Skip to content

x64: Enable load-coalescing for SSE/AVX instructions #5841

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
cfallin merged 4 commits into from
Feb 21, 2023
Merged

x64: Enable load-coalescing for SSE/AVX instructions #5841

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
4a3c539
x64: Enable load-coalescing for SSE/AVX instructions
alexcrichton Feb 19, 2023
6027c04
Fix tests
alexcrichton Feb 20, 2023
1fd55b6
Fix move-and-extend to be unaligned
alexcrichton Feb 20, 2023
37d388b
Review comments
alexcrichton Feb 21, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
206 changes: 133 additions & 73 deletions cranelift/codegen/src/isa/x64/inst.isle
View file
Open in desktop

Large diffs are not rendered by default.

205 changes: 119 additions & 86 deletions cranelift/codegen/src/isa/x64/inst/args.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,8 +31,8 @@ macro_rules! newtype_of_reg {
$newtype_reg:ident,
$newtype_writable_reg:ident,
$newtype_option_writable_reg:ident,
$newtype_reg_mem:ident,
$newtype_reg_mem_imm:ident,
reg_mem: ($($newtype_reg_mem:ident $(aligned:$aligned:ident)?),*),
reg_mem_imm: ($($newtype_reg_mem_imm:ident $(aligned:$aligned_imm:ident)?),*),
$newtype_imm8_reg:ident,
|$check_reg:ident| $check:expr
) => {
Expand Down Expand Up @@ -102,108 +102,130 @@ macro_rules! newtype_of_reg {
}
}

/// A newtype wrapper around `RegMem` for general-purpose registers.
#[derive(Clone, Debug)]
pub struct $newtype_reg_mem(RegMem);
$(
/// A newtype wrapper around `RegMem` for general-purpose registers.
#[derive(Clone, Debug)]
pub struct $newtype_reg_mem(RegMem);

impl From<$newtype_reg_mem> for RegMem {
fn from(rm: $newtype_reg_mem) -> Self {
rm.0
impl From<$newtype_reg_mem> for RegMem {
fn from(rm: $newtype_reg_mem) -> Self {
rm.0
}
}
}

impl From<$newtype_reg> for $newtype_reg_mem {
fn from(r: $newtype_reg) -> Self {
$newtype_reg_mem(RegMem::reg(r.into()))
impl From<$newtype_reg> for $newtype_reg_mem {
fn from(r: $newtype_reg) -> Self {
$newtype_reg_mem(RegMem::reg(r.into()))
}
}
}

impl $newtype_reg_mem {
/// Construct a `RegMem` newtype from the given `RegMem`, or return
/// `None` if the `RegMem` is not a valid instance of this `RegMem`
/// newtype.
pub fn new(rm: RegMem) -> Option<Self> {
match rm {
RegMem::Mem { addr: _ } => Some(Self(rm)),
RegMem::Reg { reg: $check_reg } if $check => Some(Self(rm)),
RegMem::Reg { reg: _ } => None,
impl $newtype_reg_mem {
/// Construct a `RegMem` newtype from the given `RegMem`, or return
/// `None` if the `RegMem` is not a valid instance of this `RegMem`
/// newtype.
pub fn new(rm: RegMem) -> Option<Self> {
match rm {
RegMem::Mem { addr } => {
let mut _allow = true;
$(
if $aligned {
_allow = addr.aligned();
}
)?
if _allow {
Some(Self(RegMem::Mem { addr }))
} else {
None
}
}
RegMem::Reg { reg: $check_reg } if $check => Some(Self(rm)),
RegMem::Reg { reg: _ } => None,
}
}
}

/// Convert this newtype into its underlying `RegMem`.
pub fn to_reg_mem(self) -> RegMem {
self.0
}
/// Convert this newtype into its underlying `RegMem`.
pub fn to_reg_mem(self) -> RegMem {
self.0
}

#[allow(dead_code)] // Used by some newtypes and not others.
pub(crate) fn get_operands<F: Fn(VReg) -> VReg>(
&self,
collector: &mut OperandCollector<'_, F>,
) {
self.0.get_operands(collector);
#[allow(dead_code)] // Used by some newtypes and not others.
pub(crate) fn get_operands<F: Fn(VReg) -> VReg>(
&self,
collector: &mut OperandCollector<'_, F>,
) {
self.0.get_operands(collector);
}
}
}
impl PrettyPrint for $newtype_reg_mem {
fn pretty_print(&self, size: u8, allocs: &mut AllocationConsumer<'_>) -> String {
self.0.pretty_print(size, allocs)
impl PrettyPrint for $newtype_reg_mem {
fn pretty_print(&self, size: u8, allocs: &mut AllocationConsumer<'_>) -> String {
self.0.pretty_print(size, allocs)
}
}
}
)*

/// A newtype wrapper around `RegMemImm`.
#[derive(Clone, Debug)]
pub struct $newtype_reg_mem_imm(RegMemImm);
$(
/// A newtype wrapper around `RegMemImm`.
#[derive(Clone, Debug)]
pub struct $newtype_reg_mem_imm(RegMemImm);

impl From<$newtype_reg_mem_imm> for RegMemImm {
fn from(rmi: $newtype_reg_mem_imm) -> RegMemImm {
rmi.0
}
}

impl From<$newtype_reg> for $newtype_reg_mem_imm {
fn from(r: $newtype_reg) -> Self {
$newtype_reg_mem_imm(RegMemImm::reg(r.into()))
impl From<$newtype_reg_mem_imm> for RegMemImm {
fn from(rmi: $newtype_reg_mem_imm) -> RegMemImm {
rmi.0
}
}
}

impl From<$newtype_reg_mem> for $newtype_reg_mem_imm {
fn from(r: $newtype_reg_mem) -> Self {
$newtype_reg_mem_imm(r.0.into())
impl From<$newtype_reg> for $newtype_reg_mem_imm {
fn from(r: $newtype_reg) -> Self {
$newtype_reg_mem_imm(RegMemImm::reg(r.into()))
}
}
}

impl $newtype_reg_mem_imm {
/// Construct this newtype from the given `RegMemImm`, or return
/// `None` if the `RegMemImm` is not a valid instance of this
/// newtype.
pub fn new(rmi: RegMemImm) -> Option<Self> {
match rmi {
RegMemImm::Imm { .. } => Some(Self(rmi)),
RegMemImm::Mem { addr: _ } => Some(Self(rmi)),
RegMemImm::Reg { reg: $check_reg } if $check => Some(Self(rmi)),
RegMemImm::Reg { reg: _ } => None,
impl $newtype_reg_mem_imm {
/// Construct this newtype from the given `RegMemImm`, or return
/// `None` if the `RegMemImm` is not a valid instance of this
/// newtype.
pub fn new(rmi: RegMemImm) -> Option<Self> {
match rmi {
RegMemImm::Imm { .. } => Some(Self(rmi)),
RegMemImm::Mem { addr } => {
let mut _allow = true;
$(
if $aligned_imm {
_allow = addr.aligned();
}
)?
if _allow {
Some(Self(RegMemImm::Mem { addr }))
} else {
None
}
}
RegMemImm::Reg { reg: $check_reg } if $check => Some(Self(rmi)),
RegMemImm::Reg { reg: _ } => None,
}
}
}

/// Convert this newtype into its underlying `RegMemImm`.
#[allow(dead_code)] // Used by some newtypes and not others.
pub fn to_reg_mem_imm(self) -> RegMemImm {
self.0
}
/// Convert this newtype into its underlying `RegMemImm`.
#[allow(dead_code)] // Used by some newtypes and not others.
pub fn to_reg_mem_imm(self) -> RegMemImm {
self.0
}

#[allow(dead_code)] // Used by some newtypes and not others.
pub(crate) fn get_operands<F: Fn(VReg) -> VReg>(
&self,
collector: &mut OperandCollector<'_, F>,
) {
self.0.get_operands(collector);
#[allow(dead_code)] // Used by some newtypes and not others.
pub(crate) fn get_operands<F: Fn(VReg) -> VReg>(
&self,
collector: &mut OperandCollector<'_, F>,
) {
self.0.get_operands(collector);
}
}
}

impl PrettyPrint for $newtype_reg_mem_imm {
fn pretty_print(&self, size: u8, allocs: &mut AllocationConsumer<'_>) -> String {
self.0.pretty_print(size, allocs)
impl PrettyPrint for $newtype_reg_mem_imm {
fn pretty_print(&self, size: u8, allocs: &mut AllocationConsumer<'_>) -> String {
self.0.pretty_print(size, allocs)
}
}
}
)*

/// A newtype wrapper around `Imm8Reg`.
#[derive(Clone, Debug)]
Expand Down Expand Up @@ -242,8 +264,8 @@ newtype_of_reg!(
Gpr,
WritableGpr,
OptionWritableGpr,
GprMem,
GprMemImm,
reg_mem: (GprMem),
reg_mem_imm: (GprMemImm),
Imm8Gpr,
|reg| reg.class() == RegClass::Int
);
Expand All @@ -253,8 +275,8 @@ newtype_of_reg!(
Xmm,
WritableXmm,
OptionWritableXmm,
XmmMem,
XmmMemImm,
reg_mem: (XmmMem, XmmMemAligned aligned:true),
reg_mem_imm: (XmmMemImm, XmmMemAlignedImm aligned:true),
Imm8Xmm,
|reg| reg.class() == RegClass::Float
);
Expand Down Expand Up @@ -420,6 +442,10 @@ impl Amode {
}
ret
}

pub(crate) fn aligned(&self) -> bool {
self.get_flags().aligned()
}
}

impl PrettyPrint for Amode {
Expand Down Expand Up @@ -531,6 +557,13 @@ impl SyntheticAmode {
}
}
}

pub(crate) fn aligned(&self) -> bool {
match self {
SyntheticAmode::Real(addr) => addr.aligned(),
SyntheticAmode::NominalSPOffset { .. } | SyntheticAmode::ConstantOffset { .. } => true,
}
}
}

impl Into<SyntheticAmode> for Amode {
Expand Down
34 changes: 33 additions & 1 deletion cranelift/codegen/src/isa/x64/inst/emit.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1731,7 +1731,21 @@ pub(crate) fn emit(
sink.bind_label(else_label);
}

Inst::XmmUnaryRmR {
Inst::XmmUnaryRmR { op, src, dst } => {
emit(
&Inst::XmmUnaryRmRUnaligned {
op: *op,
src: XmmMem::new(src.clone().into()).unwrap(),
dst: *dst,
},
allocs,
sink,
info,
state,
);
}

Inst::XmmUnaryRmRUnaligned {
op,
src: src_e,
dst: reg_g,
Expand Down Expand Up @@ -1843,6 +1857,24 @@ pub(crate) fn emit(
}

Inst::XmmRmR {
op,
src1,
src2,
dst,
} => emit(
&Inst::XmmRmRUnaligned {
op: *op,
dst: *dst,
src1: *src1,
src2: XmmMem::new(src2.clone().to_reg_mem()).unwrap(),
},
allocs,
sink,
info,
state,
),

Inst::XmmRmRUnaligned {
op,
src1,
src2: src_e,
Expand Down
8 changes: 4 additions & 4 deletions cranelift/codegen/src/isa/x64/inst/emit_tests.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,7 @@ impl Inst {
debug_assert!(dst.to_reg().class() == RegClass::Float);
Inst::XmmUnaryRmRImm {
op,
src: XmmMem::new(src).unwrap(),
src: XmmMemAligned::new(src).unwrap(),
imm,
dst: WritableXmm::from_writable_reg(dst).unwrap(),
}
Expand All @@ -56,7 +56,7 @@ impl Inst {
Inst::XmmRmiReg {
opcode,
src1: Xmm::new(dst.to_reg()).unwrap(),
src2: XmmMemImm::new(src).unwrap(),
src2: XmmMemAlignedImm::new(src).unwrap(),
dst: WritableXmm::from_writable_reg(dst).unwrap(),
}
}
Expand Down Expand Up @@ -96,7 +96,7 @@ impl Inst {
debug_assert!(dst.to_reg().class() == RegClass::Float);
Inst::XmmUnaryRmR {
op,
src: XmmMem::new(src).unwrap(),
src: XmmMemAligned::new(src).unwrap(),
dst: WritableXmm::from_writable_reg(dst).unwrap(),
}
}
Expand Down Expand Up @@ -136,7 +136,7 @@ impl Inst {
Inst::XmmRmRBlend {
op,
src1: Xmm::new(dst.to_reg()).unwrap(),
src2: XmmMem::new(src2).unwrap(),
src2: XmmMemAligned::new(src2).unwrap(),
mask: Xmm::new(regs::xmm0()).unwrap(),
dst: WritableXmm::from_writable_reg(dst).unwrap(),
}
Expand Down
Loading