Lowering PS inputs can eliminate some of them, which messes up
persp/linear barycentric coord usage info.
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Legacy GS has to use Wave64, so TES before GS has to use Wave64 too.
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
because legacy VS hangs.
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
We have to fix the computation so as not to break quads.
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
It only works if there are not color and no Z exports.
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Acked-by: Dave Airlie <airlied@redhat.com>
- don't pass it via a parameter if it can be derived from other parameters
- set shader_type for ac_rtld_open
- use enum pipe_shader_type instead of unsigned
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Acked-by: Dave Airlie <airlied@redhat.com>
We need to tell PA to accept edge flags generated by the input assembler,
because decomposed primitives shouldn't draw inner edges.
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
With NGG, the VGT_GS_OUT_PRIM_TYPE can change without a shader change.
The VS_STATE is required for both streamout and culling from a vertex
shader without pre-compiling outprim-specific variants.
We could consider compiling specialized variants in the future. We
could also consider compiling the NGG logic as an epilog.
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
For pipelines without API GS. We will later expand this to cover NGG
geometry shaders as well.
Note that the vtx offset passed into the GS part is just the
vertex index multiplied by VGT_ESGS_RING_ITEMSIZE.
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Also add the shader main part NGG variant, so that in principle
we can switch between legacy in NGG modes.
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
We always use NGG by default, except when tessellation is enabled with
extreme geometry shader amplification.
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
We'll have to extend this at some point, and using a bitfield union in
this way makes it easier to get the right index without excessive
branching.
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
This will make it easier to use LDS for other purposes in geometry
shaders in the future.
The lifetime of the esgs_ring variable is as follows:
- declared as [0 x i32] while compiling shader parts or monolithic shaders
- just before uploading, gfx9_get_gs_info computes (among other things)
the final ESGS ring size (this depends on both the ES and the GS shader)
- during upload, the "esgs_ring" symbol is given to ac_rtld as a shared
LDS symbol, which will lead to correctly laying out the LDS including
other LDS objects that may be defined in the future
- si_shader_gs uses shader->config.lds_size as the LDS size
This change depends on the LLVM changes for emitting LDS symbols into
the ELF file.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Upcoming changes to LLVM will emit LDS objects as symbols in the ELF
symbol table, with relocations that will be resolved with this change.
Callers will also be able to define LDS symbols that are shared between
shader parts. This will be used by radeonsi for the ESGS ring in gfx9+
merged shaders.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
If a prim discard compute shader hasn't finished compilation, we don't want
to any shader.
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
Acked-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
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.
The overall goal is to support unaligned loads from vertex buffers
natively on SI.
In the unaligned case, we fall back to the general case implementation in
ac_build_opencoded_load_format. Since this function is fully general,
we will also use it going forward for cases requiring fully manual format
conversions of dwords anyway.
This requires a different encoding of the fix_fetch array, which will now
contain the entire format information if a fixup is required.
Having to check the alignment of vertex buffers is awkward. To keep the
impact on the fast path minimal, the si_context will keep track of which
vertex buffers are (not) at least dword-aligned, while the
si_vertex_elements will note which vertex buffers have some (at most dword)
alignment requirement. Vertex buffers should be dword-aligned most of the
time, which allows a fast early-out in almost all cases.
Add the radeonsi_vs_fetch_always_opencode configuration variable for
testing purposes. Note that it can only be used reliably on LLVM >= 9,
because support for byte and short load is required.
v2:
- add a missing check to si_bind_vertex_elements
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Some of the opts are not called in the general optimastion loop
in the state trackers glsl -> nir conversion. We need to call
the radeonsi specific optimisation once before scanning over
the nir otherwise we can end up gathering info on code that
is later removed.
Fixes an assert in the piglit test:
./bin/varying-struct-centroid_gles3
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Needed to track context rolls caused by streamout and ACQUIRE_MEM.
ACQUIRE_MEM can occur outside of draw calls.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110355
v2: squashed patches and done more rework
Cc: 19.0 <mesa-stable@lists.freedesktop.org>
Force the driver thread to sync immediately with a compiler thread (but
compilation still happens in a separate thread).
This can be useful to simplify debugging compiler issues.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
This helps improve compile times. For example the shader-db dolphin
shader shaders/dolphin/ubershaders/120.shader_test goes from
~1.69 -> ~1.57 seconds on my machine with this change.
Reviewed-by: Marek Olšák <marek.olsak@amd.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>
We should get fewer context rolls with the SET_CONTEXT_REG optimization,
but it would have been for nothing if the scissor state rolled the context
anyway. Don't emit the scissor state if there is no context roll.
40% is the decrease in the LGKM counter (which includes SMEM too)
for the GFX9 LSHS stage.
This will make the LDS size slightly larger, but I wasn't able to increase
the patch stride without corruption, so I'm increasing the vertex stride.
As precursor to moving init to common code, just rename the struct
and move it.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
GS is tested, tessellation is untested.
Have outputs_written_before_ps for HW VS and outputs_written for other
stages. The reason is that COLOR and BCOLOR alias for HW VS, which
drives elimination of VS outputs based on PS inputs.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
It will contain more variables.
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Tested-by: Benedikt Schemmer <ben at besd.de>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
so that it can be removed and replaced with inline VBO descriptors,
and the pointer can be packed in unused bits of VBO descriptors.
This also removes the pointer from merged TES-GS where it's useless.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
TCS_OUT_LAYOUT has 13 unused bits. That's enough for a 32-bit address
aligned to 512KB. Hey, it's a 13-bit pointer!
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
If alpha-to-coverage is enabled, we have to compute alpha
even if color writes are disabled.
Signed-off-by: Józef Kucia <joseph.kucia@gmail.com>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
Most files in gallium/radeon now include si_pipe.h.
chip_class and family are now here:
sscreen->info.family
sscreen->info.chip_class
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
With Gallium threaded contexts, creating shader/compute states is
effectively a screen operation, so we should not use context state.
In particular, this allows us to avoid using the context's LLVM
TargetMachine.
This isn't an issue yet because u_threaded_context filters out non-async
debug callbacks, and we disable threaded contexts for debug contexts.
However, we may want to change that in the future.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
We only need the lock to guard changes in the variant linked list. The
actual compilation can happen outside the lock, since we use the ready
fence as a guard.
v2: fix double-unlock
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
There's a race condition between si_shader_select_with_key and
si_bind_XX_shader:
Thread 1 Thread 2
-------- --------
si_shader_select_with_key
begin compiling the first
variant
(guarded by sel->mutex)
si_bind_XX_shader
select first_variant by default
as state->current
si_shader_select_with_key
match state->current and early-out
Since thread 2 never takes sel->mutex, it may go on rendering without a
PM4 for that shader, for example.
The solution taken by this patch is to broaden the scope of
shader->optimized_ready to a fence shader->ready that applies to
all shaders. This does not hurt the fast path (if anything it makes
it faster, because we don't explicitly check is_optimized).
It will also allow reducing the scope of sel->mutex locks, but this is
deferred to a later commit for better bisectability.
Fixes dEQP-EGL.functional.sharing.gles2.multithread.simple.buffers.bufferdata_render
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
It's inaccurate. Instead, see the copyright and use "git log" and
"git blame" to know the authorship.
Acked-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
We'll use it in the scissors / clip / guardband state.
v2: avoid a performance regression on r600 when applied to
(pre-fork) stable branches
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Not that those are super common or useful, but hey! Fun corner cases
of the API...
Fixes dEQP-GLES31.functional.geometry_shading.emit.*
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
This marks the end of code sharing between r600 and radeonsi.
It's getting difficult to work on radeonsi without breaking r600.
A lot of functions had to be renamed to prevent linker conflicts.
There are also minor cleanups.
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
This does not take commutative blending into account yet.
R600_DEBUG=nooutoforder disables it.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
The GLSL rules for interpolateAtSample are unfortunate:
"Returns the value of the input interpolant variable at
the location of sample number sample. If
multisample buffers are not available, the input
variable will be evaluated at the center of the pixel.
If sample sample does not exist, the position used to
interpolate the input variable is undefined."
This fix will fallback to monolithic shader compilation when
interpolateAtSample is used without multisampling.
One alternative would be to always upload 16 sample positions,
filling the buffer up with repetition when the actual number of
samples is less, and then ANDing the sample ID with 0xf. However,
that punishes all well-behaving users of interpolateAtSample,
when in reality, only conformance tests should be affected by
the issue.
Fixes
dEQP-GLES31.functional.shaders.multisample_interpolation.interpolate_at_sample.non_multisample_buffer.*
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
gl_SampleMaskIn is supposed to contain set bits only for the samples that
are covered by the current fragment shader invocation, but the VGPR
initialization hardware loads the set of all bits that are covered at the
current pixel.
Fixes various tests in
dEQP-GLES31.functional.shaders.sample_variables.sample_mask_in.*
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
This removes the barrier and LDS stores and loads for tess factors
when it's possible. The removal of the barrier seems more important
to me though.
In one shader, it removes 17 * 4 bytes from the shader binary.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
When the HS wave is empty, the hardware writes the LS VGPRs starting at
v0 instead of v2. Workaround by shifting them back into place when
necessary. For simplicity, this is always done in the LS prolog.
According to the hardware team, this will be fixed in future chips,
so take that into account already.
Note that this is not a bug fix, as the bug was already worked
around by commit 166823bfd2 ("radeonsi/gfx9: add a temporary workaround
for a tessellation driver bug"). This change merely replaces the
workaround by one that should be better.
v2: add workaround code to shader only when necessary
v3: clarify the prefer_mono comment
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
This increases performance, but it was tuned for Raven, not Vega.
We don't know yet how Vega will perform, hopefully not worse.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
- don't precompile LS and ES (they don't exist on GFX9), compile as VS instead
- don't precompile HS and GS (we don't have LS and ES parts)
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
so that we don't rely on si_pm4_state_enabled_and_changed, allowing us
to move prefetches after draw calls.
v2: ckear the dirty mask after unbinding shaders
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de> (v1)
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com> (v1)
I'd like to be able to move the prefetch call site around.
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
si_build_shader_variant can actually be called directly from one of
normal-priority compiler threads. In that case, the thread_index is
only valid for the normal tm array.
v2:
- use the correct sel/shader->compiler_ctx_state
Fixes: 86cc809726 ("radeonsi: use a compiler queue with a low priority for optimized shaders")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
This adds some new helper functions to know if the current draw
call (or dispatch compute) is using bindless samplers/images,
based on TGSI analysis.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
This and the previous clip_regs commit decrease IB sizes and the number of
si_update_shaders invocations as follows:
IB size si_update_shaders calls
Borderlands 2 -10% -27%
Deus Ex: MD -5% -11%
Talos Principle -8% -30%
v2: always dirty cb_render_state in set_framebuffer_state
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
We can use a union in si_shader_key::mono.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Heaven LDS usage for LS+HS is below. The masks are "outputs_written"
for LS and HS. Note that 32K is the maximum size.
Before:
heaven_x64: ls=1f1 tcs=1f1, lds=32K
heaven_x64: ls=31 tcs=31, lds=24K
heaven_x64: ls=71 tcs=71, lds=28K
After:
heaven_x64: ls=3f tcs=3f, lds=24K
heaven_x64: ls=7 tcs=7, lds=13K
heaven_x64: ls=f tcs=f, lds=17K
All other apps have a similar decrease in LDS usage, because
the "outputs_written" masks are similar. Also, most apps don't write
POSITION in these shader stages, so there is room for improvement.
(tight per-component input/output packing might help even more)
It's unknown whether this improves performance.
Tested-by: Edmondo Tommasina <edmondo.tommasina@gmail.com>
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
LS is merged into TCS. If there is no TCS, LS is merged into fixed-func
TCS. The problem is the fixed-func TCS was ignored by scratch update
functions, so LS didn't have the scratch buffer set up.
Note that Mesa 17.1 doesn't have merged shaders.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
If we enqueue too many jobs and destroy the GL context, it may take
several seconds before the jobs finish. Just drop them instead.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
What we care about is whether PrimID is used while tessellation is
enabled; whether it's used in TCS/TES or further down the pipeline is
irrelevant.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
This builds on commit 0549ea15ec ("radeonsi: fix primitive ID in
fragment shader when using tessellation").
Fixes piglit
arb_tessellation_shader/execution/gs-primitiveid-instanced.shader_test
Cc: 17.1 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
By keeping track of fewer generics, everything can fit into 64 bits.
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
OpenGL uses at most 32 generic outputs/inputs in any stage, and they always
have a shader IO index and therefore fit into the outputs_written/
inputs_read/kill_outputs fields.
However, Nine uses semantic indices more liberally. We support that
in VS-PS pipelines, except that the optimization of killing outputs
must be skipped.
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Make it a bit clearer that the index spaces are logically seperate by
having them defined in different functions.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
In a VS->TCS->TES->PS pipeline, the primitive ID is read from TES exports,
so it is as if TES were using the primitive ID.
Specifically, this fixes a bug where the primitive ID is not reset at
the start of a new instance.
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Picked from a different branch. When we stop using the scratch patching,
this function will not be called.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
This removes s_load_dword latency for tess rings.
We need just 1 SGPR for the address if we use 64K alignment. The final asm
for recreating the descriptor is:
// s2 is (address >> 16)
s_mov_b32 s3, 0
s_lshl_b64 s[4:5], s[2:3], 16
s_mov_b32 s6, -1
s_mov_b32 s7, 0x27fac
v2: bitcast the descriptor type from v2i64 to v4i32
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
The use of PrimID in the pixel shader is too rare to deserve such
a sizable support code.
The initial idea of the VS epilog was to move the clipping code there and
remove it based on states, but optimized variants are now used to do that
and are easier to support, so the VS epilog has turned out to be not so
useful.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
The 2nd shader of merged shaders should take a reference of the 1st shader.
The next commit will do that.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
This code can be shared by radv, we bump the max to
VARYING_SLOT_MAX here, but that shouldn't have too
much fallout.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This will allow to hash additional data into the cache keys or even
change the hashing algorithm easily, should we decide to do so.
v2: don't try to compute key (and crash) if cache is disabled
Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
threaded gallium can't use pipe_context's LLVM target machine, because
create_shader_selector can be called from a non-driver thread.
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
pipe_mutex_unlock() was made unnecessary with fd33a6bcd7.
Replaced using:
find ./src -type f -exec sed -i -- \
's:pipe_mutex_unlock(\([^)]*\)):mtx_unlock(\&\1):g' {} \;
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
replace pipe_mutex_lock() was made unnecessary with fd33a6bcd7.
Replaced using:
find ./src -type f -exec sed -i -- \
's:pipe_mutex_lock(\([^)]*\)):mtx_lock(\&\1):g' {} \;
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
pipe_mutex_destroy() was made unnecessary with fd33a6bcd7.
Replace was done with:
find ./src -type f -exec sed -i -- \
's:pipe_mutex_destroy(\([^)]*\)):mtx_destroy(\&\1):g' {} \;
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
pipe_mutex_init() was made unnecessary with fd33a6bcd7.
Replace was done using:
find ./src -type f -exec sed -i -- \
's:pipe_mutex_init(\([^)]*\)):(void) mtx_init(\&\1, mtx_plain):g' {} \;
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
V2:
- when loading from disk cache also binary insert into memory cache.
- check that the binary loaded from disk is the correct size. If not
delete the cache item and skip loading from cache.
V3:
- remove unrequired variable
Reviewed-by: Grigori Goronzy <greg@chown.ath.cx>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
R600_DEBUG=mono has had no effect since:
commit 1fabb29717
Author: Marek Olšák <marek.olsak@amd.com>
Date: Tue Feb 14 22:08:32 2017 +0100
radeonsi: have separate LS and ES main shader parts in the shader selector
Also, this assertion was failing:
si_state_shaders.c:1307: si_shader_select_with_key: Assertion
`!shader->is_optimized' failed.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
We were unconditionally storing these outputs, sometimes even one component
at a time, but apps never read them in TES.
Move the TESSINNER/OUTER buffer stores into the TCS epilog where we can
easily disable them on demand.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Marek Olšák
Nicolai Hähnle
Timothy Arceri
Samuel Pitoiset
Sonny Jiang
Dave Airlie
Józef Kucia
Grazvydas Ignotas