summaryrefslogtreecommitdiff
path: root/t/helper
diff options
context:
space:
mode:
Diffstat (limited to 't/helper')
-rw-r--r--t/helper/.gitignore1
-rw-r--r--t/helper/test-config.c15
-rw-r--r--t/helper/test-lazy-init-name-hash.c264
3 files changed, 280 insertions, 0 deletions
diff --git a/t/helper/.gitignore b/t/helper/.gitignore
index d6e8b36798..758ed2e8fa 100644
--- a/t/helper/.gitignore
+++ b/t/helper/.gitignore
@@ -11,6 +11,7 @@
/test-genrandom
/test-hashmap
/test-index-version
+/test-lazy-init-name-hash
/test-line-buffer
/test-match-trees
/test-mergesort
diff --git a/t/helper/test-config.c b/t/helper/test-config.c
index 83a4f2ab86..8e3ed6a76c 100644
--- a/t/helper/test-config.c
+++ b/t/helper/test-config.c
@@ -66,6 +66,16 @@ static int iterate_cb(const char *var, const char *value, void *data)
return 0;
}
+static int early_config_cb(const char *var, const char *value, void *vdata)
+{
+ const char *key = vdata;
+
+ if (!strcmp(key, var))
+ printf("%s\n", value);
+
+ return 0;
+}
+
int cmd_main(int argc, const char **argv)
{
int i, val;
@@ -73,6 +83,11 @@ int cmd_main(int argc, const char **argv)
const struct string_list *strptr;
struct config_set cs;
+ if (argc == 3 && !strcmp(argv[1], "read_early_config")) {
+ read_early_config(early_config_cb, (void *)argv[2]);
+ return 0;
+ }
+
setup_git_directory();
git_configset_init(&cs);
diff --git a/t/helper/test-lazy-init-name-hash.c b/t/helper/test-lazy-init-name-hash.c
new file mode 100644
index 0000000000..6368a89345
--- /dev/null
+++ b/t/helper/test-lazy-init-name-hash.c
@@ -0,0 +1,264 @@
+#include "cache.h"
+#include "parse-options.h"
+
+static int single;
+static int multi;
+static int count = 1;
+static int dump;
+static int perf;
+static int analyze;
+static int analyze_step;
+
+/*
+ * Dump the contents of the "dir" and "name" hash tables to stdout.
+ * If you sort the result, you can compare it with the other type
+ * mode and verify that both single and multi produce the same set.
+ */
+static void dump_run(void)
+{
+ struct hashmap_iter iter_dir;
+ struct hashmap_iter iter_cache;
+
+ /* Stolen from name-hash.c */
+ struct dir_entry {
+ struct hashmap_entry ent;
+ struct dir_entry *parent;
+ int nr;
+ unsigned int namelen;
+ char name[FLEX_ARRAY];
+ };
+
+ struct dir_entry *dir;
+ struct cache_entry *ce;
+
+ read_cache();
+ if (single) {
+ test_lazy_init_name_hash(&the_index, 0);
+ } else {
+ int nr_threads_used = test_lazy_init_name_hash(&the_index, 1);
+ if (!nr_threads_used)
+ die("non-threaded code path used");
+ }
+
+ dir = hashmap_iter_first(&the_index.dir_hash, &iter_dir);
+ while (dir) {
+ printf("dir %08x %7d %s\n", dir->ent.hash, dir->nr, dir->name);
+ dir = hashmap_iter_next(&iter_dir);
+ }
+
+ ce = hashmap_iter_first(&the_index.name_hash, &iter_cache);
+ while (ce) {
+ printf("name %08x %s\n", ce->ent.hash, ce->name);
+ ce = hashmap_iter_next(&iter_cache);
+ }
+
+ discard_cache();
+}
+
+/*
+ * Run the single or multi threaded version "count" times and
+ * report on the time taken.
+ */
+static uint64_t time_runs(int try_threaded)
+{
+ uint64_t t0, t1, t2;
+ uint64_t sum = 0;
+ uint64_t avg;
+ int nr_threads_used;
+ int i;
+
+ for (i = 0; i < count; i++) {
+ t0 = getnanotime();
+ read_cache();
+ t1 = getnanotime();
+ nr_threads_used = test_lazy_init_name_hash(&the_index, try_threaded);
+ t2 = getnanotime();
+
+ sum += (t2 - t1);
+
+ if (try_threaded && !nr_threads_used)
+ die("non-threaded code path used");
+
+ if (nr_threads_used)
+ printf("%f %f %d multi %d\n",
+ ((double)(t1 - t0))/1000000000,
+ ((double)(t2 - t1))/1000000000,
+ the_index.cache_nr,
+ nr_threads_used);
+ else
+ printf("%f %f %d single\n",
+ ((double)(t1 - t0))/1000000000,
+ ((double)(t2 - t1))/1000000000,
+ the_index.cache_nr);
+ fflush(stdout);
+
+ discard_cache();
+ }
+
+ avg = sum / count;
+ if (count > 1)
+ printf("avg %f %s\n",
+ (double)avg/1000000000,
+ (try_threaded) ? "multi" : "single");
+
+ return avg;
+}
+
+/*
+ * Try a series of runs varying the "istate->cache_nr" and
+ * try to find a good value for the multi-threaded criteria.
+ */
+static void analyze_run(void)
+{
+ uint64_t t1s, t1m, t2s, t2m;
+ int cache_nr_limit;
+ int nr_threads_used;
+ int i;
+ int nr;
+
+ read_cache();
+ cache_nr_limit = the_index.cache_nr;
+ discard_cache();
+
+ nr = analyze;
+ while (1) {
+ uint64_t sum_single = 0;
+ uint64_t sum_multi = 0;
+ uint64_t avg_single;
+ uint64_t avg_multi;
+
+ if (nr > cache_nr_limit)
+ nr = cache_nr_limit;
+
+ for (i = 0; i < count; i++) {
+ read_cache();
+ the_index.cache_nr = nr; /* cheap truncate of index */
+ t1s = getnanotime();
+ test_lazy_init_name_hash(&the_index, 0);
+ t2s = getnanotime();
+ sum_single += (t2s - t1s);
+ the_index.cache_nr = cache_nr_limit;
+ discard_cache();
+
+ read_cache();
+ the_index.cache_nr = nr; /* cheap truncate of index */
+ t1m = getnanotime();
+ nr_threads_used = test_lazy_init_name_hash(&the_index, 1);
+ t2m = getnanotime();
+ sum_multi += (t2m - t1m);
+ the_index.cache_nr = cache_nr_limit;
+ discard_cache();
+
+ if (!nr_threads_used)
+ printf(" [size %8d] [single %f] non-threaded code path used\n",
+ nr, ((double)(t2s - t1s))/1000000000);
+ else
+ printf(" [size %8d] [single %f] %c [multi %f %d]\n",
+ nr,
+ ((double)(t2s - t1s))/1000000000,
+ (((t2s - t1s) < (t2m - t1m)) ? '<' : '>'),
+ ((double)(t2m - t1m))/1000000000,
+ nr_threads_used);
+ fflush(stdout);
+ }
+ if (count > 1) {
+ avg_single = sum_single / count;
+ avg_multi = sum_multi / count;
+ if (!nr_threads_used)
+ printf("avg [size %8d] [single %f]\n",
+ nr,
+ (double)avg_single/1000000000);
+ else
+ printf("avg [size %8d] [single %f] %c [multi %f %d]\n",
+ nr,
+ (double)avg_single/1000000000,
+ (avg_single < avg_multi ? '<' : '>'),
+ (double)avg_multi/1000000000,
+ nr_threads_used);
+ fflush(stdout);
+ }
+
+ if (nr >= cache_nr_limit)
+ return;
+ nr += analyze_step;
+ }
+}
+
+int cmd_main(int argc, const char **argv)
+{
+ const char *usage[] = {
+ "test-lazy-init-name-hash -d (-s | -m)",
+ "test-lazy-init-name-hash -p [-c c]",
+ "test-lazy-init-name-hash -a a [--step s] [-c c]",
+ "test-lazy-init-name-hash (-s | -m) [-c c]",
+ "test-lazy-init-name-hash -s -m [-c c]",
+ NULL
+ };
+ struct option options[] = {
+ OPT_BOOL('s', "single", &single, "run single-threaded code"),
+ OPT_BOOL('m', "multi", &multi, "run multi-threaded code"),
+ OPT_INTEGER('c', "count", &count, "number of passes"),
+ OPT_BOOL('d', "dump", &dump, "dump hash tables"),
+ OPT_BOOL('p', "perf", &perf, "compare single vs multi"),
+ OPT_INTEGER('a', "analyze", &analyze, "analyze different multi sizes"),
+ OPT_INTEGER(0, "step", &analyze_step, "analyze step factor"),
+ OPT_END(),
+ };
+ const char *prefix;
+ uint64_t avg_single, avg_multi;
+
+ prefix = setup_git_directory();
+
+ argc = parse_options(argc, argv, prefix, options, usage, 0);
+
+ /*
+ * istate->dir_hash is only created when ignore_case is set.
+ */
+ ignore_case = 1;
+
+ if (dump) {
+ if (perf || analyze > 0)
+ die("cannot combine dump, perf, or analyze");
+ if (count > 1)
+ die("count not valid with dump");
+ if (single && multi)
+ die("cannot use both single and multi with dump");
+ if (!single && !multi)
+ die("dump requires either single or multi");
+ dump_run();
+ return 0;
+ }
+
+ if (perf) {
+ if (analyze > 0)
+ die("cannot combine dump, perf, or analyze");
+ if (single || multi)
+ die("cannot use single or multi with perf");
+ avg_single = time_runs(0);
+ avg_multi = time_runs(1);
+ if (avg_multi > avg_single)
+ die("multi is slower");
+ return 0;
+ }
+
+ if (analyze) {
+ if (analyze < 500)
+ die("analyze must be at least 500");
+ if (!analyze_step)
+ analyze_step = analyze;
+ if (single || multi)
+ die("cannot use single or multi with analyze");
+ analyze_run();
+ return 0;
+ }
+
+ if (!single && !multi)
+ die("require either -s or -m or both");
+
+ if (single)
+ time_runs(0);
+ if (multi)
+ time_runs(1);
+
+ return 0;
+}
generated by cgit v1.2.3 (git 2.25.1) at 2024-12-13 03:35:32 +0000