These are intended to make sure that the post-RA optimizer works
correctly across control flow. The new tests emit a divergent
if-else branch (with full logical+linear CFG).
- dpp_across_cf:
Simple case of DPP optimizable across control flow. Should pass.
- dpp_across_cf_overwritten:
Similar case but the DPP source register is overwritten in CF.
This shows a bug so the test fails now (will be fixed).
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18488>
This commit adds an extra check on the emitted ppp state by
adding an EMIT_MASK.
The mask is just a copy of the header at the beginning of the emit.
Each time a word is emitted the appropriate bit in the mask is
cleared and at the end we make sure that the mask is 0. If not,
we forgot the either clear a bit somewhere or emit some words.
This is intended to make it easier to find such errors.
Signed-off-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18631>
We assume that the texturestate size is 0 so we don't emit a state
update for pds_state_ptr2 since the hw wouldn't use it based on
the 0 size. This commit adds an assert to make sure of that.
Signed-off-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18631>
This commit appends dbsc commands to be executed in secondary
command buffers on vkCmdExecuteCommands().
Signed-off-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18631>
Use pvr_csb_write_value() and pvr_csb_write_struct() to have some
extra checks when writing pre-packed command/state words in
pvr_emit_ppp_state().
Signed-off-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18631>
This commit adds two helper macros to write control stream words
into raw buffers with some extra checks.
The new macros are renamed versions of the previous CS_WRITE() and
CS_PACK_WRITE() macros.
Signed-off-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18631>
Previously `struct pvr_emit_state` was used to keep track of which
state update required emitting and the header was setup based on that.
This commit removes it so that we're instead setting up the header as
we go and we can emit it directly instead of decoding it based on the
emit_state (which is almost identical but with a few missing fields).
This commit also changes pvr_emit_ppp_state() so that each control
stream word is packed/emitted on its own instead of everything
being nested within the header pack. Making the code a bit easier
to follow.
Signed-off-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18631>
This is actually not necessary and generates a lot of superfluous
instructions at every phi (setting the value to zero).
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18566>
Prevent nir_lower_scratch accidentally turning these dead variables
into scratch. This can especially happen to arrayed outputs of
eg. tess control or mesh shaders, which become large shader temps.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18566>
This was accidentally left at the maximum possible value.
However I now tested this with the cadscene demo app and there is
hardly any benefit from going above 64. Set it to 64 for now.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18566>
Let's enable it and see what happens. This should mostly be a win.
Reviewed-by: Mihai Preda <mhpreda@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Acked-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18223>
The test can't pass because glXSwapBuffers now executes GL functions,
which the test doesn't expect. It's a test defect IMO.
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18223>
glthread decreases performance for DeusEx:MD by 7.8% and F1 2015 by 5.7%.
I think that other Feral games are affected similarly.
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18223>
glthread unmarshalling thread cannot run concurrently with code that alloc
the back bo or the code that perform the swaps.
Ensure this by running dri2_flush_drawable_for_swapbuffers early.
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18223>
It's only needed by X11/DRI2. This allows glthread to be used by android,
drm, wayland, and device (EGL backend).
This is the only solution to allow the egl/drm backend to work with glthread
without ugly hacks between libgbm and libEGL.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18223>
GLES1 only has (Multi)Draw{Array,Elements}, but glthread converts them
to the more complicated versions and then calls them through the dispatch,
which generated GL_INVALID_OPERATION.
Luckily, we can export them with the Internal prefix, so they are unlikely
to be used by apps by accident.
Acked-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18223>
If an invalid parameter is received along with an invalid pointer and we
ignore the invalid parameter and dereference the pointer, we crash.
Since we can't fully validate all parameters (such as whether "shader"
is a valid object ID), remove the custom code.
Acked-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18223>
Now that the state for each pipeline is split into pieces, we can mostly
implement it by stitching together the pieces. One TODO is that we could
do more to split up the pre-rast and FS commands into separate draw
states so that we have to emit less commands when fast linking,
currently we compile the variants but delay emitting the commands until
link time, but note that even the Gallium driver doesn't currently do
this. Given the strict SSO model (e.g. with separate VPC registers for
each stage) it may even be possible to do most of the linking ahead of
time with only a few fixups for corner cases.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18554>
Right now, we pass around the push constant state in a lot of places,
but we'll want to add other driver-managed constants. Add a struct which
we can add to, and separate out the total driver-reserved constants from
the size of push constants.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18554>
With pipeline libraries, computing this might have to be delayed because
it depends on multiple pieces of state and there's no way to disentangle
them. Therefore we have to store the requisite state in the pipeline and
combine it later.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18554>
The other state in this register comes from the FS, so determine whether
dual-source blending is enabled from the FS too, in order to avoid
entangling FS state and output state that with graphics pipeline library
can be in different pipelines.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18554>
If there are disabled attachments that are nevertheless written to, they
should already be disabled in the blend state, via the same mechanism as
color_write_enable, so there should be no reason to mix the FS state and
blend state like this. This helps unentangle state that in graphics
pipeline library can be provided in different pipelines.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18554>
This means it has to be rederived rather than passed from the place
where we actually do the optimization, but it eliminates the clutter of
having to pass it around in turnip, which will only get worse with
graphics pipeline libraries.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18554>
