mesa.git
14 years agollvmpipe: remove lp_setup_coef*.c files from Makefile
Brian Paul [Mon, 18 Oct 2010 13:59:02 +0000 (07:59 -0600)]
llvmpipe: remove lp_setup_coef*.c files from Makefile

14 years agoegl/i965: include inline_wrapper_sw_helper.h
Victor Tseng [Mon, 18 Oct 2010 13:55:37 +0000 (07:55 -0600)]
egl/i965: include inline_wrapper_sw_helper.h

Signed-off-by: Brian Paul <brianp@vmware.com>
14 years agoglsl: Don't return NULL IR for erroneous bit-shift operators.
Kenneth Graunke [Mon, 18 Oct 2010 07:24:46 +0000 (00:24 -0700)]
glsl: Don't return NULL IR for erroneous bit-shift operators.

Existing code relies on IR being generated (possibly with error type)
rather than returning NULL.  So, don't break - go ahead and generate the
operation.  As long as an error is flagged, things will work out.

Fixes fd.o bug #30914.

14 years agor600g: retrieve tiling info from kernel for shared buffers.
Dave Airlie [Sun, 17 Oct 2010 23:45:58 +0000 (09:45 +1000)]
r600g: retrieve tiling info from kernel for shared buffers.

we need to know if the back is tiled so we can blit from it properly.

14 years agor600g: fix transfer function for tiling.
Dave Airlie [Wed, 13 Oct 2010 00:44:46 +0000 (10:44 +1000)]
r600g: fix transfer function for tiling.

this makes readback with tiled back work better.

14 years agor600g: attempt to cleanup depth blit
Dave Airlie [Mon, 18 Oct 2010 02:05:27 +0000 (12:05 +1000)]
r600g: attempt to cleanup depth blit

cleanup what I'm nearly sure is unnecessary work in the depth blit code.

14 years agor600g: depth needs to bound to ds
Dave Airlie [Mon, 18 Oct 2010 03:23:34 +0000 (13:23 +1000)]
r600g: depth needs to bound to ds

14 years agor600g: fix typo in tiling setup cb code.
Dave Airlie [Mon, 18 Oct 2010 02:05:38 +0000 (12:05 +1000)]
r600g: fix typo in tiling setup cb code.

14 years agor600g: do proper tracking of views/samplers.
Dave Airlie [Mon, 18 Oct 2010 02:04:57 +0000 (12:04 +1000)]
r600g: do proper tracking of views/samplers.

we need to do pretty much what r300g does in for this, this fixes some
issues seen while working on tiling.

14 years agollvmpipe: remove unused arg from jit_setup_tri function
Keith Whitwell [Mon, 18 Oct 2010 02:23:40 +0000 (19:23 -0700)]
llvmpipe: remove unused arg from jit_setup_tri function

14 years agollvmpipe: remove unused file
Keith Whitwell [Mon, 18 Oct 2010 02:11:47 +0000 (19:11 -0700)]
llvmpipe: remove unused file

14 years agoMerge remote branch 'origin/master' into lp-setup-llvm
Keith Whitwell [Mon, 18 Oct 2010 02:03:42 +0000 (19:03 -0700)]
Merge remote branch 'origin/master' into lp-setup-llvm

Conflicts:
src/gallium/drivers/llvmpipe/lp_setup_coef.c
src/gallium/drivers/llvmpipe/lp_setup_coef.h
src/gallium/drivers/llvmpipe/lp_setup_coef_intrin.c
src/gallium/drivers/llvmpipe/lp_setup_point.c
src/gallium/drivers/llvmpipe/lp_setup_tri.c
src/gallium/drivers/llvmpipe/lp_state_derived.c
src/gallium/drivers/llvmpipe/lp_state_fs.h

14 years agollvmpipe: fail cleanly on malloc failure in lp_setup_alloc_triangle
Keith Whitwell [Mon, 18 Oct 2010 01:48:11 +0000 (18:48 -0700)]
llvmpipe: fail cleanly on malloc failure in lp_setup_alloc_triangle

14 years agollvmpipe: remove setup fallback path
Keith Whitwell [Mon, 18 Oct 2010 01:29:28 +0000 (18:29 -0700)]
llvmpipe: remove setup fallback path

14 years agogallivm: Comment lp_build_insert_new_block().
José Fonseca [Sun, 17 Oct 2010 20:41:45 +0000 (13:41 -0700)]
gallivm: Comment lp_build_insert_new_block().

14 years agor600g: set tiling bits in hw state
Dave Airlie [Fri, 8 Oct 2010 01:57:04 +0000 (11:57 +1000)]
r600g: set tiling bits in hw state

14 years agor600g: get tiling info from kernel
Dave Airlie [Fri, 8 Oct 2010 01:56:43 +0000 (11:56 +1000)]
r600g: get tiling info from kernel

14 years agor600g: add defines for tiling
Dave Airlie [Fri, 8 Oct 2010 01:56:12 +0000 (11:56 +1000)]
r600g: add defines for tiling

14 years agor600g: switch to a common formats.h file since they are in different regs
Dave Airlie [Wed, 6 Oct 2010 02:49:26 +0000 (12:49 +1000)]
r600g: switch to a common formats.h file since they are in different regs

14 years agollvmpipe: Return non-zero exit code for lp_test_round failures.
Vinson Lee [Sun, 17 Oct 2010 21:09:53 +0000 (14:09 -0700)]
llvmpipe: Return non-zero exit code for lp_test_round failures.

14 years agollvmpipe: Initialize bld ctx via lp_build_context_init instead of ad-hoc and broken...
José Fonseca [Sun, 17 Oct 2010 17:15:15 +0000 (10:15 -0700)]
llvmpipe: Initialize bld ctx via lp_build_context_init instead of ad-hoc and broken code.

14 years agollvmpipe: Fix bad refactoring.
José Fonseca [Sun, 17 Oct 2010 16:58:04 +0000 (09:58 -0700)]
llvmpipe: Fix bad refactoring.

'i' and 'chan' have random values here, which could cause a buffer
overflow in debug builds, if chan > 4.

14 years agogallivm: Fix SoA cubemap derivative computation.
José Fonseca [Sun, 17 Oct 2010 16:32:41 +0000 (09:32 -0700)]
gallivm: Fix SoA cubemap derivative computation.

Derivatives are now scalar.

Broken since 17dbd41cf23e7e7de2f27e5e9252d7f792d932f3.

14 years agollvmpipe: Ensure z_shift and z_width is initialized.
José Fonseca [Sun, 17 Oct 2010 14:45:08 +0000 (07:45 -0700)]
llvmpipe: Ensure z_shift and z_width is initialized.

14 years agollvmpipe: Fix depth-stencil regression.
José Fonseca [Sun, 17 Oct 2010 14:15:58 +0000 (07:15 -0700)]
llvmpipe: Fix depth-stencil regression.

If stencil is enabled then we need to load the z_dst, even if depth
testing is disabled.

This fixes reflect mesa demo.

14 years agor600g: add evergreen ARL support.
Dave Airlie [Sun, 17 Oct 2010 06:45:15 +0000 (16:45 +1000)]
r600g: add evergreen ARL support.

Thanks to Alex Deucher for pointing out the FLT to int conversion is necessary
and writing an initial patch, this brings about 20 piglits, and I think this
is the last piece to make evergreen and r600 equal in terms of features.

14 years agogallivm: use util_snprintf()
Brian Paul [Fri, 15 Oct 2010 23:32:23 +0000 (17:32 -0600)]
gallivm: use util_snprintf()

14 years agost/mesa: update function name, comments
Brian Paul [Fri, 15 Oct 2010 23:24:43 +0000 (17:24 -0600)]
st/mesa: update function name, comments

14 years agost/mesa: use GLuint to avoid problem w/ uint not defined on mingw32
Brian Paul [Fri, 15 Oct 2010 23:18:39 +0000 (17:18 -0600)]
st/mesa: use GLuint to avoid problem w/ uint not defined on mingw32

14 years agost/mesa: reformatting in st_cb_drawpixels.c
Brian Paul [Fri, 15 Oct 2010 23:01:56 +0000 (17:01 -0600)]
st/mesa: reformatting in st_cb_drawpixels.c

14 years agost/mesa: fix regressions in glDrawPixels(GL_STENCIL_INDEX)
Brian Paul [Fri, 15 Oct 2010 22:54:03 +0000 (16:54 -0600)]
st/mesa: fix regressions in glDrawPixels(GL_STENCIL_INDEX)

We need to keep track of three different fragment shaders: Z-only, stencil-
only, and Z+stencil.  Before, we were only keeping track of the first one
we encountered.

14 years agoglsl: add ir_unop_round_even case to silence unhandled enum warning
Brian Paul [Fri, 15 Oct 2010 21:44:01 +0000 (15:44 -0600)]
glsl: add ir_unop_round_even case to silence unhandled enum warning

14 years agogallivm: added lp_build_load_volatile()
Brian Paul [Fri, 15 Oct 2010 17:55:21 +0000 (11:55 -0600)]
gallivm: added lp_build_load_volatile()

There's no LLVM C LLVMBuildLoadVolatile() function so roll our own.
Not used anywhere at this time but can come in handy during debugging.

14 years agogallivm: added lp_build_print_vec4()
Brian Paul [Fri, 15 Oct 2010 14:41:31 +0000 (08:41 -0600)]
gallivm: added lp_build_print_vec4()

14 years agoi965: Set the type of the null register to fix gen6 FS comparisons.
Eric Anholt [Fri, 15 Oct 2010 19:04:52 +0000 (12:04 -0700)]
i965: Set the type of the null register to fix gen6 FS comparisons.

We often use reg_null as the destination when setting up the flag
regs.  However, on gen6 there aren't general implicit conversions to
destination types from src types, so the comparison to produce the
flag regs would be done on the integer result interpreted as a float.
Hilarity ensued.

Fixes 20 piglit cases.

14 years agoi965: Fix indentation after commit 3322fbaf
Ian Romanick [Fri, 15 Oct 2010 17:32:02 +0000 (10:32 -0700)]
i965: Fix indentation after commit 3322fbaf

14 years agolinker: Trivial indention fix
Ian Romanick [Fri, 15 Oct 2010 00:55:17 +0000 (17:55 -0700)]
linker: Trivial indention fix

14 years agotarget-helpers: Remove per target software wrapper check
Jakob Bornecrantz [Fri, 15 Oct 2010 14:46:10 +0000 (15:46 +0100)]
target-helpers: Remove per target software wrapper check

Instead of having a NAME_SOFTWARE check just use the GALLIUM_DRIVER
instead but set the default to native which is the same as not wrapped.

14 years agoegl: Remove unnecessary headers
Jakob Bornecrantz [Fri, 15 Oct 2010 14:58:54 +0000 (15:58 +0100)]
egl: Remove unnecessary headers

14 years agowrapper: Add a way to dewrap a pipe screen without destroying it
Jakob Bornecrantz [Fri, 15 Oct 2010 14:57:55 +0000 (15:57 +0100)]
wrapper: Add a way to dewrap a pipe screen without destroying it

14 years agowrapper: Fix spelling
Jakob Bornecrantz [Fri, 15 Oct 2010 14:38:29 +0000 (15:38 +0100)]
wrapper: Fix spelling

14 years agollvmpipe: Move makefile include to before targets
Jakob Bornecrantz [Tue, 12 Oct 2010 17:41:24 +0000 (18:41 +0100)]
llvmpipe: Move makefile include to before targets

Or plain make inside of the directory wont build libllvmpipe.a

14 years agonv50: apply layout_mask to tile_flags
Xavier Chantry [Fri, 15 Oct 2010 13:53:13 +0000 (15:53 +0200)]
nv50: apply layout_mask to tile_flags

The tile_flags now store more than just nv50 page table entry bits.

14 years agollvmpipe: validate color outputs against key->nr_cbufs
Keith Whitwell [Fri, 15 Oct 2010 12:23:30 +0000 (13:23 +0100)]
llvmpipe: validate color outputs against key->nr_cbufs

14 years agollvmpipe: check shader outputs are non-null before using
Keith Whitwell [Fri, 15 Oct 2010 12:23:05 +0000 (13:23 +0100)]
llvmpipe: check shader outputs are non-null before using

14 years agomesa: Add missing header to shaderobj.h.
Vinson Lee [Fri, 15 Oct 2010 13:12:26 +0000 (06:12 -0700)]
mesa: Add missing header to shaderobj.h.

Include compiler.h for ASSERT symbol.

14 years agollvmpipe: fix non-sse build after recent changes
Keith Whitwell [Fri, 15 Oct 2010 13:11:22 +0000 (14:11 +0100)]
llvmpipe: fix non-sse build after recent changes

14 years agollvmpipe: use aligned loads/stores for plane values
Keith Whitwell [Fri, 15 Oct 2010 12:52:00 +0000 (13:52 +0100)]
llvmpipe: use aligned loads/stores for plane values

14 years agollvmpipe: do plane calculations with intrinsics
Keith Whitwell [Tue, 12 Oct 2010 17:59:15 +0000 (18:59 +0100)]
llvmpipe: do plane calculations with intrinsics

This is a step towards moving this code into the rasterizer.

14 years agogallium: move some intrinsics helpers to u_sse.h
Keith Whitwell [Tue, 12 Oct 2010 17:58:05 +0000 (18:58 +0100)]
gallium: move some intrinsics helpers to u_sse.h

14 years agollvmpipe: don't store plane.ei value in binned data
Keith Whitwell [Fri, 15 Oct 2010 12:04:19 +0000 (13:04 +0100)]
llvmpipe: don't store plane.ei value in binned data

Further reduce the size of a binned triangle.

14 years agollvmpipe: slightly shrink the size of a binned triangle
Keith Whitwell [Fri, 15 Oct 2010 11:23:22 +0000 (12:23 +0100)]
llvmpipe: slightly shrink the size of a binned triangle

14 years agollvmpipe: don't pass frontfacing as a float
Keith Whitwell [Thu, 14 Oct 2010 23:12:19 +0000 (00:12 +0100)]
llvmpipe: don't pass frontfacing as a float

14 years agollvmpipe: reintroduce SET_STATE binner command
Keith Whitwell [Thu, 14 Oct 2010 22:28:10 +0000 (23:28 +0100)]
llvmpipe: reintroduce SET_STATE binner command

But bin lazily only into bins which are receiving geometry.

14 years agoglsl: Fix ir validation for bit logic ops
Chad Versace [Wed, 13 Oct 2010 15:29:58 +0000 (08:29 -0700)]
glsl: Fix ir validation for bit logic ops

In ir_validate::visit_leave(), the cases for
    - ir_binop_bit_and
    - ir_binop_bit_xor
    - ir_binop_bit_or
were incorrect. It was incorrectly asserted that both operands must be the
same type, when in fact one may be scalar and the other a vector. It was also
incorrectly asserted that the resultant type was the type of the left operand,
which in fact does not hold when the left operand is a scalar and the right
operand is a vector.

14 years agoglsl: Implement constant expr evaluation for bitwise logic ops
Chad Versace [Sun, 10 Oct 2010 04:27:41 +0000 (21:27 -0700)]
glsl: Implement constant expr evaluation for bitwise logic ops

Implement by adding the following cases to
ir_exporession::constant_expression_value():
    - ir_binop_bit_and
    - ir_binop_bit_or
    - ir_binop_bit_xor

14 years agoglsl: Implement constant expr evaluation for bit-shift ops
Chad Versace [Sun, 10 Oct 2010 03:56:14 +0000 (20:56 -0700)]
glsl: Implement constant expr evaluation for bit-shift ops

Implement by adding the following cases to
ir_expression::constant_expression_value():
    - ir_binop_lshfit
    - ir_binop_rshfit

14 years agoglsl: Implement constant expr evaluation for bitwise-not
Chad Versace [Sat, 9 Oct 2010 00:28:34 +0000 (17:28 -0700)]
glsl: Implement constant expr evaluation for bitwise-not

Implement by adding a case to ir_expression::constant_expression_value()
for ir_unop_bit_not.

14 years agoglsl: Implement ast-to-hir for binary shifts in GLSL 1.30
Chad Versace [Fri, 8 Oct 2010 23:22:28 +0000 (16:22 -0700)]
glsl: Implement ast-to-hir for binary shifts in GLSL 1.30

Implement by adding the following cases to ast_expression::hir():
    - ast_lshift
    - ast_rshift
Also, implement ir validation for the new operators by adding the following
cases to ir_validate::visit_leave():
    - ir_binop_lshift
    - ir_binop_rshift

14 years agoglsl: Change generated file glsl_lexer.cpp
Chad Versace [Fri, 8 Oct 2010 23:12:56 +0000 (16:12 -0700)]
glsl: Change generated file glsl_lexer.cpp

14 years agoglsl: Add lexer rules for << and >> in GLSL 1.30
Chad Versace [Fri, 8 Oct 2010 23:12:24 +0000 (16:12 -0700)]
glsl: Add lexer rules for << and >> in GLSL 1.30

Commit for generated file glsl_lexer.cpp follows this commit.

14 years agor600g: evergreen interpolation support.
Dave Airlie [Thu, 14 Oct 2010 04:40:51 +0000 (14:40 +1000)]
r600g: evergreen interpolation support.

On evergreen, interpolation has moved into the fragment shader,
with the interpolation parmaters being passed via GPRs and LDS entries.

This works out the number of interps required and reserves GPR/LDS
storage for them, it also correctly routes face/position values which
aren't interpolated from the vertex shader.

Also if we noticed nothing is to be interpolated we always setup perspective
interpolation for one value otherwise the GPU appears to lockup.

This fixes about 15 piglit tests on evergreen.

14 years agotgsi: add scanner support for centroid inputs
Dave Airlie [Thu, 14 Oct 2010 04:40:24 +0000 (14:40 +1000)]
tgsi: add scanner support for centroid inputs

14 years agoglsl: Slightly change the semantic of _LinkedShaders
Ian Romanick [Thu, 14 Oct 2010 20:28:42 +0000 (13:28 -0700)]
glsl: Slightly change the semantic of _LinkedShaders

Previously _LinkedShaders was a compact array of the linked shaders
for each shader stage.  Now it is arranged such that each slot,
indexed by the MESA_SHADER_* defines, refers to a specific shader
stage.  As a result, some slots will be NULL.  This makes things a
little more complex in the linker, but it simplifies things in other
places.

As a side effect _NumLinkedShaders is removed.

NOTE: This may be a candidate for the 7.9 branch.  If there are other
patches that get backported to 7.9 that use _LinkedShader, this patch
should be cherry picked also.

14 years agoi965: Fix texturing on pre-gen5.
Eric Anholt [Fri, 15 Oct 2010 00:01:06 +0000 (17:01 -0700)]
i965: Fix texturing on pre-gen5.

I broke it in 06fd639c519214b6ebcbf29127b6d9ed429f8641 by only testing
2 generations of hardware :(

14 years agollvmpipe: code to dump bytecode to file (disabled)
Brian Paul [Thu, 14 Oct 2010 22:31:54 +0000 (16:31 -0600)]
llvmpipe: code to dump bytecode to file (disabled)

14 years agogallivm: add compile-time option to emit inst addrs and/or line numbers
Brian Paul [Thu, 14 Oct 2010 22:21:32 +0000 (16:21 -0600)]
gallivm: add compile-time option to emit inst addrs and/or line numbers

Disabling address printing is helpful for diffing.

14 years agomesa: remove post-convolution width/height vars
Brian Paul [Thu, 14 Oct 2010 20:03:46 +0000 (14:03 -0600)]
mesa: remove post-convolution width/height vars

These were left-over bits from when convolution was removed.

14 years agoglsl: Refresh autogenerated file builtin_function.cpp.
Kenneth Graunke [Thu, 14 Oct 2010 20:41:29 +0000 (13:41 -0700)]
glsl: Refresh autogenerated file builtin_function.cpp.

14 years agoglsl: Add support for the 1.30 round() built-in.
Kenneth Graunke [Thu, 14 Oct 2010 21:33:30 +0000 (14:33 -0700)]
glsl: Add support for the 1.30 round() built-in.

This implements round() via the ir_unop_round_even opcode, rather than
adding a new opcode.  We may wish to add one in the future, since it
might enable a small performance increase on some hardware, but for now,
this should suffice.

14 years agoi965: Add support for ir_unop_round_even via the RNDE instruction.
Kenneth Graunke [Thu, 14 Oct 2010 21:31:54 +0000 (14:31 -0700)]
i965: Add support for ir_unop_round_even via the RNDE instruction.

14 years agoglsl: Add front-end support for GLSL 1.30's roundEven built-in.
Kenneth Graunke [Thu, 14 Oct 2010 20:40:26 +0000 (13:40 -0700)]
glsl: Add front-end support for GLSL 1.30's roundEven built-in.

Implemented using the op-code introduced in the previous commit.

14 years agoglsl: Add a new ir_unop_round_even opcode for GLSL 1.30's roundEven.
Kenneth Graunke [Thu, 14 Oct 2010 20:37:03 +0000 (13:37 -0700)]
glsl: Add a new ir_unop_round_even opcode for GLSL 1.30's roundEven.

Also, update ir_to_mesa's "1.30 is unsupported" case to "handle" it.

14 years agor300g: clean up warning due to unknown cap.
Dave Airlie [Thu, 14 Oct 2010 22:46:16 +0000 (08:46 +1000)]
r300g: clean up warning due to unknown cap.

14 years agor600g: handle absolute modifier in shader translator
Keith Whitwell [Thu, 14 Oct 2010 16:19:00 +0000 (17:19 +0100)]
r600g: handle absolute modifier in shader translator

This was being classed as unsupported in one place but used in others.
Enabling it seems to work fine.

Signed-off-by: Dave Airlie <airlied@redhat.com>
14 years agor600g: emit hardware linewidth
Keith Whitwell [Thu, 14 Oct 2010 15:42:39 +0000 (16:42 +0100)]
r600g: emit hardware linewidth

Tested with demos/pixeltest - line rasterization doesn't seem to be
set up for GL conventions yet, but at least width is respected now.

Signed-off-by: Dave Airlie <airlied@redhat.com>
14 years agor600/drm: fix segfaults in winsys create failure path
Keith Whitwell [Thu, 14 Oct 2010 15:42:38 +0000 (16:42 +0100)]
r600/drm: fix segfaults in winsys create failure path

Would try to destroy radeon->cman, radeon->kman both which were still
NULL.

Signed-off-by: Dave Airlie <airlied@redhat.com>
14 years agoi965: Clean up a warning in the old fragment backend.
Kenneth Graunke [Thu, 14 Oct 2010 20:19:23 +0000 (13:19 -0700)]
i965: Clean up a warning in the old fragment backend.

Hopefully this code can just go away soon.

14 years agoi965: Enable the new FS backend on pre-gen6 as well.
Eric Anholt [Thu, 14 Oct 2010 19:23:29 +0000 (12:23 -0700)]
i965: Enable the new FS backend on pre-gen6 as well.

It is now to the point where we have no regressing piglit tests.  It
also fixes Yo Frankie! and Humus DynamicBranching, probably due to the
piglit bias tests that work that didn't on the Mesa IR backend.

As a downside, performance takes about a 5-10% performance hit at the
moment (e.g. nexuiz 19.8fps -> 18.8fps), which I plan to resolve by
reintroducing 16-wide fragment shaders where possible.  It is a win,
though, for fragment shaders using flow control.

14 years agoi965: Correctly emit the RNDZ instruction.
Kenneth Graunke [Thu, 14 Oct 2010 18:40:19 +0000 (11:40 -0700)]
i965: Correctly emit the RNDZ instruction.

Simply using RNDU, RNDZ, or RNDE does not produce the desired result.
Rather, the RND* instructions place a value in the destination register
that may be 1 less than the correct answer.  They can also set per-channel
"increment bits" in a flag register, which, if set, mean dest needs to
be incremented by 1.  A second instruction - a predicated add -
completes the job.

Notably, RNDD always produces the correct answer in a single
instruction.

Fixes piglit test glsl-fs-trunc.

14 years agoi965: Use RNDZ for ir_unop_trunc in the new FS.
Kenneth Graunke [Thu, 14 Oct 2010 18:39:09 +0000 (11:39 -0700)]
i965: Use RNDZ for ir_unop_trunc in the new FS.

The existing code used RNDD, which rounds down, rather than toward zero.

14 years agoglsl: Refresh autogenerated file builtin_function.cpp.
Kenneth Graunke [Thu, 14 Oct 2010 17:51:09 +0000 (10:51 -0700)]
glsl: Refresh autogenerated file builtin_function.cpp.

14 years agoglsl: Add front-end support for the "trunc" built-in.
Kenneth Graunke [Thu, 14 Oct 2010 17:49:13 +0000 (10:49 -0700)]
glsl: Add front-end support for the "trunc" built-in.

14 years agoi965: Use logical-not when emitting ir_unop_ceil.
Kenneth Graunke [Wed, 13 Oct 2010 23:35:34 +0000 (16:35 -0700)]
i965: Use logical-not when emitting ir_unop_ceil.

Fixes piglit test glsl-fs-ceil.

14 years agoi965: Add peepholing of conditional mod generation from expressions.
Eric Anholt [Thu, 14 Oct 2010 18:27:17 +0000 (11:27 -0700)]
i965: Add peepholing of conditional mod generation from expressions.

This cuts usually 2 out of 3 instructions for flag reg generation (if
statements, conditional assignment) by producing the conditional mod
in the expression representing the boolean value.

Fixes glsl-fs-vec4-indexing-temp-dst-in-nested-loop-combined (register
allocation no longer fails for the conditional generation
proliferation)

14 years agoi965: Add a function for handling the move of boolean values to flag regs.
Eric Anholt [Thu, 14 Oct 2010 18:11:29 +0000 (11:11 -0700)]
i965: Add a function for handling the move of boolean values to flag regs.

This will be a place to peephole comparisions directly to the flag
regs, and for now avoids using MOV with conditional mod on gen6, which
is now illegal.

14 years agoOnly install vtxfmt tables for OpenGL
Kristian Høgsberg [Thu, 14 Oct 2010 18:57:47 +0000 (14:57 -0400)]
Only install vtxfmt tables for OpenGL

GLES1 and GLES2 install their own exec pointers and don't need the
Save table.  Also, the SET_* macros use different indices for the different
APIs so the offsets used in vtxfmt.c are actually wrong for the ES APIs.

14 years agoi965: Add a pass to the FS to split virtual GRFs to float channels.
Eric Anholt [Thu, 14 Oct 2010 03:17:15 +0000 (20:17 -0700)]
i965: Add a pass to the FS to split virtual GRFs to float channels.

Improves nexuiz performance 0.91% (+/- 0.54%, n=8)

14 years agoi965: Update the live interval when coalescing regs.
Eric Anholt [Thu, 14 Oct 2010 05:39:06 +0000 (22:39 -0700)]
i965: Update the live interval when coalescing regs.

14 years agoi965: Set class_sizes[] for the aligned reg pair class.
Eric Anholt [Thu, 14 Oct 2010 02:21:32 +0000 (19:21 -0700)]
i965: Set class_sizes[] for the aligned reg pair class.

So far, I've only seen this be a valgrind warning and not a real failure.

14 years agollvmpipe: don't try to emit non-existent color outputs
Keith Whitwell [Thu, 14 Oct 2010 12:15:28 +0000 (13:15 +0100)]
llvmpipe: don't try to emit non-existent color outputs

14 years agoDrop the "neutral" tnl module
Kristian Høgsberg [Wed, 13 Oct 2010 15:11:47 +0000 (11:11 -0400)]
Drop the "neutral" tnl module

Just always check for FLUSH_UPDATE_CURRENT and call Driver.BeginVertices
when necessary.  By using the unlikely() macros, this ends up as
a 10% performance improvement (for isosurf, anyway) over the old,
complicated function pointer swapping.

14 years agost/egl: Do not finish a fence that is NULL.
Chia-I Wu [Thu, 14 Oct 2010 09:13:36 +0000 (17:13 +0800)]
st/egl: Do not finish a fence that is NULL.

i915g would dereference the NULL pointer.

14 years agost/egl: Access _EGLConfig directly.
Chia-I Wu [Thu, 14 Oct 2010 09:06:21 +0000 (17:06 +0800)]
st/egl: Access _EGLConfig directly.

Drop the use of SET_CONFIG_ATTRIB.  Fix the value of EGL_SAMPLE_BUFFERS
along the way.

14 years agoegl: Access config attributes directly.
Chia-I Wu [Thu, 14 Oct 2010 08:38:58 +0000 (16:38 +0800)]
egl: Access config attributes directly.

Replace SET_CONFIG_ATTRIB/GET_CONFIG_ATTRIB by direct dereferences.

14 years agoegl: Use attribute names as the _EGLConfig member names.
Chia-I Wu [Thu, 14 Oct 2010 08:25:30 +0000 (16:25 +0800)]
egl: Use attribute names as the _EGLConfig member names.

This makes _EGLConfig more accessible and scales better when new
attributes are added.

14 years agor600g: select linear interpolate if tgsi input requests it
Dave Airlie [Thu, 14 Oct 2010 04:27:34 +0000 (14:27 +1000)]
r600g: select linear interpolate if tgsi input requests it

14 years agor600g: fixup typo in macro name
Dave Airlie [Thu, 14 Oct 2010 04:19:37 +0000 (14:19 +1000)]
r600g: fixup typo in macro name

14 years agor600g: fixup pos/face ena/address properly
Dave Airlie [Thu, 14 Oct 2010 04:14:22 +0000 (14:14 +1000)]
r600g: fixup pos/face ena/address properly

14 years agor600g: only pick centroid coordinate when asked.
Dave Airlie [Thu, 14 Oct 2010 03:29:11 +0000 (13:29 +1000)]
r600g: only pick centroid coordinate when asked.

TGSI tells us when to use this, its not hooked up from GLSL to MESA to TGSI yet though.