| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
We've assumed that we could lower per-component vector access from
vec[i] = scalar
to
vec = ir_triop_vector_insert(vec, scalar, i)
but with SSBOs (and compute shader SLM and tesselation outputs) this is
no longer valid. If a vector is "externally visible", multiple threads
can write independent components simultaneously. With lowering to
ir_triop_vector_insert, each thread read the entire vector, changes one
component, then writes out the entire vector. This is racy.
Instead of generating a ir_binop_vector_extract when we see v[i], we
generate ir_dereference_array. We then add a lowering pass to lower the
ir_dereference_array to ir_binop_vector_extract for rvalues and for to
vector_insert for lvalues in a separate lowering pass.
The resulting IR is the same as before, but we now have a window between
ast->ir conversion and the lowering pass where v[i] appears in the IR as
an array deref. This lets us run lowering passes that lower the vector
access to I/O (eg for SSBO load/store) before we lower the per-component
access to full vector writes.
Reviewed-by: Jordan Justen <[email protected]>
Signed-off-by: Kristian Høgsberg Kristensen <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
We always pass in shader->ir and we already pass in the shader, so just
drop the exec_list. Most passes either take just a exec_list or a
shader, so this seems more consistent.
Reviewed-by: Timothy Arceri <[email protected]>
Reviewed-by: Iago Toral Quiroga <[email protected]>
Reviewed-by: Jordan Justen <[email protected]>
Signed-off-by: Kristian Høgsberg Kristensen <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
With NIR, it actually hurts things.
total instructions in shared programs: 6529329 -> 6528888 (-0.01%)
instructions in affected programs: 14833 -> 14392 (-2.97%)
helped: 299
HURT: 1
In all affected programs I inspected (including the single hurt one) the
pass CSE'd some multiplies and caused some reassociation (e.g., caused
(A * B) * C to be A * (B * C)) when the original intermediate result was
reused elsewhere.
Acked-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
| |
This greatly improves generated code, especially for the snorm variants,
since it is able to get rid of the lshift/rshift for sext, as well as
replacing each shift + mask with a single op.
Signed-off-by: Ilia Mirkin <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
|
|
|
|
|
|
|
|
|
| |
It is fairly tricky to detect the proper conditions for using bitfield
insert, but easy to just use it up front. This removes a lot of
instructions on nvc0 when invoking the packing builtins.
Signed-off-by: Ilia Mirkin <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This lowers the enhanced ir_call using the lookaside table
of subroutines into an if ladder. This initially was done
at the AST level but it caused some ordering issues so a separate
pass was required.
v2: clone return value derefs.
v2.1: update for subroutine->int convert.
v2.2: add a clone for the array index
Reviewed-by: Chris Forbes <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Tessellation control outputs can be read in directly without first
having been written. Accessing these will require some special logic
anyways, so just let them through.
V2: Never lower tess control output reads, whether patch or not -- both
can be read back by other threads.
Signed-off-by: Ilia Mirkin <[email protected]>
Signed-off-by: Chris Forbes <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
| |
There is no way to lower them, because the array sizes are unknown
at compile time.
Based on a patch from: Fabian Bieler <[email protected]>
v2: add comments
Reviewed-by: Kenneth Graunke <[email protected]>
Reviewed-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
Similar to gl_ClipDistance -> gl_ClipDistanceMESA
v2: - renamed is_mesa_var to lowered_builtin_array_variable
- moved LowerTessLevel into gl_constants
- cosmetic changes in lower_tess_level.cpp
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
st_glsl_to_tgsi and ir_to_mesa have handled conditional discards for a
long time; the previous patch added that capability to i965.
i965 (Haswell) shader-db stats:
Without NIR:
total instructions in shared programs: 5792133 -> 5776360 (-0.27%)
instructions in affected programs: 737585 -> 721812 (-2.14%)
helped: 6300
HURT: 68
GAINED: 2
With NIR:
total instructions in shared programs: 5787538 -> 5769569 (-0.31%)
instructions in affected programs: 767843 -> 749874 (-2.34%)
helped: 6522
HURT: 35
GAINED: 6
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Connor Abbott <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
| |
Signed-off-by: Ilia Mirkin <[email protected]>
Reviewed-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
| |
These lowering passes are optional for the backend to request, currently
the TGSI softpipe backend most likely the r600g backend would want to use
these passes as is. They aim to hit the gallium opcodes from the standard
rounding/truncation functions.
v2: also lower floor in mod_to_floor
Signed-off-by: Dave Airlie <[email protected]>
Reviewed-by: Ilia Mirkin <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Currently, Mesa uses the lowering pass MOD_TO_FRACT to implement
mod(x,y) as y * fract(x/y). This implementation has a down side though:
it introduces precision errors due to the fract() operation. Even worse,
since the result of fract() is multiplied by y, the larger y gets the
larger the precision error we produce, so for large enough numbers the
precision loss is significant. Some examples on i965:
Operation Precision error
-----------------------------------------------------
mod(-1.951171875, 1.9980468750) 0.0000000447
mod(121.57, 13.29) 0.0000023842
mod(3769.12, 321.99) 0.0000762939
mod(3769.12, 1321.99) 0.0001220703
mod(-987654.125, 123456.984375) 0.0160663128
mod( 987654.125, 123456.984375) 0.0312500000
This patch replaces the current lowering pass with a different one
(MOD_TO_FLOOR) that follows the recommended implementation in the GLSL
man pages:
mod(x,y) = x - y * floor(x/y)
This implementation eliminates the precision errors at the expense of
an additional add instruction on some systems. On systems that can do
negate with multiply-add in a single operation this new implementation
would come at no additional cost.
v2 (Ian Romanick)
- Do not clone operands because when they are expressions we would be
duplicating them and that can lead to suboptimal code.
Fixes the following 16 dEQP tests:
dEQP-GLES3.functional.shaders.builtin_functions.precision.mod.mediump_*
dEQP-GLES3.functional.shaders.builtin_functions.precision.mod.highp_*
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Consider GLSL code such as:
const ivec2 offsets[] =
ivec2[](ivec2(-1, -1), ivec2(-1, 0), ivec2(-1, 1),
ivec2(0, -1), ivec2(0, 0), ivec2(0, 1),
ivec2(1, -1), ivec2(1, 0), ivec2(1, 1));
ivec2 offset = offsets[<non-constant expression>];
Both i965 and nv50 currently handle this very poorly. On i965, this
becomes a pile of MOVs to load the immediate constants into registers,
a pile of scratch writes to move the whole array to memory, and one
scratch read to actually access the value - effectively the same as if
it were a non-constant array.
We'd much rather upload large blocks of constant data as uniform data,
so drivers can simply upload the data via constbufs, and not have to
populate it via shader instructions.
This is currently non-optional because both i965 and nouveau benefit
from it, and according to Marek radeonsi would benefit today as well.
(According to Tom, radeonsi may want to handle this itself in the long
term, but we can always add a flag when it becomes useful.)
Improves performance in a terrain rendering microbenchmark by about 2x,
and cuts the number of instructions in about half. Helps a lot of
"Natural Selection 2" shaders, as well as one "HOARD" shader.
total instructions in shared programs: 5473459 -> 5471765 (-0.03%)
instructions in affected programs: 5880 -> 4186 (-28.81%)
v2: Use ir_var_hidden to avoid exposing the new uniform via the GL
uniform introspection API.
v3: Alphabetize Makefile.sources properly.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=77957
Signed-off-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Original patch by Petri Latvala <[email protected]>:
Add an optimization pass that drops min/max expression operands that
can be proven to not contribute to the final result. The algorithm is
similar to alpha-beta pruning on a minmax search, from the field of
AI.
This optimization pass can optimize min/max expressions where operands
are min/max expressions. Such code can appear in shaders by itself, or
as the result of clamp() or AMD_shader_trinary_minmax functions.
This optimization pass improves the generated code for piglit's
AMD_shader_trinary_minmax tests as follows:
total instructions in shared programs: 75 -> 67 (-10.67%)
instructions in affected programs: 60 -> 52 (-13.33%)
GAINED: 0
LOST: 0
All tests (max3, min3, mid3) improved.
A full shader-db run:
total instructions in shared programs: 4293603 -> 4293575 (-0.00%)
instructions in affected programs: 1188 -> 1160 (-2.36%)
GAINED: 0
LOST: 0
Improvements happen in Guacamelee and Serious Sam 3. One shader from
Dungeon Defenders is hurt by shader-db metrics (26 -> 28), because of
dropping of a (constant float (0.00000)) operand, which was
compiled to a saturate modifier.
Version 2 by Iago Toral Quiroga <[email protected]>:
Changes from review feedback:
- Squashed various cosmetic changes sent by Matt Turner.
- Make less_all_components return an enum rather than setting a class member.
(Suggested by Mat Turner). Also, renamed it to compare_components.
- Make less_all_components, smaller_constant and larger_constant static.
(Suggested by Mat Turner)
- Change mixmax_range to call its limits "low" and "high" instead of
"range[0]" and "range[1]". (Suggested by Connor Abbot).
- Use ir_builder swizzle helpers in swizzle_if_required(). (Suggested by
Connor Abbot).
- Make the logic more clearer by rearrenging the code and commenting.
(Suggested by Connor Abbot).
- Added comment to explain why we need to recurse twice. (Suggested by
Connor Abbot).
- If we cannot prune an expression, do not return early. Instead, attempt
to prune its children. (Suggested by Connor Abbot).
Other changes:
- Instead of having a global "valid" visitor member, let the various functions
that can determine this status return a boolean and check for its value
to decide what to do in each case. This is more flexible and allows to
recurse into children of parents that could not be prunned due to invalid
ranges (so related to the last bullet in the review feedback).
- Make sure we always check if a range is valid before working with it. Since
any use of get_range, combine_range or range_intersection can invalidate
a range we should check for this situation every time we use any of these
functions.
Version 3 by Iago Toral Quiroga <[email protected]>:
Changes from review feedback:
- Now we can make get_range, combine_range and range_intersection static too
(suggested by Connor Abbot).
- Do not return NULL when looking for the larger or greater constant into
mixed vector constants. Instead, produce a new constant by doing a
component-wise minmax. With this we can also remove of the validations when
we call into these functions (suggested by Connor Abbot).
- Add a comment explaining the meaning of the baserange argument in
prune_expression (suggested by Connor Abbot).
Other changes:
- Eliminate minmax expressions operating on constant vectors with mixed values
by resolving them.
No piglit regressions observed with Version 3.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=76861
Reviewed-by: Connor Abbott <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
After compilation (and before linking) we can eliminate quite a few
built-in variables. Basically, any uniform or constant (e.g.,
gl_MaxVertexTextureImageUnits) that isn't used (with one exception) can
be eliminated. System values, vertex shader inputs (with one
exception), and fragment shader outputs that are not used and not
re-declared in the shader text can also be removed.
gl_ModelViewProjectMatrix and gl_Vertex are used by the built-in
function ftransform. There are some complications with eliminating
these variables (see the comment in the patch), so they are not
eliminated.
Valgrind massif results for a trimmed apitrace of dota2:
n time(i) total(B) useful-heap(B) extra-heap(B) stacks(B)
Before (32-bit): 46 40,661,487,174 75,116,800 68,854,065 6,262,735 0
After (32-bit): 50 40,564,927,443 69,185,408 63,683,871 5,501,537 0
Before (64-bit): 64 37,200,329,700 104,872,672 96,514,546 8,358,126 0
After (64-bit): 59 36,822,048,449 96,526,888 89,113,000 7,413,888 0
A real savings of 4.9MiB on 32-bit and 7.0MiB on 64-bit.
v2: Don't remove any built-in with Transpose in the name.
v3: Fix comment typo noticed by Anuj.
Signed-off-by: Ian Romanick <[email protected]>
Suggested-by: Eric Anholt <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
Acked-by: Anuj Phogat <[email protected]>
Cc: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Converts gl_VertexID to (gl_VertexIDMESA + gl_BaseVertex). gl_VertexIDMESA
is backed by SYSTEM_VALUE_VERTEX_ID_ZERO_BASE, and gl_BaseVertex is backed
by SYSTEM_VALUE_BASE_VERTEX.
v2: Put the enum in struct gl_constants and propoerly resolve the scope
in C++ code. Fix suggested by Marek.
v3: Reabase on Matt's foreach_in_list changes (was using foreach_list).
v4 (Ken): Use a systemvalue instead of a uniform because
STATE_BASE_VERTEX has been removed.
v5: Use a boolean to select lowering, and only allow one lowering
method. Suggested by Ken.
v6 (Ken): Replace strcmp against literal "gl_BaseVertex"/"gl_VertexID"
with SYSTEM_VALUE enum checks, for efficiency.
v7: Rebase on context constant initialization work.
Signed-off-by: Ian Romanick <[email protected]>
Signed-off-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
| |
Signed-off-by: Abdiel Janulgue <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
| |
Will be used in the next commit.
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
The intention of this pass was to give us better instruction scheduling
opportunities, but it unexpectedly reduced some instruction counts as
well:
total instructions in shared programs: 1666639 -> 1666073 (-0.03%)
instructions in affected programs: 54612 -> 54046 (-1.04%)
(and trades 4 SIMD16 programs in SS3)
|
|
|
|
|
| |
Signed-off-by: Ilia Mirkin <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
In February 2013 Paul unified the values used for shader stage outputs
and shader stage inputs. See commits 8a076c5f0^..eed6baf76. Since that
time, the location_base parameters are always VARYING_SLOT_VAR0.
Instead of passing that around, just hard code it.
Signed-off-by: Ian Romanick <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
| |
Now that we pass in gl_shader_compiler_options, it makes sense to just
use options->MaxUnrollIterations, rather than passing a separate
parameter.
Half of the invocations already passed options->MaxUnrollIterations,
while the other half passed in a hardcoded value of 32.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
|
|
| |
The next patch will introduce an optimization that only works when
integers are not represented as floating point values.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
The next few patches will introduce an optimization that only works when
integers are not represented as floating point values.
v2: Re-word-wrap a line, as requested by Ian Romanick.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Tt's kind of a trap---calling do_common_optimization() after
lower_instructions() may cause opt_algebraic() to reintroduce
ir_triop_lrp expressions that were lowered, effectively defeating the
point. Because of this, nobody uses it.
v2: Delete more code (caught by Ian Romanick).
Cc: "10.1" <[email protected]>
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
Acked-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
|
| |
This lowering pass will be useful for gallium drivers as well, in order to support
the GL TG4 oddity that is textureGatherOffsets.
Reviewed-by: Chris Forbes <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
Reduces vertex shader instruction counts in DOTA2 by 6.42%, L4D2 by
4.61%, and CS:GO by 5.71%.
total instructions in shared programs: 1500153 -> 1498191 (-0.13%)
instructions in affected programs: 59919 -> 57957 (-3.27%)
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
|
|
| |
Now that loop_controls no longer creates normatively bound loops,
there is no need for ir_loop::normative_bound or the
lower_bounded_loops pass.
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Previously, all of the back-ends (ir_to_mesa, st_glsl_to_tgsi, and the
i965 fs and vec4 visitors) had nearly identical logic for handling
bounded loops. This replaces the duplicate logic with an equivalent
lowering pass that is used by all the back-ends.
Note: on i965, there is a slight increase in instruction count. For
example, a loop like this:
for (int i = 0; i < 100; i++) {
total += i;
}
would previously compile down to this (vec4) native code:
mov(8) g4<1>.xD 0D
mov(8) g8<1>.xD 0D
loop:
cmp.ge.f0(8) null g8<4;4,1>.xD 100D
(+f0) break(8)
add(8) g5<1>.xD g5<4;4,1>.xD g4<4;4,1>.xD
add(8) g8<1>.xD g8<4;4,1>.xD 1D
add(8) g4<1>.xD g4<4;4,1>.xD 1D
while(8) loop
After this patch, the "(+f0) break(8)" turns into:
(+f0) if(8)
break(8)
endif(8)
because the back-end isn't smart enough to recognize that "if
(condition) break;" can be done using a conditional break instruction.
However, it should be relatively easy for a future peephole
optimization to properly optimize this.
Reviewed-by: Jordan Justen <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This only operates on constant/uniform values for now, because otherwise I'd
have to deal with killing my available CSE entries when assignments happen,
and getting even this working in the tree ir was painful enough.
As is, it has the following effect in shader-db:
total instructions in shared programs: 1524077 -> 1521964 (-0.14%)
instructions in affected programs: 50629 -> 48516 (-4.17%)
GAINED: 0
LOST: 0
And, for tropics, that accounts for most of the effect, the FPS
improvement is 11.67% +/- 0.72% (n=3).
v2: Use read_only field of the variable, manually check the lod_info union
members, use get_num_operands(), rename cse_operands_visitor to
is_cse_candidate_visitor, move all is-a-candidate logic to that
function, and call it before checking for CSE on a given rvalue, more
comments, use private keyword.
Reviewed-by: Paul Berry <[email protected]>
|
|
|
|
| |
Reviewed-by: Paul Berry <[email protected]>
|
|
|
|
|
|
|
|
|
| |
Tested by examining generated TGSI shaders from piglit/glsl-routing.
Cc: [email protected]
Reviewed-by: Henri Verbeet <[email protected]>
Tested-by: Henri Verbeet <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
|
|
| |
Since geometry shader inputs are arrays (where the array index
indicates which vertex is being examined), varying packing needs to
treat them differently.
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
|
|
|
| |
Fixes this build error on OpenBSD 5.3.
In file included from ../../src/mesa/main/ff_fragment_shader.cpp:53:
./../glsl/ir_optimization.h:64: error: comma at end of enumerator list
Signed-off-by: Vinson Lee <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This eliminates built-in varyings such as gl_Color, gl_SecondaryColor,
gl_TexCoord, and gl_FogFragCoord if they are unused by the next stage or
not written at all (e.g. gl_TexCoord elements). The gl_TexCoord array is
broken down into separate vec4s if needed.
v2: - use a switch statement in varying_info_visitor::visit(ir_variable*)
- use snprintf
- disable the optimization for GLES2
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
|
|
| |
Convert interface blocks with instance names into flat
interface blocks without an instance name.
Signed-off-by: Jordan Justen <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This will eventually replace do_vec_index_to_cond_assign. This lowering
pass is called in all the places where do_vec_index_to_cond_assign or
do_vec_index_to_swizzle is called.
v2: Use WRITEMASK_* instead of integer literals. Use a more concise
method of generating broadcast_index. Both suggested by Eric.
v3: Use a series of scalar compares instead of a single vector compare.
Suggested by Eric and Ken. It still uses 'if (cond) v.x = y;' instead
of conditional assignments because ir_builder doesn't do conditional
assignments, and I'd rather keep the code simple.
Signed-off-by: Ian Romanick <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This pass flips (matrix * vector) operations to (vector *
matrixTranspose) for certain built-in matrices (currently
gl_ModelViewProjectionMatrix and gl_TextureMatrix).
This is equivalent, but results in dot products rather than multiplies
and adds. On some hardware, this is more efficient.
This pass is conditionalized on ctx->mvp_with_dp4, the flag drivers set
to indicate they prefer dot products.
Improves performance in Lightsmark by 1.01131% +/- 0.162069% (n = 10)
on a Haswell GT2 system. Passes Piglit on Ivybridge.
v2: Use struct gl_shader_compiler_options instead of plumbing through
another boolean flag for this purpose.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
do_common_optimization may need to make choices about whether to emit
certain kinds of instructions. gl_context::ShaderCompilerOptions
contains exactly that information, so it makes sense to pass it in.
Rather than passing the whole array, pass the structure for the stage
that's currently being worked on.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
i965/Gen7+ and Radeon/Evergreen+ have bfm/bfi instructions to implement
bitfieldInsert() from ARB_gpu_shader5.
v2: Add ir_binop_bfm and ir_triop_bfi to st_glsl_to_tgsi.cpp.
Remove spurious temporary assignment and dereference.
Reviewed-by: Chris Forbes <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
GLBenchmark 2.7's shaders contain conditional blocks like:
if (x) {
if (y) {
...
}
}
where the outer conditional's then clause contains exactly one statement
(the nested if) and there are no else clauses. This can easily be
optimized into:
if (x && y) {
...
}
This saves a few instructions in GLBenchmark 2.7:
total instructions in shared programs: 11833 -> 11649 (-1.55%)
instructions in affected programs: 8234 -> 8050 (-2.23%)
It also helps CS:GO slightly (-0.05%/-0.22%). More importantly,
however, it simplifies the control flow graph, which could enable other
optimizations.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Many GPUs have an instruction to do linear interpolation which is more
efficient than simply performing the algebra necessary (two multiplies,
an add, and a subtract).
Pattern matching or peepholing this is more desirable, but can be
tricky. By using an opcode, we can at least make shaders which use the
mix() built-in get the more efficient behavior.
Currently, all consumers lower ir_triop_lrp. Subsequent patches will
actually generate different code.
v2 [mattst88]:
- Add LRP_TO_ARITH flag to ir_to_mesa.cpp. Will be removed in a
subsequent patch and ir_triop_lrp translated directly.
v3 [mattst88]:
- Move changes from the next patch to opt_algebraic.cpp to accept
3-src operations.
Reviewed-by: Matt Turner <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Signed-off-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
| |
Lower them to arithmetic and bit manipulation expressions.
Reviewed-by: Chad Versace <[email protected]>
Reviewed-by: Paul Berry <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Lower them to arithmetic and bit manipulation expressions.
v2: Rewrite using ir_builder [for idr].
v3: Comment typos. [for mattst88]
v4: Fix arithmetic error in comments.
Factor out a shift instruction.
Don't heap allocate factory.instructions.
[for paul]
Reviewed-by: Ian Romanick <[email protected]> (v2)
Reviewed-by: Matt Tuner <[email protected]> (v3)
Reviewed-by: Paul Berry <[email protected]> (v4)
Signed-off-by: Chad Versace <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This lowering pass generates GLSL code that manually packs varyings
into vec4 slots, for the benefit of back-ends that don't support
packed varyings natively.
No functional change--the lowering pass is not yet used.
Reviewed-by: Eric Anholt <[email protected]>
v2: Don't use ir_hierarchical_visitor--just loop over instructions
directly. Also, make the names of the packed varyings include the
names of the original varyings that were packed into them.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This patch modifies the clip distance lowering pass so that the new
symbol it generates (glClipDistanceMESA) is added to the shader's
symbol table.
This will allow a later patch to modify the linker so that it finds
transform feedback varyings using the symbol table rather than having
to iterate through all the declarations in the shader.
Reviewed-by: Eric Anholt <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
v2: Reduce the impenetrable code in emit_ubo_loads() by 23 lines by keeping
the ir_variable as the variable part of the offset from handle_rvalue(),
and track the constant offsets from that with a plain old integer value,
avoiding a bunch of temporary variables in the array and struct handling.
Also, fix file description doxygen.
v3: Fix a row vs col typo, and fix spelling in a comment.
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
| |
Should fix MSVC link failure.
|
|
|
|
|
|
|
| |
Previously, I tried implementing this in the i965 driver, but did so
in a way that violated the intent of the spec, and broke Tropics.
Reviewed-by: Kenneth Graunke <[email protected]>
|