1
0
forked from M-Labs/nac3

core: add linalg methods

This commit is contained in:
abdul124 2024-07-25 12:16:53 +08:00
parent 318a675ea6
commit 2242c5af43
5 changed files with 743 additions and 3 deletions

View File

@ -1,5 +1,5 @@
use inkwell::types::BasicTypeEnum;
use inkwell::values::BasicValueEnum;
use inkwell::values::{BasicValue, BasicValueEnum, PointerValue};
use inkwell::{FloatPredicate, IntPredicate, OptimizationLevel};
use itertools::Itertools;
@ -31,7 +31,6 @@ pub fn call_int32<'ctx, G: CodeGenerator + ?Sized>(
let llvm_usize = generator.get_size_type(ctx.ctx);
let (n_ty, n) = n;
Ok(match n {
BasicValueEnum::IntValue(n) if matches!(n.get_type().get_bit_width(), 1 | 8) => {
debug_assert!(ctx.unifier.unioned(n_ty, ctx.primitives.bool));
@ -1836,3 +1835,480 @@ pub fn call_numpy_nextafter<'ctx, G: CodeGenerator + ?Sized>(
_ => unsupported_type(ctx, FN_NAME, &[x1_ty, x2_ty]),
})
}
/// Allocates a struct with the fields specified by `out_matrices` and returns a pointer to it
fn build_output_struct<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
out_matrices: Vec<BasicValueEnum<'ctx>>,
) -> PointerValue<'ctx> {
let field_ty =
out_matrices.iter().map(BasicValueEnum::get_type).collect::<Vec<BasicTypeEnum>>();
let out_ty = ctx.ctx.struct_type(&field_ty, false);
let out_ptr = ctx.builder.build_alloca(out_ty, "").unwrap();
for (i, v) in out_matrices.into_iter().enumerate() {
unsafe {
let ptr = ctx
.builder
.build_in_bounds_gep(
out_ptr,
&[
ctx.ctx.i32_type().const_zero(),
ctx.ctx.i32_type().const_int(i as u64, false),
],
"",
)
.unwrap();
ctx.builder.build_store(ptr, v).unwrap();
}
}
out_ptr
}
/// Invokes the `np_dot` using `nalgebra` crate
pub fn call_np_dot<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
x2: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "np_dot";
let (x1_ty, x1) = x1;
let (x2_ty, x2) = x2;
if let (BasicValueEnum::PointerValue(_), BasicValueEnum::PointerValue(_)) = (x1, x2) {
let (n1_elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, n1_elem_ty);
let (n2_elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x2_ty);
let n2_elem_ty = ctx.get_llvm_type(generator, n2_elem_ty);
let (BasicTypeEnum::FloatType(_), BasicTypeEnum::FloatType(_)) = (n1_elem_ty, n2_elem_ty)
else {
unimplemented!("{FN_NAME} operates on float type NdArrays only");
};
Ok(extern_fns::call_np_dot(ctx, x1, x2, None).into())
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty, x2_ty])
}
}
/// Invokes the `np_linalg_matmul` using `nalgebra` crate
pub fn call_np_linalg_matmul<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
x2: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "np_linalg_matmul";
let (x1_ty, x1) = x1;
let (x2_ty, x2) = x2;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let (BasicValueEnum::PointerValue(n1), BasicValueEnum::PointerValue(n2)) = (x1, x2) {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let (n2_elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x2_ty);
let n2_elem_ty = ctx.get_llvm_type(generator, n2_elem_ty);
let (BasicTypeEnum::FloatType(_), BasicTypeEnum::FloatType(_)) = (n1_elem_ty, n2_elem_ty)
else {
unimplemented!("{FN_NAME} operates on float type NdArrays only");
};
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let n2 = NDArrayValue::from_ptr_val(n2, llvm_usize, None);
let outdim0 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
.into_int_value()
};
let outdim1 = unsafe {
n2.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_int(1, false), None)
.into_int_value()
};
let out = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[outdim0, outdim1])
.unwrap()
.as_base_value()
.as_basic_value_enum();
extern_fns::call_np_linalg_matmul(ctx, x1, x2, out, None);
Ok(out)
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty, x2_ty])
}
}
/// Invokes the `np_linalg_cholesky` using `nalgebra` crate
pub fn call_np_linalg_cholesky<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "np_linalg_cholesky";
let (x1_ty, x1) = x1;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(n1) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let BasicTypeEnum::FloatType(_) = n1_elem_ty else {
unimplemented!("{FN_NAME} operates on float type NdArrays only");
};
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let dim0 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
.into_int_value()
};
let dim1 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_int(1, false), None)
.into_int_value()
};
let out = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim0, dim1])
.unwrap()
.as_base_value()
.as_basic_value_enum();
extern_fns::call_np_linalg_cholesky(ctx, x1, out, None);
Ok(out)
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty])
}
}
/// Invokes the `np_linalg_qr` using `nalgebra` crate
pub fn call_np_linalg_qr<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "np_linalg_qr";
let (x1_ty, x1) = x1;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(n1) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let BasicTypeEnum::FloatType(_) = n1_elem_ty else {
unimplemented!("{FN_NAME} operates on float type NdArrays only");
};
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let dim0 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
.into_int_value()
};
let dim1 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_int(1, false), None)
.into_int_value()
};
let k = llvm_intrinsics::call_int_smin(ctx, dim0, dim1, None);
let out_q = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim0, k])
.unwrap()
.as_base_value()
.as_basic_value_enum();
let out_r = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[k, dim1])
.unwrap()
.as_base_value()
.as_basic_value_enum();
extern_fns::call_np_linalg_qr(ctx, x1, out_q, out_r, None);
let out_ptr = build_output_struct(ctx, vec![out_q, out_r]);
Ok(ctx.builder.build_load(out_ptr, "QR_Factorization_result").map(Into::into).unwrap())
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty])
}
}
/// Invokes the `np_linalg_svd` using `nalgebra` crate
pub fn call_np_linalg_svd<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "np_linalg_svd";
let (x1_ty, x1) = x1;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(n1) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let BasicTypeEnum::FloatType(_) = n1_elem_ty else {
unimplemented!("{FN_NAME} operates on float type NdArrays only");
};
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let dim0 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
.into_int_value()
};
let dim1 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_int(1, false), None)
.into_int_value()
};
let k = llvm_intrinsics::call_int_smin(ctx, dim0, dim1, None);
let out_u = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim0, dim0])
.unwrap()
.as_base_value()
.as_basic_value_enum();
let out_s = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[k])
.unwrap()
.as_base_value()
.as_basic_value_enum();
let out_vh = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim1, dim1])
.unwrap()
.as_base_value()
.as_basic_value_enum();
extern_fns::call_np_linalg_svd(ctx, x1, out_u, out_s, out_vh, None);
let out_ptr = build_output_struct(ctx, vec![out_u, out_s, out_vh]);
Ok(ctx.builder.build_load(out_ptr, "SVD_Factorization_result").map(Into::into).unwrap())
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty])
}
}
/// Invokes the `np_linalg_inv` using `nalgebra` crate
pub fn call_np_linalg_inv<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "np_linalg_inv";
let (x1_ty, x1) = x1;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(n1) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let BasicTypeEnum::FloatType(_) = n1_elem_ty else {
unimplemented!("{FN_NAME} operates on float type NdArrays only");
};
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let dim0 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
.into_int_value()
};
let dim1 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_int(1, false), None)
.into_int_value()
};
let out = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim0, dim1])
.unwrap()
.as_base_value()
.as_basic_value_enum();
extern_fns::call_np_linalg_inv(ctx, x1, out, None);
Ok(out)
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty])
}
}
/// Invokes the `np_linalg_pinv` using `nalgebra` crate
pub fn call_np_linalg_pinv<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "np_linalg_pinv";
let (x1_ty, x1) = x1;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(n1) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let BasicTypeEnum::FloatType(_) = n1_elem_ty else {
unimplemented!("{FN_NAME} operates on float type NdArrays only");
};
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let dim0 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
.into_int_value()
};
let dim1 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_int(1, false), None)
.into_int_value()
};
let out = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim1, dim0])
.unwrap()
.as_base_value()
.as_basic_value_enum();
extern_fns::call_np_linalg_pinv(ctx, x1, out, None);
Ok(out)
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty])
}
}
/// Invokes the `sp_linalg_lu` using `nalgebra` crate
pub fn call_sp_linalg_lu<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "sp_linalg_lu";
let (x1_ty, x1) = x1;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(n1) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let BasicTypeEnum::FloatType(_) = n1_elem_ty else {
unimplemented!("{FN_NAME} operates on float type NdArrays only");
};
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let dim0 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
.into_int_value()
};
let dim1 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_int(1, false), None)
.into_int_value()
};
let k = llvm_intrinsics::call_int_smin(ctx, dim0, dim1, None);
let out_l = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim0, k])
.unwrap()
.as_base_value()
.as_basic_value_enum();
let out_u = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[k, dim1])
.unwrap()
.as_base_value()
.as_basic_value_enum();
extern_fns::call_sp_linalg_lu(ctx, x1, out_l, out_u, None);
let out_ptr = build_output_struct(ctx, vec![out_l, out_u]);
Ok(ctx.builder.build_load(out_ptr, "LU_Factorization_result").map(Into::into).unwrap())
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty])
}
}
/// Invokes the `sp_linalg_schur` using `nalgebra` crate
pub fn call_sp_linalg_schur<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "sp_linalg_schur";
let (x1_ty, x1) = x1;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(n1) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let BasicTypeEnum::FloatType(_) = n1_elem_ty else {
unimplemented!("{FN_NAME} operates on float type NdArrays only");
};
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let dim0 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
.into_int_value()
};
let out_t = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim0, dim0])
.unwrap()
.as_base_value()
.as_basic_value_enum();
let out_z = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim0, dim0])
.unwrap()
.as_base_value()
.as_basic_value_enum();
extern_fns::call_sp_linalg_schur(ctx, x1, out_t, out_z, None);
let out_ptr = build_output_struct(ctx, vec![out_t, out_z]);
Ok(ctx.builder.build_load(out_ptr, "Schur_Factorization_result").map(Into::into).unwrap())
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty])
}
}
/// Invokes the `sp_linalg_hessenberg` using `nalgebra` crate
pub fn call_sp_linalg_hessenberg<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
x1: (Type, BasicValueEnum<'ctx>),
) -> Result<BasicValueEnum<'ctx>, String> {
const FN_NAME: &str = "sp_linalg_hessenberg";
let (x1_ty, x1) = x1;
let llvm_usize = generator.get_size_type(ctx.ctx);
if let BasicValueEnum::PointerValue(n1) = x1 {
let (elem_ty, _) = unpack_ndarray_var_tys(&mut ctx.unifier, x1_ty);
let n1_elem_ty = ctx.get_llvm_type(generator, elem_ty);
let BasicTypeEnum::FloatType(_) = n1_elem_ty else {
unimplemented!("{FN_NAME} operates on float type NdArrays only");
};
let n1 = NDArrayValue::from_ptr_val(n1, llvm_usize, None);
let dim0 = unsafe {
n1.dim_sizes()
.get_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
.into_int_value()
};
let out_h = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim0, dim0])
.unwrap()
.as_base_value()
.as_basic_value_enum();
let out_q = numpy::create_ndarray_const_shape(generator, ctx, elem_ty, &[dim0, dim0])
.unwrap()
.as_base_value()
.as_basic_value_enum();
extern_fns::call_sp_linalg_hessenberg(ctx, x1, out_h, out_q, None);
let out_ptr = build_output_struct(ctx, vec![out_h, out_q]);
Ok(ctx
.builder
.build_load(out_ptr, "Hessenberg_decomposition_result")
.map(Into::into)
.unwrap())
} else {
unsupported_type(ctx, FN_NAME, &[x1_ty])
}
}

View File

@ -130,3 +130,91 @@ pub fn call_ldexp<'ctx>(
.map(Either::unwrap_left)
.unwrap()
}
/// Macro to generate `np_linalg` and `sp_linalg` functions
/// The function takes as input `NDArray` and returns ()
///
/// Arguments:
/// * `$fn_name:ident`: The identifier of the rust function to be generated
/// * `$extern_fn:literal`: Name of underlying extern function
/// * (2/3/4): Number of `NDArray` that function takes as input
///
/// Note:
/// The operands and resulting `NDArray` are both passed as input to the funcion
/// It is the responsibility of caller to ensure that output `NDArray` is properly allocated on stack
/// The function changes the content of the output `NDArray` in-place
macro_rules! generate_linalg_extern_fn {
($fn_name:ident, $extern_fn:literal, 2) => {
generate_linalg_extern_fn!($fn_name, $extern_fn, mat1, mat2);
};
($fn_name:ident, $extern_fn:literal, 3) => {
generate_linalg_extern_fn!($fn_name, $extern_fn, mat1, mat2, mat3);
};
($fn_name:ident, $extern_fn:literal, 4) => {
generate_linalg_extern_fn!($fn_name, $extern_fn, mat1, mat2, mat3, mat4);
};
($fn_name:ident, $extern_fn:literal $(,$input_matrix:ident)*) => {
#[doc = concat!("Invokes the linalg `", stringify!($extern_fn), " function." )]
pub fn $fn_name<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>
$(,$input_matrix: BasicValueEnum<'ctx>)*,
name: Option<&str>,
){
const FN_NAME: &str = $extern_fn;
let extern_fn = ctx.module.get_function(FN_NAME).unwrap_or_else(|| {
let fn_type = ctx.ctx.void_type().fn_type(&[$($input_matrix.get_type().into()),*], false);
let func = ctx.module.add_function(FN_NAME, fn_type, None);
for attr in ["mustprogress", "nofree", "nounwind", "willreturn", "writeonly"] {
func.add_attribute(
AttributeLoc::Function,
ctx.ctx.create_enum_attribute(Attribute::get_named_enum_kind_id(attr), 0),
);
}
func
});
ctx.builder.build_call(extern_fn, &[$($input_matrix.into(),)*], name.unwrap_or_default()).unwrap();
}
};
}
generate_linalg_extern_fn!(call_np_linalg_matmul, "np_linalg_matmul", 3);
generate_linalg_extern_fn!(call_np_linalg_cholesky, "np_linalg_cholesky", 2);
generate_linalg_extern_fn!(call_np_linalg_qr, "np_linalg_qr", 3);
generate_linalg_extern_fn!(call_np_linalg_svd, "np_linalg_svd", 4);
generate_linalg_extern_fn!(call_np_linalg_inv, "np_linalg_inv", 2);
generate_linalg_extern_fn!(call_np_linalg_pinv, "np_linalg_pinv", 2);
generate_linalg_extern_fn!(call_sp_linalg_lu, "sp_linalg_lu", 3);
generate_linalg_extern_fn!(call_sp_linalg_schur, "sp_linalg_schur", 3);
generate_linalg_extern_fn!(call_sp_linalg_hessenberg, "sp_linalg_hessenberg", 3);
/// Invokes the linalg `np_dot` function.
pub fn call_np_dot<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
mat1: BasicValueEnum<'ctx>,
mat2: BasicValueEnum<'ctx>,
name: Option<&str>,
) -> FloatValue<'ctx> {
const FN_NAME: &str = "np_dot";
let extern_fn = ctx.module.get_function(FN_NAME).unwrap_or_else(|| {
let fn_type =
ctx.ctx.f64_type().fn_type(&[mat1.get_type().into(), mat2.get_type().into()], false);
let func = ctx.module.add_function(FN_NAME, fn_type, None);
for attr in ["mustprogress", "nofree", "nounwind", "willreturn", "writeonly"] {
func.add_attribute(
AttributeLoc::Function,
ctx.ctx.create_enum_attribute(Attribute::get_named_enum_kind_id(attr), 0),
);
}
func
});
ctx.builder
.build_call(extern_fn, &[mat1.into(), mat2.into()], name.unwrap_or_default())
.map(CallSiteValue::try_as_basic_value)
.map(|v| v.map_left(BasicValueEnum::into_float_value))
.map(Either::unwrap_left)
.unwrap()
}

View File

@ -159,7 +159,7 @@ where
///
/// * `elem_ty` - The element type of the `NDArray`.
/// * `shape` - The shape of the `NDArray`, represented am array of [`IntValue`]s.
fn create_ndarray_const_shape<'ctx, G: CodeGenerator + ?Sized>(
pub fn create_ndarray_const_shape<'ctx, G: CodeGenerator + ?Sized>(
generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
elem_ty: Type,

View File

@ -556,6 +556,17 @@ impl<'a> BuiltinBuilder<'a> {
| PrimDef::FunNpLdExp
| PrimDef::FunNpHypot
| PrimDef::FunNpNextAfter => self.build_np_2ary_function(prim),
PrimDef::FunNpDot
| PrimDef::FunNpLinalgMatmul
| PrimDef::FunNpLinalgCholesky
| PrimDef::FunNpLinalgQr
| PrimDef::FunNpLinalgSvd
| PrimDef::FunNpLinalgInv
| PrimDef::FunNpLinalgPinv
| PrimDef::FunSpLinalgLu
| PrimDef::FunSpLinalgSchur
| PrimDef::FunSpLinalgHessenberg => self.build_linalg_methods(prim),
};
if cfg!(debug_assertions) {
@ -1874,6 +1885,150 @@ impl<'a> BuiltinBuilder<'a> {
}
}
/// Build `np_linalg` and `sp_linalg` functions
///
/// The input to these functions must be floating point `NDArray`
fn build_linalg_methods(&mut self, prim: PrimDef) -> TopLevelDef {
debug_assert_prim_is_allowed(
prim,
&[
PrimDef::FunNpDot,
PrimDef::FunNpLinalgMatmul,
PrimDef::FunNpLinalgCholesky,
PrimDef::FunNpLinalgQr,
PrimDef::FunNpLinalgSvd,
PrimDef::FunNpLinalgInv,
PrimDef::FunNpLinalgPinv,
PrimDef::FunSpLinalgLu,
PrimDef::FunSpLinalgSchur,
PrimDef::FunSpLinalgHessenberg,
],
);
match prim {
PrimDef::FunNpDot => create_fn_by_codegen(
self.unifier,
&self.num_or_ndarray_var_map,
prim.name(),
self.primitives.float,
&[(self.num_or_ndarray_ty.ty, "x1"), (self.num_or_ndarray_ty.ty, "x2")],
Box::new(move |ctx, _, fun, args, generator| {
let x1_ty = fun.0.args[0].ty;
let x1_val = args[0].1.clone().to_basic_value_enum(ctx, generator, x1_ty)?;
let x2_ty = fun.0.args[1].ty;
let x2_val = args[1].1.clone().to_basic_value_enum(ctx, generator, x2_ty)?;
Ok(Some(builtin_fns::call_np_dot(
generator,
ctx,
(x1_ty, x1_val),
(x2_ty, x2_val),
)?))
}),
),
PrimDef::FunNpLinalgMatmul => create_fn_by_codegen(
self.unifier,
&VarMap::new(),
prim.name(),
self.ndarray_float_2d,
&[(self.ndarray_float_2d, "x1"), (self.ndarray_float_2d, "x2")],
Box::new(move |ctx, _, fun, args, generator| {
let x1_ty = fun.0.args[0].ty;
let x1_val = args[0].1.clone().to_basic_value_enum(ctx, generator, x1_ty)?;
let x2_ty = fun.0.args[1].ty;
let x2_val = args[1].1.clone().to_basic_value_enum(ctx, generator, x2_ty)?;
Ok(Some(builtin_fns::call_np_linalg_matmul(
generator,
ctx,
(x1_ty, x1_val),
(x2_ty, x2_val),
)?))
}),
),
PrimDef::FunNpLinalgCholesky | PrimDef::FunNpLinalgInv | PrimDef::FunNpLinalgPinv => {
create_fn_by_codegen(
self.unifier,
&VarMap::new(),
prim.name(),
self.ndarray_float_2d,
&[(self.ndarray_float_2d, "x1")],
Box::new(move |ctx, _, fun, args, generator| {
let x1_ty = fun.0.args[0].ty;
let x1_val =
args[0].1.clone().to_basic_value_enum(ctx, generator, x1_ty)?;
let func = match prim {
PrimDef::FunNpLinalgCholesky => builtin_fns::call_np_linalg_cholesky,
PrimDef::FunNpLinalgInv => builtin_fns::call_np_linalg_inv,
PrimDef::FunNpLinalgPinv => builtin_fns::call_np_linalg_pinv,
_ => unreachable!(),
};
Ok(Some(func(generator, ctx, (x1_ty, x1_val))?))
}),
)
}
PrimDef::FunNpLinalgQr
| PrimDef::FunSpLinalgLu
| PrimDef::FunSpLinalgSchur
| PrimDef::FunSpLinalgHessenberg => {
let ret_ty = self.unifier.add_ty(TypeEnum::TTuple {
ty: vec![self.ndarray_float_2d, self.ndarray_float_2d],
});
create_fn_by_codegen(
self.unifier,
&VarMap::new(),
prim.name(),
ret_ty,
&[(self.ndarray_float_2d, "x1")],
Box::new(move |ctx, _, fun, args, generator| {
let x1_ty = fun.0.args[0].ty;
let x1_val =
args[0].1.clone().to_basic_value_enum(ctx, generator, x1_ty)?;
let func = match prim {
PrimDef::FunNpLinalgQr => builtin_fns::call_np_linalg_qr,
PrimDef::FunSpLinalgLu => builtin_fns::call_sp_linalg_lu,
PrimDef::FunSpLinalgSchur => builtin_fns::call_sp_linalg_schur,
PrimDef::FunSpLinalgHessenberg => {
builtin_fns::call_sp_linalg_hessenberg
}
_ => unreachable!(),
};
Ok(Some(func(generator, ctx, (x1_ty, x1_val))?))
}),
)
}
PrimDef::FunNpLinalgSvd => {
let ret_ty = self.unifier.add_ty(TypeEnum::TTuple {
ty: vec![self.ndarray_float_2d, self.ndarray_float, self.ndarray_float_2d],
});
create_fn_by_codegen(
self.unifier,
&VarMap::new(),
prim.name(),
ret_ty,
&[(self.ndarray_float_2d, "x1")],
Box::new(move |ctx, _, fun, args, generator| {
let x1_ty = fun.0.args[0].ty;
let x1_val =
args[0].1.clone().to_basic_value_enum(ctx, generator, x1_ty)?;
Ok(Some(builtin_fns::call_np_linalg_svd(generator, ctx, (x1_ty, x1_val))?))
}),
)
}
_ => {
println!("{:?}", prim.name());
unreachable!()
}
}
}
fn create_method(prim: PrimDef, method_ty: Type) -> (StrRef, Type, DefinitionId) {
(prim.simple_name().into(), method_ty, prim.id())
}

View File

@ -100,6 +100,17 @@ pub enum PrimDef {
FunNpHypot,
FunNpNextAfter,
FunNpDot,
FunNpLinalgMatmul,
FunNpLinalgCholesky,
FunNpLinalgQr,
FunNpLinalgSvd,
FunNpLinalgInv,
FunNpLinalgPinv,
FunSpLinalgLu,
FunSpLinalgSchur,
FunSpLinalgHessenberg,
// Miscellaneous Python & NAC3 functions
FunInt32,
FunInt64,
@ -270,6 +281,16 @@ impl PrimDef {
PrimDef::FunNpLdExp => fun("np_ldexp", None),
PrimDef::FunNpHypot => fun("np_hypot", None),
PrimDef::FunNpNextAfter => fun("np_nextafter", None),
PrimDef::FunNpDot => fun("np_dot", None),
PrimDef::FunNpLinalgMatmul => fun("np_linalg_matmul", None),
PrimDef::FunNpLinalgCholesky => fun("np_linalg_cholesky", None),
PrimDef::FunNpLinalgQr => fun("np_linalg_qr", None),
PrimDef::FunNpLinalgSvd => fun("np_linalg_svd", None),
PrimDef::FunNpLinalgInv => fun("np_linalg_inv", None),
PrimDef::FunNpLinalgPinv => fun("np_linalg_pinv", None),
PrimDef::FunSpLinalgLu => fun("sp_linalg_lu", None),
PrimDef::FunSpLinalgSchur => fun("sp_linalg_schur", None),
PrimDef::FunSpLinalgHessenberg => fun("sp_linalg_hessenberg", None),
// Miscellaneous Python & NAC3 functions
PrimDef::FunInt32 => fun("int32", None),