Commit graph

7731 commits

Author SHA1 Message Date
Keith Whitwell
0c047b8847 Texture copypixels, fixes to blit path also.
Need to add logic to avoid using texture engine where source, dest overlap.
2006-02-27 16:39:41 +00:00
Thomas Hellström
c64a08c286 Call kernel to update the sarea breadcrumb when we seem to be out of
batchbuffer memory.

Remove flushes from batchbuffer tails. They are not needed anymore
when we have the kernel sync flush mechanism.

Better kernel error checks.
2006-02-27 16:28:10 +00:00
Keith Whitwell
bb409e88d8 Fix drawpixels window coordinate calculation.
Make sure cliprects are emitted for metaops texture quads.
Implement PixelZoom for drawpixels.
2006-02-27 13:51:10 +00:00
Keith Whitwell
af69eefcba Fix logic for calculating regions for partial-buffer clears. 2006-02-27 13:18:13 +00:00
Keith Whitwell
d9c071adec - Use a raw buffer id rather than an intel_region struct for texture
state/relocations.
- Implement basic texture engine drawpixels.
2006-02-26 16:47:15 +00:00
Keith Whitwell
833d19c21d Make this struct less opaque. 2006-02-26 14:39:23 +00:00
Keith Whitwell
7245757174 Use spantmp2.h, based on r200 driver. 2006-02-26 14:37:54 +00:00
Keith Whitwell
3d96cb7d48 First pass at using the texture engine for copypixels. Currently only
does the same operations as are accelerated by the blitter, though this
will change...
2006-02-26 14:37:30 +00:00
Keith Whitwell
f565ab2dd5 For some reason, streaming copies to write-combined regions are
extremely sensitive to the characteristics of how the source data is
retrieved.  By reordering the source reads in the rgba to bgra
conversion, the speed of this operation increases by half.
2006-02-26 12:46:51 +00:00
Thomas Hellström
7ad0c51f48 Enable uncached TTMs by default.
Allocate batchbuffers from a 1MB local TTM pool =>
multiple clients.
2006-02-26 04:29:28 +00:00
Keith Whitwell
fea684c533 fix misleading MI_FLUSH defn and usage 2006-02-25 18:57:43 +00:00
Keith Whitwell
4e8845fc0c a few missed cases where blits can't do read/drawpixels 2006-02-25 15:33:48 +00:00
Keith Whitwell
ffcc1cdb41 Tweak the read format/type slightly so this can use the blitter. 2006-02-24 17:51:04 +00:00
Keith Whitwell
afc90075ba get PBO readpixels working also 2006-02-24 17:23:57 +00:00
Keith Whitwell
79acea2708 Get PBO drawpixels working. 2006-02-24 16:52:47 +00:00
Keith Whitwell
a12063a540 use expected argument order for intel_check_blit_format 2006-02-24 16:52:06 +00:00
Keith Whitwell
9826d83439 Add missing GetSubData callback.
Fill in various parts of the base gl_buffer_object struct as required from
each of our callbacks.
2006-02-24 16:51:33 +00:00
Keith Whitwell
84f7ded925 add bmBufferGetSubData 2006-02-24 16:50:22 +00:00
Keith Whitwell
59dd8ad752 Hack bmBufferData and bmBufferSubdata back into operation with an
implicit map/unmap.
2006-02-24 15:02:41 +00:00
Keith Whitwell
4afd26982e Hook up buffer_objects implementation to DrawPixels for testing.
Not yet working.
2006-02-24 13:58:45 +00:00
Keith Whitwell
17e4a7ea17 Bring DBG_BM() macros back to DBG(). Turn on by setting
INTEL_DEBUG=buf in the environment.
2006-02-24 13:56:28 +00:00
Keith Whitwell
3d67655cce bufferobjects implementation layered on top of bm buffers 2006-02-24 11:56:45 +00:00
Keith Whitwell
b5d7c69884 Define union fi, reenable clear-with-tris. 2006-02-24 11:05:35 +00:00
Keith Whitwell
b6759e2be7 Add an assert 2006-02-24 10:10:50 +00:00
Keith Whitwell
af853b0da7 Flush old primitive before starting new one in intelRasterPrimitive. 2006-02-24 10:10:28 +00:00
Keith Whitwell
8d349798da Add a callback to set post-fragment state in the metaops utility.
Useful for accelerating read/draw/copy pixels when blitting isn't
sufficient.
2006-02-23 14:24:39 +00:00
Keith Whitwell
4aea628194 Fix clipping for copypixel 2006-02-23 14:23:50 +00:00
Keith Whitwell
008ddf628a Remove dead code 2006-02-23 14:23:06 +00:00
Thomas Hellström
716496f263 Hack bufmgr_fake to allocate from ttms when possible. 2006-02-23 13:30:11 +00:00
Keith Whitwell
b3b325e560 get 3d textures working again 2006-02-23 11:47:36 +00:00
Keith Whitwell
382ce985ac Fix cubemap layout 2006-02-22 19:54:32 +00:00
Keith Whitwell
54b7eca7a0 Append an MI_FLUSH to each batchbuffer for synchronization. Not really
ideal from a performance viewpoint and there are mechanisms to avoid this
in future.
2006-02-22 19:26:37 +00:00
Keith Whitwell
f74e06bf11 restore swapbuffers throttling 2006-02-22 17:18:19 +00:00
Keith Whitwell
9c02649d18 Introduce fixup/relocation lists for dma buffers. 2006-02-22 15:16:01 +00:00
Keith Whitwell
2c34704e41 tweaks 2006-02-13 00:41:33 +00:00
Keith Whitwell
85d0041ff0 rename texdown to reflect behaviour changes 2006-02-13 00:40:33 +00:00
Keith Whitwell
67ff8b354e bump driver date, etc 2006-02-13 00:39:20 +00:00
Keith Whitwell
08e7957571 Fix inverted image in copytexsubimage by tickling the hardware to
do the inversion in the blitter.
2006-02-13 00:38:46 +00:00
Keith Whitwell
9c1f7089d8 Some functions had been indented by one space?! 2006-02-11 04:28:13 +00:00
Keith Whitwell
5381ac5f11 move blit functions to new file 2006-02-11 02:54:09 +00:00
Keith Whitwell
a6f78d4eee Use aligned malloc for backing store, if necessary. 2006-02-10 00:01:45 +00:00
Keith Whitwell
ae695e3566 Remove 2nd copy of do_memcpy 2006-02-09 23:09:47 +00:00
Keith Whitwell
c82c3335e1 fix refcounting and other issues 2006-02-09 23:06:23 +00:00
Keith Whitwell
f8246f504b Add basic timing for memcpys 2006-02-09 23:04:44 +00:00
Keith Whitwell
499458bcdb Add INVARIENT to the active state list 2006-02-09 23:01:34 +00:00
Keith Whitwell
37a53fd6a0 Make various changes to get a number out of texdown that better
represents realworld usage.  In particular, avoid situations where
the client texture image and the mesa copy (if there is one) are both
resident in L2 cache.
2006-02-09 22:52:20 +00:00
Keith Whitwell
5ec38fb2ea Turn DBG output on/off with INTEL_DEBUG=bufs 2006-02-09 22:21:20 +00:00
Keith Whitwell
11c0215bf8 Switch between memcpy implementations according to src/dest alignment. 2006-02-01 18:42:16 +00:00
Keith Whitwell
6a13b6c346 Debug off 2006-02-01 18:30:16 +00:00
Keith Whitwell
a103097ee5 Quieten depend 2006-02-01 18:29:56 +00:00