Based on observations of the generated assembly, this instruction appears to:
- Swizzle the 8/16 component vector in src0 according to the pattern defined in src1.
- Apply a enable mask from src2 to selectively modify elements.
I encountered this instruction while experimenting with _viv_asm and
packed types.
Here is one exmaple kernel:
kernel void k(global int* out, int a, int b) {
_viv_char2_packed s;
_viv_asm(MOV, s.x, s, a);
_viv_asm(MOV, s.y, s, b);
out[0] = s.x + s.y;
}
Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33474>
We currently assume the implicit offset shift is always 2. However, this
shift is actually based on the type, making sure the offset fields are
in units of the type size. The full offset calculation is as follows:
((SRC2<<SRC2_SHIFT) + OFF)<<TYPE_SHIFT
Where SRC2, SRC2_SHIFT, and OFF are instruction fields while TYPE_SHIFT
is implicit and derived from the TYPE field.
This commit implements (dis)assembly support for this, adopting the
syntax used by the blob.
Signed-off-by: Job Noorman <jnoorman@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33503>
The assembly syntax of certain instructions differs significantly
between generations (e.g., ldg.a/stg.a) so it's useful to be able to
generate syntax error based on the generation we are assembling for.
Signed-off-by: Job Noorman <jnoorman@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33503>
RADV won't compile without the added includes after we
stop including the full nir.h from the VK common functions.
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Acked-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33439>
This is to stop including nir related stuff in places that
actually don't need that.
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Acked-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33439>
The nir_tcs_info struct is like nir_xfb_info in the sense that
it's very specialized and not often used, so it deserves its own
header too.
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33439>
These are necessary if we want to stop including the full nir.h
in most places accross the code base.
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33439>
This will allow forward declaring these in a subsequent commit.
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33439>
This file is going to be included in some headers that are shared
between normal C++ and OpenCL. And sadly OpenCL can't handle this.
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33439>
For bin merging, we will have to first sample all bins in the pipe, then
determine which bins can be merged, then iterate over bins. Combine all
of the information required to render a bin into a tu_tile_config struct
and pass it down to tu6_emit_tile_select(). This will let us more
flexibly construct a list of bins later.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33230>
These jobs need to run in pre-merge, so let's make sure their execution
time remains in check so as not to block merges due to non-merge
pipelines.
Signed-off-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33486>
Deqp randomly dies at the beginning of VKCTS, which fails most
pipelines.
Reducing the job concurrency fixes the issue, so let's roll with that
until I go buy more RAM for this machine.
Signed-off-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33486>
This was causing a crash on
"dEQP-VK.graphicsfuzz.cov-function-large-array-max-clamp" where the
test was trying to allocate ~6GB of TLS.
Considering we were already doing something identical before those changes,
we can just add nir_lower_scratch_to_var before nir_lower_vars_to_scratch
to get the expected behavior (Cleaning LLVM spilling mess around pan_pack)
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Fixes: 1619fc596a ("bi: Optimize scratch access")
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33510>
