Fix loadtest

[Tyriar]

• Wait for parsing finish before recording time
• Use \r at end of each line so cursor doesn't move
• Report MB/s

[jerch]

@Tyriar Eww, the numbers are not very attractive (2-4 MB/s, but should be >>20 MB/s). Are you sure to measure the right thing? From the code it looks like the measuring contains the Math.random() construction as well, prolly the reason for the lowish numbers? For pure load/throughput on xterm.js I think it is better to construct the data upfront and only measure the time it takes from first to last term.write (last one with a callback with the time calculation).

[Tyriar]

Opps thought I was doing that already. I moved it down and it's similar though 🤔

[jerch]

Hmm thats weird, it misses at least a factor of 5. Were there recent perf regressions I am not aware of?

Edit: Hmm, is there something wrong with syncScrollArea? It accounts for half of the runtime here in devtools, haha.

[Tyriar]

syncScrollArea will take a decent amount of time until scrollback is filled, but after that it's still 5-10MB/s

[Tyriar]

Oh actually it is still showing up, with GCs happening as well, will investigate