nac3/nac3core/src/codegen/builtin_fns.rs

1346 lines
39 KiB
Rust

use inkwell::{FloatPredicate, IntPredicate};
use inkwell::types::BasicTypeEnum;
use inkwell::values::{BasicValueEnum, FloatValue, IntValue};
use itertools::Itertools;
use crate::codegen::{CodeGenContext, CodeGenerator, extern_fns, irrt, llvm_intrinsics};
use crate::codegen::classes::NDArrayValue;
use crate::codegen::numpy::ndarray_elementwise_unaryop_impl;
use crate::toplevel::helper::PRIMITIVE_DEF_IDS;
use crate::toplevel::numpy::unpack_ndarray_var_tys;
use crate::typecheck::typedef::Type;
/// Shorthand for [`unreachable!()`] when a type of argument is not supported.
///
/// The generated message will contain the function name and the name of the unsupported type.
fn unsupported_type(
ctx: &CodeGenContext<'_, '_>,
fn_name: &str,
tys: &[Type],
) -> ! {
unreachable!(
"{fn_name}() not supported for '{}'",
tys.iter().map(|ty| format!("'{}'", ctx.unifier.stringify(*ty))).join(", "),
)
}
/// Invokes the `int32` builtin function.
pub fn call_int32<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
n: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
let llvm_i32 = ctx.ctx.i32_type();
let llvm_usize = generator.get_size_type(ctx.ctx);
let (n_ty, n) = n;
Ok(match n.get_type() {
BasicTypeEnum::IntType(int_ty) if matches!(int_ty.get_bit_width(), 1 | 8) => {
debug_assert!(ctx.unifier.unioned(n_ty, ctx.primitives.bool));
ctx.builder
.build_int_z_extend(n.into_int_value(), llvm_i32, "zext")
.map(Into::into)
.unwrap()
}
BasicTypeEnum::IntType(int_ty) if int_ty.get_bit_width() == 32 => {
debug_assert!([
ctx.primitives.int32,
ctx.primitives.uint32,
].iter().any(|ty| ctx.unifier.unioned(n_ty, *ty)));
n
}
BasicTypeEnum::IntType(int_ty) if int_ty.get_bit_width() == 64 => {
debug_assert!([
ctx.primitives.int64,
ctx.primitives.uint64,
].iter().any(|ty| ctx.unifier.unioned(n_ty, *ty)));
ctx.builder
.build_int_truncate(n.into_int_value(), llvm_i32, "trunc")
.map(Into::into)
.unwrap()
}
BasicTypeEnum::FloatType(_) => {
debug_assert!(ctx.unifier.unioned(n_ty, ctx.primitives.float));
let to_int64 = ctx.builder
.build_float_to_signed_int(n.into_float_value(), ctx.ctx.i64_type(), "")
.unwrap();
ctx.builder
.build_int_truncate(to_int64, llvm_i32, "conv")
.map(Into::into)
.unwrap()
}
BasicTypeEnum::PointerType(_) if n_ty.obj_id(&ctx.unifier).is_some_and(|id| id == PRIMITIVE_DEF_IDS.ndarray) => {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, n_ty);
let ndarray = ndarray_elementwise_unaryop_impl(
generator,
ctx,
ctx.primitives.int32,
None,
NDArrayValue::from_ptr_val(
n.into_pointer_value(),
llvm_usize,
None,
),
|generator, ctx, val| {
call_int32(
generator,
ctx,
(elem_ty, val),
)
},
)?;
ndarray.as_ptr_value().into()
}
_ => unsupported_type(ctx, "int32", &[n_ty])
})
}
/// Invokes the `int64` builtin function.
pub fn call_int64<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
n: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
let llvm_i64 = ctx.ctx.i64_type();
let llvm_usize = generator.get_size_type(ctx.ctx);
let (n_ty, n) = n;
Ok(match n.get_type() {
BasicTypeEnum::IntType(int_ty) if matches!(int_ty.get_bit_width(), 1 | 8 | 32) => {
debug_assert!([
ctx.primitives.bool,
ctx.primitives.int32,
ctx.primitives.uint32,
].iter().any(|ty| ctx.unifier.unioned(n_ty, *ty)));
if ctx.unifier.unioned(n_ty, ctx.primitives.int32) {
ctx.builder
.build_int_s_extend(n.into_int_value(), llvm_i64, "sext")
.map(Into::into)
.unwrap()
} else {
ctx.builder
.build_int_z_extend(n.into_int_value(), llvm_i64, "zext")
.map(Into::into)
.unwrap()
}
}
BasicTypeEnum::IntType(int_ty) if int_ty.get_bit_width() == 64 => {
debug_assert!([
ctx.primitives.int64,
ctx.primitives.uint64,
].iter().any(|ty| ctx.unifier.unioned(n_ty, *ty)));
n
}
BasicTypeEnum::FloatType(_) => {
debug_assert!(ctx.unifier.unioned(n_ty, ctx.primitives.float));
ctx.builder
.build_float_to_signed_int(n.into_float_value(), ctx.ctx.i64_type(), "fptosi")
.map(Into::into)
.unwrap()
}
BasicTypeEnum::PointerType(_) if n_ty.obj_id(&ctx.unifier).is_some_and(|id| id == PRIMITIVE_DEF_IDS.ndarray) => {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, n_ty);
let ndarray = ndarray_elementwise_unaryop_impl(
generator,
ctx,
ctx.primitives.int64,
None,
NDArrayValue::from_ptr_val(
n.into_pointer_value(),
llvm_usize,
None,
),
|generator, ctx, val| {
call_int64(
generator,
ctx,
(elem_ty, val),
)
},
)?;
ndarray.as_ptr_value().into()
}
_ => unsupported_type(ctx, "int64", &[n_ty])
})
}
/// Invokes the `uint32` builtin function.
pub fn call_uint32<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
n: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
let llvm_i32 = ctx.ctx.i32_type();
let llvm_usize = generator.get_size_type(ctx.ctx);
let (n_ty, n) = n;
Ok(match n.get_type() {
BasicTypeEnum::IntType(int_ty) if matches!(int_ty.get_bit_width(), 1 | 8) => {
debug_assert!(ctx.unifier.unioned(n_ty, ctx.primitives.bool));
ctx.builder
.build_int_z_extend(n.into_int_value(), llvm_i32, "zext")
.map(Into::into)
.unwrap()
}
BasicTypeEnum::IntType(int_ty) if int_ty.get_bit_width() == 32 => {
debug_assert!([
ctx.primitives.int32,
ctx.primitives.uint32,
].iter().any(|ty| ctx.unifier.unioned(n_ty, *ty)));
n
}
BasicTypeEnum::IntType(int_ty) if int_ty.get_bit_width() == 64 => {
debug_assert!(
ctx.unifier.unioned(n_ty, ctx.primitives.int64)
|| ctx.unifier.unioned(n_ty, ctx.primitives.uint64)
);
ctx.builder
.build_int_truncate(n.into_int_value(), llvm_i32, "trunc")
.map(Into::into)
.unwrap()
}
BasicTypeEnum::FloatType(_) => {
debug_assert!(ctx.unifier.unioned(n_ty, ctx.primitives.float));
let val = n.into_float_value();
let val_gez = ctx.builder
.build_float_compare(FloatPredicate::OGE, val, val.get_type().const_zero(), "")
.unwrap();
let to_int32 = ctx.builder
.build_float_to_signed_int(val, llvm_i32, "")
.unwrap();
let to_uint64 = ctx.builder
.build_float_to_unsigned_int(val, ctx.ctx.i64_type(), "")
.unwrap();
ctx.builder
.build_select(
val_gez,
ctx.builder.build_int_truncate(to_uint64, llvm_i32, "").unwrap(),
to_int32,
"conv",
)
.unwrap()
}
BasicTypeEnum::PointerType(_) if n_ty.obj_id(&ctx.unifier).is_some_and(|id| id == PRIMITIVE_DEF_IDS.ndarray) => {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, n_ty);
let ndarray = ndarray_elementwise_unaryop_impl(
generator,
ctx,
ctx.primitives.uint32,
None,
NDArrayValue::from_ptr_val(
n.into_pointer_value(),
llvm_usize,
None,
),
|generator, ctx, val| {
call_uint32(
generator,
ctx,
(elem_ty, val),
)
},
)?;
ndarray.as_ptr_value().into()
}
_ => unsupported_type(ctx, "uint32", &[n_ty])
})
}
/// Invokes the `uint64` builtin function.
pub fn call_uint64<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
n: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
let llvm_i64 = ctx.ctx.i64_type();
let llvm_usize = generator.get_size_type(ctx.ctx);
let (n_ty, n) = n;
Ok(match n.get_type() {
BasicTypeEnum::IntType(int_ty) if matches!(int_ty.get_bit_width(), 1 | 8 | 32) => {
debug_assert!([
ctx.primitives.bool,
ctx.primitives.int32,
ctx.primitives.uint32,
].iter().any(|ty| ctx.unifier.unioned(n_ty, *ty)));
if ctx.unifier.unioned(n_ty, ctx.primitives.int32) {
ctx.builder
.build_int_s_extend(n.into_int_value(), llvm_i64, "sext")
.map(Into::into)
.unwrap()
} else {
ctx.builder
.build_int_z_extend(n.into_int_value(), llvm_i64, "zext")
.map(Into::into)
.unwrap()
}
}
BasicTypeEnum::IntType(int_ty) if int_ty.get_bit_width() == 64 => {
debug_assert!([
ctx.primitives.int64,
ctx.primitives.uint64,
].iter().any(|ty| ctx.unifier.unioned(n_ty, *ty)));
n
}
BasicTypeEnum::FloatType(_) => {
debug_assert!(ctx.unifier.unioned(n_ty, ctx.primitives.float));
let val = n.into_float_value();
let val_gez = ctx.builder
.build_float_compare(FloatPredicate::OGE, val, val.get_type().const_zero(), "")
.unwrap();
let to_int64 = ctx.builder
.build_float_to_signed_int(val, llvm_i64, "")
.unwrap();
let to_uint64 = ctx.builder
.build_float_to_unsigned_int(val, llvm_i64, "")
.unwrap();
ctx.builder
.build_select(val_gez, to_uint64, to_int64, "conv")
.unwrap()
}
BasicTypeEnum::PointerType(_) if n_ty.obj_id(&ctx.unifier).is_some_and(|id| id == PRIMITIVE_DEF_IDS.ndarray) => {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, n_ty);
let ndarray = ndarray_elementwise_unaryop_impl(
generator,
ctx,
ctx.primitives.uint64,
None,
NDArrayValue::from_ptr_val(
n.into_pointer_value(),
llvm_usize,
None,
),
|generator, ctx, val| {
call_uint64(
generator,
ctx,
(elem_ty, val),
)
},
)?;
ndarray.as_ptr_value().into()
}
_ => unsupported_type(ctx, "uint64", &[n_ty])
})
}
/// Invokes the `float` builtin function.
pub fn call_float<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
n: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
let llvm_f64 = ctx.ctx.f64_type();
let llvm_usize = generator.get_size_type(ctx.ctx);
let (n_ty, n) = n;
Ok(match n.get_type() {
BasicTypeEnum::IntType(int_ty) if matches!(int_ty.get_bit_width(), 1 | 8 | 32 | 64) => {
debug_assert!([
ctx.primitives.bool,
ctx.primitives.int32,
ctx.primitives.uint32,
ctx.primitives.int64,
ctx.primitives.uint64,
].iter().any(|ty| ctx.unifier.unioned(n_ty, *ty)));
if [
ctx.primitives.bool,
ctx.primitives.int32,
ctx.primitives.int64,
].iter().any(|ty| ctx.unifier.unioned(n_ty, *ty)) {
ctx.builder
.build_signed_int_to_float(n.into_int_value(), llvm_f64, "sitofp")
.map(Into::into)
.unwrap()
} else {
ctx.builder
.build_unsigned_int_to_float(n.into_int_value(), llvm_f64, "uitofp")
.map(Into::into)
.unwrap()
}
}
BasicTypeEnum::FloatType(_) => {
debug_assert!(ctx.unifier.unioned(n_ty, ctx.primitives.float));
n
}
BasicTypeEnum::PointerType(_) if n_ty.obj_id(&ctx.unifier).is_some_and(|id| id == PRIMITIVE_DEF_IDS.ndarray) => {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, n_ty);
let ndarray = ndarray_elementwise_unaryop_impl(
generator,
ctx,
ctx.primitives.float,
None,
NDArrayValue::from_ptr_val(
n.into_pointer_value(),
llvm_usize,
None,
),
|generator, ctx, val| {
call_float(
generator,
ctx,
(elem_ty, val),
)
},
)?;
ndarray.as_ptr_value().into()
}
_ => unsupported_type(ctx, "float", &[n_ty])
})
}
/// Invokes the `round` builtin function.
pub fn call_round<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
n: (Type, BasicValueEnum<'ctx>),
ret_ty: Type,
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "round";
let llvm_usize = generator.get_size_type(ctx.ctx);
let (n_ty, n) = n;
let llvm_ret_ty = ctx.get_llvm_abi_type(generator, ret_ty).into_int_type();
Ok(match n.get_type() {
BasicTypeEnum::FloatType(_) => {
debug_assert!(ctx.unifier.unioned(n_ty, ctx.primitives.float));
let val = llvm_intrinsics::call_float_round(ctx, n.into_float_value(), None);
ctx.builder
.build_float_to_signed_int(val, llvm_ret_ty, FN_NAME)
.map(Into::into)
.unwrap()
}
BasicTypeEnum::PointerType(_) if n_ty.obj_id(&ctx.unifier).is_some_and(|id| id == PRIMITIVE_DEF_IDS.ndarray) => {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, n_ty);
let ndarray = ndarray_elementwise_unaryop_impl(
generator,
ctx,
ret_ty,
None,
NDArrayValue::from_ptr_val(
n.into_pointer_value(),
llvm_usize,
None,
),
|generator, ctx, val| {
call_round(
generator,
ctx,
(elem_ty, val),
ret_ty,
)
},
)?;
ndarray.as_ptr_value().into()
}
_ => unsupported_type(ctx, FN_NAME, &[n_ty])
})
}
/// Invokes the `np_round` builtin function.
pub fn call_numpy_round<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
n: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (n_ty, n) = n;
if !ctx.unifier.unioned(n_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_round", &[n_ty])
}
llvm_intrinsics::call_float_roundeven(ctx, n, None)
}
/// Invokes the `bool` builtin function.
pub fn call_bool<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
n: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "bool";
let llvm_usize = generator.get_size_type(ctx.ctx);
let (n_ty, n) = n;
Ok(match n.get_type() {
BasicTypeEnum::IntType(int_ty) if matches!(int_ty.get_bit_width(), 1 | 8) => {
debug_assert!(ctx.unifier.unioned(n_ty, ctx.primitives.bool));
n
}
BasicTypeEnum::IntType(_) => {
debug_assert!([
ctx.primitives.int32,
ctx.primitives.uint32,
ctx.primitives.int64,
ctx.primitives.uint64,
].iter().any(|ty| ctx.unifier.unioned(n_ty, *ty)));
let val = n.into_int_value();
ctx.builder
.build_int_compare(IntPredicate::NE, val, val.get_type().const_zero(), FN_NAME)
.map(Into::into)
.unwrap()
}
BasicTypeEnum::FloatType(_) => {
debug_assert!(ctx.unifier.unioned(n_ty, ctx.primitives.float));
let val = n.into_float_value();
ctx.builder
.build_float_compare(FloatPredicate::UNE, val, val.get_type().const_zero(), FN_NAME)
.map(Into::into)
.unwrap()
}
BasicTypeEnum::PointerType(_) if n_ty.obj_id(&ctx.unifier).is_some_and(|id| id == PRIMITIVE_DEF_IDS.ndarray) => {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, n_ty);
let ndarray = ndarray_elementwise_unaryop_impl(
generator,
ctx,
ctx.primitives.bool,
None,
NDArrayValue::from_ptr_val(
n.into_pointer_value(),
llvm_usize,
None,
),
|generator, ctx, val| {
let elem = call_bool(
generator,
ctx,
(elem_ty, val),
)?;
Ok(generator.bool_to_i8(ctx, elem.into_int_value()).into())
},
)?;
ndarray.as_ptr_value().into()
}
_ => unsupported_type(ctx, FN_NAME, &[n_ty])
})
}
/// Invokes the `floor` builtin function.
pub fn call_floor<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
n: (Type, FloatValue<'ctx>),
llvm_ret_ty: BasicTypeEnum<'ctx>,
) -> BasicValueEnum<'ctx> {
const FN_NAME: &str = "floor";
let (n_ty, n) = n;
debug_assert!(ctx.unifier.unioned(n_ty, ctx.primitives.float));
let val = llvm_intrinsics::call_float_floor(ctx, n, None);
match llvm_ret_ty {
_ if llvm_ret_ty == val.get_type().into() => val.into(),
BasicTypeEnum::IntType(_) => {
ctx.builder
.build_float_to_signed_int(val, llvm_ret_ty.into_int_type(), FN_NAME)
.map(Into::into)
.unwrap()
}
_ => unsupported_type(ctx, FN_NAME, &[n_ty])
}
}
/// Invokes the `ceil` builtin function.
pub fn call_ceil<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
n: (Type, FloatValue<'ctx>),
llvm_ret_ty: BasicTypeEnum<'ctx>,
) -> BasicValueEnum<'ctx> {
const FN_NAME: &str = "ceil";
let (n_ty, n) = n;
debug_assert!(ctx.unifier.unioned(n_ty, ctx.primitives.float));
let val = llvm_intrinsics::call_float_ceil(ctx, n, None);
match llvm_ret_ty {
_ if llvm_ret_ty == val.get_type().into() => val.into(),
BasicTypeEnum::IntType(_) => {
ctx.builder
.build_float_to_signed_int(val, llvm_ret_ty.into_int_type(), FN_NAME)
.map(Into::into)
.unwrap()
}
_ => unsupported_type(ctx, FN_NAME, &[n_ty])
}
}
/// Invokes the `min` builtin function.
pub fn call_min<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
m: (Type, BasicValueEnum<'ctx>),
n: (Type, BasicValueEnum<'ctx>),
) -> BasicValueEnum<'ctx> {
const FN_NAME: &str = "min";
let (m_ty, m) = m;
let (n_ty, n) = n;
if !ctx.unifier.unioned(m_ty, n_ty) {
unsupported_type(ctx, FN_NAME, &[m_ty, n_ty])
}
debug_assert_eq!(m.get_type(), n.get_type());
let common_ty = m_ty;
let llvm_common_ty = m.get_type();
match llvm_common_ty {
BasicTypeEnum::IntType(_) => {
debug_assert!([
ctx.primitives.bool,
ctx.primitives.int32,
ctx.primitives.uint32,
ctx.primitives.int64,
ctx.primitives.uint64,
].iter().any(|ty| ctx.unifier.unioned(common_ty, *ty)));
let (m, n) = (m.into_int_value(), n.into_int_value());
if [
ctx.primitives.int32,
ctx.primitives.int64,
].iter().any(|ty| ctx.unifier.unioned(common_ty, *ty)) {
llvm_intrinsics::call_int_smin(ctx, m, n, Some(FN_NAME)).into()
} else {
llvm_intrinsics::call_int_umin(ctx, m, n, Some(FN_NAME)).into()
}
}
BasicTypeEnum::FloatType(_) => {
debug_assert!(ctx.unifier.unioned(common_ty, ctx.primitives.float));
let (m, n) = (m.into_float_value(), n.into_float_value());
llvm_intrinsics::call_float_minnum(ctx, m, n, Some(FN_NAME)).into()
}
_ => unsupported_type(ctx, FN_NAME, &[m_ty, n_ty])
}
}
/// Invokes the `max` builtin function.
pub fn call_max<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
m: (Type, BasicValueEnum<'ctx>),
n: (Type, BasicValueEnum<'ctx>),
) -> BasicValueEnum<'ctx> {
const FN_NAME: &str = "max";
let (m_ty, m) = m;
let (n_ty, n) = n;
if !ctx.unifier.unioned(m_ty, n_ty) {
unsupported_type(ctx, FN_NAME, &[m_ty, n_ty])
}
debug_assert_eq!(m.get_type(), n.get_type());
let common_ty = m_ty;
let llvm_common_ty = m.get_type();
match llvm_common_ty {
BasicTypeEnum::IntType(_) => {
debug_assert!([
ctx.primitives.bool,
ctx.primitives.int32,
ctx.primitives.uint32,
ctx.primitives.int64,
ctx.primitives.uint64,
].iter().any(|ty| ctx.unifier.unioned(common_ty, *ty)));
let (m, n) = (m.into_int_value(), n.into_int_value());
if [
ctx.primitives.int32,
ctx.primitives.int64,
].iter().any(|ty| ctx.unifier.unioned(common_ty, *ty)) {
llvm_intrinsics::call_int_smax(ctx, m, n, Some(FN_NAME)).into()
} else {
llvm_intrinsics::call_int_umax(ctx, m, n, Some(FN_NAME)).into()
}
}
BasicTypeEnum::FloatType(_) => {
debug_assert!(ctx.unifier.unioned(common_ty, ctx.primitives.float));
let (m, n) = (m.into_float_value(), n.into_float_value());
llvm_intrinsics::call_float_maxnum(ctx, m, n, Some(FN_NAME)).into()
}
_ => unsupported_type(ctx, FN_NAME, &[m_ty, n_ty])
}
}
/// Invokes the `abs` builtin function.
pub fn call_abs<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
n: (Type, BasicValueEnum<'ctx>),
) -> BasicValueEnum<'ctx> {
const FN_NAME: &str = "abs";
let llvm_i1 = ctx.ctx.bool_type();
let (n_ty, n) = n;
match n.get_type() {
BasicTypeEnum::IntType(_) => {
debug_assert!([
ctx.primitives.bool,
ctx.primitives.int32,
ctx.primitives.uint32,
ctx.primitives.int64,
ctx.primitives.uint64,
].iter().any(|ty| ctx.unifier.unioned(n_ty, *ty)));
if [
ctx.primitives.int32,
ctx.primitives.int64,
].iter().any(|ty| ctx.unifier.unioned(n_ty, *ty)) {
llvm_intrinsics::call_int_abs(
ctx,
n.into_int_value(),
llvm_i1.const_zero(),
Some(FN_NAME),
).into()
} else {
n
}
}
BasicTypeEnum::FloatType(_) => {
debug_assert!(ctx.unifier.unioned(n_ty, ctx.primitives.float));
llvm_intrinsics::call_float_fabs(ctx, n.into_float_value(), Some(FN_NAME)).into()
}
_ => unsupported_type(ctx, FN_NAME, &[n_ty])
}
}
/// Invokes the `np_isnan` builtin function.
pub fn call_numpy_isnan<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> IntValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_isnan", &[x_ty])
}
irrt::call_isnan(generator, ctx, x)
}
/// Invokes the `np_isinf` builtin function.
pub fn call_numpy_isinf<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> IntValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_isinf", &[x_ty])
}
irrt::call_isinf(generator, ctx, x)
}
/// Invokes the `np_sin` builtin function.
pub fn call_numpy_sin<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_sin", &[x_ty])
}
llvm_intrinsics::call_float_sin(ctx, x, None)
}
/// Invokes the `np_cos` builtin function.
pub fn call_numpy_cos<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_cos", &[x_ty])
}
llvm_intrinsics::call_float_cos(ctx, x, None)
}
/// Invokes the `np_exp` builtin function.
pub fn call_numpy_exp<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_exp", &[x_ty])
}
llvm_intrinsics::call_float_exp(ctx, x, None)
}
/// Invokes the `np_exp2` builtin function.
pub fn call_numpy_exp2<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_exp2", &[x_ty])
}
llvm_intrinsics::call_float_exp2(ctx, x, None)
}
/// Invokes the `np_log` builtin function.
pub fn call_numpy_log<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_log", &[x_ty])
}
llvm_intrinsics::call_float_log(ctx, x, None)
}
/// Invokes the `np_log10` builtin function.
pub fn call_numpy_log10<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_log10", &[x_ty])
}
llvm_intrinsics::call_float_log10(ctx, x, None)
}
/// Invokes the `np_log2` builtin function.
pub fn call_numpy_log2<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_log2", &[x_ty])
}
llvm_intrinsics::call_float_log2(ctx, x, None)
}
/// Invokes the `np_sqrt` builtin function.
pub fn call_numpy_fabs<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_fabs", &[x_ty])
}
llvm_intrinsics::call_float_fabs(ctx, x, None)
}
/// Invokes the `np_sqrt` builtin function.
pub fn call_numpy_sqrt<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_sqrt", &[x_ty])
}
llvm_intrinsics::call_float_sqrt(ctx, x, None)
}
/// Invokes the `np_rint` builtin function.
pub fn call_numpy_rint<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_rint", &[x_ty])
}
llvm_intrinsics::call_float_roundeven(ctx, x, None)
}
/// Invokes the `np_tan` builtin function.
pub fn call_numpy_tan<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_tan", &[x_ty])
}
extern_fns::call_tan(ctx, x, None)
}
/// Invokes the `np_arcsin` builtin function.
pub fn call_numpy_arcsin<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_arcsin", &[x_ty])
}
extern_fns::call_asin(ctx, x, None)
}
/// Invokes the `np_arccos` builtin function.
pub fn call_numpy_arccos<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_arccos", &[x_ty])
}
extern_fns::call_acos(ctx, x, None)
}
/// Invokes the `np_arctan` builtin function.
pub fn call_numpy_arctan<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_arctan", &[x_ty])
}
extern_fns::call_atan(ctx, x, None)
}
/// Invokes the `np_sinh` builtin function.
pub fn call_numpy_sinh<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_sinh", &[x_ty])
}
extern_fns::call_sinh(ctx, x, None)
}
/// Invokes the `np_cosh` builtin function.
pub fn call_numpy_cosh<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_cosh", &[x_ty])
}
extern_fns::call_cosh(ctx, x, None)
}
/// Invokes the `np_tanh` builtin function.
pub fn call_numpy_tanh<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_tanh", &[x_ty])
}
extern_fns::call_tanh(ctx, x, None)
}
/// Invokes the `np_asinh` builtin function.
pub fn call_numpy_asinh<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_asinh", &[x_ty])
}
extern_fns::call_asinh(ctx, x, None)
}
/// Invokes the `np_acosh` builtin function.
pub fn call_numpy_acosh<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_acosh", &[x_ty])
}
extern_fns::call_acosh(ctx, x, None)
}
/// Invokes the `np_atanh` builtin function.
pub fn call_numpy_atanh<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_atanh", &[x_ty])
}
extern_fns::call_atanh(ctx, x, None)
}
/// Invokes the `np_expm1` builtin function.
pub fn call_numpy_expm1<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_expm1", &[x_ty])
}
extern_fns::call_expm1(ctx, x, None)
}
/// Invokes the `np_cbrt` builtin function.
pub fn call_numpy_cbrt<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "np_cbrt", &[x_ty])
}
extern_fns::call_cbrt(ctx, x, None)
}
/// Invokes the `sp_spec_erf` builtin function.
pub fn call_scipy_special_erf<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
z: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (z_ty, z) = z;
if !ctx.unifier.unioned(z_ty, ctx.primitives.float) {
unsupported_type(ctx, "sp_spec_erf", &[z_ty])
}
extern_fns::call_erf(ctx, z, None)
}
/// Invokes the `sp_spec_erfc` builtin function.
pub fn call_scipy_special_erfc<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "sp_spec_erfc", &[x_ty])
}
extern_fns::call_erfc(ctx, x, None)
}
/// Invokes the `sp_spec_gamma` builtin function.
pub fn call_scipy_special_gamma<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
z: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (z_ty, z) = z;
if !ctx.unifier.unioned(z_ty, ctx.primitives.float) {
unsupported_type(ctx, "sp_spec_gamma", &[z_ty])
}
irrt::call_gamma(ctx, z)
}
/// Invokes the `sp_spec_gammaln` builtin function.
pub fn call_scipy_special_gammaln<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "sp_spec_gammaln", &[x_ty])
}
irrt::call_gammaln(ctx, x)
}
/// Invokes the `sp_spec_j0` builtin function.
pub fn call_scipy_special_j0<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "sp_spec_j0", &[x_ty])
}
irrt::call_j0(ctx, x)
}
/// Invokes the `sp_spec_j1` builtin function.
pub fn call_scipy_special_j1<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let (x_ty, x) = x;
if !ctx.unifier.unioned(x_ty, ctx.primitives.float) {
unsupported_type(ctx, "sp_spec_j1", &[x_ty])
}
extern_fns::call_j1(ctx, x, None)
}
/// Invokes the `np_arctan2` builtin function.
pub fn call_numpy_arctan2<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, FloatValue<'ctx>),
x2: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let float_t = ctx.primitives.float;
let (x1_ty, x1) = x1;
let (x2_ty, x2) = x2;
if !ctx.unifier.unioned(x1_ty, float_t) || !ctx.unifier.unioned(x2_ty, float_t) {
unsupported_type(ctx, "np_atan2", &[x1_ty, x2_ty])
}
extern_fns::call_atan2(ctx, x1, x2, None)
}
/// Invokes the `np_copysign` builtin function.
pub fn call_numpy_copysign<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, FloatValue<'ctx>),
x2: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let float_t = ctx.primitives.float;
let (x1_ty, x1) = x1;
let (x2_ty, x2) = x2;
debug_assert_eq!(x1.get_type(), x2.get_type());
if !ctx.unifier.unioned(x1_ty, float_t) || !ctx.unifier.unioned(x2_ty, float_t) {
unsupported_type(ctx, "np_copysign", &[x1_ty, x2_ty])
}
llvm_intrinsics::call_float_copysign(ctx, x1, x2, None)
}
/// Invokes the `np_fmax` builtin function.
pub fn call_numpy_fmax<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, FloatValue<'ctx>),
x2: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let float_t = ctx.primitives.float;
let (x1_ty, x1) = x1;
let (x2_ty, x2) = x2;
debug_assert_eq!(x1.get_type(), x2.get_type());
if !ctx.unifier.unioned(x1_ty, float_t) || !ctx.unifier.unioned(x2_ty, float_t) {
unsupported_type(ctx, "np_fmax", &[x1_ty, x2_ty])
}
llvm_intrinsics::call_float_maxnum(ctx, x1, x2, None)
}
/// Invokes the `np_fmin` builtin function.
pub fn call_numpy_fmin<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, FloatValue<'ctx>),
x2: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let float_t = ctx.primitives.float;
let (x1_ty, x1) = x1;
let (x2_ty, x2) = x2;
debug_assert_eq!(x1.get_type(), x2.get_type());
if !ctx.unifier.unioned(x1_ty, float_t) || !ctx.unifier.unioned(x2_ty, float_t) {
unsupported_type(ctx, "np_fmin", &[x1_ty, x2_ty])
}
llvm_intrinsics::call_float_minnum(ctx, x1, x2, None)
}
/// Invokes the `np_ldexp` builtin function.
pub fn call_numpy_ldexp<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, FloatValue<'ctx>),
x2: (Type, IntValue<'ctx>),
) -> FloatValue<'ctx> {
let (x1_ty, x1) = x1;
let (x2_ty, x2) = x2;
if !ctx.unifier.unioned(x1_ty, ctx.primitives.float) {
unsupported_type(ctx, "fp_ldexp", &[x1_ty, x2_ty])
}
if !ctx.unifier.unioned(x2_ty, ctx.primitives.int32) {
unsupported_type(ctx, "fp_ldexp", &[x1_ty, x2_ty])
}
extern_fns::call_ldexp(ctx, x1, x2, None)
}
/// Invokes the `np_hypot` builtin function.
pub fn call_numpy_hypot<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, FloatValue<'ctx>),
x2: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let float_t = ctx.primitives.float;
let (x1_ty, x1) = x1;
let (x2_ty, x2) = x2;
if !ctx.unifier.unioned(x1_ty, float_t) || !ctx.unifier.unioned(x2_ty, float_t) {
unsupported_type(ctx, "np_hypot", &[x1_ty, x2_ty])
}
extern_fns::call_hypot(ctx, x1, x2, None)
}
/// Invokes the `np_nextafter` builtin function.
pub fn call_numpy_nextafter<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, FloatValue<'ctx>),
x2: (Type, FloatValue<'ctx>),
) -> FloatValue<'ctx> {
let float_t = ctx.primitives.float;
let (x1_ty, x1) = x1;
let (x2_ty, x2) = x2;
if !ctx.unifier.unioned(x1_ty, float_t) || !ctx.unifier.unioned(x2_ty, float_t) {
unsupported_type(ctx, "np_nextafter", &[x1_ty, x2_ty])
}
extern_fns::call_nextafter(ctx, x1, x2, None)
}