Adapted from NIR's liveness analysis. This is different from our non-SSA
liveness pass for a few reasons:
1. It must handle phi nodes. This implies significant changes to the worklist
algorithm.
2. It only handles SSA. It doesn't need funny labelling schemes for
handling nir_registers in parallel with SSA defs.
3. It is scalar-only. The vector liveness information isn't interesting when
vectors are handled via COLLECT and SPLIT. This means it uses a bitset (uses
8x less memory to store livenss information, should be easier on the caches
too).
Eventually, this will become our only pre-RA liveness pass. For now, both passes
are maintained in parallel: the SSA pass used before out-of-SSA, the non-SSA
pass used after out-of-SSA and before RA, and the post-RA pass used after RA.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
Pre-RA scheduling can move instructions around in nontrivial ways, yet it must
maintain the IR's ordering invariants (for preloads, phis, etc.) Running
validation before and after makes scheduler bugs more obvious.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
We can get BRANCHZ.i16 since we lower JUMP early. This seems to have worked
before mainly by chance. With the change to how we optimize, we can get code
sequences like:
block2 {
BRANCHZ.i16.eq u256, u256 -> block5
BRANCHZ.i16.eq u256, u256 -> block4
} -> block5 from block1
which would choke the assert.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
The new SSA-based dead code elimination won't know how to deal with code of that
funny form. In actuality it doesn't have to, so we just make sure the optimizer
tests produce valid IR so this works as expected.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
Now that we have accurate source/dest counts, we don't need to worst case. Which
is good, because the worst case will become infinite once we allow phis.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
This is yet another case where we add a source, which will require reallocation.
It's easy enough to rebuild the instruction here.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
..where necessary to avoid appending sources. This removes a bunch more cases
where we would need to reallocate sources. The lifetime management is simpler if
we just reallocate the whole instruction. In theory this is a slight overhead
but it's probably worth it to avoid the complexity otherwise.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
When we add a source, we need to build a new instruction (or at the very least
reallocate sources). This is less of a hack anyway.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
Changing I->nr_srcs or I->nr_dests directly is generally unsafe, but the special
case of removing sources/destinations at the end is safe. Add and use helpers to
wrap this operation simplifying the remaining code audit before we can
dynamically allocate sources/destinations.
At this point in the series, nothing modifies I->nr_dests after allocation
except these helpers, so destinations should be safe to make dynamic. There's a
bit more work needed for sources.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
This avoids the nr_srcs/nr_dests setting dance, and will allow the builder to
handle the required memory allocation when we switch to dynamic src/dest
allocation.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
This will allow pseudo instructions to be allocated in a cleaner way when we
dynamically allocate their sources/destinationa.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
We guarantee this now, no need to check it in every pass. There's an exception
on Bifrost after clause scheduling, but few passes run after clause scheduling
so this doesn't affect much.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
There are two "shapes" of TEXC in the IR:
* Regular texturing. This TEXC writes a single set of staging registers.
* Dual texturing. This TEXC writes two sets of staging registers.
Currently we model both with a 2-destination TEXC, with a null second
destination for the usual case where dual texturing isn't used. This is awkward.
To make the "shapes" of instructions more predictable, make TEXC only write a
single set of staging registers (like the hardware instruction) and split off a
TEXC_DUAL pseudoinstruction for the second case, lowered late.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
This ensures the shape of the split is correct (no nulls in the destination).
This should not affect the register allocation, though it does "cost" extra
variable names.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
Removing dests without removing instructions only makes sense for certain
pseudo-instructions, but it makes the IR needlessly complicated for all
instructions. There's no real reason to do so, we can signal this in a different
way instead.
No shader-db changes.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
On Bifrost, BLEND writes to the link register, acting like a function call. On
Valhall, BLEND does not write anything. But the BLEND instruction in our IR is a
pseudo-instruction with Bifrost semantics, expanding to a multi-instruction
sequence on Valhall. So it's not worth special casing Valhall in instruction
selection.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
Use "r0" style to denote machine registers, consistent with the assembly. Use
instead "nr0" for NIR registers so they don't collide. This reduces noise for
the average case (either register allocated or almost SSA).
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
Now we can trust in I->nr_dests, avoiding a bunch of special cases and incorrect
printing for instructions without a destination. To compensate for those
instructions being hard to scan now, indent.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
Now that we set nr_srcs/nr_dests accurately, assert as much in the validator.
This pass will be deleted later in the series, but having it here is expected to
be useful for bisection, in case there are cases missed. Certainly running the
CTS at this point in the series is helpful to prove completeness of the
beginning of this series.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
ATEST on Valhall takes atest-param as an explicit encoded source. The source
also exists on Bifrost, but it's implied in the encoding. Rather than treating
it as a "ghost" source, model it in the IR explicitly so the count is correct.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
The builder is the primary producer of instructions, and generally the shape of
an instruction is fixed at build-time. It must set nr_dests/nr_srcs
appropriately. Likewise, when we modify sources later, we need to update
nr_srcs/nr_dests to keep everything consistent (and keep the tests passing).
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
The different combine modes form different instructions from each other and in
particular from the two-source version on Bifrost. Model them as such so we can
represent the relevant Valhall-specific lowering/optimizations accurately in the
compiler.
This requires updating the unit tests to use the new names since there's not
much point keeping around the aliases.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
Valhall-style comparisons are 3-source, allowing multiple comparisons to be
combined without bitwise arithmetic. For example, the sequence
FCMP.f32.eq t, z, w
FCMP_AND.f32.eq dest, x, y, t
calculates (z == w) && (y == t) in 2 instructions (would be 3 on Bifrost).
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
If an instruction has multiple destinations and *any* of them are needed by
helper invocations, we should keep helper invocations alive. This is a bug fix.
Consider the GLSL:
first = texture(sampler, ...);
float res = texture(sampler, vec2(first.y)).x + first.x;
Corresponding to the IR:
first = ...
x, y, z, w = SPLIT first
second = TEX y, y
x', y', z', w' = SPLIT second
FADD res, x, x'
Here, x is not required by helper invocations (the coordinates to TEX) while y
is required. If we only look at only the first destinations, we incorrectly
decide that first is not required and fail to set the .skip bit, leading to
incorrect results.
Fixes: 5febeae58e ("pan/bi: Emit collect and split")
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
...with dual source blending. Fixes shaders/dolphin/smg.1.shader_test
There are more IR sources than Valhall machine sources here.
Fixes: b48933d641 ("pan/va: Include BLEND for va_swap_12")
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17794>
../../src/gallium/auxiliary/hud/hud_context.c:1017:1: error: static declaration of 'access' follows non-static declaration
These error will shown when any header #include <io.h>
Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18333>
The GLSL code sequence:
texture2D(tex0, u_coords) + texture2D(tex1, u_coords)
will be optimized to
TEXC_DUAL tex0/tex1, u_coords, #texture_descriptor
If this optimization happens after lowering FAU, the resulting TEXC instruction
is unschedulable: both the uniform and the constant descriptor fight for the
same FAU slot.
However, if this optimization happens before lowering FAU, then the FAU lowering
will move the descriptor into a register, complicating the dual texturing fixup
in RA.
To fix this interaction, fuse dual texturing before lowering FAU and keep
texture descriptors as constants when lowering FAU of TEXC.
Fixes scheduling failure in piglit drawoverhead -test 3 with uniform reordering.
Fixes: a4d3a29647 ("pan/bi: Enable dual texture fusing pass")
Fixes: 6b2eda6b72 ("pan/bi: Reorder pushed uniforms to avoid moves")
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18378>
