Ty Overby

I don't remember contributing, but I approve

I suspect that there’s something super linear going on here because a from-scratch separate-compilation build of the same project spends ~20x less time doing dominance frontier calculation

> Can you share some code reproducing the issue? Sadly I can not > What's the size of the bytecode file? 112MB > Have you tried `--disable inline`? I have...

From debugging this a few weeks ago, I remember that the primary issue was that a region of the dominator tree is effectively just a singly-linked list (a node in...

>> Have you tried --disable inline? >I have not, though I can try it out and see what happens. I should point out that it's using direct-compilation, but doesn't use...

> remember that the primary issue was that a region of the dominator tree is effectively just a singly-linked list Because of this, I think the dominator tree depth was...

# THE PLOT THICKENS Passing `--sourcemap` makes basically _every_ pass in the compiler dramatically slower, including codegen. ## With `--sourcemap` ``` parsing js: 0.19 Start parsing... parsing: 13.50 Start Optimizing......

The same is true for jsoo link **with --sourcemap**: 5.1s **without --sourcemap**: 0.5s **without --sourcemap and all input .js files don't have a sourcemap comment**: 0.08s

This one doesn't surprise me as much; most of what I _want_ jsoo-link to do is stitch-up sourcemaps correctly, but I wasn't expecting a full order of magnitude

> Can you check the impact of https://github.com/ocsigen/js_of_ocaml/pull/1313 on these timings ? I suspect it won't change much. I'll try it out