mesa.git
14 years agonvfx: support indirect addressing in vps
Luca Barbieri [Fri, 3 Sep 2010 16:31:18 +0000 (18:31 +0200)]
nvfx: support indirect addressing in vps

Negative or huge offsets not yet supported.

14 years agor600c: add proper returns for some evergreen functions
Alex Deucher [Fri, 3 Sep 2010 16:13:47 +0000 (12:13 -0400)]
r600c: add proper returns for some evergreen functions

these weren't checked anyway.

Fixes:
https://bugs.freedesktop.org/show_bug.cgi?id=29999

14 years agonvfx: fix support for more than 8 texture units (fixes etqw crash)
Luca Barbieri [Fri, 3 Sep 2010 13:44:27 +0000 (15:44 +0200)]
nvfx: fix support for more than 8 texture units (fixes etqw crash)

14 years agor600g: fix segfault in state after rework
Dave Airlie [Fri, 3 Sep 2010 09:37:52 +0000 (19:37 +1000)]
r600g: fix segfault in state after rework

probably can improve this a bit.

14 years agor600c: emit DB_HTILE_DATA_BASE on evergreen
Alex Deucher [Fri, 3 Sep 2010 05:13:41 +0000 (01:13 -0400)]
r600c: emit DB_HTILE_DATA_BASE on evergreen

Make the hw happy.

14 years agor600g: refactor sample states into a reusable struct.
Dave Airlie [Fri, 3 Sep 2010 04:12:38 +0000 (14:12 +1000)]
r600g: refactor sample states into a reusable struct.

I will not cut-n-paste.
I will not cut-n-paste.
I will not cut-n-paste.

14 years agor600g: reduce size of r600 context structure to !insane
Dave Airlie [Fri, 3 Sep 2010 03:53:39 +0000 (13:53 +1000)]
r600g: reduce size of r600 context structure to !insane

Its now about 7.8k, and might actually fit in a cache.

14 years agor600g: add texture border state.
Dave Airlie [Fri, 3 Sep 2010 03:07:40 +0000 (13:07 +1000)]
r600g: add texture border state.

Okay I finally wrapped my head around what r600_context_state is meant to be,
maybe I should just rename all the structs so that have distinct names.

I've no idea however why 16 is a good magic number for R600_MAX_RSTATE.

14 years agor600g: deref old driver states for set entry points.
Dave Airlie [Fri, 3 Sep 2010 02:01:59 +0000 (12:01 +1000)]
r600g: deref old driver states for set entry points.

14 years agor600g: drop r600_bind_state.
Dave Airlie [Fri, 3 Sep 2010 01:55:36 +0000 (11:55 +1000)]
r600g: drop r600_bind_state.

This was another ugly function that really wasn't needed.

The 3 calls to it from the gallium api were shorter than it,
and all the calls from the set_ functions were pointless.

14 years agor600g: kill r600_context_state function
Dave Airlie [Fri, 3 Sep 2010 01:35:08 +0000 (11:35 +1000)]
r600g: kill r600_context_state function

having some sort of locality of code really matters, just create
and setup state at time. Not sure if this is just further polishing of a bad thing,
but at least it makes it more readable.

14 years agor600g: move lots of state inline helpers to separate header.
Dave Airlie [Fri, 3 Sep 2010 00:55:02 +0000 (10:55 +1000)]
r600g: move lots of state inline helpers to separate header.

this gets them out of sight of the main codeflow.

14 years agodraw: Include missing headers in draw_vs_aos.h.
Vinson Lee [Fri, 3 Sep 2010 00:00:53 +0000 (17:00 -0700)]
draw: Include missing headers in draw_vs_aos.h.

Include tgsi_exec.h for TGSI_EXEC_NUM_TEMPS.
Include draw_vs.h for draw_vs_varient.

14 years agor600g: drop magic numbers in depth state.
Dave Airlie [Thu, 2 Sep 2010 23:39:04 +0000 (09:39 +1000)]
r600g: drop magic numbers in depth state.

this also fixes occulsion queries.

14 years agoutil: Include missing header in u_linear.h.
Vinson Lee [Thu, 2 Sep 2010 23:30:34 +0000 (16:30 -0700)]
util: Include missing header in u_linear.h.

Include p_compiler.h for size_t and boolean symbols.

14 years agomesa: Fix printf-like warning.
Vinson Lee [Thu, 2 Sep 2010 23:12:58 +0000 (16:12 -0700)]
mesa: Fix printf-like warning.

14 years agomesa: Fix printf-like warnings.
Vinson Lee [Thu, 2 Sep 2010 23:03:32 +0000 (16:03 -0700)]
mesa: Fix printf-like warnings.

14 years agor600g: force unbind of previously bind sampler/sampler_view
Jerome Glisse [Thu, 2 Sep 2010 20:53:33 +0000 (16:53 -0400)]
r600g: force unbind of previously bind sampler/sampler_view

Previously bind sampler/sampler_view can be converted and endup
overwritting the current state we want to schedule. Example :
bind texA texB to sampler_view[0] & sampler_view[1], render,
bind texB to sampler_view[0] render. Now state associated to
texB are set to configure sampler_view slot 0, but as we don't
unbind sampler_view[1] still point to texB state so we end up
with sampler_view[1] overwritting sampler_view[0], which gives
wrong rendering if next rendering bind texA to sampler_view[0],
it will endup as texB is bound to sampler_view[0]. If you are
not confuse at that point give me a call i will be buying you
beer.

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
14 years agost/mesa: clamp gl_constants::MaxVarying against MAX_VARYING
Brian Paul [Thu, 2 Sep 2010 18:58:58 +0000 (12:58 -0600)]
st/mesa: clamp gl_constants::MaxVarying against MAX_VARYING

Don't try to use more generic varying vars than core Mesa supports.
Fixes fd.o bug 29959.

14 years agoglsl2: Update TODO file
Ian Romanick [Thu, 2 Sep 2010 17:11:54 +0000 (10:11 -0700)]
glsl2: Update TODO file

14 years agor600g: fix memory/bo leak
Jerome Glisse [Thu, 2 Sep 2010 15:32:32 +0000 (11:32 -0400)]
r600g: fix memory/bo leak

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
14 years agomesa: fix code generation for ir_unop_sqrt
Brian Paul [Thu, 2 Sep 2010 13:58:28 +0000 (07:58 -0600)]
mesa: fix code generation for ir_unop_sqrt

The CMP instruction needed to be flipped to properly handle
operand==0.

Fixes fd.o bug 29923.

14 years agomesa: fix some printf warnings with casts
Brian Paul [Thu, 2 Sep 2010 13:57:16 +0000 (07:57 -0600)]
mesa: fix some printf warnings with casts

14 years agor600g: fix thinko in shadow code.
Dave Airlie [Thu, 2 Sep 2010 11:25:42 +0000 (21:25 +1000)]
r600g: fix thinko in shadow code.

spotted by taiu on irc

14 years agor600g: fix logicop, the 3d ROP is the 2D rop shifted twice.
Dave Airlie [Thu, 2 Sep 2010 06:39:32 +0000 (16:39 +1000)]
r600g: fix logicop, the 3d ROP is the 2D rop shifted twice.

14 years agor600g: fix depth texture tests
Dave Airlie [Thu, 2 Sep 2010 05:51:23 +0000 (15:51 +1000)]
r600g: fix depth texture tests

14 years agoglsl: Apply implicit conversions to structure constructor parameters.
Kenneth Graunke [Thu, 2 Sep 2010 03:03:17 +0000 (20:03 -0700)]
glsl: Apply implicit conversions to structure constructor parameters.

The code for handling implicit conversions should probably get
refactored, but for now, this is easy.

Fixes piglit test constructor-26.vert.

14 years agoglsl: Convert constant record constructor parameters to ir_constants.
Kenneth Graunke [Thu, 2 Sep 2010 02:54:27 +0000 (19:54 -0700)]
glsl: Convert constant record constructor parameters to ir_constants.

I'm not sure if this is strictly necessary, but it seems wise.

14 years agoglsl: Reject structure constructors that have too many arguments.
Kenneth Graunke [Thu, 2 Sep 2010 02:44:25 +0000 (19:44 -0700)]
glsl: Reject structure constructors that have too many arguments.

Fixes piglit test constructor-27.vert.

14 years agoglsl2: Remove unnecessary glsl_symbol_table::get_function parameter return_constructors
Ian Romanick [Wed, 1 Sep 2010 21:16:53 +0000 (14:16 -0700)]
glsl2: Remove unnecessary glsl_symbol_table::get_function parameter return_constructors

Now that constructors are not generated as functions or stored in the
symbol table, there is no need to flag whether or not constructors
should be returned.

14 years agoglsl2: Remove unused method glsl_type::generate_constructor
Ian Romanick [Wed, 1 Sep 2010 21:12:24 +0000 (14:12 -0700)]
glsl2: Remove unused method glsl_type::generate_constructor

14 years agoglsl2: Remove unused 'constructor' parameter from glsl_symbol_table::add_type
Ian Romanick [Wed, 1 Sep 2010 21:10:39 +0000 (14:10 -0700)]
glsl2: Remove unused 'constructor' parameter from glsl_symbol_table::add_type

14 years agoglsl2: Don't generate constructor functions for structures
Ian Romanick [Wed, 1 Sep 2010 21:08:08 +0000 (14:08 -0700)]
glsl2: Don't generate constructor functions for structures

14 years agoglsl2: Emit structure constructors inline
Ian Romanick [Wed, 1 Sep 2010 20:46:04 +0000 (13:46 -0700)]
glsl2: Emit structure constructors inline

Fixes piglit test cases glsl-[fv]s-all-0[12].

14 years agor600g: add missing vertex fetch formats to the translation table.
Dave Airlie [Thu, 2 Sep 2010 03:32:25 +0000 (13:32 +1000)]
r600g: add missing vertex fetch formats to the translation table.

fixes at least 2 more piglits.

14 years agor600g: fix binding of same texture to several target slot
Jerome Glisse [Thu, 2 Sep 2010 01:57:43 +0000 (21:57 -0400)]
r600g: fix binding of same texture to several target slot

One can bind same texture or sampler to different slot,
each slot needs it own state. The solution implemented
here is not exactly beautifull or optimal need to think
to somethings better.

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
14 years agoglsl: Fix write mask in matrix-from-matrix constructors.
Kenneth Graunke [Wed, 1 Sep 2010 23:10:01 +0000 (16:10 -0700)]
glsl: Fix write mask in matrix-from-matrix constructors.

If the matrix being constructed was larger than the source matrix, it
would overwrite the lower-right part of the matrix with the wrong
values, rather than leaving it as the identity matrix.

For example, constructing a mat4 from a mat2 should only use a writemask
of "xy" when copying from the source, but was using "xyzw".

Fixes the code generated by piglit test constructor-23.vert.

14 years agoglsl: Add proper handling for constant matrix-from-matrix constructors.
Kenneth Graunke [Wed, 1 Sep 2010 22:37:13 +0000 (15:37 -0700)]
glsl: Add proper handling for constant matrix-from-matrix constructors.

Fixes piglit test case constructor-21.vert and changes
constructor-22.vert to give the correct output.

14 years agoglsl: Move generate_constructor_(matrix|vector) to ir_constant ctor.
Kenneth Graunke [Wed, 1 Sep 2010 22:31:06 +0000 (15:31 -0700)]
glsl: Move generate_constructor_(matrix|vector) to ir_constant ctor.

14 years agoast_function: Fix check for "too few components".
Kenneth Graunke [Wed, 1 Sep 2010 22:04:57 +0000 (15:04 -0700)]
ast_function: Fix check for "too few components".

This was triggering even for matrix-from-matrix constructors.  It is
perfectly legal to construct a mat3 from a mat2 - the rest will be
filled in by the identity matrix.

Changes piglit test constructor-23.vert from FAIL to PASS, but the
generated code is incorrect.

14 years agoast_function: Remove bogus cases from generate_constructor_matrix.
Kenneth Graunke [Wed, 1 Sep 2010 21:49:53 +0000 (14:49 -0700)]
ast_function: Remove bogus cases from generate_constructor_matrix.

There are no integer matrix types, so switching on them is silly.

14 years agor600g: fix incorrect state naming in pipe_sampler vs pipe_sampler_view
Dave Airlie [Thu, 2 Sep 2010 01:16:31 +0000 (11:16 +1000)]
r600g: fix incorrect state naming in pipe_sampler vs pipe_sampler_view

fixes problems in valgrind with uninitialised values.

14 years agoir_to_mesa: Load all the STATE_VAR elements of a builtin uniform to a temp.
Eric Anholt [Wed, 1 Sep 2010 23:06:32 +0000 (16:06 -0700)]
ir_to_mesa: Load all the STATE_VAR elements of a builtin uniform to a temp.

Like the constant handling and the handling of other uniforms, we add
the whole thing to the Parameters, avoiding messy, incomplete logic
for adding just the elements of a builting uniform that get used.
This means that a driver that relies only on ParameterValues[] for its
parameters will have an increased parameter load, but drivers
generally don't do that (since they have other params they need to
handle, too).

Fixes glsl-fs-statevar-call (testcase for Ember).  Bug #29687.

Regresses glsl-vs-array-04 on 965.  Thanks to a slight change in
register allocation, this test of undefined behavior now wraps around
the register space and unexpectedly reads the constant value it's
trying to compare to.  The test should probably not look at the
resulting color, since behavior is undefined.

14 years agoir_to_mesa: Add a little helper for emitting link failure messages.
Eric Anholt [Wed, 1 Sep 2010 21:56:43 +0000 (14:56 -0700)]
ir_to_mesa: Add a little helper for emitting link failure messages.

14 years agomesa: Fix many printf-like warnings.
Eric Anholt [Wed, 1 Sep 2010 21:46:22 +0000 (14:46 -0700)]
mesa: Fix many printf-like warnings.

Most of these are just typecasting to long to match the arg type.  I
don't really care too much about getting a GLsizei or whatever
appropriate type in.  However, there were a number of real bugs, like
missing arguments or passing floats to integer format specifiers.  My
favorite: printflike("%s, argument") is missing an argument.

14 years agomesa: Add __printf__ attribute to printf-like functions to get warnings.
Eric Anholt [Wed, 1 Sep 2010 21:39:50 +0000 (14:39 -0700)]
mesa: Add __printf__ attribute to printf-like functions to get warnings.

14 years agomesa/st: remove check for buffer/elements = 0
Dave Airlie [Wed, 1 Sep 2010 23:45:37 +0000 (09:45 +1000)]
mesa/st: remove check for buffer/elements = 0

shown by the glsl-vs-point-size failing on r600g.

the test passes on softpipe and I get a full piglit test run completing on r600g.

14 years agor600g: silence compiler warning
Jerome Glisse [Wed, 1 Sep 2010 22:09:37 +0000 (18:09 -0400)]
r600g: silence compiler warning

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
14 years agor600g: refix db/cb state
Jerome Glisse [Wed, 1 Sep 2010 22:04:38 +0000 (18:04 -0400)]
r600g: refix db/cb state

Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
14 years agomesa: fix out of bounds memory read in mipmap gen code
Brian Paul [Wed, 1 Sep 2010 19:21:51 +0000 (13:21 -0600)]
mesa: fix out of bounds memory read in mipmap gen code

Out of bounds reads could happen for reducing WxH to WxH/2
or WxH to W/2xH.

Fixes fd.o bug 29918.

14 years agomesa: more prog_execute.c debug code
Brian Paul [Wed, 1 Sep 2010 18:46:57 +0000 (12:46 -0600)]
mesa: more prog_execute.c debug code

14 years agoglsl: Add forgotten implementations of equal/notEqual on bvecs.
Kenneth Graunke [Wed, 1 Sep 2010 18:47:52 +0000 (11:47 -0700)]
glsl: Add forgotten implementations of equal/notEqual on bvecs.

14 years agor600g: fix up default state differences between r6xx and r7xx
Alex Deucher [Wed, 1 Sep 2010 17:57:52 +0000 (13:57 -0400)]
r600g: fix up default state differences between r6xx and r7xx

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
14 years agost/glx: re-order destruction of buffers, visuals
Brian Paul [Wed, 1 Sep 2010 18:39:04 +0000 (12:39 -0600)]
st/glx: re-order destruction of buffers, visuals

Free the buffers before the visuals.  Fixes valgrind warning
reported in fd.o bug 29919.

14 years agoi965: DP2 produces a scalar result like DP3, DP4, etc.
Eric Anholt [Wed, 1 Sep 2010 17:41:36 +0000 (10:41 -0700)]
i965: DP2 produces a scalar result like DP3, DP4, etc.

Fixes glsl-fs-dot-vec2-2.

14 years agoglsl2: Perform algebraic simplifications on logical binary operators
Ian Romanick [Wed, 1 Sep 2010 17:13:21 +0000 (10:13 -0700)]
glsl2: Perform algebraic simplifications on logical binary operators

Reduces glsl-vs-all-01 from 42 Mesa IR instructions (including the
END) to 17.

14 years agoglsl2: Allow ir_constant::zero to create boolean constants
Ian Romanick [Wed, 1 Sep 2010 17:12:55 +0000 (10:12 -0700)]
glsl2: Allow ir_constant::zero to create boolean constants

14 years agor600g: avoid dynamic allocation of states
Jerome Glisse [Mon, 30 Aug 2010 21:56:59 +0000 (17:56 -0400)]
r600g: avoid dynamic allocation of states

Make state statically allocated, this kills a bunch of code
and avoid intensive use of malloc/free. There is still a lot
of useless duplicate function wrapping that can be kill. This
doesn't improve yet performance, needs to avoid memcpy states
in radeon_ctx_set_draw and to avoid rebuilding vs_resources,
dsa, scissor, cb_cntl, ... states at each draw command.

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
14 years agoRevert "Revert "r600g: precompute some of the hw state""
Jerome Glisse [Wed, 1 Sep 2010 17:04:42 +0000 (13:04 -0400)]
Revert "Revert "r600g: precompute some of the hw state""

This reverts commit 1fa7245c348cb7aced81f1672140f64cb6450e2f.

Conflicts:

src/gallium/drivers/r600/r600_state.c

14 years agonouveau/nvfx: Remove enforcement of bit depth being same as front buffer
Patrice Mandin [Wed, 1 Sep 2010 16:12:11 +0000 (18:12 +0200)]
nouveau/nvfx: Remove enforcement of bit depth being same as front buffer

Signed-off-by: Patrice Mandin <patmandin@gmail.com>
14 years agomesa: initialize dummy framebuffer and renderbuffer mutexes
Vladimir Vukicevic [Wed, 1 Sep 2010 14:54:21 +0000 (08:54 -0600)]
mesa: initialize dummy framebuffer and renderbuffer mutexes

See fd.o bug 29909.

Signed-off-by: Brian Paul <brianp@vmware.com>
14 years agoCygwin: Adjust mklib so -linker and -cplusplus options are processed more like they...
Jon TURNEY [Fri, 27 Aug 2010 21:52:41 +0000 (22:52 +0100)]
Cygwin: Adjust mklib so -linker and -cplusplus options are processed more like they are for linux

It looks like we were ignoring -linker when -noprefix wasn't present, and
when -noprefix was present, -linker was mandatory and -cplusplus ignored.

Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Signed-off-by: Brian Paul <brianp@vmware.com>
14 years agoCygwin: Teach mklib/minstall to properly install libraries on cygwin
Jon TURNEY [Sat, 24 Jul 2010 11:05:34 +0000 (12:05 +0100)]
Cygwin: Teach mklib/minstall to properly install libraries on cygwin

Teach mklib/minstall more about cygwin so libraries are properly installed

Have mklib install the .dll into the lib/ staging directory as well

Have minstall install the .dll into PREFIX/bin at the same time as
installing the .dll.a link library into PREFIX/lib

mklib uses a '-' rather than a '.' as the separator before the version
number in library names on cygwin.  Change the install globs so they match
library names like that.

Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Signed-off-by: Brian Paul <brianp@vmware.com>
14 years agoCygwin: Change mklib not to report the full archname on cygwin
Jon TURNEY [Sat, 24 Jul 2010 11:06:23 +0000 (12:06 +0100)]
Cygwin: Change mklib not to report the full archname on cygwin

Change mklib not to report the full archname when building a library for cygwin
(which is something like 'CYGWIN_NT-5.1' or 'CYGWIN_NT-6.1-WOW64' and kind of
confusing), but just 'CYGWIN'.

Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Signed-off-by: Brian Paul <brianp@vmware.com>
14 years agoCygwin: Have mklib exit with error code if link fails
Jon TURNEY [Fri, 19 Feb 2010 22:38:57 +0000 (22:38 +0000)]
Cygwin: Have mklib exit with error code if link fails

Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Signed-off-by: Brian Paul <brianp@vmware.com>
14 years agoglsl2: Disallow function declarations within function definitions in GLSL 1.20
Ian Romanick [Wed, 1 Sep 2010 13:34:58 +0000 (06:34 -0700)]
glsl2: Disallow function declarations within function definitions in GLSL 1.20

The GLSL 1.20 spec specifically disallows this, but it was allowed in
GLSL 1.10.

Fixes piglit test cases local-function-0[13].frag and bugzilla #29921.

14 years agor600: cube mipmap levels are aligned to 8 faces only starting from r7xx
Andre Maasikas [Wed, 1 Sep 2010 06:19:51 +0000 (09:19 +0300)]
r600: cube mipmap levels are aligned to 8 faces only starting from r7xx

14 years agor600g: correct cb/zb offset emits.
Dave Airlie [Wed, 1 Sep 2010 04:28:55 +0000 (14:28 +1000)]
r600g: correct cb/zb offset emits.

This fixes fbo-3d and fbo-cubemap

14 years agoRevert "r600g: precompute some of the hw state"
Dave Airlie [Wed, 1 Sep 2010 04:56:04 +0000 (14:56 +1000)]
Revert "r600g: precompute some of the hw state"

This reverts commit de0b76cab22caa9fc7260f80acb8f151ccced6c5, its pre-computes the texture state wrong,

you can't just use an array of levels, since you can have FBOs to depth texture slices inside a level as well
it would get really messy quickly. Probably need to split commits like this up into pieces for each piece
of state, so we can revert bits easier in case of regressions.

This also break 5 piglit tests, and valgrind starts to warn about invalid read/writes after this.

14 years agor600g: fix typo causing segfault.
Dave Airlie [Wed, 1 Sep 2010 03:54:38 +0000 (13:54 +1000)]
r600g: fix typo causing segfault.

fixes warning that
r600_blit.c: In function ‘r600_resource_copy_region’:
r600_blit.c:136: warning: passing argument 1 of ‘util_resource_copy_region’ from incompatible pointer type

and also 7 more piglit tests.

14 years agor600g: fix glean texCube and shadows.
Dave Airlie [Wed, 1 Sep 2010 01:37:39 +0000 (11:37 +1000)]
r600g: fix glean texCube and shadows.

add cube and shadow support to the texture code.

14 years agogallivm: fix bug in nested conditionals
Brian Paul [Wed, 1 Sep 2010 01:14:18 +0000 (19:14 -0600)]
gallivm: fix bug in nested conditionals

This, plus the previous commit fix fd.o bug 29806.

14 years agoglsl2: fix bug in atan(y, x) function
Brian Paul [Wed, 1 Sep 2010 01:13:09 +0000 (19:13 -0600)]
glsl2: fix bug in atan(y, x) function

When x==0, the result was wrong.  Fixes piglit glsl-fs-atan-1.shader_test

14 years agoir_to_mesa: When emitting a pixel kill, flag that we did so.
Eric Anholt [Tue, 31 Aug 2010 20:02:59 +0000 (13:02 -0700)]
ir_to_mesa: When emitting a pixel kill, flag that we did so.

Both i965 and swrast rely on UsesKill to determine whether to do early
depth writes.  Fixes glsl-fs-discard-02.

Bug #29835.

14 years agollvmpipe: slightly simplify build_mask
Keith Whitwell [Tue, 24 Aug 2010 22:05:57 +0000 (23:05 +0100)]
llvmpipe: slightly simplify build_mask

14 years agollvmpipe: combine linear mask calculation
Keith Whitwell [Sun, 22 Aug 2010 21:56:54 +0000 (22:56 +0100)]
llvmpipe: combine linear mask calculation

14 years agollvmpipe: intrinsics versions of build_mask functions
Keith Whitwell [Sun, 15 Aug 2010 17:19:52 +0000 (18:19 +0100)]
llvmpipe: intrinsics versions of build_mask functions

14 years agoglsl2: Write vector constructor constants in a single assignment
Ian Romanick [Tue, 31 Aug 2010 21:44:13 +0000 (14:44 -0700)]
glsl2: Write vector constructor constants in a single assignment

Make two passes over the constructor parameters.  Write all of the
constants in a single write, then write the non-constants one at a
time.  This causes the fragment shader

varying float g;
void main()
{
gl_FragColor = vec4(0.0, g, 0.0, 1.0);
}

to generate

(function main
  (signature void (parameters )
    (
      (declare (temporary ) vec4 vec_ctor@0x8580058)
      (assign (constant bool (1)) (xzw) (var_ref vec_ctor@0x8580058)  (constant vec4 (0.000000 0.000000 0.000000 1.000000)) )
      (assign (constant bool (1)) (y) (var_ref vec_ctor@0x8580058)  (swiz xxxx (var_ref g@0x8580218) ))
      (assign (constant bool (1)) (xyzw) (var_ref gl_FragColor@0x84d32a0)  (var_ref vec_ctor@0x8580058) )
    ))
)

instead of

(function main
  (signature void (parameters )
    (
      (declare (temporary ) vec4 vec_ctor@0x8580058)
      (assign (constant bool (1)) (x) (var_ref vec_ctor@0x8580058)  (constant vec4 (0.000000 0.000000 0.000000 1.000000)) )
      (assign (constant bool (1)) (y) (var_ref vec_ctor@0x8580058)  (swiz xxxx (var_ref g@0x8580218) ))
      (assign (constant bool (1)) (z) (var_ref vec_ctor@0x8580058)  (constant vec4 (0.000000 0.000000 0.000000 1.000000)) )
      (assign (constant bool (1)) (w) (var_ref vec_ctor@0x8580058)  (constant vec4 (0.000000 0.000000 0.000000 1.000000)) )
      (assign (constant bool (1)) (xyzw) (var_ref gl_FragColor@0x84d32a0)  (var_ref vec_ctor@0x8580058) )
    ))
)

A similar optimization could be done for matrix constructors, but it
is a little more complicate there.

14 years agoir_to_mesa: Sort the uniform list we're adding to Parameters[] order.
Eric Anholt [Tue, 31 Aug 2010 18:16:09 +0000 (11:16 -0700)]
ir_to_mesa: Sort the uniform list we're adding to Parameters[] order.

Fixes glsl-uniform-linking-1 and failure to link a shader in Unigine.
An alternative here would be to just ditch using _mesa_add_parameter
and build the initial params list on our own, but that would require
two walks of the list as well.

Bug #29822

14 years agoir_to_mesa: Fix struct/class confusion
nobled [Sun, 22 Aug 2010 05:16:53 +0000 (05:16 +0000)]
ir_to_mesa: Fix struct/class confusion

14 years agoir_to_mesa: Set up our instruction nodes with zeroed data.
Eric Anholt [Mon, 30 Aug 2010 21:42:27 +0000 (14:42 -0700)]
ir_to_mesa: Set up our instruction nodes with zeroed data.

cond_update wasn't being set by emit_op3, leading to valgrind
complaints, and failures in several piglit tests when built with
clang.

14 years agoast_to_hir: Add support for bit-wise operators (but not shifts).
Kenneth Graunke [Tue, 31 Aug 2010 17:56:24 +0000 (10:56 -0700)]
ast_to_hir: Add support for bit-wise operators (but not shifts).

Previously, using bit-wise operators in some larger expression would
crash on a NULL pointer dereference.  This code at least doesn't crash.

Fixes piglit test bitwise-01.frag.

14 years agost/egl: Enable EGL_MESA_drm_display.
Chia-I Wu [Tue, 31 Aug 2010 07:52:26 +0000 (15:52 +0800)]
st/egl: Enable EGL_MESA_drm_display.

14 years agoegl: Mark EGL_MESA_screen_surface as obsolete.
Chia-I Wu [Tue, 31 Aug 2010 07:25:09 +0000 (15:25 +0800)]
egl: Mark EGL_MESA_screen_surface as obsolete.

EGL_MESA_drm_{display,image} can achieve the same functionality.

14 years agor600g: fix up depth write swizzles.
Dave Airlie [Tue, 31 Aug 2010 06:13:03 +0000 (16:13 +1000)]
r600g: fix up depth write swizzles.

For some reason r600c, emits extra instructions in the FP to do the depth write output swizzle,
I'm not sure this is required, so here I'm doing it in the exports.

this fixes the mesa trivial demos tri-depthwrite and tri-depthwrite2, it doesn't fix
the glsl1 gl_FragDepth writing test however.

14 years agor600g: fix fp-fragment-position test.
Dave Airlie [Tue, 31 Aug 2010 05:12:24 +0000 (15:12 +1000)]
r600g: fix fp-fragment-position test.

14 years agor600g: fix typo in last commit
Dave Airlie [Tue, 31 Aug 2010 04:57:57 +0000 (14:57 +1000)]
r600g: fix typo in last commit

14 years agor600g: fix position input to fragment shader.
Dave Airlie [Tue, 31 Aug 2010 04:52:52 +0000 (14:52 +1000)]
r600g: fix position input to fragment shader.

this fixes a few if the fs shader tests, 10 more piglits

14 years agor600g: remove unneeded function call from scs
Dave Airlie [Tue, 31 Aug 2010 03:54:19 +0000 (13:54 +1000)]
r600g: remove unneeded function call from scs

14 years agoir_to_mesa: use RSQ+MUL instead of RSQ+RCP for SQRT
Marek Olšák [Tue, 31 Aug 2010 03:38:50 +0000 (05:38 +0200)]
ir_to_mesa: use RSQ+MUL instead of RSQ+RCP for SQRT

sqrt(x) = 1/rsq(x) = x*rsq(x)

This optimization already was in the old GLSL compiler.

Acked on irc by Eric Anholt.

14 years agolinker: Require an exact matching signature when looking for prototypes.
Kenneth Graunke [Tue, 31 Aug 2010 03:42:19 +0000 (20:42 -0700)]
linker: Require an exact matching signature when looking for prototypes.

Fixes piglit test glsl-override-builtin.  The linker incorrectly found
the prototype for the float signature, rather than adding a new
prototype with the int return type.  This caused ir_calls with type int
to have their callees set to the float signature, triggering an assert.

14 years agor600g: make LIT work properly
Dave Airlie [Tue, 31 Aug 2010 01:57:04 +0000 (11:57 +1000)]
r600g: make LIT work properly

this is a bit of a workaround, something is wrong with the literal emits here
so we just use the trig copy function to copy the immd to a temp at start of op.

fix VP/FP LIT tests

14 years agoi965: fix depth test on sandybridge
Zhenyu Wang [Tue, 31 Aug 2010 01:54:44 +0000 (09:54 +0800)]
i965: fix depth test on sandybridge

This includes several corrections for fixing depth test on sandybridge.
Fix wrong bits definition in depth stencil state. Fix wrong order of
state buffer offset in 3DSTATE_CC_STATE_POINTERS command. Correctly use
buffer width parameter in depth buffer setting.

Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
14 years agor600g: fixup trig functions when input is a literal
Dave Airlie [Tue, 31 Aug 2010 01:42:01 +0000 (11:42 +1000)]
r600g: fixup trig functions when input is a literal

So as the trig functions used up the literal spots for the PI work, if the arg0 was an immediate
we'd hit failure, so copy the literal before starting.

add some tracking of max temp used to avoid trashing temp regs.

5 more piglits, fp1 COS,SCS,SIN tests

14 years agor600g: make sure LIT splits constants
Dave Airlie [Tue, 31 Aug 2010 00:43:54 +0000 (10:43 +1000)]
r600g: make sure LIT splits constants

14 years agor600g: fix constant splitting
Dave Airlie [Tue, 31 Aug 2010 00:43:04 +0000 (10:43 +1000)]
r600g: fix constant splitting

constant splitting was broken for multi-constant cases, fixes fp1 CMP+MAD, vp1 CMP.

14 years agor600g: fix LIT tests
Dave Airlie [Mon, 30 Aug 2010 23:56:35 +0000 (09:56 +1000)]
r600g: fix LIT tests

14 years agor600g: add missing literals
Dave Airlie [Mon, 30 Aug 2010 23:02:02 +0000 (09:02 +1000)]
r600g: add missing literals

Also add an error if we hit this problem again, we need to do this better
possibly tying the literal addition to the last flag.

Signed-off-by: Dave Airlie <airlied@redhat.com>
14 years agolinker: Handle varying arrays, matrices, and arrays of matrices
Ian Romanick [Mon, 30 Aug 2010 22:37:44 +0000 (15:37 -0700)]
linker: Handle varying arrays, matrices, and arrays of matrices

Fixes piglit test case glsl-array-varying-01.

14 years agoDon't pass -ffast-math to clang, since it ignores it and complains.
Eric Anholt [Mon, 30 Aug 2010 21:19:23 +0000 (14:19 -0700)]
Don't pass -ffast-math to clang, since it ignores it and complains.