Christian König
8e0c05960d
[g3dvl] cleanup empty block handling
...
doing empty block handling in the mc code is indeed faster
2010-12-05 16:46:10 +01:00
Christian König
9af3c243d9
move empty block handling back into mc for testing
2010-12-03 19:04:01 +01:00
Christian König
29840040af
add rasterizer state
2010-12-03 19:04:01 +01:00
Christian König
27016941bc
use CMP also for referenz frame fetch
2010-12-03 19:04:01 +01:00
Christian König
c8b7cf469f
cleanup and use CMP instead of IF ELSE ENDIF
2010-12-03 19:04:01 +01:00
Christian König
74c71f09f3
move to four component calculation for idct code
2010-12-03 19:04:01 +01:00
Christian König
69f53c3dc8
copy only mv really needed to vb
2010-12-03 19:04:01 +01:00
Christian König
d2888c5f2f
use vertex buffer also for mc code
2010-12-03 19:04:01 +01:00
Christian König
c7068d79a0
give each vertex element its own buffer
2010-12-03 19:04:00 +01:00
Christian König
838d109207
use vl_vb_upload_quads also for mc
2010-12-03 19:04:00 +01:00
Christian König
3e6a5077ca
move vertex elemt states into mb type handlers
2010-12-03 19:04:00 +01:00
Christian König
3bbbb3c54f
move macroblock type handling into its own structure
2010-12-03 19:04:00 +01:00
Christian König
4abe738288
use a shadow buffer for vertex data to optimize memory access
2010-12-03 19:04:00 +01:00
Christian König
a984c67b31
make nr of render targets configureable for testing
2010-12-03 19:04:00 +01:00
Christian König
336c7735ae
[g3dvl] join empty blocks to get larger slices
2010-11-28 01:21:41 +01:00
Christian König
e742a1043d
[g3dvl] use 8 zslices for idct
2010-11-27 18:20:38 +01:00
Christian König
3fd53e6c2a
[g3dvl] some more bugfixing
2010-11-27 14:08:15 +01:00
Christian König
12836fbcfa
[g3dvl] idividual vs for each stage and a bunch of bugsfixes
2010-11-27 14:01:01 +01:00
Christian König
027704db75
[g3dvl] give idct stage 1 & 2 its own sb_state and viewport
2010-11-27 11:24:24 +01:00
Christian König
13e28cff76
[g3dvl] use four component fetch also for idct source
2010-11-26 21:50:14 +01:00
Christian König
a981d62c97
r600g: disable hardware blit for stream texture
2010-11-26 21:44:22 +01:00
Christian König
cfe489b897
[g3dvl] split matrix mul into seperate functions
2010-11-26 20:25:00 +01:00
Christian König
7408a6ab89
[g3dvl] use inline constants instead of vs_const for idct
2010-11-26 19:14:55 +01:00
Christian König
3dd7bf7d39
[g3dvl] no need to keep the idct matrix multiple times
2010-11-25 22:10:21 +01:00
Christian König
9cff905343
[g3dvl] add dump option to xvmc
2010-11-25 21:23:48 +01:00
Christian König
ed8b767a8e
[g3dvl] also use four elemets on right side multiplikation
2010-11-25 19:37:12 +01:00
Christian König
c9e10c666a
[g3dvl] use four elements in matrix texture fetch
2010-11-24 23:54:21 +01:00
Christian König
de623b96ab
r600g: disable not working formats
2010-11-24 23:44:30 +01:00
Christian König
431e72984b
r600g: reenable texture uploads, but keep R16_SNORM disabled
2010-11-24 21:40:50 +01:00
Christian König
a51b0daa59
r600g: disable R32 float also in r600_translate_colorformat
2010-11-24 21:36:54 +01:00
Christian König
2c9db2484b
[g3dvl] no need for all samplers at all stages
2010-11-24 20:00:52 +01:00
Christian König
5391ef8606
[g3dvl] remove flushing between stages
2010-11-24 19:46:51 +01:00
Christian König
ed49905944
[g3dvl] spread scaling between idct stages
2010-11-24 19:40:47 +01:00
Christian König
5a8078486a
[g3dvl] remove invalid use of assert
2010-11-23 22:43:29 +01:00
Christian König
58d04f816c
[g3dvl] switch to r32 float for idct matrix
2010-11-23 21:26:26 +01:00
Christian König
21efda8687
[g3dvl] add some error handling
2010-11-23 00:19:02 +01:00
Christian König
e6b71530da
Merge remote branch 'origin/master' into pipe-video
...
Conflicts:
src/gallium/auxiliary/Makefile
src/gallium/auxiliary/SConscript
2010-11-21 19:40:52 +01:00
Brian Paul
5e3733fadf
mesa: fix get_texture_dimensions() for texture array targets
...
Fixes http://bugs.freedesktop.org/show_bug.cgi?id=31779
2010-11-21 10:05:51 -07:00
Brian Paul
0ec0f1025d
docs: update some GL 3.0 status
2010-11-21 09:40:28 -07:00
Brian Paul
5ed51e950f
mesa: hook up GL 3.x entrypoints
...
Fix up some details in the xml files and regenerate dispatch files.
2010-11-21 09:20:44 -07:00
Brian Paul
81c347ef79
glapi: rename GL3.xml to GL3x.xml as it covers all GL 3.x versions
2010-11-21 09:20:43 -07:00
Brian Paul
197b1d7898
mesa: fix error msg typo
2010-11-21 09:20:43 -07:00
Daniel Vetter
c8fca58d9d
i915g: kill idws->pool
...
The drm winsys only ever handles one gem memory manager. Rip out
the unnecessary complication.
Reviewed-by: Jakob Bornecrantz <wallbraker@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
2010-11-21 16:41:19 +01:00
Daniel Vetter
e182618853
i915g: kill buf->map_gtt
...
Not using the gtt is considered harmful for performance. And for
partial uploads there's always drm_intel_bo_subdata.
Reviewed-by: Jakob Bornecrantz <wallbraker@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
2010-11-21 16:41:19 +01:00
Daniel Vetter
d54d67499c
i915g: kill RGBA/X formats
...
It's intel, so always little endian!
Reviewed-by: Jakob Bornecrantz <wallbraker@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
2010-11-21 16:41:19 +01:00
Daniel Vetter
8624fe7a49
i915g: add pineview pci ids
...
Reviewed-by: Jakob Bornecrantz <wallbraker@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
2010-11-21 16:41:19 +01:00
Daniel Vetter
aba728eb25
i915g: s/hw_tiled/tiling
...
More in line with other intel drivers.
Change to use enum by Jakob Bornecrantz.
Reviewed-by: Jakob Bornecrantz <wallbraker@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
2010-11-21 16:41:18 +01:00
Daniel Vetter
f77a2690b4
i915g: rip out ->sw_tiled
...
It looks like this was meant to facilitate unfenced access to textures/
color/renderbuffers. It's totally incomplete and fundamentally broken
on a few levels:
- broken: The kernel needs to about every tiled bo to fix up bit17
swizzling on swap-in.
- unflexible: fenced/unfenced relocs from execbuffer2 do the same, much
simpler.
- unneeded: with relaxed fencing tiled gem bos are as memory-efficient
as this trick.
Hence kill it.
Reviewed-by: Jakob Bornecrantz <wallbraker@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
2010-11-21 16:41:18 +01:00
Joakim Sindholt
bf10055cff
r300g: silence guard band cap errors
...
Somebody should find out what these are. It can be found on Windows
getting a D3DCAPS9 from IDirect3D9::GetCaps() and reading the
GuardBand* values.
2010-11-21 15:45:20 +01:00
Christian König
42c7291d2c
[g3dvl] inverse check for iDCT
2010-11-21 14:34:38 +01:00
