Ben Skeggs [Sun, 30 Mar 2008 09:08:59 +0000 (19:08 +1000)]
nv40: vp 1/0/- swz
Ben Skeggs [Sun, 30 Mar 2008 09:00:07 +0000 (19:00 +1000)]
nv40: vp const/immd fix
Ben Skeggs [Tue, 25 Mar 2008 01:57:36 +0000 (12:57 +1100)]
nv40: respect do_flip in surface_copy()
Ben Skeggs [Tue, 25 Mar 2008 01:14:49 +0000 (12:14 +1100)]
Merge remote branch 'upstream/gallium-0.1' into nouveau-gallium-0.1
Ben Skeggs [Tue, 25 Mar 2008 01:12:26 +0000 (12:12 +1100)]
nouveau: refcount buffers on validate list.
Brian [Tue, 25 Mar 2008 00:49:56 +0000 (18:49 -0600)]
gallium: fix a few bugs, warnings in the p_debug code
added missing _ to a _debug_printf() call.
Brian [Tue, 25 Mar 2008 00:32:48 +0000 (18:32 -0600)]
gallium: move sampler_unit field to pstip_fragment_shader since it's per-shader
Also, fix another texture refcounting bug.
Brian [Tue, 25 Mar 2008 00:31:01 +0000 (18:31 -0600)]
gallium: save/restore samplers in draw_textured_quad()
Brian Paul [Mon, 24 Mar 2008 23:20:52 +0000 (17:20 -0600)]
mesa: call glutDestroyWindow (help find mem leaks)
Brian Paul [Mon, 24 Mar 2008 23:20:22 +0000 (17:20 -0600)]
gallium: fix a few texture border/mipmap glitches found w/ conform
Brian Paul [Mon, 24 Mar 2008 18:59:54 +0000 (12:59 -0600)]
gallium: check SamplersUsed to determine if texture needed, as in st_atom_sampler.c
Otherwise, we were sometimes setting texture state but not the corresponding
sampler state.
Brian [Mon, 24 Mar 2008 22:31:15 +0000 (16:31 -0600)]
gallium: make a copy of the vertex shader's token array.
This solves problems when the state tracker frees the token array when the
draw module still needs it.
Brian [Mon, 24 Mar 2008 22:26:45 +0000 (16:26 -0600)]
gallium: use pipe_texture_reference() in a few places (fixes refcounting bugs)
Brian [Mon, 24 Mar 2008 22:24:32 +0000 (16:24 -0600)]
gallium: added tgsi_num_tokens() function to return number of tokens in token array.
Maybe move to a different file someday.
José Fonseca [Mon, 24 Mar 2008 22:30:33 +0000 (22:30 +0000)]
gallium: Simple cross platform get-opt system.
Uses getenv on Linux, and a memory mapped text file on Windows.
It supports boolean options, flags, and plain strings.
José Fonseca [Mon, 24 Mar 2008 20:18:59 +0000 (20:18 +0000)]
gallium: cleanup p_debug
Now debug_printf is disabled on release builds. Use debug_error or
_debug_printf to output messages on release versions.
Brian [Mon, 24 Mar 2008 18:51:50 +0000 (12:51 -0600)]
gallium: move filter assignment out of loop
Brian [Mon, 24 Mar 2008 18:50:28 +0000 (12:50 -0600)]
gallium: fix REALLOC() to copy smaller of old/new sizes
José Fonseca [Mon, 24 Mar 2008 18:25:31 +0000 (18:25 +0000)]
gallium: Protect operator precedence in macro.
Brian Paul [Mon, 24 Mar 2008 17:55:29 +0000 (11:55 -0600)]
gallium: free bitmap fragment shaders, misc clean-up
Brian Paul [Mon, 24 Mar 2008 17:46:41 +0000 (11:46 -0600)]
gallium: add missing cso_save_samplers() call
Michal Krol [Mon, 24 Mar 2008 15:50:39 +0000 (16:50 +0100)]
gallium: Remove util_gen_mipmap_filter().
Brian [Mon, 24 Mar 2008 14:53:16 +0000 (08:53 -0600)]
gallium: pass the filter mode to util_gen_mipmap().
Remove util_gen_mipmap_filter() when no longer used.
Michal Krol [Mon, 24 Mar 2008 14:09:34 +0000 (08:09 -0600)]
gallium: tweak coords in u_gen_mipmap code
Keith Whitwell [Mon, 24 Mar 2008 10:53:26 +0000 (10:53 +0000)]
draw: pre-declare referenced structs
Keith Whitwell [Mon, 24 Mar 2008 10:52:20 +0000 (10:52 +0000)]
Revert "draw: Fix #include order to fix Linux build."
This reverts commit
a6d17bf671d6bfbb187a62ba14b9ad08fb5dafe1.
Michel Dänzer [Mon, 24 Mar 2008 09:56:12 +0000 (09:56 +0000)]
draw: Fix #include order to fix Linux build.
Ben Skeggs [Mon, 24 Mar 2008 02:45:38 +0000 (13:45 +1100)]
nv40: maintain pipe_surface status field
Ben Skeggs [Mon, 24 Mar 2008 02:41:40 +0000 (13:41 +1100)]
nv40: respect rasterizer cso bypass_clipping flag
Ben Skeggs [Mon, 24 Mar 2008 02:01:35 +0000 (13:01 +1100)]
nouveau: fix infinite wait on space for userbuf.
A recent commit removed a bug which essentially caused a sync after each
use of a user buffer. In fixing it, the scratch area can now become
fragmented under some circumstances leading to nouveau_bo_tmp() waiting
forever for a large enough block of free space.
This fixes ppracer, gl-117 and probably a heap of other apps.
Keith Whitwell [Sun, 23 Mar 2008 21:55:19 +0000 (21:55 +0000)]
draw: check need_pipeline() in passthrough
Keith Whitwell [Sun, 23 Mar 2008 21:54:44 +0000 (21:54 +0000)]
draw: tweak the definition of draw_need_pipeline, fix minor bug
Keith Whitwell [Sun, 23 Mar 2008 21:31:23 +0000 (21:31 +0000)]
draw: fix crlf
Keith Whitwell [Sun, 23 Mar 2008 21:31:14 +0000 (21:31 +0000)]
draw: rename emit functions to match pipe_format names
Keith Whitwell [Sun, 23 Mar 2008 20:36:00 +0000 (20:36 +0000)]
draw: restructure fetch/emit as a pair of function calls
Michal Krol [Sun, 23 Mar 2008 19:39:35 +0000 (20:39 +0100)]
gallium: Add util_gen_mipmap_filter().
We need a way to specify the type of minification filter
used to downsample mipmap levels.
The old util_gen_mipmap() retains its behaviour and uses
LINEAR filter.
José Fonseca [Sun, 23 Mar 2008 19:38:11 +0000 (19:38 +0000)]
gallium: Fix memory leak.
pipe cso's were being destroyed, but the hash elements themselves not.
proper fix is IMHO add a destructor callback to cso_hash.
José Fonseca [Sun, 23 Mar 2008 18:57:35 +0000 (18:57 +0000)]
gallium: Remove the debug_mask_* stuff.
Overcomplex and not much different from using a global variable...
José Fonseca [Sun, 23 Mar 2008 18:52:37 +0000 (18:52 +0000)]
gallium: Fix typo.
José Fonseca [Sun, 23 Mar 2008 18:38:10 +0000 (18:38 +0000)]
gallium: Memory debugging utilities.
There are no known tools for windows kernel memory debugging, so this is a
simple set of malloc etc wrappers.
Enabled by default on win32 debug builds
José Fonseca [Sun, 23 Mar 2008 18:30:53 +0000 (18:30 +0000)]
gallium: wrap decls in extern "C"
Keith Whitwell [Sun, 23 Mar 2008 18:21:00 +0000 (18:21 +0000)]
draw: fix some unsigned vs ushort confusion
Middle-end elements are ushort, but prior to that have to treat all
elements as unsigned to avoid wrapping and/or overruns.
Keith Whitwell [Sun, 23 Mar 2008 16:44:59 +0000 (16:44 +0000)]
gallium: beginnings of draw module vertex rework
Trying to put a structure in place that we can actually optimize.
Initially just implementing a passthrough mode, this will fairly soon
replace all the vertex_cache/prim_queue/shader_queue stuff that's so
hard to understand...
Split the vertex processing into a couple of distinct stages:
- Frontend
- Prepares two lists of elements (fetch and draw) to be processed
by the next stage. This stage doesn't fetch or draw vertices, but
makes the decision which to draw. Multiple implementations of this
will implement different strategies, currently just a vcache
implementation.
- MiddleEnd
- Takes the list of fetch elements, fetches them, runs the vertex
shader, cliptest, viewport transform on them to produce a
linear array of vertex_header vertices.
- Passes that list of vertices, plus the draw_elements (which index
into that list) onto the backend
- Backend
- Either the existing primitive/clipping pipeline, or the vbuf_render
hardware backend provided by the driver.
Currently, the middle-end is the old passthrough code, and it build hardware
vertices, not vertex_header vertices as above. It may be that passthrough
is a special case in this respect.
Ben Skeggs [Sun, 23 Mar 2008 08:08:59 +0000 (19:08 +1100)]
nv40: vp reg changes similar to recent fp changes
Ben Skeggs [Sun, 23 Mar 2008 07:20:04 +0000 (18:20 +1100)]
nv40: fix fp depth write
Ben Skeggs [Sun, 23 Mar 2008 07:11:49 +0000 (18:11 +1100)]
nv40: remove some badness from fp temp/result allocation
Should hopefully be OK now (on the fragprog size) for MRT. The hack from
a commit
01cb2cd93efe7ad94d7fd36aa5a776c2e3ab4c7d is no longer needed.
Ben Skeggs [Sun, 23 Mar 2008 05:31:31 +0000 (16:31 +1100)]
nv40: catch fp extra-const cases where both const and immd have same idx
Ben Skeggs [Sun, 23 Mar 2008 05:26:42 +0000 (16:26 +1100)]
nv40: fp: fix multiple refs to a single const withing an instruction
Ben Skeggs [Sun, 23 Mar 2008 05:11:03 +0000 (16:11 +1100)]
nv40: workaround fp result.data[0] clobberage
Temporary, the fp reg handling will get reworked at some point in the near
future. But before that, there's a few bugs to find.
Ben Skeggs [Sun, 23 Mar 2008 03:29:35 +0000 (14:29 +1100)]
nv40: add dxtn formats (disabled)
Ben Skeggs [Sun, 23 Mar 2008 03:18:06 +0000 (14:18 +1100)]
Merge remote branch 'upstream/gallium-0.1' into nouveau-gallium-0.1
Brian [Sat, 22 Mar 2008 16:29:30 +0000 (10:29 -0600)]
gallium: fix a few more shader-related mem leaks
Brian [Sat, 22 Mar 2008 16:27:55 +0000 (10:27 -0600)]
use ctx->Driver.DeleteProgram() in a few more places
Brian [Sat, 22 Mar 2008 15:42:59 +0000 (09:42 -0600)]
gallium: free samplers, textures in destroy()
Brian [Sat, 22 Mar 2008 15:15:23 +0000 (09:15 -0600)]
gallium: remove temporary _screen suffix from function names
Brian [Sat, 22 Mar 2008 15:12:52 +0000 (09:12 -0600)]
gallium: fix a few mem leaks when deleting programs
Brian [Sat, 22 Mar 2008 15:12:27 +0000 (09:12 -0600)]
gallium: fix mem leak (fee pstipple stage)
Brian [Sat, 22 Mar 2008 15:12:02 +0000 (09:12 -0600)]
delete default programs with ctx->Driver.DeleteProgram()
José Fonseca [Sat, 22 Mar 2008 10:05:55 +0000 (10:05 +0000)]
gallium: Remove pedantic asserts.
Move these to a higher level instead.
Ian Romanick [Fri, 21 Mar 2008 18:20:49 +0000 (11:20 -0700)]
cell: Generate blend / depth test code when state atom is created
Code generation should be performed when the device-specific state
atom is created, not when it is bound.
Ian Romanick [Fri, 21 Mar 2008 18:15:49 +0000 (11:15 -0700)]
cell: Change code-gen for CONST_COLOR blend factor
Previously the constant color blend factor was compiled into the
generated code. This meant that the code had to be regenerated each
time the constant color was changed. This doesn't fit with the model
used in Gallium.
As-is, the code could be better. The constant color is loaded for
every quad processed, even if it is not used. Also, if a lot of (1-x)
blend factors are used, 1.0 will be loaded and reloaded into registers
many times.
Ian Romanick [Fri, 21 Mar 2008 17:25:58 +0000 (10:25 -0700)]
Tabs to spaces
Ian Romanick [Fri, 21 Mar 2008 17:23:52 +0000 (10:23 -0700)]
cell: Remove unnecessary default_blend work-around
I suspect that there was some other bug in the blend code-gen that
made this work-around necessary.
Brian [Fri, 21 Mar 2008 21:54:27 +0000 (15:54 -0600)]
gallium: fix const buffer update bug
If only glUniform is called between two renderings, the const buffers weren't
getting updated. Need to set the _NEW_PROGRAM flag in st_upload_constants()
as that's the dirty flag set by glUniform.
Fixes glean tapi2 test.
Brian [Fri, 21 Mar 2008 20:20:07 +0000 (14:20 -0600)]
Refactor PBO validate/map code.
We always need to do PBO validation, so do that in core Mesa before calling driv
er routine.
cherry-picked from Mesa/master.
Brian Paul [Fri, 21 Mar 2008 19:44:27 +0000 (13:44 -0600)]
gallium: implement PBO operations for glDraw/ReadPixels and glBitmap
At this time there are no optimizations for directly blitting between
buffer objects and surfaces (always go through mappings).
glean pbo test passes now
Brian Paul [Fri, 21 Mar 2008 19:43:07 +0000 (13:43 -0600)]
add a number of PBO validate/map/unmap functions
Helper functions for (some) drivers, including swrast.
cherry-picked from Mesa/master
Brian [Fri, 21 Mar 2008 18:32:48 +0000 (12:32 -0600)]
Fix some PBO breakage.
In _mesa_Bitmap, can't early return if bitmap ptr is NULL, it may be an offset
into a PBO. Similarly for _mesa_GetTexImage.
Brian [Fri, 21 Mar 2008 16:03:23 +0000 (10:03 -0600)]
i915: added to-do note about setting the max_lod register to get proper min/mag filter selection
Brian Paul [Fri, 21 Mar 2008 17:06:08 +0000 (11:06 -0600)]
mesa: minor code re-org
Brian Paul [Fri, 21 Mar 2008 17:05:02 +0000 (11:05 -0600)]
gallium: use is_format_supported(PIPE_FORMAT_DXT5_RGBA) to check if s3tc is supported
The PIPE_CAP_S3TC token will be deprecated.
Brian Paul [Fri, 21 Mar 2008 17:03:53 +0000 (11:03 -0600)]
gallium: added DXT formats (preliminary, will probably change)
Brian Paul [Fri, 21 Mar 2008 16:51:21 +0000 (10:51 -0600)]
gallium: use is_format_supported() to determine if GL_EXT_texture_sRGB is supported
Brian Paul [Fri, 21 Mar 2008 16:50:26 +0000 (10:50 -0600)]
gallium: document is_format_supported()'s type param
Brian Paul [Fri, 21 Mar 2008 16:45:33 +0000 (10:45 -0600)]
gallium: PIPE_FORMAT_TYPE_ comments
Brian Paul [Fri, 21 Mar 2008 16:43:56 +0000 (10:43 -0600)]
gallium: added sRGB formats
Brian Paul [Fri, 21 Mar 2008 16:38:56 +0000 (10:38 -0600)]
gallium: additional comments, fix typos, etc
Ben Skeggs [Fri, 21 Mar 2008 12:24:16 +0000 (23:24 +1100)]
nv40: fix bug in query code
Michel Dänzer [Fri, 21 Mar 2008 12:14:06 +0000 (12:14 +0000)]
gallium: Update scons build for state tracker files added/removed.
Ben Skeggs [Fri, 21 Mar 2008 12:09:01 +0000 (23:09 +1100)]
nv40: add 16-bit SSCALED vb formats
Ben Skeggs [Fri, 21 Mar 2008 12:03:25 +0000 (23:03 +1100)]
nv40: ensure vb relocs don't end up on list for swtnl
Avoids bo code bailing out because of mapped buffers being validated.
Ben Skeggs [Fri, 21 Mar 2008 11:23:19 +0000 (22:23 +1100)]
nv40: call semi-magic 0x1d88 method
Ben Skeggs [Fri, 21 Mar 2008 11:09:29 +0000 (22:09 +1100)]
nv40: swtnl fallback on unsupported array format
Ben Skeggs [Fri, 21 Mar 2008 11:02:49 +0000 (22:02 +1100)]
nv40: oops
Ben Skeggs [Fri, 21 Mar 2008 11:01:22 +0000 (22:01 +1100)]
nv40: s/free/FREE/
Ben Skeggs [Fri, 21 Mar 2008 10:58:17 +0000 (21:58 +1100)]
nv40: align each level to 64 pixels
Ben Skeggs [Fri, 21 Mar 2008 10:17:57 +0000 (21:17 +1100)]
nv40: use num_cbufs
Ben Skeggs [Fri, 21 Mar 2008 07:28:07 +0000 (18:28 +1100)]
nouveau: do a retry if initial buffer alloc fails
Ben Skeggs [Fri, 21 Mar 2008 07:15:58 +0000 (18:15 +1100)]
nouveau: use saved fence pointer, not head of unsignalled list
It's possible for the unsignalled list head to change during fence_flush.
Fixes valgrind complaint exposed by a previous commit.
Ben Skeggs [Fri, 21 Mar 2008 06:40:01 +0000 (17:40 +1100)]
nouveau: potentially delay buffer destruction
Rather than spinning waiting for the GPU to finish with the buffer,
add a callback on the buffer's fence object instead. Gives a sizable
performance increase in OA.
Ben Skeggs [Fri, 21 Mar 2008 06:38:24 +0000 (17:38 +1100)]
nouveau: try combining prev and next resources on free
Fixes some cases where we end up with a list of many unused chunks that are
too small to be useful.
Ben Skeggs [Fri, 21 Mar 2008 01:57:05 +0000 (12:57 +1100)]
nv40: use new pipe_framebuffer width/height fields
Ben Skeggs [Fri, 21 Mar 2008 01:54:34 +0000 (12:54 +1100)]
Merge remote branch 'upstream/gallium-0.1' into nouveau-gallium-0.1
Ian Romanick [Fri, 21 Mar 2008 01:17:48 +0000 (18:17 -0700)]
cell: Fix several bugs in blend code-gen
- Alpha factor set to a _COLOR mode was mishandled
- Cases when either dst factor or src factor was ZERO were mishandled
- MIN and MAX cases were backwards
- Case when blend was disabled was mishandled
- Incorrect comments about number of instructions generated
The tests blendminmax and blendsquare run correctly.
Ian Romanick [Fri, 21 Mar 2008 00:36:31 +0000 (17:36 -0700)]
cell: Call the correct function to generate blending code
Cut-and-paste for the lose. :(
Ian Romanick [Fri, 21 Mar 2008 00:36:05 +0000 (17:36 -0700)]
cell: Fix bus error when there is no depth buffer
Ian Romanick [Thu, 20 Mar 2008 00:29:39 +0000 (17:29 -0700)]
cell: Use code-gen for alpha blend
So far this is only tested when GL_BLEND is disabled.
Brian Paul [Thu, 20 Mar 2008 23:08:07 +0000 (17:08 -0600)]
gallium: catch some out of memory conditions in the texture image code.
st_finalize_texture()'s return code now indicates success/fail instead of
presence of texture border (which we discard earlier).
Brian [Thu, 20 Mar 2008 21:22:20 +0000 (15:22 -0600)]
gallium: set fb.width/height
Brian [Thu, 20 Mar 2008 21:21:55 +0000 (15:21 -0600)]
gallium: enable vp input semantic info