See "i965/gen9: Optimize slice and subslice load balancing behavior."
for the rationale. According to Jason, improves Aztec Ruins
performance by 2.7%.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org> (v1)
v2: Undo CPU performance micro-optimization done in i965 and iris due
to lack of data justifying it on anv. Use
cmd_buffer_apply_pipe_flushes wrapper instead of emitting pipe
control command directly. (Jason)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
We stopped using this when we moved to Jason's mi_builder.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
This is the MOCS setting used for the A64 stateless messages which we
sometimes use for SSBO operations.
Fixes: 48ed2a7bb0 "anv: Implement VK_EXT_buffer_device_address"
Fixes: 79fb0d27f3 "anv: Implement SSBOs bindings with GPU addr..."
Reviewed-by: Chad Versace <chadversary@chromium.org>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
This should fix floating-point border color on all gen7 HW. Integer is
still thoroughly busted on gen7 because it doesn't exist on IVB and it's
crazy on HSW.
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Intel hardware didn't get support for sampling from W-tiled (required
for stencil) images until Broadwell so we can't directly sample from
stencil. Instead, if we want to support stencil texturing on gen7
hardware, we have to keep a texture-capable shadow copy around and use
BLORP to update when stencil changes. The one thing this commit does
not implement is self-dependencies with stencil input attachments.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99493
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Modern DXVK requires event support [1], but looks like it only
uses vkCmdSetEvent() + vkGetEventStatus(). So we can just
borrow the relevant code from gen8, leaving CmdWaitEvents still
unimplemented.
[1] 8c3900c533
v2: Also move CmdWaitEvents into genX_cmd_buffer.c (Jason)
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
On CNL+, the clear color struct is composed of RGBA channel values and
fields which are either reserved by the HW or used to control
fast-clears. Currently anv initializes the channel values to zero and
allows the other fields to be undefined.
Satisfy the MBZ field requirements by removing an optimization that
doesn't hold true for CNL+ and pulling in the number of dwords to
initialize from ISL.
Cc: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
The key reason for that mechanism is gone: all the extra optional data
that could be in the anv_push_constants was moved elsewhere. At this
point, just put anv_push_constants directly in anv_cmd_state (part of
anv_cmd_buffer).
v2: Remove a NULL check we don't need anymore in
anv_cmd_buffer_push_constants(). (Lionel)
Fix size we consider for valid push params. (Lionel)
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
This commit changes anv to put bindless handles and sampler pointers
into the descriptor buffer and use those instead of bindful when we run
out of binding table space. This "spilling" of descriptors allows to to
advertise an almost unbounded number of images and samplers.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
This is really where they belong; not push constants. The one downside
here is that we can't push them anymore for compute shaders. However,
that's a general problem and we should figure out how to push descriptor
sets for compute shaders. This lets us bump MAX_IMAGES to 64 on BDW and
earlier platforms because we no longer have to worry about push constant
overhead limits.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
We spend a lot of time in the driver adding things to hash sets to track
residency. The reality is that a properly built Vulkan app uses large
memory objects and sub-allocates from them. In a typical frame, most of
if not all of those allocations are going to be resident for the entire
frame so we're really not saving ourselves much by tracking fine-grained
residency. Just throwing everything in the validation list does make it
a little bit more expensive inside the kernel to walk the list and
ensure that all our VA is in order. However, without relocations, the
overhead of that is pretty small.
If we ever do run into a memory pressure situation where the fine-
grained residency could even potentially help, we would likely be
swapping one page out to make room for another within the draw call and
performance is totally lost at that point. We're better off swapping
out other apps and just letting ours run a whole frame.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Vulkan spec doesn't explicitly forbid zero size transform
feedback buffers.
Having zero size xfb caused SurfaceSize overflow and
triggered assert in debug build.
The only way to have zero size SO_BUFFER is to disable
SO_BUFFER as stated in hardware spec.
From SKL PRM, Vol 2a, "3DSTATE_SO_BUFFER":
"If set, stream output to SO Buffer is enabled,
if 3DSTATE_STREAMOUT::SO Function ENABLE is also enabled.
If clear, the SO Buffer is considered "not bound" and effectively
treated as a zero- length buffer for the purposes of SO output and
overflow detection. If an enabled stream's Stream to Buffer Selects
includes this buffer it is by definition an overflow condition.
That stream will cause no writes to occur,
and only SO_PRIM_STORAGE_NEEDED[<stream>] will increment."
Fixes: 36ee2fd61c "anv: Implement the basic form of VK_EXT_transform_feedback"
Signed-off-by: Danylo Piliaiev <danylo.piliaiev@globallogic.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
This buffer goes along side the CPU data structure and may contain
pointers, bindless handles, or any other descriptor information.
Currently, all descriptors are size zero and nothing goes in the buffer
but this commit sets up the framework we will need later.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
This is what we're actually storing in the descriptor set and consuming
when we bind surface states. This commit renames image_count to
image_param_count a few places and moves the decision to not count image
params on gen9+ into anv_descriptor_set.c when we build the layout.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
This commit moves our handling of gl_NumWorkgroups over to work like our
handling of other special bindings in the Vulkan driver. We give it a
magic descriptor set number and teach emit_binding_tables to handle it.
This is better than the bias mechanism we were using because it allows
us to do proper accounting through the bind map mechanism.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Annoyingly, this requires that we implement integer division on the
command streamer. Fortunately, we're only ever dividing by constants so
we can use the mulh+add+shift trick and it's not as bad as it sounds.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Doesn't save us a great deal of lines but at least they get decoded in
aubinators.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
In commit 9a7b319903 ("anv/query: flush render target before
copying results") we tracked all the render target writes to apply a
flushes in the vkCopyQueryResults(). But we can narrow this down to
only when we write a buffer (which is the only input of
vkCopyQueryResults).
v2: Drop newer render target write flags introduce by 1952fd8d2c
("anv: Implement VK_EXT_conditional_rendering for gen 7.5+")
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net> (v1)
Conditional rendering affects next functions:
- vkCmdDraw, vkCmdDrawIndexed, vkCmdDrawIndirect, vkCmdDrawIndexedIndirect
- vkCmdDrawIndirectCountKHR, vkCmdDrawIndexedIndirectCountKHR
- vkCmdDispatch, vkCmdDispatchIndirect, vkCmdDispatchBase
- vkCmdClearAttachments
Value from conditional buffer is cached into designated register,
MI_PREDICATE is emitted every time conditional rendering is enabled
and command requires it.
v2: by Jason Ekstrand
- Use vk_find_struct_const instead of manually looping
- Move draw count loading to prepare function
- Zero the top 32-bits of MI_ALU_REG15
v3: Apply pipeline flush before accessing conditional buffer
(The issue was found by Samuel Iglesias)
v4: - Remove support of Haswell due to possible hardware bug
- Made TMP_REG_PREDICATE and TMP_REG_DRAW_COUNT defines to
define registers in one place.
v5: thanks to Jason Ekstrand and Lionel Landwerlin
- Workaround the fact that MI_PREDICATE_RESULT is not
accessible on Haswell by manually calculating
MI_PREDICATE_RESULT and re-emitting MI_PREDICATE
when necessary.
v6: suggested by Lionel Landwerlin
- Instead of calculating the result of predicate once - re-emit
MI_PREDICATE to make it easier to investigate error states.
v7: suggested by Jason
- Make anv_pipe_invalidate_bits_for_access_flag add CS_STALL
if VK_ACCESS_CONDITIONAL_RENDERING_READ_BIT is set.
v8: suggested by Lionel
- Precompute conditional predicate's result to
support secondary command buffers.
- Make prepare_for_draw_count_predicate more readable.
Signed-off-by: Danylo Piliaiev <danylo.piliaiev@globallogic.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
v2: by Jason Ekstrand
- Move out of the draw loop population of registers
which aren't changed in it.
- Remove dependency on ALU registers.
- Clarify usage of PIPE_CONTROL
- Without usage of ALU registers patch works for gen7+
v3: set pending_pipe_bits |= ANV_PIPE_RENDER_TARGET_WRITES
Signed-off-by: Danylo Piliaiev <danylo.piliaiev@globallogic.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Change block_pool->bo to be a pointer, and update its usage everywhere.
This makes it simpler to switch it later to a list of BOs.
v3:
- Use a static "bos" field in the struct, instead of malloc'ing it.
This will be later changed to a fixed length array of BOs.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
The ++ operator strikes again.
Fixes: f92c5bc8f3 ("anv/device: fix maximum number of images supported")
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
We had defined MAX_IMAGES as 8, which we used to size the array for
image push constant data. The comment there stated that this was for
gen8, but anv_nir_apply_pipeline_layout runs for all gens and writes
that array, asserting that we don't exceed that number of images,
which imposes a limit of MAX_IMAGES on all gens.
Furthermore, despite this, we are exposing up to 64 images per shader
stage on all gens, gen8 included.
This patch lowers the number of images we expose in gen8 to 8 and
keeps 64 images for gen9+ while making sure that only pre-SKL gens
use push constant space to handle images.
v2:
- <= instead of < in the assert (Eric, Lionel)
- Change the way the assertion is written (Eric)
v3:
- Revert the way the assertion is written to the form it had in v1,
the version in v2 was not equivalent and was incorrect. (Lionel)
v4:
- gen9+ doesn't need push constants for images at all (Jason)
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com> (v3)
In the following scenario :
1. Create image format R8G8B8A8_UNORM
2. Create image view format R8G8B8A8_SRGB
3. Clear the view through a sub pass to a particular color
4. Barrier on the image to from color attachment to source transfer
5. Copy the image into a linear buffer to check the content
The step 4 resolving the clear color is unaware of the SRGB format of
the view, because the blorp resolve operations operate on images the
color associated with the resolve will not operate on SRGB format but
UNORM. Leading to the wrong color being written into surfaces.
This change forces a clear color resolve at the end of the render pass
so following resolves won't have to deal with the clear color with a
format that doesn't match the image's format.
On gfxbench vulkan_5_normal 1280x720, this appear to cost us ~0.5fps,
from 49.316 down to 48.949.
v2: Only fast clear resolve when image & view have different formats
(Lionel)
v3: Update warning (Jason)
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108911
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Suggested-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: mesa-stable@lists.freedesktop.org
Resolve operations can happen when dealing with view (begin/end
subpasses) in which case the view's format needs to apply, not the
image's format.
v2: Relayout arguments of a ccs_op() call (Jason)
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Suggested-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108911
Cc: mesa-stable@lists.freedesktop.org
We've made the choice not to use fast clears on layer > 0 with
multilayer images. This is partly because we would need to store
multiple clear colors for each layer, making the existing memory
layout, already including aux surfaces, fast clear color, image state,
etc... even more complex.
Partial resolves are the operations transfering the clear colors into
the auxiliary buffers. This operation is currently implemented in
Blorp by loading the clear color from the image's BO, into a shader
that then samples from the auxiliary buffer and writes the color only
if it isn't there already.
The problem here is that because we store only one clear color for all
layers and it is used for partial resolves. If you trigger a partial
clear on a layer > 0, then you're likely to deal with a color that is
not what you actually want. In the particular issues below, we have
multiple layers, each cleared with a different color but the partial
resolve just writes the wrong color into the auxiliary buffers for
layers > 0.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108910
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108911
Cc: mesa-stable@lists.freedesktop.org
When we first started using genxml, we decided to represent MOCS as an
actual structure, and pack values. However, in many places, it was more
convenient to use a numeric value rather than treating it as a struct,
so we added secondary setters in a bunch of places as well.
We were not entirely consistent, either. Some places only had one.
Gen6 had both kinds of setters for STATE_BASE_ADDRESS, but newer gens
only had the struct-based setters. The names were sometimes "Constant
Buffer Object Control State" instead of "Memory", making it harder to
find. Many had prefixes like "Vertex Buffer MOCS"...in a vertex buffer
packet...which is a bit redundant.
On modern hardware, MOCS is simply an index into a table, but we were
still carrying around the structure with an "Index to MOCS Table" field,
in addition to the direct numeric setters. This is clunky - we really
just want a number on new hardware.
This patch eliminates the struct-based setters, and makes the numeric
setters be consistently called "MOCS". We leave the struct definition
around on Gen7-8 for reference purposes, but it is unused.
v2: Drop bonus "Depth Buffer MOCS" fields on Gen7.5 and Gen9
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
This change tracks render target writes in the pipeline and applies a
render target flush before copying the query results to make sure the
preceding operations have landed in memory before the command streamer
initiates the copy.
v2: Simplify logic in CopyQueryResults (Jason)
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108909
Fixes: 37f9788e9a ("anv: flush pipeline before query result copies")
Cc: mesa-stable@lists.freedesktop.org
L3 allocation table in h/w specification recommends using 4 KB
granularity for programming allocation fields in L3CNTLREG.
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
The default setting of this bit is not the desirable behavior.
WA_1406697149
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Ref: 263b584d5e "i965/skl: Emit extra zeros in STATE_BASE_ADDRESS on Skylake."
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
On Broadwell and above, we have to use different MOCS settings to allow
the kernel to take over and disable caching when needed for external
buffers. On Broadwell, this is especially important because the kernel
can't disable eLLC so we have to do it in userspace. We very badly
don't want to do that on everything so we need separate MOCS for
external and internal BOs.
In order to do this, we add an anv-specific BO flag for "external" and
use that to distinguish between buffers which may be shared with other
processes and/or display and those which are entirely internal. That,
together with an anv_mocs_for_bo helper lets us choose the right MOCS
settings for each BO use.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99507
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
