fscache: fscache takes an initial size

Update enable_fscache() to take an optional initial size parameter which is
used to initialize the hashmap so that it can avoid having to rehash as
additional entries are added.

Add a separate disable_fscache() macro to make the code clearer and easier
to read.

Signed-off-by: Ben Peart <benpeart@microsoft.com>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>

Author: benpeart

builtin/add.c

@@ -597,7 +597,7 @@ int cmd_add(int argc, const char **argv, const char *prefix)
 	die_in_unpopulated_submodule(&the_index, prefix);
 	die_path_inside_submodule(&the_index, &pathspec);
 
-	enable_fscache(1);
+	enable_fscache(0);
 	/* We do not really re-read the index but update the up-to-date flags */
 	preload_index(&the_index, &pathspec, 0);
 

builtin/checkout.c

@@ -391,7 +391,7 @@ static int checkout_worktree(const struct checkout_opts *opts,
 	if (pc_workers > 1)
 		init_parallel_checkout();
 
-	enable_fscache(1);
+	enable_fscache(active_nr);
 	for (pos = 0; pos < active_nr; pos++) {
 		struct cache_entry *ce = active_cache[pos];
 		if (ce->ce_flags & CE_MATCHED) {
@@ -416,7 +416,7 @@ static int checkout_worktree(const struct checkout_opts *opts,
 		errs |= run_parallel_checkout(&state, pc_workers, pc_threshold,
 					      NULL, NULL);
 	mem_pool_discard(&ce_mem_pool, should_validate_cache_entries());
-	enable_fscache(0);
+	disable_fscache();
 	remove_marked_cache_entries(&the_index, 1);
 	remove_scheduled_dirs();
 	errs |= finish_delayed_checkout(&state, &nr_checkouts, opts->show_progress);

builtin/commit.c

@@ -1540,7 +1540,7 @@ int cmd_status(int argc, const char **argv, const char *prefix)
 		       PATHSPEC_PREFER_FULL,
 		       prefix, argv);
 
-	enable_fscache(1);
+	enable_fscache(0);
 	if (status_format != STATUS_FORMAT_PORCELAIN &&
 	    status_format != STATUS_FORMAT_PORCELAIN_V2)
 		progress_flag = REFRESH_PROGRESS;
@@ -1581,7 +1581,7 @@ int cmd_status(int argc, const char **argv, const char *prefix)
 	wt_status_print(&s);
 	wt_status_collect_free_buffers(&s);
 
-	enable_fscache(0);
+	disable_fscache();
 	return 0;
 }
 

compat/win32/fscache.c

@@ -403,7 +403,7 @@ static struct fsentry *fscache_get(struct fsentry *key)
  * Enables or disables the cache. Note that the cache is read-only, changes to
  * the working directory are NOT reflected in the cache while enabled.
  */
-int fscache_enable(int enable)
+int fscache_enable(int enable, size_t initial_size)
 {
 	int result;
 
@@ -419,7 +419,11 @@ int fscache_enable(int enable)
 		InitializeCriticalSection(&mutex);
 		lstat_requests = opendir_requests = 0;
 		fscache_misses = fscache_requests = 0;
-		hashmap_init(&map, (hashmap_cmp_fn) fsentry_cmp, NULL, 0);
+		/*
+		 * avoid having to rehash by leaving room for the parent dirs.
+		 * '4' was determined empirically by testing several repos
+		 */
+		hashmap_init(&map, (hashmap_cmp_fn) fsentry_cmp, NULL, initial_size * 4);
 		initialized = 1;
 	}
 

compat/win32/fscache.h

@@ -1,8 +1,9 @@
 #ifndef FSCACHE_H
 #define FSCACHE_H
 
-int fscache_enable(int enable);
-#define enable_fscache(x) fscache_enable(x)
+int fscache_enable(int enable, size_t initial_size);
+#define enable_fscache(initial_size) fscache_enable(1, initial_size)
+#define disable_fscache() fscache_enable(0, 0)
 
 int fscache_enabled(const char *path);
 #define is_fscache_enabled(path) fscache_enabled(path)

fetch-pack.c

@@ -700,7 +700,7 @@ static void mark_complete_and_common_ref(struct fetch_negotiator *negotiator,
 	save_commit_buffer = 0;
 
 	trace2_region_enter("fetch-pack", "parse_remote_refs_and_find_cutoff", NULL);
-	enable_fscache(1);
+	enable_fscache(0);
 	for (ref = *refs; ref; ref = ref->next) {
 		struct commit *commit;
 
@@ -727,7 +727,7 @@ static void mark_complete_and_common_ref(struct fetch_negotiator *negotiator,
 		if (!cutoff || cutoff < commit->date)
 			cutoff = commit->date;
 	}
-	enable_fscache(0);
+	disable_fscache();
 	trace2_region_leave("fetch-pack", "parse_remote_refs_and_find_cutoff", NULL);
 
 	/*

git-compat-util.h

@@ -1405,6 +1405,10 @@ static inline int is_missing_file_error(int errno_)
 #define enable_fscache(x) /* noop */
 #endif
 
+#ifndef disable_fscache
+#define disable_fscache() /* noop */
+#endif
+
 #ifndef is_fscache_enabled
 #define is_fscache_enabled(path) (0)
 #endif

preload-index.c

@@ -126,7 +126,7 @@ void preload_index(struct index_state *index,
 		pthread_mutex_init(&pd.mutex, NULL);
 	}
 
-	enable_fscache(1);
+	enable_fscache(index->cache_nr);
 	for (i = 0; i < threads; i++) {
 		struct thread_data *p = data+i;
 		int err;
@@ -157,7 +157,7 @@ void preload_index(struct index_state *index,
 	trace2_data_intmax("index", NULL, "preload/sum_lstat", t2_sum_lstat);
 	trace2_region_leave("index", "preload", NULL);
 
-	enable_fscache(0);
+	disable_fscache();
 }
 
 int repo_read_index_preload(struct repository *repo,

read-cache.c

@@ -1625,7 +1625,7 @@ int refresh_index(struct index_state *istate, unsigned int flags,
 	typechange_fmt = in_porcelain ? "T\t%s\n" : "%s: needs update\n";
 	added_fmt      = in_porcelain ? "A\t%s\n" : "%s: needs update\n";
 	unmerged_fmt   = in_porcelain ? "U\t%s\n" : "%s: needs merge\n";
-	enable_fscache(1);
+	enable_fscache(0);
 	/*
 	 * Use the multi-threaded preload_index() to refresh most of the
 	 * cache entries quickly then in the single threaded loop below,
@@ -1720,7 +1720,7 @@ int refresh_index(struct index_state *istate, unsigned int flags,
 	display_progress(progress, istate->cache_nr);
 	stop_progress(&progress);
 	trace_performance_leave("refresh index");
-	enable_fscache(0);
+	disable_fscache();
 	return has_errors;
 }