On 05/04/2017 10:30, Oscar Mateo wrote:
From: Daniele Ceraolo Spurio <daniele.ceraolospurio@xxxxxxxxx>
Technically speaking, the context size is per engine class, not per
instance.
It is very nice to have the code match the documentation!
Cc: Tvrtko Ursulin <tvrtko.ursulin@xxxxxxxxx>
Cc: Paulo Zanoni <paulo.r.zanoni@xxxxxxxxx>
Cc: Rodrigo Vivi <rodrigo.vivi@xxxxxxxxx>
Cc: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@xxxxxxxxx>
Signed-off-by: Oscar Mateo <oscar.mateo@xxxxxxxxx>
---
drivers/gpu/drm/i915/intel_lrc.c | 34 ++++++++++++++++++++++------------
drivers/gpu/drm/i915/intel_lrc.h | 7 ++++++-
2 files changed, 28 insertions(+), 13 deletions(-)
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index 0dc1cc4..6b1fc4a 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -1908,8 +1908,10 @@ static void execlists_init_reg_state(u32 *regs,
}
/**
- * intel_lr_context_size() - return the size of the context for an engine
- * @engine: which engine to find the context size for
+ * intel_lr_class_context_size() - return the size of the context for a given
+ * engine class
+ * @dev_priv: i915 device private
+ * @class: which engine class to find the context size for
*
* Each engine may require a different amount of space for a context image,
* so when allocating (or copying) an image, this function can be used to
@@ -1921,25 +1923,33 @@ static void execlists_init_reg_state(u32 *regs,
* in LRC mode, but does not include the "shared data page" used with
* GuC submission. The caller should account for this if using the GuC.
*/
-uint32_t intel_lr_context_size(struct intel_engine_cs *engine)
+uint32_t intel_lr_class_context_size(struct drm_i915_private *dev_priv,
+ enum intel_engine_class class)
{
int ret = 0;
- WARN_ON(INTEL_GEN(engine->i915) < 8);
+ WARN_ON(INTEL_GEN(dev_priv) < 8);
- switch (engine->id) {
- case RCS:
- if (INTEL_GEN(engine->i915) >= 9)
+ switch (class) {
+ case RENDER_CLASS:
+ switch (INTEL_GEN(dev_priv)) {
+ default:
+ DRM_ERROR("Unknown context size for GEN\n");
MISSING_CASE I think, otherwise it is too easy to miss in the noise. And
it is MISSING_CASE for the class below which looks appropriate to me for
this layer.
+ case 9:
ret = GEN9_LR_CONTEXT_RENDER_SIZE;
- else
+ break;
+ case 8:
ret = GEN8_LR_CONTEXT_RENDER_SIZE;
+ break;
+ }
break;
- case VCS:
- case BCS:
- case VECS:
- case VCS2:
+ case VIDEO_DECODE_CLASS:
+ case VIDEO_ENHANCEMENT_CLASS:
+ case COPY_ENGINE_CLASS:
ret = GEN8_LR_CONTEXT_OTHER_SIZE;
break;
+ default:
+ MISSING_CASE(class);
}
return ret;
diff --git a/drivers/gpu/drm/i915/intel_lrc.h b/drivers/gpu/drm/i915/intel_lrc.h
index e8015e7..b3a4331 100644
--- a/drivers/gpu/drm/i915/intel_lrc.h
+++ b/drivers/gpu/drm/i915/intel_lrc.h
@@ -78,7 +78,12 @@ enum {
struct drm_i915_private;
struct i915_gem_context;
-uint32_t intel_lr_context_size(struct intel_engine_cs *engine);
+uint32_t intel_lr_class_context_size(struct drm_i915_private *dev_priv,
+ enum intel_engine_class class);
+static inline uint32_t intel_lr_context_size(struct intel_engine_cs *engine)
+{
+ return intel_lr_class_context_size(engine->i915, engine->class);
+}
void intel_lr_context_resume(struct drm_i915_private *dev_priv);
uint64_t intel_lr_context_descriptor(struct i915_gem_context *ctx,
Otherwise looks fine to me. I also really hope to be able to use the
class/instance cleanup in scope of the better VCS load balancing which
is currently being looked at.
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
