211-refactor-logical-planner - Branch - jzombie/rust-llkv

Refactor planning and execution

#223

Comparing

211-refactor-logical-planner

(

3b66a0a

) with

main

(

8977772

)

-54%

Improvements: 9

Regressions: 9

Untouched: 57

Skipped: 2

Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data. For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

Benchmarks

Skipped (2)

Failed

hash_join_anti_join[1000]

criterion::hash_join_anti_join::1000

Regression

-54%

35 µs75.2 µs

hash_join_left_join[10000]

criterion::hash_join_left_join::10000

Regression

-18%

508.8 µs617.4 µs

hash_join_semi_join[1000]

criterion::hash_join_semi_join::1000

Regression

-25%

30.4 µs40.6 µs

hash_join_inner_join[10000]

criterion::hash_join_inner_join::10000

Regression

-24%

605.3 µs796.5 µs

hash_join_anti_join[10000]

criterion::hash_join_anti_join::10000

Regression

-39%

271.2 µs445.5 µs

hash_join_semi_join[10000]

criterion::hash_join_semi_join::10000

Regression

-29%

219.4 µs310.2 µs

hash_join_semi_join[100000]

criterion::hash_join_semi_join::100000

Regression

-11%

3.1 ms3.5 ms

hash_join_semi_join[1000000]

criterion::hash_join_semi_join::1000000

Regression

-15%

36.7 ms43.2 ms

ingest_mixed_1M[batches=64]

criterion::ingest_mixed_1M::batches=64

Regression

-11%

340.9 ms381.6 ms

Improved

hash_join_many_to_many[1000]

criterion::hash_join_many_to_many::1000

+68%

129.9 µs77.5 µs

hash_join_no_matches[1000]

criterion::hash_join_no_matches::1000

+60%

114.1 µs71.3 µs

hash_join_left_join[1000]

criterion::hash_join_left_join::1000

+47%

149.9 µs101.9 µs

select4_full

criterion::slturl_cases

+35%

7.6 s5.6 s

storage_fused_both_case_sensitive

criterion::string_substring_storage_only

+17%

7.5 ms6.4 ms

scan_stream_expr_add_one

criterion::llkv_table_math_1M

+16%

3.5 ms3 ms

storage_fused_starts_case_insensitive

criterion::string_substring_storage_only

+13%

8.4 ms7.5 ms

table_layer_sum

criterion::table_layer_overhead_analysis

+13%

2.4 ms2.1 ms

storage_sequential_both_case_sensitive

criterion::string_substring_storage_only

+11%

54.9 ms49.4 ms

Passed

storage_sequential_starts_case_insensitive

criterion::string_substring_storage_only

+10%

94.6 ms86.1 ms

hash_join_left_join[1000000]

criterion::hash_join_left_join::1000000

+9%

142.6 ms130.7 ms

hash_join_many_to_many[100000]

criterion::hash_join_many_to_many::100000

+9%

12.7 ms11.7 ms

hash_join_inner_join[1000]

criterion::hash_join_inner_join::1000

+8%

75.5 µs69.6 µs

hash_join_many_to_many[100]

criterion::hash_join_many_to_many::100

+8%

26.4 µs24.4 µs

storage_sequential_contains_case_insensitive

criterion::string_substring_storage_only

+8%

116.4 ms107.6 ms

delete1000_full

criterion::slturl_cases

+7%

375.6 ms351 ms

hash_join_no_matches[10000]

criterion::hash_join_no_matches::10000

+7%

632.8 µs594.1 µs

direct_columnstore_sum

criterion::table_layer_overhead_analysis

+5%

102.2 µs97.5 µs

store_sequential_contains_case_insensitive

criterion::string_substring_fused_vs_sequential

+4%

110.4 ms105.8 ms

scan_stream_sum_u64

criterion::llkv_table_math_1M

+4%

2.2 ms2.1 ms

scan_stream_expr_subtract_one

criterion::llkv_table_math_1M

+4%

3.1 ms3 ms

storage_fused_contains_case_insensitive

criterion::string_substring_storage_only

+3%

88.6 ms85.6 ms

hash_join_inner_join[1000000]

criterion::hash_join_inner_join::1000000

+3%

276.3 ms267.1 ms

hash_join_left_join[100000]

criterion::hash_join_left_join::100000

+3%

7.6 ms7.4 ms

scan_stream_expr_multiply_two

criterion::llkv_table_math_1M

+3%

3.1 ms3 ms

storage_fused_both_case_insensitive

criterion::string_substring_storage_only

+2%

86.4 ms84.5 ms

store_sequential_starts_case_insensitive

criterion::string_substring_fused_vs_sequential

+2%

90.4 ms88.5 ms

lowfrag/range/without_index

criterion::index_matrix_1M::lowfrag_range_without_index

+2%

3.9 ms3.8 ms

sum_u64_scan_only

criterion::column_store_sum_1M

+2%

103.1 µs101 µs

planner_fused_both_case_insensitive

criterion::string_substring_fused_vs_sequential

+2%

167.9 ms164.7 ms

planner_fused_starts_case_insensitive

criterion::string_substring_fused_vs_sequential

+1%

111.5 ms109.9 ms

planner_fused_contains_case_insensitive

criterion::string_substring_fused_vs_sequential

+1%

130 ms128.7 ms

store_sequential_both_case_insensitive

criterion::string_substring_fused_vs_sequential

+1%

146 ms145.1 ms

storage_sequential_both_case_insensitive

criterion::string_substring_storage_only

146 ms145.4 ms

multi_column_batched

criterion::gather_rows

2 ms2 ms

multi_column_sequential

criterion::gather_rows

2 ms2 ms

planner_fused_both_case_sensitive

criterion::string_substring_fused_vs_sequential

73.7 ms73.6 ms

store_sequential_both_case_sensitive

criterion::string_substring_fused_vs_sequential

49.9 ms49.9 ms

sum_i32_scan_only

criterion::column_store_sum_1M

49.1 µs49.1 µs

single_column

criterion::gather_rows

667.6 µs667.9 µs

scan_sorted_stream_join

criterion::random_probe_u64

20.5 ms20.5 ms

scan_stream_expr_divide_two

criterion::llkv_table_math_1M

4.3 ms4.3 ms

multi_column_batched_with_nulls

criterion::gather_rows

770.7 µs771.3 µs

highfrag/full/without_index

criterion::index_matrix_1M::highfrag_full_without_index

14 ms14 ms

sorted_range_u64

criterion::scan_builder_1M

22.7 ms22.8 ms

ingest_mixed_1M[batches=16]

criterion::ingest_mixed_1M::batches=16

346.3 ms347.6 ms

ingest_mixed_1M[batches=4]

criterion::ingest_mixed_1M::batches=4

316.5 ms317.9 ms

highfrag/full/with_index

criterion::index_matrix_1M::highfrag_full_with_index

41.1 ms41.3 ms

hash_join_no_matches[100000]

criterion::hash_join_no_matches::100000

-1%

10.4 ms10.5 ms

lowfrag/range/with_index

criterion::index_matrix_1M::lowfrag_range_with_index

-1%

1.9 ms1.9 ms

scan_contains_needle_case_insensitive

criterion::string_substring_search

-1%

102.9 ms103.6 ms

lowfrag/full/without_index

criterion::index_matrix_1M::lowfrag_full_without_index

-1%

14.5 ms14.6 ms

highfrag/range/with_index

criterion::index_matrix_1M::highfrag_range_with_index

-1%

2.8 ms2.8 ms

sorted_with_row_ids_u64

criterion::scan_builder_1M

-1%

19.7 ms19.9 ms

lowfrag/full/with_index

criterion::index_matrix_1M::lowfrag_full_with_index

-1%

20 ms20.3 ms

highfrag/range/without_index

criterion::index_matrix_1M::highfrag_range_without_index

-2%

5.2 ms5.3 ms

ingest_mixed_1M[batches=1]

criterion::ingest_mixed_1M::batches=1

-2%

241 ms245.5 ms

scan_contains_needle_case_sensitive

criterion::string_substring_search

-2%

46.4 ms47.4 ms

unsorted_sum_u64

criterion::scan_builder_1M

-3%

961.6 µs987.1 µs

sum_u64_fragmented_scan_only

criterion::column_store_fragmented_1M

-7%

88 µs94.2 µs

hash_join_inner_join[100000]

criterion::hash_join_inner_join::100000

-7%

11.2 ms12.1 ms

hash_join_no_matches[1000000]

criterion::hash_join_no_matches::1000000

-9%

246.6 ms270.4 ms

hash_join_anti_join[1000000]

criterion::hash_join_anti_join::1000000

-9%

42.5 ms46.7 ms

hash_join_many_to_many[10000]

criterion::hash_join_many_to_many::10000

-9%

752.4 µs828.5 µs

scan_unsorted_multiset

criterion::random_probe_u64

-10%

1.4 ms1.6 ms

hash_join_anti_join[100000]

criterion::hash_join_anti_join::100000

-10%

3.7 ms4.1 ms

Commits

Click on a commit to change the comparison range

Base

main

8977772

-53.53%

211-refactor-logical-planner -- perf monitor (#224) * Prototype granular perf calcs * Add TODOs * Prototype batched row counts * Unify total row count handling * Prototype perf monitoring * Apply cargo formatting * Prototype query-based perf monitoring * Add hit count * Add TODOs for multi-table physical planning * Simplify query context * Update comment * Verify zero-cost of timing code * Verify assembly is different if perf enabled * Add comment * Remove erroneous side effect * Manually feature-gate macros Also add `measure_with_duration` which always times the internal code * Remove assertion which conflicts with SLT tests * Add TODOs * Update SQL tests to match current SLT expectations * Set `LLKV_PERF_SLOW_MS` to 1 second in CI * Rename macros for clarity * Fix bench compilation * Temporarily bypass failing test * Reimplement support for correlated subqueries * Add TODO * Initial conversion to Arc-based plan vectors * Add TODOs * Remove older, duplicated physical planner structs * Transition many Vec to Arc-wrapped slices * Fix failing pager tests This is unrelated to the previous commit's Arc update, and likely a slight regression after reimplementing support for correlated subqueries (c70c1), though I haven't confirmed. * Resolve Clippy warnings * Debug Windows failure with TOML-safe path handling * Rename tests for consistency * Add TODO * Prototype hash perf optimizations * Replace std HashMap with FxHashMap * Lint deps * Add'l dep lint * `cargo fmt --all` * Comment-out log spam * Debug failing lint * Fix incorrect empty projection assumption * Resolve lint warnings * Resolve Clippy warnings * Ensure `run_script_at_path` respects `SLT_HARNESS_STACK_SIZE` * `cargo fmt --all`

3b66a0a

28 days ago

by jzombie

Home Terms Privacy Docs