warp-lang 1.6.1__py3-none-win_amd64.whl → 1.7.0__py3-none-win_amd64.whl

warp/builtins.py

@@ -1,15 +1,24 @@
-# Copyright (c) 2022 NVIDIA CORPORATION.  All rights reserved.
-# NVIDIA CORPORATION and its licensors retain all intellectual property
-# and proprietary rights in and to this software, related documentation
-# and any modifications thereto.  Any use, reproduction, disclosure or
-# distribution of this software and related documentation without an express
-# license agreement from NVIDIA CORPORATION is strictly prohibited.
+# SPDX-FileCopyrightText: Copyright (c) 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
 import builtins
 import functools
-import tempfile
-from pathlib import Path
 from typing import Any, Callable, Mapping, Sequence
 
+import warp.build
+import warp.context
 from warp.codegen import Reference, Var, strip_reference
 from warp.types import *
 
@@ -32,7 +41,7 @@ def sametypes(arg_types: Mapping[str, Any]):
     return all(types_equal(arg_type_0, t) for t in arg_types_iter)
 
 
-def sametypes_create_value_func(default):
+def sametypes_create_value_func(default: TypeVar):
     def fn(arg_types, arg_values):
         if arg_types is None:
             return default
@@ -390,7 +399,7 @@ add_builtin(
 )
 
 
-def scalar_infer_type(arg_types: Mapping[str, type]):
+def scalar_infer_type(arg_types: Union[Mapping[str, type], Tuple[type, ...], None]):
     if arg_types is None:
         return Scalar
 
@@ -941,6 +950,12 @@ def matrix_value_func(arg_types: Mapping[str, type], arg_values: Mapping[str, An
             raise RuntimeError("the `shape` argument must be specified when initializing a matrix by value")
 
         if all(type_is_vector(x) for x in variadic_arg_types):
+            warp.utils.warn(
+                "the built-in `wp.matrix()` won't support taking column vectors as input "
+                "in the future. Use `wp.matrix_from_rows()` or `wp.matrix_from_cols()` instead.",
+                DeprecationWarning,
+            )
+
             if shape[1] != variadic_arg_count:
                 raise RuntimeError(
                     f"incompatible number of column vectors given ({variadic_arg_count}) "
@@ -1021,6 +1036,86 @@ add_builtin(
 )
 
 
+def matrix_from_vecs_create_value_func(cols: bool):
+    def fn(arg_types: Mapping[str, type], arg_values: Mapping[str, Any]):
+        if arg_types is None:
+            return matrix(shape=(Any, Any), dtype=Scalar)
+
+        variadic_arg_types = arg_types.get("args", ())
+        variadic_arg_count = len(variadic_arg_types)
+
+        if not all(type_is_vector(x) for x in variadic_arg_types):
+            raise RuntimeError("all arguments are expected to be vectors")
+
+        length = variadic_arg_types[0]._length_
+        if any(x._length_ != length for x in variadic_arg_types):
+            raise RuntimeError("all vectors are expected to have the same length")
+
+        dtype = variadic_arg_types[0]._wp_scalar_type_
+        if any(x._wp_scalar_type_ != dtype for x in variadic_arg_types):
+            raise RuntimeError("all vectors are expected to have the same dtype")
+
+        shape = (length, variadic_arg_count) if cols else (variadic_arg_count, length)
+        return matrix(shape=shape, dtype=dtype)
+
+    return fn
+
+
+def matrix_from_vecs_dispatch_func(input_types: Mapping[str, type], return_type: Any, args: Mapping[str, Var]):
+    # We're in the codegen stage where we emit the code calling the built-in.
+    # Further validate the given argument values if needed and map them
+    # to the underlying C++ function's runtime and template params.
+
+    shape = return_type._shape_
+    dtype = return_type._wp_scalar_type_
+
+    variadic_args = args.get("args", ())
+
+    func_args = variadic_args
+
+    if shape in ((2, 2), (3, 3), (4, 4)):
+        # Template specializations exist for these shapes, don't pass them
+        # as template parameters.
+        template_args = (dtype,)
+    else:
+        template_args = (*shape, dtype)
+
+    return (func_args, template_args)
+
+
+def matrix_from_vecs_initializer_list_func(args, return_type):
+    shape = return_type._shape_
+
+    return shape[0] != shape[1] or shape[0] > 4
+
+
+add_builtin(
+    "matrix_from_cols",
+    input_types={"*args": vector(length=Any, dtype=Scalar)},
+    variadic=True,
+    value_func=matrix_from_vecs_create_value_func(cols=True),
+    dispatch_func=matrix_from_vecs_dispatch_func,
+    initializer_list_func=matrix_from_vecs_initializer_list_func,
+    native_func="matrix_from_cols",
+    doc="Construct a matrix from column vectors.",
+    group="Vector Math",
+    export=False,
+)
+
+add_builtin(
+    "matrix_from_rows",
+    input_types={"*args": vector(length=Any, dtype=Scalar)},
+    variadic=True,
+    value_func=matrix_from_vecs_create_value_func(cols=False),
+    dispatch_func=matrix_from_vecs_dispatch_func,
+    initializer_list_func=matrix_from_vecs_initializer_list_func,
+    native_func="matrix_from_rows",
+    doc="Construct a matrix from row vectors.",
+    group="Vector Math",
+    export=False,
+)
+
+
 def identity_value_func(arg_types: Mapping[str, type], arg_values: Mapping[str, Any]):
     if arg_types is None:
         return matrix(shape=(Any, Any), dtype=Scalar)
@@ -1132,6 +1227,21 @@ add_builtin(
     while the left and right basis vectors are returned in ``U`` and ``V``.""",
 )
 
+add_builtin(
+    "svd2",
+    input_types={
+        "A": matrix(shape=(2, 2), dtype=Float),
+        "U": matrix(shape=(2, 2), dtype=Float),
+        "sigma": vector(length=2, dtype=Float),
+        "V": matrix(shape=(2, 2), dtype=Scalar),
+    },
+    value_type=None,
+    group="Vector Math",
+    export=False,
+    doc="""Compute the SVD of a 2x2 matrix ``A``. The singular values are returned in ``sigma``,
+    while the left and right basis vectors are returned in ``U`` and ``V``.""",
+)
+
 add_builtin(
     "qr3",
     input_types={
@@ -1323,7 +1433,18 @@ add_builtin(
     input_types={"mat": matrix(shape=(3, 3), dtype=Float)},
     value_func=lambda arg_types, arg_values: quaternion(dtype=float_infer_type(arg_types)),
     group="Quaternion Math",
-    doc="Construct a quaternion from a 3x3 matrix.",
+    doc="""Construct a quaternion from a 3x3 matrix.
+
+    If the matrix is not a pure rotation, but for example includes scaling or skewing, the result is undefined.""",
+)
+add_builtin(
+    "quat_from_matrix",
+    input_types={"mat": matrix(shape=(4, 4), dtype=Float)},
+    value_func=lambda arg_types, arg_values: quaternion(dtype=float_infer_type(arg_types)),
+    group="Quaternion Math",
+    doc="""Construct a quaternion from a 4x4 matrix.
+
+    If the top-left 3x3 block of the matrix is not a pure rotation, but for example includes scaling or skewing, the result is undefined.""",
 )
 add_builtin(
     "quat_rpy",
@@ -2366,7 +2487,7 @@ add_builtin(
 
     This function converts values computed using scalar kernel code to a tile representation for input into collective operations.
 
-    * If the input value is a scalar, then the resulting tile has ``shape=(1, block_dim)``
+    * If the input value is a scalar, then the resulting tile has ``shape=(block_dim,)``
     * If the input value is a vector, then the resulting tile has ``shape=(length(vector), block_dim)``
 
     :param x: A per-thread local value, e.g. scalar, vector, or matrix.
@@ -2660,11 +2781,9 @@ def tile_broadcast_value_func(arg_types, arg_values):
 def tile_broadcast_dispatch_func(arg_types: Mapping[str, type], return_type: Any, arg_values: Mapping[str, Var]):
     tile = arg_values["a"]
 
-    template_args = []
-    template_args.append(return_type.shape[0])
-    template_args.append(return_type.shape[1])
-    template_args.append(return_type.strides[0])
-    template_args.append(return_type.strides[1])
+    assert len(return_type.shape) == len(return_type.strides)
+    assert 1 <= len(return_type.shape) <= 4
+    template_args = [*return_type.shape, *return_type.strides]
 
     return ((tile,), template_args)
 
@@ -2677,56 +2796,17 @@ add_builtin(
     variadic=False,
     doc="""Broadcast a tile.
 
-    This function will attempt to broadcast the input tile ``a`` to the destination shape (m, n).
-
+    Broadcasts the input tile ``a`` to the destination shape.
     Broadcasting follows NumPy broadcast rules.
 
     :param a: Tile to broadcast
     :param shape: The shape to broadcast to
-    :returns: Tile with broadcast ``shape=(m, n)``""",
+    :returns: Tile with broadcast shape""",
     group="Tile Primitives",
     export=False,
 )
 
 
-def tile_matmul_value_func(arg_types, arg_values):
-    # return generic type (for doc builds)
-    if arg_types is None:
-        return Tile(dtype=Any, shape=Any)
-
-    if len(arg_types) != 3:
-        raise TypeError(f"tile_matmul() takes exactly 3 positional arguments but {len(arg_types)} were given")
-
-    return None
-
-
-def tile_matmul_dispatch_func(arg_types: Mapping[str, type], return_type: Any, arg_values: Mapping[str, Var]):
-    a = arg_values["a"]
-    b = arg_values["b"]
-    out = arg_values["out"]
-
-    # force the storage type of the input variables to shared memory
-    a.type.storage = "shared"
-    b.type.storage = "shared"
-    out.type.storage = "shared"
-
-    template_args = []
-    return ((a, b, out), template_args)
-
-
-add_builtin(
-    "tile_matmul_scalar",
-    input_types={"a": Tile, "b": Tile, "out": Tile},
-    value_func=tile_matmul_value_func,
-    dispatch_func=tile_matmul_dispatch_func,
-    variadic=True,
-    doc="Compute matrix product and accumulate out += a*b.",
-    group="Tile Primitives",
-    hidden=True,
-    export=False,
-)
-
-
 def tile_sum_value_func(arg_types, arg_values):
     # return generic type (for doc builds)
     if arg_types is None:
@@ -3021,7 +3101,7 @@ def tile_binary_map_value_func(arg_types, arg_values):
 
     for i in range(len(a.shape)):
         if a.shape[i] != b.shape[i]:
-            raise ValueError(f"tile_map() shapes do not match on dimension {i}, got {a.shape[i]} and {b.shape[i]}")
+            raise ValueError(f"tile_map() shapes do not match on dimension {i}, got {a.shape} and {b.shape}")
 
     return TileBinaryMap(a, b)
 
@@ -3798,6 +3878,18 @@ _volume_supported_value_types = {
 }
 
 
+def _is_volume_type_supported(dtype):
+    for typ in _volume_supported_value_types:
+        if types_equal(typ, dtype):
+            return True
+    return False
+
+
+def _check_volume_type_is_supported(dtype):
+    if not _is_volume_type_supported(dtype):
+        raise RuntimeError(f"unsupported volume type `{type_repr(dtype)}`")
+
+
 def check_volume_value_grad_compatibility(dtype, grad_dtype):
     if type_is_vector(dtype):
         expected = matrix(shape=(type_length(dtype), 3), dtype=type_scalar_type(dtype))
@@ -3813,9 +3905,7 @@ def volume_value_func(arg_types: Mapping[str, type], arg_values: Mapping[str, An
         return Any
 
     dtype = arg_values["dtype"]
-
-    if dtype not in _volume_supported_value_types:
-        raise RuntimeError(f"unsupported volume type `{dtype.__name__}`")
+    _check_volume_type_is_supported(dtype)
 
     return dtype
 
@@ -3851,9 +3941,7 @@ def volume_sample_grad_value_func(arg_types: Mapping[str, type], arg_values: Map
         return Any
 
     dtype = arg_values["dtype"]
-
-    if dtype not in _volume_supported_value_types:
-        raise RuntimeError(f"unsupported volume type `{dtype.__name__}`")
+    _check_volume_type_is_supported(dtype)
 
     check_volume_value_grad_compatibility(dtype, arg_types["grad"])
 
@@ -3891,9 +3979,7 @@ def volume_lookup_value_func(arg_types: Mapping[str, type], arg_values: Mapping[
         return Any
 
     dtype = arg_values["dtype"]
-
-    if dtype not in _volume_supported_value_types:
-        raise RuntimeError(f"unsupported volume type `{dtype.__name__}`")
+    _check_volume_type_is_supported(dtype)
 
     return dtype
 
@@ -3930,9 +4016,7 @@ def volume_store_value_func(arg_types: Mapping[str, type], arg_values: Mapping[s
         return None
 
     dtype = arg_types["value"]
-
-    if dtype not in _volume_supported_value_types:
-        raise RuntimeError(f"unsupported volume type `{dtype.__name__}`")
+    _check_volume_type_is_supported(dtype)
 
     return None
 
@@ -4182,6 +4266,20 @@ add_builtin(
     group="Random",
     doc="Return a random integer between [low, high).",
 )
+add_builtin(
+    "randu",
+    input_types={"state": uint32},
+    value_type=uint32,
+    group="Random",
+    doc="Return a random unsigned integer in the range [0, 2^32).",
+)
+add_builtin(
+    "randu",
+    input_types={"state": uint32, "low": uint32, "high": uint32},
+    value_type=uint32,
+    group="Random",
+    doc="Return a random unsigned integer between [low, high).",
+)
 add_builtin(
     "randf",
     input_types={"state": uint32},
@@ -4490,11 +4588,31 @@ add_builtin(
     export=False,
     group="Utility",
 )
+
+
+def select_dispatch_func(input_types: Mapping[str, type], return_type: Any, args: Mapping[str, Var]):
+    warp.utils.warn(
+        "wp.select() is deprecated and will be removed in a future\n"
+        "version. Use wp.where(cond, value_if_true, value_if_false) instead.",
+        category=DeprecationWarning,
+    )
+
+    func_args = tuple(args.values())
+    template_args = ()
+
+    return (func_args, template_args)
+
+
 add_builtin(
     "select",
     input_types={"cond": builtins.bool, "value_if_false": Any, "value_if_true": Any},
     value_func=lambda arg_types, arg_values: Any if arg_types is None else arg_types["value_if_false"],
-    doc="Select between two arguments, if ``cond`` is ``False`` then return ``value_if_false``, otherwise return ``value_if_true``",
+    dispatch_func=select_dispatch_func,
+    doc="""Select between two arguments, if ``cond`` is ``False`` then return ``value_if_false``, otherwise return ``value_if_true``.
+
+    .. deprecated:: 1.7
+         Use :func:`where` instead, which has the more intuitive argument order:
+         ``where(cond, value_if_true, value_if_false)``.""",
     group="Utility",
 )
 for t in int_types:
@@ -4502,14 +4620,47 @@ for t in int_types:
         "select",
         input_types={"cond": t, "value_if_false": Any, "value_if_true": Any},
         value_func=lambda arg_types, arg_values: Any if arg_types is None else arg_types["value_if_false"],
-        doc="Select between two arguments, if ``cond`` is ``False`` then return ``value_if_false``, otherwise return ``value_if_true``",
+        dispatch_func=select_dispatch_func,
+        doc="""Select between two arguments, if ``cond`` is ``False`` then return ``value_if_false``, otherwise return ``value_if_true``.
+
+    .. deprecated:: 1.7
+         Use :func:`where` instead, which has the more intuitive argument order:
+         ``where(cond, value_if_true, value_if_false)``.""",
         group="Utility",
     )
 add_builtin(
     "select",
     input_types={"arr": array(dtype=Any), "value_if_false": Any, "value_if_true": Any},
     value_func=lambda arg_types, arg_values: Any if arg_types is None else arg_types["value_if_false"],
-    doc="Select between two arguments, if ``arr`` is null then return ``value_if_false``, otherwise return ``value_if_true``",
+    dispatch_func=select_dispatch_func,
+    doc="""Select between two arguments, if ``arr`` is null then return ``value_if_false``, otherwise return ``value_if_true``.
+
+    .. deprecated:: 1.7
+         Use :func:`where` instead, which has the more intuitive argument order:
+         ``where(arr, value_if_true, value_if_false)``.""",
+    group="Utility",
+)
+
+add_builtin(
+    "where",
+    input_types={"cond": builtins.bool, "value_if_true": Any, "value_if_false": Any},
+    value_func=lambda arg_types, arg_values: Any if arg_types is None else arg_types["value_if_false"],
+    doc="Select between two arguments, if ``cond`` is ``True`` then return ``value_if_true``, otherwise return ``value_if_false``.",
+    group="Utility",
+)
+for t in int_types:
+    add_builtin(
+        "where",
+        input_types={"cond": t, "value_if_true": Any, "value_if_false": Any},
+        value_func=lambda arg_types, arg_values: Any if arg_types is None else arg_types["value_if_false"],
+        doc="Select between two arguments, if ``cond`` is ``True`` then return ``value_if_true``, otherwise return ``value_if_false``.",
+        group="Utility",
+    )
+add_builtin(
+    "where",
+    input_types={"arr": array(dtype=Any), "value_if_true": Any, "value_if_false": Any},
+    value_func=lambda arg_types, arg_values: Any if arg_types is None else arg_types["value_if_false"],
+    doc="Select between two arguments, if ``arr`` is not null then return ``value_if_true``, otherwise return ``value_if_false``.",
     group="Utility",
 )
 
@@ -5103,33 +5254,51 @@ add_builtin(
 )
 
 
+# implements vector[index] = value
+add_builtin(
+    "assign_inplace",
+    input_types={"a": vector(length=Any, dtype=Scalar), "i": int, "value": Scalar},
+    value_type=None,
+    hidden=True,
+    group="Utility",
+)
+
+# implements quaternion[index] = value
+add_builtin(
+    "assign_inplace",
+    input_types={"a": quaternion(dtype=Scalar), "i": int, "value": Scalar},
+    value_type=None,
+    hidden=True,
+    group="Utility",
+)
+
+
 def vector_assign_value_func(arg_types: Mapping[str, type], arg_values: Mapping[str, Any]):
     vec_type = arg_types["a"]
     return vec_type
 
 
-# implements vector[index] = value
+# implements vector[index] = value, performs a copy internally if wp.config.enable_vector_component_overwrites is True
 add_builtin(
-    "assign",
+    "assign_copy",
     input_types={"a": vector(length=Any, dtype=Scalar), "i": int, "value": Scalar},
     value_func=vector_assign_value_func,
     hidden=True,
     group="Utility",
 )
 
-# implements quaternion[index] = value
+# implements quaternion[index] = value, performs a copy internally if wp.config.enable_vector_component_overwrites is True
 add_builtin(
-    "assign",
+    "assign_copy",
     input_types={"a": quaternion(dtype=Scalar), "i": int, "value": Scalar},
     value_func=vector_assign_value_func,
     hidden=True,
     group="Utility",
 )
 
-
 # implements vector[idx] += scalar
 add_builtin(
-    "augassign_add",
+    "add_inplace",
     input_types={"a": vector(length=Any, dtype=Scalar), "i": int, "value": Scalar},
     value_type=None,
     hidden=True,
@@ -5138,7 +5307,7 @@ add_builtin(
 
 # implements quaternion[idx] += scalar
 add_builtin(
-    "augassign_add",
+    "add_inplace",
     input_types={"a": quaternion(dtype=Scalar), "i": int, "value": Scalar},
     value_type=None,
     hidden=True,
@@ -5147,7 +5316,7 @@ add_builtin(
 
 # implements vector[idx] -= scalar
 add_builtin(
-    "augassign_sub",
+    "sub_inplace",
     input_types={"a": vector(length=Any, dtype=Scalar), "i": int, "value": Scalar},
     value_type=None,
     hidden=True,
@@ -5156,7 +5325,7 @@ add_builtin(
 
 # implements quaternion[idx] -= scalar
 add_builtin(
-    "augassign_sub",
+    "sub_inplace",
     input_types={"a": quaternion(dtype=Scalar), "i": int, "value": Scalar},
     value_type=None,
     hidden=True,
@@ -5200,11 +5369,6 @@ add_builtin(
 )
 
 
-def matrix_assign_value_func(arg_types: Mapping[str, type], arg_values: Mapping[str, Any]):
-    mat_type = arg_types["a"]
-    return mat_type
-
-
 def matrix_vector_sametype(arg_types: Mapping[str, Any]):
     mat_size = arg_types["a"]._shape_[0]
     vec_size = arg_types["value"]._length_
@@ -5215,7 +5379,33 @@ def matrix_vector_sametype(arg_types: Mapping[str, Any]):
 
 # implements matrix[i,j] = scalar
 add_builtin(
-    "assign",
+    "assign_inplace",
+    input_types={"a": matrix(shape=(Any, Any), dtype=Scalar), "i": int, "j": int, "value": Scalar},
+    value_type=None,
+    hidden=True,
+    group="Utility",
+)
+
+
+# implements matrix[i] = vector
+add_builtin(
+    "assign_inplace",
+    input_types={"a": matrix(shape=(Any, Any), dtype=Scalar), "i": int, "value": vector(length=Any, dtype=Scalar)},
+    constraint=matrix_vector_sametype,
+    value_type=None,
+    hidden=True,
+    group="Utility",
+)
+
+
+def matrix_assign_value_func(arg_types: Mapping[str, type], arg_values: Mapping[str, Any]):
+    mat_type = arg_types["a"]
+    return mat_type
+
+
+# implements matrix[i,j] = scalar
+add_builtin(
+    "assign_copy",
     input_types={"a": matrix(shape=(Any, Any), dtype=Scalar), "i": int, "j": int, "value": Scalar},
     value_func=matrix_assign_value_func,
     hidden=True,
@@ -5225,7 +5415,7 @@ add_builtin(
 
 # implements matrix[i] = vector
 add_builtin(
-    "assign",
+    "assign_copy",
     input_types={"a": matrix(shape=(Any, Any), dtype=Scalar), "i": int, "value": vector(length=Any, dtype=Scalar)},
     constraint=matrix_vector_sametype,
     value_func=matrix_assign_value_func,
@@ -5236,7 +5426,7 @@ add_builtin(
 
 # implements matrix[i,j] += scalar
 add_builtin(
-    "augassign_add",
+    "add_inplace",
     input_types={"a": matrix(shape=(Any, Any), dtype=Scalar), "i": int, "j": int, "value": Scalar},
     value_type=None,
     hidden=True,
@@ -5244,9 +5434,20 @@ add_builtin(
 )
 
 
+# implements matrix[i] += vector
+add_builtin(
+    "add_inplace",
+    input_types={"a": matrix(shape=(Any, Any), dtype=Scalar), "i": int, "value": vector(length=Any, dtype=Scalar)},
+    constraint=matrix_vector_sametype,
+    value_type=None,
+    hidden=True,
+    group="Utility",
+)
+
+
 # implements matrix[i,j] -= scalar
 add_builtin(
-    "augassign_sub",
+    "sub_inplace",
     input_types={"a": matrix(shape=(Any, Any), dtype=Scalar), "i": int, "j": int, "value": Scalar},
     value_type=None,
     hidden=True,
@@ -5254,6 +5455,16 @@ add_builtin(
 )
 
 
+# implements matrix[i] -= vector
+add_builtin(
+    "sub_inplace",
+    input_types={"a": matrix(shape=(Any, Any), dtype=Scalar), "i": int, "value": vector(length=Any, dtype=Scalar)},
+    value_type=None,
+    hidden=True,
+    group="Utility",
+)
+
+
 for t in scalar_types + vector_types + (bool,):
     if "vec" in t.__name__ or "mat" in t.__name__:
         continue
@@ -5401,7 +5612,27 @@ add_builtin(
 )
 add_builtin(
     "expect_near",
-    input_types={"a": vec3, "b": vec3, "tolerance": float},
+    input_types={"a": vector(length=Any, dtype=Float), "b": vector(length=Any, dtype=Float), "tolerance": Float},
+    defaults={"tolerance": 1.0e-6},
+    value_type=None,
+    doc="Prints an error to stdout if any element of ``a`` and ``b`` are not closer than tolerance in magnitude",
+    group="Utility",
+)
+add_builtin(
+    "expect_near",
+    input_types={"a": quaternion(dtype=Float), "b": quaternion(dtype=Float), "tolerance": Float},
+    defaults={"tolerance": 1.0e-6},
+    value_type=None,
+    doc="Prints an error to stdout if any element of ``a`` and ``b`` are not closer than tolerance in magnitude",
+    group="Utility",
+)
+add_builtin(
+    "expect_near",
+    input_types={
+        "a": matrix(shape=(Any, Any), dtype=Float),
+        "b": matrix(shape=(Any, Any), dtype=Float),
+        "tolerance": Float,
+    },
     defaults={"tolerance": 1.0e-6},
     value_type=None,
     doc="Prints an error to stdout if any element of ``a`` and ``b`` are not closer than tolerance in magnitude",
@@ -5980,7 +6211,7 @@ add_builtin(
 ##
 ## Matmul
 ##
-def tile_matmul_generic_value_func(arg_types, arg_values):
+def tile_matmul_value_func(arg_types, arg_values):
     # return generic type (for doc builds)
     if arg_types is None:
         return Tile(dtype=Any, shape=Any)
@@ -6006,7 +6237,7 @@ def tile_matmul_generic_value_func(arg_types, arg_values):
     return None
 
 
-def tile_matmul_generic_lto_dispatch_func(
+def tile_matmul_lto_dispatch_func(
     arg_types: Mapping[str, type],
     return_type: Any,
     return_values: List[Var],
@@ -6045,142 +6276,82 @@ def tile_matmul_generic_lto_dispatch_func(
     out.type.storage = "shared"
     template_args = [accumulate]
 
-    # Maps Python/Warp types to C++ types and enums
-    def cublasdx_type_map(dtype):
-        if dtype == float16:
-            return ("wp::float16", 3, 0)
-        if dtype == float32:
-            return ("wp::float32", 5, 0)
-        if dtype == float64:
-            return ("wp::float64", 6, 0)
-        if dtype == vec2h:
-            return ("wp::vec2h", 3, 1)
-        if dtype == vec2f:
-            return ("wp::vec2f", 5, 1)
-        if dtype == vec2d:
-            return ("wp::vec2d", 6, 1)
-        raise TypeError("Unsupported input type in tile_matmul")
-
-    def cublasdx_arrangement_map(layout):
-        if layout == "colmajor":
-            return 0  # CUBLASDX_ARRANGEMENT_COL_MAJOR
-        if layout == "rowmajor":
-            return 1  # CUBLASDX_ARRANGEMENT_ROW_MAJOR
-        raise ValueError("Unsupported layout in tile_matmul")
-
-    # generate the LTO
     M, K = a.type.shape[0], a.type.shape[1]
     _, N = b.type.shape[0], b.type.shape[1]
     num_threads = options["block_dim"]
     arch = options["output_arch"]
 
-    def make_function(M, N, K, adtype, bdtype, cdtype, alayout, blayout, clayout):
-        (a_dtype, a_prec, a_type) = cublasdx_type_map(adtype)
-        (b_dtype, b_prec, b_type) = cublasdx_type_map(bdtype)
-        (c_dtype, c_prec, c_type) = cublasdx_type_map(cdtype)
-        a_arrangement = cublasdx_arrangement_map(alayout)
-        b_arrangement = cublasdx_arrangement_map(blayout)
-        c_arrangement = cublasdx_arrangement_map(clayout)
-
-        if a_type != b_type or a_type != c_type:
-            raise TypeError("time_matmul(A, B, C) requires all inputs to be real or complex")
-
-        element_type = a_type
-
-        lto_symbol = f"dot_{M}_{N}_{K}_{arch}_{num_threads}_{a_arrangement}_{b_arrangement}_{c_arrangement}_{a_prec}_{b_prec}_{c_prec}_{element_type}"
+    if arch is None or not warp.context.runtime.core.is_mathdx_enabled():
+        # CPU/no-MathDx dispatch
+        return ((0, 0, 0, a, b, out), template_args, [], 0)
+    else:
 
-        # early out if LTO for this combination already exists for this module
-        if lto_symbol in builder.ltoirs:
-            return lto_symbol, builder.ltoirs[lto_symbol]
+        def tile_flip_layout(layout):
+            if layout == "rowmajor":
+                return "colmajor"
+            elif layout == "colmajor":
+                return "rowmajor"
 
-        # otherwise compile LTO
-        lto_code = tempfile.NamedTemporaryFile(prefix="warp", delete=False)
-        result = warp.context.runtime.core.cuda_compile_dot(
-            lto_code.name.encode("utf-8"),
-            lto_symbol.encode("utf-8"),
-            0,
-            None,
-            None,
+        # generate the LTOs
+        #    C += A * B
+        (fun_forward, lto_forward) = warp.build.build_lto_dot(
+            M,
+            N,
+            K,
+            a.type.dtype,
+            b.type.dtype,
+            out.type.dtype,
+            a.type.layout,
+            b.type.layout,
+            out.type.layout,
             arch,
+            num_threads,
+            builder,
+        )
+        # adjA += adjC * B^T - Transpose ~= flipped layout
+        (fun_backward_A, lto_backward_A) = warp.build.build_lto_dot(
             M,
+            K,
             N,
+            out.type.dtype,
+            b.type.dtype,
+            a.type.dtype,
+            out.type.layout,
+            tile_flip_layout(b.type.layout),
+            a.type.layout,
+            arch,
+            num_threads,
+            builder,
+        )
+        # adjB += A^T * adjC - Transpose ~= flipped layout
+        (fun_backward_B, lto_backward_B) = warp.build.build_lto_dot(
             K,
-            a_prec,
-            b_prec,
-            c_prec,
-            element_type,
-            a_arrangement,
-            b_arrangement,
-            c_arrangement,
+            N,
+            M,
+            a.type.dtype,
+            out.type.dtype,
+            b.type.dtype,
+            tile_flip_layout(a.type.layout),
+            out.type.layout,
+            b.type.layout,
+            arch,
             num_threads,
+            builder,
         )
-        lto_code_path = Path(lto_code.name)
-        if not result:
-            lto_code.close()
-            if lto_code_path.exists():
-                lto_code_path.unlink()
-            raise RuntimeError("Failed to compile tile_matmul")
-        else:
-            with open(lto_code.name, "rb") as f:
-                lto_code_data = f.read()
-            lto_code.close()
-            lto_code_path.unlink()
-
-            builder.ltoirs[lto_symbol] = lto_code_data
-            builder.ltoirs_decl[lto_symbol] = (
-                f"void {lto_symbol}({c_dtype}, {a_dtype}*, {b_dtype}*, {c_dtype}, {c_dtype}*);"
-            )
-
-            return lto_symbol, lto_code_data
 
-    def tile_flip_layout(layout):
-        if layout == "rowmajor":
-            return "colmajor"
-        elif layout == "colmajor":
-            return "rowmajor"
-
-    #    C += A * B
-    (fun_forward, lto_forward) = make_function(
-        M, N, K, a.type.dtype, b.type.dtype, out.type.dtype, a.type.layout, b.type.layout, out.type.layout
-    )
-    # adjA += adjC * B^T - Transpose ~= flipped layout
-    (fun_backward_A, lto_backward_A) = make_function(
-        M,
-        K,
-        N,
-        out.type.dtype,
-        b.type.dtype,
-        a.type.dtype,
-        out.type.layout,
-        tile_flip_layout(b.type.layout),
-        a.type.layout,
-    )
-    # adjB += A^T * adjC - Transpose ~= flipped layout
-    (fun_backward_B, lto_backward_B) = make_function(
-        K,
-        N,
-        M,
-        a.type.dtype,
-        out.type.dtype,
-        b.type.dtype,
-        tile_flip_layout(a.type.layout),
-        out.type.layout,
-        b.type.layout,
-    )
-
-    return (
-        (
-            Var(fun_forward, str, False, True, False),
-            Var(fun_backward_A, str, False, True, False),
-            Var(fun_backward_B, str, False, True, False),
-            a,
-            b,
-            out,
-        ),
-        template_args,
-        [lto_forward, lto_backward_A, lto_backward_B],
-        0,
-    )
+        return (
+            (
+                Var(fun_forward, str, False, True, False),
+                Var(fun_backward_A, str, False, True, False),
+                Var(fun_backward_B, str, False, True, False),
+                a,
+                b,
+                out,
+            ),
+            template_args,
+            [lto_forward, lto_backward_A, lto_backward_B],
+            0,
+        )
 
 
 add_builtin(
@@ -6190,8 +6361,8 @@ add_builtin(
         "b": Tile(dtype=Any, shape=Any),
         "out": Tile(dtype=Any, shape=Any),
     },
-    value_func=tile_matmul_generic_value_func,
-    lto_dispatch_func=tile_matmul_generic_lto_dispatch_func,
+    value_func=tile_matmul_value_func,
+    lto_dispatch_func=tile_matmul_lto_dispatch_func,
     variadic=False,
     doc="""Computes the matrix product and accumulates ``out += a*b``.
 
@@ -6199,7 +6370,7 @@ add_builtin(
         * fp16, fp32, fp64 (real)
         * vec2h, vec2f, vec2d (complex)
 
-    All input and output tiles must have the same datatype. Tile data will be automatically be migrated
+    All input and output tiles must have the same datatype. Tile data will automatically be migrated
     to shared memory if necessary and will use TensorCore operations when available.
 
     :param a: A tile with ``shape=(M, K)``
@@ -6213,8 +6384,8 @@ add_builtin(
 add_builtin(
     "tile_matmul",
     input_types={"a": Tile(dtype=Any, shape=Any), "b": Tile(dtype=Any, shape=Any)},
-    value_func=tile_matmul_generic_value_func,
-    lto_dispatch_func=tile_matmul_generic_lto_dispatch_func,
+    value_func=tile_matmul_value_func,
+    lto_dispatch_func=tile_matmul_lto_dispatch_func,
     variadic=False,
     doc="""Computes the matrix product ``out = a*b``.
 
@@ -6222,7 +6393,7 @@ add_builtin(
         * fp16, fp32, fp64 (real)
         * vec2h, vec2f, vec2d (complex)
 
-    Both input tiles must have the same datatype. Tile data will be automatically be migrated
+    Both input tiles must have the same datatype. Tile data will automatically be migrated
     to shared memory if necessary and will use TensorCore operations when available.
 
     :param a: A tile with ``shape=(M, K)``
@@ -6294,59 +6465,29 @@ def tile_fft_generic_lto_dispatch_func(
     num_threads = options["block_dim"]
     arch = options["output_arch"]
     ept = size // num_threads
-    lto_symbol = f"fft_{size}_{ept}_{arch}_{direction}_{precision}"
-
-    # early out if LTO for this combination already exists for this module
-    if lto_symbol in builder.ltoirs:
-        return lto_symbol, builder.ltoirs[lto_symbol]
-
-    # otherwise compile LTO
-    lto_code = tempfile.NamedTemporaryFile(prefix="warp", delete=False)
-    shared_memory_size = ctypes.c_int(0)
-
-    result = warp.context.runtime.core.cuda_compile_fft(
-        lto_code.name.encode("utf-8"),
-        lto_symbol.encode("utf-8"),
-        0,
-        None,
-        None,
-        arch,
-        size,
-        ept,
-        dir,
-        precision,
-        ctypes.byref(shared_memory_size),
-    )
-    lto_code_path = Path(lto_code.name)
-    if not result:
-        lto_code.close()
-        if lto_code_path.exists():
-            lto_code_path.unlink()
-        raise RuntimeError("Failed to compile tile_fft")
-
-    with open(lto_code.name, "rb") as f:
-        lto_code_data = f.read()
-
-    lto_code.close()
-    lto_code_path.unlink()
-
-    builder.ltoirs[lto_symbol] = lto_code_data
-
-    shared_memory_bytes = Tile.round_up(shared_memory_size.value)
-
-    return (
-        (
-            Var(lto_symbol, str, False, True, False),
-            Var(dtype, str, False, True, False),
-            Var(str(shared_memory_bytes), str, False, True, False),
-            Var(str(batch), str, False, True, False),
-            Var(str(ept), str, False, True, False),
-            inout,
-        ),
-        [],
-        [lto_code_data],
-        shared_memory_bytes,
-    )
+
+    if arch is None or not warp.context.runtime.core.is_mathdx_enabled():
+        # CPU/no-MathDx dispatch
+        return ([], [], [], 0)
+    else:
+        # generate the LTO
+        lto_symbol, lto_code_data, shared_memory_bytes = warp.build.build_lto_fft(
+            arch, size, ept, direction, dir, precision, builder
+        )
+
+        return (
+            (
+                Var(lto_symbol, str, False, True, False),
+                Var(dtype, str, False, True, False),
+                Var(str(shared_memory_bytes), str, False, True, False),
+                Var(str(batch), str, False, True, False),
+                Var(str(ept), str, False, True, False),
+                inout,
+            ),
+            [],
+            [lto_code_data],
+            shared_memory_bytes,
+        )
 
 
 add_builtin(
@@ -6408,7 +6549,7 @@ def tile_cholesky_generic_value_func(arg_types, arg_values):
         raise TypeError(f"tile_cholesky() argument must be a tile, got {a!r}")
 
     if len(a.shape) != 2:
-        raise ValueError("tile_cholesky() argumust must be a 2D tile")
+        raise ValueError("tile_cholesky() argument must be a 2D tile")
 
     if a.shape[0] != a.shape[1]:
         raise ValueError("tile_cholesky() argument must be square")
@@ -6449,57 +6590,36 @@ def tile_cholesky_generic_lto_dispatch_func(
     if out.type.shape[0] != M or out.type.shape[1] != M:
         raise ValueError("tile_cholesky() output tile must be square")
 
-    num_threads = options["block_dim"]
-    arch = options["output_arch"]
-    lto_symbol = f"potrf_{M}_{N}_{arch}_{precision_enum}"
-
-    # early out if LTO for this combination already exists for this module
-    if lto_symbol in builder.ltoirs:
-        return lto_symbol, builder.ltoirs[lto_symbol]
-
-    # otherwise compile LTO
-    lto_code = tempfile.NamedTemporaryFile(prefix="warp", delete=False)
-    universal_fatbin_code = tempfile.NamedTemporaryFile(prefix="warp", delete=False)
+    solver = "potrf"
+    solver_enum = cusolver_function_map[solver]
 
-    # cuSOLVERDx only support col-major input/outputs,
+    # cuSOLVERDx only supports col-major input/outputs,
     # so we use upper to mimic a row-major input
-    result = warp.context.runtime.core.cuda_compile_solver(
-        universal_fatbin_code.name.encode("utf-8"),
-        lto_code.name.encode("utf-8"),
-        lto_symbol.encode("utf-8"),
-        0,
-        None,
-        None,
-        arch,
-        M,
-        N,
-        cusolver_function_map["potrf"],
-        precision_enum,
-        cusolver_fill_mode_map["upper"],
-        num_threads,
-    )
+    fill_mode = cusolver_fill_mode_map["upper"]
 
-    if not result:
-        for f in [lto_code, universal_fatbin_code]:
-            f.close()
-            if Path(f.name).exists():
-                Path(f.name).unlink()
-        raise RuntimeError("Failed to compile tile_cholesky")
+    arch = options["output_arch"]
+    num_threads = options["block_dim"]
+    parameter_list = f"({dtype}*, unsigned)"
 
+    if arch is None or not warp.context.runtime.core.is_mathdx_enabled():
+        # CPU/no-MathDx dispatch
+        return ((0, a, out), [], [], 0)
     else:
-        with open(lto_code.name, "rb") as f:
-            lto_code_data = f.read()
-        with open(universal_fatbin_code.name, "rb") as f:
-            universal_fatbin_code_data = f.read()
-        for f in [lto_code, universal_fatbin_code]:
-            f.close()
-            Path(f.name).unlink()
-
-    builder.ltoirs[lto_symbol] = lto_code_data
-    builder.ltoirs_decl[lto_symbol] = f"void {lto_symbol}({dtype}*, unsigned);"
-    builder.fatbins["cholesky"] = universal_fatbin_code_data
+        # generate the LTO
+        lto_symbol, lto_code_data = warp.build.build_lto_solver(
+            M,
+            N,
+            solver,
+            solver_enum,
+            fill_mode,
+            arch,
+            precision_enum,
+            num_threads,
+            parameter_list,
+            builder,
+        )
 
-    return ((Var(lto_symbol, str, False, True, False), a, out), [], [lto_code_data], 0)
+        return ((Var(lto_symbol, str, False, True, False), a, out), [], [lto_code_data], 0)
 
 
 add_builtin(
@@ -6593,57 +6713,36 @@ def tile_cholesky_solve_generic_lto_dispatch_func(
             f"got {y.type.shape[0]} elements in output and {M} rows in 'L'"
         )
 
-    num_threads = options["block_dim"]
-    arch = options["output_arch"]
-    lto_symbol = f"potrs_{M}_{N}_{arch}_{precision_enum}"
-
-    # early out if LTO for this combination already exists for this module
-    if lto_symbol in builder.ltoirs:
-        return lto_symbol, builder.ltoirs[lto_symbol]
-
-    # otherwise compile LTO
-    lto_code = tempfile.NamedTemporaryFile(prefix="warp", delete=False)
-    universal_fatbin_code = tempfile.NamedTemporaryFile(prefix="warp", delete=False)
+    solver = "potrs"
+    solver_enum = cusolver_function_map[solver]
 
-    # cuSOLVERDx only support col-major input/outputs,
+    # cuSOLVERDx only supports col-major input/outputs,
     # so we use upper to mimic a row-major input
-    result = warp.context.runtime.core.cuda_compile_solver(
-        universal_fatbin_code.name.encode("utf-8"),
-        lto_code.name.encode("utf-8"),
-        lto_symbol.encode("utf-8"),
-        0,
-        None,
-        None,
-        arch,
-        M,
-        N,
-        cusolver_function_map["potrs"],
-        precision_enum,
-        cusolver_fill_mode_map["upper"],
-        num_threads,
-    )
+    fill_mode = cusolver_fill_mode_map["upper"]
 
-    if not result:
-        for f in [lto_code, universal_fatbin_code]:
-            f.close()
-            if Path(f.name).exists():
-                Path(f.name).unlink()
-        raise RuntimeError("Failed to compile tile_cholesky_solve")
+    arch = options["output_arch"]
+    num_threads = options["block_dim"]
+    parameter_list = f"({dtype}*, {dtype}*)"
 
+    if arch is None or not warp.context.runtime.core.is_mathdx_enabled():
+        # CPU/no-MathDx dispatch
+        return ((0, L, x, y), [], [], 0)
     else:
-        with open(lto_code.name, "rb") as f:
-            lto_code_data = f.read()
-        with open(universal_fatbin_code.name, "rb") as f:
-            universal_fatbin_code_data = f.read()
-        for f in [lto_code, universal_fatbin_code]:
-            f.close()
-            Path(f.name).unlink()
-
-    builder.ltoirs[lto_symbol] = lto_code_data
-    builder.ltoirs_decl[lto_symbol] = f"void {lto_symbol}({dtype}*, {dtype}*);"
-    builder.fatbins["cholesky"] = universal_fatbin_code_data
-
-    return ((Var(lto_symbol, str, False, True, False), L, x, y), [], [lto_code_data], 0)
+        # generate the LTO
+        lto_symbol, lto_code_data = warp.build.build_lto_solver(
+            M,
+            N,
+            solver,
+            solver_enum,
+            fill_mode,
+            arch,
+            precision_enum,
+            num_threads,
+            parameter_list,
+            builder,
+        )
+
+        return ((Var(lto_symbol, str, False, True, False), L, x, y), [], [lto_code_data], 0)
 
 
 add_builtin(