@matheusbsbs / matheus-bagliano / commits / 950a234cbd

string-list: use ALLOC_GROW macro when reallocing string_list

Use ALLOC_GROW() macro when reallocing a string_list array rather than simply increasing it by 32. This is a performance optimization. During status on a very large repo and there are many changes, a significant percentage of the total run time is spent reallocing the wt_status.changes array. This change decreases the time in wt_status_collect_changes_worktree() from 125 seconds to 45 seconds on my very large repository. This produced a modest gain on my 1M file artificial repo, but broke even on linux.git. Test HEAD^^ HEAD --------------------------------------------------------------------------------------- 0005.2: read-tree status br_ballast (1000001) 8.29(5.62+2.62) 8.22(5.57+2.63) -0.8% Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>

Jeff Hostetler committed Apr 14, 2017 at 19:51 UTC 950a234cbd781021d69fcfaa40ab6fd258b1d917
2 files changed +50 -4
string-list.c
+1 -4
@@ -41,10 +41,7 @@ static int add_entry(int insert_at, struct string_list *list, const char *string
41 if (exact_match)
42 return -1 - index;
43
44 - if (list->nr + 1 >= list->alloc) {
45 - list->alloc += 32;
46 - REALLOC_ARRAY(list->items, list->alloc);
47 - }
44 + ALLOC_GROW(list->items, list->nr+1, list->alloc);
45 if (index < list->nr)
46 memmove(list->items + index + 1, list->items + index,
47 (list->nr - index)
t/perf/p0005-status.sh new
+49
@@ -0,0 +1,49 @@
1 +#!/bin/sh
2 +#
3 +# This test measures the performance of various read-tree
4 +# and status operations. It is primarily interested in
5 +# the algorithmic costs of index operations and recursive
6 +# tree traversal -- and NOT disk I/O on thousands of files.
7 +
8 +test_description="Tests performance of read-tree"
9 +
10 +. ./perf-lib.sh
11 +
12 +test_perf_default_repo
13 +
14 +# If the test repo was generated by ./repos/many-files.sh
15 +# then we know something about the data shape and branches,
16 +# so we can isolate testing to the ballast-related commits
17 +# and setup sparse-checkout so we don't have to populate
18 +# the ballast files and directories.
19 +#
20 +# Otherwise, we make some general assumptions about the
21 +# repo and consider the entire history of the current
22 +# branch to be the ballast.
23 +
24 +test_expect_success "setup repo" '
25 + if git rev-parse --verify refs/heads/p0006-ballast^{commit}
26 + then
27 + echo Assuming synthetic repo from many-files.sh
28 + git branch br_base master
29 + git branch br_ballast p0006-ballast
30 + git config --local core.sparsecheckout 1
31 + cat >.git/info/sparse-checkout <<-EOF
32 + /*
33 + !ballast/*
34 + EOF
35 + else
36 + echo Assuming non-synthetic repo...
37 + git branch br_base $(git rev-list HEAD | tail -n 1)
38 + git branch br_ballast HEAD
39 + fi &&
40 + git checkout -q br_ballast &&
41 + nr_files=$(git ls-files | wc -l)
42 +'
43 +
44 +test_perf "read-tree status br_ballast ($nr_files)" '
45 + git read-tree HEAD &&
46 + git status
47 +'
48 +
49 +test_done