1
0
forked from M-Labs/nac3

core/expr: Add support for multi-dim slicing of NDArrays

This commit is contained in:
David Mak 2024-05-30 16:08:15 +08:00
parent c35ad06949
commit ed79d5bb9e
2 changed files with 216 additions and 111 deletions

View File

@ -1667,6 +1667,7 @@ fn gen_ndarray_subscript_expr<'ctx, G: CodeGenerator>(
slice: &Expr<Option<Type>>,
) -> Result<Option<ValueEnum<'ctx>>, String> {
let llvm_i1 = ctx.ctx.bool_type();
let llvm_i32 = ctx.ctx.i32_type();
let llvm_usize = generator.get_size_type(ctx.ctx);
let TypeEnum::TLiteral { values, .. } = &*ctx.unifier.get_ty_immutable(ndims) else {
@ -1712,32 +1713,11 @@ fn gen_ndarray_subscript_expr<'ctx, G: CodeGenerator>(
slice.location,
);
if let ExprKind::Slice { lower, upper, step } = &slice.node {
let dim0_sz = unsafe {
v.dim_sizes().get_typed_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
};
let Some((start, stop, step)) = handle_slice_indices(
lower,
upper,
step,
ctx,
generator,
dim0_sz,
)? else { return Ok(None) };
return Ok(Some(numpy::ndarray_sliced_copy(
generator,
ctx,
ty,
v,
&[(start, stop, step)],
)?.as_ptr_value().into()))
}
let index = if let Some(index) = generator.gen_expr(ctx, slice)? {
let index = index.to_basic_value_enum(ctx, generator, slice.custom.unwrap())?.into_int_value();
// Normalizes a possibly-negative index to its corresponding positive index
let normalize_index = |generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
index: IntValue<'ctx>,
dim: u64| {
gen_if_else_expr_callback(
generator,
ctx,
@ -1757,7 +1737,7 @@ fn gen_ndarray_subscript_expr<'ctx, G: CodeGenerator>(
v.dim_sizes().get_typed_unchecked(
ctx,
generator,
&llvm_usize.const_zero(),
&llvm_usize.const_int(dim, true),
None,
)
};
@ -1770,97 +1750,194 @@ fn gen_ndarray_subscript_expr<'ctx, G: CodeGenerator>(
Ok(Some(ctx.builder.build_int_truncate(index, llvm_i32, "").unwrap()))
},
)?.map(BasicValueEnum::into_int_value).unwrap()
} else {
return Ok(None)
).map(|v| v.map(BasicValueEnum::into_int_value))
};
let index_addr = generator.gen_var_alloc(ctx, index.get_type().into(), None)?;
ctx.builder.build_store(index_addr, index).unwrap();
if ndims.len() == 1 && ndims[0] == 1 {
// Accessing an element from a 1-dimensional `ndarray`
// Converts a slice expression into a slice-range tuple
let expr_to_slice = |generator: &mut G,
ctx: &mut CodeGenContext<'ctx, '_>,
node: &ExprKind<Option<Type>>,
dim: u64| {
match node {
ExprKind::Constant { value: Constant::Int(v), .. } => {
let Some(index) = normalize_index(
generator, ctx, llvm_i32.const_int(*v as u64, true), dim,
)? else {
return Ok(None)
};
Ok(Some(v.data()
.get(
Ok(Some((index, index, llvm_i32.const_int(1, true))))
}
ExprKind::Slice { lower, upper, step } => {
let dim_sz = unsafe {
v.dim_sizes()
.get_typed_unchecked(
ctx,
generator,
&llvm_usize.const_int(dim, false),
None,
)
};
handle_slice_indices(lower, upper, step, ctx, generator, dim_sz)
}
_ => {
let Some(index) = generator.gen_expr(ctx, slice)? else {
return Ok(None)
};
let index = index
.to_basic_value_enum(ctx, generator, slice.custom.unwrap())?
.into_int_value();
let Some(index) = normalize_index(generator, ctx, index, dim)? else {
return Ok(None)
};
Ok(Some((index, index, llvm_i32.const_int(1, true))))
}
}
};
Ok(Some(match &slice.node {
ExprKind::Tuple { elts, .. } => {
let slices = elts.iter().enumerate()
.map(|(dim, elt)| expr_to_slice(generator, ctx, &elt.node, dim as u64))
.take_while_inclusive(|slice| slice.as_ref().is_ok_and(Option::is_some))
.collect::<Result<Vec<_>, _>>()?;
if slices.len() < elts.len() {
return Ok(None)
}
let slices = slices.into_iter()
.map(Option::unwrap)
.collect_vec();
numpy::ndarray_sliced_copy(
generator,
ctx,
ty,
v,
&slices,
)?.as_ptr_value().into()
}
ExprKind::Slice { .. } => {
let Some(slice) = expr_to_slice(generator, ctx, &slice.node, 0)? else {
return Ok(None)
};
numpy::ndarray_sliced_copy(
generator,
ctx,
ty,
v,
&[slice],
)?.as_ptr_value().into()
}
_ => {
let index = if let Some(index) = generator.gen_expr(ctx, slice)? {
index.to_basic_value_enum(ctx, generator, slice.custom.unwrap())?.into_int_value()
} else {
return Ok(None)
};
let Some(index) = normalize_index(generator, ctx, index, 0)? else {
return Ok(None)
};
let index_addr = generator.gen_var_alloc(ctx, index.get_type().into(), None)?;
ctx.builder.build_store(index_addr, index).unwrap();
if ndims.len() == 1 && ndims[0] == 1 {
// Accessing an element from a 1-dimensional `ndarray`
return Ok(Some(v.data()
.get(
ctx,
generator,
&ArraySliceValue::from_ptr_val(
index_addr,
llvm_usize.const_int(1, false),
None,
),
None,
)
.into()))
}
// Accessing an element from a multi-dimensional `ndarray`
// Create a new array, remove the top dimension from the dimension-size-list, and copy the
// elements over
let subscripted_ndarray = generator.gen_var_alloc(
ctx,
llvm_ndarray_t.into(),
None
)?;
let ndarray = NDArrayValue::from_ptr_val(
subscripted_ndarray,
llvm_usize,
None
);
let num_dims = v.load_ndims(ctx);
ndarray.store_ndims(
ctx,
generator,
ctx.builder.build_int_sub(num_dims, llvm_usize.const_int(1, false), "").unwrap(),
);
let ndarray_num_dims = ndarray.load_ndims(ctx);
ndarray.create_dim_sizes(ctx, llvm_usize, ndarray_num_dims);
let ndarray_num_dims = ndarray.load_ndims(ctx);
let v_dims_src_ptr = unsafe {
v.dim_sizes().ptr_offset_unchecked(
ctx,
generator,
&llvm_usize.const_int(1, false),
None,
)
};
call_memcpy_generic(
ctx,
ndarray.dim_sizes().base_ptr(ctx, generator),
v_dims_src_ptr,
ctx.builder
.build_int_mul(ndarray_num_dims, llvm_usize.size_of(), "")
.map(Into::into)
.unwrap(),
llvm_i1.const_zero(),
);
let ndarray_num_elems = call_ndarray_calc_size(
generator,
ctx,
&ndarray.dim_sizes().as_slice_value(ctx, generator),
(None, None),
);
ndarray.create_data(ctx, llvm_ndarray_data_t, ndarray_num_elems);
let v_data_src_ptr = v.data().ptr_offset(
ctx,
generator,
&ArraySliceValue::from_ptr_val(index_addr, llvm_usize.const_int(1, false), None),
None,
)
.into()))
} else {
// Accessing an element from a multi-dimensional `ndarray`
// Create a new array, remove the top dimension from the dimension-size-list, and copy the
// elements over
let subscripted_ndarray = generator.gen_var_alloc(
ctx,
llvm_ndarray_t.into(),
None
)?;
let ndarray = NDArrayValue::from_ptr_val(
subscripted_ndarray,
llvm_usize,
None
);
let num_dims = v.load_ndims(ctx);
ndarray.store_ndims(
ctx,
generator,
ctx.builder.build_int_sub(num_dims, llvm_usize.const_int(1, false), "").unwrap(),
);
let ndarray_num_dims = ndarray.load_ndims(ctx);
ndarray.create_dim_sizes(ctx, llvm_usize, ndarray_num_dims);
let ndarray_num_dims = ndarray.load_ndims(ctx);
let v_dims_src_ptr = unsafe {
v.dim_sizes().ptr_offset_unchecked(
None
);
call_memcpy_generic(
ctx,
generator,
&llvm_usize.const_int(1, false),
None,
)
};
call_memcpy_generic(
ctx,
ndarray.dim_sizes().base_ptr(ctx, generator),
v_dims_src_ptr,
ctx.builder
.build_int_mul(ndarray_num_dims, llvm_usize.size_of(), "")
.map(Into::into)
.unwrap(),
llvm_i1.const_zero(),
);
ndarray.data().base_ptr(ctx, generator),
v_data_src_ptr,
ctx.builder
.build_int_mul(ndarray_num_elems, llvm_ndarray_data_t.size_of().unwrap(), "")
.map(Into::into)
.unwrap(),
llvm_i1.const_zero(),
);
let ndarray_num_elems = call_ndarray_calc_size(
generator,
ctx,
&ndarray.dim_sizes().as_slice_value(ctx, generator),
(None, None),
);
ndarray.create_data(ctx, llvm_ndarray_data_t, ndarray_num_elems);
let v_data_src_ptr = v.data().ptr_offset(
ctx,
generator,
&ArraySliceValue::from_ptr_val(index_addr, llvm_usize.const_int(1, false), None),
None
);
call_memcpy_generic(
ctx,
ndarray.data().base_ptr(ctx, generator),
v_data_src_ptr,
ctx.builder
.build_int_mul(ndarray_num_elems, llvm_ndarray_data_t.size_of().unwrap(), "")
.map(Into::into)
.unwrap(),
llvm_i1.const_zero(),
);
Ok(Some(ndarray.as_ptr_value().into()))
}
ndarray.as_ptr_value().into()
}
}))
}
/// See [`CodeGenerator::gen_expr`].

View File

@ -2,6 +2,7 @@ use std::collections::{HashMap, HashSet};
use std::convert::{From, TryInto};
use std::iter::once;
use std::{cell::RefCell, sync::Arc};
use std::ops::Not;
use super::typedef::{Call, FunSignature, FuncArg, RecordField, Type, TypeEnum, Unifier, VarMap};
use super::{magic_methods::*, type_error::TypeError, typedef::CallId};
@ -554,7 +555,10 @@ impl<'a> Fold<()> for Inferencer<'a> {
ExprKind::ListComp { .. }
| ExprKind::Lambda { .. }
| ExprKind::Call { .. } => expr.custom, // already computed
ExprKind::Slice { .. } => None, // we don't need it for slice
ExprKind::Slice { .. } => {
// slices aren't exactly ranges, but for our purposes this should suffice
Some(self.primitives.range)
}
_ => return report_error("not supported", expr.location),
};
Ok(ast::Expr { custom, location: expr.location, node: expr.node })
@ -1642,6 +1646,30 @@ impl<'a> Inferencer<'a> {
}
}
}
ExprKind::Tuple { elts, .. } => {
if value.custom
.unwrap()
.obj_id(self.unifier)
.is_some_and(|id| id == PRIMITIVE_DEF_IDS.ndarray)
.not() {
return report_error("Tuple slices are only supported for ndarrays", slice.location)
}
for elt in elts {
if let ExprKind::Slice { lower, upper, step } = &elt.node {
for v in [lower.as_ref(), upper.as_ref(), step.as_ref()].iter().flatten() {
self.constrain(v.custom.unwrap(), self.primitives.int32, &v.location)?;
}
} else {
self.constrain(elt.custom.unwrap(), self.primitives.int32, &elt.location)?;
}
}
let (_, ndims) = unpack_ndarray_var_tys(self.unifier, value.custom.unwrap());
let ndarray_ty = make_ndarray_ty(self.unifier, self.primitives, Some(ty), Some(ndims));
self.constrain(value.custom.unwrap(), ndarray_ty, &value.location)?;
Ok(ndarray_ty)
}
_ => {
if let TypeEnum::TTuple { .. } = &*self.unifier.get_ty(value.custom.unwrap()) {
return report_error("Tuple index must be a constant (KernelInvariant is also not supported)", slice.location)