Investigate increased CI time

[alexytsu]

The change in #1259 has caused a significant increase in CI times for this repo.

https://github.com/filecoin-project/builtin-actors/actions/runs/4527273009/jobs/7972963894 took around 18 minutes and has doubled (34 minutes) both when it landed on master: https://github.com/filecoin-project/builtin-actors/actions/runs/4538508299/jobs/7997478197 and when it ran on the PR branch (35 minutes) https://github.com/filecoin-project/builtin-actors/actions/runs/4528315067/jobs/7974981200

This PR reverts commit 424669b to see if that commit caused the increase in CI times. (It does)

By switching to a generic trait, the compiler will monomorphize the TestVM code. This should not in fact affect test execution times but running this PR to make sure. Locally I see that the time taken to build the TestVM tests has more than doubled ~2:30 -> 6 minutes whilst time taken to run the tests remains roughly the same ~4 minutes.

[codecov-commenter]

Codecov Report

Merging #1262 (308b6b7) into master (424669b) will decrease coverage by 0.03%.
The diff coverage is 100.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1262      +/-   ##
==========================================
- Coverage   90.90%   90.88%   -0.03%     
==========================================
  Files         133      133              
  Lines       26754    26721      -33     
==========================================
- Hits        24322    24285      -37     
- Misses       2432     2436       +4     

Impacted Files	Coverage Δ
actors/market/src/testing.rs	96.03% <100.00%> (-0.05%)	⬇️
state/src/check.rs	86.35% <100.00%> (+0.26%)	⬆️
test_vm/src/lib.rs	82.66% <100.00%> (-0.50%)	⬇️
test_vm/src/util.rs	99.91% <100.00%> (-0.01%)	⬇️

... and 2 files with indirect coverage changes

[alexytsu]

From my local timings, almost all the increase in time can be attributed to building the tests in the test_vm package

	master	nv19	delta
cargo test —workspace —no-run	cargo test --workspace --no-run 2716.88s user 116.09s system 830% cpu 5:40.96 total cargo test --workspace --no-run 2948.45s user 141.17s system 774% cpu 6:38.89 total	cargo test --workspace --no-run 1002.52s user 87.83s system 718% cpu 2:31.72 total cargo test --workspace --no-run 995.24s user 89.34s system 705% cpu 2:33.75 total	~4 minutes increase
with prebuilt tests cargo test --workspace	cargo test --workspace 336.27s user 6.37s system 153% cpu 3:43.14 total cargo test --workspace 339.78s user 7.77s system 123% cpu 4:40.51 total	cargo test --workspace 338.90s user 6.96s system 125% cpu 4:35.64 total cargo test --workspace 337.49s user 6.47s system 154% cpu 3:43.32 total	~no increase
only building tests in test_vm package	cargo test --no-run 2148.08s user 60.56s system 909% cpu 4:02.95 total	cargo test --no-run 411.78s user 32.10s system 662% cpu 1:06.96 total	~3 minutes increase

[Stebalien]

Try running cargo bloat.

[Stebalien]

Hm. And twiggy monos. (wasm only, unfortunately)

[Stebalien]

It looks like removing the opt-level=2 on the test vm (

builtin-actors/Cargo.toml

Line 125 in 424669b

opt-level = 2

) reduces the build time in half, at least for me. However, that'll probably slow down the tests.

[alexytsu]

removing the opt-level=2

Significantly decreases build time for me too but it is far outweighed by increase in test running time

(for both opt-level 0 and 1 neither seems net faster)