Commit Graph

7 Commits

Author SHA1 Message Date
Eric Anholt eda3e4e055 nir/builder: Add a name format arg to nir_builder_init_simple_shader().
This cleans up a bunch of gross sprintfs and keeps the caller from needing
to remember to ralloc_strdup.  I added a couple of '"%s", name ? name :
""' to radv where I didn't fully trace through whether a non-null name was
being passed in.

I also took the liberty of adding a basic name to a few shaders (pan_blit,
unit tests)

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7323>
2020-11-11 08:50:29 -08:00
Eric Anholt 5f992802f5 nir/builder: Drop the mem_ctx arg from nir_builder_init_simple_shader().
This looks a lot more simple now!

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7323>
2020-11-11 08:50:29 -08:00
Eric Anholt 4e9328e3b6 nir_builder: Return a new builder from nir_builder_init_simple_shader().
It's a little inline function, so we can just RAII it for better
ergonomics.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7323>
2020-11-11 08:49:49 -08:00
Lionel Landwerlin 8b913bd1ce nir/tests: take reference on glsl types
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
2019-08-21 09:44:10 +02:00
Ian Romanick 12217de08c nir: Handle swizzle in nir_alu_srcs_negative_equal
When I added this function, I was not sure if swizzles of immediate
values were a thing that occurred in NIR.  The only existing user of
these functions is the partial redundancy elimination for compares.
Since comparison instructions are inherently scalar, this does not
occur.

However, a couple later patches, "nir/algebraic: Recognize
open-coded flrp(-1, 1, a) and flrp(1, -1, a)" combined with "intel/vec4:
Try to emit a single load for multiple 3-src instruction operands",
collaborate to create a few thousand instances.

No shader-db changes on any Intel platform.

v2: Handle the swizzle in nir_alu_srcs_negative_equal and leave
nir_const_value_negative_equal unchanged.  Suggested by Jason.

v3: Correctly handle write masks.  Add note (and assertion) that the
caller is responsible for various compatibility checks.  The single
existing caller only calls this for combinations of scalar fadd and
float comparison instructions, so all of the requirements are met.  A
later patch (intel/vec4: Try to emit a single load for multiple 3-src
instruction operands) will call this for sources of the same
instruction, so all of the requirements are met.

v4: Add unit test for nir_opt_comparison_pre that is fixed by this
commit.

Reviewed-by: Matt Turner <mattst88@gmail.com>
2019-07-08 11:30:11 -07:00
Ian Romanick 0ac5ff9ecb nir: Use nir_src_bit_size instead of alu1->dest.dest.ssa.bit_size
This is important because, for example nir_op_fne has
dest.dest.ssa.bit_size == 1, but the source operands can be 16-, 32-, or
64-bits.  Fixing this helps partial redundancy elimination for compares
in a few more shaders.

v2: Add unit tests for nir_opt_comparison_pre that are fixed by this
commit.

All Intel platforms had similar results.
total instructions in shared programs: 17179408 -> 17179081 (<.01%)
instructions in affected programs: 43958 -> 43631 (-0.74%)
helped: 118
HURT: 2
helped stats (abs) min: 1 max: 5 x̄: 2.87 x̃: 2
helped stats (rel) min: 0.06% max: 4.12% x̄: 1.19% x̃: 0.81%
HURT stats (abs)   min: 6 max: 6 x̄: 6.00 x̃: 6
HURT stats (rel)   min: 5.83% max: 6.06% x̄: 5.94% x̃: 5.94%
95% mean confidence interval for instructions value: -3.08 -2.37
95% mean confidence interval for instructions %-change: -1.30% -0.85%
Instructions are helped.

total cycles in shared programs: 360959066 -> 360942386 (<.01%)
cycles in affected programs: 774274 -> 757594 (-2.15%)
helped: 111
HURT: 4
helped stats (abs) min: 1 max: 1591 x̄: 169.49 x̃: 36
helped stats (rel) min: <.01% max: 24.43% x̄: 8.86% x̃: 2.24%
HURT stats (abs)   min: 1 max: 2068 x̄: 533.25 x̃: 32
HURT stats (rel)   min: 0.02% max: 5.10% x̄: 3.06% x̃: 3.56%
95% mean confidence interval for cycles value: -200.61 -89.47
95% mean confidence interval for cycles %-change: -10.32% -6.58%
Cycles are helped.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net> [v1]
Suggested-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Fixes: be1cc3552b ("nir: Add nir_const_value_negative_equal")
2019-07-08 11:30:10 -07:00
Ian Romanick b08d704051 nir: Add unit tests for nir_opt_comparison_pre
Each tests has a comment with the expected before and after NIR.  The
tests don't actually check this.  The tests only check whether or not
the optimization pass reported progress.  I couldn't think of a robust,
future-proof way to check the before and after code.

Reviewed-by: Matt Turner <mattst88@gmail.com>
2019-07-08 11:30:10 -07:00