Not to be confused with variablePointersStorageBuffer which is the
subset of VK_KHR_variable_pointers required to enable the extension.
This means we now have "full" support for variable pointers.
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
The SPIR-V spec is a bit underspecified when it comes to exactly how
you're allowed to use OpPtrAccessChain and what it means in certain edge
cases. In particular, what if the base pointer of the OpPtrAccessChain
points to the base struct of an SSBO instead of an element in that SSBO.
The original variable pointers implementation in mesa assumed that you
weren't allowed to do an OpPtrAccessChain that adjusted the block index
and asserted such. However, there are some CTS tests that do this and,
if the CTS does it, someone will do it in the wild so we should probably
handle it. With this commit, we significantly reduce our assumptions
and should be able to handle more-or-less anything.
The one assumption we still make for correctness is that if we see an
OpPtrAccessChain on a pointer to a struct decorated block that the block
index should be adjusted. In theory, someone could try to put an array
stride on such a pointer and try to make the SSBO an implicit array of
the base struct and we would not give them what they want. That said,
any index other than 0 would count as an out-of-bounds access which is
invalid.
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
This is required for being able to handle OpPtrAccessChain in SPIR-V
where the base type of the incoming pointer requires us to add to the
block index instead of the byte offset.
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Before, we always left workgroup variables as shared nir_variables and
let the driver call nir_lower_io. This adds an option to do the
lowering directly in spirv_to_nir. To do this, we implicitly assign the
variables a std430 layout and then treat them like a UBO or SSBO and
immediately lower all the way to an offset.
As a side-effect, the spirv_to_nir pass now handles variable pointers
for workgroup variables.
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Up until now, all pointers have been ivec2s. We're about to add support
for pointers to workgroup storage and those are going to be uints.
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
There is no good reason why we should have the same logic repeated in
get_vulkan_resource_index and vtn_ssa_offset_pointer_dereference. If
we're a bit more careful about how we do things, we can just use the one
function and get rid of the other entirely. This also makes the push
constant special case a lot more clear.
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
This commit moves them both into vtn_variables.c towards the top, makes
them take a vtn_builder, and replaces a hand-rolled instance of
is_external_block with a function call.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
This makes us key off of !offset instead of !block_index. It also puts
the guts inside a switch statement so that we can handle more than just
UBOs and SSBOs.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
This parallels what we do for vtn_block_load except that we don't yet
support anything except SSBO loads through this path.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
This is equivalent and means we don't have resource index code scattered
about.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
The layer parameter is signed. Fixes the error message seen when
running the arb_texture_multisample-errors test which checks a
negative layer value.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
After the mesa/st nir linking support, we start to see inputs/outputs
like:
decl_var shader_out INTERP_MODE_NONE float packed:uv (VARYING_SLOT_VAR9.x, 1, 0)
decl_var shader_out INTERP_MODE_NONE float packed:uv@0 (VARYING_SLOT_VAR9.y, 1, 0)
(ie. were location_frac != .x)
Unfortunately I overlooked the addition of the component parameter to
load_input/store_output, so when we started encountering inputs/outputs
with component other than .x, we'd end up loading/storing the wrong
input/output.
Signed-off-by: Rob Clark <robdclark@gmail.com>
Since in the NIR case, driver takes ownership of the NIR shader, we need
to clone what is passed to the driver. Normally this is done as part of
creating the shader variant (where is clone is anyways needed). But
compute shaders have no variants, so we were cloning earlier.
The problem is that after the NIR linking optimizations, we ended up
cloning *before* all the lowering passes where done.
So move this into st_get_cp_variant(), to make compute shaders work more
like other shader stages.
Signed-off-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
The atomic rat path has a bug in the ssbo path, refactor out the
address calcs from the load/store paths and reuse to fix the bug
in the buffer rat atomic path.
Signed-off-by: Dave Airlie <airlied@redhat.com>
This just changes how thread id loading is done, it makes
smaller shaders if we don't use thread id gprs.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Commit 94ca8e04ad ("spirv: Add vtn_fail and vtn_assert helpers") broke
Android builds which have -Werror enabled with the following errors:
external/mesa3d/src/compiler/spirv/spirv_to_nir.c:272:1: error: control may reach end of non-void function [-Werror,-Wreturn-type]
external/mesa3d/src/compiler/spirv/spirv_to_nir.c:810:1: error: control may reach end of non-void function [-Werror,-Wreturn-type]
...
The problem is the noreturn attribute is not enabled and we to define
HAVE_FUNC_ATTRIBUTE_NORETURN.
Auditing src/util/macros.h, we're also missing
HAVE_FUNC_ATTRIBUTE_RETURNS_NONNULL and HAVE_FUNC_ATTRIBUTE_WARN_UNUSED_RESULT,
so add them too.
Fixes: 94ca8e04ad ("spirv: Add vtn_fail and vtn_assert helpers")
Cc: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Signed-off-by: Rob Herring <robh@kernel.org>
libmesa_amd_common static dependency is added in Android build
to avoid the following building errors:
In file included from external/mesa/src/gallium/drivers/radeon/r600_buffer_common.c:24:
In file included from external/mesa/src/gallium/drivers/radeonsi/si_pipe.h:26:
external/mesa/src/gallium/drivers/radeonsi/si_shader.h:138:10: fatal error: 'ac_binary.h' file not found
^~~~~~~~~~~~~
1 error generated.
...
In file included from external/mesa/src/gallium/drivers/radeon/r600_gpu_load.c:34:
In file included from external/mesa/src/gallium/drivers/radeonsi/si_pipe.h:26:
external/mesa/src/gallium/drivers/radeonsi/si_shader.h:138:10: fatal error: 'ac_binary.h' file not found
^~~~~~~~~~~~~
1 error generated.
Fixes: 950221f923 ("radeonsi: remove r600_common_screen")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Just reuse the cs atomics bit and emit the hw atomic state.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
On Cayman we don't use the append/consume counters (fglrx doesn't)
and they don't seem to work well with compute shaders.
This just uses GDS instead to do the atomic operations.
v1.1: remove unused line.
v2: use EOS on cayman, it appears to work.
Acked-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
v2: - set d3d_drivers_path instead of dri_drivers_path
- Fix nine guard to check for all relavent gallium drivers
- Link with libswdri and libswkmsdri when necessary
- Fix pkg-config generation
- Add missing comma
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>