdas=indy" | 0:23.13 | 1.27 | 102436 | |
| 1.8.20-RC | 0.8.0 | no | yes | yes | none | 0:36.71 | 1.35 | 102324 |
Some conclusions:
- kotlin 1.6.20 -> recent: 34.67%
- hacked kotlinx-html-jvm: 38.00%
- use compiler flags incl K2: 20.45%
- reduce nesting: 5.71%
Combining all improvements (fastest vs slowest): 69.44% speedup