This should help avoid stalls in the pixel mask array in certain
non-promoted depth cases. It especially helps for Z16, as each bit
in the PMA corresponds to two pixels when using Z16, as opposed to
the usual one pixel.
Improves performance in GFXBench5 TRex by 22% (n=1).
From the MEDIA_VFE_STATE docs:
"Starting with this configuration, the Maximum Number of Threads must
be set to (#EU * 8) for GPGPU dispatches.
Although there are only 7 threads per EU in the configuration, the
FFTID is calculated as if there are 8 threads per EU, which in turn
requires a larger amount of Scratch Space to be allocated by the
driver."
It's pretty clear that we need to increase this for scratch address
calculations, because the FFTID has a certain bit-pattern. The quote
above seems to indicate that we should increase the actual thread count
programmed in MEDIA_VFE_STATE as well, but we think the intention is to
only bump the scratch space.
Fixes GPU hangs in Bioshock Infinite and Synmark's CSDof on Icelake 8x8.
Fixes: 5ac804bd9a ("intel: Add a preliminary device for Ice Lake")
Reviewed-by: Matt Turner <mattst88@gmail.com>
If we can entirely push uniform data, we don't need a SURFACE_STATE
descriptor for pulling data. Since constant uploads are a very common
operation, and being able to push all data is also very common, we would
like to avoid the overhead in this case.
This patch defers uploading new descriptors. Instead of handling that
at iris_set_constant_buffer, we do it at iris_update_compiled_shaders,
where we can see the currently bound shader variants. If any need pull
descriptors, and descriptors are missing, we update them and flag that
the binding table also needs to be refreshed.
Improves performance in GFXBench5 gl_driver2 on an i7-6770HQ by
31.9774% +/- 1.12947% (n=15).
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
The compiler now sets the "Null Render Target" bit in the RT write
extended message descriptor, causing it to write to an implicit null
surface without us needing to set one up in the binding table.
Together with the last patch, this improves performance in Car Chase on
an Icelake 8x8 (locked to 700Mhz) by 0.0445526% +/- 0.0132736% (n=832).
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Otherwise it's impossible to know the maximum SSBO index for both
internal TGSI shaders from TTN (which don't have any notion of atomic
counters and no offset) as well as shaders from GLSL.
I fixed everything I could find while grepping for num_ssbos and
num_abos, which hopefully is everything (iris was the only user I could
find that uses it in a meaningful way).
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
It's not used by anything anymore now that so much lowering has been
moved into NIR. Sadly, we still need on in brw_compile_gs() for
geometry shaders on Sandy Bridge. Short of a lot of pointless work,
that one's probably not going away.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This better matches all the other atomic intrinsics such as those for
SSBOs and shared variables where the sign is part of the intrinsic
opcode. Both generators (GLSL and SPIR-V) know the sign from the type
of the image variable or handle. In SPIR-V, signed min/max are separate
opcodes from unsigned.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
This will be used in next patches for supporting non coherent
framebuffer fetch on Broadwell.
v2: Fix comment (Kenneth Graunke)
v3: 1) Fix a few nits (Caio)
2) Add comment (Caio)
Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This commit is all annoying plumbing work which just adds support for a
new brw_compile_stats struct. This struct provides a binary driver
readable form of the same statistics we dump out to stderr when we
INTEL_DEBUG is set with a shader stage.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
v5: add patch
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Iris advertises support for PIPE_CAP_TGSI_VS_WINDOW_SPACE_POSITION
so let's actually implement it.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110657
Signed-off-by: Danylo Piliaiev <danylo.piliaiev@globallogic.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
The rules for gl_SubgroupSize in Vulkan require that it be a constant
that can be queried through the API. However, all GL requires is that
it's a uniform. Instead of always claiming that the subgroup size in
the shader is 32 in GL like we have to do for Vulkan, claim 8 for
geometry stages, the maximum for fragment shaders, and the actual size
for compute.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
We were failing to flag the program dirty when it changed. Also, we
were unnecessarily setting key->input_vertices for SINGLE_PATCH mode,
which would reduce program cache hits. Only set it if needed.
Right now, all keys have two things in common: a program string ID and a
sampler_prog_key_data. I'd like to add another thing or two and need a
place to put it. This commit adds a new brw_base_prog_key struct which
contains those two common bits.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This commit moves the sysvals to a separate, new constant buffer
at the end (before the shader constants). It also allows us to
remove the special handling we had for cbuf0, and enables all
constant buffers to support user-specified resources and user
buffers.
v2: (by Kenneth Graunke)
- Rebase on the previous patch to fix system value uploading.
- Fix disk cache num_cbufs calculation
- Fix passthrough TCS to report num_cbufs = 1 so upload actually occurs
- Change upload_sysvals to assert that num_cbufs > 0 when
num_system_values > 0.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
I neglected to mark cbuf0_needs_upload = false after uploading it.
The obvious fix regressed user clip plane tests, because of a second
bug: we also forgot to mark that they may need re-uploading when
changing shader programs (which may have more or less system values).
Thanks to Timur Kristóf for catching the original issue.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Fixes valgrind errors when running two CTS tests back to back:
- KHR-GL45.shader_image_load_store.basic-allTargets-loadStoreT*
(The first test has an actual TCS, the second uses passthrough.)
We run a ton of backend specific passes here (mostly brw_preprocess_nir)
and ought to sweep up any unused memory at this point, since we're going
to hang on to this NIR for as long as the linked program lives.
Now that NIR_TEST_* doesn't swap the shader out from under us, it's
sufficient to just modify the shader rather than having to return in
case we're testing serialization or cloning.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Mesa measures in DWords. The hardware also claims to measure in DWords.
Except the SO_WRITE_OFFSET field is actually bits 31:2, with 1:0 MBZ.
Which means that it really measures in bytes. So, convert to bytes.
Without this, our offset / stride denominator was 1/4th the size it
should be, leading to 4x the vertex count that we should have had.
Fixes GTF-GL46.gtf40.GL3Tests.transform_feedback2.transform_feedback2_two_buffers
Don't have a separate mechanism for NIR constants to be removed from
the table. If unused, we will compact it away. The use_null_surface
is needed when INTEL_DISABLE_COMPACT_BINDING_TABLE is set.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Change the iris_binding_table to keep track of what surfaces are
actually going to be used, then assign binding table indices just for
those. Reducing unused bytes on those are valuable because we use a
reduced space for those tables in Iris.
The rest of the driver can go from "group indices" (i.e. UBO #2) to
BTI and vice-versa using helper functions. The value
IRIS_SURFACE_NOT_USED is returned to indicate a certain group index is
not used or a certain BTI is not valid.
The environment variable INTEL_DISABLE_COMPACT_BINDING_TABLE can be
set to skip compacting binding table.
v2: (all from Ken)
Use BITFIELD64_MASK helper. Improve comments.
Assert all group is marked as used when we have indirects.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Stop using brw_compiler to lower the final binding table indices for
surface access. This is done by simply not setting the
'prog_data->binding_table.*_start' fields. Then make the driver
perform this lowering.
This is a better place to perfom the binding table assignments, since
the driver has more information and will also later consume those
assignments to upload resources.
This also prepares us for two changes: use ibc without having to
implement binding table logic there; and remove unused entries from
the binding table.
Since the `block` field in brw_ubo_range now refers to the final
binding table index, we need to adjust it before using to index
shs->constbuf.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
We'll change iris to perform lowering of the binding table indices
earlier (before the backend kick in), but the backend compiler uses
the result of the analysis to identify load_ubo intrinsics, so we do
the analysis after the lowering to have the right indices.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Shader-db results on Kaby Lake:
total instructions in shared programs: 15306230 -> 15304726 (<.01%)
instructions in affected programs: 4570 -> 3066 (-32.91%)
helped: 16
HURT: 0
total cycles in shared programs: 361703436 -> 361680041 (<.01%)
cycles in affected programs: 129388 -> 105993 (-18.08%)
helped: 16
HURT: 0
LOST: 0
GAINED: 2
The helped programs were in XCom 2, Deus Ex: Mankind Divided, and Kerbal
Space Program
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
It will be true for the constant/system value buffer because they use a
constant zero but it's not true in general. If we ever got here when
the source wasn't constant, nir_src_as_uint would assert.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: mesa-stable@lists.freedesktop.org
Jason pointed out that we don't need to keep an entire copy of the
serialized NIR around, we just need the SHA1. This does change our
disk cache key to be taking a SHA1 of a SHA1, which is a bit odd,
but should work out and be faster and use less memory.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
We were checking this based on nir->info.name, but with the shader
cache enabled, nir_strip throws out the name, causing us to use IEEE
mode for ARB programs.
gl-1.0-spot-light regressed because it wants ALT mode for 0^0 behavior.
Fixes: dc5dc727d5 iris: Serialize the NIR to a blob we can use for shader cache purposes.
We will use a hash of the serialized NIR together with brw_prog_*_key
(for NOS) as the disk cache key, where the disk cache contains actual
assembly shaders.
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
It had been internal to iris_program.c, but with the upcoming disk cache
code, the "program module" is going to be spread across a couple source
files. Into a header it goes!
Now it lives alongside iris_compiled_shader, which makes sense.
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
Our tessellation control shaders can be dispatched in several modes.
- SINGLE_PATCH (Gen7+) processes a single patch per thread, with each
channel corresponding to a different patch vertex. PATCHLIST_N will
launch (N / 8) threads. If N is less than 8, some channels will be
disabled, leaving some untapped hardware capabilities. Conditionals
based on gl_InvocationID are non-uniform, which means that they'll
often have to execute both paths. However, if there are fewer than
8 vertices, all invocations will happen within a single thread, so
barriers can become no-ops, which is nice. We also burn a maximum
of 4 registers for ICP handles, so we can compile without regard for
the value of N. It also works in all cases.
- DUAL_PATCH mode processes up to two patches at a time, where the first
four channels come from patch 1, and the second group of four come
from patch 2. This tries to provide better EU utilization for small
patches (N <= 4). It cannot be used in all cases.
- 8_PATCH mode processes 8 patches at a time, with a thread launched per
vertex in the patch. Each channel corresponds to the same vertex, but
in each of the 8 patches. This utilizes all channels even for small
patches. It also makes conditions on gl_InvocationID uniform, leading
to proper jumps. Barriers, unfortunately, become real. Worse, for
PATCHLIST_N, the thread payload burns N registers for ICP handles.
This can burn up to 32 registers, or 1/4 of our register file, for
URB handles. For Vulkan (and DX), we know the number of vertices at
compile time, so we can limit the amount of waste. In GL, the patch
dimension is dynamic state, so we either would have to waste all 32
(not reasonable) or guess (badly) and recompile. This is unfortunate.
Because we can only spawn 16 thread instances, we can only use this
mode for PATCHLIST_16 and smaller. The rest must use SINGLE_PATCH.
This patch implements the new 8_PATCH TCS mode, but leaves us using
SINGLE_PATCH by default. A new INTEL_DEBUG=tcs8 flag will switch to
using 8_PATCH mode for testing and benchmarking purposes. We may
want to consider using 8_PATCH mode in Vulkan in some cases.
The data I've seen shows that 8_PATCH mode can be more efficient in
some cases, but SINGLE_PATCH mode (the one we use today) is faster
in other cases. Ultimately, the TES matters much more than the TCS
for performance, so the decision may not matter much.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
I was setting it based off a pipe_rasterizer_state field that appears
to be entirely dead outside of the draw module respecting it.
I should be setting it when the primitive type reaching the SF is
neither points nor lines. This is, unfortunately, rather dirty,
as we have to look at the rasterizer state, the geometry shader state,
the tessellation evaluation shader state, and the primitive type...
Order of operations is important, otherwise we'll find the program we
just uploaded as the "old" compile and get confused why nothing is
different between the two keys.
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
I was lazy earlier and hadn't bothered typing / refactoring this.
Now I'm hitting some extra recompiles and would like to see why.
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
