Ian Romanick [Wed, 31 Mar 2010 00:04:48 +0000 (17:04 -0700)]
Forbid array comparisons in GLSL 1.10
This causes the following tests to pass:
glslparsertest/shaders/array3.frag
Ian Romanick [Tue, 30 Mar 2010 23:59:27 +0000 (16:59 -0700)]
Begin handling array declarations
This causes the following tests to pass:
glslparsertest/shaders/array4.frag
glslparsertest/shaders/array5.frag
This causes the following tests to fail. These shaders were
previously failing to compile, but they were all failing for the wrong
reasons.
glslparsertest/shaders/array3.frag
Ian Romanick [Tue, 30 Mar 2010 23:58:19 +0000 (16:58 -0700)]
Implement array type handling
Since all glsl_type objects are flyweights, support is added to track all
known array types. This accounts for most of the changes.
Ian Romanick [Tue, 30 Mar 2010 23:57:29 +0000 (16:57 -0700)]
Add proper wrappers so that C++ code can use hash_table type
Ian Romanick [Tue, 30 Mar 2010 23:56:50 +0000 (16:56 -0700)]
Initial bits of constant expression evaluator
Currently only works for constants. The rest will be added later.
Ian Romanick [Tue, 30 Mar 2010 23:56:22 +0000 (16:56 -0700)]
Add some array declaration tests
Ian Romanick [Tue, 30 Mar 2010 00:40:11 +0000 (17:40 -0700)]
Arrays are not allowed as vertex shader inputs in GLSL 1.30 either
Ian Romanick [Tue, 30 Mar 2010 00:19:13 +0000 (17:19 -0700)]
Add several tests for vertex shader attributes
Ian Romanick [Tue, 30 Mar 2010 00:16:35 +0000 (17:16 -0700)]
Add some checking for vertex shader inputs / attributes
This causes the following tests to pass:
glslparsertest/shaders/attribute.vert
glslparsertest/shaders/attribute1.vert
glslparsertest/shaders/attribute2.vert
Ian Romanick [Mon, 29 Mar 2010 23:44:29 +0000 (16:44 -0700)]
Add parser support for texture rectangle types
Ian Romanick [Mon, 29 Mar 2010 23:32:55 +0000 (16:32 -0700)]
Use call-by-reference for apply_implicit_conversion
I'm not sure if this is a win or not. It makes the code in
apply_implicit_conversion more clear, but it obscures the fact that it
may change the pointers.
Ian Romanick [Mon, 29 Mar 2010 23:25:56 +0000 (16:25 -0700)]
Apply implicit conversions to ?: operator
Ian Romanick [Mon, 29 Mar 2010 23:22:38 +0000 (16:22 -0700)]
Apply implicit conversions to equality operators
Ian Romanick [Mon, 29 Mar 2010 23:20:07 +0000 (16:20 -0700)]
Apply implicit conversions to relational operators
Ian Romanick [Mon, 29 Mar 2010 23:17:56 +0000 (16:17 -0700)]
Apply implicit conversions to arithmetic operators
Ian Romanick [Mon, 29 Mar 2010 23:17:15 +0000 (16:17 -0700)]
Add glsl_type::is_array and glsl_type::is_float queries
Ian Romanick [Mon, 29 Mar 2010 22:36:02 +0000 (15:36 -0700)]
Allow single-component constructors
This causes the following tests to pass:
glslparsertest/shaders/CorrectVersion.V110.frag
shaders/glsl-vs-sqrt-zero.frag
shaders/glsl-vs-sqrt-zero.vert
This causes the following tests to fail. These shaders were
previously failing to compile, but they were all failing for the wrong
reasons.
glslparsertest/shaders/attribute1.vert
glslparsertest/shaders/attribute2.vert
glslparsertest/shaders/main2.vert
Ian Romanick [Mon, 29 Mar 2010 22:34:21 +0000 (15:34 -0700)]
Add tests for :? operator
Ian Romanick [Mon, 29 Mar 2010 22:33:54 +0000 (15:33 -0700)]
Implement HIR conversion for ?: operator
Ian Romanick [Mon, 29 Mar 2010 22:20:42 +0000 (15:20 -0700)]
Add generate_temporary to generate an anonymous temporary
Ian Romanick [Mon, 29 Mar 2010 22:11:05 +0000 (15:11 -0700)]
Implement HIR conversion for ast_nequal and ast_equal
The following tests now pass:
shaders/glsl-unused-varying.frag
shaders/glsl-fs-sqrt-branch.frag
Ian Romanick [Mon, 29 Mar 2010 21:15:05 +0000 (14:15 -0700)]
Add missing break-statements in ast_expression::hir
The ast_conditional case was flowing right into ast_pre_inc.
Ian Romanick [Mon, 29 Mar 2010 21:11:25 +0000 (14:11 -0700)]
Implement ir_if (for if-statments) and conversion from AST
The following tests now pass:
glslparsertest/shaders/if1.frag
glslparsertest/shaders/if2.frag
The following tests that used to pass now fail. It appears that most
of these fail because ast_nequal and ast_equal are not converted to HIR.
shaders/glsl-unused-varying.frag
shaders/glsl-fs-sqrt-branch.frag
Ian Romanick [Mon, 29 Mar 2010 21:11:10 +0000 (14:11 -0700)]
Add several simple if-statement tests
Ian Romanick [Mon, 29 Mar 2010 19:59:02 +0000 (12:59 -0700)]
Add a variable to the symbol table after processing the initializer
Ian Romanick [Mon, 29 Mar 2010 19:52:00 +0000 (12:52 -0700)]
Trivial code cleanup in ir_dereference::is_lvalue
Eric Anholt [Sun, 28 Mar 2010 08:55:38 +0000 (01:55 -0700)]
Add support for builtin radians() and degrees().
Eric Anholt [Sun, 28 Mar 2010 08:46:48 +0000 (01:46 -0700)]
Add definitions for 1.10 built-in uniforms for ff state.
Eric Anholt [Sun, 28 Mar 2010 08:29:18 +0000 (01:29 -0700)]
Don't let swizzles with duplicated components be considered as lvalues.
Fixes swizzle2.frag.
Eric Anholt [Sun, 28 Mar 2010 08:24:55 +0000 (01:24 -0700)]
Reject main() declarations with a non-void return value.
Fixes main3.vert.
Eric Anholt [Sun, 28 Mar 2010 07:56:22 +0000 (00:56 -0700)]
Reject variables marked with attribute in the fragment shader.
Fixes attribute.frag.
Eric Anholt [Sun, 28 Mar 2010 07:36:06 +0000 (00:36 -0700)]
Add the instruction for the parameter variable declarations of builtin funcs.
Matches constructor setup, but I'm not really sure why we make the
variable twice.
Eric Anholt [Sat, 27 Mar 2010 21:04:43 +0000 (14:04 -0700)]
Implement dot() builtin.
Eric Anholt [Sat, 27 Mar 2010 20:56:35 +0000 (13:56 -0700)]
Implement exp2() and log2(), and make ir_unop_exp and ir_unop_log be base e.
Making the base e functions IR operations is not a clear win. i965
doesn't support it, it doesn't look like r600 supports it, but r500
does. It should be easily supportable as a lowering pass, though.
Eric Anholt [Sat, 27 Mar 2010 20:55:04 +0000 (13:55 -0700)]
Implement builtin length() function.
Eric Anholt [Sat, 27 Mar 2010 20:01:51 +0000 (13:01 -0700)]
Add sqrt() builtin as an IR operation.
Following a discussion in #dri-devel, I think this makes more sense
than implementing it as RSQ RCP CMP as Mesa did. The i965 has a
hardware sqrt that should work, and AMD is suppposed to be able to
implement it as RSQ RCP with an alternate floating point mode so that
the 0.0 case is handled like we want.
Eric Anholt [Sat, 27 Mar 2010 19:59:42 +0000 (12:59 -0700)]
Add builtin pow() function.
Eric Anholt [Sat, 27 Mar 2010 19:48:57 +0000 (12:48 -0700)]
Add support for builtin deprecated fs varyings.
Fixes glsl-color.frag.
Eric Anholt [Sat, 27 Mar 2010 19:43:13 +0000 (12:43 -0700)]
Implement some binary gentype builtin functions.
Eric Anholt [Sat, 27 Mar 2010 19:25:20 +0000 (12:25 -0700)]
Implement additional unary gentype builtins.
Eric Anholt [Sat, 27 Mar 2010 01:20:30 +0000 (18:20 -0700)]
Implement the first builtin function: exp().
Ian Romanick [Mon, 29 Mar 2010 00:03:16 +0000 (17:03 -0700)]
Generate more correctly diagnostics from some invalid initializers
Ian Romanick [Sun, 28 Mar 2010 01:56:53 +0000 (18:56 -0700)]
Initial bits to process initializers in variable declarations
As a result, the following tests pass:
glslparsertest/array3.frag
glslparsertest/CGStandardLibrary.frag
glslparsertest/ConstantConversions.frag
glslparsertest/constructor1.frag
glslparsertest/constructor2.frag
glslparsertest/constructor3.V110.frag
glslparsertest/dataType4.frag
glslparsertest/dataType5.frag
glslparsertest/dataType13.frag
glslparsertest/dataType19.frag
glslparsertest/matrix.V110.frag
glslparsertest/parser7.frag
glslparsertest/swizzle3.frag
The following tests also pass, but it is just by dumb luck. In these
cases the shader fails to compile, but it fails for the wrong reason:
glslparsertest/array6.frag
glslparsertest/comma2.frag
glslparsertest/conditional1.frag
glslparsertest/conditional2.frag
glslparsertest/conditional3.frag
glslparsertest/constFunc.frag
glslparsertest/ParseTest3.frag
glslparsertest/ParseTest4.frag
glslparsertest/varying3.frag
glslparsertest/parser8.frag (also segfaults)
glslparsertest/parser9.frag (also segfaults)
The following tests now fail. As far as I can tell, these are all
cases where the shader was failing to compile, but it was failing for
the wrong reason.
glslparsertest/CorrectMatComma.frag
glslparsertest/CorrectModule.frag
glslparsertest/CorrectSwizzle2.vert
glslparsertest/shaders/glsl-fs-bug25902.frag
Ian Romanick [Sat, 27 Mar 2010 01:05:27 +0000 (18:05 -0700)]
Reject uniform initializers in GLSL 1.10 mode
Now both glslparsertest/dataType3.frag and glslparsertest/dataType2.frag pass.
Eric Anholt [Fri, 26 Mar 2010 23:37:53 +0000 (16:37 -0700)]
Set variables with the sampler base type to read only.
Fixes increment3.frag.
Eric Anholt [Fri, 26 Mar 2010 23:37:22 +0000 (16:37 -0700)]
Set the var_type for arrays to error_type until it's implemented.
"Fixes" segfaults in array tests.
Eric Anholt [Fri, 26 Mar 2010 19:16:54 +0000 (12:16 -0700)]
Add support for ast_to_hir of pre inc/dec.
Eric Anholt [Fri, 26 Mar 2010 19:14:54 +0000 (12:14 -0700)]
Add support for ast_to_hir of post inc/dec.
Eric Anholt [Fri, 26 Mar 2010 22:51:45 +0000 (15:51 -0700)]
Make read-only variables not be considered lvalues.
This should fix tests trying to assign to builtin "in" variables, and
will also be relied on for post_inc/dec handling.
Eric Anholt [Fri, 26 Mar 2010 19:07:44 +0000 (12:07 -0700)]
Add constructors for immediate hir constants.
This will make ast_to_hir for inc/dec easier.
Eric Anholt [Fri, 26 Mar 2010 18:57:46 +0000 (11:57 -0700)]
Add support for %= in ast_to_hir.
Eric Anholt [Fri, 26 Mar 2010 18:53:37 +0000 (11:53 -0700)]
Factor out assignment setup in ast_to_hir.
Eric Anholt [Fri, 26 Mar 2010 18:00:07 +0000 (11:00 -0700)]
IR print visitor: Print out something for the operator.
Ian Romanick [Sat, 27 Mar 2010 00:42:10 +0000 (17:42 -0700)]
IR print visitor: Print expressions a little better
Ian Romanick [Sat, 27 Mar 2010 00:38:58 +0000 (17:38 -0700)]
Initial implementation of constructor handling code
All of the scalar, vector, and matrix constructors *except* "from
bool" constructors should be handled. Array and structure
constructors are also not yet handled.
Ian Romanick [Sat, 27 Mar 2010 00:30:30 +0000 (17:30 -0700)]
IR print visitor: print function calls
Ian Romanick [Sat, 27 Mar 2010 00:29:29 +0000 (17:29 -0700)]
Add hack ir_call::callee_name to get the name of the called function
Ian Romanick [Sat, 27 Mar 2010 00:28:52 +0000 (17:28 -0700)]
Be sure to set ir_function_signature::definition for constructors
Ian Romanick [Sat, 27 Mar 2010 00:19:47 +0000 (17:19 -0700)]
Add ir_call::iterator to iterate over actual parameters
Ian Romanick [Sat, 27 Mar 2010 00:17:41 +0000 (17:17 -0700)]
Add missing parenthesis in foreach_iter
This allows uses like 'foreach_iter(exec_list_iterator, iter, *list_ptr)'.
Ian Romanick [Fri, 26 Mar 2010 23:41:43 +0000 (16:41 -0700)]
Add some simple constructor tests
Ian Romanick [Fri, 26 Mar 2010 23:11:48 +0000 (16:11 -0700)]
Add unary operator to convert unsigned integer to float
Ian Romanick [Fri, 26 Mar 2010 18:43:36 +0000 (11:43 -0700)]
Slightly change the representation of numeric types
For numeric types, vector_elements and matrix_columns must be at least
1. Previously matrix_columns was 0 for vectors, and both were 0 for
scalars. This change simplifies things in some places.
Ian Romanick [Fri, 26 Mar 2010 18:13:43 +0000 (11:13 -0700)]
Add glsl_type::components to query total number of components in a type
Ian Romanick [Fri, 26 Mar 2010 21:41:32 +0000 (14:41 -0700)]
Replace remaining use of is_error_type with glsl_type::is_error
Ian Romanick [Fri, 26 Mar 2010 21:40:37 +0000 (14:40 -0700)]
Replace remaining uses of is_numeric_base_type with glsl_type::is_numeric
Ian Romanick [Fri, 26 Mar 2010 21:38:37 +0000 (14:38 -0700)]
Replace is_integer_base_type macro with glsl_type::is_integer method
Ian Romanick [Fri, 26 Mar 2010 21:33:41 +0000 (14:33 -0700)]
Make glsl_*_type glsl_type class static data
Ian Romanick [Fri, 26 Mar 2010 21:27:23 +0000 (14:27 -0700)]
Use glsl_type::is_error instead of comparison with glsl_error_type pointer
Kenneth Graunke [Fri, 26 Mar 2010 08:20:08 +0000 (01:20 -0700)]
Move swizzles out of ir_dereference and into their own class.
Also turn generate_swizzle into a static "create" method of the new
class; we'll want to use it for the IR reader as well.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Kenneth Graunke [Fri, 26 Mar 2010 07:25:36 +0000 (00:25 -0700)]
Add new abstract ir_rvalue class; rework accordingly.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Kenneth Graunke [Fri, 26 Mar 2010 06:30:28 +0000 (23:30 -0700)]
Replace "mode" type tag with virtual as_foo() downcasting functions.
These should work well even in a non-flat IR hierarchy.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Ian Romanick [Fri, 26 Mar 2010 01:41:41 +0000 (18:41 -0700)]
Merge branch 'anholt'
Ian Romanick [Fri, 26 Mar 2010 01:38:28 +0000 (18:38 -0700)]
IR print visitor: Finish printing constants
Ian Romanick [Fri, 26 Mar 2010 01:29:25 +0000 (18:29 -0700)]
IR print visitor: Remove most of the newlines from the printed output
This makes it a lot easier to read... if you have a really wide display.
Ian Romanick [Fri, 26 Mar 2010 01:25:37 +0000 (18:25 -0700)]
IR print visitor: print expressions
Not quite complete. The operator is not yet printed.
Ian Romanick [Fri, 26 Mar 2010 00:38:13 +0000 (17:38 -0700)]
IR print visitor: print ir_dereference instructions
Also make a slight change to ir_variable. The ir_dereference tracks
the number of nested dereferences. If an ir_variable is visited and
the count is non-zero, just print the name of the variable.
Ian Romanick [Fri, 26 Mar 2010 00:04:26 +0000 (17:04 -0700)]
IR print visitor: Less newlines when printing ir_variables
Eric Anholt [Fri, 26 Mar 2010 00:02:15 +0000 (17:02 -0700)]
Fix assignment operators: *=, /=, +=, -=.
Basically, replace everything different from operator_assign other
than the creation of the rhs value from the lvalue and rvalue with the
contents of operator_assign. Fixes a segfault in
CorrectSwizzle1.frag, and fixes parser10.frag.
Ian Romanick [Fri, 26 Mar 2010 00:02:22 +0000 (17:02 -0700)]
Before generating HIR for user code, generate constructors for built-in types
Ian Romanick [Fri, 26 Mar 2010 00:01:15 +0000 (17:01 -0700)]
Add dereference constructor for array element dereferences
This should have gone in before the previous commit.
Ian Romanick [Thu, 25 Mar 2010 23:58:45 +0000 (16:58 -0700)]
Add functions to generate constructors for built-in types.
Currently only vector and matrix types are supported. Structure types will be
added later.
Eric Anholt [Thu, 25 Mar 2010 23:47:49 +0000 (16:47 -0700)]
Fix up the operator printing for the logic or and beyond, check array bounds.
Ian Romanick [Thu, 25 Mar 2010 23:02:41 +0000 (16:02 -0700)]
Merge branch 'anholt'
Eric Anholt [Thu, 25 Mar 2010 21:48:25 +0000 (14:48 -0700)]
Set up fragment shader builtin variables.
Eric Anholt [Thu, 25 Mar 2010 21:37:25 +0000 (14:37 -0700)]
Make the standalone parser return an exit code so we can automate testing.
Ian Romanick [Thu, 25 Mar 2010 20:41:32 +0000 (13:41 -0700)]
More matrix compilation tests
Ian Romanick [Thu, 25 Mar 2010 20:22:19 +0000 (13:22 -0700)]
Merge branch 'kwg'
Ian Romanick [Thu, 25 Mar 2010 20:19:13 +0000 (13:19 -0700)]
Fix matrix dimensioning
Newb GL mistake: matrices in GL are column-major. This means that
vector_elements is the number of rows. Making these changes causes
matrix-08.glsl to pass.
Ian Romanick [Thu, 25 Mar 2010 20:06:58 +0000 (13:06 -0700)]
Use glsl_type::row_type and glsl_type::column type in arithmetic_result_type
This substantially clarifies the code for matching matrix types. It
also eliminates some uses of glsl_type member data.
Ian Romanick [Thu, 25 Mar 2010 20:05:43 +0000 (13:05 -0700)]
Replace several glsl_type field comparisons with a single pointer comparison
This simplifies the process of matching function parameter types.
More simplifications are probably possible here, but arrays and
structures need to be implemented first.
Ian Romanick [Thu, 25 Mar 2010 18:42:45 +0000 (11:42 -0700)]
Add queries to get the glsl_type of a row or column of a matrix
Kenneth Graunke [Thu, 25 Mar 2010 18:22:42 +0000 (11:22 -0700)]
Widen num_components to handle vec4 correctly.
Kenneth Graunke [Tue, 23 Mar 2010 22:17:49 +0000 (15:17 -0700)]
Typo fixes: de/re/ference.
Ian Romanick [Thu, 25 Mar 2010 00:53:53 +0000 (17:53 -0700)]
Replace several field comparisons with a single pointer comparison
The only way the specified type fields can match is if the types are the
same. Previous tests (and assertions) have filtered away all other possible
cases.
Ian Romanick [Thu, 25 Mar 2010 00:47:20 +0000 (17:47 -0700)]
Remove unused _mesa_glsl_get_vector_type
This function has been completely replaced by glsl_type::get_instance.
Ian Romanick [Thu, 25 Mar 2010 00:46:39 +0000 (17:46 -0700)]
Use glsl_type::get_instance instead of _mesa_glsl_get_vector_type
Ian Romanick [Thu, 25 Mar 2010 00:45:14 +0000 (17:45 -0700)]
Use glsl_type::get_instance instead of symbol table look-up
Ian Romanick [Thu, 25 Mar 2010 00:42:59 +0000 (17:42 -0700)]
Add some matrix math tests
Ian Romanick [Thu, 25 Mar 2010 00:11:30 +0000 (17:11 -0700)]
Add glsl_type::get_instance method
Gets the singleton corresponding to a particular scalar, vector, or
matrix type.