Vinson Lee [Sun, 21 Nov 2010 06:01:30 +0000 (22:01 -0800)]
mesa: Clean up header file inclusion in scissor.h.
Vinson Lee [Sun, 21 Nov 2010 05:32:07 +0000 (21:32 -0800)]
mesa: Clean up header file inclusion in renderbuffer.h.
Vinson Lee [Sun, 21 Nov 2010 05:23:35 +0000 (21:23 -0800)]
mesa: Clean up header file inclusion in readpix.h.
Vinson Lee [Sun, 21 Nov 2010 05:14:06 +0000 (21:14 -0800)]
mesa: Clean up header file inclusion in rastpos.h.
Vinson Lee [Sun, 21 Nov 2010 05:06:09 +0000 (21:06 -0800)]
mesa: Clean up header file inclusion in polygon.h.
Vinson Lee [Sun, 21 Nov 2010 04:13:50 +0000 (20:13 -0800)]
intel: Remove unnecessary header.
Vinson Lee [Sun, 21 Nov 2010 03:04:30 +0000 (19:04 -0800)]
r600: Remove unnecesary header.
Vinson Lee [Sun, 21 Nov 2010 03:00:18 +0000 (19:00 -0800)]
swrast: Remove unnecessary header.
Vinson Lee [Sun, 21 Nov 2010 02:48:09 +0000 (18:48 -0800)]
st/mesa: Remove unnecessary headers.
Chia-I Wu [Sun, 21 Nov 2010 01:47:11 +0000 (17:47 -0800)]
scons: Define IN_DRI_DRIVER.
The define is required for DRI drivers. It is not needed for
libgl-xlib, but the overhead it introduces should be minor.
Xavier Chantry [Sat, 20 Nov 2010 21:51:12 +0000 (22:51 +0100)]
nvfx: only expose one rt on nv30
We do not know how to use more, GL_ARB_draw_buffers is not exposed on blob.
Owen W. Taylor [Sat, 20 Nov 2010 17:18:56 +0000 (12:18 -0500)]
r600g: Fix location for clip plane registers
The stride between the different clip plane registers was incorrect.
https://bugs.freedesktop.org/show_bug.cgi?id=31788
agd5f: fix evergreen as well.
Marek Olšák [Sun, 14 Nov 2010 17:57:14 +0000 (18:57 +0100)]
r300g: fix rendering with no vertex elements
Fixes glsl-vs-point-size, although I meant to fix glsl-novertexdata.
Since swrast fails glsl-novertexdata too, I guess it's a core issue.
Eric Anholt [Fri, 19 Nov 2010 07:57:05 +0000 (15:57 +0800)]
i965: Remove duplicate MRF writes in the FS backend.
This is quite common for multitexture sampling, and not only cuts down
on the second and later set of MOVs, but typically also allows
compute-to-MRF on the first set.
No statistically siginficant performance difference in nexuiz (n=3),
but it reduces instruction count in one of its shaders and seems like
a good idea.
Eric Anholt [Thu, 18 Nov 2010 07:03:50 +0000 (15:03 +0800)]
i965: Improve compute-to-mrf.
We were skipping it if the instruction producing the value we were
going to compute-to-mrf used its result reg as a source reg. This
meant that the typical "write interpolated color to fragment color" or
"texture from interpolated texcoord" shader didn't compute-to-MRF.
Just don't check for the interference cases until after we've checked
if this is the instruction we wanted to compute-to-MRF.
Improves nexuiz high-settings performance on my laptop 0.48% +- 0.08%
(n=3).
Eric Anholt [Fri, 19 Nov 2010 10:50:05 +0000 (18:50 +0800)]
ir_to_mesa: Detect and emit MOV_SATs for saturate constructs.
The goal here is to avoid regressing performance on ir_to_mesa drivers
for fixed function fragment shaders requiring saturates.
Eric Anholt [Fri, 19 Nov 2010 02:36:06 +0000 (10:36 +0800)]
i965: Recognize saturates and turn them into a saturated mov.
On pre-gen6, this turns 4 instructions into 1. We could still do
better by folding the saturate into the instruction generating the
value if nobody else uses it, but that should be a separate pass.
Eric Anholt [Fri, 19 Nov 2010 10:27:41 +0000 (18:27 +0800)]
glsl: Add a helper function for determining if an rvalue could be a saturate.
Hardware pretty commonly has saturate modifiers on instructions, and
this can be used in codegen to produce those, without everyone else
needing to understand clamping other than min and max.
Eric Anholt [Thu, 18 Nov 2010 03:34:54 +0000 (11:34 +0800)]
i965: Fold constants into the second arg of BRW_SEL as well.
This hits a common case with min/max operations.
Eric Anholt [Thu, 18 Nov 2010 03:48:47 +0000 (11:48 +0800)]
i965: Remove extra \n at the end of every instruction in INTEL_DEBUG=wm.
Eric Anholt [Fri, 19 Nov 2010 09:44:35 +0000 (17:44 +0800)]
i965: Just use memset() to clear most members in FS constructors.
This should make it a lot harder to forget to zero things.
Eric Anholt [Fri, 19 Nov 2010 05:53:28 +0000 (13:53 +0800)]
i965: Fix compute_to_mrf to not move a MRF write up into another live range.
Fixes glsl-fs-copy-propagation-texcoords-1.
Eric Anholt [Fri, 19 Nov 2010 09:19:38 +0000 (17:19 +0800)]
mesa: Include C++ files in the makedepend of DRI drivers.
Vinson Lee [Sat, 20 Nov 2010 01:28:22 +0000 (17:28 -0800)]
glsl: Fix type of label 'default' in switch statement.
Vinson Lee [Sat, 20 Nov 2010 01:22:23 +0000 (17:22 -0800)]
glsl: Add lower_vector.cpp to SConscript.
Ian Romanick [Sat, 20 Nov 2010 01:16:12 +0000 (17:16 -0800)]
glsl: Fix matrix constructors with vector parameters
When the semantics of write masks in assignments were changed, this
code was not correctly updated.
Fixes piglit test glsl-mat-from-vec-ctor-01.
Kenneth Graunke [Fri, 19 Nov 2010 01:54:07 +0000 (17:54 -0800)]
glsl: Combine many instruction lowering passes into one.
This should save on the overhead of tree-walking and provide a
convenient place to add more instruction lowering in the future.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Kenneth Graunke [Fri, 19 Nov 2010 01:15:13 +0000 (17:15 -0800)]
glsl: Simplify a type check by using type->is_integer().
Ian Romanick [Tue, 16 Nov 2010 20:01:42 +0000 (12:01 -0800)]
glsl: Add ir_quadop_vector expression
The vector operator collects 2, 3, or 4 scalar components into a
vector. Doing this has several advantages. First, it will make
ud-chain tracking for components of vectors much easier. Second, a
later optimization pass could collect scalars into vectors to allow
generation of SWZ instructions (or similar as operands to other
instructions on R200 and i915). It also enables an easy way to
generate IR for SWZ instructions in the ARB_vertex_program assembler.
Ian Romanick [Tue, 9 Nov 2010 22:19:10 +0000 (14:19 -0800)]
glsl: Add unary ir_expression constructor
Ian Romanick [Tue, 16 Nov 2010 19:59:22 +0000 (11:59 -0800)]
glsl: Add ir_rvalue::is_negative_one predicate
Ian Romanick [Thu, 11 Nov 2010 00:33:10 +0000 (16:33 -0800)]
glsl: Eliminate assumptions about size of ir_expression::operands
This may grow in the near future.
Ian Romanick [Thu, 18 Nov 2010 19:05:32 +0000 (11:05 -0800)]
glsl: Add ir_unop_sin_reduced and ir_unop_cos_reduced
The operate just like ir_unop_sin and ir_unop_cos except that they
expect their inputs to be limited to the range [-pi, pi]. Several
GPUs require this limited range for their sine and cosine
instructions, so having these as operations (along with a to-be-written
lowering pass) helps this architectures.
These new operations also matche the semantics of the
GL_ARB_fragment_program SCS instruction. Having these as operations
helps in generating GLSL IR directly from assembly fragment programs.
Alex Deucher [Fri, 19 Nov 2010 20:51:24 +0000 (15:51 -0500)]
r600g: use full range of VS resources for vertex samplers
Now that we have fetch shaders, the full range of VS resources
can be used for sampling.
Alex Deucher [Fri, 19 Nov 2010 20:32:02 +0000 (15:32 -0500)]
r600g: use meaningful defines for chiprev
Makes the code much clearer.
Alex Deucher [Fri, 19 Nov 2010 20:19:39 +0000 (15:19 -0500)]
r600g: translate ARR instruction for evergreen
evergreen variant of:
9f7ec103e26c67cb077fd7d94d2fb68562b86c40
Jerome Glisse [Thu, 18 Nov 2010 19:29:16 +0000 (14:29 -0500)]
r600g: add fetch shader capabilities
Use fetch shader instead of having fetch instruction in the vertex
shader. Allow to restrict shader update to a smaller part when
vertex buffer input layout changes.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Alex Deucher [Fri, 19 Nov 2010 18:34:22 +0000 (13:34 -0500)]
r600g: All EVENT_WRITE packets need the EVENT_INDEX field
6xx-evergreen
Viktor Novotný [Tue, 16 Nov 2010 22:22:33 +0000 (23:22 +0100)]
dri/nouveau: Clean up magic numbers in get_rt_format
Signed-off-by: Viktor Novotný <noviktor@seznam.cz>
Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Jerome Glisse [Fri, 19 Nov 2010 16:51:37 +0000 (11:51 -0500)]
r600g: fix occlusion query on evergreen (avoid lockup)
Occlusion query on evergreen need the event index field to be
set otherwise we endup locking up the GPU.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Ian Romanick [Fri, 19 Nov 2010 00:11:25 +0000 (16:11 -0800)]
ir_to_mesa: Generate smarter code for some conditional moves
Condiation moves with a condition of (a < 0), (a > 0), (a <= 0), or (a
>= 0) can be generated with "a" directly as an operand of the CMP
instruction. This doesn't help much now, but it will help with
assembly shaders that use the CMP instruction.
Ian Romanick [Fri, 19 Nov 2010 01:11:17 +0000 (17:11 -0800)]
glsl: Make is_zero and is_one virtual methods of ir_rvalue
This eliminates the need in some cames to validate that an rvalue is
an ir_constant before checking to see if it's 0 or 1.
Brian Paul [Thu, 18 Nov 2010 23:15:33 +0000 (16:15 -0700)]
mesa: pass gl_format to _mesa_init_teximage_fields()
This should prevent the field going unset in the future. See bug
http://bugs.freedesktop.org/show_bug.cgi?id=31544 for background.
Also remove unneeded calls to clear_teximage_fields().
Finally, call _mesa_set_fetch_functions() from the
_mesa_init_teximage_fields() function so callers have one less
thing to worry about.
José Fonseca [Thu, 18 Nov 2010 13:02:36 +0000 (13:02 +0000)]
scons: Use inline wrap helpers more consistently.
Dave Airlie [Thu, 18 Nov 2010 08:10:55 +0000 (18:10 +1000)]
gallium/noop: report GL 2.1
this should at least make app use the same paths as they would for a real
driver.
Vinson Lee [Thu, 18 Nov 2010 06:42:51 +0000 (22:42 -0800)]
glsl: Fix 'control reaches end of non-void function' warning.
Fix this GCC warning.
ir.cpp: In static member function
'static unsigned int ir_expression::get_num_operands(ir_expression_operation)':
ir.cpp:199: warning: control reaches end of non-void function
Chia-I Wu [Wed, 17 Nov 2010 16:08:01 +0000 (00:08 +0800)]
mesa: Clean up core.h.
Remove version.h and context.h from core.h.
Chia-I Wu [Wed, 17 Nov 2010 15:43:01 +0000 (23:43 +0800)]
st/glx: Replace MESA_VERSION_STRING by xmesa_get_name.
xmesa_get_name returns the name of the st_api, which is the same as
MESA_VERSION_STRING.
Chia-I Wu [Wed, 10 Nov 2010 16:21:12 +0000 (00:21 +0800)]
st/wgl: Use st_context_iface::share for DrvShareLists.
Chia-I Wu [Wed, 10 Nov 2010 16:09:09 +0000 (00:09 +0800)]
gallium: Add st_context_iface::share to st_api.
It will be used to implement wglShareLists. Fill st_context_iface::copy
for glXCopyContext as well.
Chia-I Wu [Wed, 17 Nov 2010 15:56:42 +0000 (23:56 +0800)]
gallium: Add st_api::name.
It is the name of the rendering API. This field is informative.
Chia-I Wu [Wed, 17 Nov 2010 15:35:02 +0000 (23:35 +0800)]
st/vega: Do not wait NULL fences.
Eric Anholt [Thu, 18 Nov 2010 02:44:34 +0000 (10:44 +0800)]
i965: Eliminate dead code more aggressively.
If an instruction writes reg but nothing later uses it, then we don't
need to bother doing it. Before, we were just killing code that was
never read after it was ever written.
This removes many interpolation instructions for attributes with only
a few comopnents used. Improves nexuiz high-settings performance .46%
+/- .12% (n=3) on my Ironlake.
Brian Paul [Wed, 17 Nov 2010 15:10:26 +0000 (08:10 -0700)]
mesa: upgrade to glext.h version 66
The type of the num/count parameter to glProgramParameters4[df]vNV()
changed so some API dispatch code needed updates too.
Alex Deucher [Thu, 18 Nov 2010 02:30:09 +0000 (21:30 -0500)]
r600g: fix buffer alignment
This should fix the remaining buffer alignment issues in r600g.
Eric Anholt [Wed, 17 Nov 2010 06:59:08 +0000 (14:59 +0800)]
i965: Fail on loops on gen6 for now until we write the EU emit code for it.
Eric Anholt [Mon, 15 Nov 2010 23:41:40 +0000 (07:41 +0800)]
i965: Add dumping of the sampler default color.
Eric Anholt [Mon, 15 Nov 2010 23:34:32 +0000 (07:34 +0800)]
i965: Add state dumping for sampler state.
Eric Anholt [Sat, 13 Nov 2010 22:30:01 +0000 (14:30 -0800)]
mesa: Don't spam the console in a debug build unless some spam is requested.
It's annoying to use test suites under a Mesa debug build because
pretty output is cluttered with stderr's continuous reports that
you're still using the debug driver.
Eric Anholt [Wed, 17 Nov 2010 07:06:22 +0000 (15:06 +0800)]
i965: Shut up spurious gcc warning about GLSL_TYPE enums.
Jakob Bornecrantz [Wed, 10 Nov 2010 22:49:40 +0000 (22:49 +0000)]
gallium: Remove redundant sw and debug target helpers
Jakob Bornecrantz [Wed, 10 Nov 2010 22:02:39 +0000 (22:02 +0000)]
graw: Use inline debug helper instead of non-inline version
Jakob Bornecrantz [Wed, 10 Nov 2010 23:42:17 +0000 (23:42 +0000)]
libgl-xlib: Use inline debug helper instead of non-inline version
Chad Versace [Wed, 17 Nov 2010 23:28:36 +0000 (15:28 -0800)]
glsl: Improve usage message for glsl_compiler
The new usage message lists possible command line options. (Newcomers to Mesa
currently have to trawl through the source to find the command line options,
and we should save them from that trouble.)
Example Output
--------------
usage: ./glsl_compiler [options] <file.vert | file.geom | file.frag>
Possible options are:
--glsl-es
--dump-ast
--dump-hir
--dump-lir
--link
Kenneth Graunke [Wed, 17 Nov 2010 23:31:35 +0000 (15:31 -0800)]
glsl: Refactor get_num_operands.
This adds sentinel values to the ir_expression_operation enum type:
ir_last_unop, ir_last_binop, and ir_last_opcode. They are set to the
previous one so they don't trigger "unhandled case in switch statement"
warnings, but should never be handled directly.
This allows us to remove the huge array of 1s and 2s in
ir_expression::get_num_operands().
Jerome Glisse [Wed, 17 Nov 2010 22:20:59 +0000 (17:20 -0500)]
r600g: code cleanup (indent, trailing space, empty line ...)
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Kenneth Graunke [Wed, 17 Nov 2010 21:59:17 +0000 (13:59 -0800)]
glsl: Remove the ir_binop_cross opcode.
Kenneth Graunke [Wed, 17 Nov 2010 21:37:16 +0000 (13:37 -0800)]
Refresh autogenerated file builtin_function.cpp.
Kenneth Graunke [Wed, 17 Nov 2010 21:20:30 +0000 (13:20 -0800)]
glsl: Reimplement the "cross" built-in without ir_binop_cross.
We are not aware of any GPU that actually implements the cross product
as a single instruction. Hence, there's no need for it to be an opcode.
Future commits will remove it entirely.
Kenneth Graunke [Wed, 17 Nov 2010 20:53:07 +0000 (12:53 -0800)]
Regenerate glcpp parser.
Kenneth Graunke [Wed, 17 Nov 2010 20:50:35 +0000 (12:50 -0800)]
glsl: Unconditionally define GL_FRAGMENT_PRECISION_HIGH in ES2 shaders.
This is really supposed to be defined only if the driver supports highp
in the fragment shader - but all of our current ES2 implementations do.
So, just define it. In the future, we'll need to add a flag to
gl_context and only define the macro if the flag is set.
"Fixes" freedesktop.org bug #31673.
Robert Hooker [Tue, 16 Nov 2010 20:10:20 +0000 (15:10 -0500)]
egl_dri2: Add missing intel chip ids.
Signed-off-by: Robert Hooker <robert.hooker@canonical.com>
Chad Versace [Wed, 17 Nov 2010 18:43:10 +0000 (10:43 -0800)]
glsl: Fix Doxygen tag \file in recently renamed files
Chad Versace [Wed, 17 Nov 2010 18:28:01 +0000 (10:28 -0800)]
glsl: Fix erroneous cast in ast_jump_statement::hir()
Return values were erroneously cast from (ir_rvalue*) to (ir_expression*).
NOTE: This is a candidate for the 7.9 branch.
Kenneth Graunke [Wed, 17 Nov 2010 18:40:28 +0000 (10:40 -0800)]
glsl: Fix constant expression handling for <, >, <=, >= on vectors.
ir_binop_less, ir_binop_greater, ir_binop_lequal, and ir_binop_gequal
are defined to work on vectors as well as scalars, as long as the two
operands have the same type.
This is evident from both ir_validate.cpp and our use of these opcodes
in the GLSL lessThan, greaterThan, lessThanEqual, greaterThanEqual
built-in functions.
Found by code inspection. Not known to fix any bugs. Presumably, our
tests for the built-in comparison functions must pass because C.E.
handling is done on the ir_call of "greaterThan" rather than the inlined
opcode. The C.E. handling of the built-in function calls is correct.
NOTE: This is a candidate for the 7.9 branch.
Marek Olšák [Wed, 17 Nov 2010 16:42:34 +0000 (17:42 +0100)]
r300g: print FS inputs uninitialized due to hardware limits to stderr
Alex Deucher [Wed, 17 Nov 2010 16:30:52 +0000 (11:30 -0500)]
r600c/evergreen: texture align is group_bytes just like 6xx/7xx
Default group bytes to 512 on evergreen. Don't query
tiling config yet for evergreen, the current info returned is not
adequate for evergreen (no way to get bank info).
Brian Paul [Tue, 16 Nov 2010 22:47:11 +0000 (15:47 -0700)]
mesa: minor clean-ups in context code
Brian Paul [Tue, 16 Nov 2010 22:32:52 +0000 (15:32 -0700)]
mesa: reorder texture_error_check() params
To better match other functions.
Brian Paul [Tue, 16 Nov 2010 22:30:18 +0000 (15:30 -0700)]
mesa: 80-column wrapping
Brian Paul [Tue, 16 Nov 2010 22:23:51 +0000 (15:23 -0700)]
mesa: whitespace cleanups
Brian Paul [Tue, 16 Nov 2010 22:15:58 +0000 (15:15 -0700)]
mesa: fix error messages and minor reindenting
Kenneth Graunke [Wed, 17 Nov 2010 00:21:54 +0000 (16:21 -0800)]
Refresh autogenerated glcpp parser.
Kenneth Graunke [Wed, 17 Nov 2010 00:21:20 +0000 (16:21 -0800)]
glcpp: Define GL_FRAGMENT_PRECISION_HIGH if GLSL version >= 1.30.
Per section 4.5.4 of the GLSL 1.30 specification.
Henri Verbeet [Tue, 16 Nov 2010 23:29:03 +0000 (00:29 +0100)]
r600g: Synchronize supported color formats between Evergreen and r600/r700.
Henri Verbeet [Tue, 16 Nov 2010 23:29:02 +0000 (00:29 +0100)]
r600g: Swizzle vertex data only once.
Vertex data swizzles are already done in the vertex shader. Doing them twice
breaks BGRA vertex arrays for example.
Marek Olšák [Tue, 16 Nov 2010 21:19:47 +0000 (22:19 +0100)]
r300g: remove the hack with OPCODE_RET
RET was interpreted as END, which was wrong. Instead, if a shader contains RET
in the main function, it will fail to compile with an error message
from now on.
The hack is from early days.
Ian Romanick [Mon, 15 Nov 2010 21:41:06 +0000 (13:41 -0800)]
glsl: Simplify generation of swizzle for vector constructors
Ian Romanick [Fri, 12 Nov 2010 18:19:08 +0000 (10:19 -0800)]
glsl: Refactor is_vec_{zero,one} to be methods of ir_constant
These predicates will be used in other places soon.
José Fonseca [Tue, 16 Nov 2010 18:56:39 +0000 (18:56 +0000)]
libgl-gdi: Allow to pick softpipe/llvmpipe on runtime.
Vinson Lee [Tue, 16 Nov 2010 06:00:32 +0000 (22:00 -0800)]
mesa: Add definitions for inverse hyperbolic function on MSVC.
Vinson Lee [Tue, 16 Nov 2010 04:54:12 +0000 (20:54 -0800)]
glsl: Add ir_constant_expression.cpp to SConscript.
This was accidentally removed in commit
32aaf89823de11e98cb59d5ec78c66cd3e74bcd4.
Fixes SCons builds.
Brian Paul [Tue, 16 Nov 2010 01:59:39 +0000 (18:59 -0700)]
glsl: remove opt_constant_expression.cpp from SConscript
And alphabetize the opt_* files.
Brian Paul [Tue, 16 Nov 2010 01:50:58 +0000 (18:50 -0700)]
mesa: add more work-arounds for acoshf(), asinhf(), atahf()
Brian Paul [Tue, 16 Nov 2010 01:48:43 +0000 (18:48 -0700)]
glsl: fix assorted MSVC warnings
Brian Paul [Tue, 16 Nov 2010 01:36:50 +0000 (18:36 -0700)]
st/mesa: fix glDrawPixels(depth/stencil) bugs
When drawing GL_DEPTH_COMPONENT the usual fragment pipeline steps apply
so don't override the depth state.
When drawing GL_STENCIL_INDEX (or GL_DEPTH_STENCIL) the fragment pipeline
does not apply (only the stencil and Z writemasks apply) so disable writes
to the color buffers.
Fixes some regressions from commit
ef8bb7ada98f1ddc8e2554a7336af5d669cb1290
Kenneth Graunke [Mon, 15 Nov 2010 22:35:46 +0000 (14:35 -0800)]
glsl: Rename various ir_* files to lower_* and opt_*.
This helps distinguish between lowering passes, optimization passes, and
other compiler code.
Kenneth Graunke [Mon, 15 Nov 2010 22:45:40 +0000 (14:45 -0800)]
glsl: Remove unused and out of date Makefile.am.
This was from when glsl2 lived in a separate repository and used
automake.
Kenneth Graunke [Mon, 15 Nov 2010 22:08:58 +0000 (14:08 -0800)]
glsl: Add constant expression handling for asinh, acosh, and atanh.
Kenneth Graunke [Mon, 15 Nov 2010 22:02:13 +0000 (14:02 -0800)]
glsl: Refresh autogenerated file builtin_function.cpp.