This new pass is for lowering explicitly laid out memory coming in from
SPIR-V or a similar source.  It works differently because it computes
addresses on-the-fly so that it can properly handle casts and addresses
going through phi webs and the like.

This commit adds a new dereference type: ptr_as_array as well as new
explicit stride and offset parameters that can be used to for external
memory access such as a UBO or SSBO dereference.

We already detect any incomplete deref chains (where the deref is used
for something other than another deref or a load/store) and flag the
variable as used thanks to deref_used_for_not_store.  All that's left to
do is to properly skip casts when cleaning up.

This commit adds a new num_components value for intrinsic sources of -1
which means that it consumes everything and the number of components
effectively isn't validated.  This is useful for deref sources which
just take the result of the deref and we leave it up to the driver to
decide what that size should be.

This simplifies our deref handling by emitting the actual NIR deref
instructions on-the-fly instead of of building up a deref chain and then
emitting them at the last moment.  In order for this to work with the
parts of the compiler that assume they can chase deref chains, we have
to run nir_rematerialize_derefs_in_use_blocks_impl to put the derefs
back in the right places in cases like loops where dominance may not
quite work out the way SPIR-V thinks it does.

Pipeline state pending bits should be taken into account when copying
results.

In the particular bug below, the results of the
vkCmdCopyQueryPoolResults() command was being overwritten by the
preceding vkCmdCopyBuffer() with a same destination buffer. This is
because we copy the buffers using the 3D pipeline whereas we copy the
query results using the command streamer. Those pieces of HW work in
parallel and the results are somewhat undefined.

v2: Unconditionally flush the pipeline before copying the results
    (Jason)

v3: Wrap & expressions (Jason)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Suggested-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108894
Cc: mesa-stable@lists.freedesktop.org

I didn't mean to push this. I don't think it makes any difference.

This reverts commit f737fe00.

The calculated length of a line may be infinite, if the coords we
get are bogus. This leads to an infinite loop in line stippling.
To prevent this test for this explicitly (although technically
on at least x86 sse it would actually work without the explicit
test, as long as we use the int-converted length value).
While here also get rid of some always-true condition.

Note this does not actually solve the root cause, which is that
the coords we receive are bogus after clipping. This seems a difficult
problem to solve. One issue is that due to float arithmetic, clip w
may become 0 after clipping if the incoming geometry is
"sufficiently degenerate", hence x/y/z ndc (and window) coords will
be all inf (or nan). Even with w not quite 0, I believe it's possible
we produce values which are actually outside the view volume.
(Also, x=y=z=w=0 coords in clipspace would be not considered subject
to clipping, and similarly result in all NaN coords.) We just hope for
now other draw stages (and rasterizers) can handle those relatively
safely (llvmpipe itself should be sort of robust against this, certainly
converstion to fixed point will produce garbage, it might fail a couple
assertions but should neither hang nor crash otherwise).

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>

Józef Kucia authored Nov 26, 2018 and Lionel Landwerlin committed Nov 29, 2018

Signed-off-by: Józef Kucia <joseph.kucia@gmail.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

Update to the internal master as of 2018-11-15.

This has a lot of gratuitous whitespace change, but on the plus
side it's built using the same tooling that's used for AMDVLK,
which should help going forward.

Our choices here are simply redundant as long as sin.flags is set
correctly.

(v2:
- remove unused function parameter)
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

v2:
- use SI_CONTEXT_REG_OFFSET

Reviewed-by: Dave Airlie <airlied@redhat.com>

The ioctl.cap_3d member was never freed.

Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Sinclair Yeh <syeh@vmware.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

Free the context after destruction.

Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Sinclair Yeh <syeh@vmware.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

Viewport/scissor don't need to be updated for array textures.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

If all layers are bound we can perform a fast color or depth clear
instead of iterating over all layers. This has the advantage
to avoid trashing the framebuffer for nothing if you we end up by
doing a fast clear when calling radv_clear_image_layer(), and
clearing all layers in one shot is obviously faster.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

Currently only true if RADV_PERFTEST=dccmsaa is set.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

For further optimisations.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

The driver doesn't support DCC/CMASK for mipmapped textures.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

Fix build error.

  CXXLD    pipe_msm.la
../../../../src/gallium/drivers/freedreno/.libs/libfreedreno.a(freedreno_batch.o): In function `batch_init':
src/gallium/drivers/freedreno/freedreno_batch.c:54: undefined reference to `fd_device_version'
src/gallium/drivers/freedreno/freedreno_batch.c:59: undefined reference to `fd_submit_new'
src/gallium/drivers/freedreno/freedreno_batch.c:61: undefined reference to `fd_submit_new_ringbuffer'
src/gallium/drivers/freedreno/freedreno_batch.c:64: undefined reference to `fd_submit_new_ringbuffer'
src/gallium/drivers/freedreno/freedreno_batch.c:66: undefined reference to `fd_submit_new_ringbuffer'
src/gallium/drivers/freedreno/freedreno_batch.c:70: undefined reference to `fd_submit_new_ringbuffer'

Fixes: b4476138 ("freedreno: move drm to common location")
Fixes: aa0fed10 ("freedreno: move ir3 to common location")
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Rob Clark <robdclark@gmail.com>

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

Acked-by: Christian König <christian.koenig@amd.com>

Window system buffers didn't use the optimal alignment.

Acked-by: Christian König <christian.koenig@amd.com>

Acked-by: Christian König <christian.koenig@amd.com>

Sadly, the 3 games I tested (DeusEx:MD, DiRT Rally, DOTA 2) are unaffected
by the overallocation, because I guess their buffers don't fall into
the small range below a power-of-two size.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>