[browser][perf] Blazor startup regression 2023-04-21

[maraf]

• https://msit.powerbi.com/groups/me/reports/10265790-7e2e-41d3-9388-86ab72be3fe9/ReportSection5ddef44dec0ed1074279?experience=power-bi
• Affected run from commit e795ec77ab3737b54c8ab3e92f87a02d29eaeabb, previous run from commit e21ecb3dc51338f4d8288a844cb8491f550e4cd1
• Runtime changes between 2023-04-14 - 2023-04-21
• Suspicious changes
  • [wasm] Bump emscripten to 3.1.34 #83998
  • [browser] Minimal blazor.boot.json integration #84296
  • [blazor] Boot handled by dotnet.js aspnetcore#47752

cc @radekdoulik

[ghost]

Tagging subscribers to 'arch-wasm': @lewing
See info in area-owners.md if you want to be subscribed.

Issue Details

---

• https://msit.powerbi.com/groups/me/reports/10265790-7e2e-41d3-9388-86ab72be3fe9/ReportSection5ddef44dec0ed1074279?experience=power-bi
• Affected run from commit e795ec77ab3737b54c8ab3e92f87a02d29eaeabb, previous run from commit e21ecb3dc51338f4d8288a844cb8491f550e4cd1
• Runtime changes between 2023-04-14 - 2023-04-21
• Suspicious changes
  • [wasm] Bump emscripten to 3.1.34 #83998
  • [browser] Minimal blazor.boot.json integration #84296
  • [blazor] Boot handled by dotnet.js aspnetcore#47752

cc @radekdoulik

Author:	maraf
Assignees:	-
Labels:	arch-wasm, tenet-performance
Milestone:	8.0.0

[maraf]

I wasn't able to reproduce the regression locally. Local runs for commits from mentioned PRs had variations

[lewing]

Were there any System.Text.Json commits nearby in the history but outside the range?

[pavelsavara]

Right now I think that the regression could be caused by postponing the start of the DLL downloads until dotnet.runtime.js is loaded , which happened in dotnet/aspnetcore#47752

I hope I mostly fixed that in my #89857

[pavelsavara]

In Net7 Blazor didn't have any download throttling at all.
After unification of the loader, we throttle downloads to max 16 in parallel.
We do it in order to make the startup more stable in low resource environments, such as codespaces and helix.
Maybe also mobile phones benefit, but I don't have data for that.

runtime/src/mono/wasm/runtime/loader/globals.ts

Line 70 in c6db89c

maxParallelDownloads: 16,

But maybe it's impacting the perf benchmark because it's running on beefy machine with local web server.

We could

• change the default to something more aggressive, like 32 parallel or even unlimited (and just set it low for our unit tests)
• admit that measuring startup time on fast machine with local server is not realistic. Leave it as is, or even add server side network throttling.
• disable our throttling just for Blazor

Note, I'm not 100% sure that this is the regression we see here.

[maraf]

The throttling was still "disabled" in this regression https://github.com/maraf/aspnetcore/blob/4b7087f7dd984b2a7d882ff8ad3305a5959f7f66/src/Components/Web.JS/src/Platform/Mono/MonoPlatform.ts#L252C27-L252C34

EDIT: It is STILL disabled https://github.com/dotnet/aspnetcore/blob/a0db11ba7c3ae217e9745976056f10cb2a7dafde/src/Components/Web.JS/src/Platform/Mono/MonoPlatform.ts#L179

[maraf]

Changes from #90388 seem promising

[maraf]

We have gained something back with mentioned PR matching values from March