Automatic toolchain upgrade to nightly-2023-12-21

[github-actions]

Update Rust toolchain from nightly-2023-12-20 to nightly-2023-12-21 without any other source changes.
This is an automatically generated pull request. If any of the CI checks fail, manual intervention is required. In such a case, review the changes at https://github.com/rust-lang/rust from rust-lang/rust@f704f3b up to rust-lang/rust@5ac4c8a. The log for this commit range is:
rust-lang/rust@5ac4c8a63e Auto merge of #119037 - RalfJung:repr-c-abi-mismatch, r=scottmcm
rust-lang/rust@92ad4b433a Auto merge of #119166 - GuillaumeGomez:rollup-qfgj76w, r=GuillaumeGomez
rust-lang/rust@c76841928d Rollup merge of #119159 - petrochenkov:llvmup, r=nikic
rust-lang/rust@5906d8f0cf Rollup merge of #119155 - Zalathar:async-fn, r=compiler-errors
rust-lang/rust@0bbcdb2aa3 Rollup merge of #119115 - GuillaumeGomez:env-docs, r=Nilstrieb
rust-lang/rust@2679bca97a Add link to explanations about dep-info files
rust-lang/rust@8fca8295cb Auto merge of #119136 - petrochenkov:feedvis3, r=WaffleLapkin
rust-lang/rust@8468840a56 Update LLVM submodule
rust-lang/rust@f9d52dc594 Auto merge of #119134 - petrochenkov:feedvis2, r=compiler-errors
rust-lang/rust@5e5d82e803 resolve: Eagerly feed closure visibilities
rust-lang/rust@3e4a15ea06 Auto merge of #119156 - matthiaskrgr:rollup-482ow65, r=matthiaskrgr
rust-lang/rust@d6656af16a Rollup merge of #119107 - lcnr:uwuwu, r=compiler-errors
rust-lang/rust@9058791813 Rollup merge of #119102 - chrisnc:arm-none-eabi-docs, r=wesleywiser
rust-lang/rust@8b3914ce48 Rollup merge of #119094 - celinval:smir-layout, r=compiler-errors
rust-lang/rust@f9c752d931 Rollup merge of #119089 - fmease:dont-ice-on-tilde-const-non-const-trait, r=fee1-dead
rust-lang/rust@16a231d8eb Rollup merge of #119071 - lcnr:overflowo, r=compiler-errors
rust-lang/rust@df4d56307b Rollup merge of #118973 - Enselic:fix-IncorrectCguReuseType, r=michaelwoerister
rust-lang/rust@f3f9b3043e Rollup merge of #118691 - chfogelman:improve-cstr-error, r=fmease
rust-lang/rust@cf6dc7adb3 coverage: Check for async fn explicitly, without needing a heuristic
rust-lang/rust@2a0290a802 coverage: Pass around &ExtractedHirInfo instead of individual fields
rust-lang/rust@4ae792036e coverage: Add a test for async blocks
rust-lang/rust@3095d31a75 Auto merge of #119118 - arttet:feature/fix-arm64e-targets, r=petrochenkov
rust-lang/rust@5810deef69 Auto merge of #119096 - compiler-errors:yeet-unnecessary-param-envs, r=lcnr
rust-lang/rust@51c0db6a91 Auto merge of #106790 - the8472:rawvec-niche, r=scottmcm
rust-lang/rust@2c96025874 Improve compiler error for c-strings in pre-2021
rust-lang/rust@f6150db78f update natvis to match changed RawVec structure
rust-lang/rust@7571f6f685 resolve: Feed visibilities for unresolved trait impl items
rust-lang/rust@d46df80c73 rustc_codegen_ssa: Don't let IncorrectCguReuseType errors get lost
rust-lang/rust@76b3e6de55 Fix c_variadic flag and add opaque info to PassMode
rust-lang/rust@a75d0021ed Remove param env from relation altogether
rust-lang/rust@17705ea0a8 add comment
rust-lang/rust@009cd9c815 Remove unnecessary param-env from lexical region resolution and fully structural relations
rust-lang/rust@e6d6b1d92c Do not evaluate in structurally_relate_tys
rust-lang/rust@ffdc6c7b65 effects: fix comment
rust-lang/rust@fd0033c777 Use LLVM features for arm64e_apple_ios target
rust-lang/rust@3f8704355b Remove legacy bitcode defaults
rust-lang/rust@bd984ddcc1 Update documentation for --env compilation flag
rust-lang/rust@90893e4850 Add arm-none-eabi and armv7r-none-eabi platform-support documentation.
rust-lang/rust@df5b604945 subtype_predicate: remove unnecessary probe
rust-lang/rust@1a83c5b55b Add function ABI and type layout to StableMIR
rust-lang/rust@4a38442c90 dont discard overflow from normalizes-to goals
rust-lang/rust@ca718ffd2d track the source of nested goals
rust-lang/rust@c7e3b3f84c do not allow ABI mismatches inside repr(C) types
rust-lang/rust@81997094e6 update debug providers to match new RawVec capacity field
rust-lang/rust@6a2f44e9d8 add comment to RawVec::cap field
rust-lang/rust@502df1b7d4 add more niches to rawvec

[zhassan-aws]

I ran the harness that is failing in the performance regressions (recovery::rtt_estimator::test::weighted_average_test) several times locally with the current toolchain (2023-12-20), and the next one (2023-12-21), and confirmed that the slowdown is consistent. So it seems that some change in the toolchain is resulting in worse performance.

[adpaco-aws]

Thanks for letting us know, @zhassan-aws . I'm going to compare the code that's being generated for this test case to see where the differences are. I haven't been able to execute the test case in isolation yet (i.e., without benchcomp), but I'm hoping to eventually get there so I can make the comparison.

[zhassan-aws]

@feliperodri you can cd into tests/perf/s2n-quic/quic/s2n-quic-core and run:

cargo kani --tests --harness weighted_average_test

to run this specific harness.

[adpaco-aws]

Thanks, @zhassan-aws ! I haven't looked into the specific code differences but I could notice a significant degradation just by looking at the logs.

Current toolchain nightly-2023-12-20:

size of program expression: 14586 steps
slicing removed 9093 assignments
Generated 423 VCC(s), 160 remaining after simplification
[...]
Verification Time: 123.503876s

Next toolchain nightly-2023-12-21:

size of program expression: 17012 steps
slicing removed 10408 assignments
Generated 611 VCC(s), 230 remaining after simplification
[...]
Verification Time: 243.44136s
[...]

In my opinion, this is a legit regression and we shouldn't merge the update until we know more about the changes.

[zhassan-aws]

Subsumed by #2969.