criterion: initial benchmarks

[hinto-janai]

What

Adds initial Criterion benchmarks for:

• cuprate-helper
• cuprate-database
• cuprate-cryptonight
• cuprate-rpc-types
• cuprate-blockchain

The location of these is benches/criterion/$CRATE e.g. benches/criterion/cuprate-helper.

This PR combines #322 #324 #327 #328 #332 to prevent merge conflicts.

In-memory disk ops

cuprate-{database,blockchain} are setup to use tempfile::tempdir() which uses std::env::temp_dir() which should be in-memory (/tmp) on most Linuxes, not sure about Windows/macOS. This should remove disk-related variables and provide more accurate function runtime data.

When broader benches/benchmarks are created, those could use disk to provide more realistic real-world data.

cuprate-helper

cargo bench -p cuprate-criterion-helper

benchmark	average time per call
integer	1.2848 ns
unsigned	1.2876 ns
combine_low_high_bits_to_u128	846.72 ps
split_u128_into_low_high_bits	647.49 ps
timelock_to_u64	1.6886 ns
u64_to_timelock	1.7356 ns
cmp_float	1.0841 ns
cmp_float_nan	659.60 ps
get_mid	3.9615 ns
median	55.030 ns
tx_fee	6.9573 ns

cuprate-database

cargo bench -p cuprate-criterion-database

• redb values are only shown for >3% diffs
• code formatting is faster

benchmark	heed	redb	diff
ro_get	24.193 ns	21.809 ns	-9.98%
ro_len	4.2254 ns
ro_first	29.797 ns
ro_last	31.675 ns	33.186 ns	+4.39%
ro_is_empty	7.6418 ns
ro_contains	22.448 ns	21.696 ns	-3.96%
rw_get	23.204 ns	22.107 ns	-4.63%
rw_len	4.6742 ns	4.4666 ns	-4.63%
rw_first	31.337 ns
rw_last	33.221 ns	34.402 ns	+3.90%
rw_is_empty	7.7049 ns
rw_contains	22.874 ns	22.017 ns	-3.72%
get_range	1.6075 µs
iter	1.9882 µs	1.9254 µs	-3.54%
keys	766.35 ns
values	1.5660 µs
put	3.2995 µs	3.5970 µs	+8.76%
delete	422.90 ns
pop_first	313.42 ns	286.33 ns	-8.70%
pop_last	329.65 ns	305.60 ns	-7.11%
take	4.8301 µs
env_inner	3.7474 ns
tx_ro	77.387 ns	65.183 ns	-15.73%
tx_rw	33.163 ns	35.133 ns	+6.00%
open_db_ro	18.273 ns
open_db_rw	17.722 ns
create_db	18.222 ns
resize	15.658 µs
current_map_size	4.2498 ns
disk_size_bytes	2.2181 µs
pre_rct_output_id_as_bytes	649.28 ps
pre_rct_output_id_from_bytes	782.82 ps
output_as_bytes	650.27 ps
output_from_bytes	5.8138 ns

cuprate-cryptonight

Benchmark naming scheme: format!("{hash_fn_name}_{}", input.len())

E.g. cuprate_cryptonight::cryptonight_hash_v0(&[0; 8]) is v0_8.

benchmark	average time per call
r_8	57.091 ms
r_64	58.804 ms
r_512	58.892 ms
r_4096	58.645 ms
r_65536	58.925 ms
v0_8	25.941 ms
v0_64	26.790 ms
v0_512	26.750 ms
v0_4096	26.300 ms
v0_65536	26.787 ms
v1_8	3.4884 ns
v1_64	26.920 ms
v1_512	26.406 ms
v1_4096	26.896 ms
v1_65536	27.092 ms
v2_64	28.900 ms
v2_512	29.331 ms
v2_4096	28.683 ms
v2_65536	29.345 ms

cuprate-rpc-types

benchmark	average time per call
epee_from_bytes_get_blocks_request	86.675 ns
epee_to_bytes_get_blocks_request	108.94 ns
serde_from_str_tx_entry	928.48 ns
serde_to_string_tx_entry	190.51 ns
serde_from_str_get_last_block_header_response	651.84 ns
serde_to_string_get_last_block_header_response	507.25 ns
serde_from_str_calc_pow_request	183.96 ns
serde_to_string_calc_pow_request	60.611 ns
serde_from_str_sync_info_response	1.5147 µs
serde_to_string_sync_info_response	180.23 ns
serde_from_str_get_info_response	997.70 ns
serde_to_string_get_info_response	843.08 ns
serde_from_str_get_block_response	1.7308 µs
serde_to_string_get_block_response	576.69 ns
serde_from_str_get_connections_response	1.0853 µs
serde_to_string_get_connections_response	46.413 ns
serde_from_str_get_block_template_response	469.21 ns
serde_to_string_get_block_template_response	267.68 ns
serde_from_str_get_block_headers_range_response	1.1637 µs
serde_to_string_get_block_headers_range_response	74.125 ns

cuprate-blockchain

benchmark	average time per call
add_block_v1_tx2	244.05 µs
add_block_v9_tx3	54.324 µs
add_block_v16_tx0	5.4705 µs
add_alt_block_v1_tx2	1.5027 µs
add_alt_block_v9_tx3	1.7813 µs
add_alt_block_v16_tx0	1.4159 µs

[hinto-janai]

See #354.