| Age | Commit message (Collapse) | Author |
|---|
|
|
|
|
|
Signed-off-by: Francisco Jerez <currojerez@riseup.net>
|
|
Previously they depended on format blocks, but after removing those
they started depending on format encoding.
|
|
- Added flush notify functions for NV30 and NV40.
- Flushing mid frame will call flush notify, which will resubmit all
relocs.
- We don't try to recover from reloc failure yet.
|
|
|
|
|
|
|
|
width/height/depth arrays
|
|
Mip-mapped 3D textures are not arrays of 2D layers
with a mip-map layout like 2D textures, therefore we
cannot use image_nr == depth for them.
Making use of "volume tiling" modes now, the allowed
modes are 0xZY where Z <= 5 and y <= 5.
|
|
First, using width * block size as pitch is evidently
wrong if a block contains more than 1 texel.
For tiled textures, since a block occupies a contiguous
area of memory, y addressing in m2mf has to be done by
block index, not the y coordinate itself.
This should fix compressed textures.
|
|
Adds a more generic SIFC transfer function.
|
|
Always test for PIPE_TRANSFER_READ/WRITE using the bit-wise and operator, and
add a pipe_transfer_buffer_flags() helper for getting the buffer usage flags
corresponding to them.
|
|
|
|
The winsys once again has to know about textures it seems, so we need a
common representation between all our pipe drivers to store some
information the winsys will need.
Only the nv50 driver has been fixed so far.
|
|
|
|
The hardware expects a texture's tile mode to change with
the mipmap level.
Also, only multiply by block size once to obtain size.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
The core reference counting code is centralized in p_refcnt.h.
This has some consequences related to struct pipe_buffer:
* The screen member of struct pipe_buffer must be initialized, or
pipe_buffer_reference() will crash trying to destroy a buffer with reference
count 0. u_simple_screen takes care of this, but I may have missed some of
the drivers not using it.
* Except for rare exceptions deep in winsys code, buffers must always be
allocated via pipe_buffer_create() or via screen->*buffer_create() rather
than via winsys->*buffer_create().
|
|
It's possible a state tracker will map/unmap a transfer object many times.
Delaying upload until the object is destroyed will prevent unnecessary
uploads to the GPU.
Also fixing a typo here, was unmapping the wrong buffer on transfer_unmap!
|
|
|
