lib: Spin fast, retire early

When using the pollable spinner, we often want to use it as a means of
ensuring the task is running on the GPU before switching to something
else. In which case we don't want to add extra delay inside the spinner,
but the current 1000 NOPs add on order of 5us, which is often larger
than the target latency.

v2: Don't change perf_pmu as that is sensitive to the extra CPU latency
from a tight GPU spinner.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Antonio Argenziano <antonio.argenziano@intel.com> #v1
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> #v1
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

1 files changed, 2 insertions, 1 deletions

diff --git a/lib/igt_dummyload.c b/lib/igt_dummyload.c
index 94efdf74..7beb6624 100644
--- a/lib/igt_dummyload.c
+++ b/lib/igt_dummyload.c
@@ -199,7 +199,8 @@ emit_recursive_batch(igt_spin_t *spin,
 	 * between function calls, that appears enough to keep SNB out of
 	 * trouble. See https://bugs.freedesktop.org/show_bug.cgi?id=102262
 	 */
-	batch += 1000;
+	if (!(opts->flags & IGT_SPIN_FAST))
+		batch += 1000;
 
 	/* recurse */
 	r = &relocs[obj[BATCH].relocation_count++];