Brian Paul [Mon, 27 Apr 2009 16:45:41 +0000 (10:45 -0600)]
i965: only upload constant buffer data when we actually need the const buffer
Make the use_const_buffer field per-program and only call the code which
updates the constant buffer's data if the flag is set.
This should undo the perf regression from
20f3497e4b6756e330f7b3f54e8acaa1d6c92052
(cherry picked from master, commit
dc9705d12d162ba6d087eb762e315de9f97bc456)
Brian Paul [Fri, 24 Apr 2009 22:33:46 +0000 (16:33 -0600)]
i965: rework GLSL/WM register allocation
Use a bitvector of used/free flags.
If we run out of temps, examine the live intervals of the temp regs in
the program and free those which are no longer alive.
Also, enable the new WM const buffer code.
Brian Paul [Fri, 24 Apr 2009 22:28:36 +0000 (16:28 -0600)]
mesa: refactor code and make _mesa_find_temp_intervals() public
Brian Paul [Fri, 24 Apr 2009 15:43:44 +0000 (09:43 -0600)]
mesa: signal _NEW_PROGRAM_CONSTANTS instead of _NEW_PROGRAM
Use _NEW_PROGRAM_CONSTANTS when changing constant/uniform buffer values.
Binding a new program/shader sets both _NEW_PROGRAM and _NEW_PROGRAM_CONSTANTS.
Brian Paul [Wed, 22 Apr 2009 19:46:58 +0000 (13:46 -0600)]
i965: disable debug printf
Brian Paul [Wed, 22 Apr 2009 17:52:16 +0000 (11:52 -0600)]
i965: enable VS constant buffers
In the VS constants can now be handled in two different ways:
1. If there's room in the GRF, put constants there. They're preloaded from
the CURBE prior to VS execution. This is the historical approach. The
problem is the GRF may not have room for all the shader's constants and
temps and misc registers. Hence...
2. Use a separate constant buffer which is read from using a READ message.
This allows a very large number of constants and frees up GRF regs for
shader temporaries. This is the new approach. May be a little slower
than 1.
1 vs. 2 is chosen according to how many constants and temps the shader needs.
Brian Paul [Wed, 22 Apr 2009 17:47:59 +0000 (11:47 -0600)]
i965: define BRW_MAX_GRF
Brian Paul [Wed, 22 Apr 2009 17:29:49 +0000 (11:29 -0600)]
i965: remove old code to init surface-related cache IDs
These types are only found in the new surface state cache now.
Brian Paul [Wed, 22 Apr 2009 17:28:24 +0000 (11:28 -0600)]
i965: comments, reformatting
Brian Paul [Wed, 22 Apr 2009 17:12:07 +0000 (11:12 -0600)]
i965: actually use the new, second surface state cache
Brian Paul [Wed, 22 Apr 2009 17:08:46 +0000 (11:08 -0600)]
i965: checkpoint commit: use two state caches instead of one
The new, second cache will only be used for surface-related items.
Since we can create many surfaces the original, single cache could get
filled quickly. When we cleared it, we had to regenerate shaders, etc.
With two caches, we can avoid doing that.
Brian Paul [Wed, 22 Apr 2009 16:44:36 +0000 (10:44 -0600)]
i965: remove unused state atom entries
Brian Paul [Wed, 22 Apr 2009 15:23:15 +0000 (09:23 -0600)]
i965: the brw_constant_buffer state atom is no longer dynamic
No more dynamic atoms so we can simplify the state validation code a little.
Brian Paul [Wed, 22 Apr 2009 15:16:21 +0000 (09:16 -0600)]
i965: add _NEW_PROGRAM_CONSTANTS to mesa_bits[] list
Brian Paul [Wed, 22 Apr 2009 15:14:53 +0000 (09:14 -0600)]
i915: check the new _NEW_PROGRAM_CONSTANT flag
Brian Paul [Wed, 22 Apr 2009 15:14:16 +0000 (09:14 -0600)]
i965: use _NEW_PROGRAM_CONSTANTS and always create new const buffers
When program constants change we create a new VS constant buffer
instead of re-using the old one. This allows us to have several
const buffers in flight with vertex rendering.
Brian Paul [Wed, 22 Apr 2009 15:07:46 +0000 (09:07 -0600)]
i965: updates to some debug code
Brian Paul [Wed, 22 Apr 2009 15:07:02 +0000 (09:07 -0600)]
i965: use new _NEW_PROGRAM_CONSTANTS flag instead of dynamic flags
Brian Paul [Wed, 22 Apr 2009 14:59:47 +0000 (08:59 -0600)]
r200/r300/r500: add _NEW_PROGRAM_CONSTANTS flag
Make sure we detect constant buffer changes indicated by the new flag.
Should be able to remove _NEW_PROGRAM (and _NEW_MODELVIEW, _NEW_LIGHT, etc)
from several places (someday.
Brian Paul [Wed, 22 Apr 2009 13:58:25 +0000 (07:58 -0600)]
st: also check _NEW_PROGRAM flag for vertex shader constant buffers
This is a follow-on to commit
c1a3b852807fb160f0cd246c1364b7336b4b947e.
Note that (at this time) wherever _NEW_PROGRAM_CONSTANTS is set we're still
setting _NEW_PROGRAM so this won't really make any difference (for now).
Thomas Hellstrom [Mon, 20 Apr 2009 18:43:56 +0000 (20:43 +0200)]
gallium: Reinstate unconditional flushes.
Lost in commit
e50dd26ca6d0eb0d0f97c2780020ea16e3d4a687.
Signed-off-by: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
Keith Whitwell [Tue, 21 Apr 2009 15:55:57 +0000 (16:55 +0100)]
mesa: protect driver.flush() with FLUSH_CURRENT
Need to do this to ensure vbo code unmaps its buffers before calling
the driver, which may be sitting on top of a memory manager which
objects to firing commands from a mapped buffer.
Brian Paul [Tue, 21 Apr 2009 23:00:54 +0000 (17:00 -0600)]
st: play it safe for now and check _NEW_PROGRAM for shader const buffer atom
When a new program is bound but no constants are updated we still need
to update the Gallium const buffer.
Brian Paul [Tue, 21 Apr 2009 23:00:01 +0000 (17:00 -0600)]
swrast: simplify state update logic for fragment shader const buffers
Brian Paul [Tue, 21 Apr 2009 22:50:34 +0000 (16:50 -0600)]
st: use the static atoms[] array directly
We can simplify this now that we no longer have any dynamic atoms.
Brian Paul [Tue, 21 Apr 2009 22:47:30 +0000 (16:47 -0600)]
st: do away with dynamic state atom for const buffers
Just use the new _NEW_PROGRAM_CONSTANTS flag instead.
Brian Paul [Tue, 21 Apr 2009 21:56:45 +0000 (15:56 -0600)]
mesa: new _NEW_PROGRAM_CONSTANTS flag
This state flag will be used to indicate that vertex/fragment program
constants have changed. _NEW_PROGRAM will be used to indicate changes
to the vertex/fragment shader itself, or misc related state.
_NEW_PROGRAM_CONSTANTS is also set whenever a program parameter that's
tracking GL state has changed. For example, if the projection matrix is
in the parameter list, calling glFrustum() will cause _NEW_PROGRAM_CONSTANTS
to be set. This will let to remove the need for dynamic state atoms in
some drivers.
For now, we still set _NEW_PROGRAM in all the places we used to. We'll no
longer set _NEW_PROGRAM in glUniform() after drivers/etc have been updated.
Brian Paul [Tue, 21 Apr 2009 20:47:57 +0000 (14:47 -0600)]
mesa: print internal.current[i] attrib
Brian Paul [Tue, 21 Apr 2009 20:44:03 +0000 (14:44 -0600)]
mesa: print parameter list dirty state flag mask
Brian Paul [Mon, 20 Apr 2009 17:58:09 +0000 (11:58 -0600)]
i965: const correctness
Michel Dänzer [Tue, 21 Apr 2009 17:32:42 +0000 (18:32 +0100)]
Update GALLIUM_AUXILIARY_DIRS in configure.ac to match configs/default.
Maciej Cencora [Mon, 20 Apr 2009 23:55:06 +0000 (01:55 +0200)]
r300: r300 hw doesn't support any input modifiers in tex insts
Mathias Gottschlag [Tue, 21 Apr 2009 15:52:30 +0000 (09:52 -0600)]
r300-gallium: Fix CS size mismatch
This fixes some warnings which appear because the driver assumes a wrong
cs size (13 vs 16 register writes in some cases).
Brian Paul [Tue, 21 Apr 2009 13:27:12 +0000 (07:27 -0600)]
demos: check that GL version is 2.0 or higher
Brian Paul [Mon, 20 Apr 2009 21:03:38 +0000 (15:03 -0600)]
st: report GL_OUT_OF_MEMORY instead of asserting
Keith Whitwell [Tue, 21 Apr 2009 10:40:59 +0000 (11:40 +0100)]
trivial/tri-viewport: add keys for frustrum/ortho and z coordinate
Keith Whitwell [Tue, 21 Apr 2009 09:59:54 +0000 (10:59 +0100)]
trivial/tri-viewport: add more out-of-bounds background quads
Keith Whitwell [Mon, 20 Apr 2009 16:32:15 +0000 (17:32 +0100)]
trivial/tri_viewport: add space==reset key
Keith Whitwell [Mon, 20 Apr 2009 16:30:53 +0000 (17:30 +0100)]
trivial/tri_viewport: add width/height keys
Brian Paul [Mon, 20 Apr 2009 20:57:00 +0000 (14:57 -0600)]
softpipe: fix softpipe_is_buffer/texture_referenced() regression
Return the conservative PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE
value for now.
This fixes a bunch of regressions seen in piglit and glean.
Brian Paul [Mon, 20 Apr 2009 20:14:26 +0000 (14:14 -0600)]
swrast: fix pointer arithmetic error in get_texel_array()
This came from commit
1b2ab023673261b4b942e1126c0b599d02fbd4a0
José Fonseca [Mon, 20 Apr 2009 15:44:15 +0000 (16:44 +0100)]
gdi: Don't implement broken gl_dispatch_stub_xxx.
José Fonseca [Mon, 20 Apr 2009 15:43:44 +0000 (16:43 +0100)]
wgl: Don't implement broken gl_dispatch_stub_xxx.
These don't respect the stdcall, so they crash upon return.
José Fonseca [Mon, 20 Apr 2009 15:40:12 +0000 (16:40 +0100)]
mesa: Correct the gl_dispatch_stub_xxx prototypes.
José Fonseca [Mon, 20 Apr 2009 13:53:08 +0000 (14:53 +0100)]
mesa: Handle failure to create a transfer.
Keith Whitwell [Mon, 20 Apr 2009 15:17:50 +0000 (16:17 +0100)]
mesa/progs: fix scons build after recent demo moves
Keith Whitwell [Mon, 20 Apr 2009 14:50:44 +0000 (15:50 +0100)]
trivial/tri-viewport.c - add guide lines, more triangles, make interactive
This is becoming more like a test than a trivial/ example.
Keith Whitwell [Fri, 17 Apr 2009 16:11:09 +0000 (17:11 +0100)]
st: assert on pipe_buffer_create failure
This needs a proper fix to propogate the out-of-memory condition back
up to Mesa and the app as a GL error. Until then, at least catch the
problem at its source.
Keith Whitwell [Fri, 17 Apr 2009 14:54:57 +0000 (15:54 +0100)]
tests/mipmap_view: add linear/nearest key
Keith Whitwell [Fri, 17 Apr 2009 10:29:39 +0000 (11:29 +0100)]
util: don't set unused blend state
Try to avoid creating multiple blend atoms.
Jakob Bornecrantz [Mon, 20 Apr 2009 14:05:01 +0000 (16:05 +0200)]
trace: Add initializer for static variables
Maciej Cencora [Mon, 20 Apr 2009 13:31:44 +0000 (07:31 -0600)]
r300: fix register-negate branch merge regression
Michel Dänzer [Mon, 20 Apr 2009 10:52:56 +0000 (12:52 +0200)]
gallium: Fix glDraw/CopyPixels fragment program leak.
Jakob Bornecrantz [Sun, 19 Apr 2009 15:09:10 +0000 (16:09 +0100)]
trace: Enable dumping to be turned on and off
Jakob Bornecrantz [Sat, 18 Apr 2009 22:14:58 +0000 (23:14 +0100)]
st: Wait to create bitmap transfer until needed
Jakob Bornecrantz [Sat, 18 Apr 2009 22:13:56 +0000 (23:13 +0100)]
autoconf: Build gallium softpipe when driver=xlib
Jakob Bornecrantz [Sun, 19 Apr 2009 15:22:43 +0000 (16:22 +0100)]
progs/demos: Update ignore
Jakob Bornecrantz [Sat, 18 Apr 2009 22:16:54 +0000 (23:16 +0100)]
progs/glsl: Update ignore
José Fonseca [Sat, 18 Apr 2009 22:14:42 +0000 (23:14 +0100)]
softpipe: Simplify softpipe_create's prototype.
José Fonseca [Sat, 18 Apr 2009 22:06:11 +0000 (23:06 +0100)]
softpipe: Fix softpipe_is_texture_referenced prototype.
José Fonseca [Sat, 18 Apr 2009 20:22:32 +0000 (21:22 +0100)]
softpipe: Remove softpipe_winsys.
Not used by softpipe anyway.
Brian Paul [Sat, 18 Apr 2009 20:18:59 +0000 (14:18 -0600)]
demos: added glsl/texaaline.c program and overhaul the Makefile
Brian Paul [Sat, 18 Apr 2009 19:18:44 +0000 (13:18 -0600)]
demos: updated .gitignore list
Brian Paul [Sat, 18 Apr 2009 19:16:35 +0000 (13:16 -0600)]
docs: removed/added demo programs
Brian Paul [Sat, 18 Apr 2009 19:12:50 +0000 (13:12 -0600)]
demos: move demos/texobj.c to tests/
Brian Paul [Sat, 18 Apr 2009 19:10:51 +0000 (13:10 -0600)]
demos: move demos/occlude.c (old HP extension) to tests
Brian Paul [Sat, 18 Apr 2009 19:08:48 +0000 (13:08 -0600)]
demos: move tests/dinoshade.c to demos/
Brian Paul [Sat, 18 Apr 2009 19:05:51 +0000 (13:05 -0600)]
demos: move tests/projtex.c to demos/
And fix compiler warnings.
Brian Paul [Sat, 18 Apr 2009 19:00:48 +0000 (13:00 -0600)]
demos: move tests/fbotexture.c to demos/
Brian Paul [Sat, 18 Apr 2009 18:58:00 +0000 (12:58 -0600)]
demos: move texdown.c to tests/
Brian Paul [Sat, 18 Apr 2009 18:57:13 +0000 (12:57 -0600)]
demos: move glutfx demo to tests/
Brian Paul [Sat, 18 Apr 2009 18:55:55 +0000 (12:55 -0600)]
demos: move streaming_rect.c demo to tests/
Brian Paul [Sat, 18 Apr 2009 18:54:27 +0000 (12:54 -0600)]
demos: move glslnoise.c demo to glsl/noise2.c
Brian Paul [Sat, 18 Apr 2009 16:42:15 +0000 (10:42 -0600)]
i965: use region width, height in brw_update_renderbuffer_surface()
Fixes a regression from commit
2c30fd84dfa052949a117c78d932b58c1f88b446
seen with DRI1.
Brian Paul [Sat, 18 Apr 2009 16:40:21 +0000 (10:40 -0600)]
intel: #include polygon.h to silence warning
Brian Paul [Sat, 18 Apr 2009 16:20:26 +0000 (10:20 -0600)]
demos: fix incorrect assertion
Brian Paul [Sat, 18 Apr 2009 16:08:54 +0000 (10:08 -0600)]
mesa: add switch case for GL_VERTEX_STATE_PROGRAM_NV in _mesa_new_program()
Fixes bug seen in progs/tests/vptest1.c
Brian Paul [Sat, 18 Apr 2009 16:08:15 +0000 (10:08 -0600)]
demos: fix usage text
Michel Dänzer [Sat, 18 Apr 2009 13:47:14 +0000 (15:47 +0200)]
intel: Handle ARB_vertex_buffer_object state in intel_clear_tris().
Fixes gearsvbo app by Michael Clark.
Brian Paul [Fri, 17 Apr 2009 22:23:33 +0000 (16:23 -0600)]
demos: new glsl/array.c demo
Test variable indexing into a uniform array in a vertex shader.
Brian Paul [Fri, 17 Apr 2009 22:11:05 +0000 (16:11 -0600)]
intel: make sure polygon mode is set properly in intel_clear_tris()
Fixes progs/glsl/skinning.c demo.
Brian Paul [Fri, 17 Apr 2009 22:04:41 +0000 (16:04 -0600)]
mesa: suppress extra newline
Brian Paul [Fri, 17 Apr 2009 22:01:20 +0000 (16:01 -0600)]
i915: fix broken indirect constant buffer reads
The READ message's msg_control value can be 0 or 1 to indicate that the
Oword should be read into the lower or upper half of the target register.
It seems that the other half of the register gets clobbered though. So
we read into two dest registers then use a MOV to combine the upper/lower
halves.
Brian Paul [Fri, 17 Apr 2009 19:55:58 +0000 (13:55 -0600)]
mesa: build a float[4] value in _mesa_add_sampler() to avoid random values
Brian Paul [Thu, 16 Apr 2009 22:41:07 +0000 (16:41 -0600)]
dri: build DRI swrast driver too
Brian Paul [Thu, 16 Apr 2009 22:37:53 +0000 (16:37 -0600)]
dri: __driUtilMessage(): not all messages are errors
Brian Paul [Thu, 16 Apr 2009 22:29:45 +0000 (16:29 -0600)]
mesa: minor tweak to error message
Brian Paul [Thu, 16 Apr 2009 21:43:32 +0000 (15:43 -0600)]
i965: updated CURBE allocation code
Now that we have real constant buffers, the demands on the CURBE are lessened.
When we use real VS/WM constant buffers we only use the CURBE for clip planes.
Jakob Bornecrantz [Fri, 17 Apr 2009 19:23:14 +0000 (21:23 +0200)]
xlib/trace: Fixup xlib trace
José Fonseca [Fri, 17 Apr 2009 17:40:46 +0000 (18:40 +0100)]
pipe: Get the p_atomic_dec_zero logic right this time.
Brian Paul [Fri, 17 Apr 2009 15:15:58 +0000 (09:15 -0600)]
demos: move glewInit() call, fixes crash/bug 21247
Michel Dänzer [Fri, 17 Apr 2009 15:02:34 +0000 (17:02 +0200)]
gallium: Fix PIPE_ATOMIC_GCC_INTRINSIC build.
Brian Paul [Fri, 17 Apr 2009 14:09:17 +0000 (08:09 -0600)]
demos: move glewInit() after glutCreateWindow()
Fixes segfault. See bug 21239. However, the demo doesn't render
properly. Probably a bug in the GL_ATI_fragment_shader code.
Jakob Bornecrantz [Fri, 17 Apr 2009 13:55:51 +0000 (15:55 +0200)]
trace: Keep screen objects on lists
Jakob Bornecrantz [Fri, 17 Apr 2009 14:00:25 +0000 (16:00 +0200)]
trace: Fix is_referenced functions
Jakob Bornecrantz [Fri, 17 Apr 2009 13:52:05 +0000 (15:52 +0200)]
trace: Simplify trace_buffer function
José Fonseca [Fri, 17 Apr 2009 14:01:38 +0000 (15:01 +0100)]
pipe: Fix InterlockedDecrement usage.
José Fonseca [Fri, 17 Apr 2009 13:55:49 +0000 (14:55 +0100)]
wgl: Put commas in the right places.
José Fonseca [Fri, 17 Apr 2009 13:53:10 +0000 (14:53 +0100)]
wgl: Flip the pixel format order.
List 32bit pixel formats first, as nobody is interested in 16bit pixel
formats nowadays.
Thomas Hellstrom [Fri, 17 Apr 2009 09:47:30 +0000 (11:47 +0200)]
gallium: Create OGL state tracker wrappers for various CPU access operations.
There are two usage types of buffer CPU accesses:
One where we try to use the buffer contents for multiple draw commands in
a batch. (batch := sequence of commands that are flushed together),
like incrementally adding bitmaps to a bitmap texture that is reallocated
on flush.
And one where we assume we can safely overwrite the old buffer contexts, like
glTexSubImage. In this case we need to make sure all old drawing commands
referencing the buffer are flushed before we map the buffer.
This is easily forgotten.
Add wrappers for the most common of these operations. The first type is
prefixed with "st_no_flush" and the second type is prefixed with
"st_cond_flush", where "cond" indicates that we attmpt to only flush
if there is indeed unflushed draw commands referencing the buffer.
Prefixed functions are
screen::get_tex_transfer
pipe_buffer_write
pipe_buffer_read
pipe_buffer_map
Please use the wrappers whenever possible.
Signed-off-by: Thomas Hellstrom <thellstrom-at-vmware-dot-com>