[Intel-gfx] [PATCH 6/8] [drm/i915] implement fastpath for overlay flip waiting
Daniel Vetter
daniel.vetter at ffwll.ch
Tue Sep 15 22:57:36 CEST 2009
As long as the gpu can keep up, neither the cpu (waiting for gpu)
nore the gpu (waiting for vblank to do an overlay flip) stalls.
Signed-off-by: Daniel Vetter <daniel.vetter at ffwll.ch>
---
drivers/gpu/drm/i915/i915_drv.h | 3 ++
drivers/gpu/drm/i915/i915_gem.c | 4 +-
drivers/gpu/drm/i915/intel_drv.h | 2 +
drivers/gpu/drm/i915/intel_overlay.c | 43 ++++++++++++++++++++++++++-------
4 files changed, 41 insertions(+), 11 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index d31198b..8b17942 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -724,6 +724,9 @@ void i915_gem_cleanup_ringbuffer(struct drm_device *dev);
int i915_gem_do_init(struct drm_device *dev, unsigned long start,
unsigned long end);
int i915_gem_idle(struct drm_device *dev);
+uint32_t i915_add_request(struct drm_device *dev, struct drm_file *file_priv,
+ uint32_t flush_domains);
+int i915_do_wait_request(struct drm_device *dev, uint32_t seqno, int interruptible);
int i915_lp_ring_sync(struct drm_device *dev);
int i915_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf);
int i915_gem_object_set_to_gtt_domain(struct drm_gem_object *obj,
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index a3d8fa8..4793766 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1518,7 +1518,7 @@ i915_gem_object_move_to_inactive(struct drm_gem_object *obj)
*
* Returned sequence numbers are nonzero on success.
*/
-static uint32_t
+uint32_t
i915_add_request(struct drm_device *dev, struct drm_file *file_priv,
uint32_t flush_domains)
{
@@ -1738,7 +1738,7 @@ i915_gem_retire_work_handler(struct work_struct *work)
mutex_unlock(&dev->struct_mutex);
}
-static int
+int
i915_do_wait_request(struct drm_device *dev, uint32_t seqno, int interruptible)
{
drm_i915_private_t *dev_priv = dev->dev_private;
diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h
index f84a0cf..9eea5c7 100644
--- a/drivers/gpu/drm/i915/intel_drv.h
+++ b/drivers/gpu/drm/i915/intel_drv.h
@@ -125,7 +125,9 @@ struct intel_overlay {
u32 flip_addr;
struct drm_i915_gem_object *reg_bo;
void *virt_addr;
+ /* flip handling */
int hw_wedged;
+ uint32_t last_flip_req;
};
struct intel_crtc {
diff --git a/drivers/gpu/drm/i915/intel_overlay.c b/drivers/gpu/drm/i915/intel_overlay.c
index 4e88abb..85e07e4 100644
--- a/drivers/gpu/drm/i915/intel_overlay.c
+++ b/drivers/gpu/drm/i915/intel_overlay.c
@@ -251,7 +251,6 @@ static void intel_overlay_continue(struct intel_overlay *overlay,
drm_i915_private_t *dev_priv = dev->dev_private;
u32 flip_addr = overlay->flip_addr;
u32 tmp;
- int ret;
RING_LOCALS;
BUG_ON(!overlay->active);
@@ -264,11 +263,40 @@ static void intel_overlay_continue(struct intel_overlay *overlay,
if (tmp & (1 << 17))
DRM_DEBUG("overlay underrun, DOVSTA: %x\n", tmp);
- BEGIN_LP_RING(6);
+ BEGIN_LP_RING(4);
OUT_RING(MI_FLUSH);
OUT_RING(MI_NOOP);
OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE);
OUT_RING(flip_addr);
+ ADVANCE_LP_RING();
+
+ overlay->last_flip_req = i915_add_request(dev, NULL, 0);
+}
+
+static int intel_overlay_wait_flip(struct intel_overlay *overlay)
+{
+ struct drm_device *dev = overlay->dev;
+ drm_i915_private_t *dev_priv = dev->dev_private;
+ int ret;
+ u32 tmp;
+ RING_LOCALS;
+
+ if (overlay->last_flip_req != 0) {
+ ret = i915_do_wait_request(dev, overlay->last_flip_req, 0);
+
+ if (ret != 0)
+ return ret;
+
+ overlay->last_flip_req = 0;
+
+ tmp = I915_READ(ISR);
+
+ if (!(tmp & I915_OVERLAY_PLANE_FLIP_PENDING_INTERRUPT))
+ return 0;
+ }
+
+ /* synchronous slowpath */
+ BEGIN_LP_RING(2);
OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
OUT_RING(MI_NOOP);
ADVANCE_LP_RING();
@@ -279,13 +307,8 @@ static void intel_overlay_continue(struct intel_overlay *overlay,
DRM_ERROR("intel overlay: ring sync failed, hw likely wedged\n");
overlay->hw_wedged = 1;
}
-}
-static int intel_overlay_wait_flip(struct intel_overlay *overlay)
-{
- /* don't overcomplicate things for now with asynchronous operations
- * see comment above */
- return 0;
+ return ret;
}
/* overlay needs to be disabled in OCMD reg */
@@ -344,7 +367,9 @@ static int intel_overlay_off(struct intel_overlay *overlay)
return ret;
}
-/* wait for pending overlay flip and release old frame */
+/* Wait for pending overlay flip and release old frame.
+ * Needs to be called before the overlay register are changed
+ * via intel_overlay_(un)map_regs_atomic */
static int intel_overlay_release_old_vid(struct intel_overlay *overlay)
{
int ret;
--
1.6.3.3
More information about the Intel-gfx
mailing list