drm/amd/display: Prevent premature pageflip when comitting in vblank. (v3)
Subsystems (3)
PATCH 2/2 drm/amd/display: Prevent premature pageflip when comitting in vblank. (v3)
Make sure we do not program a hw pageflip inside vblank 'n' iff the atomic flip is comitted while inside the same vblank 'n'. We must defer such a flip by one refresh cycle to vblank 'n+1', unless this is a DRM_MODE_PAGE_FLIP_ASYNC async pageflip, which must always execute as soon as possible.
Without this, pageflips programmed via X11 GLX_OML_sync_control extensions glXSwapBuffersMscOML(..., target_msc, ...); call and/or via DRI3/Present PresentPixmap(..., target_msc, ...); request will complete one vblank too early whenever target_msc > current_msc + 1, ie. more than 1 vblank in the future. In such a case, the call of the pageflip ioctl() would be triggered by a queued drmWaitVblank() vblank event, which itself gets dispatched inside the vblank one frame before the target_msc vblank.
Testing with this patch does no longer show any problems with OML_sync_control swap scheduling or flip completion timestamps. Tested on R9 380 Tonga.
v2: Add acked/r-b by Harry and Michel.
v3: Feedback from Andrey: Must not wait an extra frame for
DRM_MODE_PAGE_FLIP_ASYNC flips.
--- .../gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_types.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-)
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_types.c +10 -10
@@ -2460,6 +2460,9 @@ static void amdgpu_dm_do_flip(struct amdgpu_device *adev = crtc->dev->dev_private;bool async_flip = (acrtc->flip_flags &DRM_MODE_PAGE_FLIP_ASYNC ) !=0 ; +/* Prepare wait for target vblank early - before the fence-waits */ + target_vblank = target -drm_crtc_vblank_count (crtc) + +amdgpu_get_vblank_counter_kms (crtc->dev, acrtc->crtc_id);/*TODO This might fail and hence better not used, wait * explicitly on fences instead
@@ -2478,13 +2481,9 @@ static void amdgpu_dm_do_flip(amdgpu_bo_unreserve (abo); -/* Wait for target vblank */ /* Wait until we're out of the vertical blank period before the one * targeted by the flip */ - target_vblank = target -drm_crtc_vblank_count (crtc) + -amdgpu_get_vblank_counter_kms (crtc->dev, acrtc->crtc_id); - while ((acrtc->enabled && (amdgpu_get_crtc_scanoutpos(adev->ddev, acrtc->crtc_id, 0, &vpos, &hpos, NULL, NULL,
@@ -2760,14 +2759,15 @@ void amdgpu_dm_atomic_commit_tail( pflip_needed = !state->allow_modeset;if (pflip_needed) { + wait_for_vblank = + acrtc->flip_flags &DRM_MODE_PAGE_FLIP_ASYNC ? + false : true; +amdgpu_dm_do_flip ( - crtc, - fb, -drm_crtc_vblank_count (crtc)); + crtc, + fb, +drm_crtc_vblank_count (crtc) + wait_for_vblank); - wait_for_vblank = - acrtc->flip_flags &DRM_MODE_PAGE_FLIP_ASYNC ? - false : true;/*clean up the flags for next usage*/ acrtc->flip_flags =0 ; }
Signature · 6 lines
2.7.4 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Quoted 105 lines
-----Original Message-----
From: Mario Kleiner [mailto:[email protected]]
Sent: Monday, April 24, 2017 12:54 PM
To: [email protected]
Cc: [email protected]; Grodzovsky, Andrey; [email protected]; Wentland, Harry; Deucher, Alexander; Daenzer, Michel
Subject: [PATCH 2/2] drm/amd/display: Prevent premature pageflip when comitting in vblank. (v3)Make sure we do not program a hw pageflip inside vblank 'n' iff the atomic flip is comitted while inside the same vblank 'n'. We must defer such a flip by one refresh cycle to vblank 'n+1', unless this is a
DRM_MODE_PAGE_FLIP_ASYNCasync pageflip, which must always execute as soon as possible.Without this, pageflips programmed via X11
GLX_OML_sync_controlextensionsglXSwapBuffersMscOML(..., target_msc, ...); call and/or via DRI3/PresentPresentPixmap(..., target_msc, ...); request will complete one vblank too early whenevertarget_msc>current_msc+ 1, ie. more than 1 vblank in the future. In such a case, the call of the pageflipioctl()would be triggered by a queueddrmWaitVblank()vblank event, which itself gets dispatched inside the vblank one frame before thetarget_mscvblank.Testing with this patch does no longer show any problems with
OML_sync_controlswap scheduling or flip completion timestamps. Tested on R9 380 Tonga.v2: Add acked/r-b by Harry and Michel.
v3: Feedback from Andrey: Must not wait an extra frame for
DRM_MODE_PAGE_FLIP_ASYNCflips.Signed-off-by Mario Kleiner <[email protected]>Acked-by Harry Wentland <[email protected]>Reviewed-by Michel Dänzer <[email protected]>Cc Harry Wentland <[email protected]>Cc Alex Deucher <[email protected]>Cc Michel Dänzer <[email protected]>--- .../gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_types.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-)drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_types.c +3
-0@@ -2460,6 +2460,9 @@ static void amdgpu_dm_do_flip(struct amdgpu_device *adev = crtc->dev->dev_private;bool async_flip = (acrtc->flip_flags &DRM_MODE_PAGE_FLIP_ASYNC ) !=0 ; +/* Prepare wait for target vblank early - before the fence-waits */ + target_vblank = target -drm_crtc_vblank_count (crtc) + +amdgpu_get_vblank_counter_kms (crtc->dev, acrtc- >crtc_id);drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_types.c +0
-4/*TODO This might fail and hence better not used, wait * explicitly on fences instead @@ -2478,13 +2481,9 @@ static void amdgpu_dm_do_flip(amdgpu_bo_unreserve (abo); -/* Wait for target vblank */ /* Wait until we're out of the vertical blank period before the one * targeted by the flip */ - target_vblank = target -drm_crtc_vblank_count (crtc) + -amdgpu_get_vblank_counter_kms (crtc->dev, acrtc- >crtc_id); - while ((acrtc->enabled && (amdgpu_get_crtc_scanoutpos(adev->ddev, acrtc->crtc_id,0,
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_types.c +2
-0&vpos, &hpos, NULL, NULL,@@ -2760,14 +2759,15 @@ void amdgpu_dm_atomic_commit_tail( pflip_needed = !state->allow_modeset;if (pflip_needed) { + wait_for_vblank = + acrtc->flip_flags &
DRM_MODE_PAGE_FLIP_ASYNC?drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_types.c +5
-3+ false : true; + amdgpu_dm_do_flip( - crtc, - fb, -drm_crtc_vblank_count (crtc)); + crtc, + fb, + drm_crtc_vblank_count(crtc) +
wait_for_vblank);drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_types.c +0
-2- wait_for_vblank = - acrtc->flip_flags &
DRM_MODE_PAGE_FLIP_ASYNC?drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_types.c +0
-2- false : true;/*clean up the flags for next usage*/ acrtc->flip_flags =0 ; } --
Quoted 1 line
2.7.4
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
A message that is not in the archive (sent off-list or not yet received): 1493052860-15178-1-git-send-email-mario.kleiner.de-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org
Quoted 18 lines · Mario Kleiner
Make sure we do not program a hw pageflip inside vblank 'n' iff the atomic flip is comitted while inside the same vblank 'n'. We must defer such a flip by one refresh cycle to vblank 'n+1', unless this is a
DRM_MODE_PAGE_FLIP_ASYNCasync pageflip, which must always execute as soon as possible.Without this, pageflips programmed via X11
GLX_OML_sync_controlextensionsglXSwapBuffersMscOML(..., target_msc, ...); call and/or via DRI3/PresentPresentPixmap(..., target_msc, ...); request will complete one vblank too early whenevertarget_msc>current_msc+ 1, ie. more than 1 vblank in the future. In such a case, the call of the pageflipioctl()would be triggered by a queueddrmWaitVblank()vblank event, which itself gets dispatched inside the vblank one frame before thetarget_mscvblank.Testing with this patch does no longer show any problems with
OML_sync_controlswap scheduling or flip completion timestamps. Tested on R9 380 Tonga.
Quoted 7 lines
v2: Add acked/r-b by Harry and Michel.
v3: Feedback from Andrey: Must not wait an extra frame for
DRM_MODE_PAGE_FLIP_ASYNCflips.Signed-off-by Mario Kleiner <[email protected]>Acked-by Harry Wentland <[email protected]>Reviewed-by Michel Dänzer <[email protected]>
This patch and v2 of patch 1 pushed to the internal amd-staging-4.9 branch, thanks!
Signature · 6 lines
Earthling Michel Dänzer | http://www.amd.com Libre software enthusiast | Mesa and X developer _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx