Commit graph

1048 commits

Author SHA1 Message Date
Alyssa Rosenzweig
5bb99b1104 agx: add helper to pad 16-bit to 32-bit
this is a bit tricky to get right

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31532>
2024-10-05 18:30:12 +00:00
Alyssa Rosenzweig
92d7ecb98c agx: fix spilling around exports
need to hoist the reloads.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31532>
2024-10-05 18:30:12 +00:00
Alyssa Rosenzweig
eae4c50be9 agx: add spilling debug helpers
I keep typing these

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31532>
2024-10-05 18:30:12 +00:00
Alyssa Rosenzweig
4ca52cde7a agx: increase lower vars to scratch thresh
should reduce silliness.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31532>
2024-10-05 18:30:12 +00:00
Alyssa Rosenzweig
8f3caec5dc agx: handle hw tess eval shaders
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31532>
2024-10-05 18:30:11 +00:00
Alyssa Rosenzweig
9b3216f61b agx: clean up packs from bounds check
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31532>
2024-10-05 18:30:11 +00:00
Alyssa Rosenzweig
5b86a55f10 agx: optimize robust access
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31532>
2024-10-05 18:30:11 +00:00
Alyssa Rosenzweig
6287346482 agx: handle bounds_agx
fallback when optimization fails.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31532>
2024-10-05 18:30:11 +00:00
Alyssa Rosenzweig
169acb1170 agx: clarify spill lowering math
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31532>
2024-10-05 18:30:11 +00:00
Alyssa Rosenzweig
e16ad50bfa agx: allocate scratch in spilling unit test
for assertion

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31532>
2024-10-05 18:30:11 +00:00
Alyssa Rosenzweig
2600224f77 agx: clarify scratch size units
bytes, not words.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31532>
2024-10-05 18:30:11 +00:00
Alyssa Rosenzweig
d21bc62f4d agx: fix =spill
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31532>
2024-10-05 18:30:11 +00:00
Alyssa Rosenzweig
4e3a5f0e60 agx: fix spilling in i686 builds
../src/asahi/compiler/agx_pack.c: In function ‘agx_pack_instr’:
../src/asahi/compiler/agx_pack.c:1074:38: warning: left shift count >= width of type [-Wshift-count-overflow]
 1074 |          ((uint64_t)q4 << 42) | (1UL << 47) | // XXX
      |                                      ^~
../src/asahi/compiler/agx_pack.c:1110:43: warning: left shift count >= width of type [-Wshift-count-overflow]
 1110 |          ((uint64_t)i5 << 44) | (L ? (1UL << 47) : 0) |
      |                                           ^~
../src/asahi/compiler/agx_pack.c:1131:64: warning: left shift count >= width of type [-Wshift-count-overflow]
 1131 |          ((uint64_t)((I->stack_size >> 4) & 0xF) << 32) | (1UL << 47) | // XXX

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31532>
2024-10-05 18:30:11 +00:00
Alyssa Rosenzweig
cd66a6813b agx: plumb COHERENT
set the magic caching bits. this fixes memory model fails on g13d.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31532>
2024-10-05 18:30:10 +00:00
Alyssa Rosenzweig
d4e34bc829 agx: pull out the big hammers for device-barriers
g13d.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31532>
2024-10-05 18:30:10 +00:00
Alyssa Rosenzweig
f977c52b84 ail: swallow up formats
ail is a more sensible place for the format tables to live. this does create a
bit of dependency soup but hey.

nfc

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30981>
2024-09-02 23:27:14 +00:00
Alyssa Rosenzweig
06a48375c8 agx: fix wait_pix scheduling with eMRT
unclear how this didn't blow up before.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30981>
2024-09-02 23:27:14 +00:00
Alyssa Rosenzweig
05d535cb87 agx: support bindless block image store
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30981>
2024-09-02 23:27:14 +00:00
Alyssa Rosenzweig
afc7557cb6 nir,agx: make block image store an image() intrinsic
so we can do a bindless version

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30981>
2024-09-02 23:27:14 +00:00
Alyssa Rosenzweig
f690592692 agx: lower exact frcp
the hardware is off by 1 ULP. fixes nir_lower_idiv brokenness.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30633>
2024-08-12 19:11:23 -04:00
Alyssa Rosenzweig
e3e82f7a6c agx: add a comment about an unknown bit
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30633>
2024-08-12 19:11:23 -04:00
Alyssa Rosenzweig
8b29f0a762 agx: handle explicit coordinate tib instructions
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30633>
2024-08-12 19:11:23 -04:00
Alyssa Rosenzweig
f706079dc5 agx: allow st_tile in compute shaders
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30633>
2024-08-12 19:11:23 -04:00
Alyssa Rosenzweig
9b39c5427f agx: plumb imageblock stride
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30633>
2024-08-12 19:11:23 -04:00
Alyssa Rosenzweig
290820e8b0 agx: fix block_image_store indirect ts
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30633>
2024-08-12 18:46:31 -04:00
Alyssa Rosenzweig
c50e0934e7 agx: let if-fusing opts interact
total instructions in shared programs: 2115031 -> 2114800 (-0.01%)
instructions in affected programs: 46937 -> 46706 (-0.49%)
helped: 147
HURT: 30
Instructions are helped.

total alu in shared programs: 1669893 -> 1669655 (-0.01%)
alu in affected programs: 38134 -> 37896 (-0.62%)
helped: 147
HURT: 28
Alu are helped.

total fscib in shared programs: 1666017 -> 1665779 (-0.01%)
fscib in affected programs: 38134 -> 37896 (-0.62%)
helped: 147
HURT: 28
Fscib are helped.

total bytes in shared programs: 14059380 -> 14057364 (-0.01%)
bytes in affected programs: 306294 -> 304278 (-0.66%)
helped: 147
HURT: 28
Bytes are helped.

total regs in shared programs: 656483 -> 656491 (<.01%)
regs in affected programs: 257 -> 265 (3.11%)
helped: 3
HURT: 4
Inconclusive result (value mean confidence interval includes 0).

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30633>
2024-08-12 18:46:31 -04:00
Alyssa Rosenzweig
64239ac12c agx: fuse not into if
total instructions in shared programs: 2115466 -> 2115031 (-0.02%)
instructions in affected programs: 68776 -> 68341 (-0.63%)
helped: 222
HURT: 5
Instructions are helped.

total alu in shared programs: 1670328 -> 1669893 (-0.03%)
alu in affected programs: 54224 -> 53789 (-0.80%)
helped: 222
HURT: 5
Alu are helped.

total fscib in shared programs: 1666452 -> 1666017 (-0.03%)
fscib in affected programs: 54224 -> 53789 (-0.80%)
helped: 222
HURT: 5
Fscib are helped.

total bytes in shared programs: 14061990 -> 14059380 (-0.02%)
bytes in affected programs: 458702 -> 456092 (-0.57%)
helped: 222
HURT: 5
Bytes are helped.

total regs in shared programs: 656515 -> 656483 (<.01%)
regs in affected programs: 432 -> 400 (-7.41%)
helped: 7
HURT: 1
Inconclusive result (value mean confidence interval includes 0).

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30633>
2024-08-12 18:46:31 -04:00
Alyssa Rosenzweig
2b1859cd06 agx: tweak jmp_exec_none heuristic
to make sure border colour code is skipped.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30633>
2024-08-12 18:46:31 -04:00
Alyssa Rosenzweig
907fe00da1 agx: handle min_lod sources
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30633>
2024-08-12 18:46:31 -04:00
Alyssa Rosenzweig
03830c9523 agx: optimize popcount(ballot(true))
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30633>
2024-08-12 18:46:31 -04:00
Alyssa Rosenzweig
f04ae930d9 nir,agx: add "active threads in subgroup" intrinsic
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30633>
2024-08-12 18:45:58 -04:00
Alyssa Rosenzweig
dc3288dcb1 agx: exploit soft fault info
for preambles and for peephole selection.

total instructions in shared programs: 2159359 -> 2114124 (-2.09%)
instructions in affected programs: 359763 -> 314528 (-12.57%)
helped: 814
HURT: 6
Instructions are helped.

total alu in shared programs: 1685059 -> 1670200 (-0.88%)
alu in affected programs: 217210 -> 202351 (-6.84%)
helped: 589
HURT: 45
Alu are helped.

total fscib in shared programs: 1681202 -> 1666324 (-0.88%)
fscib in affected programs: 217477 -> 202599 (-6.84%)
helped: 590
HURT: 45
Fscib are helped.

total ic in shared programs: 460856 -> 455502 (-1.16%)
ic in affected programs: 41350 -> 35996 (-12.95%)
helped: 174
HURT: 8
Ic are helped.

total bytes in shared programs: 14302484 -> 14053982 (-1.74%)
bytes in affected programs: 2380614 -> 2132112 (-10.44%)
helped: 814
HURT: 7
Bytes are helped.

total regs in shared programs: 662302 -> 656517 (-0.87%)
regs in affected programs: 26979 -> 21194 (-21.44%)
helped: 432
HURT: 9
Regs are helped.

total uniforms in shared programs: 1651909 -> 1687077 (2.13%)
uniforms in affected programs: 95383 -> 130551 (36.87%)
helped: 17
HURT: 783
Uniforms are HURT.

total threads in shared programs: 20324608 -> 20326592 (<.01%)
threads in affected programs: 16192 -> 18176 (12.25%)
helped: 17
HURT: 3
Threads are helped.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30633>
2024-08-12 18:36:26 -04:00
Alyssa Rosenzweig
f0cd599913 agx: factor out agx_device_key
we will have other device-dependent key members soon.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30633>
2024-08-12 18:36:26 -04:00
Alyssa Rosenzweig
5d06944665 agx: fix clang-format
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30565>
2024-08-08 15:26:08 +00:00
Alyssa Rosenzweig
2282abecec agx: convert to ddx intrinsics
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30565>
2024-08-08 15:26:07 +00:00
Zan Dobersek
7fd5f76393 nir/lower_vars_to_scratch: calculate threshold-limited variable size separately
ir3's lowering of variables to scratch memory has to treat 8-bit values as
16-bit ones when comparing such value's size against the given threshold
since those values are handled through 16-bit half-registers. But those
values can still use natural 8-bit size and alignment for storing inside
scratch memory.

nir_lower_vars_to_scratch now accepts two size-and-alignment functions,
one used for calculating the variable size and the other for calculating
the size and alignment needed for storing inside scratch memory. Non-ir3
uses of this pass can just duplicate the currently-used function. ir3
provides a separate variable-size function that special-cases 8-bit types.

Signed-off-by: Zan Dobersek <zdobersek@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29875>
2024-08-07 14:32:28 +00:00
Alyssa Rosenzweig
c40c723336 agx: use opt_uniform_atomics
Apple does something similar.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30488>
2024-08-06 11:48:18 -04:00
Alyssa Rosenzweig
39e7d06eea agx: add some SRs
the subgroup one seen in metal uniform atomic code, the quad one is by
symmetry.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30488>
2024-08-06 11:48:18 -04:00
Alyssa Rosenzweig
75d07cc3d0 agx: fix ballot extend packing
hit with uniform atomic ops with tessellation.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30488>
2024-08-06 11:48:03 -04:00
Daniel Stone
e05415a82e format: Generate endian-independent format aliases
Instead of having a hardcoded list of endian-independent format aliases
in the header, generate them from the format definitions.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29649>
2024-07-19 13:50:42 +00:00
Alyssa Rosenzweig
e8b673a109 agx: do not flush denorms for fp16 fmin/fmax
total instructions in shared programs: 2164639 -> 2158940 (-0.26%)
instructions in affected programs: 319475 -> 313776 (-1.78%)
helped: 1200
HURT: 6
Instructions are helped.

total alu in shared programs: 1690198 -> 1684653 (-0.33%)
alu in affected programs: 272173 -> 266628 (-2.04%)
helped: 1181
HURT: 6
Alu are helped.

total fscib in shared programs: 1686497 -> 1680797 (-0.34%)
fscib in affected programs: 272922 -> 267222 (-2.09%)
helped: 1200
HURT: 6
Fscib are helped.

total bytes in shared programs: 14334550 -> 14300314 (-0.24%)
bytes in affected programs: 2075546 -> 2041310 (-1.65%)
helped: 1200
HURT: 6
Bytes are helped.

total regs in shared programs: 662332 -> 662302 (<.01%)
regs in affected programs: 1103 -> 1073 (-2.72%)
helped: 14
HURT: 15
Inconclusive result (value mean confidence interval includes 0).

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30075>
2024-07-15 19:29:00 +00:00
Alyssa Rosenzweig
6ac289dade agx: set lower_fminmax_signed_zero
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30075>
2024-07-15 19:29:00 +00:00
Alyssa Rosenzweig
27e3495902 agx: set discard_is_demote
this is simpler/more correct.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29861>
2024-06-22 10:09:45 -04:00
Alyssa Rosenzweig
7dd73290fb agx: add unit test for ballot bug
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29861>
2024-06-22 10:09:45 -04:00
Alyssa Rosenzweig
6628f24e4d agx: fix insidious ballot optimizer bug
see next test.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29861>
2024-06-22 10:09:45 -04:00
Alyssa Rosenzweig
15257b65c6 treewide: use nir_metadata_control_flow
Via Coccinelle patch:

    @@
    @@

    -nir_metadata_block_index | nir_metadata_dominance
    +nir_metadata_control_flow

...plus some manual fixups for call sites missed by coccinelle.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Acked-by: Karol Herbst <kherbst@redhat.com>
Acked-by: Juan A. Suarez Romero <jasuarez@igalia.com> [broadcom]
Acked-by: Vasily Khoruzhick <anarsoul@gmail.com> [lima]
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29745>
2024-06-17 16:28:14 -04:00
Daniel Schürmann
9b1a748b5e nir: remove nir_intrinsic_discard
The semantics of discard differ between GLSL and HLSL and
their various implementations. Subsequently, numerous application
bugs occurred and SPV_EXT_demote_to_helper_invocation was written
in order to clarify the behavior. In NIR, we now have 3 different
intrinsics for 2 things, and while demote and terminate have clear
semantics, discard still doesn't and can mean either of the two.

This patch entirely removes nir_intrinsic_discard and
nir_intrinsic_discard_if and replaces all occurences either with
nir_intrinsic_terminate{_if} or nir_intrinsic_demote{_if} in the
case that the NIR option 'discard_is_demote' is being set.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27617>
2024-06-17 19:37:16 +00:00
Alyssa Rosenzweig
88cdcd8f72 agx: fix fmin/fmax with (-0, 0) pair
We need additional lowering to handle negzero properly. fixes float_controls2
fails but strictly the bug was already present!

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29742>
2024-06-16 10:10:33 -04:00
Alyssa Rosenzweig
ab21d179d6 agx: fix 64-bit bcsel ingestion
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29742>
2024-06-16 10:01:46 -04:00
Alyssa Rosenzweig
761d79ec3e agx: fix indirect CF accounting
if we have a non-compact indirectly indexed array of scalars

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29607>
2024-06-07 16:57:03 +00:00