Item operation perf in _GetCopyToOutputDirectoryItemsFromTransitiveProjectReferences about msbuild HOT 8 OPEN

On a large build with 900 projects this target consumes 28 minutes cumulatively (wall clock time is less because the build is massively parallel)

For the largest project it takes 15 seconds:

from msbuild.

Is AssignTargetPath called twice for @(_CompileItemsToCopy)? Once for transitive, once for this project? I suppose it's benign and idempotent?

from msbuild.

To clarify, out of 15 seconds for that target, 12 seconds is calling the MSBuild task on the project references, so the remaining 3.3 seconds must be the overhead, perhaps partially attributable to item manipulation:

I might be totally wrong though, need to measure, and perhaps in a tight loop in a unit-test using BenchmarkDotNet.

from msbuild.

The way to measure perf here would be to build a large-ish solution such as MSBuild itself, or Roslyn, and query for $target GetCopyToOutputDirectoryItems $time in both binlogs, and see what difference does it make. We might or might not see a difference in the total build duration, and I'd be curious if we see a diff or not. But even if the time for the targets comes down, it'll be a good thing (because we'll be doing less compute and less allocations).

from msbuild.

Apparently the fact that this makes the condition more complicated is more impactful than removing a line of code:

(Current MSBuild on the right)

from msbuild.

Huh, that's counter-intuitive, but thanks for checking!

from msbuild.

if you still have both binlogs, can you send them to me? and your branch with the prototype

from msbuild.

Reopening because I forgot to build -c Release until I've done more testing

from msbuild.