These require a lot more handholding in the IR than we can deal with at
this stage; we need to restrict ourselves to frcp/sqrt.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4470>
We've had alignment parameters on these operations for a while but a
bunch of places weren't setting them. That should be resolved now so we
can start validating that they're always set.
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4441>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4441>
All buffers have Name != 0.
Note that there is no longer the pointer dereference to get Name,
so it's faster.
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4466>
Since VAOs don't use NullBufferObj for vertex attribs anymore, let's remove
more uses of NullBufferObj.
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4466>
This improves performance by 5% in the game "torcs", FPS: 98.83 -> 103.73
It does a lot of glPush/PopClientAttrib, which exacerbates the overhead
of setting NullBufferObj.
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4466>
Precompute the default state in gl_context, and just copy it when we create
a VAO.
This also helps glPushClientAttrib function, which always creates a VAO,
which has a substantial CPU overhead in profiles.
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4466>
Fixes a building error due to compiler/aco_statistics.cpp
missing in src/amd/Makefile.sources
Fixes: b1544352 ("aco: add various compiler statistics")
Signed-off-by: Mauro Rossi <issor.oruam@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Programs can be shared among many contexts and each program holds a
variant list which has context specific variants. When context gets
destroyed it must make sure it relases all variants, otherwise remaining
context that utilizes same program will attempt to save a zombie shader
for already deleted context when releasing program and its variants.
Fixes:
dEQP-EGL.functional.sharing.gles2.program.render
and other flaky multihread dEQP-EGL failures.
v2: pass program pointer via & (Marek)
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Cc: mesa-stable@lists.freedesktop.org
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4386>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4386>
These raise another set of issues -- indeed, not all of these tests are
passing, since it turns out I have an actual bug in the packing code. So
after all this work, test bringup has identified an actual issue :)
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4458>
We'd like to dump both the generated IR (so we know exactly what's being
tested) as well as the compiled program (so we know what's running for
comparison).
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4458>
Given an instruction, we'd like to wrap it in a clause with some I/O on
each end so we can pack it up and send it to the hardware to compare
against the simulator.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4458>
Missing some details about modifiers but the core structure should
look like this for 32 and 16-bit, I think. My sincerest apologies for
the macro magic, I tried to make it the least bad I could but trying to
keep down repitition.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4458>
We'd like to step through a BIR program to evaluate it for testing.
Let's stub out some infrastructure for modeling Bifrost.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4458>
It turns out ports need to be in order of the arguments of an
instruction (port 3, that is), which breaks on instructions whose IR
argument order is different from the packed order, like csel. So fix
that.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4458>
It's a special case for unclear reasons, and if you mess it up you get
INSTR_INVALID_ENC. Isn't hardware fun?
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4458>
This change removes the basic infrastructure to work with perfmon
from the perfmon query impl and puts it into its own place.
Makes the whole series easier to review and ends smaller changes.
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/1530>
For the upcoming conversion of perfmon queries to the acc query
framework we need a way to tell that the data is not ready.
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/1530>
Saves us from doing an extra flush in !wait case and seems more
logical now. Also evaluate etna_bo_cpu_prep(..) retun value.
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/1530>
Alyssa Rosenzweig
Rhys Perry
Rob Clark
Jason Ekstrand
Marek Olšák
Mauro Rossi
Hyunjun Ko
Tapani Pälli
Vasily Khoruzhick
Christian Gmeiner