fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-01-03 18:00:10 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

nak: Rework isetp64 on sm50 again

Browse source 
Unlike on Volta, isetp.x on Maxwell seems to take the accumulator into
account and I don't know exactly how.  Fortunately, we can avoid it by
emitting a different 3 instructions.  The result is actually a good bit
simpler.

Fixes: efd86d6daa ("nak/builder: Fix isetp64 for pre-Volta")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30402>
This commit is contained in:
Faith Ekstrand 2024-07-27 22:04:43 -05:00 • committed by Marge Bot
parent 3d4a4a68a0
commit c695fd7cd2
1 changed files with 27 additions and 40 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

67
src/nouveau/compiler/nak/builder.rs
View file

@ -574,50 +574,37 @@ pub trait SSABuilder: Builder {
low_cmp: low.into(),
});
} else {
// On Maxwell, iset.ex only has one source for both accum
// and low_cmp and it does a weird truth table dance. (See
// Foldable for OpISetP for details.)
let low_or_accum = self.alloc_ssa(RegFile::Pred, 1);
let set_op = match cmp_op {
IntCmpOp::Ge | IntCmpOp::Gt => {
// When x != y, we want low_or_accum == false
self.push_op(OpISetP {
dst: low_or_accum.into(),
set_op: PredSetOp::And,
cmp_op: IntCmpOp::Eq,
cmp_type: IntCmpType::U32,
ex: false,
srcs: [x[1].into(), y[1].into()],
accum: low.into(),
low_cmp: true.into(),
});
PredSetOp::Or
}
IntCmpOp::Le | IntCmpOp::Lt => {
// When x != y, we want low_or_accum == true
self.push_op(OpISetP {
dst: low_or_accum.into(),
set_op: PredSetOp::Or,
cmp_op: IntCmpOp::Ne,
cmp_type: IntCmpType::U32,
ex: false,
srcs: [x[1].into(), y[1].into()],
accum: low.into(),
low_cmp: true.into(),
});
PredSetOp::And
}
_ => panic!("Not an integer inequality"),
};
// On Maxwell, iset.ex doesn't do what we want so we need to
// do it with 3 comparisons. Fortunately, we can chain them
// together and don't need the extra logic that the NIR
// lowering would emit.
let low_and_high_eq = self.alloc_ssa(RegFile::Pred, 1);
self.push_op(OpISetP {
dst: low_and_high_eq.into(),
set_op: PredSetOp::And,
cmp_op: IntCmpOp::Eq,
cmp_type: IntCmpType::U32,
ex: false,
srcs: [x[1].into(), y[1].into()],
accum: low.into(),
low_cmp: true.into(),
});
self.push_op(OpISetP {
dst: dst.into(),
set_op,
cmp_op,
set_op: PredSetOp::Or,
// We always want a strict inequality for the high part
// so it's false when the two are equal and safe to OR
// with the low part.
cmp_op: match cmp_op {
IntCmpOp::Lt | IntCmpOp::Le => IntCmpOp::Lt,
IntCmpOp::Gt | IntCmpOp::Ge => IntCmpOp::Gt,
_ => panic!("Not an integer inequality"),
},
cmp_type,
ex: true,
ex: false,
srcs: [x[1].into(), y[1].into()],
accum: low_or_accum.into(),
low_cmp: low_or_accum.into(),
accum: low_and_high_eq.into(),
low_cmp: true.into(),
});
}
}