This late optimization pass is only affected by nir_opt_if() and handles all cases
in a single pass. It's enough to call it once after the optimization loop.
No changes on vkpipeline-db.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
For es_vgpr_comp_cnt.
Fixes: 795adbbadd "radv/gfx10: Add pipeline state support for tess."
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Enabling tracing, and then having a vmfault, can leads to a segfault
before we print out the traces, as if a meta shader is executing
and we don't have the NIR for it.
Just pass the stage and give back a default.
Fixes: 9b9ccee4d6 ("radv: take LDS into account for compute shader occupancy stats")
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
I did implement this extension a while ago but it didn't work
on pre GFX10 for some reasons. Now all CTS pass.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
This is unsupported and hangs.
This fixes GPU hangs with
dEQP-VK.tessellation.geometry_interaction.limits.output_required_*.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Only VS needs that. We shouldn't hardcode these values but
that's complicated to not do that for now.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Currently this is done rather late in radv, after lowering booleans, so
it isn't safe to run additional optimizations that may add e.g. 1-bit
booleans. We could move the lowering parts earlier, but since right now
we only lower FS inputs and by this point all indirects have been
lowered away, there's no reason we should need to optimize anything.
One shader from Devil May Cry 5 was getting optimized, but only because
the optimization loop was working on 32-bit booleans which revealed an
opportunity that was hidden with 1-bit booleans, and we generated a
1-bit boolean which is invalid.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111092
Fixes: 118a66df99
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
We have to add a few lowering to deal with things that used to be dealt
with inline when creating inputs. We also move the code that fills out
the radv_shader_variant_info struct for linking purposes to
radv_shader.c, as it's no longer tied to the NIR->LLVM lowering.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
load_fragcoord is already handled in common code for radeonsi, so we
don't need to do anything to handle it. However, there were some passes
creating NIR with the varying, so we switch them over to the sysval. In
the case of nir_lower_input_attachments which is used by both radv and
anv, we add handling for both until intel switches to using a sysval.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
This needs to be cleaned up a bit, and it probably contains
missing stuff and/or bugs.
This doesn't fix the "half of the triangles" issue.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Transform feedback is really different on GFX10.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
This simplifies a bunch of stuff by
(1) Keeping all the things in a single allocation, making things easier
for the cache.
(2) creating a shader_variant creation helper.
This is immediately put to use by creating rtld shader binaries. This
is the main reason for the binaries, as we need to do the linking at
upload time, i.e. post caching. We do not enable rtld yet.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Just a cleanup, it shouldn't change anything.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Just move around the switch case. GFX9+ is handled below.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
spirv_to_nir() returned the nir_function corresponding to the
entrypoint, as a way to identify it. There's now a bool is_entrypoint
in nir_function and also a helper function to get the entry_point from
a nir_shader.
The return type reflects better what the function name suggests. It
also helps drivers avoid the mistake of reusing internal shader
references after running NIR_PASS on it. When using NIR_TEST_CLONE or
NIR_TEST_SERIALIZE, those would be invalidated right in the first pass
executed.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Replace its uses with checking for is_entrypoint and calling
nir_shader_get_entrypoint().
This is a preparation to change spirv_to_nir() return type.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Instead of setting the glsl types of the pointers for each resource,
set the nir_address_format, from which we can derive the glsl_type,
and in the future the bit pattern representing a NULL pointer.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
The load/store optimizer pass doesn't handle WaW hazards correctly
and this is the root cause of the reflection issue with Monster
Hunter World. AFAIK, it's the only game that are affected by this
issue.
This is fixed with LLVM r361008, but we need a workaround for older
LLVM versions unfortunately.
Cc: "19.0" "19.1" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
We already use GFX9 and I don't want us to have confusing naming
in the driver. GFXn naming is better from the driver perspective,
because it's the real version of the gfx portion of the hw. Also,
CIK means Bonaire-Kaveri-Kabini, it doesn't mean CI.
It shouldn't confuse our SDMA, UVD, VCE etc. code much. Those have
nothing to do with GFXn and they have their own version numbers.
This can be used by both etnaviv and freedreno/a2xx as they are both vec4
architectures with some instructions being scalar-only.
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
I don't know why I thought NIR_PASS always set the progress variable.
Derp.
Fixes: d41cdef2a5 ("nir: Use the flrp lowering pass instead of nir_opt_algebraic")
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Coverity CID: 1444996
Coverity CID: 1444995
Coverity CID: 1444994
Coverity CID: 1444993
Coverity CID: 1444991
Coverity CID: 1444989
I tried to be very careful while updating all the various drivers, but I
don't have any of that hardware for testing. :(
i965 is the only platform that sets always_precise = true, and it is
only set true for fragment shaders. Gen4 and Gen5 both set lower_flrp32
only for vertex shaders. For fragment shaders, nir_op_flrp is lowered
during code generation as a(1-c)+bc. On all other platforms 64-bit
nir_op_flrp and on Gen11 32-bit nir_op_flrp are lowered using the old
nir_opt_algebraic method.
No changes on any other Intel platforms.
v2: Add panfrost changes.
Iron Lake and GM45 had similar results. (Iron Lake shown)
total cycles in shared programs: 188647754 -> 188647748 (<.01%)
cycles in affected programs: 5096 -> 5090 (-0.12%)
helped: 3
HURT: 0
helped stats (abs) min: 2 max: 2 x̄: 2.00 x̃: 2
helped stats (rel) min: 0.12% max: 0.12% x̄: 0.12% x̃: 0.12%
Reviewed-by: Matt Turner <mattst88@gmail.com>
This enables the remaining capabilities in SPV_EXT_descriptor_indexing.
Fixes: 0e10790558 "radv: Enable VK_EXT_descriptor_indexing."
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Only computeDerivativeGroupLinear is supported for now.
All crucible tests pass.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
No support for 64-bit compare&swap atomic operations.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
When I implemented opt_if_loop_last_continue() I had restricted
this pass from moving other if-statements inside the branch opposite
the continue. At the time it was causing a bunch of spilling in
shader-db for i965.
However Samuel Pitoiset noticed that making this pass more aggressive
significantly improved the performance of Doom on RADV. Below are
the statistics he gathered.
28717 shaders in 14931 tests
Totals:
SGPRS: 1267317 -> 1267549 (0.02 %)
VGPRS: 896876 -> 895920 (-0.11 %)
Spilled SGPRs: 24701 -> 26367 (6.74 %)
Code Size: 48379452 -> 48507880 (0.27 %) bytes
Max Waves: 241159 -> 241190 (0.01 %)
Totals from affected shaders:
SGPRS: 23584 -> 23816 (0.98 %)
VGPRS: 25908 -> 24952 (-3.69 %)
Spilled SGPRs: 503 -> 2169 (331.21 %)
Code Size: 2471392 -> 2599820 (5.20 %) bytes
Max Waves: 586 -> 617 (5.29 %)
The codesize increases is related to Wolfenstein II it seems largely
due to an increase in phis rather than the existing jumps.
This gives +10% FPS with Doom on my Vega56.
Rhys Perry also benchmarked Doom on his VEGA64:
Before: 72.53 FPS
After: 80.77 FPS
v2: disable pass on non-AMD drivers
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com> (v1)
Acked-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Only 8-bit integers for now, float16 requires a bit more work.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
This lowering isn't needed for RADV because AMDGCN has two
instructions. It will be disabled for RADV in an upcoming series.
While we are at it, factorize a little bit.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Fixes: 58bcebd987 ("spirv: Allow [i/u]mulExtended to use new nir opcode")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
the naming is a bit confusing no matter how you look at it. Within SPIR-V
"global" memory is memory accessible from all threads. glsl "global" memory
normally refers to shader thread private memory declared at global scope. As
we already use "shared" for memory shared across all thrads of a work group
the solution where everybody could be happy with is to rename "global" to
"private" and use "global" later for memory usually stored within system
accessible memory (be it VRAM or system RAM if keeping SVM in mind).
glsl "local" memory is memory only accessible within a function, while SPIR-V
"local" memory is memory accessible within the same workgroup.
v2: rename local to function as well
v3: rename vtn_variable_mode_local as well
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
For now, it's hidden behind a cap. Hopefully, we can eventually drop
that along with all the manual offset code in spirv_to_nir.
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Tested-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Instead of baking in uvec2 for UBO and SSBO pointers and uint for push
constant and shared memory pointers, make it configurable.
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
We're going to want to do more deref optimizations going forward and
this gives us a central place to do them. Also, cast propagation will
get a bit more complicated with the addition of ptr_as_array derefs.
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
This workaround has been introduced by 135e4d434f for fixing
DXVK GPU hangs with many games. It is no longer needed since
LLVM r345718.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
On some GPUs, especially older Intel GPUs, some math instructions are
very expensive. On those architectures, don't reduce flow control to a
csel if one of the branches contains one of these expensive math
instructions.
This prevents a bunch of cycle count regressions on pre-Gen6 platforms
with a later patch (intel/compiler: More peephole select for pre-Gen6).
v2: Remove stray #if block. Noticed by Thomas.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Thomas Helland <thomashelland90@gmail.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
That flow control may be trying to avoid invalid loads. On at least
some platforms, those loads can also be expensive.
No shader-db changes on any Intel platform (even with the later patch
"intel/compiler: More peephole select").
v2: Add a 'indirect_load_ok' flag to nir_opt_peephole_select. Suggested
by Rob. See also the big comment in src/intel/compiler/brw_nir.c.
v3: Use nir_deref_instr_has_indirect instead of deref_has_indirect (from
nir_lower_io_arrays_to_elements.c).
v4: Fix inverted condition in brw_nir.c. Noticed by Lionel.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
User are encouraged to switch to LLVM 7.0 released in September 2018.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Helpful for debugging compiler backend problems: this allows us to
easily retrieve the LLVM IR from RenderDoc.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
This implementation should work and potential bugs can be
fixed during the release candidates window anyway.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Totals from affected shaders:
SGPRS: 1112 -> 1112 (0.00 %)
VGPRS: 1492 -> 1196 (-19.84 %)
Spilled SGPRs: 0 -> 0 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Private memory VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 112172 -> 101316 (-9.68 %) bytes
LDS: 0 -> 0 (0.00 %) blocks
Max Waves: 93 -> 98 (5.38 %)
Wait states: 0 -> 0 (0.00 %)
All affected shaders are from "Batman: Arkham City" over DXVK.
The pass detects that the temporary array created by DXVK for
storing TCS inputs is a copy of the input arrays and allows
us to avoid copying all of the input data and then indirecting
on it with if-ladders, instead we just do indirect indexing.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
This allows NIR to CSE more operations. LLVM does this also so the
impact is limited, however doing this in NIR allows other opts to
make progress. For example in radeonsi more loops are unrolled in
Civilization Beyond Earth.
The actual pipeline-db stats are not overwhelming but even in the
negatively affected shaders the NIR is clearly better. It just
happens that the code shuffling and in some cases calls to max
rather than a flt result in the final output from LLVM not
giving as good numbers.
However this is an incremental opt that further passes build off
so the change should be made IMO.
Totals from affected shaders:
SGPRS: 20192 -> 20184 (-0.04 %)
VGPRS: 19516 -> 19524 (0.04 %)
Spilled SGPRs: 437 -> 444 (1.60 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Private memory VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 1527444 -> 1522276 (-0.34 %) bytes
LDS: 6 -> 6 (0.00 %) blocks
Max Waves: 1018 -> 1016 (-0.20 %)
Wait states: 0 -> 0 (0.00 %)
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
All CTS pass on Polaris/Vega with LLVM 6, 7 and master, so
I think it's safe to enable the feature.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Not sure if this is all wired up. CTS does pass and the Tangrams
demo works fine on Vega. There are corruption issues on Polaris
but not sure if that related to 16-bit support.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
It was very inconsistently handled; the only things that made use of it
were glsl_to_nir, glspirv, and nir_gather_info. In particular,
nir_lower_io completely ignored it so anyone using nir_lower_io on
64-bit vertex attributes was going to be in for a shock. Also, as of
the previous commit, it's set by every driver that supports 64-bit
vertex attributes. There's no longer any reason to have it be an option
so let's just delete it.
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Otherwise, nir_lower_clip_cull_distance_arrays might report
wrong number of output clips/culls because it relies on
shader output variables and some of them might be dead.
This fixes a rendering issue with Dolphin and Super Mario
Sunshine.
Fixes: b0c643d8f5 ("spirv: Use NIR per-member splitting")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107610
CC: 18.2 <mesa-stable@lists.freedesktop.org>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
The code sizes return here get passed to the cache shader insert function,
which then memcpy from the code ptr, and causes all sorts of valgrind
errors like:
==6755== Invalid read of size 8
==6755== at 0x4C32FEE: memcpy@GLIBC_2.2.5 (vg_replace_strmem.c:1021)
==6755== by 0x2305D4C7: radv_pipeline_cache_insert_shaders (radv_pipeline_cache.c:416)
==6755== by 0x2305791D: radv_create_shaders (radv_pipeline.c:2158)
==6755== by 0x2305C523: radv_pipeline_init (radv_pipeline.c:3404)
==6755== by 0x2305C890: radv_graphics_pipeline_create (radv_pipeline.c:3515)
==6755== by 0x230188AB: radv_device_init_meta_blit_color (radv_meta_blit.c:871)
==6755== by 0x2301D50E: radv_device_init_meta_blit_state (radv_meta_blit.c:1278)
==6755== by 0x23011893: radv_device_init_meta (radv_meta.c:352)
==6755== by 0x2300744B: radv_CreateDevice (radv_device.c:1576)
==6755== by 0x5187D0F: ??? (in /usr/lib64/libvulkan.so.1.1.77)
==6755== by 0x518F6A3: ??? (in /usr/lib64/libvulkan.so.1.1.77)
==6755== by 0x5192A42: vkCreateDevice (in /usr/lib64/libvulkan.so.1.1.77)
==6755== Address 0x22a58548 is 4 bytes after a block of size 116 alloc'd
==6755== at 0x4C2EBAB: malloc (vg_replace_malloc.c:299)
==6755== by 0x23089DC4: ac_elf_read (ac_binary.c:144)
==6755== by 0x23090A60: ac_compile_module_to_binary (ac_llvm_helper.cpp:162)
==6755== by 0x23053F06: compile_to_memory_buffer (radv_llvm_helper.cpp:58)
==6755== by 0x23053F06: radv_compile_to_binary (radv_llvm_helper.cpp:98)
==6755== by 0x23052769: ac_llvm_compile (radv_nir_to_llvm.c:3394)
==6755== by 0x23052823: ac_compile_llvm_module (radv_nir_to_llvm.c:3418)
==6755== by 0x23053C05: radv_compile_nir_shader (radv_nir_to_llvm.c:3542)
==6755== by 0x23061B4E: shader_variant_create (radv_shader.c:580)
==6755== by 0x23061CFD: radv_shader_variant_create (radv_shader.c:634)
==6755== by 0x23057765: radv_create_shaders (radv_pipeline.c:2123)
==6755== by 0x2305C523: radv_pipeline_init (radv_pipeline.c:3404)
==6755== by 0x2305C890: radv_graphics_pipeline_create (radv_pipeline.c:3515)
Since we are just inserting the code into the cache, we can avoid these
bad reads and data in the cache by just using the binary code size here.
Fixes: 939e5a382 (radv: add padding for the UMR disassembler)
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
This uses the common compiler passes abstraction to help radv
avoid fixed cost compiler overheads. This uses a linked list per
thread stored in thread local storage, with an entry in the list
for each target machine.
This should remove all the fixed overheads setup costs of creating
the pass manager each time.
This takes a demo app time to compile the radv meta shaders on nocache
and exit from 1.7s to 1s. It also has been reported to take the startup
time of uncached shaders on RoTR from 12m24s to 11m35s (Alex)
v2: fix llvm6 build, inline emit function, handle multiple targets
in one thread
v3: rebase and port onto new structure
v4: rename some vars (Bas)
v5: drag all code into radv for now, we can refactor it out later
for radeonsi if we make it shareable
v6: use a bit more C++ in the wrapper
v7: logic bugs fixed so it actually runs again.
v8: rebase on top of radeonsi changes.
v9: drop some C++ headers, cleanup list entry
v10: use pop_back (didn't have enough caffeine)
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
This ports radv to the shared code, however due to a bug in LLVM
version prior to 7, radv cannot add target info at this stage,
as it would leak one for every shader compile, however I'd prefer
to keep this llvm damage in the shared code, since it isn't the
driver at fault here. We just add a flag to denote if the driver
can support leaking the target info or not, and the common code
does the right thing depending on the llvm version.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
This is prep work for moving this to a per-thread struct
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
This just splits out the non-shared code and reuses ac_get_llvm_target in radv.
v2: rebase on Marek's patch - fixup brace position/whitespace
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
The driver already supports exporting the stencil value.
The following CTS test now pass:
dEQP-VK.pipeline.shader_stencil_export.op_replace
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Rob Clark <robdclark@gmail.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This commit completely reworks function calls in NIR. Instead of having
a set of variables for the parameters and return value, nir_call_instr
now has simply has a number of sources which get mapped to load_param
intrinsics inside the functions. It's up to the client API to build an
ABI on top of that. In SPIR-V, out parameters are handled by passing
the result of a deref through as an SSA value and storing to it.
This virtue of this approach can be seen by how much it allows us to
delete from core NIR. In particular, nir_inline_functions gets halved
and goes from a fairly difficult pass to understand in detail to almost
trivial. It also simplifies spirv_to_nir somewhat because NIR functions
never were a good fit for SPIR-V.
Unfortunately, there is no good way to do this without a mega-commit.
Core NIR and SPIR-V have to be changed at the same time. This also
requires changes to anv and radv because nir_inline_functions couldn't
handle deref instructions before this change and can't work without them
after this change.
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Before, we were doing structure splitting in spirv_to_nir.
Unfortunately, this doesn't really work when you think about passing
struct pointers into functions. Doing it later in NIR is a much better
plan.
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This inserts a call to nir_lower_deref_instrs at every call site of
glsl_to_nir, spirv_to_nir, and prog_to_nir.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
It's a bit late to round up after an integer division.
Fixes: de88979413 "radv: Implement VK_AMD_shader_info"
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Alex Smith <asmith@feralinteractive.com>
This allows to run the LLVM verifier pass.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Workaround for bug in llvm that causes the GPU to hang in presence
of nested loops because there is an exec mask issue. The proper
solution is to fix LLVM but this might require a bunch of work.
This fixes a bunch of GPU hangs that happen with DXVK.
Vega10:
Totals from affected shaders:
SGPRS: 110456 -> 110456 (0.00 %)
VGPRS: 122800 -> 122800 (0.00 %)
Spilled SGPRs: 7478 -> 7478 (0.00 %)
Spilled VGPRs: 36 -> 36 (0.00 %)
Code Size: 9901104 -> 9922928 (0.22 %) bytes
Max Waves: 7143 -> 7143 (0.00 %)
Code size slightly increases because it inserts more branch
instructions but that's expected. I don't see any real performance
changes.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=105613
Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Errors are not that common of a case so we can eat a slight perf
hit in having to call a function and do a runtime check.
In turn this makes debugging random errors happening for end users
easier, because they don't have to have a debug build on hand.
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
This doesn't nothing special currently because we don't create
any copy_var instructions, but this is needed for the next patch.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Having an entrypoint different than "main" doesn't mean we
have multiple shaders per module.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
When VK_PIPELINE_CREATE_DISABLE_OPTIMIZATION_BIT is set we skip NIR
linking optimisations and only run over the NIR optimisation loop
once similar to the GLSLOptimizeConservatively constant used by
some GL drivers.
We need to run over the opts at least once to avoid errors in LLVM
(e.g. dead vars it can't handle) and also to reduce the time spent
compiling the IR in LLVM.
With this change the Blacksmith Unity demos compilation times
go from 329760 ms -> 299881 ms when using Wine and DXVK.
V2: add bit to radv_pipeline_key
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=106246
The driver already supports exporting the Layer and ViewportIndex
built-ins from vertex or tessellation shaders.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Noticed while passing by. Not sure if it impacts anything, but
likely to impact GFX9 more than anything else since we lower
inputs, outputs and locals there.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
If a shader only writes to an output via a constant initializer we
need to lower it before we call nir_remove_dead_variables so that
this pass sees the stores from the initializer and doesn't kill the
output.
Fixes test failures in new work-in-progress CTS tests:
dEQP-VK.spirv_assembly.instruction.graphics.variable_init.output.float
This is ported from anv:
99b57daf4a anv/pipeline: lower constant initializers on output variables earlier
from Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
If AMD_shader_info or RADV_TRACE_FILE is used we might need to
keep trace of LLVM IR.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
So now, during spirv_to_nir, it uses the capability instead of the
extension. Note that we are really doing here is treating
SPV_AMD_gcn_shader as other supported extensions. SPV_AMD_gcn_shader
is not the first SPV extension supported. For example, the capability
draw_parameters infers if the extension SPV_KHR_shader_draw_parameters
is supported or not.
This could be seen as counter-intuitive, and that it would be easier
to define which extensions are supported, and based our checks on
that, but we need to take into account that some capabilities are
optional from core, and others came from new extensions.
Also this commit would make the implementation of ARB_spirv_extensions
easier.
v2: AMD_gcn_shader capability renamed to gcn_shader (Daniel Schürmann)
Reviewed-by: Daniel Schürmann <daniel.schuermann@campus.tu-berlin.de>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Now the "ac/nir" prefix will really be the shared code between
RadeonSI and RADV, that might avoid confusions in the future.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Until llvm handles indirects better we will need to use these
workarounds in the radeonsi backend also.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Otherwise new local variables can cause hangs on vega.
CC: <mesa-stable@lists.freedesktop.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=105098
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
LLVM can't shrink loads.
Polaris10:
Totals from affected shaders:
SGPRS: 62528 -> 59955 (-4.11 %)
VGPRS: 44708 -> 44616 (-0.21 %)
Spilled SGPRs: 16 -> 8 (-50.00 %)
Code Size: 1355504 -> 1355172 (-0.02 %) bytes
Max Waves: 11710 -> 11670 (-0.34 %)
Vega10:
Totals from affected shaders:
SGPRS: 51448 -> 50371 (-2.09 %)
VGPRS: 39140 -> 39048 (-0.24 %)
Spilled SGPRs: 16 -> 16 (0.00 %)
Code Size: 1307188 -> 1304296 (-0.22 %) bytes
Max Waves: 11312 -> 11292 (-0.18 %)
This reduces SGPRs spilling in MadMax, and it also reduces
number of SGPRs in DOW3 and F12017. The number of waves slightly
decreases in F1 but I don't see any performance changes after
benchmarking it. Talos and Serious Sam are not affected because
they don't use any push constants.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Allows nir drivers to either use a single or dual locations for
vs double inputs.
i965 uses dual locations for both OpenGL and Vulkan drivers, for
now gallium OpenGL drivers only use a single location.
The following patch will also make use of this option when
calling nir_shader_gather_info().
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Fixes a bunch of arb_gpu_shader_fp64 piglit tests for example:
generated_tests/spec/arb_gpu_shader_fp64/execution/built-in-functions/fs-mix-double-double-double.shader_test
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
VGPR1 is only needed for topology that needs 3 offsets like
triangles or quads.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
This can still be improved, but let's start with this.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
This reverts commit 2294d35b24.
We can't do this without adjusting the input SGPRs/VGPRs logic.
For now, just revert it. I will send a proper solution later.
It fixes a rendering issue in F1 2017 that CTS didn't catch up.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Tested-by: Alex Smith <asmith@feralinteractive.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
We should also not load the input SGPRs and VGPRS, but
let's start with this for now.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
This is a bit more general and lets us pass additional options into the
spirv_to_nir pass beyond what capabilities we support.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
It's really annoying and this pollutes the output especially
when a bunch of non-meta shaders are compiled.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
For consistency and it might help for debugging purposes.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
There's no point recalculating these the whole time on descriptor
emission, just store them at pipeline creation.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103513
Fixes: de88979413 ("radv: Implement VK_AMD_shader_info")
Signed-off-by: Alex Smith <asmith@feralinteractive.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
This allows an app to query shader statistics and get a disassembly of
a shader. RenderDoc git has support for it, so this allows you to view
shader disassembly from a capture.
When this extension is enabled on a device (or when tracing), we now
disable pipeline caching, since we don't get the shader debug info when
we retrieve cached shaders.
v2: Improvements to resource usage reporting
v3: Disassembly string must be null terminated (string_buffer's length
does not include the terminator)
v4: Fixed LDS reporting. (Bas)
Signed-off-by: Alex Smith <asmith@feralinteractive.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
This uses the new kernel interfaces for reduced cs overhead,
We only set the local flag for memory allocations that don't have
a dedicated allocation and ones that aren't imports.
v2: add to all the internal buffer creation paths.
v3: missed some command submission paths, handle 0/empty bo lists.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Dave Airlie <airlied@redhat.com>
We now have linking optimisations so we want to delay dumping the
nir until after these are complete.
Fixes: 06f05040eb (radv: Link shaders)
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Since it also uses the output vector before writing to memory.
Fixes: e38685cc62 'Revert "radv: disable support for VEGA for now."'
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Due to LLVM bugs. Fixes a bunch of dEQP-VK.glsl.indexing.*
tests.
Fixes: e38685cc62 'Revert "radv: disable support for VEGA for now."'
Reviewed-by: Dave Airlie <airlied@redhat.com>
It looks the original indirect mask was probably copied from
ANV.
Sascha Willems demo results:
tessellation ~4000 -> ~4200 fps
V2: continue lowering local indirects due to llvm deficiencies.
Tested-by: Alex Smith <asmith@feralinteractive.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Here we make use of NIR the linking helpers to remove unused
varyings.
Sascha Willems demo results:
computecullandlod 39 -> 41 fps
pipelines ~6100 -> ~6200 fps
Signed-off-by: Bas Nieuwenhuizen <basni@google.com>
Signed-off-by: Timothy Arceri <tarceri@itsqueeze.com>
Acked-by: Dave Airlie <airlied@redhat.com>
So it appears the Vulkan SPIR-V fma opcode can be equivalent to a
mad operation, and the fma hw opcode on AMD hw is issued like a double
opcode so is slower. Also the radeonsi stack does this.
This appears to improve performance on a number of games from Feral,
and thanks to Feral for noticing the problem.
I'm reposting this one as Marek indicated he thinks this is what
we should be doing on AMD hw.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Cc: "17.2" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This looks a bit ugly to me, but the existing codepath
is not terribly elegant as well.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
To dump the shader stats when a hang is detected.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
With the shaders in the ssao demo, the nir_opt_if wasn't
working properly without this, after this the if gets optimised
so that loop unrolling gets called.
(loop unrolling fails due to instruction count, but at least
it gets to do that.)
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Cc: "17.2" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This will allow to dump the active shaders when a hang is
detected. Only the ASM will be dumped for now.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
The device object contains the debug flags.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reduce size of radv_pipeline.c and improve code isolation. More
code can probably moved but it's a start.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>