| Age | Commit message (Collapse) | Author |
|---|
|
There's no LLVM C LLVMBuildLoadVolatile() function so roll our own.
Not used anywhere at this time but can come in handy during debugging.
|
|
|
|
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.
|
|
|
|
|
|
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.
|
|
|
|
|
|
|
|
Or plain make inside of the directory wont build libllvmpipe.a
|
|
The tile_flags now store more than just nv50 page table entry bits.
|
|
|
|
|
|
Include compiler.h for ASSERT symbol.
|
|
|
|
|
|
This is a step towards moving this code into the rasterizer.
|
|
|
|
Further reduce the size of a binned triangle.
|
|
|
|
|
|
But bin lazily only into bins which are receiving geometry.
|
|
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.
|
|
Implement by adding the following cases to
ir_exporession::constant_expression_value():
- ir_binop_bit_and
- ir_binop_bit_or
- ir_binop_bit_xor
|
|
Implement by adding the following cases to
ir_expression::constant_expression_value():
- ir_binop_lshfit
- ir_binop_rshfit
|
|
Implement by adding a case to ir_expression::constant_expression_value()
for ir_unop_bit_not.
|
|
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
|
|
|
|
Commit for generated file glsl_lexer.cpp follows this commit.
|
|
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.
|
|
|
|
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.
|
|
I broke it in 06fd639c519214b6ebcbf29127b6d9ed429f8641 by only testing
2 generations of hardware :(
|
|
|
|
Disabling address printing is helpful for diffing.
|
|
These were left-over bits from when convolution was removed.
|
|
|
|
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.
|
|
|
|
Implemented using the op-code introduced in the previous commit.
|
|
Also, update ir_to_mesa's "1.30 is unsupported" case to "handle" it.
|
|
|
|
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>
|
|
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>
|
|
Would try to destroy radeon->cman, radeon->kman both which were still
NULL.
Signed-off-by: Dave Airlie <airlied@redhat.com>
|
|
Hopefully this code can just go away soon.
|
|
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.
|
|
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.
|
|
The existing code used RNDD, which rounds down, rather than toward zero.
|
|
|
