aed4c0b5a9
nir: Drop the unused instr arg for src/dest copy functions.
...
Now that we don't use ralloc, we don't need this arg to get at the right
ralloc ctx.
Reviewed-by: Matt Turner <mattst88@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11776 >
2021-09-14 17:53:06 +00:00
cfc4433015
nir,glsl_to_nir: use nir_fdot()
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com >
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8056 >
2021-08-16 17:19:45 +00:00
174a4f36f9
nir: create ffma from builders more often
...
We will not be able to combine instructions into ffma later if they are
exact, so create them from the start. They can be lowered later if they
are unwanted.
fossil-db (GFX10.3):
Totals from 16589 (11.34% of 146267) affected shaders:
VGPRs: 938872 -> 938704 (-0.02%)
SpillSGPRs: 11334 -> 10785 (-4.84%)
CodeSize: 96551964 -> 96498040 (-0.06%); split: -0.08%, +0.02%
MaxWaves: 338760 -> 338772 (+0.00%)
Instrs: 18356857 -> 18350486 (-0.03%); split: -0.06%, +0.02%
Latency: 561563310 -> 561414360 (-0.03%); split: -0.08%, +0.05%
InvThroughput: 145629673 -> 145594740 (-0.02%); split: -0.04%, +0.01%
fossil-db (GFX10):
Totals from 16252 (11.11% of 146267) affected shaders:
VGPRs: 893820 -> 893744 (-0.01%)
SpillSGPRs: 11334 -> 10785 (-4.84%)
CodeSize: 95890244 -> 95839124 (-0.05%); split: -0.08%, +0.02%
MaxWaves: 367704 -> 367734 (+0.01%)
Instrs: 18199741 -> 18194437 (-0.03%); split: -0.06%, +0.03%
Latency: 560912971 -> 560854179 (-0.01%); split: -0.07%, +0.06%
InvThroughput: 142899814 -> 142877939 (-0.02%); split: -0.03%, +0.02%
fossil-db (GFX9):
Totals from 16287 (11.12% of 146401) affected shaders:
SGPRs: 1312784 -> 1312768 (-0.00%); split: -0.05%, +0.05%
VGPRs: 931440 -> 931444 (+0.00%); split: -0.00%, +0.00%
SpillSGPRs: 14623 -> 14597 (-0.18%)
CodeSize: 94428788 -> 94344404 (-0.09%); split: -0.10%, +0.01%
MaxWaves: 90105 -> 90109 (+0.00%)
Instrs: 18486905 -> 18473434 (-0.07%); split: -0.08%, +0.01%
Latency: 720947295 -> 720818323 (-0.02%); split: -0.07%, +0.05%
InvThroughput: 365240104 -> 365224659 (-0.00%); split: -0.02%, +0.01%
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com >
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8056 >
2021-08-16 17:19:45 +00:00
82d0600ba2
nir: swap fadd operands in nir_atan()
...
This shouldn't do anything but will make testing a later patch easier.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com >
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8056 >
2021-08-16 17:19:44 +00:00
09440ce3fb
nir: Fix MSVC warning C4334 (32bit shift cast to 64bit)
...
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Reviewed-By: Bill Kristiansen <billkris@microsoft.com >
Cc: mesa-stable@lists.freedesktop.org
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10331 >
2021-04-20 00:28:34 +00:00
b2da598ff9
nir: Use sized types for nir_tex_instr::dest_type
...
Revieweeviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7989 >
2021-01-25 11:21:48 +01:00
a42c846d24
nir: Fix nextafter() for hardware that don't support denorms
...
We need to make sure we never return a denorm float, either by flushing
the denorm to 0 or by adjusting the minimum non-zero value.
v2 (Rhys): Use shader float controls execution mode instead of a dedicated option
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com >
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com >
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7565 >
2020-11-18 04:05:37 +00:00
09bca4cb95
vtn/opencl: Switch some nir-sequence ops to use libclc
...
All of these are pretty well-defined. Rather than implementing them
as a sequence of nir ops, we can just use the libclc implementation.
v2 (idr): Delete functions that are now unused.
Acked-by: Jason Ekstrand <jason@jlekstrand.net >
Reviewed-by: Dave Airlie <airlied@redhat.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6035 >
2020-09-25 13:14:45 -07:00
89401e5867
nir: More NIR_MAX_VEC_COMPONENTS fixes
...
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6655 >
2020-09-09 20:19:42 +00:00
1598370aca
nir/builder: Return an integer from nir_get_texture_size
...
It's convenient in a bunch of cases for nir_get_texture_size to return a
float but it's very unexpected. This fixes a bug in the R600 NIR code.
Fixes: f718ac6268eric@anholt.net >
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3897 >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3897 >
2020-02-21 18:48:03 +00:00
6c9495b392
nir: make nir_get_texture_size/lod available outside nir_lower_tex
...
This functions can be useful in other places.
Signed-off-by: Gert Wollny <gert.wollny@collabora.com >
Reviewed-by: Eric Anholt <eric@anholt.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3286 >
2020-01-04 16:22:40 +00:00
2098ae16c8
nir/builder: Move nir_atan and nir_atan2 from SPIR-V translator
...
Moves build_atan and build_atan2 into nir_builtin_builder. The goal is
to be able to use this from the GLSL translator too.
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com >
2019-10-12 09:43:17 +02:00
ddd08e1888
nir/builder: Remove the use_fmov parameter from nir_swizzle
...
This flag has caused more confusion than good in most cases. You can
validly use imov for floats or fmov for integers because, without source
modifiers, neither modify their input in any way. Using imov for floats
is more reliable so we go that direction.
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com >
Acked-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
2019-05-24 08:38:11 -05:00
272e927d0e
nir/spirv: initial handling of OpenCL.std extension opcodes
...
Not complete, mostly just adding things as I encounter them in CTS. But
not getting far enough yet to hit most of the OpenCL.std instructions.
Anyway, this is better than nothing and covers the most common builtins.
v2: add hadd proof from Jason
move some of the lowering into opt_algebraic and create new nir opcodes
simplify nextafter lowering
fix normalize lowering for inf
rework upsample to use nir_pack_bits
add missing files to build systems
v3: split lines of iadd/sub_sat expressions
Signed-off-by: Karol Herbst <kherbst@redhat.com >
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
2019-03-05 22:28:29 +01:00
2083cfb6eb
nir: add builtin builder
...
also move some of the GLSL builtins over we will need for implementing
some OpenCL builtins
v2: replace NIR_IMM_FP by nir_imm_floatN_t in ported code
fix up changes caused by swizzle rework
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Signed-off-by: Karol Herbst <kherbst@redhat.com >
2018-07-24 20:40:05 +02:00
Emma Anholt