glbench: Reduce max iteration duration to 0.4s
On trogdor, the longer duration is bumping into the kernel's hangcheck
timer, which means the higher iteration counts are not necessarily
measuring anything useful.
(glbench is already not really measuring anything useful on trogdor, due
to minigbm disabling UBWC bandwidth compression for anything that uses
waffle, to work around waffle limitations.. but lets make glbench
measure something that is at least a bit less useless.)
BUG=b:179122157
TEST=Run on trogdor
Change-Id: I29dbf9726f633599297ff789fb95a0b0bf25d6fc
Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/glbench/+/2737682
Tested-by: Rob Clark <robdclark@chromium.org>
Auto-Submit: Rob Clark <robdclark@chromium.org>
Reviewed-by: Ilja H. Friedel <ihf@chromium.org>
Commit-Queue: Rob Clark <robdclark@chromium.org>
diff --git a/src/testbase.cc b/src/testbase.cc
index 3279ad9..9b77bf4 100644
--- a/src/testbase.cc
+++ b/src/testbase.cc
@@ -35,10 +35,10 @@
return time2 - time1;
}
-// Target minimum iteration duration of 1s. This means the final/longest
-// iteration is between 1s and 2s and the machine is active for 2s to 4s.
+// Target minimum iteration duration of 0.4s. This means the final/longest
+// iteration is between 0.4s and 0.8s and the machine is active for 0.8s to 1.6s.
// Notice as of March 2014 the BVT suite has a hard limit per job of 20 minutes.
-#define MIN_ITERATION_DURATION_US 1000000
+#define MIN_ITERATION_DURATION_US 400000
#define MAX_TESTNAME 46
