[compiler-v2] Test cases reduced from the framework showcasing need for stack optimizations

[vineethk]

Description

In this PR, we add a bunch of test cases that show opportunities for optimization of generated stack-based bytecode. These tests were reduced from aptos-framework.

I have put them in a folder called eager-pushes because they can generally be resolved by eagerly pushing the operand on to the stack at the appropriate point.

How Has This Been Tested?

New tests are added.

Key Areas to Review

Tests and the generated bytecode. It might be helpful to look at the stackless bytecode generated as well.

Type of Change

• Tests

Which Components or Systems Does This Change Impact?

• Move Compiler

[trunk-io]

⏱️ 55m total CI duration on this PR

Job	Cumulative Duration	Recent Runs
rust-move-unit-coverage	16m	🟩
rust-move-unit-coverage	10m	🟩
rust-move-tests	9m	🟩
rust-move-tests	9m	🟩
check	4m	🟩
rust-cargo-deny	4m	🟩 🟩
check-dynamic-deps	1m	🟩 🟩
general-lints	1m	🟩 🟩
semgrep/ci	41s	🟩 🟩
file_change_determinator	22s	🟩 🟩
permission-check	5s	🟩 🟩
permission-check	5s	🟩 🟩
permission-check	4s	🟩 🟩

_{settings ⋅ feedback ⋅ docs ⋅ learn more about trunk.io}

[vineethk]

• [compiler-v2] Test cases reduced from the framework showcasing need for stack optimizations #14800 👈
• [compiler-v2] [waiting for comparison testing before merging] Inefficient loads: window peephole optimization #14796
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

Join @vineethk and the rest of your teammates on Graphite

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 59.8%. Comparing base (3c9937d) to head (f384f11).

Additional details and impacted files

@@                      Coverage Diff                       @@
##           vk/variable-window-peephole   #14800     +/-   ##
==============================================================
- Coverage                         59.8%    59.8%   -0.1%     
==============================================================
  Files                              853      853             
  Lines                           207923   207923             
==============================================================
- Hits                            124365   124348     -17     
- Misses                           83558    83575     +17     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[brmataptos]

The eager_load_03.exp disassembly looks as good as we can do without IPA. I guess having another test doesn't really hurt, but I'm not sure why it's here.

Others do look like they could benefit from a more stack-aware stack-code-generation phase, that builds expression trees and rearranges computations to push values on the stack in the order needed for the next operation.

[brmataptos]

Without interprocedural analysis, this code looks pretty optimal. You can't load Arg0 earlier because of the call to bar(&mut Arg0). Meanwhile, loc0 comes from the call to one() and you have to rearrange the two arguments before calling baz. A swap operation would be useful here, but without that this looks good.

Of course, if you can inline the other calls you can do much better.

[vineethk]

Yes, this was a test case to show that you should not pre-load Arg0 because of the &mut to it, like you point out.

I should have mentioned in the PR description that there are some negative test cases as well - this one is actually handcrafted and not reduced from framework code.

[github-actions]

✅ Forge suite realistic_env_max_load success on a9afe5bbd161f86281a6a8ca898905e3133f5e8a

two traffics test: inner traffic : committed: 14388.56 txn/s, latency: 2762.45 ms, (p50: 2700 ms, p70: 2700, p90: 3000 ms, p99: 3300 ms), latency samples: 5470960
two traffics test : committed: 99.94 txn/s, latency: 1603.66 ms, (p50: 1400 ms, p70: 1400, p90: 1500 ms, p99: 9400 ms), latency samples: 1800
Latency breakdown for phase 0: ["MempoolToBlockCreation: max: 2.021, avg: 1.585", "ConsensusProposalToOrdered: max: 0.326, avg: 0.294", "ConsensusOrderedToCommit: max: 0.362, avg: 0.349", "ConsensusProposalToCommit: max: 0.653, avg: 0.643"]
Max non-epoch-change gap was: 0 rounds at version 0 (avg 0.00) [limit 4], 1.00s no progress at version 29367 (avg 0.20s) [limit 15].
Max epoch-change gap was: 0 rounds at version 0 (avg 0.00) [limit 4], 8.48s no progress at version 2603800 (avg 8.48s) [limit 15].
Test Ok

• Grafana dashboard
• Humio Logs
• Axiom Logs
• Validator CPU Profile
• Fullnode CPU Profile
• Test runner output
• Test run is land-blocking

[github-actions]

✅ Forge suite framework_upgrade success on f38ec72e975a4dff4e9919e7a1d8118a75858bab ==> a9afe5bbd161f86281a6a8ca898905e3133f5e8a

Compatibility test results for f38ec72e975a4dff4e9919e7a1d8118a75858bab ==> a9afe5bbd161f86281a6a8ca898905e3133f5e8a (PR)
Upgrade the nodes to version: a9afe5bbd161f86281a6a8ca898905e3133f5e8a
framework_upgrade::framework-upgrade::full-framework-upgrade : committed: 1213.46 txn/s, submitted: 1216.20 txn/s, failed submission: 2.74 txn/s, expired: 2.74 txn/s, latency: 2548.96 ms, (p50: 2400 ms, p70: 2700, p90: 3600 ms, p99: 4800 ms), latency samples: 106460
framework_upgrade::framework-upgrade::full-framework-upgrade : committed: 1339.98 txn/s, submitted: 1342.63 txn/s, failed submission: 2.65 txn/s, expired: 2.65 txn/s, latency: 2245.65 ms, (p50: 2100 ms, p70: 2400, p90: 3000 ms, p99: 4800 ms), latency samples: 121180
5. check swarm health
Compatibility test for f38ec72e975a4dff4e9919e7a1d8118a75858bab ==> a9afe5bbd161f86281a6a8ca898905e3133f5e8a passed
Upgrade the remaining nodes to version: a9afe5bbd161f86281a6a8ca898905e3133f5e8a
framework_upgrade::framework-upgrade::full-framework-upgrade : committed: 1245.77 txn/s, submitted: 1249.51 txn/s, failed submission: 3.74 txn/s, expired: 3.74 txn/s, latency: 2360.90 ms, (p50: 2100 ms, p70: 2400, p90: 3300 ms, p99: 5000 ms), latency samples: 113200
Test Ok

• Grafana dashboard
• Humio Logs
• Axiom Logs
• Validator CPU Profile
• Fullnode CPU Profile
• Test runner output
• Test run is land-blocking

[github-actions]

✅ Forge suite compat success on f38ec72e975a4dff4e9919e7a1d8118a75858bab ==> a9afe5bbd161f86281a6a8ca898905e3133f5e8a

Compatibility test results for f38ec72e975a4dff4e9919e7a1d8118a75858bab ==> a9afe5bbd161f86281a6a8ca898905e3133f5e8a (PR)
1. Check liveness of validators at old version: f38ec72e975a4dff4e9919e7a1d8118a75858bab
compatibility::simple-validator-upgrade::liveness-check : committed: 14543.84 txn/s, latency: 2243.56 ms, (p50: 1900 ms, p70: 2100, p90: 2700 ms, p99: 8500 ms), latency samples: 559600
2. Upgrading first Validator to new version: a9afe5bbd161f86281a6a8ca898905e3133f5e8a
compatibility::simple-validator-upgrade::single-validator-upgrading : committed: 6070.01 txn/s, latency: 4666.37 ms, (p50: 5200 ms, p70: 5500, p90: 5700 ms, p99: 5900 ms), latency samples: 114760
compatibility::simple-validator-upgrade::single-validator-upgrade : committed: 6092.89 txn/s, latency: 5326.57 ms, (p50: 5700 ms, p70: 6000, p90: 6900 ms, p99: 7100 ms), latency samples: 207120
3. Upgrading rest of first batch to new version: a9afe5bbd161f86281a6a8ca898905e3133f5e8a
compatibility::simple-validator-upgrade::half-validator-upgrading : committed: 6217.15 txn/s, latency: 4591.84 ms, (p50: 5300 ms, p70: 5500, p90: 5600 ms, p99: 5700 ms), latency samples: 113640
compatibility::simple-validator-upgrade::half-validator-upgrade : committed: 6375.59 txn/s, latency: 5156.66 ms, (p50: 5700 ms, p70: 5700, p90: 5900 ms, p99: 6000 ms), latency samples: 216380
4. upgrading second batch to new version: a9afe5bbd161f86281a6a8ca898905e3133f5e8a
compatibility::simple-validator-upgrade::rest-validator-upgrading : committed: 8945.92 txn/s, latency: 3091.68 ms, (p50: 3200 ms, p70: 3500, p90: 4600 ms, p99: 4900 ms), latency samples: 158440
compatibility::simple-validator-upgrade::rest-validator-upgrade : committed: 8538.19 txn/s, latency: 3752.34 ms, (p50: 3400 ms, p70: 4700, p90: 5200 ms, p99: 6500 ms), latency samples: 280040
5. check swarm health
Compatibility test for f38ec72e975a4dff4e9919e7a1d8118a75858bab ==> a9afe5bbd161f86281a6a8ca898905e3133f5e8a passed
Test Ok

• Grafana dashboard
• Humio Logs
• Axiom Logs
• Validator CPU Profile
• Fullnode CPU Profile
• Test runner output
• Test run is land-blocking