Author: stefan2
Date: Thu Apr 12 18:57:19 2012
New Revision: 1325445
URL: http://svn.apache.org/viewvc?rev=1325445&view=rev
Log:
On the revprop-cache branch: Test code cleanup
* subversion/tests/libsvn_subr/named_atomic-test.c
(calibrate_iterations): drop overhead calculation; run larger intervals
(calibrate_concurrency): print results; use stricter performance limits
(test_multithreaded, test_multiprocess): remove printf
Modified:
subversion/branches/revprop-cache/subversion/tests/libsvn_subr/named_atomic-test.c
Modified:
subversion/branches/revprop-cache/subversion/tests/libsvn_subr/named_atomic-test.c
URL:
http://svn.apache.org/viewvc/subversion/branches/revprop-cache/subversion/tests/libsvn_subr/named_atomic-test.c?rev=1325445&r1=1325444&r2=1325445&view=diff
==============================================================================
---
subversion/branches/revprop-cache/subversion/tests/libsvn_subr/named_atomic-test.c
(original)
+++
subversion/branches/revprop-cache/subversion/tests/libsvn_subr/named_atomic-test.c
Thu Apr 12 18:57:19 2012
@@ -42,7 +42,7 @@
static int hw_thread_count = 0;
/* number of iterations that we should perform on concurrency tests
- * (will be calibrated to about 2s runtime)*/
+ * (will be calibrated to about 1s runtime)*/
static int suggested_iterations = 0;
/* Bring shared memory to a defined state. This is very useful in case of
@@ -249,39 +249,30 @@ run_procs(apr_pool_t *pool, const char *
}
/* Set SUGGESTED_ITERATIONS to a value that COUNT workers will take
- * about 2 seconds to execute.
+ * about 1 second to execute.
*/
static svn_error_t *
calibrate_iterations(apr_pool_t *pool, int count)
{
apr_time_t start;
- apr_time_t overhead;
int calib_iterations;
double taken = 0.0;
- /* measure start-up overhead */
-
- SVN_ERR(init_concurrency_test_shm(pool, count));
-
- start = apr_time_now();
- SVN_ERR(run_procs(pool, TEST_PROC, count, 10));
- overhead = apr_time_now() - start;
-
/* increase iterations until we pass the 10ms mark */
- for (calib_iterations = 100; taken < 10000.0; calib_iterations *= 2)
+ for (calib_iterations = 10; taken < 100000.0; calib_iterations *= 2)
{
SVN_ERR(init_concurrency_test_shm(pool, count));
start = apr_time_now();
- SVN_ERR(run_procs(pool, TEST_PROC, count, calib_iterations + 10));
+ SVN_ERR(run_procs(pool, TEST_PROC, count, calib_iterations));
- taken = (int)(apr_time_now() - start - overhead);
+ taken = apr_time_now() - start;
}
- /* scale that to 2s */
+ /* scale that to 1s */
- suggested_iterations = (int)(2000000.0 / taken * calib_iterations);
+ suggested_iterations = (int)(1000000.0 / taken * calib_iterations);
return SVN_NO_ERROR;
}
@@ -303,13 +294,16 @@ calibrate_concurrency(apr_pool_t *pool)
{
int saved_suggestion = suggested_iterations;
SVN_ERR(calibrate_iterations(pool, hw_thread_count * 2));
- if (suggested_iterations < 10000)
+ if (suggested_iterations < 100000)
{
suggested_iterations = saved_suggestion;
break;
}
}
- }
+
+ printf("using %d cores for %d iterations\n", hw_thread_count,
+ suggested_iterations);
+ }
return SVN_NO_ERROR;
}
@@ -565,7 +559,6 @@ test_multithreaded(apr_pool_t *pool)
{
SVN_ERR(calibrate_concurrency(pool));
- printf("%d %d\n", hw_thread_count, suggested_iterations);
SVN_ERR(init_concurrency_test_shm(pool, hw_thread_count));
SVN_ERR(run_threads(pool, hw_thread_count, suggested_iterations,
test_pipeline));
@@ -578,7 +571,6 @@ test_multiprocess(apr_pool_t *pool)
{
SVN_ERR(calibrate_concurrency(pool));
- printf("%d %d\n", hw_thread_count, suggested_iterations);
SVN_ERR(init_concurrency_test_shm(pool, hw_thread_count));
SVN_ERR(run_procs(pool, TEST_PROC, hw_thread_count, suggested_iterations));
