string-list: use ALLOC_GROW macro when reallocing string_list

jeffhostetler · gitster · commit 950a234cbd78 · 2017-04-15T02:04:41.000-07:00
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 &lt;jeffhost@microsoft.com&gt;
Signed-off-by: Junio C Hamano &lt;gitster@pobox.com&gt;
diff --git a/string-list.c b/string-list.c
@@ -41,10 +41,7 @@ static int add_entry(int insert_at, struct string_list *list, const char *string
 	if (exact_match)
 		return -1 - index;
 
-	if (list->nr + 1 >= list->alloc) {
-		list->alloc += 32;
-		REALLOC_ARRAY(list->items, list->alloc);
-	}
+	ALLOC_GROW(list->items, list->nr+1, list->alloc);
 	if (index < list->nr)
 		memmove(list->items + index + 1, list->items + index,
 				(list->nr - index)
diff --git a/t/perf/p0005-status.sh b/t/perf/p0005-status.sh
@@ -0,0 +1,49 @@
+#!/bin/sh
+#
+# This test measures the performance of various read-tree
+# and status operations.  It is primarily interested in
+# the algorithmic costs of index operations and recursive
+# tree traversal -- and NOT disk I/O on thousands of files.
+
+test_description="Tests performance of read-tree"
+
+. ./perf-lib.sh
+
+test_perf_default_repo
+
+# If the test repo was generated by ./repos/many-files.sh
+# then we know something about the data shape and branches,
+# so we can isolate testing to the ballast-related commits
+# and setup sparse-checkout so we don't have to populate
+# the ballast files and directories.
+#
+# Otherwise, we make some general assumptions about the
+# repo and consider the entire history of the current
+# branch to be the ballast.
+
+test_expect_success "setup repo" '
+	if git rev-parse --verify refs/heads/p0006-ballast^{commit}
+	then
+		echo Assuming synthetic repo from many-files.sh
+		git branch br_base            master
+		git branch br_ballast         p0006-ballast
+		git config --local core.sparsecheckout 1
+		cat >.git/info/sparse-checkout <<-EOF
+		/*
+		!ballast/*
+		EOF
+	else
+		echo Assuming non-synthetic repo...
+		git branch br_base            $(git rev-list HEAD | tail -n 1)
+		git branch br_ballast         HEAD
+	fi &&
+	git checkout -q br_ballast &&
+	nr_files=$(git ls-files | wc -l)
+'
+
+test_perf "read-tree status br_ballast ($nr_files)" '
+	git read-tree HEAD &&
+	git status
+'
+
+test_done
