In the case of SPIRV tessellation shaders, the execution mode can be
specified in the tessellation control shader. So we need a way to know the domain
when compiling the tessellation evaluation shader.
Acked-by: M Henning <drawoc@darkrefraction.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24327>
Get rid of our own FMT_NONE and just use a NULL format in that case.
We're already checking su->tex.format != NULL in a bunch of cases but
none of those NULL checks are doing anything right now execpt detecting
when its an actual texture op and not an image op. We'd like them to
start also applying for image load/store with an unspecified format.
Reviewed-by: M Henning <drawoc@darkrefraction.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24327>
This came from CTS clipping tests with geometry shaders. Maybe can be
done as a lowering operation instead.
Reviewed-by: M Henning <drawoc@darkrefraction.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24327>
This instructs NIR to lower DeviceIndex to zero, which is needed
for a no-op implementation of VK_KHR_device_group.
Signed-off-by: Rebecca Mckeever <rebecca.mckeever@collabora.com>
Reviewed-by: M Henning <drawoc@darkrefraction.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24327>
gp has explicit load and store reg instructions so it can implement
the new nir_intrinsic_{decl,load,store}_reg intrinsics.
The implementation reuses the gpir_node_find and register_node functions
instead of just emitting load and store instructions in order to
preserve the behavior that there will never be a store_reg followed by
a load_reg of the same register in the same block.
Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Vasily Khoruzhick <anarsoul@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24246>
With the upcoming transition away from nir_register and modifiers in
nir, ppir needs to prepare for those features to be removed.
The most straightforward way for ppir without major reworks at this time
is to use the new nir_legacy helpers so that most of the backend can be
kept untouched (particularly due to use of abs/neg/sat modifiers).
Do this for now to unblock the transition with little impact; ppir can
be optimized at a later moment to have its own passes which will allow
it to drop the nir_legacy use and use just the chasing helpers.
Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Vasily Khoruzhick <anarsoul@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24246>
Even if loads are initially duplicated for each user, with new
optimizations such as folding modifiers, it may happen that loads
end up with only users in different blocks.
Handle that and don't delete or pipeline those nodes.
Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Vasily Khoruzhick <anarsoul@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24246>
This is the GLES3 sampler object API trivially backported to GLES2,
because it allows for simpler/better support in glamor and mutter.
The only code change we need is adding these to the generated dispatch
tables for ES2 contexts. The interactions with EXT_shadow_samplers,
EXT_texture_filter_anisotropic and EXT_texture_sRGB_decode are already
in place before this change, and OES_texture_border_clamp is always
supported in Mesa.
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2440>
ffloor(f2i(x)) can't be optimized to f2i(x) due to differing behaviour for
negative x, but u_blitter only uses this with nonnegative x so we can instead
use ftrunc(f2i(x)) which NIR will optimize to f2i(x) for us. This gets rid of
the silly ffloor instructions in blit shaders.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24319>
True for all users. I intentionally didn't add is_ssa asserts because they're
pointless and will be deleted, like, next week and will just make that churn
even more annoying.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24319>
This dri config is no longer necessary since the case where it was
required is being handled in the previous commit.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Igor Torrente <igor.torrente@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24222>
When we try to run a compositor on top of Zink, we hit a lockup
when enumerating the Vulkan devices. The vulkan_device_select
tries to reorder the devices and gets stuck waiting for the Xserver.
With this patch, we avoid this issue by detecting when we are running
a compositor and disabling the X and Wayland instance extensions.
And code is added to try to pick the same device as the display.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Igor Torrente <igor.torrente@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24222>
Make sure that all platform specific implementations of
eglGetPlatformDisplay() overwrite the Device field.
Fixes: c237539d62 ("egl: Implement EGL_EXT_explicit_device")
Signed-off-by: Robert Foss <rfoss@kernel.org>
Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24394>
_eglGetSurfacelessDisplay(), _eglGetWaylandDisplay() and
_eglGetGbmDisplay() handle the attrib_list[0] == EGL_NONE
case incorrectly by returning an EGL_BAD_ATTRIBUTE error.
Fix this bug, and switch the functions to use switch/case
in order to have the same structure as the sibling
_eglGetXXXDisplay() functions which support multiple attributes.
Fixes: c237539d62 ("egl: Implement EGL_EXT_explicit_device")
Signed-off-by: Robert Foss <rfoss@kernel.org>
Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24394>
It was not taken into account that without Offset decoration
the output is not written into XFB.
Aside from eliminating more outputs this change prevents gl_PerVertex
builtins generated by glslang from being kept alive in case when XFB
is enabled. Keeping such outputs alive may upset a driver.
VUID-StandaloneSpirv-Offset-04716:
"Only variables or block members in the output interface decorated
with Offset can be captured for transform feedback, and those
variables or block members must also be decorated with XfbBuffer
and XfbStride, or inherit XfbBuffer and XfbStride decorations from
a block containing them"
Additional info about glslang behavior could be found at:
https://github.com/KhronosGroup/glslang/issues/1526
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24318>
It was not taken into account that without Offset decoration
the output is not written into XFB.
Aside from eliminating more outputs this change prevents gl_PerVertex
builtins generated by glslang from being kept alive in case when XFB
is enabled. Keeping such outputs alive may upset a driver.
VUID-StandaloneSpirv-Offset-04716:
"Only variables or block members in the output interface decorated
with Offset can be captured for transform feedback, and those
variables or block members must also be decorated with XfbBuffer
and XfbStride, or inherit XfbBuffer and XfbStride decorations from
a block containing them"
Additional info about glslang behavior could be found at:
https://github.com/KhronosGroup/glslang/issues/1526
Fixes: e95531e101
("radv: fix gathering XFB info if there is dead outputs")
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24318>
Unlike some other core structs, I didn't bother to copy the entire
VkSamplerCreateInfo into the sampler. There isn't typcially a lot of
pre-processing we can or need to do here and we may as well keep the
struct small for now. It does, however, handle most things involving
chain-in structs.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24335>
Each of these are just themselves in a single plane. This is copied
from the 8888 version, these are just the 10, 12, and 16-bit versions.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24335>
This follows the pipeline libraries method of including SPIR-V and lets
you provide the NIR shader without wrapping it in a VkShaderModule.
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24335>
Xe KMD only requires userptr to be bound to VM, so here returning
UINT32_MAX as gem_handle in Xe version of gem_create_userptr() for all
userptr bos.
As no bo is created it was also necessary to add additional handling
to xe_gem_close().
The vm bind side of userptr was already implemented, so it was only
necessary add the special handling and the kmd vm bind call.
This fixes piglit@amd_pinned_memory subtests that makes uses of
userptr.
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23817>
Xe support of userptr will be implemented in the next patch,
this is just moving the i915 function to KMD backend.
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23817>
The next patch will need a special handling when closing userptr bos
in Xe KMD, so here moving iris_bufmgr_bo_close() to kmd backend
and changing the gem_handle parameter to iris_bo.
There still one DRM_IOCTL_GEM_CLOSE call left in iris_bufmgr, that
is used to close exported gem handles.
iris_bufmgr_get_for_fd() could be used to get the iris_bufmgr but
we would still have problems with bo_export != iris_bo, so leaving
as is until a better solution is found.
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23817>
This was being called from the common code, so it was being executed
for Xe KMD.
Luckily Xe don't have any uAPI at 0x1f offset.
There is still one user of i915_gem_set_domain() in iris_bufmgr.c
so it was duplicated in i915 backend but a future patch in this series
will take to remove it when the userptr code moves to backend.
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23817>
