There are a lot of optimizations in opt_algebraic that match ('ine', a,
0), but there are almost none that match i2b. Instead of adding a huge
pile of additional patterns (including variations that include both ine
and i2b), always lower i2b to a != 0.
At this point in the series, it should be impossible for anything to
generate i2b, so there /should not/ be any changes.
The failing test on d3d12 is a pre-existing bug that is triggered by
this change. I talked to Jesse about it, and, after some analysis, he
suggested just adding it to the list of known failures.
v2: Don't rematerialize i2b instructions in dxil_nir_lower_x2b.
v3: Don't rematerialize i2b instructions in zink_nir_algebraic.py.
v4: Fix zink-on-TGL CI failures by calling nir_opt_algebraic after
nir_lower_doubles makes progress. The latter can generate b2i
instructions, but nir_lower_int64 can't handle them (anymore).
v5: Add back most of the hunk at line 2125 of nir_opt_algebraic.py. I
had accidentally removed the f2b(bf2(x)) optimization.
v6: Just eliminate the i2b instruction.
v7: Remove missed i2b32 in midgard_compile.c. Remove (now unused)
emit_alu_i2orf2_b1 function from sfn_instr_alu.cpp. Previously this
function was still used. 🤷
No shader-db changes on any Intel platform.
All Intel platforms had similar results. (Ice Lake shown)
Instructions in all programs: 141165875 -> 141165873 (-0.0%)
Instructions helped: 2
Cycles in all programs: 9098956382 -> 9098956350 (-0.0%)
Cycles helped: 2
The two Vulkan shaders are helped because of the "new" (('b2i32',
('ine', ('ubfe', a, b, 1), 0)), ('ubfe', a, b, 1)) algebraic pattern.
Acked-by: Jesse Natalie <jenatali@microsoft.com> [earlier version]
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Daniel Schürmann <daniel@schuermann.dev> [earlier version]
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15121>
Starting with !19748 lowered 64 bit shifts were showing wrong results for
shifts with insignificant bits set.
nir shifts are defined to only look at the least significant bits. The
lowering has take this into account.
So there are two things going on:
1. the `ieq` and `uge` further down depend on `y` being masked.
2. the calculation of `reverse_count` actually depends on a masked `y` as
well, due to the `(iabs (iadd y -32))` giving a different result for
shifts > 31;
Fixes: 41f3e9e5f5 ("nir: Implement lowering of 64-bit shift operations")
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19995>
Currently float16 to int64 conversions don't work correctly, because
the "div" variable has an infinite value, since 2^32 isn't
representable as a 16-bit float, which causes the result of of rem(x,
div) to be NaN for all inputs, leading to an incorrect result. Since
no values of magnitude greater than 2^32 are representable as a
float16 we don't actually need to do the fdiv/frem operations, the
conversion is equivalent to f2u32 with the result padded to 64 bits.
Rework:
* Jordan: Handle f16 in if/else rather than conditional
Fixes: 936c58c8fc ("nir: Extend nir_lower_int64() to support i2f/f2i lowering")
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19391>
This involves computing the significand with a 64-bit precision type,
and implementing the normalization and packing manually instead of
relying on u2f32, since the significand can no longer be represented
as a 32-bit integer. This fixes 64-bit integer to 64-bit float
conversions on devices that support 64-bit float natively but lack
64-bit integer support, like Intel MTL hardware.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com> (v1)
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19128>
The existing code for this appears to work okay for conversions
involving 64-bit floats, relax the assert and enable the lowering
path. This fixes 64-bit float to 64-bit integer integer conversions
on devices that have native support for 64-bit floats but lack 64-bit
integer support, like Intel MTL hardware.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19128>
One exception is src/amd/addrlib/, for which -Wimplicit-fallthrough is
explicitly disabled.
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Juan A. Suarez <jasuarez@igalia.com>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10220>
These are all pretty trivial because we can just split the op into one
subgroup op per half of the value. There's some question as to whether
these belong in lower_int64 or lower_subgroups but, on Intel, they key
decider of whether or not we need the lowering is based on whether or
not we have hardware int64 support.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7329>
These consist of the variations nir_op_{i|u|f}{min|max|med}3 which are either
lowered in the backend (LLVM) anyway or can be recombined by the backend (ACO).
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6421>
I was doing math manually in a lowering pass for converting a division to
a ushr, and this will let the pass be expressed more naturally.
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6378>
The round-to-nearest-even implementation found in lower_2f() is incorrect
for any value having a significand that is not directly representable
and whose non-representable part lies between 1 and half the minimum
representable value. In this case, the significand is rounded up instead
of being rounded down.
Fixes: 936c58c8fc ("nir: Extend nir_lower_int64() to support i2f/f2i lowering")
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Tested-by: Jesse Natalie <jenatali@microsoft.com>
Acked-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6290>
That's an attempt at replacing the complex __int64_to_float() and
__float_to_int64() implementations found in float64.glsl by a simpler
native NIR equivalent.
Thanks to that, we can have lower those conversion without having to
compile a GLSL shader, which would be quite annoying for OpenCL
kernels.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5588>
This information is exposed through shader->options->lower_int64_options.
Removing the extra arg forces drivers to initialize this field correctly.
This also allows us to check the int64 lowering options from each int64
lowering helper and decide if we should lower the instructions we
introduce.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Rob Clark <robdclark@chromium.org>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5588>
Fixes an issue with Renderdoc's shader debugging with ACO.
If nir_opt_algebraic isn't called in-between nir_lower_explicit_io and
nir_lower_int64, we can end up with 64-bit multiplications.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Fixes: 6320e37d4b ('nir: add amul instruction')
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5709>
Fixes dEQP-VK.spirv_assembly.instruction.amd_trinary_minmax.*.i64.*
with ACO because this backend compiler expects most of the 64-bit
operations to be lowered.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4570>
We have the code to do the lowering, we were just missing the
boilerplate bits to make should_lower_int64_alu_instr return true.
Fixes: 62d55f1281 "nir: Wire up int64 lowering functions"
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4365>
This adds the option to lower 64-bit ufind_msb opcodes.
v2: use split_x/y removes component loops (Jason)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
One advantage of this is that we no longer need to run in a loop because
the new framework handles lowering instructions added by lowering.
Reviewed-by: Eric Anholt <eric@anholt.net>
We need this when doing full software 64-bit emulation.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110309
Fixes: cbad201c2b "nir/algebraic: Add missing 64-bit extract_[iu]8..."
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
We already have one internally for int64 but we don't have a similar one
for doubles so we'll have to make one.
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
On Gen 8 and 9, "mul" instruction supports 64 bit destination type. We
can reduce our 64x64 int multiplication from 4 instructions to 3.
Also instead of emitting two mul instructions, we can emit single mul
instuction and extract low/high 32 bits from 64 bit result for
[i/u]mulExtended
v2: 1) Allow lower_mul_high64 to use new opcode (Jason Ekstrand)
2) Add lower_mul_2x32_64 flag (Matt Turner)
3) Remove associative property as bit size is different (Connor
Abbott)
v3: Fix indentation and variable naming convention (Jason Ekstrand)
Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
NIR metadata validation verifies that the debug bit was unset (by a call
to nir_metadata_preserve) if a NIR optimization pass made progress on
the shader. With the expectation that the NIR shader consists of only a
single main function, it has been safe to call nir_metadata_preserve()
iff progress was made.
However, most optimization passes calculate progress per-function and
then return the union of those calculations. In the case that an
optimization pass makes progress only on a subset of the functions in
the shader metadata validation will detect the debug bit is still set on
any unchanged functions resulting in a failed assertion.
This patch offers a quick solution (short of a larger scale refactoring
which I do not wish to undertake as part of this series) that simply
unsets the debug bit on unchanged functions.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Instead of a single i2b and b2i, we now have i2b32 and b2iN where N is
one if 8, 16, 32, or 64. This leads to having a few more opcodes but
now everything is consistent and booleans aren't a weird special case
anymore.
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
v2: use nir_metadata_preserve
preserve metadata in case of !progress
Fixes: 074f5ba0b5
"nir: Add a simple int64 lowering pass"
Signed-off-by: Karol Herbst <kherbst@redhat.com>
The previous implementation was fine for GLSL which doesn't really have
a signed modulus/remainder. They just leave the behavior undefined
whenever either source is negative. However, in SPIR-V, there is a
defined behavior for negative arguments. This commit beefs up the pass
so that it handles both correctly. Tested using a hacked up version of
the Vulkan CTS test to get 64-bit support.
Reviewed-by: Matt Turner <mattst88@gmail.com>
The algorithms used by this pass, especially for division, are heavily
based on the work Ian Romanick did for the similar int64 lowering pass
in the GLSL compiler.
v2: Properly handle vectors
v3: Get rid of log2_denom stuff. Since we're using bcsel, we do all the
calculations anyway and this is just extra instructions.
v4:
- Add back in the log2_denom stuff since it's needed for ensuring that
the shifts don't overflow.
- Rework the looping part of the pass to be easier to expand.
Reviewed-by: Matt Turner <mattst88@gmail.com>
