Eric Anholt [Sat, 31 Jul 2010 00:04:49 +0000 (17:04 -0700)]
glsl2: Add new tree grafting optimization pass.
Eric Anholt [Fri, 30 Jul 2010 23:05:27 +0000 (16:05 -0700)]
glsl2: Factor out the variable refcounting part of ir_dead_code.cpp.
Aras Pranckevicius [Thu, 29 Jul 2010 12:35:22 +0000 (15:35 +0300)]
glsl2: Fix stack smash when ternary selection is used.
Eric Anholt [Fri, 30 Jul 2010 17:20:34 +0000 (10:20 -0700)]
glsl2: Fix the implementation of atan(y, x).
So many problems here. One is that we can't do the quadrant handling
for all the channels at the same time, so we call the float(y, x)
version multiple times. I'd also left out the x == 0 handling. Also,
the quadrant handling was broken for y == 0, so there was a funny
discontinuity on the +x side if you plugged in obvious values to test.
I generated the atan(float y, float x) code from a short segment of
GLSL and pasted it in by hand. It would be nice to automate that
somehow.
Fixes:
glsl-fs-atan-1
glsl-fs-atan-2
Carl Worth [Thu, 29 Jul 2010 23:39:36 +0000 (16:39 -0700)]
ast: Initialize location data in constructor of all ast_node objects.
This prevents using uninitialized data in _msea_glsl_error in some
cases, (including at least 6 piglit tests). Thanks to valgrind for
pointing out the problem!
Eric Anholt [Fri, 30 Jul 2010 21:44:09 +0000 (14:44 -0700)]
ir_to_mesa: Add the function name as a comment to BGNSUB and ENDSUB.
Eric Anholt [Fri, 30 Jul 2010 18:24:23 +0000 (11:24 -0700)]
glsl2: Update the callee pointer of calls to newly-linked-in functions.
Otherwise, ir_function_inlining will see the body of the function from
the unlinked version of the shader, which won't have had the lowering
passes done on it or linking's variable remapping.
Kenneth Graunke [Fri, 30 Jul 2010 20:30:11 +0000 (13:30 -0700)]
glsl2: Initialize ir_function_signature::is_built_in.
Fixes a valgrind error.
Kenneth Graunke [Fri, 30 Jul 2010 20:24:50 +0000 (13:24 -0700)]
glcpp: Don't look for backslashes before the beginning of the string.
Fixes a valgrind error.
Eric Anholt [Thu, 29 Jul 2010 22:49:14 +0000 (15:49 -0700)]
glsl2: Do ir_if_return on the way out, not the way in.
The problem with doing it on the way in is that for a function with
multiple early returns, we'll move an outer block in, then restart the
pass, then move the two inside returns out, then never move outer
blocks in again because the remaining early returns are inside an else
block and they don't know that there's a return just after their
block. By going inside-out, we get the early returns stacked up so
that they all move out with a series of
move_returns_after_block().
Fixes (on i965):
glsl-fs-raytrace-bug27060
glsl-vs-raytrace-bug26691
Eric Anholt [Thu, 29 Jul 2010 20:29:17 +0000 (13:29 -0700)]
glsl2: Make sure functions end with a return before doing ir_if_return.
This catches a few remaining functions that weren't getting inlined,
generally operating on global or out variables and using an early
return to skip work when possible.
Fixes for i965:
glsl1-function with early return (3)
Eric Anholt [Thu, 29 Jul 2010 19:36:06 +0000 (12:36 -0700)]
glsl2: Make ir_if_return handle if () { return } else { not return }
This makes many remaining functions inlinable.
Fixes for i965:
glsl1-function with early return (1)
glsl1-function with early return (2)
Eric Anholt [Thu, 29 Jul 2010 19:15:04 +0000 (12:15 -0700)]
glsl2: Refactor a bit of ir_if_return for the next changes.
Eric Anholt [Thu, 29 Jul 2010 22:17:23 +0000 (15:17 -0700)]
ir_to_mesa: Don't emit a duplicate return at the end of a function.
It was harmless, but ugly.
Eric Anholt [Thu, 29 Jul 2010 22:04:45 +0000 (15:04 -0700)]
glsl2: Allow use of _mesa_print_ir without a parse state on hand.
Eric Anholt [Thu, 29 Jul 2010 21:54:01 +0000 (14:54 -0700)]
ir_constant_variable: Don't mark variable from outside our scope as constant.
Fixes (with software, except for alpha):
glsl1-function with early return(3)
Eric Anholt [Thu, 29 Jul 2010 21:36:59 +0000 (14:36 -0700)]
glsl2: When dumping IR for debug, indent nested blocks.
No more trying to match parens in my head when looking at the body of
a short function containing an if statement.
Eric Anholt [Thu, 29 Jul 2010 21:20:39 +0000 (14:20 -0700)]
glsl2: When dumping IR for debug, skip all the empty builtin prototypes.
Eric Anholt [Thu, 29 Jul 2010 20:52:25 +0000 (13:52 -0700)]
glsl2: Fix spelling of "sentinel."
Eric Anholt [Thu, 29 Jul 2010 20:50:17 +0000 (13:50 -0700)]
glsl2: Fix spelling of "initializer."
Eric Anholt [Thu, 29 Jul 2010 20:42:39 +0000 (13:42 -0700)]
glsl2: Remove an inlined unvalued return statement.
We already have asserts that it was the last call in the function, so
it's safe to remove after it got cloned in.
Fixes:
glsl-fs-functions-4.
Eric Anholt [Thu, 29 Jul 2010 00:36:07 +0000 (17:36 -0700)]
glsl2: Actually fix glsl-version-define.
Eric Anholt [Wed, 28 Jul 2010 23:58:39 +0000 (16:58 -0700)]
glcpp: Add __VERSION__ define to the current language version.
Fixes:
glsl-version-define
glsl-version-define-110
glsl-version-define-120
Eric Anholt [Wed, 28 Jul 2010 23:53:51 +0000 (16:53 -0700)]
glcpp: Print integer tokens as decimal, not hex.
Eric Anholt [Wed, 28 Jul 2010 23:11:26 +0000 (16:11 -0700)]
glsl2: Make lowp, mediump, highp, and precision identifiers pre-1.20.
Fixes glsl-precision-110.
Kenneth Graunke [Wed, 28 Jul 2010 23:44:56 +0000 (16:44 -0700)]
glsl2/Makefile: Append to DEFINES rather than replacing them.
Otherwise, we lose DEBUG, which causes mtypes.h to set NDEBUG, which
causes assertions to not happen, which is no fun for anyone.
Eric Anholt [Wed, 28 Jul 2010 22:42:35 +0000 (15:42 -0700)]
ir_to_mesa: Respect the driver if it rejects a shader.
Kenneth Graunke [Wed, 28 Jul 2010 18:58:27 +0000 (11:58 -0700)]
glsl2: Fix outerProduct builtin.
The type signatures were completely backwards.
Kenneth Graunke [Fri, 23 Jul 2010 20:24:09 +0000 (13:24 -0700)]
ir_constant_expression: Add support for the "outerProduct" builtin.
Kenneth Graunke [Fri, 23 Jul 2010 19:36:50 +0000 (12:36 -0700)]
ir_constant_expression: Add support for the "mix" builtin.
Both 1.10 and 1.30 variants.
Kenneth Graunke [Thu, 22 Jul 2010 22:34:49 +0000 (15:34 -0700)]
ir_constant_expression: Add support for the "transpose" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 05:37:50 +0000 (22:37 -0700)]
ir_constant_expression: Add support for the "smoothstep" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 05:23:17 +0000 (22:23 -0700)]
ir_constant_expression: Add support for the "clamp" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 04:53:40 +0000 (21:53 -0700)]
ir_constant_expression: Add support for the "step" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 04:46:16 +0000 (21:46 -0700)]
ir_constant_expression: Add support for the "faceforward" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 03:33:42 +0000 (20:33 -0700)]
ir_constant_expression: Add support for the "refract" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 03:24:29 +0000 (20:24 -0700)]
ir_constant_expression: Add support for the "reflect" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 03:19:32 +0000 (20:19 -0700)]
ir_constant_expression: Add support for the "normalize" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 03:03:55 +0000 (20:03 -0700)]
ir_constant_expression: Add support for the "matrixCompMult" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 03:12:43 +0000 (20:12 -0700)]
ir_constant_expression: Simplify code that implements the "dot" builtin.
There's no need to use an ir_expression; we have a handy C function.
Kenneth Graunke [Thu, 22 Jul 2010 01:26:00 +0000 (18:26 -0700)]
ir_constant_expression: Add support for the "length" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 03:09:21 +0000 (20:09 -0700)]
ir_constant_expression: Extract dot product calculation for reuse.
Kenneth Graunke [Thu, 22 Jul 2010 01:18:16 +0000 (18:18 -0700)]
ir_constant_expression: Remove support for dot products of integers.
This shouldn't be required since dot is only defined for floating point
types, even in GLSL 4.0.
Kenneth Graunke [Thu, 22 Jul 2010 01:08:05 +0000 (18:08 -0700)]
ir_constant_expression: Add support for the "greaterThanEqual" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 01:07:43 +0000 (18:07 -0700)]
ir_constant_expression: Add support for the "greaterThan" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 01:07:09 +0000 (18:07 -0700)]
ir_constant_expression: Add support for the "lessThanEqual" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 01:06:18 +0000 (18:06 -0700)]
ir_constant_expression: Add support for the "lessThan" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 01:05:36 +0000 (18:05 -0700)]
ir_constant_expression: Add support for the "notEqual" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 01:04:22 +0000 (18:04 -0700)]
ir_constant_expression: Add support for the "equal" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 01:01:22 +0000 (18:01 -0700)]
ir_constant_expression: Add support for the "distance" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 00:59:00 +0000 (17:59 -0700)]
ir_constant_expression: Add support for the "degrees" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 00:58:03 +0000 (17:58 -0700)]
ir_constant_expression: Add support for the "radians" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 00:51:23 +0000 (17:51 -0700)]
ir_constant_expression: Add support for the "tanh" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 00:50:50 +0000 (17:50 -0700)]
ir_constant_expression: Add support for the "tan" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 00:49:56 +0000 (17:49 -0700)]
ir_constant_expression: Add support for the "sinh" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 00:47:55 +0000 (17:47 -0700)]
ir_constant_expression: Add support for the "cosh" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 00:42:23 +0000 (17:42 -0700)]
ir_constant_expression: Add support for "atan" builtins.
Kenneth Graunke [Thu, 22 Jul 2010 00:39:47 +0000 (17:39 -0700)]
ir_constant_expression: Add support for "acos" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 00:38:38 +0000 (17:38 -0700)]
ir_constant_expression: Add support for "asin" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 00:35:52 +0000 (17:35 -0700)]
ir_constant_expression: Add support for "any" builtin.
Kenneth Graunke [Thu, 22 Jul 2010 00:34:32 +0000 (17:34 -0700)]
ir_constant_expression: Add support for "all" builtin.
Kenneth Graunke [Wed, 21 Jul 2010 23:57:53 +0000 (16:57 -0700)]
ir_constant_expression: Implement builtins that wrap an expression.
These builtin functions are represented by ir_expression_operations, so
we can just create one of those and ask for its value.
Kenneth Graunke [Wed, 21 Jul 2010 23:57:10 +0000 (16:57 -0700)]
ir_constant_expression: Add support for builtins dFdx, dFdy, and fwidth.
These always return zero (the derivative of a constant).
Kenneth Graunke [Fri, 23 Jul 2010 01:29:29 +0000 (18:29 -0700)]
ast_function: Set constant_value on return value temporaries in 1.20+.
Kenneth Graunke [Wed, 21 Jul 2010 23:55:46 +0000 (16:55 -0700)]
ir_constant_expression: Stub out support for constant builtins.
Kenneth Graunke [Thu, 22 Jul 2010 04:56:13 +0000 (21:56 -0700)]
ir_constant_expression: Use Mesa's MIN2/MAX2 instead of our own.
Kenneth Graunke [Wed, 28 Jul 2010 19:20:38 +0000 (12:20 -0700)]
ir_algebraic: Use ir_constant::zero.
Kenneth Graunke [Wed, 21 Jul 2010 22:54:15 +0000 (15:54 -0700)]
glsl2: Add new ir_constant::zero static method.
This conveniently creates a zero value of whatever type you want.
Eric Anholt [Wed, 28 Jul 2010 21:58:31 +0000 (14:58 -0700)]
glsl2: Add the define for ARB_fragment_coord_conventions when present.
Fixes:
glsl-arb-fragment-coord-conventions-define
Eric Anholt [Wed, 28 Jul 2010 21:41:51 +0000 (14:41 -0700)]
glsl2: Add support for redeclaring layout of gl_FragCoord for ARB_fcc.
Fixes:
glsl-arb-fragment-coord-conventions
Ian Romanick [Thu, 1 Jul 2010 00:48:09 +0000 (17:48 -0700)]
glsl2: Perform some semantic checking of ARB_fcc layout qualifiers
The rest cannot be handled until built-in variables (i.e.,
gl_FragCoord) can be redeclared to add qualifiers.
Ian Romanick [Thu, 1 Jul 2010 00:30:03 +0000 (17:30 -0700)]
glsl2: Parser support for GL_ARB_fragment_coord_conventions
Eric Anholt [Wed, 28 Jul 2010 20:52:23 +0000 (13:52 -0700)]
glsl2: Fail linking where the FS reads a varying that the VS doesn't write.
Fixes:
glsl1-varying read but not written
glsl1-varying var mismatch
Eric Anholt [Wed, 28 Jul 2010 20:42:36 +0000 (13:42 -0700)]
glsl2: Unmark unwritten varyings as varying.
This fixes an assertion failure in ir_to_mesa, and the varying won't
take up varying space.
Carl Worth [Wed, 28 Jul 2010 19:33:56 +0000 (12:33 -0700)]
glcpp: Add generated source files.
This is now consistent with other usage of flex/bison througout mesa,
(which is that these generated files are added to source control so
that the build system does not require external tools like flex/bison
for non-developers).
Carl Worth [Wed, 28 Jul 2010 19:30:07 +0000 (12:30 -0700)]
glsl: Ignore glsl_compiler and glsl_parser.output files.
These are generated files where we can do the sane thing, and keep
them out of version control.
Carl Worth [Wed, 28 Jul 2010 19:27:33 +0000 (12:27 -0700)]
glsl: Add generated files from flex/bison.
The mesa build environment does not (currently) accept external
dependencies such as flex and bison. The compromise is to commit the
generated output files, (in spite of the pain that comes from having
generated files under version control).
Carl Worth [Wed, 28 Jul 2010 19:19:44 +0000 (12:19 -0700)]
glcpp: Add expected output for a recently-added test.
I simply forgot to add this file when adding the test case originally.
Eric Anholt [Wed, 28 Jul 2010 15:13:53 +0000 (08:13 -0700)]
ir_to_mesa: Add remaining state variable (builtin uniforms) support.
Fixes:
glsl1-GL state variable reference (diffuse product)
glsl1-GL state variable reference (gl_FrontMaterial.ambient)
glsl1-GL state variable reference (gl_LightSource[0].diffuse)
glsl1-GL state variable reference (point attenuation)
glsl1-GL state variable reference (point size)
glsl1-linear fog
Eric Anholt [Wed, 28 Jul 2010 19:23:51 +0000 (12:23 -0700)]
glsl2: Add the remaining builtin uniforms.
Eric Anholt [Wed, 28 Jul 2010 15:18:59 +0000 (08:18 -0700)]
glsl2: Size builtin arrays according to the context constants.
Cleans up some of the FINISHMEs in this file.
Carl Worth [Wed, 28 Jul 2010 18:10:52 +0000 (11:10 -0700)]
glcpp: Explicitly expect 0 shift/reduce conflicts.
The "%expect 0" construct will make bison emit an error if any future
changes to the grammar introduce shift/reduce conflicts, (without also
increasing the number after "%expect").
Carl Worth [Wed, 28 Jul 2010 18:07:46 +0000 (11:07 -0700)]
glcpp: Remove 2 shift/reduce conflicts from the grammar.
Since we have productions to turn "defined FOO" and "defined ( FOO )"
into a conditional_token we don't need to list DEFINED as an operator
as well. Doing so just introduces the shift/reduce ambiguity with no
benefit.
Eric Anholt [Tue, 27 Jul 2010 22:25:07 +0000 (15:25 -0700)]
glsl2: When stealing var->constant_value, steal its children as well.
Fixes:
glsl1-GLSL 1.20 uniform array constructor
Eric Anholt [Tue, 27 Jul 2010 22:10:38 +0000 (15:10 -0700)]
ir_to_mesa: Provide a restricted type size to _mesa_add_uniform.
Fixes:
glsl-uniform-out-of-bounds.
Eric Anholt [Tue, 27 Jul 2010 21:32:21 +0000 (14:32 -0700)]
glsl2: Actually use the linked dead code eliminator.
I managed to revert the change from unlinked at some point while
cleaning up the changes. glsl-fs-raytrace-bug27060 drops from 389
instructions to 370.
Eric Anholt [Tue, 27 Jul 2010 19:10:50 +0000 (12:10 -0700)]
glsl2: Don't dereference a NULL var in CE handling during a compile error.
If an undeclared variable was dereferenced in an expression that
needed constant expression handling, we would walk off a null ir->var
pointer.
Fixes:
glsl1-TIntermediate::addUnaryMath
Eric Anholt [Tue, 27 Jul 2010 18:58:32 +0000 (11:58 -0700)]
ir_to_mesa: Add support for array constants.
Fixes:
glsl1-GLSL 1.20 array constructor 1
glsl1-GLSL 1.20 array constructor 2
glsl1-GLSL 1.20 array.length()
glsl1-GLSL 1.20 const array constructor 1
glsl1-GLSL 1.20 const array constructor 2
Eric Anholt [Tue, 27 Jul 2010 18:31:12 +0000 (11:31 -0700)]
glsl2: Use ir_dead_code's linked version after linking.
glsl-fs-raytrace-bug27060 goes from 485 Mesa IR instructions to 389
before Mesa IR optimization.
Eric Anholt [Tue, 27 Jul 2010 18:29:17 +0000 (11:29 -0700)]
glsl2: Fix the linked version of ir_dead_code.
If we don't walk into functions, we won't see any usage of variables
in the functions.
Eric Anholt [Tue, 27 Jul 2010 18:28:26 +0000 (11:28 -0700)]
glsl2: Make the dead code handler make its own talloc context.
This way, we don't need to pass in a parse state, and the context
doesn't grow with the number of passes through optimization.
Eric Anholt [Tue, 27 Jul 2010 18:45:27 +0000 (11:45 -0700)]
mesa: Allow large temporary indices coming into the temporary reg allocator.
This gets glsl-vs-raytrace, glsl-fs-raytrace running on the new
compiler.
Eric Anholt [Tue, 27 Jul 2010 18:14:59 +0000 (11:14 -0700)]
glsl2: Talloc type names.
Otherwise, we end up losing structure names after compile time, and
dumping IR often ends up reporting some other mysterious string.
Eric Anholt [Tue, 27 Jul 2010 17:39:40 +0000 (10:39 -0700)]
ir_to_mesa: Set the swizzle on constant struct src regs.
MESA_GLSL=nopt now produces believable output for glsl-fs-raytrace.
Eric Anholt [Tue, 27 Jul 2010 17:08:38 +0000 (10:08 -0700)]
ir_to_mesa: Fix stray "break" that broke functions of >1 argument.
Eric Anholt [Tue, 27 Jul 2010 05:50:29 +0000 (22:50 -0700)]
glsl2: Add optimization pass for algebraic simplifications.
This cleans up the assembly output of almost all the non-logic tests
glsl-algebraic-*. glsl-algebraic-pow-two needs love (basically,
flattening to a temporary and squaring it).
Eric Anholt [Tue, 27 Jul 2010 16:01:30 +0000 (09:01 -0700)]
ir_validate: Check the types of expression operations.
Eric Anholt [Tue, 27 Jul 2010 07:18:33 +0000 (00:18 -0700)]
glsl2: Fix missing visit_continue return in ir_validate.
Eric Anholt [Tue, 27 Jul 2010 02:36:53 +0000 (19:36 -0700)]
ir_to_mesa: Fix up handling of void function returns.
void functions have a type of glsl_type::void_type, not a null type.
Eric Anholt [Tue, 27 Jul 2010 02:41:23 +0000 (19:41 -0700)]
ir_to_mesa: Actually allocate the right size for constant matrix temps.