2 source instruction only support immediate for src1 operand, so no
point in adding optimization condition for src0 oprand.
v2:
- Update commit message and don't remove ADD optimization (Matt Turner)
Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5341>
Gen9 and Cherryview have the ability to mark texture instructions with
the End-of-thread bit under some conditions, which allows the texture
result to be written to the render target directly, rather than
returning to the EU.
In order to handle overlapping primitives correctly, we have to use the
'sendc' instruction which stalls until other threads potentially writing
to the same locations in the render target are retired. Unfortunately,
this stall happens before the texture is sampled (rather than in
parallel with stall), so for some literal edge cases (like the diagonal
edge between two triangles forming a rectangle) there can be a
performance penalty. As a result, it's probably not a good idea to use
this optimization in general.
I had planned to leave it enabled only for BLORP, where we use rectangle
primitives and are typically clearing/blitting an entire render target
without any overlapping primitives, but I noticed that the optimization
wasn't applied in some normal cases anyway. For example, in the piglit
test tests/shaders/glsl-fs-texture2d-bias.shader_test it is applied to
one BLORP-blit shader but not another due to some kind of mishandling of
register types (the destination register type of the texture operation
is UD while the color source of the render target write is F).
Additionally the instruction scheduler assumed that the combined texture
and render target write operation took 0 cycles, leading to cycle
estimates that are wildly inaccurate. Since the optimization was not
implemented for SIMD32 and our decision whether to use the SIMD32
program is made by comparing the estimated performance with that of the
SIMD16 shader, we wrongly threw out a bunch of SIMD32 programs that are
likely profitable.
total cycles in shared programs: 472807891 -> 473784245 (0.21%)
cycles in affected programs: 108277 -> 1084631 (901.72%)
helped: 0
HURT: 1290
total sends in shared programs: 998955 -> 1000245 (0.13%)
sends in affected programs: 1400 -> 2690 (92.14%)
helped: 0
HURT: 1290
LOST: 0
GAINED: 33
This patch shows no performance changes in Intel's Mesa performance CI.
Given the problems, the lack of evidence that the pass improves
performance, and the fact that the hardware feature was removed from
subsequent GPU generations, I think that the pass is not valuable and
should be removed.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5412>
Just to clarify the missing break is intentional.
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5365>
Add a lowering pass extracting all control barriers embedded in scoped
barriers into proper control barriers so we can get rid of the logic
inserting control barriers when an SpvOpControlBarrier with WorkGroup
scope is parsed in spirv_to_nir().
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4900>
SPIRV OpControlBarrier can have both a memory and a control barrier
which some hardware can handle with a single instruction. Let's
turn the scoped_memory_barrier into a scoped barrier which can embed
both barrier types. Note that control-only or memory-only barriers can
be supported through this new intrinsic by passing NIR_SCOPE_NONE to the
unused barrier type.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Suggested-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4900>
This allows us to do API specific checks before removing variable
without filling nir_remove_dead_variables() with API specific code.
In the following patches we will use this to support the removal
of dead uniforms in GLSL.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4797>
This uses a meson builtin to handle -fvisibility=hidden. This is nice
because we don't need to track which languages are used, if C++ is
suddenly added meson just does the right thing.
Acked-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4740>
Using HALT to immediately jump to the end of the shader is required to
implement GL_EXT_gpu_shader4 and OpenGL 3.0. However, vanilla OpenGL
1.2 doesn't forbid it and it likely makes something somewhere faster.
We should be consistent and implement the same discard behavior on all
hardware if we can.
The rules for HALT on Gen4-5 are a bit different from Gen6+. On the
older hardware, there is no stack for HALT; instead it's up to software
to save and restore mask registers. However, there's no real saving
needed since we only use HALT to jump to the end of the program where
we're about about to do our FB writes. All we need to do is reset AMask
to DMask, the value it was initialized to at the start of the thread.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5244>
This will make the GL drivers pick the right SIMD variant for a given
group size set during dispatch. The heuristic implemented in
brw_cs_simd_size_for_group_size() is the same as in brw_compile_cs().
The cs_prog_data::simd_size field was removed. The generated SIMD
sizes are marked in a bitmask, which is already used via
brw_cs_simd_size_for_group_size() by the drivers.
When in variable group size, it is OK if larger SIMD shader spill,
since we'd need it for the cases where the smaller one can't hold all
the invocations.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5142>
Move the decision one level up, let brw_compile_*() functions use the
spilling information to decide whether or not a certain width
compilation can spill (passed via run_*() functions).
The min_dispatch_width was used to compare with the dispatch_width and
decide whether "a previous shader is already available, so don't
accept spill".
This is replaced by:
- Not calling run_*() functions if it is know beforehand a smaller width
already spilled -- since the larger width will spill and fail;
- Explicitly passing whether or not a shader is allowed to spill. For
the cases where the smaller width is available and haven't spilled,
the larger width will be compiled but is only useful if it won't
spill.
Moving the decision to this level will be useful later for variable
group size, which is a case where we want all the widths to be allowed
to spill.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5142>
It was found that dual-source blending hangs with SIMD16 dispatch in some
specific but unknown situation. Which in the wild happen when rgba
anti-aliasing is enabled for fonts.
Cc: <mesa-stable@lists.freedesktop.org>
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/2183
Signed-off-by: Danylo Piliaiev <danylo.piliaiev@globallogic.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5037>
After 2663759af0 ("intel/fs: Add and use a new load_simd_width_intel
intrinsic") the local_workgroup_size is not used anymore except for
assertions at the pass' start, so drop it from state struct.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5213>
The nir_intrinsic_load_simd_width_intel is always lowered by the
brw_nir_lower_simd() pass before the emission happens. This is likely
a "leftover" from patch rewriting/squashing that happened when this
intrinsic was added.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5213>
Issue description from Matt's commit e7c376ad:
"var_range_end(v, n) loops over the n components of variable number v and
finds the maximum value, giving the last use of any component of v.
Therefore it expects v to correspond to the variable associated with the
.x channel of the VGRF.
var_from_reg() however returns the variable for the first channel of the
VGRF, post-swizzle.
So, if the last register had a swizzle with y, z, or w in the swizzle
component, we would read out of bounds. For any other register, we would
read liveness information from the next register.
The fix is to convert the src_reg to a dst_reg in order to call the
dst_reg version of var_from_reg() that doesn't consider the swizzle."
Closes: #3003
Fixes: 48dfb30f ('intel/compiler: Move all live interval analysis results into vec4_live_variables')
Reviewed-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Andrii Simiklit <asimiklit.work@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4941>
This prevents vectorization for loads/stores that can overflow if
the low offset is negative and the range greater or equal than 0.
The caller can pass the list of variable modes that matter for
robust access.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4881>
The motivating factor is: this lowering may cause
nir_intrinsic_load_local_group_size intrinsics to be added to the
shader, and by moving this around we make possible for the drivers to
lower that intrinsic by themselves.
Iris will do just that in a later patch for implementing variable
group size.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4794>
Intrinsic to get the SIMD width, which not always the same as subgroup
size. Starting with a small scope (Intel), but we can rename it later
to generalize if this turns out useful for other drivers.
Change brw_nir_lower_cs_intrinsics() to use this intrinsic instead of
a width will be passed as argument. The pass also used to optimized
load_subgroup_id for the case that the workgroup fitted into a single
thread (it will be constant zero). This optimization moved together
with lowering of the SIMD.
This is a preparation for letting the drivers call it before the
brw_compile_cs() step.
No shader-db changes in BDW, SKL, ICL and TGL.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4794>
Adding this since Iris will handle variable group size parameters by
itself.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4794>
gen12 does away with the single patch dispatch mode for tcs, and
increases some limits so that 8_patch mode can always work. Make the
necessary changes so we don't try to fall back to single patch mode.
Fixes KHR-GL46.tessellation_shader.single.max_patch_vertices and others
Fixes: 44754279ac ("intel/fs/gen12: Use TCS 8_PATCH mode.")
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4843>
Render Target Array Index has moved from R0.0[26:16] to
R1.1[26:16] on gen12.
Fixes dEQP-VK.multiview.input_attachments.*
Cc: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4836>
We don't care about full IA coherency since we always have the
opportunity in GL or Vulkan to flush the data cache. Using IA-coherent
mode is likely just making A64 access slower than it needs to be.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4819>
The SFID field of the SHADER_OPCODE_MEMORY_FENCE and
SHADER_OPCODE_INTERLOCK instructions now indicates the target function
of the memory fence. Account the cycle-count cost to the right shared
unit.
Fixes: f858fa26b4 ("intel/fs,vec4: Pull stall logic for memory fences up into the IR")
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4817>
Shader-db results on ICL:
total instructions in shared programs: 17133088 -> 17133287 (<.01%)
instructions in affected programs: 61300 -> 61499 (0.32%)
helped: 0
HURT: 199
This means it's likely fixing 199 bugs. :-) All the changed shaders are
in Mad Max. It's surprisingly difficult to get the back-end compiler to
generate a pattern that hits this we don't tend to emit a lot coalescable
MOVs. The pattern in Mad Max that's able to hit is fsign(fsat(x)) under
the right conditions.
Closes: #2820
Cc: mesa-stable@lists.freedesktop.org
Tested-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4773>
In Gen11+, when emitting a fence for both L3 and SLM, the generated
code would look like
SEND, MOV (for stall), SEND, MOV (for stall)
This commit change that so two SENDs are emitted before the MOVs for
stall. This is similar to the approach used in Ivy Bridge for the
render fence.
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3278>
Instead of emitting the stall MOV "inside" the
SHADER_OPCODE_MEMORY_FENCE generation, use the scheduling fences when
creating the IR.
For IvyBridge, every (data cache) fence is accompained by a render
cache fence, that now is explicit in the IR, two
SHADER_OPCODE_MEMORY_FENCEs are emitted (with different SFIDs).
Because Begin and End interlock intrinsics are effectively memory
barriers, move its handling alongside the other memory barrier
intrinsics. The SHADER_OPCODE_INTERLOCK is still used to distinguish
if we are going to use a SENDC (for Begin) or regular SEND (for End).
This change is a preparation to allow emitting both SENDs in Gen11+
before we can stall on them.
Shader-db results for IVB (i965):
total instructions in shared programs: 11971190 -> 11971200 (<.01%)
instructions in affected programs: 11482 -> 11492 (0.09%)
helped: 0
HURT: 8
HURT stats (abs) min: 1 max: 3 x̄: 1.25 x̃: 1
HURT stats (rel) min: 0.03% max: 0.50% x̄: 0.14% x̃: 0.10%
95% mean confidence interval for instructions value: 0.66 1.84
95% mean confidence interval for instructions %-change: 0.01% 0.27%
Instructions are HURT.
Unlike the previous code, that used the `mov g1 g2` trick to force
both `g1` and `g2` to stall, the scheduling fence will generate `mov
null g1` and `mov null g2`. During review it was decided it was not
worth keeping the special codepath for the small effect will have.
Shader-db results for HSW (i965), BDW and SKL don't have a change
on instruction count, but do report changes in cycles count, showing
SKL results below
total cycles in shared programs: 341738444 -> 341710570 (<.01%)
cycles in affected programs: 7240002 -> 7212128 (-0.38%)
helped: 46
HURT: 5
helped stats (abs) min: 14 max: 1940 x̄: 676.22 x̃: 154
helped stats (rel) min: <.01% max: 2.62% x̄: 1.28% x̃: 0.95%
HURT stats (abs) min: 2 max: 1768 x̄: 646.40 x̃: 362
HURT stats (rel) min: <.01% max: 0.83% x̄: 0.28% x̃: 0.08%
95% mean confidence interval for cycles value: -777.71 -315.38
95% mean confidence interval for cycles %-change: -1.42% -0.83%
Cycles are helped.
This seems to be the effect of allocating two registers separatedly
instead of a single one with size 2, which causes different register
allocation, affecting the cycle estimates.
while ICL also has not change on instruction count but report changes
negative changes in cycles
total cycles in shared programs: 352665369 -> 352707484 (0.01%)
cycles in affected programs: 9608288 -> 9650403 (0.44%)
helped: 4
HURT: 104
helped stats (abs) min: 24 max: 128 x̄: 88.50 x̃: 101
helped stats (rel) min: <.01% max: 0.85% x̄: 0.46% x̃: 0.49%
HURT stats (abs) min: 2 max: 2016 x̄: 408.36 x̃: 48
HURT stats (rel) min: <.01% max: 3.31% x̄: 0.88% x̃: 0.45%
95% mean confidence interval for cycles value: 256.67 523.24
95% mean confidence interval for cycles %-change: 0.63% 1.03%
Cycles are HURT.
AFAICT this is the result of the case above.
Shader-db results for TGL have similar cycles result as ICL, but also
affect instructions
total instructions in shared programs: 17690586 -> 17690597 (<.01%)
instructions in affected programs: 64617 -> 64628 (0.02%)
helped: 55
HURT: 32
helped stats (abs) min: 1 max: 16 x̄: 4.13 x̃: 3
helped stats (rel) min: 0.05% max: 2.78% x̄: 0.86% x̃: 0.74%
HURT stats (abs) min: 1 max: 65 x̄: 7.44 x̃: 2
HURT stats (rel) min: 0.05% max: 4.58% x̄: 1.13% x̃: 0.69%
95% mean confidence interval for instructions value: -2.03 2.28
95% mean confidence interval for instructions %-change: -0.41% 0.15%
Inconclusive result (value mean confidence interval includes 0).
Now that more is done in the IR, more dependencies are visible and
more SWSB annotations are emitted. Mixed with different register
allocation decisions like above, some shaders will see more `sync
nops` while others able to avoid them.
Most of the new `sync nops` are also redundant and could be dropped,
which will be fixed in a separate change.
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3278>
The cycle count estimation logic part of the scheduler is now
redundant with the shader performance modeling pass, and the estimates
can be consolidated into the brw::performance analysis result object
instead of being part of the CFG, which guarantees that the estimates
cannot be accessed without previously calling the
performance_analysis::require() method, which makes sure that the
right analysis pass is executed at the right time if we don't already
have up-to-date cached results.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
So we can eventually remove the cycle count estimates from the CFG
data structure and consolidate performance information in the
brw::performance object.
It would be cleaner to pass the brw::performance object directly to
the disassembler but that isn't straightforward since the disassembler
is built as a plain C file unlike the rest of the compiler back-end.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
These should be more accurate than the current cycle counts, since
among other things they consider the effect of post-scheduling passes
like the software scoreboard on TGL. In addition it will enable us to
clean up some of the now redundant cycle-count estimation
functionality in the instruction scheduler.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This is useful in order to identify codegen issues caused by SIMD32.
It doesn't currently have any effect on compute shaders since SIMD32
dispatch is only enabled for CS when it's strictly necessary to do so
in order to support the workgroup size requested for the shader --
That might change in the future though when we hook up the SIMD32
heuristic to CS compilation.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
