internal: add integrated completion benchmark

[matklad]

bors r+
🤖

[matklad]

$ cargo test --release --package rust-analyzer --lib -- integrated_benchmarks::integrated_completion_benchmark --exact --nocapture
running 1 test
workspace loading: 329.92ms
initial: 6.87s
change: 33.93µs
cpu profiling is disabled, uncomment `default = [ "cpu_profiler" ]` in Cargo.toml to enable.
   23ms - SourceBinder::to_module_def
       23ms - crate_def_map:wait
           16ms - item_tree_query (197 calls)
            0ms - parse_macro_expansion (1 calls)
            7ms - ???
   20ms - descend_into_macros
       20ms - Semantics::analyze_impl
           15ms - infer:wait @ name
               15ms - infer_query
                   15ms - deref
                       15ms - deref_by_trait
                           15ms - trait_solve::wait
                                5ms - impl_data_query (272 calls)
                                9ms - ???
                    0ms - crate_def_map:wait (36 calls)
                    0ms - deref (5 calls)
                    0ms - resolve_obligations_as_possible (3 calls)
                    0ms - trait_solve::wait (5 calls)
            0ms - SourceBinder::to_module_def (1 calls)
            0ms - body_with_source_map_query (1 calls)
            0ms - crate_def_map:wait (2 calls)
            0ms - impl_data_query (1 calls)
            1ms - macro_expand (108 calls)
            1ms - parse_macro_expansion (152 calls)
   47ms - import_on_the_fly @ sel
       28ms - import_assets::search_for_imports
           28ms - import_assets::search_for
               28ms - import_assets::path_applicable_imports
                   23ms - items_with_name @ Name: sel, crate: Exclude, assoc items: Some("hir"), limit: Some(40)
                       23ms - find_items
                           20ms - search_dependencies @ Query { query: "sel", lowercased: "sel", name_only: true, assoc_items_only: false, search_mode: Fuzzy, case_sensitive: false, limit: 40, exclude_import_kinds: {AssociatedItem} }
                               19ms - import_map_query (17 calls)
                            0ms - crate_def_map:wait (1 calls)
                    2ms - find_path_prefixed (43 calls)
                    1ms - get_name_definition (40 calls)
                0ms - import_assets::scope_definitions (1 calls)
        9ms - render_resolution
            9ms - render_fn
                9ms - parse_query @ FileId(2993)
                0ms - crate_def_map:wait (6 calls)
                0ms - generic_params_query (1 calls)
                0ms - item::Builder::build (1 calls)
        0ms - crate_def_map:wait (1 calls)
        9ms - render_resolution (38 calls)
unqualified path completion: 137.42ms
change: 41.78µs
cpu profiling is disabled, uncomment `default = [ "cpu_profiler" ]` in Cargo.toml to enable.
   21ms - SourceBinder::to_module_def
       21ms - crate_def_map:wait
           13ms - item_tree_query (197 calls)
            0ms - macro_expand (34 calls)
            0ms - parse_macro_expansion (43 calls)
            7ms - ???
   19ms - descend_into_macros
       19ms - Semantics::analyze_impl
           15ms - infer:wait @ name
               15ms - infer_query
                   14ms - deref
                       14ms - deref_by_trait
                           14ms - trait_solve::wait
                                5ms - impl_data_query (272 calls)
                                9ms - ???
                    0ms - crate_def_map:wait (35 calls)
                    0ms - deref (6 calls)
                    0ms - resolve_obligations_as_possible (3 calls)
                    0ms - trait_solve::wait (5 calls)
            0ms - SourceBinder::to_module_def (1 calls)
            0ms - body_with_source_map_query (1 calls)
            0ms - crate_def_map:wait (2 calls)
            0ms - impl_data_query (1 calls)
            1ms - macro_expand (99 calls)
            1ms - parse_macro_expansion (148 calls)
    8ms - import_on_the_fly @ 
        8ms - import_assets::search_for_imports
            8ms - import_assets::search_for
                7ms - import_assets::trait_applicable_items
                    0ms - crate_def_map:wait (155 calls)
                    0ms - deref (2 calls)
                    0ms - generic_params_query (72 calls)
                    1ms - get_name_definition (40 calls)
                    4ms - items_with_name (1 calls)
                    0ms - trait_solve::wait (44 calls)
                0ms - import_assets::scope_definitions (1 calls)
        0ms - Semantics::analyze_impl (1 calls)
dot completion: 71.34ms
test integrated_benchmarks::integrated_completion_benchmark ... ok

test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 40 filtered out; finished in 7.68s

[matklad]

Things that jump at me -- there should be no item_tree or impl_data queries in there.

[lnicola]

Wait, how does this work?

[matklad]

I uncomment this manually when I want to capture profile :-)

[lnicola]

Might be worth checking for an env variable instead.

[matklad]

bors r+

[bors]

Canceled.

[bors]

Build succeeded:

• Rust (macos-latest)
• Rust (ubuntu-latest)
• Rust (windows-latest)
• TypeScript (ubuntu-latest)
• TypeScript (windows-latest)

[SomeoneToIgnore]

47ms - import_on_the_fly @ sel
8ms - import_on_the_fly @

I've expected complete opposite results, IMO we should exaggerate the latter by calling it on some struct with multiple applicable trait impls.

[matklad]

Filed #8721. I am exctremely satisfied that it took like 15 minutes to find the first WTF behavior after having the benchmark :3

@SomeoneToIgnore this type (hir::Struct) has a fair number of impls: HasSource, HasAttrs, HirDisplay, a bunch of derived ones. I do expect "plain" lookups to be costlier, as they look at all the symbols, without any kind of type filtering at all.