| Age | Commit message (Collapse) | Author |
|---|
|
This patch cleans up many of the extra copies in GLSL IR introduced by
i965's scalarizing passes. It doesn't result in a statistically
significant performance difference on nexuiz high settings (n=3) or my
demo (n=10), due to brw_fs.cpp's register coalescing covering most of
those extra moves anyway. However, it does make the debug of wine's
GLSL shaders much more tractable, and reduces instruction count of
glsl-fs-convolution-2 from 376 to 288.
|
|
This adds a new optional max_depth parameter (defaulting to 0) to
lower_if_to_cond_assign, and makes the pass only flatten if-statements
nested deeper than that.
By default, all if-statements will be flattened, just like before.
This patch also renames do_if_to_cond_assign to lower_if_to_cond_assign,
to match the new naming conventions.
|
|
|
|
This should allow lower_if_to_cond_assign to work in the presence of
discards, fixing bug #31690 and likely #31983.
NOTE: This is a candidate for the 7.9 branch.
|
|
NOTE: This is a candidate for the 7.9 branch.
|
|
This should save on the overhead of tree-walking and provide a
convenient place to add more instruction lowering in the future.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
|
|
The vector operator collects 2, 3, or 4 scalar components into a
vector. Doing this has several advantages. First, it will make
ud-chain tracking for components of vectors much easier. Second, a
later optimization pass could collect scalars into vectors to allow
generation of SWZ instructions (or similar as operands to other
instructions on R200 and i915). It also enables an easy way to
generate IR for SWZ instructions in the ARB_vertex_program assembler.
|
|
|
|
|
|
Currenly GLSL happily generates indirect addressing of any kind of
arrays.
Unfortunately DirectX 9 GPUs are not guaranteed to support any of them in
general.
This pass fixes that by lowering such constructs to a binary search on the
values, followed at the end by vectorized generation of equality masks, and
4 conditional assignments for each mask generation.
Note that this requires the ir_binop_equal change so that we can emit SEQ
to generate the boolean masks.
Unfortunately, ir_structure_splitting is too dumb to turn the resulting
constant array references to individual variables, so this will need to
be added too before this pass can actually be effective for temps.
Several patches in the glsl2-lower-variable-indexing were squashed
into this commit. These patches fix bugs in Luca's original
implementation, and the individual patches can be seen in that branch.
This was done to aid bisecting in the future.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
|
|
|
|
Changes in v2:
- Base class renamed to ir_control_flow_visitor
- Tried to comply with coding style
This is a new pass that supersedes ir_if_return and "lowers" jumps
to if/else structures.
Currently it causes no regressions on softpipe and nv40, but I'm not sure
whether the piglit glsl tests are thorough enough, so consider this
experimental.
It can be asked to:
1. Pull jumps out of ifs where possible
2. Remove all "continue"s, replacing them with an "execute flag"
3. Replace all "break" with a single conditional one at the end of the loop
4. Replace all "return"s with a single return at the end of the function,
for the main function and/or other functions
This gives several great benefits:
1. All functions can be inlined after this pass
2. nv40 and other pre-DX10 chips without "continue" can be supported
3. nv30 and other pre-DX10 chips with no control flow at all are better supported
Note that for full effect we should also teach the unroller to unroll
loops with a fixed maximum number of iterations but with the canonical
conditional "break" that this pass will insert if asked to.
Continues are lowered by adding a per-loop "execute flag", initialized to
TRUE, that when cleared inhibits all execution until the end of the loop.
Breaks are lowered to continues, plus setting a "break flag" that is checked
at the end of the loop, and trigger the unique "break".
Returns are lowered to breaks/continues, plus adding a "return flag" that
causes loops to break again out of their enclosing loops until all the
loops are exited: then the "execute flag" logic will ignore everything
until the end of the function.
Note that "continue" and "return" can also be implemented by adding
a dummy loop and using break.
However, this is bad for hardware with limited nesting depth, and
prevents further optimization, and thus is not currently performed.
|
|
|
|
This increases the chance that GLSL programs will actually work.
Note that continues and returns are not yet lowered, so linking
will just fail if not supported.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
|
|
With the glsl2-965 branch, the optimization of glsl-algebraic-rcp-rcp
regressed due to noop swizzles hiding information from ir_algebraic.
This cleans up those noop swizzles for us.
|
|
These are passes that we expect all codegen to be happy with. The
other lowering passes for Mesa IR are moved to the Mesa IR generator.
|
|
All the current HW backends transform subtract to adding the negation,
so I haven't bothered peepholing it back out in Mesa IR. This allows
some subtract of subtract to get removed in ir_algebraic.
|
|
Whereas constant folding evaluates constant expressions at rvalue
nodes, constant propagation tracks constant components of vectors
across execution to replace (possibly swizzled) variable dereferences
with constant values, triggering possible constant folding or reduced
variable liveness.
|
|
Fixes ir_to_mesa handling of unop_log, which used the weird ARB_vp LOG
opcode that doesn't do what we want. This also lets the multiplication
coefficients in there get constant-folded, possibly.
Fixes:
glsl-fs-log
|
|
This doesn't do anything if your structure goes through an uninlined
function call or if whole-structure assignment occurs. As such, the
impact is limited, at least until we do some global copy propagation
to reduce whole-structure assignment.
|
|
For a shader involving many small functions, this avoids running
optimization across all of them after they've been inlined
post-linking.
Reduces the runtime of linking and running a fragment shader from Yo
Frankie from 1.6 seconds to 0.9 seconds (-44.9%, +/- 3.3%).
|
|
|
|
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.
|
|
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).
|
|
This will be used on 915 and similar hardware of that generation.
|
|
This will be used by the Mesa IR and likely most HW backends, as it
allows other optimizations to occur that might not otherwise.
Fixes glsl-vs-mat-sub-1, glsl-vs-mat-div-1.
|
|
This allows function inlining making the following tests work even
without function calls implemented:
glsl-fs-functions-2
glsl-fs-functions-3
glsl-vs-functions
glsl-vs-functions-2
glsl-vs-functions-3
glsl-vs-vec4-indexing-5
(Note that those tests were designed to trigger actual function calls,
and this defeats them. However, those testcases ended up catching the
bug in the previous commit.)
|
|
The Mesa IR needs this to support vector indexing correctly, and
hardware backends such as 915 would want this behavior as well.
Fixes glsl-vs-vec4-indexing-2.
|
|
This results in constant folding of a constant divisor.
|
|
This is used by the Mesa IR backend to implement mod, fixing glsl-fs-mod.
|
|
This cuts runtime by around 20% from talloc_parent() lookups.
|
|
|
