mesa.git
13 years agoglsl2: Create new talloc contexts the "right" way.
Kenneth Graunke [Fri, 25 Jun 2010 20:36:14 +0000 (13:36 -0700)]
glsl2: Create new talloc contexts the "right" way.

13 years agoglsl2: Use _mesa_glsl_parse_state as the talloc parent, not glsl_shader.
Kenneth Graunke [Fri, 25 Jun 2010 20:14:37 +0000 (13:14 -0700)]
glsl2: Use _mesa_glsl_parse_state as the talloc parent, not glsl_shader.

_mesa_glsl_parse_state should be the parent for all temporary allocation
done while compiling a shader.  glsl_shader should only be used as the
parent for the shader's final IR---the _result_ of compilation.

Since many IR instructions may be added or discarded during optimization
passes, IR should not ever be allocated to glsl_shader directly.

Done via sed -i s/talloc_parent(state)/state/g and s/talloc_parent(st)/st/g.

This also removes a ton of talloc_parent calls, which may help performance.

13 years agoglsl2: Steal the live IR and free the rest of the junk.
Kenneth Graunke [Fri, 25 Jun 2010 21:10:01 +0000 (14:10 -0700)]
glsl2: Steal the live IR and free the rest of the junk.

13 years agolinker: Don't dynamically allocate slots for linked shaders
Ian Romanick [Wed, 30 Jun 2010 00:58:43 +0000 (17:58 -0700)]
linker: Don't dynamically allocate slots for linked shaders

The can be at most one shader per stage.  There are currently only two
stages.  There is zero reason to dynamically size this array.

13 years agoir_to_mesa: Tell Mesa about our choices for vertex attribute locations.
Eric Anholt [Wed, 30 Jun 2010 01:23:05 +0000 (18:23 -0700)]
ir_to_mesa: Tell Mesa about our choices for vertex attribute locations.

13 years agolinker: Don't automatically allocate VERT_ATTRIB_GENERIC0
Ian Romanick [Wed, 30 Jun 2010 01:58:20 +0000 (18:58 -0700)]
linker: Don't automatically allocate VERT_ATTRIB_GENERIC0

13 years agoglsl2: Use Mesa's gl_shader_program instead of our own struct glsl_program.
Eric Anholt [Wed, 30 Jun 2010 18:49:17 +0000 (11:49 -0700)]
glsl2: Use Mesa's gl_shader_program instead of our own struct glsl_program.

This avoids more allocation and shuffling of data around.

13 years agoglsl2: Make function names and variable names be children of the node.
Eric Anholt [Wed, 30 Jun 2010 18:51:54 +0000 (11:51 -0700)]
glsl2: Make function names and variable names be children of the node.

This avoids losing their memory when the parser state is freed.

13 years agoglsl2: Move our data from a glsl_shader* on the side to the main gl_shader *.
Eric Anholt [Wed, 30 Jun 2010 18:05:43 +0000 (11:05 -0700)]
glsl2: Move our data from a glsl_shader* on the side to the main gl_shader *.

This saves recompiling at link time.  gl_shader->ir is made a pointer
so that we don't have to bring exec_list into mtypes.h.

13 years agoir_to_mesa: Add support for ir_unop_abs.
Eric Anholt [Tue, 29 Jun 2010 23:39:38 +0000 (16:39 -0700)]
ir_to_mesa: Add support for ir_unop_abs.

13 years agoir_to_mesa: Add support for dFdx, dFdy.
Eric Anholt [Tue, 29 Jun 2010 23:36:42 +0000 (16:36 -0700)]
ir_to_mesa: Add support for dFdx, dFdy.

13 years agoir_to_mesa: Start adding support for texture instructions.
Eric Anholt [Tue, 29 Jun 2010 21:16:11 +0000 (14:16 -0700)]
ir_to_mesa: Start adding support for texture instructions.

Fixes:
glsl-fs-bug25902
glsl-fs-sampler-numbering
glsl-lod-bias

13 years agoglsl2: Keep the same number of components in implicit conversions.
Kenneth Graunke [Tue, 29 Jun 2010 22:59:27 +0000 (15:59 -0700)]
glsl2: Keep the same number of components in implicit conversions.

Fixes piglit test glsl-implicit-conversion-01.

13 years agoglsl2: Make gl_MaxDrawBuffers available in the vertex shader
Ian Romanick [Tue, 29 Jun 2010 22:29:56 +0000 (15:29 -0700)]
glsl2: Make gl_MaxDrawBuffers available in the vertex shader

13 years agoglsl2: Make gl_MaxDrawBuffers available in the fragment shader
Ian Romanick [Tue, 29 Jun 2010 22:19:11 +0000 (15:19 -0700)]
glsl2: Make gl_MaxDrawBuffers available in the fragment shader

13 years agoglsl2: Make gl_FragData be available in GLSL 1.10 too
Ian Romanick [Tue, 29 Jun 2010 22:10:09 +0000 (15:10 -0700)]
glsl2: Make gl_FragData be available in GLSL 1.10 too

13 years agoglsl2: Pass MaxDrawBuffers from core Mesa into the GLSL compiler
Ian Romanick [Tue, 29 Jun 2010 21:21:05 +0000 (14:21 -0700)]
glsl2: Pass MaxDrawBuffers from core Mesa into the GLSL compiler

13 years agoglsl_type: Add _mesa_glsl_release_types to release all type related storage
Ian Romanick [Mon, 28 Jun 2010 20:17:18 +0000 (13:17 -0700)]
glsl_type: Add _mesa_glsl_release_types to release all type related storage

13 years agoglsl_type: All glsl_type objects live in their own talloc context
Ian Romanick [Mon, 28 Jun 2010 19:19:52 +0000 (12:19 -0700)]
glsl_type: All glsl_type objects live in their own talloc context

13 years agoglsl_type: Record type constructors are private
Ian Romanick [Mon, 28 Jun 2010 18:57:38 +0000 (11:57 -0700)]
glsl_type: Record type constructors are private

13 years agoglsl_type: Add get_record_instance method
Ian Romanick [Mon, 28 Jun 2010 18:54:57 +0000 (11:54 -0700)]
glsl_type: Add get_record_instance method

13 years agoglsl_type: Vector, matrix, and sampler type constructors are private
Ian Romanick [Mon, 28 Jun 2010 18:09:40 +0000 (11:09 -0700)]
glsl_type: Vector, matrix, and sampler type constructors are private

13 years agoglsl_type: Make all static objects be class private
Ian Romanick [Mon, 28 Jun 2010 18:01:57 +0000 (11:01 -0700)]
glsl_type: Make all static objects be class private

13 years agoglsl2: Use talloc_strdup when generating constructor temporary names
Ian Romanick [Mon, 28 Jun 2010 20:22:55 +0000 (13:22 -0700)]
glsl2: Use talloc_strdup when generating constructor temporary names

13 years agoglsl_type: Remove vector and matrix constructor generators
Ian Romanick [Sat, 26 Jun 2010 00:58:16 +0000 (17:58 -0700)]
glsl_type: Remove vector and matrix constructor generators

All scalar, vector, and matrix constructors are generated in-line
during AST-to-HIR translation.  There is no longer any need to
generate function versions of the constructors.

13 years agoglsl2: Don't flatten constructor parameters to scalars
Ian Romanick [Sat, 26 Jun 2010 00:36:17 +0000 (17:36 -0700)]
glsl2: Don't flatten constructor parameters to scalars

Now that all scalar, vector, and matrix constructors are emitted
in-line, the parameters to these constructors should not be flattened
to a pile of scalars.  Instead, the functions that emit the in-line
constructor bodies can directly write the parameters to the correct
locations in the objects being constructed.

13 years agoglsl2: Always emit matrix constructors inline
Ian Romanick [Fri, 25 Jun 2010 23:10:43 +0000 (16:10 -0700)]
glsl2: Always emit matrix constructors inline

13 years agoglsl2: Always emit vector constructors inline
Ian Romanick [Wed, 23 Jun 2010 22:19:40 +0000 (15:19 -0700)]
glsl2: Always emit vector constructors inline

13 years agoir_swizzle: Add new constructor, refactor constructors
Ian Romanick [Fri, 25 Jun 2010 22:25:27 +0000 (15:25 -0700)]
ir_swizzle: Add new constructor, refactor constructors

Adds a new constructor that takes an array of component values.  Refactors
the meat of the two constructors to an init_mask method.

13 years agoglsl2: Update TODO.
Kenneth Graunke [Tue, 29 Jun 2010 17:02:01 +0000 (10:02 -0700)]
glsl2: Update TODO.

13 years agoglsl2: Check for non-void functions that don't have a return statement.
Kenneth Graunke [Tue, 29 Jun 2010 16:59:40 +0000 (09:59 -0700)]
glsl2: Check for non-void functions that don't have a return statement.

This doesn't do any control flow analysis to ensure that the return
statements are actually reached.

Fixes piglit tests function5.frag and function-07.vert.

13 years agoglsl2: Reject return types with qualifiers.
Kenneth Graunke [Tue, 29 Jun 2010 07:48:10 +0000 (00:48 -0700)]
glsl2: Reject return types with qualifiers.

Fixes piglit test return-qualifier.frag.

13 years agoglsl2: Add a method for querying if an AST type has any qualifiers.
Kenneth Graunke [Tue, 29 Jun 2010 07:47:44 +0000 (00:47 -0700)]
glsl2: Add a method for querying if an AST type has any qualifiers.

13 years agoglsl2: Check that returned expressions match the function return type.
Kenneth Graunke [Tue, 29 Jun 2010 06:38:04 +0000 (23:38 -0700)]
glsl2: Check that returned expressions match the function return type.

From my reading of the specification, implicit conversions are not
allowed.  ATI seems to agree, though nVidia allows it without warning.

13 years agoglsl2: Invoke preprocessor before calling the compiler proper
Ian Romanick [Tue, 29 Jun 2010 17:02:36 +0000 (10:02 -0700)]
glsl2: Invoke preprocessor before calling the compiler proper

13 years agoir_to_mesa: Actually initialize the undef register for scalar_op1.
Eric Anholt [Mon, 28 Jun 2010 22:02:51 +0000 (15:02 -0700)]
ir_to_mesa: Actually initialize the undef register for scalar_op1.

Fixes glsl-sin, glsl-cos on 965, where we rely on unused src arguments
in the VS having a file of PROGRAM_UNDEFINED.

13 years agoir_to_mesa: Support user-defined varyings using the linker's locations.
Eric Anholt [Mon, 28 Jun 2010 21:47:43 +0000 (14:47 -0700)]
ir_to_mesa: Support user-defined varyings using the linker's locations.

Fixes glsl-reload-source.

13 years agoir_to_mesa: Actually add the header file for the interface.
Eric Anholt [Mon, 28 Jun 2010 20:45:43 +0000 (13:45 -0700)]
ir_to_mesa: Actually add the header file for the interface.

13 years agoir_to_mesa: Fix binop_sqrt for multi-channel and negative source channels.
Eric Anholt [Mon, 28 Jun 2010 20:01:49 +0000 (13:01 -0700)]
ir_to_mesa: Fix binop_sqrt for multi-channel and negative source channels.

Fixes glsl-fs-sqrt-branch.

13 years agoir_to_mesa: Fix indexes of temps used in expressions.
Eric Anholt [Mon, 28 Jun 2010 19:48:47 +0000 (12:48 -0700)]
ir_to_mesa: Fix indexes of temps used in expressions.

It looks like I managed to horribly mangle this in some rebase of the
branch.  Fixes:
glsl-fs-fragcoord
glsl-fs-mix

13 years agoir_to_mesa: Notify the driver when we generate new Mesa programs for GLSL.
Eric Anholt [Mon, 28 Jun 2010 19:35:54 +0000 (12:35 -0700)]
ir_to_mesa: Notify the driver when we generate new Mesa programs for GLSL.

Fixes glsl-fs-if-*.

13 years agoir_to_mesa: Add support for the pow expression.
Eric Anholt [Mon, 28 Jun 2010 19:26:19 +0000 (12:26 -0700)]
ir_to_mesa: Add support for the pow expression.

Fixes glsl-algebraic-pow-two.

13 years agoir_to_mesa: Fix EmitCondCodes for boolean vars as condition.
Eric Anholt [Mon, 28 Jun 2010 19:16:03 +0000 (12:16 -0700)]
ir_to_mesa: Fix EmitCondCodes for boolean vars as condition.

Fixes glsl-vs-if-bool.

13 years agoir_to_mesa: Respect EmitCondCodes for IF statements.
Eric Anholt [Mon, 28 Jun 2010 18:59:13 +0000 (11:59 -0700)]
ir_to_mesa: Respect EmitCondCodes for IF statements.

Fixes glsl-vs-if-* for the 965 driver.

13 years agoir_to_mesa: Traverse the "else" instrs after "else", instead of "then" again.
Eric Anholt [Mon, 28 Jun 2010 18:44:30 +0000 (11:44 -0700)]
ir_to_mesa: Traverse the "else" instrs after "else", instead of "then" again.

13 years agoir_to_mesa: Fix matrix * scalar multiplication.
Eric Anholt [Tue, 29 Jun 2010 02:56:53 +0000 (19:56 -0700)]
ir_to_mesa: Fix matrix * scalar multiplication.

We're accessing in terms of columns, so we need to do MUL/MAD/MAD/MAD
instead of DP4s.

Fixes:
glsl-fs-exp2
glsl-fs-log2
glsl-fs-mix-constant
glsl-fs-sqrt-zero
glsl-vs-sqrt-zero

13 years agoir_to_mesa: Check the right element for matrix * scalar multiplication.
Eric Anholt [Mon, 28 Jun 2010 04:07:21 +0000 (21:07 -0700)]
ir_to_mesa: Check the right element for matrix * scalar multiplication.

13 years agoglsl2: Add support for some builtin matrices.
Eric Anholt [Mon, 28 Jun 2010 03:36:41 +0000 (20:36 -0700)]
glsl2: Add support for some builtin matrices.

13 years agoir_to_mesa: Fix copy and wasted InputsRead/OutputsWritten setup.
Eric Anholt [Mon, 28 Jun 2010 03:43:09 +0000 (20:43 -0700)]
ir_to_mesa: Fix copy and wasted InputsRead/OutputsWritten setup.

13 years agoUse a more sensible context in copy propagation.
Kenneth Graunke [Fri, 25 Jun 2010 20:44:09 +0000 (13:44 -0700)]
Use a more sensible context in copy propagation.

13 years agoUse more sensible contexts in ir_dead_code_local.
Kenneth Graunke [Fri, 25 Jun 2010 20:36:14 +0000 (13:36 -0700)]
Use more sensible contexts in ir_dead_code_local.

13 years agoglsl2: Add option to stand-alone GLSL compiler to dump IR before optimizations
Ian Romanick [Fri, 25 Jun 2010 22:27:47 +0000 (15:27 -0700)]
glsl2: Add option to stand-alone GLSL compiler to dump IR before optimizations

13 years agoglsl2: Use i2b and f2b IR opcodes for casting int or float to bool
Ian Romanick [Fri, 25 Jun 2010 23:19:45 +0000 (16:19 -0700)]
glsl2: Use i2b and f2b IR opcodes for casting int or float to bool

13 years agoglsl2: Add support for non-float constants in Mesa IR.
Eric Anholt [Sat, 26 Jun 2010 00:50:25 +0000 (17:50 -0700)]
glsl2: Add support for non-float constants in Mesa IR.

Fixes glsl-vs-vec4-indexing-4.

13 years agoglsl2: Take out the spamming of debug flags I'd added.
Eric Anholt [Sat, 26 Jun 2010 00:23:31 +0000 (17:23 -0700)]
glsl2: Take out the spamming of debug flags I'd added.

13 years agoglsl2: Don't clear swizzles for Mesa IR constants after fetching them.
Eric Anholt [Sat, 26 Jun 2010 00:20:46 +0000 (17:20 -0700)]
glsl2: Don't clear swizzles for Mesa IR constants after fetching them.

Missed this while hacking in constants support.  Fixes:
glsl-algebraic-mul-*
glsl-algebraic-rcp-*
glsl-vs-swizzle-swizzle-lhs
glsl-vs-vec4-indexing-6

13 years agoir_reader: Free memory for S-Expressions earlier.
Kenneth Graunke [Fri, 25 Jun 2010 20:10:37 +0000 (13:10 -0700)]
ir_reader: Free memory for S-Expressions earlier.

There's no point in keeping it around once we've read the IR.

Also, remove an unnecessary talloc_parent call.

13 years agoglsl2: Start trying to hook up uniforms.
Eric Anholt [Fri, 25 Jun 2010 21:35:48 +0000 (14:35 -0700)]
glsl2: Start trying to hook up uniforms.

This should be resolved with linker.cpp's location assignment, as
currently we drop that location assignment on the ground.  However,
this gets basic programs using uniforms working for now.

13 years agoglsl2: Associate the GLenum for the type with builtin GLSL types.
Eric Anholt [Fri, 25 Jun 2010 21:27:07 +0000 (14:27 -0700)]
glsl2: Associate the GLenum for the type with builtin GLSL types.

13 years agoglsl2: Use the parser state as the talloc context for dead code elimination.
Eric Anholt [Fri, 25 Jun 2010 20:38:38 +0000 (13:38 -0700)]
glsl2: Use the parser state as the talloc context for dead code elimination.

This cuts runtime by around 20% from talloc_parent() lookups.

13 years agoglsl2: Emit OPCODE_END at the end of the Mesa program.
Eric Anholt [Fri, 25 Jun 2010 20:00:38 +0000 (13:00 -0700)]
glsl2: Emit OPCODE_END at the end of the Mesa program.

The 965 driver can now run a glsl2-generated shader!

13 years agoglsl2: Hook up constant parameters in ir_to_mesa.
Eric Anholt [Fri, 25 Jun 2010 19:59:10 +0000 (12:59 -0700)]
glsl2: Hook up constant parameters in ir_to_mesa.

13 years agoglsl2: Set InputsRead and OutputsWritten on the generated programs.
Eric Anholt [Fri, 25 Jun 2010 19:52:01 +0000 (12:52 -0700)]
glsl2: Set InputsRead and OutputsWritten on the generated programs.

13 years agoglsl2: Start integrating ir_to_mesa.cpp into shader_api.h
Eric Anholt [Fri, 25 Jun 2010 19:37:21 +0000 (12:37 -0700)]
glsl2: Start integrating ir_to_mesa.cpp into shader_api.h

The compiler is now called by the driver, and generates program
instructions.  Parameter lists are still not set up, so the driver
chokes on it shortly thereafter.

13 years agoglsl2: Use Mesa types instead of duping them into our program.h.
Eric Anholt [Fri, 25 Jun 2010 19:23:38 +0000 (12:23 -0700)]
glsl2: Use Mesa types instead of duping them into our program.h.

13 years agoglsl2: Fix dependencies. (at least partially)
Eric Anholt [Fri, 25 Jun 2010 19:23:20 +0000 (12:23 -0700)]
glsl2: Fix dependencies.  (at least partially)

13 years agoglsl2: Replace the GLSL compiler with the glsl2 project.
Eric Anholt [Mon, 21 Jun 2010 18:29:15 +0000 (11:29 -0700)]
glsl2: Replace the GLSL compiler with the glsl2 project.

13 years agoglsl2: Wrap includes of C interfaces with extern "C".
Eric Anholt [Fri, 25 Jun 2010 00:08:53 +0000 (17:08 -0700)]
glsl2: Wrap includes of C interfaces with extern "C".

13 years agoglsl2: Remove files that had been imported for standalone.
Eric Anholt [Thu, 24 Jun 2010 23:34:43 +0000 (16:34 -0700)]
glsl2: Remove files that had been imported for standalone.

13 years agoglsl2: Stop .gitignoring the old standalone build system.
Eric Anholt [Thu, 24 Jun 2010 22:52:07 +0000 (15:52 -0700)]
glsl2: Stop .gitignoring the old standalone build system.

13 years agoglsl2: Move the Mesa IR codegen into mesa/shader/
Eric Anholt [Thu, 24 Jun 2010 22:49:18 +0000 (15:49 -0700)]
glsl2: Move the Mesa IR codegen into mesa/shader/

13 years agoMerge branch 'glsl2-head' into glsl2
Eric Anholt [Thu, 24 Jun 2010 22:47:38 +0000 (15:47 -0700)]
Merge branch 'glsl2-head' into glsl2

This brings in the standalone GLSL compiler that we are planning on
replacing the existing Mesa GLSL compiler.  It currently targets GLSL
1.20 and the Mesa IR.

13 years agoglsl2: Add a README file for the new compiler.
Eric Anholt [Thu, 24 Jun 2010 22:41:40 +0000 (15:41 -0700)]
glsl2: Add a README file for the new compiler.

13 years agoglsl2: Move the compiler to the subdirectory it will live in in Mesa.
Eric Anholt [Thu, 24 Jun 2010 22:32:15 +0000 (15:32 -0700)]
glsl2: Move the compiler to the subdirectory it will live in in Mesa.

13 years agoMerge branch 'mesa'
Eric Anholt [Thu, 24 Jun 2010 22:21:51 +0000 (15:21 -0700)]
Merge branch 'mesa'

This brings in the ir_to_mesa.cpp code I've been developing to codegen
to the Mesa IR.  It does not actually generate a complete Mesa
fragment/vertex program yet.

13 years agoMove the talloc_parent lookup down in a few hot paths.
Eric Anholt [Thu, 24 Jun 2010 22:18:39 +0000 (15:18 -0700)]
Move the talloc_parent lookup down in a few hot paths.

talloc_parent is still 80% of our runtime, but likely talloc_parent
lookups will be reduced as we improve the handling of memory
ownership.

13 years agoMerge remote branch 'cworth/master'
Eric Anholt [Thu, 24 Jun 2010 22:13:03 +0000 (15:13 -0700)]
Merge remote branch 'cworth/master'

Conflicts:
ast_to_hir.cpp
ir.cpp

This brings in the talloc-based memory management work, so that the
compiler (almost) no longer leaks memory.

13 years agoir_to_mesa: Handle a limited subset of matrix multiplication.
Eric Anholt [Thu, 3 Jun 2010 23:37:17 +0000 (16:37 -0700)]
ir_to_mesa: Handle a limited subset of matrix multiplication.

glsl-mvp.vert now generates believable code, and mesa mode fails only
5 tests that master doesn't.  I must have left out some asserts...

13 years agoir_to_mesa: Handle constant matrices.
Eric Anholt [Thu, 3 Jun 2010 23:31:14 +0000 (16:31 -0700)]
ir_to_mesa: Handle constant matrices.

There's not much to it since we're not actually storing constant data yet.

13 years agoir_to_mesa: Fix copy-and-wasted second argument to compare expresssion ops.
Eric Anholt [Thu, 3 Jun 2010 18:18:51 +0000 (11:18 -0700)]
ir_to_mesa: Fix copy-and-wasted second argument to compare expresssion ops.

Fixes CorrectParse2.vert assertion due to uninitialized values.

13 years agoir_to_mesa: Don't allocate temps for swizzles.
Eric Anholt [Thu, 3 Jun 2010 17:13:18 +0000 (10:13 -0700)]
ir_to_mesa: Don't allocate temps for swizzles.

We do them in place by actually, you know, swizzling.

13 years agoir_to_mesa: Set up storage for uniform vars.
Eric Anholt [Thu, 3 Jun 2010 16:39:54 +0000 (09:39 -0700)]
ir_to_mesa: Set up storage for uniform vars.

13 years agoir_to_mesa: Move the classes into the file now that we don't have the burg.
Eric Anholt [Thu, 3 Jun 2010 16:31:46 +0000 (09:31 -0700)]
ir_to_mesa: Move the classes into the file now that we don't have the burg.

At 1kloc, it doesn't look like I'll want to split the ir_to_mesa file
up even once it's feature-complete.  Move definitions closer to usage,
and prevent rebuilding the world when changing the definitions.

13 years agoir_to_mesa: Remove old monoburg structure.
Eric Anholt [Thu, 3 Jun 2010 16:29:29 +0000 (09:29 -0700)]
ir_to_mesa: Remove old monoburg structure.

13 years agoir_to_mesa: Restrict dst writemasks like we did in the monoburg setup.
Eric Anholt [Thu, 3 Jun 2010 16:17:54 +0000 (09:17 -0700)]
ir_to_mesa: Restrict dst writemasks like we did in the monoburg setup.

13 years agoir_to_mesa: Fix copy-and-wasted DIV instruction sequence.
Eric Anholt [Thu, 3 Jun 2010 16:04:57 +0000 (09:04 -0700)]
ir_to_mesa: Fix copy-and-wasted DIV instruction sequence.

13 years agoir_to_mesa: Remove the BURG code.
Eric Anholt [Thu, 3 Jun 2010 00:43:43 +0000 (17:43 -0700)]
ir_to_mesa: Remove the BURG code.

The promise of the BURG was to recognize multi-instruction sequences
and emit reduced sequences for them.  It would have worked well for
recognizing MUL+ADD -> MAD and possibly even MIN(MAX(val, 0), 1) ->
MOV_SAT with some grammar changes.  However, that potential benefit in
making those optimizations easy is outweighed by the fragility of
monoburg, the amount of (incorrect, as I wrote it) code for using it,
and the burden it was going to cause for handling operations on
aggregate types.

13 years agoir_to_mesa: Fix mapping of FS texcoord inputs and color output.
Eric Anholt [Tue, 1 Jun 2010 23:32:46 +0000 (16:32 -0700)]
ir_to_mesa: Fix mapping of FS texcoord inputs and color output.

13 years agoir_to_mesa: Try to fix up the dereference handling for the visitor rework.
Eric Anholt [Tue, 1 Jun 2010 23:23:57 +0000 (16:23 -0700)]
ir_to_mesa: Try to fix up the dereference handling for the visitor rework.

One of the gstreamer shaders I play with now compiles, but input
mappings are wrong.

13 years agoir_to_mesa: Implement min and max expressions.
Eric Anholt [Wed, 19 May 2010 23:10:37 +0000 (16:10 -0700)]
ir_to_mesa: Implement min and max expressions.

fixes glsl-orangebook-ch06-bump.frag.

13 years agoir_to_mesa: Don't assert over assignments with a constant-true condition.
Eric Anholt [Wed, 19 May 2010 23:06:37 +0000 (16:06 -0700)]
ir_to_mesa: Don't assert over assignments with a constant-true condition.

13 years agoir_to_mesa: Add support for trunc/ceil/floor.
Eric Anholt [Wed, 19 May 2010 23:02:00 +0000 (16:02 -0700)]
ir_to_mesa: Add support for trunc/ceil/floor.

13 years agoir_to_mesa: Implement neg expression.
Eric Anholt [Wed, 19 May 2010 22:54:28 +0000 (15:54 -0700)]
ir_to_mesa: Implement neg expression.

13 years agoir_to_mesa: Add sin/cos.
Eric Anholt [Wed, 19 May 2010 22:50:02 +0000 (15:50 -0700)]
ir_to_mesa: Add sin/cos.

13 years agoir_to_mesa: Start trying to support struct storage.
Eric Anholt [Wed, 12 May 2010 17:16:11 +0000 (10:16 -0700)]
ir_to_mesa: Start trying to support struct storage.

13 years agoir_to_mesa: Fix up array indexing.
Eric Anholt [Tue, 11 May 2010 23:20:21 +0000 (16:20 -0700)]
ir_to_mesa: Fix up array indexing.

The grammar for array_reference_vec4_vec4 was set up wrong, so we
weren't generating instructions if necessary for the array index.

13 years agoir_to_mesa: Remove stale comment about monoburg.
Eric Anholt [Tue, 11 May 2010 07:00:35 +0000 (00:00 -0700)]
ir_to_mesa: Remove stale comment about monoburg.

13 years agoir_to_mesa: Add support for variable indexing of temporary arrays.
Eric Anholt [Mon, 10 May 2010 17:06:36 +0000 (10:06 -0700)]
ir_to_mesa: Add support for variable indexing of temporary arrays.

Fixes loop-01.vert, loop-02.vert.

13 years agoir_to_mesa: Clean up some handling of builtins and arrays.
Eric Anholt [Tue, 11 May 2010 01:15:33 +0000 (18:15 -0700)]
ir_to_mesa: Clean up some handling of builtins and arrays.

Constant-index dereferences of arrays should work now.  One test is
regressed, but it should have been failing before this commit, too.

13 years agoir_to_mesa: Add support for loops.
Eric Anholt [Fri, 7 May 2010 19:59:08 +0000 (12:59 -0700)]
ir_to_mesa: Add support for loops.

Fixes CorrectParse1 and the glsl2 loop tests that don't use arrays.