mii/wasmer
1
0
Fork 0
mirror of https://github.com/mii443/wasmer.git synced 2025-08-24 01:09:26 +00:00
Code Issues Packages Projects Releases Wiki Activity

Implement all but 2 new SIMD ops in compiler LLVM

Browse Source
This commit is contained in:
Mark McCaskey
2021-05-11 11:14:25 -07:00
parent 2f5f9d9e48
commit da86e69a45
3 changed files with 243 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files

246
lib/compiler-llvm/src/translator/code.rs
View File

@ -344,7 +344,7 @@ impl<'ctx, 'a> LLVMFunctionCodeGenerator<'ctx, 'a> {
// Convert floating point vector to integer and saturate when out of range.
// https://github.com/WebAssembly/nontrapping-float-to-int-conversions/blob/master/proposals/nontrapping-float-to-int-conversion/Overview.md
fn trunc_sat<T: FloatMathType<'ctx>>(
fn trunc_sat_inner<T: FloatMathType<'ctx>>(
&self,
fvec_ty: T,
ivec_ty: T::MathConvType,
@ -353,7 +353,7 @@ impl<'ctx, 'a> LLVMFunctionCodeGenerator<'ctx, 'a> {
int_min_value: u64,
int_max_value: u64,
value: IntValue<'ctx>,
) -> IntValue<'ctx> {
) -> VectorValue<'ctx> {
// a) Compare vector with itself to identify NaN lanes.
// b) Compare vector with splat of inttofp(upper_bound) to identify
// lanes that need to saturate to max.
@ -454,10 +454,32 @@ impl<'ctx, 'a> LLVMFunctionCodeGenerator<'ctx, 'a> {
.builder
.build_select(above_upper_bound_cmp, int_max_value, value, "")
.into_vector_value();
let res = self
.builder
self.builder
.build_select(below_lower_bound_cmp, int_min_value, value, "")
.into_vector_value();
.into_vector_value()
}
// Convert floating point vector to integer and saturate when out of range.
// https://github.com/WebAssembly/nontrapping-float-to-int-conversions/blob/master/proposals/nontrapping-float-to-int-conversion/Overview.md
fn trunc_sat<T: FloatMathType<'ctx>>(
&self,
fvec_ty: T,
ivec_ty: T::MathConvType,
lower_bound: u64, // Exclusive (least representable value)
upper_bound: u64, // Exclusive (greatest representable value)
int_min_value: u64,
int_max_value: u64,
value: IntValue<'ctx>,
) -> IntValue<'ctx> {
let res = self.trunc_sat_inner(
fvec_ty,
ivec_ty,
lower_bound,
upper_bound,
int_min_value,
int_max_value,
value,
);
self.builder
.build_bitcast(res, self.intrinsics.i128_ty, "")
.into_int_value()
@ -3158,17 +3180,73 @@ impl<'ctx, 'a> LLVMFunctionCodeGenerator<'ctx, 'a> {
let res = self.builder.build_bitcast(res, self.intrinsics.i128_ty, "");
self.state.push1(res);
}
Operator::I8x16Bitmask => {
let (v, i) = self.state.pop1_extra()?;
let (v, _) = self.v128_into_i8x16(v, i);
let zeros = self.intrinsics.i8x16_ty.const_zero();
let res = self
.builder
.build_int_compare(IntPredicate::SLT, v, zeros, "");
let res = self
.builder
.build_bitcast(res, self.intrinsics.i16_ty, "")
.into_int_value();
let res = self
.builder
.build_int_z_extend(res, self.intrinsics.i32_ty, "");
self.state.push1(res);
}
Operator::I16x8Bitmask => {
// WIP
/*let (v, i) = self.state.pop1_extra()?;
let (v, i) = self.state.pop1_extra()?;
let (v, _) = self.v128_into_i16x8(v, i);
let one = self.intrinsics.i16_ty.const_int(1, false);
let one = VectorType::const_vector(&[one; 8]);
let sign_bits = self.builder.build_right_shift(v, one, true, "");
let res = self.builder.build_and(v, sign_bits, "");
*/
todo!()
let zeros = self.intrinsics.i16x8_ty.const_zero();
let res = self
.builder
.build_int_compare(IntPredicate::SLT, v, zeros, "");
let res = self
.builder
.build_bitcast(res, self.intrinsics.i8_ty, "")
.into_int_value();
let res = self
.builder
.build_int_z_extend(res, self.intrinsics.i32_ty, "");
self.state.push1(res);
}
Operator::I32x4Bitmask => {
let (v, i) = self.state.pop1_extra()?;
let (v, _) = self.v128_into_i32x4(v, i);
let zeros = self.intrinsics.i32x4_ty.const_zero();
let res = self
.builder
.build_int_compare(IntPredicate::SLT, v, zeros, "");
let res = self
.builder
.build_bitcast(res, self.intrinsics.i4_ty, "")
.into_int_value();
let res = self
.builder
.build_int_z_extend(res, self.intrinsics.i32_ty, "");
self.state.push1(res);
}
Operator::I64x2Bitmask => {
let (v, i) = self.state.pop1_extra()?;
let (v, _) = self.v128_into_i64x2(v, i);
let zeros = self.intrinsics.i64x2_ty.const_zero();
let res = self
.builder
.build_int_compare(IntPredicate::SLT, v, zeros, "");
let res = self
.builder
.build_bitcast(res, self.intrinsics.i2_ty, "")
.into_int_value();
let res = self
.builder
.build_int_z_extend(res, self.intrinsics.i32_ty, "");
self.state.push1(res);
}
Operator::I32Shl => {
let ((v1, i1), (v2, i2)) = self.state.pop2_extra()?;
@ -4741,6 +4819,19 @@ impl<'ctx, 'a> LLVMFunctionCodeGenerator<'ctx, 'a> {
self.state
.push1_extra(res, info | ExtraInfo::pending_f32_nan());
}
Operator::F32x4Ceil => {
let (v, i) = self.state.pop1_extra()?;
let (v, _) = self.v128_into_f32x4(v, i);
let res = self
.builder
.build_call(self.intrinsics.ceil_f32x4, &[v.as_basic_value_enum()], "")
.try_as_basic_value()
.left()
.unwrap();
let res = self.builder.build_bitcast(res, self.intrinsics.i128_ty, "");
self.state
.push1_extra(res, i | ExtraInfo::pending_f32_nan());
}
Operator::F64Ceil => {
let (input, info) = self.state.pop1_extra()?;
let res = self
@ -4752,6 +4843,19 @@ impl<'ctx, 'a> LLVMFunctionCodeGenerator<'ctx, 'a> {
self.state
.push1_extra(res, info | ExtraInfo::pending_f64_nan());
}
Operator::F64x2Ceil => {
let (v, i) = self.state.pop1_extra()?;
let (v, _) = self.v128_into_f64x2(v, i);
let res = self
.builder
.build_call(self.intrinsics.ceil_f64x2, &[v.as_basic_value_enum()], "")
.try_as_basic_value()
.left()
.unwrap();
let res = self.builder.build_bitcast(res, self.intrinsics.i128_ty, "");
self.state
.push1_extra(res, i | ExtraInfo::pending_f64_nan());
}
Operator::F32Floor => {
let (input, info) = self.state.pop1_extra()?;
let res = self
@ -4763,6 +4867,19 @@ impl<'ctx, 'a> LLVMFunctionCodeGenerator<'ctx, 'a> {
self.state
.push1_extra(res, info | ExtraInfo::pending_f32_nan());
}
Operator::F32x4Floor => {
let (v, i) = self.state.pop1_extra()?;
let (v, _) = self.v128_into_f32x4(v, i);
let res = self
.builder
.build_call(self.intrinsics.floor_f32x4, &[v.as_basic_value_enum()], "")
.try_as_basic_value()
.left()
.unwrap();
let res = self.builder.build_bitcast(res, self.intrinsics.i128_ty, "");
self.state
.push1_extra(res, i | ExtraInfo::pending_f32_nan());
}
Operator::F64Floor => {
let (input, info) = self.state.pop1_extra()?;
let res = self
@ -4774,6 +4891,19 @@ impl<'ctx, 'a> LLVMFunctionCodeGenerator<'ctx, 'a> {
self.state
.push1_extra(res, info | ExtraInfo::pending_f64_nan());
}
Operator::F64x2Floor => {
let (v, i) = self.state.pop1_extra()?;
let (v, _) = self.v128_into_f64x2(v, i);
let res = self
.builder
.build_call(self.intrinsics.floor_f64x2, &[v.as_basic_value_enum()], "")
.try_as_basic_value()
.left()
.unwrap();
let res = self.builder.build_bitcast(res, self.intrinsics.i128_ty, "");
self.state
.push1_extra(res, i | ExtraInfo::pending_f64_nan());
}
Operator::F32Trunc => {
let (v, i) = self.state.pop1_extra()?;
let res = self
@ -4785,6 +4915,19 @@ impl<'ctx, 'a> LLVMFunctionCodeGenerator<'ctx, 'a> {
self.state
.push1_extra(res, i | ExtraInfo::pending_f32_nan());
}
Operator::F32x4Trunc => {
let (v, i) = self.state.pop1_extra()?;
let (v, _) = self.v128_into_f32x4(v, i);
let res = self
.builder
.build_call(self.intrinsics.trunc_f32x4, &[v.as_basic_value_enum()], "")
.try_as_basic_value()
.left()
.unwrap();
let res = self.builder.build_bitcast(res, self.intrinsics.i128_ty, "");
self.state
.push1_extra(res, i | ExtraInfo::pending_f32_nan());
}
Operator::F64Trunc => {
let (v, i) = self.state.pop1_extra()?;
let res = self
@ -4796,6 +4939,19 @@ impl<'ctx, 'a> LLVMFunctionCodeGenerator<'ctx, 'a> {
self.state
.push1_extra(res, i | ExtraInfo::pending_f64_nan());
}
Operator::F64x2Trunc => {
let (v, i) = self.state.pop1_extra()?;
let (v, _) = self.v128_into_f64x2(v, i);
let res = self
.builder
.build_call(self.intrinsics.trunc_f64x2, &[v.as_basic_value_enum()], "")
.try_as_basic_value()
.left()
.unwrap();
let res = self.builder.build_bitcast(res, self.intrinsics.i128_ty, "");
self.state
.push1_extra(res, i | ExtraInfo::pending_f64_nan());
}
Operator::F32Nearest => {
let (v, i) = self.state.pop1_extra()?;
let res = self
@ -4811,6 +4967,23 @@ impl<'ctx, 'a> LLVMFunctionCodeGenerator<'ctx, 'a> {
self.state
.push1_extra(res, i | ExtraInfo::pending_f32_nan());
}
Operator::F32x4Nearest => {
let (v, i) = self.state.pop1_extra()?;
let (v, _) = self.v128_into_f32x4(v, i);
let res = self
.builder
.build_call(
self.intrinsics.nearbyint_f32x4,
&[v.as_basic_value_enum()],
"",
)
.try_as_basic_value()
.left()
.unwrap();
let res = self.builder.build_bitcast(res, self.intrinsics.i128_ty, "");
self.state
.push1_extra(res, i | ExtraInfo::pending_f32_nan());
}
Operator::F64Nearest => {
let (v, i) = self.state.pop1_extra()?;
let res = self
@ -4826,6 +4999,23 @@ impl<'ctx, 'a> LLVMFunctionCodeGenerator<'ctx, 'a> {
self.state
.push1_extra(res, i | ExtraInfo::pending_f64_nan());
}
Operator::F64x2Nearest => {
let (v, i) = self.state.pop1_extra()?;
let (v, _) = self.v128_into_f64x2(v, i);
let res = self
.builder
.build_call(
self.intrinsics.nearbyint_f64x2,
&[v.as_basic_value_enum()],
"",
)
.try_as_basic_value()
.left()
.unwrap();
let res = self.builder.build_bitcast(res, self.intrinsics.i128_ty, "");
self.state
.push1_extra(res, i | ExtraInfo::pending_f64_nan());
}
Operator::F32Abs => {
let (v, i) = self.state.pop1_extra()?;
let v = self.apply_pending_canonicalization(v, i);
@ -6381,37 +6571,25 @@ impl<'ctx, 'a> LLVMFunctionCodeGenerator<'ctx, 'a> {
let (v, i) = self.state.pop1_extra()?;
let v = self.apply_pending_canonicalization(v, i);
let v = v.into_int_value();
let res = self.trunc_sat(
let res = self.trunc_sat_inner(
self.intrinsics.f64x2_ty,
self.intrinsics.i64x2_ty,
self.intrinsics.i32_ty.vec_type(2),
cmp_min,
cmp_max,
min,
max,
v,
);
let res = self
.builder
.build_bitcast(res, self.intrinsics.i32x4_ty, "")
.into_vector_value();
let all_ones = self
.intrinsics
.i32_ty
.const_int(u32::max_value() as u64, false);
let zero = self.intrinsics.i32_ty.const_int(0, false);
let mask = VectorType::const_vector(&[all_ones, zero, all_ones, zero]);
let res = self.builder.build_and(res, mask, "");
/*let res = self.builder.build_int_truncate(res,
self.intrinsics.i32_ty.vec_type(4) , "");*/
let zeros = VectorType::const_vector(&[zero; 2]);
let res = self.builder.build_shuffle_vector(
res,
res.get_type().get_undef(),
zeros,
VectorType::const_vector(&[
self.intrinsics.i32_ty.const_int(0, false),
self.intrinsics.i32_ty.const_int(1, false),
self.intrinsics.i32_ty.const_int(2, false),
// the 3rd element should always be 0
self.intrinsics.i32_ty.const_int(3, false),
self.intrinsics.i32_ty.const_int(3, false),
]),
"",
@ -7683,13 +7861,15 @@ impl<'ctx, 'a> LLVMFunctionCodeGenerator<'ctx, 'a> {
ExtraInfo::arithmetic_f32() | ExtraInfo::arithmetic_f64(),
);
}
Operator::I8x16AllTrue | Operator::I16x8AllTrue | Operator::I32x4AllTrue => {
// | Operator::I64x2AllTrue
Operator::I8x16AllTrue
| Operator::I16x8AllTrue
| Operator::I32x4AllTrue
| Operator::I64x2AllTrue => {
let vec_ty = match op {
Operator::I8x16AllTrue => self.intrinsics.i8x16_ty,
Operator::I16x8AllTrue => self.intrinsics.i16x8_ty,
Operator::I32x4AllTrue => self.intrinsics.i32x4_ty,
// Operator::I64x2AllTrue => self.intrinsics.i64x2_ty,
Operator::I64x2AllTrue => self.intrinsics.i64x2_ty,
_ => unreachable!(),
};
let (v, i) = self.state.pop1_extra()?;