drm/i915: Avoid the branch in computing intel_ring_space()
Exploit the power-of-two ring size to compute the space across the
wraparound using a mask rather than a if. Convert to unsigned integers
so the operation is well defined.
References: https://bugs.freedesktop.org/show_bug.cgi?id=99671
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170504130846.4807-1-chris@chris-wilson.co.uk
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
index 600713b..650ab88 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.h
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -17,17 +17,6 @@
#define CACHELINE_BYTES 64
#define CACHELINE_DWORDS (CACHELINE_BYTES / sizeof(uint32_t))
-/*
- * Gen2 BSpec "1. Programming Environment" / 1.4.4.6 "Ring Buffer Use"
- * Gen3 BSpec "vol1c Memory Interface Functions" / 2.3.4.5 "Ring Buffer Use"
- * Gen4+ BSpec "vol1c Memory Interface and Command Stream" / 5.3.4.5 "Ring Buffer Use"
- *
- * "If the Ring Buffer Head Pointer and the Tail Pointer are on the same
- * cacheline, the Head Pointer must not be greater than the Tail
- * Pointer."
- */
-#define I915_RING_FREE_SPACE 64
-
struct intel_hw_status_page {
struct i915_vma *vma;
u32 *page_addr;
@@ -145,9 +134,9 @@ struct intel_ring {
u32 tail;
u32 emit;
- int space;
- int size;
- int effective_size;
+ u32 space;
+ u32 size;
+ u32 effective_size;
};
struct i915_gem_context;
@@ -548,6 +537,25 @@ assert_ring_tail_valid(const struct intel_ring *ring, unsigned int tail)
*/
GEM_BUG_ON(!IS_ALIGNED(tail, 8));
GEM_BUG_ON(tail >= ring->size);
+
+ /*
+ * "Ring Buffer Use"
+ * Gen2 BSpec "1. Programming Environment" / 1.4.4.6
+ * Gen3 BSpec "1c Memory Interface Functions" / 2.3.4.5
+ * Gen4+ BSpec "1c Memory Interface and Command Stream" / 5.3.4.5
+ * "If the Ring Buffer Head Pointer and the Tail Pointer are on the
+ * same cacheline, the Head Pointer must not be greater than the Tail
+ * Pointer."
+ *
+ * We use ring->head as the last known location of the actual RING_HEAD,
+ * it may have advanced but in the worst case it is equally the same
+ * as ring->head and so we should never program RING_TAIL to advance
+ * into the same cacheline as ring->head.
+ */
+#define cacheline(a) round_down(a, CACHELINE_BYTES)
+ GEM_BUG_ON(cacheline(tail) == cacheline(ring->head) &&
+ tail < ring->head);
+#undef cacheline
}
static inline unsigned int