fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-08 11:18:08 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

aco: fix byte_align_scalar for 3 dword vectors

Browse source 
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Fixes: fe08f0ccf9
   ('aco: add byte_align_scalar() & trim_subdword_vector() helper functions')

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4710>
(cherry picked from commit 8faf85f687)
This commit is contained in:
Rhys Perry 2020-08-24 20:00:10 +01:00 • committed by Dylan Baker
parent 4c6cc7277f
commit e98a278dcd
2 changed files with 12 additions and 5 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
.pick_status.json
View file

@ -2065,7 +2065,7 @@
"description": "aco: fix byte_align_scalar for 3 dword vectors",
"nominated": true,
"nomination_type": 1,
"resolution": 0,
"resolution": 1,
"master_sha": null,
"because_sha": "fe08f0ccf94a7315bded5868b4f6a8bae744de79"
},

15
src/amd/compiler/aco_instruction_selection.cpp
View file

@ -430,10 +430,17 @@ void byte_align_scalar(isel_context *ctx, Temp vec, Operand offset, Temp dst)
emit_split_vector(ctx, dst, 2);
else
emit_extract_vector(ctx, tmp, 0, dst);
} else if (vec.size() == 4) {
Temp lo = bld.tmp(s2), hi = bld.tmp(s2);
bld.pseudo(aco_opcode::p_split_vector, Definition(lo), Definition(hi), vec);
hi = bld.pseudo(aco_opcode::p_extract_vector, bld.def(s1), hi, Operand(0u));
} else if (vec.size() == 3 || vec.size() == 4) {
Temp lo = bld.tmp(s2), hi;
if (vec.size() == 3) {
/* this can happen if we use VMEM for a uniform load */
hi = bld.tmp(s1);
bld.pseudo(aco_opcode::p_split_vector, Definition(lo), Definition(hi), vec);
} else {
hi = bld.tmp(s2);
bld.pseudo(aco_opcode::p_split_vector, Definition(lo), Definition(hi), vec);
hi = bld.pseudo(aco_opcode::p_extract_vector, bld.def(s1), hi, Operand(0u));
}
if (select != Temp())
hi = bld.sop2(aco_opcode::s_cselect_b32, bld.def(s1), hi, Operand(0u), bld.scc(select));
lo = bld.sop2(aco_opcode::s_lshr_b64, bld.def(s2), bld.def(s1, scc), lo, shift);