Performance of LLVM 17 vs LLVM 16

commit

Filtered for unodb::db:

• micro_benchmark_key_prefix: 2% slowdown (unpredictable_cut_key_prefix) to 0% slowdown (unpredictable_leaf_key_prefix_split)
• micro_benchmark_n4: 5% slowdown (full_n4_random_deletes/512) to 1% speedup (minimal_n4_sequential_insert/255)
• micro_benchmark_n16: 3% slowdown (full_n16_tree_sequential_delete/32768) to 9% speedup (full_n16_tree_sequential_delete/64)
• micro_benchmark_n48: 4% slowdown (full_n48_tree_random_delete/196608) to 5% speedup (shrink_n256_to_n48_sequentially/8)
• micro_benchmark_n256: 6% slowdown (full_n256_tree_sequential_delete/512) to 10% speedup (grow_n48_to_n256_randomly/8)

Filtered for unodb::olc_db:

• micro_benchmark_key_prefix: 2% speedup (unpredictable_get_shared_length)
• micro_benchmark_n4: 1% speedup (full_n4_sequential_insert/65535) to 4% speedup (n4_full_scan/4096)
• micro_benchmark_n16: 1% slowdown (n16_sequential_add/10) to 5% speedup (grow_n4_to_n16_sequentially/20)
• micro_benchmark_n48: 6% slowdown (shrink_n256_to_n48_randomly/64) to 4% speedup (full_n48_tree_sequential_delete/192)
• micro_benchmark_n256: 10% slowdown (grow_n48_to_n256_randomly/8) to 6% speedup (full_n256_tree_random_delete/196608)