Parallel VTI diagnostics

[yoyo930021]

In my Macbook pro 13 2020, It's 10x faster.
The VTI can use in production when pairing up #2325 #2328

gtime result

old:
834.59 user 29.64 system 11:21.12 elapsed 126% CPU (0avgtext+0avgdata 1409020maxresident)k

new:
0.26 user 0.12 system 1:47.03 elapsed 0% CPU (0avgtext+0avgdata 37980maxresident)k

[andrewisaburden]

@yoyo930021

I tested on Ubuntu, and also #2328, and also combined both.

Old:
1185.37user 16.35system 12:55.15elapsed 155%CPU (0avgtext+0avgdata 1034768maxresident)k

#2329 only:
0.16user 0.03system 2:20.26elapsed 0%CPU (0avgtext+0avgdata 40492maxresident)k

#2328 only:
1108.55user 15.17system 11:56.96elapsed 156%CPU (0avgtext+0avgdata 1024188maxresident)k

Both #2329 #2328:
0.20user 0.02system 2:15.25elapsed 0%CPU (0avgtext+0avgdata 40328maxresident)k

The performance from #2329 is very nice, but #2328 seems quite minor (maybe non-existent, if there is variation from run to run).

Good work.

[yoyo930021]

@yoyo930021

I tested on Ubuntu, and also #2328, and also combined both.

Old:
1185.37user 16.35system 12:55.15elapsed 155%CPU (0avgtext+0avgdata 1034768maxresident)k

#2329 only:
0.16user 0.03system 2:20.26elapsed 0%CPU (0avgtext+0avgdata 40492maxresident)k

#2328 only:
1108.55user 15.17system 11:56.96elapsed 156%CPU (0avgtext+0avgdata 1024188maxresident)k

Both #2329 #2328:
0.20user 0.02system 2:15.25elapsed 0%CPU (0avgtext+0avgdata 40328maxresident)k

The performance from #2329 is very nice, but #2328 seems quite minor (maybe non-existent, if there is variation from run to run).

Good work.

Thanks your test.
Do you yarn link vls in vti package when merging two PR and testing?

The VTI package is not using local vls in default.

[andrewisaburden]

@yoyo930021

I tried again:
cd vetur
yarn && yarn compile
cd server
yarn link
cd vetur/vti
yarn link vls
yarn compile

Then ran again:
0.15user 0.05system 1:44.48elapsed 0%CPU (0avgtext+0avgdata 42360maxresident)k

That's for both #2329 and #2328, so maybe the first time I missed a step (there is some improvement from ~2.15 to ~1.44).

Both looking good!

[yoyo930021]

@yoyo930021

I tried again:
cd vetur
yarn && yarn compile
cd server
yarn link
cd vetur/vti
yarn link vls
yarn compile

Then ran again:
0.15user 0.05system 1:44.48elapsed 0%CPU (0avgtext+0avgdata 42360maxresident)k

That's for both #2329 and #2328, so maybe the first time I missed a step (there is some improvement from ~2.15 to ~1.44).

Both looking good!

This PR is violent.
Otherwise only #2328 , It should be twice as fast. 🤣

[yoyo930021]

I will close this PR.
This PR takes a lot of resources, and now we have a better way to do it.

I test a branch from merge #2328 and #2374 .

92.62user 3.35system 1:16.66elapsed 125%CPU (0avgtext+0avgdata 1169040maxresident)k
0inputs+0outputs (112major+688386minor)pagefaults 0swaps

It's an amazing result.