[PATCH 10/40] drm/i915/execlists: Reduce barrier on context switch to a wmb()

Chris Wilson chris at chris-wilson.co.uk
Sun Nov 10 17:31:21 UTC 2019


Having been forced to reduce Braswell back to using the aliasing ppgtt,
the coherency issue we previously observed cannot impact us. Reduce the
performance penalty imposed on all platforms from using the mfence to a
mere sfence.

References: cf66b8a0ba14 ("drm/i915/execlists: Apply a full mb before execution for Braswell")
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala at linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen at linux.intel.com>
---
 drivers/gpu/drm/i915/gt/intel_lrc.c | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index a06074a26fb5..4b6d9e6b1bfd 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -1218,13 +1218,8 @@ static u64 execlists_update_context(const struct i915_request *rq)
 	 * may not be visible to the HW prior to the completion of the UC
 	 * register write and that we may begin execution from the context
 	 * before its image is complete leading to invalid PD chasing.
-	 *
-	 * Furthermore, Braswell, at least, wants a full mb to be sure that
-	 * the writes are coherent in memory (visible to the GPU) prior to
-	 * execution, and not just visible to other CPUs (as is the result of
-	 * wmb).
 	 */
-	mb();
+	wmb();
 
 	desc = ce->lrc_desc;
 	ce->lrc_desc &= ~CTX_DESC_FORCE_RESTORE;
-- 
2.24.0



More information about the Intel-gfx-trybot mailing list