Having NIR-based instruction selection tests instead of GLSL lets us be
more precise with the input NIR.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28301>
These don't really do anything. They're just a source and user of SSA
defs.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28301>
For
if (uniform) {
if (divergent)
continue
} else {
break
}
we don't need to consider the continue to be divergent.
No fossil-db changes.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28301>
This simplifies instruction selection and fixes the case where the loop
ends with a continue instruction.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28301>
STACK_ARRAY() is used in a lot of places. When games are running we
see STACK_ARRAY() arrays being used all the time: each queue
submission uses 6, WaitSemaphores and syncobj waiting also uses them:
they're constantly present in Vulkan runtime.
There's no need for STACK_ARRAY()'s stack array to be initialized,
callers cannot not depend on it. If the number of elements is greater
than STACK_ARRAY_SIZE, then STACK_ARRAY() will just malloc() the array
and return it not initialized: anybody depending of
zero-initialization is going to break when the array is big.
The reason why we're zero-intializing STACK_ARRAY()'s stack array is
to silence -Wmaybe-uninitialized warnings: see commit d7957df318
("vulkan: fix uninitialized variables"). I don't think that commit is
the ideal way to deal with the problem, so this patch proposes a
better solution.
The problem here is that zero-initializing it adds code we don't need
for every single caller. STACK_ARRAY() already has 63 callers and only
3 of them are affected by the -Wmaybe-uninitialized warining. So here
we undo what commit d7957df318 did and instead we fix the 3 cases
that actually generate the -Wmaybe-uninitialized warnings.
Gcc is only emitting those warinings because it knows that the number
of elements in the array may be zero, so the loops we have that set
elements to the array may end up do nothing, and then we pass the
array uninitialized to other functions.
For the cases related to vk_sync this is just returning VK_SUCCESS
earlier, instead of relying on the check that eventually happens at
__vk_sync_wait_many(). For the vkCmdWaitEvents() function, the Vulkan
spec says that "eventCount must be greater than 0", so the early
return doesn't hurt anybody either. In both cases we make the zero
case faster by not defining an 8-sized array, zero-initializing it,
then returning success without using it.
Reference: d7957df318 ("vulkan: fix uninitialized variables")
Acked-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28288>
The padded width/height is stored in samples, while the blit box
dimensions need to be specified in pixels. Use the unpadded
width/height of the resource levels to generate the blit box
dimensions used to copy a resource. The blit code already extends
those sizes to the padded sizes when necessary and possible.
This commit needed to be reverted previously, as it triggered
a bug in the RS blit path. This bug has been fixed so now we
can reapply this change to get the blit dimensions corrected.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <cgmeiner@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28598>
When trying to increase the height alignment to unlock multi-pipe resolve for
better performance we need to be careful to not overstep the source dimensions
as this would cause the blit to be rejected.
Do so and also rearrange the code a bit to make it more obvious what is being
done.
Fixes: 797454edfc ("etnaviv: rs: fix blits with insufficient alignment for dual pipe operation")
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <cgmeiner@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28598>
Not setting this bits, it seems we get incorrect depth values (i.e
not zero) for null depth/stencil tiles.
Fixes vkd3d-proton's test_sparse_depth_stencil_rendering
CTS doesn´t seem to exercise any depth/stencil format.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: mesa-stable
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28611>
This has been a long and painful extension to implement but finally
RADV has support for it on all supported generations. It's passing
VKCTS and GLCTS through Zink but the number of application using ESO
is currently very limited.
This adds RADV_DEBUG=noeso to disable it for debugging purposes.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28374>
We currently allocate 10 MB up-front for tiler heaps. That can sometimes
be a lot, depending on the system. So let's add some DRIconf variables to
allow tweaking these.
Until we have incremental-rendering implemented, it's a bit risky to lower
these too much. Be warned.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28594>
copy_image calls blit now for multisampled images, including
depth. But blit explicitly uses only PIPE_MASK_RGBA, so it is
incapable of copying depth buffers.
This patch checks the destination format and uses PIPE_MASK_ZS if
it is a depth or stencil. Ideally we would simply use PIPE_MASK_RGBAZS
always, but not all drivers actually handle getting this mask
(they probably should, but that's another story).
The change to copy_image was in 5027b5aa2, so in some sense this
patch "fixes" that. In fact though the issue wasn't in the copy_image
change, it was always latent in blit().
Fixes: 5027b5aa28 ("gallium: stop calling resource_copy_region for multisampled copy_image")
Signed-off-by: Eric R. Smith <eric.smith@collabora.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28585>
This is a more logical place for them and will avoid us adding more
function validtion calls to the st glsl to nir file in future
patches.
Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28538>
Commit 719bf30165 added this removal code with the following
justification:
"The GLSL 4.10 rules for redeclaration of built-in interface blocks
(which we've chosen to regard as clarifications of GLSL 1.50) only
require gl_PerVertex blocks to match in shaders that actually use
those blocks. The easiest way to implement this is to detect
situations where a compiled shader doesn't refer to any elements of
gl_PerVertex, and remove all the associated ir_variables from the
shader at the end of ast-to-ir conversion."
However the intention is to avoid matching a redeclared block with
gl's default block if unused. We are still required to do block
matching in the shader should the block be redeclared, even if unused.
So with this change we only remove the block if it is both unused
and not redeclared.
The existing glsl IR code managed to avoid failing CTS tests for this
due to seemingly magical or hacky use of the symbol table but fixing
it will make things much clearer, and also allow a nir version of
this validation in a following patch.
Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28538>
This is a perpetual bug that hits Windows. In the MSVC CRT, qsort
is unstable, where the glibc qsort is stable. So apps run fine on
Windows IHV drivers, and on Linux Mesa drivers, and only break down
when running on Windows Mesa drivers.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/10922
Reviewed-by: Sil Vilerino <sivileri@microsoft.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28586>
When I bumped the max size of VGRFs, I should have bumped the values
in the scheduler too.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: d33aff783d ("intel/fs: add support for sparse accesses")
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28188>
These calls need a different preparation and result type handling than
the calls that query resource info, so we move this to an extra
function that can then be called from the default case of the
texop switch.
v2: drop ALWAYS_INLINE (zmike)
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28511>
