This lets us write the Z directly from the FTOI for computed Z, and may
let us coalesce color writes in the future.
No change in my shader-db, but clearly drops an instruction in piglit's
early-z test.
It wasn't correctly flagged everywhere, and QPU generation now handles the
only remaining case that was paying attention to it.
No change on shader-db.
Now that the field exists in the instruction, we can make discards less
special. As a bonus, that means that we should be able to merge some more
.sf instructions together when we get around to that.
This causes some scheduling changes, as it allows tlb_color_reads to be
delayed past the discard condition setup. Since the tlb_color_read ends
up later, this may mean performance improvements, but I haven't tested.
total instructions in shared programs: 78114 -> 78035 (-0.10%)
instructions in affected programs: 1922 -> 1843 (-4.11%)
total estimated cycles in shared programs: 234318 -> 234329 (0.00%)
estimated cycles in affected programs: 8200 -> 8211 (0.13%)
The register allocator doesn't really do anything about the temp, so it
doesn't seem like it should matter. However, the scheduler would think
that a new def is being created.
This doesn't change anything yet, but it avoids a bunch of regressions in
the next commit.
This is done instead of copy propagating the VPM reads into the
instructions using them, because VPM reads have to stay in order.
shader-db results:
total instructions in shared programs: 78509 -> 78114 (-0.50%)
instructions in affected programs: 5203 -> 4808 (-7.59%)
total estimated cycles in shared programs: 234670 -> 234318 (-0.15%)
estimated cycles in affected programs: 5345 -> 4993 (-6.59%)
Signed-off-by: Varad Gautam <varadgautam@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Tested-by: Rhys Kidd <rhyskidd@gmail.com>
The two consumers want to know that the destination will be exactly the
source, which is not true if we might not set the destination.
Signed-off-by: Eric Anholt <eric@anholt.net>
I'm moving away from QIR being SSA (since NIR is doing lots of SSA
optimization for us now) and instead having QIR just be QPU operations
with virtual registers. By making our SELs be composed of two MOVs, we
could potentially coalesce the registers for the MOV's src and dst and
eliminate the MOV.
total instructions in shared programs: 88448 -> 88028 (-0.47%)
instructions in affected programs: 39845 -> 39425 (-1.05%)
total estimated cycles in shared programs: 246306 -> 245762 (-0.22%)
estimated cycles in affected programs: 162887 -> 162343 (-0.33%)
If you want the SF of the value of a register produced from a series of
packing MOVs or conditional MOVs, we can't just SF on the last MOV into
the register.
The caller isn't going to expect it from a return, so it would probably
get misinterpreted. If the caller had an unpack in its reg, that's fine,
but don't lose track of it.
This includes GL_SAMPLE_COVERAGE, GL_SAMPLE_ALPHA_TO_ONE, and
GL_SAMPLE_ALPHA_TO_COVAGE.
I haven't implemented a dithering function yet, and gallium doesn't give
me a good chance to do so for GL_SAMPLE_COVERAGE.
It looks like nir_lower_idiv is going to use it soon, so add support.
With Ilia's change, this fixes one case in fs-op-div-large-uint-uint (with
GL 3.0 forced on).
Cc: "11.0" <mesa-stable@lists.freedesktop.org>
This paves the way for copy propagating our unpacks. We end up with a
small change on shader-db:
total instructions in shared programs: 89390 -> 89251 (-0.16%)
instructions in affected programs: 19041 -> 18902 (-0.73%)
which appears to be because we no longer convert MOVs for an FMAX dst,
r4.unpack, r4.unpack (instead of the previous MOV dst, r4.unpack), and
this ends up with a slightly better schedule.
(originally part of previous patch, split out to separate patch by Rob)
v2: squash in some fixes from Eric
v3: Another fix from Eric for point coords.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
This avoids generation of undefined packing in qir and qpu instructions,
fixing a lot of rendering errors.
Fixes 8b36d107fd (vc4: Pack the unorm-packing bits into a src MUL
instruction when possible.)
Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Boyan Ding <boyan.j.ding@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Now that we do non-SSA QIR instructions, we can take a NIR SSA src that's
only used by the unorm packing and just stuff the pack bits into it.
total instructions in shared programs: 98136 -> 97974 (-0.17%)
instructions in affected programs: 4149 -> 3987 (-3.90%)
This helps ensure that the register allocator doesn't force the later pack
operations to insert extra MOVs.
total instructions in shared programs: 98170 -> 98159 (-0.01%)
instructions in affected programs: 2134 -> 2123 (-0.52%)
Now that we have NIR, most of the optimization we still need to do is
peepholes on instruction selection rather than general dataflow
operations. This means we want to be able to have QIR be a lot closer to
the actual QPU instructions, just with virtual registers. Allowing
multiple instructions writing the same register opens up a lot of
possibilities.
Previously, SFU values always moved to a temporary, and TLB color reads
and texture reads always lived in r4. Instead, we can have these results
just be normal temporaries, and the register allocator can leave the
values in r4 when they don't interfere with anything else using r4.
shader-db results:
total instructions in shared programs: 100809 -> 100040 (-0.76%)
instructions in affected programs: 42383 -> 41614 (-1.81%)
NIR brings us better optimization than I would have bothered to write
within the driver, developers sharing future optimization work, and the
ability to share device-specific lowering code that we and other
GLES2-level drivers need.
total uniforms in shared programs: 13421 -> 13422 (0.01%)
uniforms in affected programs: 62 -> 63 (1.61%)
total instructions in shared programs: 39961 -> 39707 (-0.64%)
instructions in affected programs: 15494 -> 15240 (-1.64%)
v2: Add missing imov support, and assert that there are no dest saturates.
v3: Rebase on the target-specific algebraic series.
v4: Rebase on gallium-includes-from-NIR changes in mater.
v5: Rebase on variables being in lists instead of hash tables.
v6: Squash in intermediate changes that used the NIR-to-TGSI pass (which
I'm not committing)
This cleans up some pointless operations generated by the in-driver mul24
lowering (commonly generated by making a vec4 index for a matrix in a
uniform array).
I could fill in other operations, but pretty much anything else ought to
be getting handled at the NIR level, I think.
total uniforms in shared programs: 13423 -> 13421 (-0.01%)
uniforms in affected programs: 346 -> 344 (-0.58%)
This lets us more intelligently decide which uniform values should be put
into temporaries, by choosing the most reused values to push to temps
first.
total uniforms in shared programs: 13457 -> 13433 (-0.18%)
uniforms in affected programs: 1524 -> 1500 (-1.57%)
total instructions in shared programs: 40198 -> 40019 (-0.45%)
instructions in affected programs: 6027 -> 5848 (-2.97%)
I noticed this opportunity because with the NIR work, some programs were
happening to make different uniform copy propagation choices that
significantly increased instruction counts.
Right now the places that used to emit a mov.sf just put the SF on the
previous instruction when it generated the source of the SF value. Even
without optimization to push the sf up further (and kill thus potentially
kill more MOVs), this gets us:
total uniforms in shared programs: 13455 -> 13457 (0.01%)
uniforms in affected programs: 3 -> 5 (66.67%)
total instructions in shared programs: 40296 -> 40198 (-0.24%)
instructions in affected programs: 12595 -> 12497 (-0.78%)
I'm using this in some WIP commits for doing blending in 8888 instead of
vec4. But it also gives us these results immediately, thanks to allowing
more uniforms/immediates in the arguments:
total instructions in shared programs: 41027 -> 40960 (-0.16%)
instructions in affected programs: 4381 -> 4314 (-1.53%)
Small immediates have the downside of taking over the raddr B field, so
you might have less chance to pack instructions together thanks to raddr B
conflicts. However, it also reduces some register pressure since it lets
you load 2 "uniform" values in one instruction (avoiding a previous load
of the constant value to a register), and increases some pairing for the
same reason.
total uniforms in shared programs: 16231 -> 13374 (-17.60%)
uniforms in affected programs: 10280 -> 7423 (-27.79%)
total instructions in shared programs: 40795 -> 41168 (0.91%)
instructions in affected programs: 25551 -> 25924 (1.46%)
In a previous version of this patch I had a reduction in instruction count
by forcing the other args alongside a SMALL_IMM to be in the A file or
accumulators, but that increases register pressure and had a bug in
handling FRAG_Z. In this patch is I just use raddr conflict resolution,
which is more expensive. I think I'd rather tweak allocation to have some
way to slightly prefer good choices for files in general, rather than risk
failing to register allocate by forcing things into register classes.