Prevent the accidental passing of 0 components or bits, as it makes no sense.
Cc: mesa-stable
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Suggested-by: Karol Herbst <kherbst@redhat.com>
Signed-off-by: David Heidelberg <david@ixit.cz>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31103>
By default we have to take into account that the application could set
offsets, or that one kernel launch won't fit into a single hw dispatch.
In order to mitigate the overhead it causes at kernel runtime, and because
those things are in most cases irrelevant, we compile an optimized kernel
making a few assumptions.
We also make use of the the workgroup_size_hint as an additional
optimization.
This should speed up relatively small kernels significantly as it can cut
the instruction count in half for those.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30152>
If we allow this whenever the extension is supported by the driver, we
end up allowing out-of-spec behavior for instance on GLES 1.x.
So let's tighten these checks to consider the current API, by using the
new helpers.
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31039>
We don't want to expose GLSL versions above what should be exposed for
the current API, otherwise we allow out-of-spec behavior.
This is unlikely to have any real-world effect, because most of the
time, the version is the highest supported. It's only if the version is
artificially limited that this should matter.
But it's also a bit cleaner this way.
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31039>
Instead of recreating the packets in the DGC prepare shader, the best
solution is to emit them to a temporary CS object at pipeline creation
time. Then in the DGC prepare shader, the driver just needs to copy
the packets.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31101>
When VK_PIPELINE_CREATE_2_CAPTURE_DATA_BIT_KHR is set, implementations
shouldn't store pipeline data to an internal cache.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30812>
The Vulkan spec says:
"disableInternalCache can be used to disable the driver’s internal
cache, allowing an application to take full control of both memory
and disk usage."
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30812>
This change was wrong but there is nothing testing this. For example,
if we have a pipeline with VS+GS+FS and DGC only updates push constants
for VS. On GFX9+, VS is merged to GS, so the VS push constants info
would be zero and nothing would be emitted.
This reverts commit 45319cb253.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31099>
When use LA464(3A5000 pc), LLVM18 and kernel(which not supports lsx/lasx),
it will get a illegal instruction error. LA464 will return LA464 and
LA664 will return a generic when calling getHostCPUName(). Return LA464
name will add all feature(+f,+d,+lsx,+lasx,lvz) by default and generate
vector instruction, generic will add null. Hence we use util_get_cpu_caps
to judge if kernel supports lsx/lasx or not.
Set lsx and lasx by util_get_cpu_caps. Only LLVM 18 on loongarch
supports lsx and lasx completely. LLVM 16 isn't supported and
LLVM 17's LSX support is incomplete.
Signed-off-by: Jiale Zhao <zhaojiale@loongson.cn>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30220>
Although we don't want to rely on hwconfig for devinfo->verx10 == 120,
due to the dependence on closed source software, we do check to see if
hwconfig reports different values in the DEVINFO_HWCONFIG macro.
Matt was seeing this warning on 8086:a7a0:
> MESA: warning: INTEL_HWCONFIG_TOTAL_PS_THREADS (128) != devinfo->max_threads_per_psd (64)
Reported-by: Matt Turner <mattst88@gmail.com>
Fixes: 3e4f73b3a0 ("intel/dev: Update hwconfig => max_threads_per_psd for Xe2")
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31077>
This caused GLX applications to segfault under zink.
This technically also fixes cleanup when any driver fails init
Signed-off-by: Sid Pranjale <mail@sidonthe.net>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31072>
Now there's a single vtn_type associated with a vtn_pointer, so
discard the qualifier.
After this and previous changes, here's a summary of where/what types
are:
```
struct vtn_pointer *p;
p->type; // type of this pointer
p->deref; // NIR deref of this pointer (unchanged)
p->type->pointed; // type of the object pointed by this pointer
```
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31069>
This can be obtained by the pointer type (currently at ptr_type). For
the cases where there wasn't an user provided type for that, now create
an internal vtn_type. This can happen when creating intermediate
vtn_pointer for complex loads/stores/copies.
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31069>
To avoid confusion with the vtn_pointer::deref that is a NIR deref. New
name comes from description of OpTypePointer, where is described as the
"type of the object pointed to".
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31069>
The TGL PRM says,
This bit should never be programmed to 0
So, set it to true. I chose not to use the MBO attribute in genxml
because the field lacks the "Format: MBO" line in the PRM.
We previously made this programming conditional with commit 2e1be771e4
because of tests failing in
dEQP-GLES3.functional.texture.specification.tex*depth*. However, those
failures were fixed when we started using gl_FragDepth for depth buffer
copies in commit 6cec618e82.
Note: when bisecting this, I cherry-picked commit 7a68045b5d in order
to get past build failures related to a deprecated python function.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31066>
If there's difference between scan_inst dest type and inst src type we
should be more careful, because difference in signedness can cause
incorrect results after the propagation.
Updated ror-default.trace hash, as the change fixes misrendering there.
Fixes: b23432c5 ("intel/fs: Fix a cmod prop bug when the source type of a mov doesn't match the dest type of scan_inst")
Signed-off-by: Sviatoslav Peleshko <sviatoslav.peleshko@globallogic.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30998>
