This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
[PATCH] benchtests: Rename argument to TIMING_INIT macro.
- From: Will Newton <will dot newton at linaro dot org>
- To: libc-alpha at sourceware dot org
- Cc: patches at linaro dot org
- Date: Wed, 11 Sep 2013 13:27:51 +0100
- Subject: [PATCH] benchtests: Rename argument to TIMING_INIT macro.
- Authentication-results: sourceware.org; auth=none
The TIMING_INIT macro currently sets the number of loop iterations
to 1000, which limits usefulness. Make the argument a clock
resolution value and multiply by 1000 in bench-skeleton.c instead
to allow easier reuse.
ChangeLog:
2013-09-11 Will Newton <will.newton@linaro.org>
* benchtests/bench-timing.h: Rename iters parameter
to res and remove hardcoded 1000 value.
* benchtests/bench-skeleton.c: Pass res parameter to
TIMING_INIT and multiply by 1000.
---
benchtests/bench-skeleton.c | 6 ++++--
benchtests/bench-timing.h | 14 ++++++--------
2 files changed, 10 insertions(+), 10 deletions(-)
diff --git a/benchtests/bench-skeleton.c b/benchtests/bench-skeleton.c
index 4e3a507..05edc69 100644
--- a/benchtests/bench-skeleton.c
+++ b/benchtests/bench-skeleton.c
@@ -53,9 +53,11 @@ main (int argc, char **argv)
memset (&runtime, 0, sizeof (runtime));
- unsigned long iters;
+ unsigned long iters, res;
- TIMING_INIT (iters);
+ TIMING_INIT (res);
+
+ iters = 1000 * res;
for (int v = 0; v < NUM_VARIANTS; v++)
{
diff --git a/benchtests/bench-timing.h b/benchtests/bench-timing.h
index 009813b..48a8521 100644
--- a/benchtests/bench-timing.h
+++ b/benchtests/bench-timing.h
@@ -25,10 +25,10 @@
hp_timing_t _dl_hp_timing_overhead;
typedef hp_timing_t timing_t;
-# define TIMING_INIT(iters) \
+# define TIMING_INIT(res) \
({ \
HP_TIMING_DIFF_INIT(); \
- (iters) = 1000; \
+ (res) = 1; \
})
# define TIMING_NOW(var) HP_TIMING_NOW (var)
@@ -43,15 +43,13 @@ typedef hp_timing_t timing_t;
#else
typedef uint64_t timing_t;
-/* Measure 1000 times the resolution of the clock. So for a 1ns
- resolution clock, we measure 1000 iterations of the function call at a
- time. Measurements close to the minimum clock resolution won't make
- much sense, but it's better than having nothing at all. */
-# define TIMING_INIT(iters) \
+/* Measure the resolution of the clock so we can scale the number of
+ benchmark iterations by this value. */
+# define TIMING_INIT(res) \
({ \
struct timespec start; \
clock_getres (CLOCK_PROCESS_CPUTIME_ID, &start); \
- (iters) = 1000 * start.tv_nsec; \
+ (res) = start.tv_nsec; \
})
# define TIMING_NOW(var) \
--
1.8.1.4