afiqzudinhadi/zmk-steno-engine
1
0
Fork 0
Code Issues 3 Pull requests Projects Releases Packages Wiki Activity Actions

Fix CHD MPHF: increase bucket ratio for large partitions

Browse source 
bucket_count = entry_count/3 caused CHD construction failure on
~113K entry right partition (bucket too crowded → displacement
search exhausted 65536 tries). entry_count/2 gives more buckets,
trades ~19KB displacement overhead for reliable construction.
This commit is contained in:
afiqzudinhadi 2026-07-02 15:23:25 +08:00
parent 0908dc7c85
commit 99bf19180e
1 changed files with 3 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

6
tools/compile_mphf.py
View file

@ -170,7 +170,7 @@ def build_chd(keys_and_bytes, entry_count):
displacements[bucket] = d value displacements[bucket] = d value
slot_to_entry_idx[slot] = index into keys_and_bytes, or -1 if empty slot_to_entry_idx[slot] = index into keys_and_bytes, or -1 if empty
""" """
bucket_count = max(entry_count // 3, min(entry_count, 16)) bucket_count = max(entry_count // 2, min(entry_count, 16))
# Assign keys to buckets # Assign keys to buckets
buckets = defaultdict(list) buckets = defaultdict(list)
@ -271,7 +271,7 @@ def compile_mphf(entries, max_size=None, block_size=4096):
unique_count = len(unique_translations) unique_count = len(unique_translations)
# Estimate size # Estimate size
bucket_count = max(entry_count // 3, min(entry_count, 16)) bucket_count = max(entry_count // 2, min(entry_count, 16))
est_value_bits = max(1, math.ceil(math.log2(max(unique_count, 2)))) est_value_bits = max(1, math.ceil(math.log2(max(unique_count, 2))))
est_disp_bits = 16 # conservative est_disp_bits = 16 # conservative
est_disp_bytes = (bucket_count * est_disp_bits + 7) // 8 est_disp_bytes = (bucket_count * est_disp_bits + 7) // 8
@ -318,7 +318,7 @@ def compile_mphf(entries, max_size=None, block_size=4096):
file=sys.stderr) file=sys.stderr)
entry_count = len(keys_and_bytes) entry_count = len(keys_and_bytes)
bucket_count = max(entry_count // 3, min(entry_count, 16)) bucket_count = max(entry_count // 2, min(entry_count, 16))
print(f" Building CHD MPHF: {entry_count} entries, {bucket_count} buckets...", print(f" Building CHD MPHF: {entry_count} entries, {bucket_count} buckets...",
file=sys.stderr) file=sys.stderr)