From: Derrick Stolee <dstolee@xxxxxxxxxxxxx>
The changed-path Bloom filters record an entry in the filter for
every path that was changed. This includes every add and delete,
regardless of whther a rename was detected. Detecting renames
causes significant performance issues, but also will trigger
downloading missing blobs in partial clone.
The simple fix is to disable rename detection when computing a
changed-path Bloom filter.
Signed-off-by: Derrick Stolee <dstolee@xxxxxxxxxxxxx>
---
bloom: ignore renames when computing changed paths
I promised [1] I would adapt the commit that was dropped from
gs/commit-graph-path-filter [2] on top of gs/commit-graph-path-filter
and jt/avoid-prefetch-when-able-in-diff. However, I noticed that the
change was extremely simple and has value without basing it on
jt/avoid-prefetch-when-able-in-diff.
This change applied to gs/commit-graph-path-filter has obvious CPU time
improvements for computing changed-path Bloom filters (that I did not
measure). The partial clone improvements require
jt/avoid-prefetch-when-able-in-diff to be included, too, but the code
does not depend on it at compile time.
Thanks, -Stolee
[1]
https://lore.kernel.org/git/7de2f54b-8704-a0e1-12aa-0ca9d3d70f6f@xxxxxxxxx/
[2]
https://lore.kernel.org/git/55824cda89c1dca7756c8c2d831d6e115f4a9ddb.1585528298.git.gitgitgadget@xxxxxxxxx/
Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-601%2Fderrickstolee%2Fdiff-and-bloom-filters-v1
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-601/derrickstolee/diff-and-bloom-filters-v1
Pull-Request: https://github.com/gitgitgadget/git/pull/601
bloom.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/bloom.c b/bloom.c
index c5b461d1cfe..dd9bab9bbd6 100644
--- a/bloom.c
+++ b/bloom.c
@@ -189,6 +189,7 @@ struct bloom_filter *get_bloom_filter(struct repository *r,
repo_diff_setup(r, &diffopt);
diffopt.flags.recursive = 1;
+ diffopt.detect_rename = 0;
diffopt.max_changes = max_changes;
diff_setup_done(&diffopt);
base-commit: d5b873c832d832e44523d1d2a9d29afe2b84c84f
--
gitgitgadget
