Blender Git Loki

Cleanup: Remove BKE_MESH_BATCH_DIRTY_MAYBE_ALL

Cleanup: draw_common: Use switch instead of ifs

Cleanup: Mesh Batch Cache: Remove unused BKE_MESH_BATCH_DIRTY_SCULPT_COORDS

Cleanup: GPUShader: Use switch instead of ifs

DRW: Copy matrix in case of ModelMatrixInverse being used

ob->imat is garanteed to be valid by the depsgraph.

Wireframe: Refactor vertex shader to use less matrices

This means less matrices needs to be prepared by the CPU.

DRW: Speedup: Only do batch cache request and validation once for duplis

Mesh Batch Cache: Speedup: Only alloc gpu_attrs if needed and don't clear

No need to clear since GPU_material_vertex_attrs is overriding

Mesh Batch Cache: Speedup: Use flags to early remove unecessary checks

DRW: Make batch validation run first during iteration

This reduces the cost of querying the batches to the batch cache.

DRW: Move all batch request functions to own header

This is in order to using BLI_INLINE for thoses functions and keep headers
cleany separated.

DRW: Change function orders to improve CPU cache efficiency

These changes seems to impact cache efficiency as it uses shgroup just
after creating a DRWCall before shgroup.

Object Overlay: Add dupli fast path

Wireframe: Create dupli fast path using DRW_duplidata_get

DRW: Add DRW_duplidata_get to create fast path for duplis

This is in order to create less shading group when using duplis.

Data for dupli objects keep in draw manager state until the source object
changes so retrieval is fast.

Note that this system could be extended to all meshes.

GPUViewport: Fix possible hash colision with enabled engines

Also fix engine data validation that was not previously not working.

DRW: Remove uneeded calls to drw_viewport_engine_data_ensure when iterating

This remove a bit of overhead specially in scene with lots of objects.

DRW: Replace BLI_mempool by BLI_memblock

This remove a avoid the big overhead present in BLI_mempool when it is
cleared.

DRW: Speedup: Don't call GPU_shader_uniform_vector if not needed

This seems to remove a bit of overhead of going into the function even if
the uniform is not needed.

BLI_memblock: New memory allocator

This is really close to BLI_mempool but uses an array to keep track of the
chunks of memory. There is no tagging necessary to clear the whole
structure so reuse is fast.

Naturally supports iteration but does not support freeing.