Since requests can no longer be generated as a side-effect of
intel_ring_begin(), we know that the seqno will be unchanged during
ring-emission. This predicatablity then means we do not have to check
for the seqno wrapping around whilst emitting the semaphore for
engine->sync_to().
Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
---
drivers/gpu/drm/i915/i915_gem.c | 13 ++-----
drivers/gpu/drm/i915/intel_ringbuffer.c | 67 ++++++++++++++-------------------
drivers/gpu/drm/i915/intel_ringbuffer.h | 5 +--
3 files changed, 33 insertions(+), 52 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 235a3de6e0a0..b0230e7151ce 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2572,22 +2572,15 @@ __i915_gem_object_sync(struct drm_i915_gem_object *obj,
i915_gem_object_retire_request(obj, from);
} else {
int idx = intel_engine_sync_index(from->engine, to->engine);
- u32 seqno = i915_gem_request_get_seqno(from);
-
- if (seqno <= from->engine->semaphore.sync_seqno[idx])
+ if (from->fence.seqno <= from->engine->semaphore.sync_seqno[idx])
return 0;
trace_i915_gem_ring_sync_to(to, from);
- ret = to->engine->semaphore.sync_to(to, from->engine, seqno);
+ ret = to->engine->semaphore.sync_to(to, from);
if (ret)
return ret;
- /* We use last_read_req because sync_to()
- * might have just caused seqno wrap under
- * the radar.
- */
- from->engine->semaphore.sync_seqno[idx] =
- i915_gem_request_get_seqno(obj->last_read_req[from->engine->id]);
+ from->engine->semaphore.sync_seqno[idx] = from->fence.seqno;
}
return 0;
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index 556e9e2c1fec..d37cdb2f9073 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -1384,69 +1384,58 @@ static inline bool i915_gem_has_seqno_wrapped(struct drm_i915_private *dev_priv,
*/
static int
-gen8_ring_sync(struct drm_i915_gem_request *waiter_req,
- struct intel_engine_cs *signaller,
- u32 seqno)
+gen8_ring_sync(struct drm_i915_gem_request *wait,
+ struct drm_i915_gem_request *signal)
{
- struct intel_ring *waiter = waiter_req->ring;
- struct drm_i915_private *dev_priv = waiter_req->i915;
+ struct intel_ring *waiter = wait->ring;
+ struct drm_i915_private *dev_priv = wait->i915;
int ret;
- ret = intel_ring_begin(waiter_req, 4);
+ ret = intel_ring_begin(wait, 4);
if (ret)
return ret;
- intel_ring_emit(waiter, MI_SEMAPHORE_WAIT |
- MI_SEMAPHORE_GLOBAL_GTT |
- MI_SEMAPHORE_POLL |
- MI_SEMAPHORE_SAD_GTE_SDD);
- intel_ring_emit(waiter, seqno);
intel_ring_emit(waiter,
- lower_32_bits(GEN8_WAIT_OFFSET(waiter_req->engine,
- signaller->id)));
+ MI_SEMAPHORE_WAIT |
+ MI_SEMAPHORE_GLOBAL_GTT |
+ MI_SEMAPHORE_POLL |
+ MI_SEMAPHORE_SAD_GTE_SDD);
+ intel_ring_emit(waiter, signal->fence.seqno);
intel_ring_emit(waiter,
- upper_32_bits(GEN8_WAIT_OFFSET(waiter_req->engine,
- signaller->id)));
+ lower_32_bits(GEN8_WAIT_OFFSET(wait->engine,
+ signal->engine->id)));
+ intel_ring_emit(waiter,
+ upper_32_bits(GEN8_WAIT_OFFSET(wait->engine,
+ signal->engine->id)));
intel_ring_advance(waiter);
return 0;
}
static int
-gen6_ring_sync(struct drm_i915_gem_request *waiter_req,
- struct intel_engine_cs *signaller,
- u32 seqno)
+gen6_ring_sync(struct drm_i915_gem_request *wait,
+ struct drm_i915_gem_request *signal)
{
- struct intel_ring *waiter = waiter_req->ring;
+ struct intel_ring *waiter = wait->ring;
u32 dw1 = MI_SEMAPHORE_MBOX |
MI_SEMAPHORE_COMPARE |
MI_SEMAPHORE_REGISTER;
- u32 wait_mbox = signaller->semaphore.mbox.wait[waiter_req->engine->id];
+ u32 wait_mbox = signal->engine->semaphore.mbox.wait[wait->engine->id];
int ret;
- /* Throughout all of the GEM code, seqno passed implies our current
- * seqno is >= the last seqno executed. However for hardware the
- * comparison is strictly greater than.
- */
- seqno -= 1;
-
WARN_ON(wait_mbox == MI_SEMAPHORE_SYNC_INVALID);
- ret = intel_ring_begin(waiter_req, 4);
+ ret = intel_ring_begin(wait, 4);
if (ret)
return ret;
- /* If seqno wrap happened, omit the wait with no-ops */
- if (likely(!i915_gem_has_seqno_wrapped(waiter_req->i915, seqno))) {
- intel_ring_emit(waiter, dw1 | wait_mbox);
- intel_ring_emit(waiter, seqno);
- intel_ring_emit(waiter, 0);
- intel_ring_emit(waiter, MI_NOOP);
- } else {
- intel_ring_emit(waiter, MI_NOOP);
- intel_ring_emit(waiter, MI_NOOP);
- intel_ring_emit(waiter, MI_NOOP);
- intel_ring_emit(waiter, MI_NOOP);
- }
+ intel_ring_emit(waiter, dw1 | wait_mbox);
+ /* Throughout all of the GEM code, seqno passed implies our current
+ * seqno is >= the last seqno executed. However for hardware the
+ * comparison is strictly greater than.
+ */
+ intel_ring_emit(waiter, signal->fence.seqno - 1);
+ intel_ring_emit(waiter, 0);
+ intel_ring_emit(waiter, MI_NOOP);
intel_ring_advance(waiter);
return 0;
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
index 8147ce1379fb..fc9c1e453be1 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.h
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -283,9 +283,8 @@ struct intel_engine_cs {
};
/* AKA wait() */
- int (*sync_to)(struct drm_i915_gem_request *to_req,
- struct intel_engine_cs *from,
- u32 seqno);
+ int (*sync_to)(struct drm_i915_gem_request *to,
+ struct drm_i915_gem_request *from);
int (*signal)(struct drm_i915_gem_request *signaller_req,
/* num_dwords needed by caller */
unsigned int num_dwords);
--
2.7.0.rc3
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
