Haswell GT3e has the unique feature of supporting Write-Through cacheing
of objects within the eLLC/LLC. The purpose of this is to enable the display
plane to remain coherent whilst objects lie resident in the eLLC/LLC - so
that we, in theory, get the best of both worlds, perfect display and fast
access.
v2: Actually do the clflush on transition to WT, nagging by Ville.
Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
Cc: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx>
Cc: Kenneth Graunke <kenneth@xxxxxxxxxxxxx>
---
drivers/gpu/drm/i915/i915_dma.c | 3 +++
drivers/gpu/drm/i915/i915_drv.h | 4 +++-
drivers/gpu/drm/i915/i915_gem.c | 29 +++++++++++++++--------------
drivers/gpu/drm/i915/i915_gem_gtt.c | 11 ++++++++++-
include/uapi/drm/i915_drm.h | 1 +
5 files changed, 32 insertions(+), 16 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
index 8da0b3d..75989fc 100644
--- a/drivers/gpu/drm/i915/i915_dma.c
+++ b/drivers/gpu/drm/i915/i915_dma.c
@@ -976,6 +976,9 @@ static int i915_getparam(struct drm_device *dev, void *data,
case I915_PARAM_HAS_LLC:
value = HAS_LLC(dev);
break;
+ case I915_PARAM_HAS_WT:
+ value = HAS_WT(dev);
+ break;
case I915_PARAM_HAS_ALIASING_PPGTT:
value = dev_priv->mm.aliasing_ppgtt ? 1 : 0;
break;
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 34d2b9d..d27a82a 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -452,6 +452,7 @@ enum i915_cache_level {
I915_CACHE_NONE = 0,
I915_CACHE_LLC,
I915_CACHE_LLC_MLC, /* gen6+, in docs at least! */
+ I915_CACHE_WT,
};
typedef uint32_t gen6_gtt_pte_t;
@@ -1344,7 +1345,7 @@ struct drm_i915_gem_object {
unsigned int pending_fenced_gpu_access:1;
unsigned int fenced_gpu_access:1;
- unsigned int cache_level:2;
+ unsigned int cache_level:3;
unsigned int has_aliasing_ppgtt_mapping:1;
unsigned int has_global_gtt_mapping:1;
@@ -1547,6 +1548,7 @@ struct drm_i915_file_private {
#define HAS_BLT(dev) (INTEL_INFO(dev)->has_blt_ring)
#define HAS_VEBOX(dev) (INTEL_INFO(dev)->has_vebox_ring)
#define HAS_LLC(dev) (INTEL_INFO(dev)->has_llc)
+#define HAS_WT(dev) (IS_HASWELL(dev) && ((struct drm_i915_private *)(dev)->dev_private)->ellc_size)
#define I915_NEED_GFX_HWS(dev) (INTEL_INFO(dev)->need_gfx_hws)
#define HAS_HW_CONTEXTS(dev) (INTEL_INFO(dev)->gen >= 5)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 99362f7..6b33494 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -3447,27 +3447,27 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
i915_gem_obj_ggtt_set_color(obj, cache_level);
}
- if (cache_level == I915_CACHE_NONE) {
- u32 old_read_domains, old_write_domain;
-
+ if (cache_level == I915_CACHE_NONE ||
+ cache_level == I915_CACHE_WT) {
/* If we're coming from LLC cached, then we haven't
* actually been tracking whether the data is in the
* CPU cache or not, since we only allow one bit set
* in obj->write_domain and have been skipping the clflushes.
- * Just set it to the CPU cache for now.
+ * Do them now.
*/
- WARN_ON(obj->base.write_domain & ~I915_GEM_DOMAIN_CPU);
- WARN_ON(obj->base.read_domains & ~I915_GEM_DOMAIN_CPU);
+ if (obj->pages && !obj->stolen) {
+ u32 old_write_domain;
- old_read_domains = obj->base.read_domains;
- old_write_domain = obj->base.write_domain;
+ trace_i915_gem_object_clflush(obj);
+ drm_clflush_sg(obj->pages);
- obj->base.read_domains = I915_GEM_DOMAIN_CPU;
- obj->base.write_domain = I915_GEM_DOMAIN_CPU;
+ old_write_domain = obj->base.write_domain;
+ obj->base.write_domain &= ~I915_GEM_DOMAIN_CPU;
- trace_i915_gem_object_change_domain(obj,
- old_read_domains,
- old_write_domain);
+ trace_i915_gem_object_change_domain(obj,
+ obj->base.read_domains,
+ old_write_domain);
+ }
}
obj->cache_level = cache_level;
@@ -3565,7 +3565,8 @@ i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj,
* of uncaching, which would allow us to flush all the LLC-cached data
* with that bit in the PTE to main memory with just one PIPE_CONTROL.
*/
- ret = i915_gem_object_set_cache_level(obj, I915_CACHE_NONE);
+ ret = i915_gem_object_set_cache_level(obj,
+ HAS_WT(obj->base.dev) ? I915_CACHE_WT : I915_CACHE_NONE);
if (ret)
return ret;
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 0522d00..072a348 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -54,6 +54,7 @@
(((bits) & 0x8) << (11 - 3)))
#define HSW_WB_LLC_AGE0 HSW_CACHEABILITY_CONTROL(0x3)
#define HSW_WB_ELLC_LLC_AGE0 HSW_CACHEABILITY_CONTROL(0xb)
+#define HSW_WT_ELLC_LLC_AGE0 HSW_CACHEABILITY_CONTROL(0x6)
static gen6_gtt_pte_t gen6_pte_encode(dma_addr_t addr,
enum i915_cache_level level)
@@ -116,8 +117,16 @@ static gen6_gtt_pte_t iris_pte_encode(dma_addr_t addr,
gen6_gtt_pte_t pte = GEN6_PTE_VALID;
pte |= HSW_PTE_ADDR_ENCODE(addr);
- if (level != I915_CACHE_NONE)
+ switch (level) {
+ case I915_CACHE_NONE:
+ break;
+ case I915_CACHE_WT:
+ pte |= HSW_WT_ELLC_LLC_AGE0;
+ break;
+ default:
pte |= HSW_WB_ELLC_LLC_AGE0;
+ break;
+ }
return pte;
}
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index e47cf00..e831292 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -338,6 +338,7 @@ typedef struct drm_i915_irq_wait {
#define I915_PARAM_HAS_PINNED_BATCHES 24
#define I915_PARAM_HAS_EXEC_NO_RELOC 25
#define I915_PARAM_HAS_EXEC_HANDLE_LUT 26
+#define I915_PARAM_HAS_WT 27
typedef struct drm_i915_getparam {
int param;
--
1.8.3.2
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
