If any of the perf tests run into 0 time, not only are we liable to
divide by zero, but the result would be highly questionable.
Nevertheless, let's not have a div-by-zero error.
Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
Cc: Andi Shyti <andi.shyti@xxxxxxxxx>
Reviewed-by: Andi Shyti <andi.shyti@xxxxxxxxx>
---
.../drm/i915/selftests/intel_memory_region.c | 20 +++++++++++++------
1 file changed, 14 insertions(+), 6 deletions(-)
diff --git a/drivers/gpu/drm/i915/selftests/intel_memory_region.c b/drivers/gpu/drm/i915/selftests/intel_memory_region.c
index 75839db63bea..ce7adfa3bca0 100644
--- a/drivers/gpu/drm/i915/selftests/intel_memory_region.c
+++ b/drivers/gpu/drm/i915/selftests/intel_memory_region.c
@@ -852,14 +852,22 @@ static int _perf_memcpy(struct intel_memory_region *src_mr,
}
sort(t, ARRAY_SIZE(t), sizeof(*t), wrap_ktime_compare, NULL);
+ if (t[0] <= 0) {
+ /* ignore the impossible to protect our sanity */
+ pr_debug("Skipping %s src(%s, %s) -> dst(%s, %s) %14s %4lluKiB copy, unstable measurement [%lld, %lld]\n",
+ __func__,
+ src_mr->name, repr_type(src_type),
+ dst_mr->name, repr_type(dst_type),
+ tests[i].name, size >> 10,
+ t[0], t[4]);
+ continue;
+ }
+
pr_info("%s src(%s, %s) -> dst(%s, %s) %14s %4llu KiB copy: %5lld MiB/s\n",
__func__,
- src_mr->name,
- repr_type(src_type),
- dst_mr->name,
- repr_type(dst_type),
- tests[i].name,
- size >> 10,
+ src_mr->name, repr_type(src_type),
+ dst_mr->name, repr_type(dst_type),
+ tests[i].name, size >> 10,
div64_u64(mul_u32_u32(4 * size,
1000 * 1000 * 1000),
t[1] + 2 * t[2] + t[3]) >> 20);
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
