fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-26 14:38:13 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 2
221427 commits 66 branches 936 tags 1 GiB
Commit graph

50 commits

Author SHA1 Message Date
Alyssa Rosenzweig
28cf0f52c1 jay/to_binary: handle packing accumulators 
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:09 +00:00
Alyssa Rosenzweig
aa37d8b248 jay/print: deal with bare r0 copies 
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:09 +00:00
Kenneth Graunke
e55af8793f jay: Add missing ROR case 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:09 +00:00
Alyssa Rosenzweig
6c862b1951 jay: fix SEL types 
SEL.f32 flushes denorms but SEL.u32 does not. That means changing the type of
the SEL is only justified if we know we're used as a float. This fixes
miscompilation in cases like:

   ieq(1, bcsel(a, fneg(b), c))

Previously we'd be too greedy and form

   (a) SEL.f32 t, -b, c
   cmp.u32 t, 1

But that would inadvertently flush c which is an integer here. So just set the
type based on what we're used as. Some regressions due to is_only_used_as_float
not seeing through phis (..could probably be fixed?).

Totals:
Instrs: 2760796 -> 2761525 (+0.03%); split: -0.06%, +0.08%
CodeSize: 44244128 -> 44222384 (-0.05%); split: -0.13%, +0.08%

Totals from 945 (35.70% of 2647) affected shaders:
Instrs: 1968645 -> 1969374 (+0.04%); split: -0.08%, +0.11%
CodeSize: 31721968 -> 31700224 (-0.07%); split: -0.17%, +0.11%

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:09 +00:00
Alyssa Rosenzweig
b5898a418b jay: relax mov type check 
prevents regression with next patch which turns u32 into s32.

Totals:
Instrs: 2764288 -> 2760796 (-0.13%)
CodeSize: 44299920 -> 44244128 (-0.13%); split: -0.13%, +0.00%

Totals from 193 (7.29% of 2647) affected shaders:
Instrs: 255455 -> 251963 (-1.37%)
CodeSize: 4160400 -> 4104608 (-1.34%); split: -1.34%, +0.00%

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:07 +00:00
Alyssa Rosenzweig
1b648326ac jay: refuse to propagate ADDRESS copies 
at least until we have address RA..

Totals:
Instrs: 2764282 -> 2764288 (+0.00%)
CodeSize: 44299872 -> 44299920 (+0.00%)

Totals from 2 (0.08% of 2647) affected shaders:
Instrs: 4215 -> 4221 (+0.14%)
CodeSize: 67456 -> 67504 (+0.07%)

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:07 +00:00
Alyssa Rosenzweig
56ffad0c3a jay: call DCE an extra time 
Totals:
Instrs: 2767235 -> 2765908 (-0.05%); split: -0.10%, +0.05%
CodeSize: 44349488 -> 44328688 (-0.05%); split: -0.10%, +0.06%

Totals from 347 (13.11% of 2647) affected shaders:
Instrs: 718067 -> 716740 (-0.18%); split: -0.39%, +0.20%
CodeSize: 11626032 -> 11605232 (-0.18%); split: -0.39%, +0.21%

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:06 +00:00
Alyssa Rosenzweig
d85eb51e17 jay/register_allocate: don't depend on indexing 
this can get messed up by optimizations.

Totals:
Instrs: 2768612 -> 2764317 (-0.16%); split: -0.29%, +0.13%
CodeSize: 44367648 -> 44300352 (-0.15%); split: -0.28%, +0.13%

Totals from 867 (32.75% of 2647) affected shaders:
Instrs: 1694745 -> 1690450 (-0.25%); split: -0.47%, +0.22%
CodeSize: 27387648 -> 27320352 (-0.25%); split: -0.46%, +0.21%

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:06 +00:00
Alyssa Rosenzweig
a964f321a5 jay: don't print internal without the flag 
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:06 +00:00
Alyssa Rosenzweig
3a73c76373 jay: fix spiller coupling code 
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:05 +00:00
Alyssa Rosenzweig
cd6c5a2f90 jay: improve spiller debug 
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:05 +00:00
Alyssa Rosenzweig
d637554418 jay: fix simd32 deswizzle 
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:05 +00:00
Alyssa Rosenzweig
f728e3cb05 jay: test logic op fusing 
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:04 +00:00
Alyssa Rosenzweig
698223ccd1 jay/test-optimizer: fuse before/after cases 
new macro to DRY.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:04 +00:00
Alyssa Rosenzweig
99796bff04 jay: fold logic ops 
Totals:
Instrs: 2798036 -> 2784419 (-0.49%); split: -0.58%, +0.10%
CodeSize: 44815024 -> 44614000 (-0.45%); split: -0.56%, +0.11%
Number of fill instructions: 2270 -> 2280 (+0.44%)

Totals from 1298 (49.04% of 2647) affected shaders:
Instrs: 2165338 -> 2151721 (-0.63%); split: -0.75%, +0.13%
CodeSize: 34865440 -> 34664416 (-0.58%); split: -0.72%, +0.15%
Number of fill instructions: 1571 -> 1581 (+0.64%)

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:04 +00:00
Alyssa Rosenzweig
5d22e9d2a5 jay: allow predication of pure-flag instrs 
i.e. compares

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:03 +00:00
Alyssa Rosenzweig
2ab8a614dd jay/register_allocate: tie predicated-defaults 
(if we can)

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:03 +00:00
Alyssa Rosenzweig
d74ada78c0 jay/assign_flags: handle predicated CMP 
the optimizer will generate this soon, so make sure flag RA can deal.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:02 +00:00
Alyssa Rosenzweig
375945ea0b jay/lower_pre_ra: skip predication 
otherwise the assert blows up

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:02 +00:00
Alyssa Rosenzweig
176b9a0f0c jay/opt_dead_code: handle predication 
otherwise we'll get validation splat soon.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:02 +00:00
Alyssa Rosenzweig
f63ab3eea5 jay/register_allocate: use standard builder name 
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:02 +00:00
Alyssa Rosenzweig
f98e2b24bc jay: fix the source pinning code 
I was just trying to get rid of the loop but it also generates better code.

Totals:
Instrs: 2806469 -> 2798036 (-0.30%); split: -0.33%, +0.02%
CodeSize: 44950448 -> 44815024 (-0.30%); split: -0.32%, +0.02%

Totals from 143 (5.40% of 2647) affected shaders:
Instrs: 665554 -> 657121 (-1.27%); split: -1.37%, +0.10%
CodeSize: 10611344 -> 10475920 (-1.28%); split: -1.37%, +0.10%

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:02 +00:00
Alyssa Rosenzweig
ad731766d3 jay: fix SEL implied pipe 
Pretty obscure, but this is not valid:

< (1&W)       sync.nop _.0                                        | $2.dst
< (32&f2.0)   sel.f32 g48, g48, -g40                              | I@7
---
> (32&f2.0)   sel.f32 g48, g48, -g40                              | @7 $2.dst

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:01 +00:00
Alyssa Rosenzweig
9b423bfe94 jay: reduce calloc to malloc when memsetting after 
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:01 +00:00
Alyssa Rosenzweig
b46d677aab jay: reduce zeroing 
this is fully initialized when constructing phi webs anyway.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:01 +00:00
Alyssa Rosenzweig
e20f8ab2b2 jay: generalize last kill code 
Totals:
Instrs: 2815692 -> 2806469 (-0.33%); split: -0.44%, +0.11%
CodeSize: 45100624 -> 44950448 (-0.33%); split: -0.44%, +0.11%

Totals from 1292 (48.81% of 2647) affected shaders:
Instrs: 2427684 -> 2418461 (-0.38%); split: -0.51%, +0.13%
CodeSize: 38993984 -> 38843808 (-0.39%); split: -0.51%, +0.13%

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:00 +00:00
Alyssa Rosenzweig
70bfa005a5 jay: drop dead code 
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:32:00 +00:00
Alyssa Rosenzweig
f39f6ce7ba jay: strengthen assert 
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
 2026-04-20 22:31:59 +00:00
Mauro Rossi
cc44922048 intel/jay: fix static_assert expression
Some checks are pending
macOS-CI / macOS-CI (dri) (push) Waiting to run
Details
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Details 
Fixes the following building error:

FAILED: src/intel/compiler/jay/libintel_compiler_jay.a.p/jay_assign_flags.c.o
...
In file included from ../src/intel/compiler/jay/jay_assign_flags.c:6:
../src/intel/compiler/jay/jay_builder.h:184:24: error: static_assert expression is not an integral constant expression
         static_assert(sizeof(uintptr_t) <= sizeof(uint64_t) &&
                       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1 error generated.

Fixes: e42e3193 ("intel: add Jay")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Alyssa Rosenzweig
1b029f3279 jay: allow cmod on cvt 
it's just a MOV

saves an instruction on dEQP-GLES31.functional.compute.basic.image_atomic_op_local_size_8

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Alyssa Rosenzweig
9df62df55e jay: fix bfn cmod 
affects dEQP-GLES31.functional.compute.basic.image_atomic_op_local_size_8

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Alyssa Rosenzweig
b9f8467855 jay: fix a bunch of opcode properties 
really need a full audit..

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Alyssa Rosenzweig
8afcbbe77d jay: load_simd_width_intel 
dEQP-GLES31.functional.shaders.arrays_of_arrays.es31.array_access.dynamic_expression_access_fragment

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Alyssa Rosenzweig
4eb838eb48 jay: split up jay_from_nir.c 
Big monolithic file, split it up into the relevant pieces.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Alyssa Rosenzweig
6925d9ee23 jay: move deswizzle hack outside of swsb 
this will eventually enable better swsb for the simd32 payload code.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Alyssa Rosenzweig
48a24f3c27 jay: fix instr counts 
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Kenneth Graunke
3a1227f36b jay: Clear default group for quad swizzles 
Quad swizzles should always execute with NoMask and group 0.

We skipped initializing the group, and so inherited whatever the state
from the previous instruction was.  This led to incorrect behavior if
the previous instruction was SIMD split:

   (16)        mov.u32 g84<2>, g2        |
   (16|M16)    mov.u32 g86<2>, g3        |
   (32|M16&W)  mov.u32 g2, g126.2<4,4,0> | I@1

Oops.  The final quad swizzle shouldn't have had M16 set.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Kenneth Graunke
8bfb139033 jay: Make lower_immediates bail if there are no sources 
For example, JAY_OPCODE_PRELOAD has no sources.  Reading src[other]
would read uninitialized data.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Kenneth Graunke
a79931421c jay: Assert that source is not null in jay_copy_strided 
Catch bugs earlier.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Alyssa Rosenzweig
e84cba531b jay: fix simd split swsb bugs 
this is subtle, but the relevant igc:

  // In case of shooting down of this instruction, we need to add sync to
  // preserve the swsb id sync, so that it's safe to clear the dep
  if (currInst.hasPredication() ||
      (currInst.getExecSize() != dep.getInstruction()->getExecSize()) ||
      (currInst.getChannelOffset() != dep.getInstruction()->getChannelOffset()))
    needSyncForShootDownInst = true;

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Alyssa Rosenzweig
e66bfbd691 jay: fix SEND scoreboarding 
by inspection in a glmark2 shader.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Alyssa Rosenzweig
10fbfe1f65 jay: improve vector affinities 
Totals:
Instrs: 2807558 -> 2805545 (-0.07%); split: -0.10%, +0.02%
CodeSize: 45051504 -> 45015888 (-0.08%); split: -0.11%, +0.03%

Totals from 631 (23.84% of 2647) affected shaders:
Instrs: 736113 -> 734100 (-0.27%); split: -0.37%, +0.10%
CodeSize: 11666608 -> 11630992 (-0.31%); split: -0.42%, +0.12%

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Alyssa Rosenzweig
dd6efffb47 jay: generalize alignment heuristic 
Totals:
Instrs: 2807560 -> 2805375 (-0.08%); split: -0.27%, +0.20%
CodeSize: 45051536 -> 45011488 (-0.09%); split: -0.28%, +0.19%

Totals from 1292 (48.81% of 2647) affected shaders:
Instrs: 2154123 -> 2151938 (-0.10%); split: -0.36%, +0.26%
CodeSize: 34658096 -> 34618048 (-0.12%); split: -0.36%, +0.25%

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Alyssa Rosenzweig
e54898fdc8 jay: tweak roundrobin 
Totals:
Instrs: 2809726 -> 2807560 (-0.08%); split: -0.39%, +0.31%
CodeSize: 45085472 -> 45051536 (-0.08%); split: -0.41%, +0.33%

Totals from 2229 (84.21% of 2647) affected shaders:
Instrs: 2726162 -> 2723996 (-0.08%); split: -0.40%, +0.32%
CodeSize: 43775120 -> 43741184 (-0.08%); split: -0.42%, +0.34%

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Alyssa Rosenzweig
8aeeaa4d85 jay: marginally improve send splitting heuristic 
Instrs: 2810815 -> 2809726 (-0.04%); split: -0.07%, +0.04%
CodeSize: 45101440 -> 45085472 (-0.04%); split: -0.07%, +0.04%

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Alyssa Rosenzweig
820e3a9403 jay: roundrobin RA 
Totals:
Instrs: 2827788 -> 2810815 (-0.60%); split: -1.12%, +0.52%
CodeSize: 45449488 -> 45101440 (-0.77%); split: -1.28%, +0.51%
Number of spill instructions: 1984 -> 1982 (-0.10%)
Number of fill instructions: 2272 -> 2270 (-0.09%)

Totals from 2449 (92.52% of 2647) affected shaders:
Instrs: 2818824 -> 2801851 (-0.60%); split: -1.13%, +0.53%
CodeSize: 45314880 -> 44966832 (-0.77%); split: -1.28%, +0.51%
Number of spill instructions: 1984 -> 1982 (-0.10%)
Number of fill instructions: 2272 -> 2270 (-0.09%)

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Alyssa Rosenzweig
57a00707a1 jay: drop GRF reg stats 
doesn't match what brw does, will revisit for Xe3.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Alyssa Rosenzweig
4ab67f80e9 jay: rematerialize address regs 
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Alyssa Rosenzweig
9a9365ff9b jay: fix W-entry calcs 
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40960>
 2026-04-14 23:14:07 +00:00
Alyssa Rosenzweig
e42e319313 intel: add Jay 
Jay is a new SSA-based compiler for Intel GPUs. This is an early
work-in-progress. It isn't ready to ship, but we'd like to move development in
tree rather than rebasing the world every week. Please don't bother testing yet
- we know the status and we're working on it!

Jay's design is similar to other modern NIR backends, particularly ACO, NAK and
AGX. It is fully SSA, deconstructing phis after RA. We use a Colombet register
allocator similar to NAK, allowing us to handle Intel's complex register
regioning restrictions in a straightforward way. Spilling logical registers is
straightforward with Braun-Hack.

Thanks to the SSA-based design, the entire backend is essentially linear time,
regardless of register pressure, addressing brw's excessive compile time when
especially spilling with brw.

In this current early draft, we support a limited subset of all three APIs on
Xe2. A lot works and a lot doesn't. The core compiler is there (spilling,
scoreboarding, SIMD32, etc should more or less work), but there are details to
fill in for both performance and correctness. We essentially pass conformance on
OpenGL ES 3.0 and OpenCL 3.0, and we're busy iterating on Vulkan.

Likewise, additional hardware support will come down the line. There's nothing
fundamentally Xe2-specific here. I just have a Lunarlake laptop on my desk, Ken
has a Battlemage card, and we had to pick _something_ as the first target.

Co-authored-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40835>
 2026-04-10 18:21:21 +00:00