That's not necessarily true for actual blits, so let's pass the min/max
coords to pan_blitter_emit_viewport().
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
The type has already been selected when forging the key, no need to do
it again pan_blitter_emit_bifrost_blend().
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
There's clearly no need to pass RTs/ZS views around, pass only what we
really need.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
This function will be used for blits too.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
We will emit one viewport and attach it to several DCDs for multi-layer
blits. Make that possible by adjusting the prototype and rename the
function along the way.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
The texture descriptors will be emitted once and re-used in several
DCDs when blitting more than one layer. Rename the functions and make
them return a GPU pointer instead of filling the DCD directly.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
It doesn't make sense to have it done in pan_preload_emit_varying().
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
The sampler descriptors will be emitted once and re-used in
several DCDs for the multi-layer blit case. We will also need
to select the filter. Let's adjust the pan_preload_emit_*_sampler()
functions to support that and rename them along the way.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
It should speed up a bit hash calculation and allow us to add extra
info without changing the structure size.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
first should be initialized to true if we want to get rid of the leading
';' in the shader name.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Disabling it that made us start to fail register allocation for a
few tests. Re-enable it until we figure out what is causing that.
Fixes: ca9e0871fb ('v3d: enable NIR loop unrolling')
Reviewed-by: Juan A. Suarez <jasuarez@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10684>
We have to unwind the lod into the scalar path correctly.
Fixes a crash with renderdoc demo
Fixes: e168d148d7 ("gallivm/nir: handle non-uniform texture offsets")
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10676>
this doesn't actually do anything other than test the codepath since
the hashed pipeline state will still change on every frontface change,
but it's a start
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10508>
by creating a separate hash value for vertex state, changes to the vbos and
bindings requires massively less overhead as it doesn't require the rest of
the pipeline state to be rehashed
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10508>
in theory we could have multiple pools per layout, but really it just makes
more sense to make the layouts persistent since they're more likely to be
reused
32bit compile fixes by Witold Baryluk
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9965>
Remove the PAN_MESA_DEBUG=fp16 flag that was hiding it.
Skip two buggy dEQP tests. See linked discussion. We'll need to make
sure this gets sorted out before submitting conformance, but I don't see
a test with a fix in the pipeline as valid reason to hold back valid
code.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9239>
Needed for constant folding to be effective. But don't copyprop into
instructions already reading from FAU, that will just end up adding more
moves!
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9239>
We don't vectorize transcendentals, since those are scalar only in
hardware. Also don't vectorize a few places where impedance mismatches
between NIR and the hardware make handling vectors infeasible for now.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9239>
gcc 11 warns:
[846/1506] Compiling C object src/mesa/libmesa_common.a.p/main_shaderapi.c.o
In function ‘shader_source’,
inlined from ‘_mesa_ShaderSource_no_error’ at ../src/mesa/main/shaderapi.c:2137:4:
../src/mesa/main/shaderapi.c:2095:25: warning: ‘*offsets_10 + _130’ may be used uninitialized [-Wmaybe-uninitialized]
2095 | totalLength = offsets[count - 1] + 2;
I can't really see how it's getting to that conclusion, but allocating
`offsets` with calloc is both natural to do here and guarantees
initialization.
Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10671>
gcc 11 warns:
../src/util/format/u_format_fxt1.c:940:22: warning: ‘fxt1_variance.constprop’ accessing 128 bytes in a region of size 64 [-Wstringop-overflow=]
940 | int32_t maxVarR = fxt1_variance(NULL, &input[N_TEXELS / 2], n_comp);
But, suspiciously, if you inline fxt1_variance the warning goes away.
What's happening is that the 2nd arg is uint8_t[N_TEXELS][MAX_COMP], so
it looks like we're passing too small of an array in since gcc knows
that `input` is also [N_TEXELS][MAX_COMP]. Fair enough. Fix the
signature to reflect what's actually going on, and remove some unused
arguments while we're at it.
Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10671>
This only implements dynamic primitive restart enable, depth bias
enable and rasterizer discard enable. I leave logic op and patch
control points for later.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10340>
