Userspace is able to hint to the kernel that its command stream and auxiliary state buffers already hold the correct presumed addresses and so the relocation process may be skipped if the kernel does not need to move any buffers in preparation for the execbuffer. Thus for the common case where the allotment of buffers is static between batches, we can avoid the overhead of individually checking the relocation entries. Note that this requires userspace to supply the domain tracking and requests for workarounds itself that would otherwise be computed based upon the relocation entries. Using copywinwin10 as an example that is dependent upon emitting a lot of relocations (2 per operation), we see improvements of: c2d/gm45: 618000.0/sec to 632000.0/sec. i3-330m: 748000.0/sec to 830000.0/sec. (measured relative to a baseline with neither optimisations applied).
Showing the changes to include/uapi (1 of 3 files). Show all
Unified · Split
| @@ -308,6 +308,7 @@ typedef struct drm_i915_irq_wait { · 307 unmodified lines | |||
| 308 | #define I915_PARAM_RSVD_FOR_FUTURE_USE 22 | 308 | #define I915_PARAM_RSVD_FOR_FUTURE_USE 22 |
| 309 | #define I915_PARAM_HAS_SECURE_BATCHES 23 | 309 | #define I915_PARAM_HAS_SECURE_BATCHES 23 |
| 310 | #define I915_PARAM_HAS_PINNED_BATCHES 24 | 310 | #define I915_PARAM_HAS_PINNED_BATCHES 24 |
| 311 | #define I915_PARAM_HAS_EXEC_NO_RELOC 25 | ||
| 311 | 312 | ||
| 312 | typedef struct drm_i915_getparam { | 313 | typedef struct drm_i915_getparam { |
| 313 | int param; | 314 | int param; |
| @@ -628,7 +629,11 @@ struct drm_i915_gem_exec_object2 { · 314 unmodified lines | |||
| 628 | __u64 offset; | 629 | __u64 offset; |
| 629 | 630 | ||
| 630 | #define EXEC_OBJECT_NEEDS_FENCE (1<<0) | 631 | #define EXEC_OBJECT_NEEDS_FENCE (1<<0) |
| 632 | #define EXEC_OBJECT_NEEDS_GTT (1<<1) | ||
| 633 | #define EXEC_OBJECT_WRITE (1<<2) | ||
| 634 | #define __EXEC_OBJECT_UNKNOWN_FLAGS -(EXEC_OBJECT_WRITE<<1) | ||
| 631 | __u64 flags; | 635 | __u64 flags; |
| 636 | |||
| 632 | __u64 rsvd1; | 637 | __u64 rsvd1; |
| 633 | __u64 rsvd2; | 638 | __u64 rsvd2; |
| 634 | }; | 639 | }; |
| @@ -687,6 +692,15 @@ struct drm_i915_gem_execbuffer2 { · 52 unmodified lines | |||
| 687 | */ | 692 | */ |
| 688 | #define I915_EXEC_IS_PINNED (1<<10) | 693 | #define I915_EXEC_IS_PINNED (1<<10) |
| 689 | 694 | ||
| 695 | /** Provide a hint to the kernel that the command stream and auxilliary | ||
| 696 | * state buffers already holds the correct presumed addresses and so the | ||
| 697 | * relocation process may be skipped if no buffers need to be moved in | ||
| 698 | * preparation for the execbuffer. | ||
| 699 | */ | ||
| 700 | #define I915_EXEC_NO_RELOC (1<<11) | ||
| 701 | |||
| 702 | #define __I915_EXEC_UNKNOWN_FLAGS -(I915_EXEC_NO_RELOC<<1) | ||
| 703 | |||
| 690 | #define I915_EXEC_CONTEXT_ID_MASK (0xffffffff) | 704 | #define I915_EXEC_CONTEXT_ID_MASK (0xffffffff) |
| 691 | #define i915_execbuffer2_set_context_id(eb2, context) \ | 705 | #define i915_execbuffer2_set_context_id(eb2, context) \ |
| 692 | (eb2).rsvd1 = context & I915_EXEC_CONTEXT_ID_MASK | 706 | (eb2).rsvd1 = context & I915_EXEC_CONTEXT_ID_MASK |