Brian [Tue, 25 Sep 2007 20:29:11 +0000 (14:29 -0600)]
Translate mesa vertex/fragment programs to TGSI programs at same time to do proper linking.
Previously, programs were translated independently during validation.
The problem is the translation to TGSI format, which packs shader
input/outputs into continuous slots, depends on which vertex program is
being paired with which fragment shader. Now, we look at the outputs
of the vertex program in conjunction with the inputs of the fragment shader
to be sure the attributes match up correctly.
The new 'linked_program_pair' class keeps track of the associations
between vertex and fragment shaders. It's also the place where the TGSI
tokens are kept since they're no longer per-program state but per-linkage.
Still a few loose ends, like implementing some kind of hash/lookup table
for linked_program_pairs.
Brian [Tue, 25 Sep 2007 20:22:13 +0000 (14:22 -0600)]
translate Mesa programs to TGSI programs (formerly in st_atom_[fv]s.c)
Brian [Tue, 25 Sep 2007 20:21:18 +0000 (14:21 -0600)]
vertex/fragment shader state and linking
Brian [Fri, 21 Sep 2007 18:41:24 +0000 (12:41 -0600)]
test for presence of both front and back color vertex slots
Brian [Fri, 21 Sep 2007 18:39:26 +0000 (12:39 -0600)]
remove unused lookup array
Keith Whitwell [Tue, 25 Sep 2007 12:20:53 +0000 (13:20 +0100)]
First attempt at building vertex buffers post-clip.
Build a buffer of contigous vertices and indices at the backend of our
software transformation/clipping path. This will become the mechanism
for emitting buffers of vertices to rasterization hardware.
This is similar to but not the same as the post-transform vertex cache.
In particular, these vertices are subject to clipping, culling, poly offset,
etc. The vertices emitted will all be used by hardware.
TODOs include the actual transformation to hardware vertex formats, moving
this out of softpipe to somewhere more useful and allowing >1 primitive to
share the generated VB.
keithw [Tue, 25 Sep 2007 12:16:25 +0000 (13:16 +0100)]
turn off shader debug
keithw [Tue, 25 Sep 2007 11:08:16 +0000 (12:08 +0100)]
test against -1 for front/back attrib
michal [Mon, 24 Sep 2007 11:32:26 +0000 (12:32 +0100)]
Enable SSE2 for FS.
Zack Rusin [Mon, 24 Sep 2007 09:43:07 +0000 (05:43 -0400)]
In here we actually do want to assign it. Fix compilation.
Brian [Fri, 21 Sep 2007 18:37:13 +0000 (12:37 -0600)]
invert draw_vertex_cache_check_space()
Brian [Fri, 21 Sep 2007 18:36:59 +0000 (12:36 -0600)]
added PSIZE
Brian [Fri, 21 Sep 2007 18:21:05 +0000 (12:21 -0600)]
fix some post-transform vertex layout issues (bcolors, psize emitted last
Brian [Fri, 21 Sep 2007 18:06:08 +0000 (12:06 -0600)]
Fix up some point size breakage. Start on fogcoord too.
Brian [Fri, 21 Sep 2007 17:55:28 +0000 (11:55 -0600)]
reenable some clear code that was temporarily disabled
Brian [Fri, 21 Sep 2007 17:54:57 +0000 (11:54 -0600)]
remove debug printfs
Brian [Fri, 21 Sep 2007 17:34:28 +0000 (11:34 -0600)]
comments
Brian [Fri, 21 Sep 2007 17:34:20 +0000 (11:34 -0600)]
use VERT_ATTRIB_MAX for loop limit
Brian [Fri, 21 Sep 2007 17:22:34 +0000 (11:22 -0600)]
remove SEMANTIC_TEXCOORD
Brian [Fri, 21 Sep 2007 17:20:29 +0000 (11:20 -0600)]
Get rid of TGSI_SEMANTIC_TEXCOORD and just use TGSI_SEMANTIC_GENERIC.
Brian [Fri, 21 Sep 2007 17:01:26 +0000 (11:01 -0600)]
rename index var from i to attr, use FRAG_ATTRIB_MAX instead of 16
Brian [Fri, 21 Sep 2007 16:57:30 +0000 (10:57 -0600)]
minor renaming
Brian [Fri, 21 Sep 2007 16:56:32 +0000 (10:56 -0600)]
pass vs output semantics to tgsi_mesa_compile_vp_program() to generate output DECs
Brian [Fri, 21 Sep 2007 16:53:45 +0000 (10:53 -0600)]
fix in/out typos
Brian [Fri, 21 Sep 2007 16:40:18 +0000 (10:40 -0600)]
Pass vertex program input semantics into tgsi_mesa_compile_vp_program() to produce DCLs.
Brian [Fri, 21 Sep 2007 16:38:18 +0000 (10:38 -0600)]
fix 1/i typos
José Fonseca [Fri, 21 Sep 2007 15:47:00 +0000 (16:47 +0100)]
Fix the doxygen target in the top-level Makefile.
Zack Rusin [Fri, 21 Sep 2007 11:06:19 +0000 (07:06 -0400)]
Convert i915 to use cso semantics of alpha_test.
Zack Rusin [Fri, 21 Sep 2007 11:00:20 +0000 (07:00 -0400)]
Make the alpha test state a cso.
Zack Rusin [Fri, 21 Sep 2007 08:56:39 +0000 (04:56 -0400)]
Silence a few warnings.
Brian [Fri, 21 Sep 2007 01:06:26 +0000 (19:06 -0600)]
remove :5 form format field in pipe_surface, makes no difference at this time
Brian [Thu, 20 Sep 2007 21:32:18 +0000 (15:32 -0600)]
remove old code in src_vector()
Brian [Thu, 20 Sep 2007 20:47:22 +0000 (14:47 -0600)]
Clean-up the TGSI_SEMANTIC tokens, introduce semantic indexes.
Still need to produce decl instructions for vertex shaders...
Brian [Thu, 20 Sep 2007 20:08:09 +0000 (14:08 -0600)]
remove old/used code
Brian [Thu, 20 Sep 2007 20:03:00 +0000 (14:03 -0600)]
remove lots of dead code related to program input/output mapping
Brian [Thu, 20 Sep 2007 19:58:47 +0000 (13:58 -0600)]
obsolete
Brian [Thu, 20 Sep 2007 19:58:12 +0000 (13:58 -0600)]
remove #includes of tgsi_attribs.h
Brian [Thu, 20 Sep 2007 19:50:53 +0000 (13:50 -0600)]
checkpoint: TGSI_ATTRIB_x tokens no longer used
Brian [Thu, 20 Sep 2007 19:42:37 +0000 (13:42 -0600)]
Checkpoint: vertex attribute clean-up.
Remove/disable the attrib/slot mapping arrays in a few places.
Work in progress...
Brian [Thu, 20 Sep 2007 19:39:17 +0000 (13:39 -0600)]
Always update st->state.vs, not just when the program is dirty.
This fixes a regression in the cubemap.c demo which alternates between
two different vertex shaders.
Brian [Thu, 20 Sep 2007 15:55:56 +0000 (09:55 -0600)]
remove unused lookup[]
Brian [Thu, 20 Sep 2007 15:43:51 +0000 (09:43 -0600)]
fix input/output typos
Zack Rusin [Thu, 20 Sep 2007 16:34:31 +0000 (12:34 -0400)]
Cache the i915 sampler state.
Zack Rusin [Thu, 20 Sep 2007 15:33:35 +0000 (11:33 -0400)]
Make the rasterizer state in i915 use the cso semantics.
Zack Rusin [Thu, 20 Sep 2007 15:33:05 +0000 (11:33 -0400)]
Allow drivers to cache the template in case they want to
fallback through softpipe which will require the template.
Zack Rusin [Thu, 20 Sep 2007 14:28:20 +0000 (10:28 -0400)]
Rewrite the depth_stencil state handling in i915.
Done to match the new cso semantics. translate in create,
use in bind and later delete.
Zack Rusin [Thu, 20 Sep 2007 14:07:10 +0000 (10:07 -0400)]
Fix failover state binding and convert the sampler to use the new
state constant state object semantics.
Zack Rusin [Thu, 20 Sep 2007 12:35:10 +0000 (08:35 -0400)]
Convert depth_stencil state to the new semantics.
Zack Rusin [Thu, 20 Sep 2007 11:50:33 +0000 (07:50 -0400)]
Switch fragment/vertex shaders to the new caching semantics.
Allow driver custom allocation within cached objects. The shaders
are currently twiced (by cso layer and by the program itself).
Brian [Thu, 20 Sep 2007 00:53:36 +0000 (18:53 -0600)]
Checkpoint: replacement of TGSI_ATTRIB_x tokens with input/output semantics.
TGSI_ATTRIB_x tokens still present and used in a few places.
Expanded set of TGSI_SEMANTIC_x tokens for describing the meaning
of inputs/outputs. These tokens are in a crude state ATM.
Lots of #if 0 / disabled code to be removed yet, etc...
Softpipe and i915 drivers should be in working condition but not heavily tested.
José Fonseca [Wed, 19 Sep 2007 22:43:36 +0000 (23:43 +0100)]
Update linux-profile config.
Zack Rusin [Wed, 19 Sep 2007 19:29:17 +0000 (15:29 -0400)]
Fix a really stupid mistake that Michel hit. Return proper hash.
Michel Dänzer [Wed, 19 Sep 2007 19:03:08 +0000 (20:03 +0100)]
Fix window resizes.
The memcmp is insufficient for eliminating redundant framebuffer state changes.
Michel Dänzer [Wed, 19 Sep 2007 16:59:51 +0000 (17:59 +0100)]
Pad surface dimensions to multiples of 2.
Avoids an assertion failure with softpipe if requested width or height is odd.
Zack Rusin [Wed, 19 Sep 2007 18:14:03 +0000 (14:14 -0400)]
Adding proper includes often helps with the warnings,
or so i hear..
Zack Rusin [Wed, 19 Sep 2007 18:06:19 +0000 (14:06 -0400)]
Avoid redundant reallocation of the template.
cso already allocated the template for us. Returning 0 means
we have no driver specific representation and just want
the template on the bind.
Zack Rusin [Wed, 19 Sep 2007 18:01:18 +0000 (14:01 -0400)]
Convert the rasterizer cso to the new semantics.
Basically make cso hold the driver specific struct, while
managing the template.
Zack Rusin [Wed, 19 Sep 2007 17:05:58 +0000 (13:05 -0400)]
Convert the i915 driver blend state to the new semantics.
Zack Rusin [Wed, 19 Sep 2007 16:35:29 +0000 (12:35 -0400)]
Redo the cso cache to map driver data in a lot more pleasing way.
Drivers can now create whatever they want from the state template. We
use cso_state object to store the template (necessary during lookups),
and the driver data. Convert blend state to the new semantics.
Brian [Wed, 19 Sep 2007 16:16:46 +0000 (10:16 -0600)]
grab vertex program pointer after state validation
Michel Dänzer [Wed, 19 Sep 2007 15:39:04 +0000 (16:39 +0100)]
Only generate vertex shader code when necessary.
Zack Rusin [Wed, 19 Sep 2007 10:46:32 +0000 (06:46 -0400)]
Finish up conversions of shaders to immutable objects.
Create/Delete calls should be split since in create we'll be
compiling them so we want to know which one it is (vertex/fragment).
Brian [Wed, 19 Sep 2007 01:37:36 +0000 (19:37 -0600)]
Checkpoint: rework shader input/output register mapping.
This is a step toward removing TGSI_ATTRIB_ tokens.
Basically, when translating Mesa programs to TGSI programs, pass in input and
output register re-maps, plus interpolation info.
There's some known breakage (cubemap.c) so more to be done...
Brian [Wed, 19 Sep 2007 01:29:26 +0000 (19:29 -0600)]
temporarily set the FRAG_BIT_FOGC bit in InputsRead when fog is enabled
Michel Dänzer [Tue, 18 Sep 2007 20:50:31 +0000 (21:50 +0100)]
intel_winsys: Disable scheduled buffer swaps for now.
They don't work with private back buffers yet.
This gets vsync working.
Michel Dänzer [Tue, 18 Sep 2007 20:49:43 +0000 (21:49 +0100)]
Fix linux-dri-x86 target.
Brian [Tue, 18 Sep 2007 18:55:53 +0000 (12:55 -0600)]
use perspective projection, press r/R to rotate the quad
Brian [Tue, 18 Sep 2007 18:55:20 +0000 (12:55 -0600)]
Fix some issues with perspective-corrected interpolation.
In mesa_to_tgsi.c, use TGSI_INTERPOLATE_PERSPECTIVE by default (to match
post-transform vertex info convention). More to be done there...
In sp_quad_fs.c, interpolate W in addition to Z. This fixes the divide
by zero happening in perspective_interpolation() tgsi_exec.c
As it was, we were only getting perspective correction of texture coords
used by the TGSI_TEX instruction since it does a homogeneous divide.
Other coords/varyings were incorrect.
Brian [Tue, 18 Sep 2007 18:50:20 +0000 (12:50 -0600)]
additional comments for tri_persp_coeff()
Zack Rusin [Tue, 18 Sep 2007 17:24:44 +0000 (13:24 -0400)]
Convert shader to an immutable state object.
Zack Rusin [Tue, 18 Sep 2007 14:02:51 +0000 (10:02 -0400)]
Add emacs dirvars file with Mesa indention definition.
Zack Rusin [Tue, 18 Sep 2007 14:02:16 +0000 (10:02 -0400)]
Finishing up rename of the setup state to the rasterizer state.
Zack Rusin [Tue, 18 Sep 2007 10:24:54 +0000 (06:24 -0400)]
move cso cache to the pipe directory
Zack Rusin [Mon, 17 Sep 2007 17:49:12 +0000 (13:49 -0400)]
Fix a silly bug on setting samplers.
Bind when the currently bound state is /not/ the one
we just found/created.
Zack Rusin [Mon, 17 Sep 2007 16:59:50 +0000 (12:59 -0400)]
converting the setup state to immutable object and renaming it to rasterizer state
Zack Rusin [Mon, 17 Sep 2007 15:55:18 +0000 (11:55 -0400)]
Combing depth and stencil objects and making them immutable.
Converting depth and stencil objects into a single state object
(d3d10 like) and making it immutable.
Zack Rusin [Mon, 17 Sep 2007 13:50:08 +0000 (09:50 -0400)]
Fix the warning.
The const is there to enforce the immutable state of the object,
which is in reality owned by the pipe so just cast away the
constness.
Zack Rusin [Mon, 17 Sep 2007 13:47:41 +0000 (09:47 -0400)]
Make sampler an immutable state object.
Switch the sample to be an immutable state object.
Zack Rusin [Mon, 17 Sep 2007 11:56:56 +0000 (07:56 -0400)]
Implementing a better hash, removing state_tracker dependency from the cache.
Replacing mesa's main hash with one that handles collisions, moving state_tracker
related caching to the state tracker to keep cso cache independent of it. Cleanups.
Zack Rusin [Fri, 14 Sep 2007 08:08:58 +0000 (04:08 -0400)]
First stab at immutable state objects (create/bind/delete)
We want our state objects to be immutable, handled via the
create/bind/delete calls instead of struct propagation.
Only implementing the blend state to see how it would look like
and work.
Brian [Mon, 17 Sep 2007 21:44:59 +0000 (15:44 -0600)]
Replace GL_TEXTURE_* tokens with PIPE_TEXTURE_*
Must have missed this a while back. This fixes cube mapping.
Brian [Mon, 17 Sep 2007 20:24:11 +0000 (14:24 -0600)]
Plug in selection/feedback code.
Not quite finished yet.
Selection/feedback are done with a private instance of the 'draw' module
in the state tracker. Not quite all the draw context's state is set yet,
namely vertex format info. Hold off on that for a bit...
Brian [Mon, 17 Sep 2007 20:22:49 +0000 (14:22 -0600)]
selection/feedback support
Brian [Mon, 17 Sep 2007 18:18:32 +0000 (12:18 -0600)]
s/bugmgr/bufgr/
Brian [Mon, 17 Sep 2007 18:07:29 +0000 (12:07 -0600)]
Merge branch 'softpipe_0_1_branch' of git+ssh://brianp@git.freedesktop.org/git/mesa/mesa into softpipe_0_1_branch
Keith Whitwell [Mon, 17 Sep 2007 17:02:46 +0000 (18:02 +0100)]
Remove non-meaningful region_idle() interface
This is already handled by the semanics of map(), etc.
Brian [Mon, 17 Sep 2007 16:05:34 +0000 (10:05 -0600)]
silence warning
Brian [Mon, 17 Sep 2007 16:05:21 +0000 (10:05 -0600)]
update begin/end_query() functions
Michel Dänzer [Mon, 17 Sep 2007 12:22:51 +0000 (13:22 +0100)]
Fix linux-dri-x86 build.
Michel Daenzer [Fri, 14 Sep 2007 14:51:00 +0000 (15:51 +0100)]
intel_winsys: Fix intel_buffer_reference.
Correctly clear the pointer to the old buffer (not sure how this could build at
all before...) and only reference the new one when its pointer is non-NULL.
Michel Daenzer [Fri, 14 Sep 2007 14:48:54 +0000 (15:48 +0100)]
intel_winsys: Adapt to recent drm changes (s/pipes/planes/).
Brian [Tue, 11 Sep 2007 22:01:17 +0000 (16:01 -0600)]
Implement query object interface.
This replaces the temporary occlusion counter functions we had before.
Added new ctx->Driver.WaitQuery() function which should block until the result is ready.
Sketch out some code for vertex transformation feedback counters.
Brian [Mon, 10 Sep 2007 23:02:58 +0000 (17:02 -0600)]
vertex feedback functions
Brian [Mon, 10 Sep 2007 22:52:05 +0000 (16:52 -0600)]
merge buffer_unreference(), buffer_reference()
Brian [Mon, 10 Sep 2007 22:35:16 +0000 (16:35 -0600)]
clean-ups, comments
Brian [Mon, 10 Sep 2007 22:32:24 +0000 (16:32 -0600)]
s/xm_softpipe.c/xm_winsys.c/
Brian [Mon, 10 Sep 2007 22:31:24 +0000 (16:31 -0600)]
Rename
Brian [Mon, 10 Sep 2007 22:29:23 +0000 (16:29 -0600)]
More work on vertex feedback / glRasterPos. Basic rasterpos works now.
Brian [Mon, 10 Sep 2007 22:28:39 +0000 (16:28 -0600)]
added st_cb_rasterpos.c
Brian [Mon, 10 Sep 2007 22:28:27 +0000 (16:28 -0600)]
plug in rasterpos/feedback code