fluencelabs/wasmer
1
0
Fork 0
mirror of https://github.com/fluencelabs/wasmer synced 2025-03-18 09:00:51 +00:00
Code Issues Projects Releases Wiki Activity

Finish implementation of trunc_sat in singlepass x86-64.

Browse Source
This commit is contained in:
Nick Lewycky 2019-12-16 14:29:56 -08:00
parent 442c40f5f1
commit d52c193e84
1 changed files with 309 additions and 98 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

407
lib/singlepass-backend/src/codegen_x64.rs
View File

@ -2028,7 +2028,9 @@ impl X64FunctionCode {
reg: XMM,
lower_bound: f32,
upper_bound: f32,
fail_label: <Assembler as Emitter>::Label,
underflow_label: <Assembler as Emitter>::Label,
overflow_label: <Assembler as Emitter>::Label,
nan_label: <Assembler as Emitter>::Label,
succeed_label: <Assembler as Emitter>::Label,
) {
let lower_bound = f32::to_bits(lower_bound);
@ -2043,7 +2045,7 @@ impl X64FunctionCode {
a.emit_vcmpless(reg, XMMOrMemory::XMM(tmp_x), tmp_x);
a.emit_mov(Size::S32, Location::XMM(tmp_x), Location::GPR(tmp));
a.emit_cmp(Size::S32, Location::Imm32(0), Location::GPR(tmp));
a.emit_jmp(Condition::NotEqual, fail_label);
a.emit_jmp(Condition::NotEqual, underflow_label);
// Overflow.
a.emit_mov(Size::S32, Location::Imm32(upper_bound), Location::GPR(tmp));
@ -2051,13 +2053,13 @@ impl X64FunctionCode {
a.emit_vcmpgess(reg, XMMOrMemory::XMM(tmp_x), tmp_x);
a.emit_mov(Size::S32, Location::XMM(tmp_x), Location::GPR(tmp));
a.emit_cmp(Size::S32, Location::Imm32(0), Location::GPR(tmp));
a.emit_jmp(Condition::NotEqual, fail_label);
a.emit_jmp(Condition::NotEqual, overflow_label);
// NaN.
a.emit_vcmpeqss(reg, XMMOrMemory::XMM(reg), tmp_x);
a.emit_mov(Size::S32, Location::XMM(tmp_x), Location::GPR(tmp));
a.emit_cmp(Size::S32, Location::Imm32(0), Location::GPR(tmp));
a.emit_jmp(Condition::Equal, fail_label);
a.emit_jmp(Condition::Equal, nan_label);
a.emit_jmp(Condition::None, succeed_label);
@ -2076,28 +2078,70 @@ impl X64FunctionCode {
let trap = a.get_label();
let end = a.get_label();
Self::emit_f32_int_conv_check(a, m, reg, lower_bound, upper_bound, trap, end);
Self::emit_f32_int_conv_check(a, m, reg, lower_bound, upper_bound, trap, trap, trap, end);
a.emit_label(trap);
a.emit_ud2();
a.emit_label(end);
}
fn emit_f32_int_conv_check_zero(
fn emit_f32_int_conv_check_sat<
F1: FnOnce(&mut Assembler, &mut Machine),
F2: FnOnce(&mut Assembler, &mut Machine),
F3: FnOnce(&mut Assembler, &mut Machine),
F4: FnOnce(&mut Assembler, &mut Machine),
>(
a: &mut Assembler,
m: &mut Machine,
reg: XMM,
lower_bound: f32,
upper_bound: f32,
underflow_cb: F1,
overflow_cb: F2,
nan_cb: Option<F3>,
convert_cb: F4,
) {
let trap = a.get_label();
// As an optimization nan_cb is optional, and when set to None we turn
// use 'underflow' as the 'nan' label. This is useful for callers who
// set the return value to zero for both underflow and nan.
let underflow = a.get_label();
let overflow = a.get_label();
let nan = if nan_cb.is_some() {
a.get_label()
} else {
underflow
};
let convert = a.get_label();
let end = a.get_label();
Self::emit_f32_int_conv_check(a, m, reg, lower_bound, upper_bound, trap, end);
a.emit_label(trap);
let gpr = m.acquire_temp_gpr().unwrap();
a.emit_mov(Size::S32, Location::Imm32(0), Location::GPR(gpr));
a.emit_mov(Size::S32, Location::GPR(gpr), Location::XMM(reg));
m.release_temp_gpr(gpr);
Self::emit_f32_int_conv_check(
a,
m,
reg,
lower_bound,
upper_bound,
underflow,
overflow,
nan,
convert,
);
a.emit_label(underflow);
underflow_cb(a, m);
a.emit_jmp(Condition::None, end);
a.emit_label(overflow);
overflow_cb(a, m);
a.emit_jmp(Condition::None, end);
if let Some(cb) = nan_cb {
a.emit_label(nan);
cb(a, m);
a.emit_jmp(Condition::None, end);
}
a.emit_label(convert);
convert_cb(a, m);
a.emit_label(end);
}
@ -2108,7 +2152,9 @@ impl X64FunctionCode {
reg: XMM,
lower_bound: f64,
upper_bound: f64,
fail_label: <Assembler as Emitter>::Label,
underflow_label: <Assembler as Emitter>::Label,
overflow_label: <Assembler as Emitter>::Label,
nan_label: <Assembler as Emitter>::Label,
succeed_label: <Assembler as Emitter>::Label,
) {
let lower_bound = f64::to_bits(lower_bound);
@ -2123,7 +2169,7 @@ impl X64FunctionCode {
a.emit_vcmplesd(reg, XMMOrMemory::XMM(tmp_x), tmp_x);
a.emit_mov(Size::S32, Location::XMM(tmp_x), Location::GPR(tmp));
a.emit_cmp(Size::S32, Location::Imm32(0), Location::GPR(tmp));
a.emit_jmp(Condition::NotEqual, fail_label);
a.emit_jmp(Condition::NotEqual, underflow_label);
// Overflow.
a.emit_mov(Size::S64, Location::Imm64(upper_bound), Location::GPR(tmp));
@ -2131,13 +2177,13 @@ impl X64FunctionCode {
a.emit_vcmpgesd(reg, XMMOrMemory::XMM(tmp_x), tmp_x);
a.emit_mov(Size::S32, Location::XMM(tmp_x), Location::GPR(tmp));
a.emit_cmp(Size::S32, Location::Imm32(0), Location::GPR(tmp));
a.emit_jmp(Condition::NotEqual, fail_label);
a.emit_jmp(Condition::NotEqual, overflow_label);
// NaN.
a.emit_vcmpeqsd(reg, XMMOrMemory::XMM(reg), tmp_x);
a.emit_mov(Size::S32, Location::XMM(tmp_x), Location::GPR(tmp));
a.emit_cmp(Size::S32, Location::Imm32(0), Location::GPR(tmp));
a.emit_jmp(Condition::Equal, fail_label);
a.emit_jmp(Condition::Equal, nan_label);
a.emit_jmp(Condition::None, succeed_label);
@ -2156,28 +2202,70 @@ impl X64FunctionCode {
let trap = a.get_label();
let end = a.get_label();
Self::emit_f64_int_conv_check(a, m, reg, lower_bound, upper_bound, trap, end);
Self::emit_f64_int_conv_check(a, m, reg, lower_bound, upper_bound, trap, trap, trap, end);
a.emit_label(trap);
a.emit_ud2();
a.emit_label(end);
}
fn emit_f64_int_conv_check_zero(
fn emit_f64_int_conv_check_sat<
F1: FnOnce(&mut Assembler, &mut Machine),
F2: FnOnce(&mut Assembler, &mut Machine),
F3: FnOnce(&mut Assembler, &mut Machine),
F4: FnOnce(&mut Assembler, &mut Machine),
>(
a: &mut Assembler,
m: &mut Machine,
reg: XMM,
lower_bound: f64,
upper_bound: f64,
underflow_cb: F1,
overflow_cb: F2,
nan_cb: Option<F3>,
convert_cb: F4,
) {
let trap = a.get_label();
// As an optimization nan_cb is optional, and when set to None we turn
// use 'underflow' as the 'nan' label. This is useful for callers who
// set the return value to zero for both underflow and nan.
let underflow = a.get_label();
let overflow = a.get_label();
let nan = if nan_cb.is_some() {
a.get_label()
} else {
underflow
};
let convert = a.get_label();
let end = a.get_label();
Self::emit_f64_int_conv_check(a, m, reg, lower_bound, upper_bound, trap, end);
a.emit_label(trap);
let gpr = m.acquire_temp_gpr().unwrap();
a.emit_mov(Size::S64, Location::Imm64(0), Location::GPR(gpr));
a.emit_mov(Size::S64, Location::GPR(gpr), Location::XMM(reg));
m.release_temp_gpr(gpr);
Self::emit_f64_int_conv_check(
a,
m,
reg,
lower_bound,
upper_bound,
underflow,
overflow,
nan,
convert,
);
a.emit_label(underflow);
underflow_cb(a, m);
a.emit_jmp(Condition::None, end);
a.emit_label(overflow);
overflow_cb(a, m);
a.emit_jmp(Condition::None, end);
if let Some(cb) = nan_cb {
a.emit_label(nan);
cb(a, m);
a.emit_jmp(Condition::None, end);
}
a.emit_label(convert);
convert_cb(a, m);
a.emit_label(end);
}
@ -4657,17 +4745,29 @@ impl FunctionCodeGenerator<CodegenError> for X64FunctionCode {
loc,
Location::XMM(tmp_in),
);
Self::emit_f32_int_conv_check_zero(
Self::emit_f32_int_conv_check_sat(
a,
&mut self.machine,
tmp_in,
-1.0,
4294967296.0,
|a, _m| {
a.emit_mov(Size::S32, Location::Imm32(0), Location::GPR(tmp_out));
},
|a, _m| {
a.emit_mov(
Size::S32,
Location::Imm32(std::u32::MAX),
Location::GPR(tmp_out),
);
},
None::<fn(_a: &mut Assembler, _m: &mut Machine)>,
|a, _m| {
a.emit_cvttss2si_64(XMMOrMemory::XMM(tmp_in), tmp_out);
},
);
a.emit_cvttss2si_64(XMMOrMemory::XMM(tmp_in), tmp_out);
a.emit_mov(Size::S32, Location::GPR(tmp_out), ret);
self.machine.release_temp_xmm(tmp_in);
self.machine.release_temp_gpr(tmp_out);
}
@ -4752,17 +4852,35 @@ impl FunctionCodeGenerator<CodegenError> for X64FunctionCode {
loc,
Location::XMM(tmp_in),
);
Self::emit_f32_int_conv_check_zero(
Self::emit_f32_int_conv_check_sat(
a,
&mut self.machine,
tmp_in,
-2147483904.0,
2147483648.0,
|a, _m| {
a.emit_mov(
Size::S32,
Location::Imm32(std::i32::MIN as u32),
Location::GPR(tmp_out),
);
},
|a, _m| {
a.emit_mov(
Size::S32,
Location::Imm32(std::i32::MAX as u32),
Location::GPR(tmp_out),
);
},
Some(|a: &mut Assembler, _m: &mut Machine| {
a.emit_mov(Size::S32, Location::Imm32(0), Location::GPR(tmp_out));
}),
|a, _m| {
a.emit_cvttss2si_32(XMMOrMemory::XMM(tmp_in), tmp_out);
},
);
a.emit_cvttss2si_32(XMMOrMemory::XMM(tmp_in), tmp_out);
a.emit_mov(Size::S32, Location::GPR(tmp_out), ret);
self.machine.release_temp_xmm(tmp_in);
self.machine.release_temp_gpr(tmp_out);
}
@ -4847,16 +4965,35 @@ impl FunctionCodeGenerator<CodegenError> for X64FunctionCode {
loc,
Location::XMM(tmp_in),
);
Self::emit_f32_int_conv_check_zero(
Self::emit_f32_int_conv_check_sat(
a,
&mut self.machine,
tmp_in,
-9223373136366403584.0,
9223372036854775808.0,
|a, _m| {
a.emit_mov(
Size::S64,
Location::Imm64(std::i64::MIN as u64),
Location::GPR(tmp_out),
);
},
|a, _m| {
a.emit_mov(
Size::S64,
Location::Imm64(std::i64::MAX as u64),
Location::GPR(tmp_out),
);
},
Some(|a: &mut Assembler, _m: &mut Machine| {
a.emit_mov(Size::S64, Location::Imm64(0), Location::GPR(tmp_out));
}),
|a, _m| {
a.emit_cvttss2si_64(XMMOrMemory::XMM(tmp_in), tmp_out);
},
);
a.emit_cvttss2si_64(XMMOrMemory::XMM(tmp_in), tmp_out);
a.emit_mov(Size::S64, Location::GPR(tmp_out), ret);
a.emit_mov(Size::S64, Location::GPR(tmp_out), ret);
self.machine.release_temp_xmm(tmp_in);
self.machine.release_temp_gpr(tmp_out);
}
@ -4965,41 +5102,54 @@ impl FunctionCodeGenerator<CodegenError> for X64FunctionCode {
loc,
Location::XMM(tmp_in),
);
Self::emit_f32_int_conv_check_zero(
Self::emit_f32_int_conv_check_sat(
a,
&mut self.machine,
tmp_in,
-1.0,
18446744073709551616.0,
|a, _m| {
a.emit_mov(Size::S64, Location::Imm64(0), Location::GPR(tmp_out));
},
|a, _m| {
a.emit_mov(
Size::S64,
Location::Imm64(std::u64::MAX),
Location::GPR(tmp_out),
);
},
None::<fn(_a: &mut Assembler, _m: &mut Machine)>,
|a, m| {
let tmp = m.acquire_temp_gpr().unwrap(); // r15
let tmp_x1 = m.acquire_temp_xmm().unwrap(); // xmm1
let tmp_x2 = m.acquire_temp_xmm().unwrap(); // xmm3
a.emit_mov(
Size::S32,
Location::Imm32(1593835520u32),
Location::GPR(tmp),
); //float 9.22337203E+18
a.emit_mov(Size::S32, Location::GPR(tmp), Location::XMM(tmp_x1));
a.emit_mov(Size::S32, Location::XMM(tmp_in), Location::XMM(tmp_x2));
a.emit_vsubss(tmp_in, XMMOrMemory::XMM(tmp_x1), tmp_in);
a.emit_cvttss2si_64(XMMOrMemory::XMM(tmp_in), tmp_out);
a.emit_mov(
Size::S64,
Location::Imm64(0x8000000000000000u64),
Location::GPR(tmp),
);
a.emit_xor(Size::S64, Location::GPR(tmp_out), Location::GPR(tmp));
a.emit_cvttss2si_64(XMMOrMemory::XMM(tmp_x2), tmp_out);
a.emit_ucomiss(XMMOrMemory::XMM(tmp_x1), tmp_x2);
a.emit_cmovae_gpr_64(tmp, tmp_out);
m.release_temp_xmm(tmp_x2);
m.release_temp_xmm(tmp_x1);
m.release_temp_gpr(tmp);
},
);
let tmp = self.machine.acquire_temp_gpr().unwrap(); // r15
let tmp_x1 = self.machine.acquire_temp_xmm().unwrap(); // xmm1
let tmp_x2 = self.machine.acquire_temp_xmm().unwrap(); // xmm3
a.emit_mov(
Size::S32,
Location::Imm32(1593835520u32),
Location::GPR(tmp),
); //float 9.22337203E+18
a.emit_mov(Size::S32, Location::GPR(tmp), Location::XMM(tmp_x1));
a.emit_mov(Size::S32, Location::XMM(tmp_in), Location::XMM(tmp_x2));
a.emit_vsubss(tmp_in, XMMOrMemory::XMM(tmp_x1), tmp_in);
a.emit_cvttss2si_64(XMMOrMemory::XMM(tmp_in), tmp_out);
a.emit_mov(
Size::S64,
Location::Imm64(0x8000000000000000u64),
Location::GPR(tmp),
);
a.emit_xor(Size::S64, Location::GPR(tmp_out), Location::GPR(tmp));
a.emit_cvttss2si_64(XMMOrMemory::XMM(tmp_x2), tmp_out);
a.emit_ucomiss(XMMOrMemory::XMM(tmp_x1), tmp_x2);
a.emit_cmovae_gpr_64(tmp, tmp_out);
a.emit_mov(Size::S64, Location::GPR(tmp_out), ret);
self.machine.release_temp_xmm(tmp_x2);
self.machine.release_temp_xmm(tmp_x1);
self.machine.release_temp_gpr(tmp);
self.machine.release_temp_xmm(tmp_in);
self.machine.release_temp_gpr(tmp_out);
}
@ -5085,17 +5235,29 @@ impl FunctionCodeGenerator<CodegenError> for X64FunctionCode {
loc,
Location::XMM(tmp_in),
);
Self::emit_f64_int_conv_check_zero(
Self::emit_f64_int_conv_check_sat(
a,
&mut self.machine,
tmp_in,
-1.0,
4294967296.0,
|a, _m| {
a.emit_mov(Size::S32, Location::Imm32(0), Location::GPR(tmp_out));
},
|a, _m| {
a.emit_mov(
Size::S32,
Location::Imm32(std::u32::MAX),
Location::GPR(tmp_out),
);
},
None::<fn(_a: &mut Assembler, _m: &mut Machine)>,
|a, _m| {
a.emit_cvttsd2si_64(XMMOrMemory::XMM(tmp_in), tmp_out);
},
);
a.emit_cvttsd2si_64(XMMOrMemory::XMM(tmp_in), tmp_out);
a.emit_mov(Size::S32, Location::GPR(tmp_out), ret);
self.machine.release_temp_xmm(tmp_in);
self.machine.release_temp_gpr(tmp_out);
}
@ -5191,17 +5353,35 @@ impl FunctionCodeGenerator<CodegenError> for X64FunctionCode {
}
};
Self::emit_f64_int_conv_check_zero(
Self::emit_f64_int_conv_check_sat(
a,
&mut self.machine,
real_in,
-2147483649.0,
2147483648.0,
|a, _m| {
a.emit_mov(
Size::S32,
Location::Imm32(std::i32::MIN as u32),
Location::GPR(tmp_out),
);
},
|a, _m| {
a.emit_mov(
Size::S32,
Location::Imm32(std::i32::MAX as u32),
Location::GPR(tmp_out),
);
},
Some(|a: &mut Assembler, _m: &mut Machine| {
a.emit_mov(Size::S32, Location::Imm32(0), Location::GPR(tmp_out));
}),
|a, _m| {
a.emit_cvttsd2si_32(XMMOrMemory::XMM(real_in), tmp_out);
},
);
a.emit_cvttsd2si_32(XMMOrMemory::XMM(real_in), tmp_out);
a.emit_mov(Size::S32, Location::GPR(tmp_out), ret);
self.machine.release_temp_xmm(tmp_in);
self.machine.release_temp_gpr(tmp_out);
}
@ -5287,17 +5467,35 @@ impl FunctionCodeGenerator<CodegenError> for X64FunctionCode {
loc,
Location::XMM(tmp_in),
);
Self::emit_f64_int_conv_check_zero(
Self::emit_f64_int_conv_check_sat(
a,
&mut self.machine,
tmp_in,
-9223372036854777856.0,
9223372036854775808.0,
|a, _m| {
a.emit_mov(
Size::S64,
Location::Imm64(std::i64::MIN as u64),
Location::GPR(tmp_out),
);
},
|a, _m| {
a.emit_mov(
Size::S64,
Location::Imm64(std::i64::MAX as u64),
Location::GPR(tmp_out),
);
},
Some(|a: &mut Assembler, _m: &mut Machine| {
a.emit_mov(Size::S64, Location::Imm64(0), Location::GPR(tmp_out));
}),
|a, _m| {
a.emit_cvttsd2si_64(XMMOrMemory::XMM(tmp_in), tmp_out);
},
);
a.emit_cvttsd2si_64(XMMOrMemory::XMM(tmp_in), tmp_out);
a.emit_mov(Size::S64, Location::GPR(tmp_out), ret);
self.machine.release_temp_xmm(tmp_in);
self.machine.release_temp_gpr(tmp_out);
}
@ -5407,41 +5605,54 @@ impl FunctionCodeGenerator<CodegenError> for X64FunctionCode {
loc,
Location::XMM(tmp_in),
);
Self::emit_f64_int_conv_check_zero(
Self::emit_f64_int_conv_check_sat(
a,
&mut self.machine,
tmp_in,
-1.0,
18446744073709551616.0,
|a, _m| {
a.emit_mov(Size::S64, Location::Imm64(0), Location::GPR(tmp_out));
},
|a, _m| {
a.emit_mov(
Size::S64,
Location::Imm64(std::u64::MAX),
Location::GPR(tmp_out),
);
},
None::<fn(_a: &mut Assembler, _m: &mut Machine)>,
|a, m| {
let tmp = m.acquire_temp_gpr().unwrap(); // r15
let tmp_x1 = m.acquire_temp_xmm().unwrap(); // xmm1
let tmp_x2 = m.acquire_temp_xmm().unwrap(); // xmm3
a.emit_mov(
Size::S64,
Location::Imm64(4890909195324358656u64),
Location::GPR(tmp),
); //double 9.2233720368547758E+18
a.emit_mov(Size::S64, Location::GPR(tmp), Location::XMM(tmp_x1));
a.emit_mov(Size::S64, Location::XMM(tmp_in), Location::XMM(tmp_x2));
a.emit_vsubsd(tmp_in, XMMOrMemory::XMM(tmp_x1), tmp_in);
a.emit_cvttsd2si_64(XMMOrMemory::XMM(tmp_in), tmp_out);
a.emit_mov(
Size::S64,
Location::Imm64(0x8000000000000000u64),
Location::GPR(tmp),
);
a.emit_xor(Size::S64, Location::GPR(tmp_out), Location::GPR(tmp));
a.emit_cvttsd2si_64(XMMOrMemory::XMM(tmp_x2), tmp_out);
a.emit_ucomisd(XMMOrMemory::XMM(tmp_x1), tmp_x2);
a.emit_cmovae_gpr_64(tmp, tmp_out);
m.release_temp_xmm(tmp_x2);
m.release_temp_xmm(tmp_x1);
m.release_temp_gpr(tmp);
},
);
let tmp = self.machine.acquire_temp_gpr().unwrap(); // r15
let tmp_x1 = self.machine.acquire_temp_xmm().unwrap(); // xmm1
let tmp_x2 = self.machine.acquire_temp_xmm().unwrap(); // xmm3
a.emit_mov(
Size::S64,
Location::Imm64(4890909195324358656u64),
Location::GPR(tmp),
); //double 9.2233720368547758E+18
a.emit_mov(Size::S64, Location::GPR(tmp), Location::XMM(tmp_x1));
a.emit_mov(Size::S64, Location::XMM(tmp_in), Location::XMM(tmp_x2));
a.emit_vsubsd(tmp_in, XMMOrMemory::XMM(tmp_x1), tmp_in);
a.emit_cvttsd2si_64(XMMOrMemory::XMM(tmp_in), tmp_out);
a.emit_mov(
Size::S64,
Location::Imm64(0x8000000000000000u64),
Location::GPR(tmp),
);
a.emit_xor(Size::S64, Location::GPR(tmp_out), Location::GPR(tmp));
a.emit_cvttsd2si_64(XMMOrMemory::XMM(tmp_x2), tmp_out);
a.emit_ucomisd(XMMOrMemory::XMM(tmp_x1), tmp_x2);
a.emit_cmovae_gpr_64(tmp, tmp_out);
a.emit_mov(Size::S64, Location::GPR(tmp_out), ret);
self.machine.release_temp_xmm(tmp_x2);
self.machine.release_temp_xmm(tmp_x1);
self.machine.release_temp_gpr(tmp);
self.machine.release_temp_xmm(tmp_in);
self.machine.release_temp_gpr(tmp_out);
}