[PATCH] drm/amdgpu: Improve MES responsiveness during oversubscription

Jay Cornwall jay.cornwall at amd.com
Wed Oct 4 15:59:50 UTC 2023


When MES is oversubscribed it may not frequently check for new
command submissions from driver if the scheduling load is high.
Response latency as high as 5 seconds has been observed.

Enable a flag which adds a check for new commands between
scheduling quantums.

Signed-off-by: Jay Cornwall <jay.cornwall at amd.com>
Cc: Alexandru Tudor <alexandru.tudor at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/mes_v11_0.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c
index 4a3020b5b30f..31b26e6f0b30 100644
--- a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c
@@ -406,6 +406,7 @@ static int mes_v11_0_set_hw_resources(struct amdgpu_mes *mes)
 	mes_set_hw_res_pkt.disable_mes_log = 1;
 	mes_set_hw_res_pkt.use_different_vmid_compute = 1;
 	mes_set_hw_res_pkt.enable_reg_active_poll = 1;
+	mes_set_hw_res_pkt.enable_level_process_quantum_check = 1;
 	mes_set_hw_res_pkt.oversubscription_timer = 50;
 
 	return mes_v11_0_submit_pkt_and_poll_completion(mes,
-- 
2.25.1



More information about the amd-gfx mailing list