The core problem was that _mesa_generate_mipmap was not respecting RowStride
of the source image. Additionally, the intel private data associated with the
images (level and face) was not being initialized for the
_mesa_generate_mipmap-generated images.
mipmap pitches must account for the device alignment requirements, which
used to be fairly simple; just align to a 4-byte boundary. However, to allow
textures to be drawn to under TTM, they now need to be aligned to a 64-byte
boundary. Placing all of the alignment constraints in a single function
allows this new constraint to be applied uniformly.
There was some pitch constraining code in intel_miptree_create, but that was
modifying the pitch long after the miptree had been layed out, so it only
served to wreck the mipmap and cause rendering errors.
With FBOs, we end up wanting to do 3D metaops against one or the other without
having to find the other one to fill in if we're not going to draw to it.
This removes the delayed texture upload optimization from 965, in exchange for
bringing us closer to PBO support. It also disables SGIS_generate_mipmap,
which didn't seem to be working before anyway, according to the lodbias demo.
Each buffer object now has a relocation buffer pointer, which contains the
relocations for the buffer if there are any. At the point where we have to
create a new type of relocation entry, we can change the code over to allowing
multiple relocation lists, but trying to anticipate what that'll look like
now just increases complexity.
This is a 30% performance improvement on 965.
The user-space suballocator that was used avoided relocation computations by
using the general and surface state base registers and allocating those types
of buffers out of pools built on top of single buffer objects. It also
avoided calls into the buffer manager for these small state allocations, since
only one buffer object was being used.
However, the buffer allocation cost appears to be low, and with relocation
caching, computing relocations for buffers is essentially free. Additionally,
implementing the suballocator required a don't-fence-subdata flag to disable
waiting on buffer maps so that writing new data didn't block on rendering using
old data, and careful handling when mapping to update old data (which we need
to do for unavoidable relocations with FBOs). More importantly, when the
suballocator filled, it had no replacement algorithm and just threw out all
of the contents and forced them to be recomputed, which is a significant cost.
This is the first step, which just changes the buffer type, but doesn't yet
improve the hash table to not result in full recompute on overflow. Because
the buffers are all allocated out of the general buffer allocator, we can
no longer use the general/surface state bases to avoid relocations, and they
are set to 0 instead.
Xiang, Haihao