[Intel-gfx] [PATCH 24/42] drm/i915/bdw: Workload submission mechanism for Execlists
oscar.mateo at intel.com
oscar.mateo at intel.com
Fri Jul 11 17:49:16 CEST 2014
From: Oscar Mateo <oscar.mateo at intel.com>
This is what i915_gem_do_execbuffer calls when it wants to execute some
worload in an Execlists world.
Signed-off-by: Oscar Mateo <oscar.mateo at intel.com>
---
drivers/gpu/drm/i915/i915_drv.h | 6 ++
drivers/gpu/drm/i915/i915_gem_execbuffer.c | 4 +-
drivers/gpu/drm/i915/intel_lrc.c | 128 ++++++++++++++++++++++++++++-
3 files changed, 135 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 24faec3..9a3adff 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2232,6 +2232,12 @@ int i915_gem_set_domain_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_sw_finish_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
+void i915_gem_execbuffer_move_to_active(struct list_head *vmas,
+ struct intel_engine_cs *ring);
+void i915_gem_execbuffer_retire_commands(struct drm_device *dev,
+ struct drm_file *file,
+ struct intel_engine_cs *ring,
+ struct drm_i915_gem_object *obj);
int i915_gem_ringbuffer_submission(struct drm_device *dev,
struct drm_file *file,
struct intel_engine_cs *ring,
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index 491ef7e..41e6467 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -962,7 +962,7 @@ i915_gem_validate_context(struct drm_device *dev, struct drm_file *file,
return ctx;
}
-static void
+void
i915_gem_execbuffer_move_to_active(struct list_head *vmas,
struct intel_engine_cs *ring)
{
@@ -994,7 +994,7 @@ i915_gem_execbuffer_move_to_active(struct list_head *vmas,
}
}
-static void
+void
i915_gem_execbuffer_retire_commands(struct drm_device *dev,
struct drm_file *file,
struct intel_engine_cs *ring,
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index f4b3834..40fc671 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -89,6 +89,57 @@ int intel_sanitize_enable_execlists(struct drm_device *dev, int enable_execlists
return 0;
}
+static int logical_ring_invalidate_all_caches(struct intel_ringbuffer *ringbuf)
+{
+ struct intel_engine_cs *ring = ringbuf->ring;
+ uint32_t flush_domains;
+ int ret;
+
+ flush_domains = 0;
+ if (ring->gpu_caches_dirty)
+ flush_domains = I915_GEM_GPU_DOMAINS;
+
+ ret = ring->emit_flush(ringbuf, I915_GEM_GPU_DOMAINS, flush_domains);
+ if (ret)
+ return ret;
+
+ ring->gpu_caches_dirty = false;
+ return 0;
+}
+
+static int execlists_move_to_gpu(struct intel_ringbuffer *ringbuf,
+ struct list_head *vmas)
+{
+ struct intel_engine_cs *ring = ringbuf->ring;
+ struct i915_vma *vma;
+ uint32_t flush_domains = 0;
+ bool flush_chipset = false;
+ int ret;
+
+ list_for_each_entry(vma, vmas, exec_list) {
+ struct drm_i915_gem_object *obj = vma->obj;
+ ret = i915_gem_object_sync(obj, ring);
+ if (ret)
+ return ret;
+
+ if (obj->base.write_domain & I915_GEM_DOMAIN_CPU)
+ flush_chipset |= i915_gem_clflush_object(obj, false);
+
+ flush_domains |= obj->base.write_domain;
+ }
+
+ if (flush_chipset)
+ i915_gem_chipset_flush(ring->dev);
+
+ if (flush_domains & I915_GEM_DOMAIN_GTT)
+ wmb();
+
+ /* Unconditionally invalidate gpu caches and ensure that we do flush
+ * any residual writes from the previous batch.
+ */
+ return logical_ring_invalidate_all_caches(ringbuf);
+}
+
int intel_execlists_submission(struct drm_device *dev, struct drm_file *file,
struct intel_engine_cs *ring,
struct intel_context *ctx,
@@ -97,7 +148,82 @@ int intel_execlists_submission(struct drm_device *dev, struct drm_file *file,
struct drm_i915_gem_object *batch_obj,
u64 exec_start, u32 flags)
{
- /* TODO */
+ struct drm_i915_private *dev_priv = dev->dev_private;
+ struct intel_ringbuffer *ringbuf = ctx->engine[ring->id].ringbuf;
+ int instp_mode;
+ u32 instp_mask;
+ int ret;
+
+ if (args->num_cliprects != 0) {
+ DRM_DEBUG("clip rectangles are only valid on pre-gen5\n");
+ return -EINVAL;
+ } else {
+ if (args->DR4 == 0xffffffff) {
+ DRM_DEBUG("UXA submitting garbage DR4, fixing up\n");
+ args->DR4 = 0;
+ }
+
+ if (args->DR1 || args->DR4 || args->cliprects_ptr) {
+ DRM_DEBUG("0 cliprects but dirt in cliprects fields\n");
+ return -EINVAL;
+ }
+ }
+
+ ret = execlists_move_to_gpu(ringbuf, vmas);
+ if (ret)
+ return ret;
+
+ instp_mode = args->flags & I915_EXEC_CONSTANTS_MASK;
+ instp_mask = I915_EXEC_CONSTANTS_MASK;
+ /* The HW changed the meaning on this bit on gen6 */
+ instp_mask &= ~I915_EXEC_CONSTANTS_REL_SURFACE;
+ switch (instp_mode) {
+ case I915_EXEC_CONSTANTS_REL_GENERAL:
+ break;
+
+ case I915_EXEC_CONSTANTS_ABSOLUTE:
+ if (ring != &dev_priv->ring[RCS]) {
+ DRM_DEBUG("non-0 rel constants mode on non-RCS\n");
+ return -EINVAL;
+ }
+ break;
+
+ case I915_EXEC_CONSTANTS_REL_SURFACE:
+ DRM_DEBUG("rel surface constants mode invalid on gen5+\n");
+ return -EINVAL;
+
+ default:
+ DRM_DEBUG("execbuf with unknown constants: %d\n", instp_mode);
+ return -EINVAL;
+ }
+
+ if (ring == &dev_priv->ring[RCS] &&
+ instp_mode != dev_priv->relative_constants_mode) {
+ ret = intel_logical_ring_begin(ringbuf, 4);
+ if (ret)
+ return ret;
+
+ intel_logical_ring_emit(ringbuf, MI_NOOP);
+ intel_logical_ring_emit(ringbuf, MI_LOAD_REGISTER_IMM(1));
+ intel_logical_ring_emit(ringbuf, INSTPM);
+ intel_logical_ring_emit(ringbuf, instp_mask << 16 | instp_mode);
+ intel_logical_ring_advance(ringbuf);
+
+ dev_priv->relative_constants_mode = instp_mode;
+ }
+
+ if (args->flags & I915_EXEC_GEN7_SOL_RESET) {
+ DRM_DEBUG("sol reset is gen7 only\n");
+ return -EINVAL;
+ }
+
+ ret = ring->emit_bb_start(ringbuf, exec_start, flags);
+ if (ret)
+ return ret;
+
+ i915_gem_execbuffer_move_to_active(vmas, ring);
+ i915_gem_execbuffer_retire_commands(dev, file, ring, batch_obj);
+
return 0;
}
--
1.9.0
More information about the Intel-gfx
mailing list