Fixes excessive (and failing) spilling in dEQP-GLES31.functional.ssbo.layout.*
shader-db results are a toss up (I suspect we'd see better results if we
tracked register pressure directly):
total instructions in shared programs: 161377 -> 161377 (0.00%)
total nops in shared programs: 121159 -> 121203 (0.04%)
nops in affected programs: 1839 -> 1883 (2.39%)
Nops are HURT.
total clauses in shared programs: 31604 -> 31606 (<.01%)
clauses in affected programs: 38 -> 40 (5.26%)
Inconclusive result (value mean confidence interval includes 0).
total quadwords in shared programs: 130847 -> 130845 (<.01%)
quadwords in affected programs: 1246 -> 1244 (-0.16%)
Inconclusive result (value mean confidence interval includes 0).
total loops in shared programs: 18 -> 18 (0.00%)
total spills in shared programs: 705 -> 705 (0.00%)
total fills in shared programs: 1645 -> 1645 (0.00%)
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9164>
How many times can I break such a small pass?
Fixes: a805d999c0 ("pan/bi: Fix jumps to terminal block again")
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9137>
Here's another edge case: there could be instructions in the last block
after NIR->BIR but they could be optimized out by backend DCE, causing
the block to become a terminal block.
Noticed while toying with geometry shaders.
Fixes: a805d999c0 ("pan/bi: Fix jumps to terminal block again")
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9137>
Otherwise we get a bad RA if RT 0 = RT 3 (for example), fixes
dEQP-GLES31.functional.shaders.opaque_type_indexing.sampler.const_literal.fragment.sampler2d
Fixes: a6f1500bed ("pan/bi: Workaround BLEND precolour with explicit moves")
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9105>
Passes the relevant tests of
dEQP-GLES31.functional.shaders.opaque_type_indexing.sampler.*, a few
failures that seem to relate to MRT instead of this.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9105>
Needs to support non-blend shader operation (conversion descriptor
sourced from a sysval), as well as MRT. Fixes fbfetch on Bifrost.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9105>
Don't hardcode the RT to 0. Affects ES3.0 which already exposes MRT --
despite no dEQP coverage of this particular corner case, apps could hit
this in the wild on 21.0. Fixes
dEQP-GLES31.functional.draw_buffers_indexed.overwrite_indexed.common_blend_func_buffer_blend_func
Fixes: c7e1ef7c0c ("panfrost: Advertise ES3.0 on Bifrost")
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9105>
All the same formula: calculate an address, emit a pseudoinstruction for
the atomic, emit a postprocess that can be DCE'd if not needed.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Icecream95 <ixn@disroot.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9105>
Multiply by the destination bit size to get the number of bits to
load instead of assuming 32 bits.
Fixes: 2e57684d2d ("pan/bi: Implement load_ubo with the builder")
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9071>
While at it, rework the code to avoid copies between intermediate
structures: the pan_shader_info is passed to the compiler context so
the compiler can fill shader information directly.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8963>
So we can re-use the panfrost_sysvals definition outside of the
compiler without dragging the sysval_to_id hash table.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8963>
Conceptually same lowering as the DDK, although we're missing a number
of relevant compiler optimizations so the generated code is awful.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8774>
Lowered to #(sz - 1) - clz(x), taking advantage of the machine's
8-bit and 16-bit variants of clz and the widening on the second argument
of ISUB to implement neatly in two instructions.
Note that in NIR, ufind_msb can take any integer type but always output
i32.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8774>
Fixes INSTR_INVALID_ENC fault on dEQP-GLES31.functional.compute.basic.empty
Fixes: bfcdc8f174 ("pan/bi: Add some zero bytes after shaders on Bifrost")
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9011>
New scheduler broke this. We need to shuffle some code around so we do
the lower pre-schedule instead of post-schedule (no clauses to work
with).
Fixes: 77933d16d8 ("pan/bi: Switch to new scheduler")
Reported-by: Icecream95 <ixn@disroot.org>
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9011>
There are blit shader specific optimizations available.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9002>
Now that we lower FAU correctly, we don't need to write the extra move
explicitly, it will be lowered in later.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8973>
Move and reshape bi_lower_fau to bi_schedule.c. This generalizes the
pass for FAU reads, allowing copyprop to work with FAU without problems.
The pass must run immediately before scheduling. Its post-conditions are
directly specified as the scheduler's pre-conditions. It momentarily
will depend on internal scheduler predicates. It is, for all intents and
purposes, part of the scheduler. Keep it all together.
Finally, adjust the 0 handling to avoid a move at the expense of
constrained scheduling of something like `FADD.v2f16.clamp_0_1 u0, #0`
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8973>
This makes UBO 0 less special, allowing us to generalize uniform
optimization. Note this disables RMU on Midgard as we're about to
rewrite the RMU mechanism.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8973>