Rework and optimize match MIR generation

[xacrimon]

left to do:

• cleanup
• code comments
• tcx arena strategy
• const eq
• handle non u8 types
• less pricky about prefixes
• also handle slice midsections and suffixes
• probably more?
• tidy
• ci
• tests
• perf improvements?
• handle type changing automatic deref/ref properly
• const eq handling for `user_ty constants in patterns
• more generic const subsequence handling involving bindings or wildcard patterns
• inspect generated mir
• phf? trie? generate an efficient decision tree to decide between patterns?
• and patterns and other things?
• subslicing/indexing bounds & check them so we don't UB
• look at earlier prs?
• reduce code size?
• handle indirect constant patterns
• inspect the mir, are we doing more work than needed? can we generate match trees in a better way? sorting?
• fix duplicated slice eq arms
• tests?

example code:

use std::{num::NonZero, ffi::CStr};
const STANDARD_BASE: &[u8] = b"org.freedesktop.DBus.Error.";
#[no_mangle]
pub(crate) fn to_errno(s: &CStr) -> Option<NonZero<i32>> {
    let (pfx, sfx) = s.to_bytes().split_at_checked(STANDARD_BASE.len())?;
    if pfx != STANDARD_BASE {
        return None;
    }
    match sfx {
        b"NoMemory" => NonZero::new(1),
        b"ServiceUnknown" | b"UnknownObject" | b"UnknownInterface" | b"UnknownMethod"
        | b"UnknownProperty" | b"NameHasNoOwner" => NonZero::new(18),
        b"AccessDenied" | b"AuthFailed" | b"InteractiveAuthorizationRequired" => {
            NonZero::new(1245)
        }
        b"LimitsExceeded" => NonZero::new(9),
        b"Timeout" | b"TimedOut" => NonZero::new(420),
        b"UnixProcessIdUnknown" => NonZero::new(69),
        b"NotSupported" => NonZero::new(228),
        b"IOError" | b"NoNetwork" | b"AddressInUse" => NonZero::new(100500),
        _ => None,
    }
}

duplicate arms example

#[no_mangle]
pub(crate) fn has_magic(s: &[u8]) -> bool {
    matches!(
        s,
        [b'm', b'a', b'g', b'i', b'c', .., b'e', b'n', b'd']
            | [b's', b't', b'a', b'r', b't', .., b'e', b'n', b'd']
    )
}

without this patch: mir and aarch64asm,opt-level=3
with this patch: mir and aarch64asm,opt-level=3

note to self: should probably be looked at by compiler-errors and scottmcm, oli?

[rustbot]

Thanks for the pull request, and welcome! The Rust team is excited to review your changes, and you should hear from @compiler-errors (or someone else) some time within the next two weeks.

Please see the contribution instructions for more information. Namely, in order to ensure the minimum review times lag, PR authors and assigned reviewers should ensure that the review label (S-waiting-on-review and S-waiting-on-author) stays updated, invoking these commands when appropriate:

• @rustbot author: the review is finished, PR author should check the comments and take action accordingly
• @rustbot review: the author is ready for a review, this PR will be queued again in the reviewer's queue

[xacrimon]

r? @ghost

[rustbot]

Failed to set assignee to ghost: invalid assignee

Note: Only org members with at least the repository "read" role, users with write permissions, or people who have commented on the PR may be assigned.

[bors]

☔ The latest upstream changes (presumably #136227) made this pull request unmergeable. Please resolve the merge conflicts.

[xacrimon]

@rustbot label +S-waiting-on-author -S-waiting-on-review

[rust-log-analyzer]

The job x86_64-gnu-tools failed! Check out the build log: (web) (plain)

Click to see the possible cause of the failure (guessed by this bot)

   Compiling rustc_codegen_llvm v0.0.0 (/checkout/compiler/rustc_codegen_llvm)
error[E0023]: this pattern has 2 fields, but the corresponding tuple variant has 1 field
   --> compiler/rustc_mir_build/src/builder/matches/match_pair.rs:170:41
    |
170 |             && let ty::ConstKind::Value(_, valtree) = const_.kind()
    |                                         ^  ^^^^^^^ expected 1 field, found 2
   ::: /checkout/compiler/rustc_type_ir/src/const_kind.rs:34:11
    |
34  |     Value(I::ValueConst),
    |           ------------- tuple variant has 1 field
    |           ------------- tuple variant has 1 field

error[E0023]: this pattern has 2 fields, but the corresponding tuple variant has 1 field
   --> compiler/rustc_mir_build/src/builder/matches/match_pair.rs:182:41
    |
182 |             && let ty::ConstKind::Value(_, valtree) = const_.kind()
    |                                         ^  ^^^^^^^ expected 1 field, found 2
   ::: /checkout/compiler/rustc_type_ir/src/const_kind.rs:34:11
    |
34  |     Value(I::ValueConst),
    |           ------------- tuple variant has 1 field
    |           ------------- tuple variant has 1 field

error[E0061]: this enum variant takes 1 argument but 2 arguments were supplied
   --> compiler/rustc_mir_build/src/builder/matches/match_pair.rs:214:44
    |
214 |         let ty_const = ty::Const::new(tcx, ty::ConstKind::Value(const_ty, valtree));
    |                                            ^^^^^^^^^^^^^^^^^^^^ --------  ------- unexpected argument #2 of type `ValTree<'tcx>`
    |                                                                 expected `Value<'_>`, found `Ty<'_>`
    |
note: tuple variant defined here
   --> /checkout/compiler/rustc_type_ir/src/const_kind.rs:34:5
   --> /checkout/compiler/rustc_type_ir/src/const_kind.rs:34:5
    |
34  |     Value(I::ValueConst),
    |     ^^^^^
help: remove the extra argument
    |
214 -         let ty_const = ty::Const::new(tcx, ty::ConstKind::Value(const_ty, valtree));
214 +         let ty_const = ty::Const::new(tcx, ty::ConstKind::Value(/* rustc_middle::ty::Value<'_> */));

Some errors have detailed explanations: E0023, E0061.
For more information about an error, try `rustc --explain E0023`.
error: could not compile `rustc_mir_build` (lib) due to 3 previous errors

[xacrimon]

Closing in favour of #136417