The previous change gave us only two modes, one which looped over the batch
per cliprect (3d drawing) and one that didn't (state updeast).
However, we really want 4:
- Batch doesn't care about cliprects (state updates)
- Batch needs DRAWING_RECTANGLE looping per cliprect (3d drawing)
- Batch needs to be executed just once (region fills, copies, etc.)
- Batch already includes cliprect handling, and must be flushed by unlock time
(copybuffers, clears).
All callers should now be fixed to use one of these states for any batchbuffer
emits. Thanks to Keith Whitwell for pointing out the failure.
This may allow better concurrency (noop in openarena performance now), but is
also important for the previous commit -- otherwise, we may end up with
BufferData, draw_prims, BufferData and the draw_prims would use the new VBO
data instead of old. This could still occur with user-supplied VBOs and poor
use of MapBuffer without BufferData.
The comment about (vbo)_exec_api.c appeared to be stale, as the VBO code seems
to only use non-named VBOs (not actual VBOs) or freshly-allocated VBO data.
This brings a 2x speedup to openarena, because we can submit nearly-full
batchbuffers instead of many 450-byte ones.
This allows us to avoid re-emitting some state when validate_state happens
multiple times per batchbuffer. Even though we flush batch per primitive
currently, that may still happen already if the primitive changed (this should
probably be fixed as well).
Both drivers have ended up relying on lost_hardware being called after each
batch buffer, so update the name. This removes one of the calls on 965 whic
h was outside of the batchbuffer handling code and just duplicating what had
already happened through batchbuffer handling.
In particular, batch buffers are no longer flushed when switching from
CLIPRECTS to NO_CLIPRECTS or vice versa, and 965 just uses DRM cliprect
handling for primitives instead of trying to sneak in its own to avoid the
DRM stuff. The disadvantage is that we will re-execute state updates per
cliprect, but the advantage is that we will be able to accumulate larger
batch buffers, which were proving to be a major overhead.
Each array element is now a BUFFER_x token rather than a BUFFER_BIT_x bitmask.
The number of active color buffers is specified by _NumColorDrawBuffers.
This builds on the previous DrawBuffer changes and will help with drivers
implementing GL_ARB_draw_buffers.
These fields are no longer indexed by shader output. Now, we just have
a simple array of renderbuffer pointers.
If the shader writes to gl_FragData[i], send those colors to the N
_ColorDrawBuffers. Otherwise, replicate the single gl_FragColor (or
the fixed-function color) to the N _ColorDrawBuffers.
A few more changes and simplifications can follow from this...
By avoiding the repeated relocation buffer creation/map/unmap/destroy for each
new batch buffer, this improves OpenArena framerates by 30%. Caching batch
buffers themselves doesn't appear to be a significant performance win over
this change.
We have two consumers of relocations. One is static state buffers, which
want the same relocation every time. The other is the batchbuffer, which gets
thrown out immediately after submit. This lets us reduce repeated computation
for static state buffers, and clean up the code by moving relocations nearer
to where the state buffer is computed.
This reverts commit 8bb9ae3693.
Validating our kernel buffers with the caching off in flags but on in mask
means that the kernel migrates the buffer to be uncached, which is undesired.
The 'mask' value used in the validation operation specifies which of the
'flags' bits are being modified. Buffer validation wants to pass the memory
type and access mode (rwx) to the kernel so that the buffer will be placed
correctly, and so that the right kind of fence will be created (read vs
write). That means we actually want a constant mask for these operations,
and not something computed from the bits coming in. The constant we want is
DRM_BO_MASK_MEM | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE | DRM_BO_FLAG_EXE.
