rt/snapraid
1
0
Fork 0
mirror of https://github.com/amadvance/snapraid.git synced 2026-10-25 18:40:19 -04:00
Code Issues Projects Releases Packages Wiki Activity Actions

Defer file hash table insertions and pre-reserve bucket capacity in state_read_content

Browse source
This commit is contained in:
Andrea Mazzoleni 2026-10-03 21:19:56 +02:00
parent 93c39d5526
commit 8589327114
1 changed files with 38 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

43
cmdline/state.c
View file

@ -2527,6 +2527,7 @@ static void state_read_content(struct snapraid_state* state, const char* path, S
ssize_t ret;
tommy_array disk_mapping;
uint32_t mapping_max;
tommy_size_t disk_file_count[RAID_DATA_MAX];
tommy_hashdyn bucket_hash;
blockmax = 0;
@ -2547,6 +2548,7 @@ static void state_read_content(struct snapraid_state* state, const char* path, S
has_info = 0;
has_invalid_parity = 0;
mapping_max = 0;
memset(disk_file_count, 0, sizeof(disk_file_count));
tommy_array_init(&disk_mapping);
tommy_hashdyn_init(&bucket_hash);
@ -2730,12 +2732,13 @@ static void state_read_content(struct snapraid_state* state, const char* path, S
*/
file = file_alloc_len(state->block_size, sub, sub_len, v_size, v_mtime_sec, v_mtime_nsec, v_inode);
/* insert the file in the file containers */
if (file->inode != INODE_INVALID)
tommy_hashdyn_insert(&disk->inodeset, &file->nodeset, file, file_inode_hash(file->inode));
tommy_hashdyn_insert(&disk->pathset, &file->pathset, file, file_path_hash_len(file->sub, sub_len));
tommy_hashdyn_insert(&disk->stampset, &file->stampset, file, file_stamp_hash(file->size, file->mtime_sec, file->mtime_nsec));
/*
* Append the file to the disk list. Hash table insertions into
* inodeset, pathset, and stampset are deferred until all records
* are parsed to avoid cache thrashing and allow pre-reserving buckets.
*/
tommy_list_insert_tail(&disk->filelist, &file->nodelist, file);
++disk_file_count[mapping];
/* read all the blocks */
v_idx = 0;
@ -3923,6 +3926,36 @@ static void state_read_content(struct snapraid_state* state, const char* path, S
}
}
/*
* Reserve hash table bucket capacity upfront and populate containers for
* all loaded files. Deferring this until after stream parsing avoids cache
* thrashing during I/O and ensures zero dynamic resizes in tommy_hashdyn.
*/
for (uint32_t i = 0; i < mapping_max; ++i) {
struct snapraid_disk* disk = tommy_array_get(&disk_mapping, i);
tommy_size_t count = disk_file_count[i];
tommy_node* node;
if (count == 0)
continue;
tommy_hashdyn_reserve(&disk->pathset, count);
tommy_hashdyn_reserve(&disk->inodeset, count);
tommy_hashdyn_reserve(&disk->stampset, count);
node = tommy_list_head(&disk->filelist);
while (node) {
struct snapraid_file* file = node->data;
if (file->inode != INODE_INVALID)
tommy_hashdyn_insert(&disk->inodeset, &file->nodeset, file, file_inode_hash(file->inode));
tommy_hashdyn_insert(&disk->pathset, &file->pathset, file, file_path_hash(file->sub));
tommy_hashdyn_insert(&disk->stampset, &file->stampset, file, file_stamp_hash(file->size, file->mtime_sec, file->mtime_nsec));
node = node->next;
}
}
tommy_array_done(&disk_mapping);
if (serror(f)) {