Zack Rusin [Mon, 24 Dec 2007 12:57:34 +0000 (07:57 -0500)]
i965: a little better way of handling immediates
Zack Rusin [Thu, 20 Dec 2007 17:54:23 +0000 (12:54 -0500)]
i965: very crude and hacky way of handling immediates
Keith Whitwell [Thu, 20 Dec 2007 13:47:46 +0000 (13:47 +0000)]
gallium: make state tracker explictly ask for rendercache flushes
Keith Whitwell [Thu, 20 Dec 2007 13:47:11 +0000 (13:47 +0000)]
965: respect pipe flush flags
Now we emit way too many flushes instead of none at all.
Keith Whitwell [Thu, 20 Dec 2007 13:19:56 +0000 (13:19 +0000)]
gallium: translate ARB fp/vp immediates consistently to tgsi immediates
Zack Rusin [Thu, 20 Dec 2007 12:05:52 +0000 (07:05 -0500)]
965: fix the constant buffers
Brian [Wed, 19 Dec 2007 21:06:22 +0000 (14:06 -0700)]
Add some prototype code for converting RET to END for main(). Disabled for now.
Brian [Wed, 19 Dec 2007 20:53:28 +0000 (13:53 -0700)]
temporarily defeat an assertion
Brian [Wed, 19 Dec 2007 20:45:00 +0000 (13:45 -0700)]
Fix problem with initial viewport/scissor size.
If an app never called glViewport, the viewport size was always 0 by 0 pixels.
Now pass initial size to st_create_framebuffer() and initialize the viewport
and scissor bounds in st_make_current().
This could also be fixed by ensuring the gl_framebuffers passed to
_mesa_make_current() were initialized to the right size. But that involves
allocating the renderbuffers/pipe_surfaces earlier and that runs into some
other issues ATM.
Also remove obsolete createRenderbuffers param to st_create_framebuffer().
Brian [Wed, 19 Dec 2007 18:50:50 +0000 (11:50 -0700)]
convert Mesa OPCODE_END to TGSI_OPCODE_END, not TGSI_OPCOD_RET
Brian [Wed, 19 Dec 2007 15:51:17 +0000 (08:51 -0700)]
remove obsolete TXP, add some sanity checks
Brian [Wed, 19 Dec 2007 15:50:52 +0000 (08:50 -0700)]
special-case PSIZE too
Brian [Wed, 19 Dec 2007 15:50:35 +0000 (08:50 -0700)]
move st_make_current() before buffer size check so renderbuffer alloc storage works
Brian [Wed, 19 Dec 2007 15:49:59 +0000 (08:49 -0700)]
new assertions
Keith Whitwell [Wed, 19 Dec 2007 19:05:19 +0000 (19:05 +0000)]
965: dump curbe contents to stderr
Keith Whitwell [Wed, 19 Dec 2007 19:04:35 +0000 (19:04 +0000)]
965: handle BRW_CONSTANT_BUFFER data type
Zack Rusin [Wed, 19 Dec 2007 18:22:27 +0000 (13:22 -0500)]
consts
Keith Whitwell [Wed, 19 Dec 2007 15:48:53 +0000 (15:48 +0000)]
965: count grf allocation correctly for wm prog, first aub triangle
Keith Whitwell [Wed, 19 Dec 2007 15:16:25 +0000 (15:16 +0000)]
965: hardwire correct behaviour for vp-tri setup (for now...)
Keith Whitwell [Wed, 19 Dec 2007 15:13:20 +0000 (15:13 +0000)]
vp-tri: match vertex/color data of other tests
Keith Whitwell [Wed, 19 Dec 2007 14:20:23 +0000 (14:20 +0000)]
965: align buffer allocations to 4k
Keith Whitwell [Wed, 19 Dec 2007 13:13:42 +0000 (13:13 +0000)]
use a depth buffer temporarily
Keith Whitwell [Wed, 19 Dec 2007 13:12:34 +0000 (13:12 +0000)]
965: make sure stipple state gets uploaded
Keith Whitwell [Wed, 19 Dec 2007 13:11:56 +0000 (13:11 +0000)]
965: allocate buffer space to hold batch commands
Zack Rusin [Wed, 19 Dec 2007 12:49:42 +0000 (07:49 -0500)]
actually set the max_index. useful in the driver
Keith Whitwell [Wed, 19 Dec 2007 12:23:27 +0000 (12:23 +0000)]
965: fill unused surface pointers with zero
Zack Rusin [Wed, 19 Dec 2007 12:08:13 +0000 (07:08 -0500)]
i965: emit fb write on RET for now (until we get END back)
Keith Whitwell [Wed, 19 Dec 2007 11:49:56 +0000 (11:49 +0000)]
965: pitch is in bytes not pixels
Keith Whitwell [Wed, 19 Dec 2007 11:36:03 +0000 (11:36 +0000)]
965: fix off-by-one in scissor rect
Keith Whitwell [Wed, 19 Dec 2007 11:35:21 +0000 (11:35 +0000)]
965: fix off-by-one in surface dimensions
Keith Whitwell [Wed, 19 Dec 2007 11:32:56 +0000 (11:32 +0000)]
965: respect surface width when dumping bitmap
Zack Rusin [Wed, 19 Dec 2007 11:00:30 +0000 (06:00 -0500)]
i965: actually detect whether a sampler is enabled/disabled
Keith Whitwell [Wed, 19 Dec 2007 11:10:30 +0000 (11:10 +0000)]
965: disable clipping more, and restore drawing rect packet
Keith Whitwell [Wed, 19 Dec 2007 11:10:06 +0000 (11:10 +0000)]
965: disable clipping for now
Keith Whitwell [Wed, 19 Dec 2007 10:55:49 +0000 (10:55 +0000)]
965: fix vertex pointsize state, match default cull mode
Keith Whitwell [Wed, 19 Dec 2007 10:49:16 +0000 (10:49 +0000)]
965: fix colormask state
Keith Whitwell [Wed, 19 Dec 2007 10:20:14 +0000 (10:20 +0000)]
965: populate fp_input_count in setup key
Zack Rusin [Wed, 19 Dec 2007 10:28:28 +0000 (05:28 -0500)]
i965: stuff the outputs into mrf registers when possible
Zack Rusin [Wed, 19 Dec 2007 09:58:29 +0000 (04:58 -0500)]
i965: be smarter about register allocation
Brian [Tue, 18 Dec 2007 23:01:25 +0000 (16:01 -0700)]
turn off TGSI_DEBUG
Brian [Tue, 18 Dec 2007 23:00:58 +0000 (16:00 -0700)]
fix bug on GL_VERTEX_PROGRAM_TWO_SIDE path
Brian [Tue, 18 Dec 2007 23:00:31 +0000 (16:00 -0700)]
setup the frontface register (fog.y, ATM)
Brian [Tue, 18 Dec 2007 22:59:54 +0000 (15:59 -0700)]
fix some semantic info mix-ups in calculate_vertex_layout()
Keith Whitwell [Tue, 18 Dec 2007 17:25:33 +0000 (17:25 +0000)]
gallium: rationalize vertex_element state packet
Remove dst_offset (not used)
Add nr_components, which could be calculated from format, but would be
too much effort.
Update i965 driver to cope.
Keith Whitwell [Tue, 18 Dec 2007 16:57:17 +0000 (16:57 +0000)]
gallium: give userbuffers some storage in the aub buffer pool
Keith Whitwell [Tue, 18 Dec 2007 16:56:22 +0000 (16:56 +0000)]
vbo: unmap and remap immediate vbo before/after each draw.
Also use BufferData(NULL) to get fresh storage and avoid synchronous
operation where we would have to flush and wait for the fence after each
draw because of the map.
This will chew through a whole load of buffer space on small draws, so
it isn't a proper solution. Need to support a no-fence or append mapping
mode to do this right, or use user buffers.
Zack Rusin [Tue, 18 Dec 2007 14:59:54 +0000 (09:59 -0500)]
i965: don't treat swz differently and upload vertex buffers
Brian [Tue, 18 Dec 2007 03:41:20 +0000 (20:41 -0700)]
obsolete
Brian [Mon, 17 Dec 2007 21:05:53 +0000 (14:05 -0700)]
updates for depth/stencil/alpha state
Brian [Mon, 17 Dec 2007 20:59:55 +0000 (13:59 -0700)]
Emit declarations for constants (references to the const buffer)
Zack Rusin [Mon, 17 Dec 2007 19:36:31 +0000 (14:36 -0500)]
i965: index the destination offset with regards to the current index
Zack Rusin [Mon, 17 Dec 2007 19:05:46 +0000 (14:05 -0500)]
i965: return false when the function fails, not succeeds
Zack Rusin [Mon, 17 Dec 2007 18:21:45 +0000 (13:21 -0500)]
Add the new test program for fp's.
Keith Whitwell [Mon, 17 Dec 2007 16:14:29 +0000 (16:14 +0000)]
gallium: incorporate alpha state into depth_stencil state object.
Keith Whitwell [Fri, 14 Dec 2007 20:15:22 +0000 (20:15 +0000)]
965: remove dead state atom decls
Keith Whitwell [Fri, 14 Dec 2007 20:14:58 +0000 (20:14 +0000)]
965: fix for non-indexed draw path
Keith Whitwell [Fri, 14 Dec 2007 20:14:30 +0000 (20:14 +0000)]
965: fix/hack check for NULL texunit in state upload
Keith Whitwell [Fri, 14 Dec 2007 20:14:05 +0000 (20:14 +0000)]
965: fix ExtDivide check
Keith Whitwell [Fri, 14 Dec 2007 20:13:44 +0000 (20:13 +0000)]
965: put brw_constant_buffer atom directly in update list
Brian [Sun, 16 Dec 2007 17:00:51 +0000 (10:00 -0700)]
switch on cpp instead of format
Brian [Sat, 15 Dec 2007 23:51:57 +0000 (16:51 -0700)]
rename some 'mesa' functions
Brian [Sat, 15 Dec 2007 23:44:29 +0000 (16:44 -0700)]
Use tile functions in new p_tile.[ch]
This removes quite a bit of code duplicated in the drivers.
Brian [Sat, 15 Dec 2007 23:42:22 +0000 (16:42 -0700)]
Re-usable tile get/put functions
Brian [Sat, 15 Dec 2007 16:49:20 +0000 (09:49 -0700)]
don't allocate scratch tile in sp_tile_cache_flush_clear() to avoid stack overflow
Brian [Fri, 14 Dec 2007 19:25:25 +0000 (12:25 -0700)]
Added origin_lower_left field to pipe_rasterizer_state
This controls whether the window origin is considered to be the lower-left
or upper-left corner.
This effects computation of gl_FragCoord and the application of polygon stipple.
Brian [Fri, 14 Dec 2007 18:43:15 +0000 (11:43 -0700)]
update_samplers() didn't respect the sampler->texunit mapping.
This fixes the glsl/texdemo1.c program.
Brian [Fri, 14 Dec 2007 18:42:28 +0000 (11:42 -0700)]
set SamplerUnit[] entry in load_texture() just to be safe
Brian [Fri, 14 Dec 2007 18:17:48 +0000 (11:17 -0700)]
emit tgsi declarations for texture samplers
Brian [Fri, 14 Dec 2007 18:16:49 +0000 (11:16 -0700)]
set program->SamplersUsed bit when using a texture instruction
Brian [Fri, 14 Dec 2007 18:00:46 +0000 (11:00 -0700)]
Don't always declare frag shader INPUT[0] as fragment position.
We were doing this for the sake of softpipe and the tgsi intergrepter since
we always need the fragment position and W-coordinate information in order
to compute fragment interpolants.
But that's not appropriate for hardware drivers.
The tgsi interpreter now get x,y,w information from a separate tgsi_exec_vector
variable setup by softpipe.
The new pipe_shader_state->input_map[] defines how vert shader outputs map
to frag shader inputs. It may go away though, since one can also examine
the semantic label on frag shader input[0] to figure things out.
Brian [Fri, 14 Dec 2007 17:46:29 +0000 (10:46 -0700)]
update comment for TGSI_TOKEN_TYPE_IMMEDIATE case
Brian [Fri, 14 Dec 2007 01:26:48 +0000 (18:26 -0700)]
add missing code for PIPE_FORMAT_S8_UNORM renderbuffer
Brian [Fri, 14 Dec 2007 01:26:10 +0000 (18:26 -0700)]
remove unused var
Keith Whitwell [Thu, 13 Dec 2007 21:00:10 +0000 (21:00 +0000)]
965: get brw_wm_surface_state compiling again
Keith Whitwell [Thu, 13 Dec 2007 20:38:56 +0000 (20:38 +0000)]
965: get fragment shader compiler compiling
Don't think that it will run though.
Zack Rusin [Thu, 13 Dec 2007 16:42:30 +0000 (11:42 -0500)]
i965: make the wm_surface compile
Keith Whitwell [Thu, 13 Dec 2007 16:39:49 +0000 (16:39 +0000)]
965: implement magic buffer offset callback
Keith Whitwell [Thu, 13 Dec 2007 16:35:27 +0000 (16:35 +0000)]
965: magic callback to get buffer offset
Brian [Thu, 13 Dec 2007 16:05:15 +0000 (09:05 -0700)]
Add QuadX,Y fields to tgsi_exec_machine to pass quad's position to tgsi interpreter.
The tgsi linear/perspective attribute code now uses these values rather than
input[0]. Need to update SSE path to take quad x,y as function params.
Then, we can remove additional code.
Keith Whitwell [Thu, 13 Dec 2007 15:39:34 +0000 (15:39 +0000)]
965: missing code from last commit
Keith Whitwell [Thu, 13 Dec 2007 14:52:31 +0000 (14:52 +0000)]
965: remove dead code
Keith Whitwell [Thu, 13 Dec 2007 11:12:44 +0000 (11:12 +0000)]
965: remove dead code
Keith Whitwell [Thu, 13 Dec 2007 10:57:02 +0000 (10:57 +0000)]
965: remove dead code
Keith Whitwell [Thu, 13 Dec 2007 10:56:09 +0000 (10:56 +0000)]
965: remove dead code
Keith Whitwell [Thu, 13 Dec 2007 10:55:47 +0000 (10:55 +0000)]
965: remove reference to metaops
Keith Whitwell [Thu, 13 Dec 2007 10:55:16 +0000 (10:55 +0000)]
965: use state atom array directly
Keith Whitwell [Thu, 13 Dec 2007 10:54:31 +0000 (10:54 +0000)]
965: calculate nr enabled vbos on the fly
Zack Rusin [Thu, 13 Dec 2007 15:32:36 +0000 (10:32 -0500)]
Actually allocate the output registers.
In tgsi the output's will (hopefully) consecutive so this
check (besides not being currently implemented) is not
necessary.
Zack Rusin [Thu, 13 Dec 2007 15:31:39 +0000 (10:31 -0500)]
Don't generate spurious vs, move the enable a bit to do that.
Brian [Wed, 12 Dec 2007 23:30:49 +0000 (16:30 -0700)]
added r5g5b5_put_tile_rgba()
Brian [Wed, 12 Dec 2007 23:05:38 +0000 (16:05 -0700)]
Remove get/put_tile_rgba() functions.
Temporarily, use the functions from sp_rgba_tile.c
Brian [Wed, 12 Dec 2007 23:04:52 +0000 (16:04 -0700)]
remove #include
Brian [Wed, 12 Dec 2007 23:00:11 +0000 (16:00 -0700)]
Move float/rgba tile get/put functions into new file.
This should be moved into a gallium util lib (location tbd) so it can also
be used by the state tracker.
Then, remove the pipe->get/put_tile_rgba() functions.
Brian [Wed, 12 Dec 2007 22:24:01 +0000 (15:24 -0700)]
PIPE_SCREEN_SURFACE no longer needed
Brian [Wed, 12 Dec 2007 21:55:57 +0000 (14:55 -0700)]
Re-org of st_create_framebuffer() and renderbuffer format selection.
st_create_framebuffer() now takes pipe_formats for the color, depth, stencil
buffers.
This avoids a round-about chain of calls to pipe->is_format_supported() for
window renderbuffers (their format never changes).
Renderbuffer format selection code in st_format.c is simpler now too.
Brian [Wed, 12 Dec 2007 21:45:22 +0000 (14:45 -0700)]
move some code to avoid deref of NULL ptr
Keith Whitwell [Wed, 12 Dec 2007 20:07:35 +0000 (20:07 +0000)]
965: remove complex compiler for simple fragment programs.
Keep the simple compiler for complex fragment programs...
Keith Whitwell [Wed, 12 Dec 2007 19:53:41 +0000 (19:53 +0000)]
965: sketch changes to brw_wm_surface_state.c
Keith Whitwell [Wed, 12 Dec 2007 18:44:03 +0000 (18:44 +0000)]
965: delete dead files
Keith Whitwell [Wed, 12 Dec 2007 17:24:10 +0000 (17:24 +0000)]
965: delete dead files