-
Notifications
You must be signed in to change notification settings - Fork 3
Performance of LLVM 15 vs LLVM 14
Laurynas Biveinis edited this page Sep 23, 2022
·
5 revisions
Filtered for unodb::db
:
-
micro_benchmark_key_prefix
: 1% speedup (unpredictable_get_shared_length
) to 4% speedup (unpredictable_prepend_key_prefix
) -
micro_benchmark_n4
: 4% slowdown (full_n4_sequential_delete/512
) to 7% speedup (full_n4_sequential_insert/4096
) -
micro_benchmark_n16
: 0% slowdown (n16_sequential_add/512
) to 7% speedup (n16_random_add/10
) -
micro_benchmark_n48
: 7% slowdown (full_n48_tree_full_scan/128
) to 13% speedup (n48_sequential_add/2
) -
micro_benchmark_n256
: 1% slowdown (minimal_n256_tree_full_scan/4
) to 9% speedup (grow_n48_to_n256_sequentially/64
)
Filtered for unodb::olc_db
:
-
micro_benchmark_key_prefix
: 0% speedup (unpredictable_get_shared_length
) to 2% speedup (unpredictable_leaf_key_prefix_split
) -
micro_benchmark_n4
: 0% speedup (n4_full_scan/100
) to 5% speedup (shrink_node16_to_n4_sequentially/4096
) -
micro_benchmark_n16
: 1% slowdown (n16_sequential_add/64
) to 11% speedup (full_n16_tree_random_delete/32768
) -
micro_benchmark_n48
: 1% slowdown (minimal_n48_tree_full_scan/6144
) to 7% speedup (full_n48_tree_random_delete/196608
) -
micro_benchmark_n256
: 2% slowdown (full_n256_tree_random_delete/192
) to 7% speedup (grow_n48_to_n256_sequentially/8
)