Feat: Tensor, DType, PatternMatcher, etc (#21)

* wip

* concepts

* wip

* wip

* wip

* facet ideas

* allocation graph

* allocation graph

* allocation graph

* allocation graph

* concrete semantics

* dataclass

* dataclass

* dataclass

* dataclass

* tuple

* concrete semantics

* concrete semantics

* concrete semantics

* WIP

* some clean ups

* just an idea

* feat: reshape

* feat: reshape

* feat: expand

* feat: expand

* Cache, Type Inference, Verification

* Cache, Type Inference, Verification

* Cache, Type Inference, Verification

* Cache, Type Inference, Verification

* Feat: Simplifier

* Feat: Simplifier

* Feat: Simplifier

* Reshape symbolic

* docs

* docs

* smax

* write specs on def

* write specs on def

* Fix scope issue

* Fix scope issue

* fix ruff

* fix ruff

* fix ruff

* helpers

* helpers

* helpers

* helpers

* helpers

* helpers

* helpers

* helpers

* helpers

Author: hikettei

AGENTS.md

@@ -37,27 +37,4 @@
 
 ## Polyhedral DSL Guidelines
 - Prefer using Mixin operator overloads (e.g., `A | B` instead of `A.union(B)`) for cleaner code in user scripts and DSL implementations.
-
-## 作業計画と進捗 (2025-11-16)
-直近のギャップ集計: `docs/ISL_missing_apis.md`（2025-11-16 再生成、欠落API 2047件）。map 残 2 件（tuple_name系シンボル未提供のみ、libisl非存在）。
-優先順とステータス（✅完了 / 🚧着手中 / ⏳未着手）
-- Identifier / Id: 🚧（基本APIは揃うが欠落検証 継続）
-- Space / LocalSpace: 🚧（dim/tuple系以外の抜け有り）
-- Constraint / Equality-Constraint / Inequality-Constraint: 🚧
-- BasicSet / Set: 🚧（missing計: set 105, basic_set 63）
-- UnionSet: 🚧（missing計: union_set 52）
-- BasicMap / Map: 🚧（missing計: basic_map 85, map 190）
-- UnionMap: 🚧（missing計: union_map 112）
-- Aff / PwAff / MultiAff / PwMultiAff: 🚧（missing計: aff 73, pw_aff 96, multi_aff 90, pw_multi_aff 89）
-- MultiVal: 🚧（missing計: multi_val 37, val 66）
-- MultiUnionPwAff / UnionPwAff / UnionPwMultiAff / MultiUnionPwAff: 🚧（missing計: multi_union_pw_aff 75 ほか）
-- ScheduleConstraint / Schedule / ScheduleNode: ✅（schedule_node 0）
-- UnionAccessInfo / UnionFlow: ⏳
-- ASTExpr / ASTNode / ASTBuild: 🚧（Expr系クラス不足・missing計: ast_expr 0, ast_node 0）
-- Mat: ✅（要素参照系API実装済・missing計: mat 0）
-- その他: misc 71, options 29 など多数。
-
-次に着手する対象:
-1) ScheduleNode / ASTExpr / Mat のクラス追加・アクセサ補完
-2) UnionAccessInfo / UnionFlow ラッパ実装
-3) map / set 系を皮切りに `docs/ISL_missing_apis.md` に基づく欠落API埋め
+- Do not write shit code, be respectful to existing codes.

caten/__init__.py

@@ -0,0 +1,4 @@
+from .dtype import *  # noqa: F403, I001
+from .simplifier import *  # noqa: F403, I001
+from .tensor import *  # noqa: F403, I001
+from .runtime import cpu  # noqa: I001, F401

caten/dtype.py

@@ -0,0 +1,40 @@
+from __future__ import annotations
+
+from dataclasses import dataclass
+from typing import Any
+
+
+class DTypeMetaClass(type):
+    dcache: dict[tuple, DType] = {}
+    def __call__(cls, *args: Any, **kwargs: Any) -> DType:
+        if (ret:=DTypeMetaClass.dcache.get(args, None)) is not None: return ret
+        DTypeMetaClass.dcache[args] = ret = super().__call__(*args)
+        return ret
+
+# TODO: Vector/Packed DType
+@dataclass(frozen=True, eq=False)
+class DType:
+    name: str
+    @staticmethod
+    def new(name:str) -> DType: return DType(name)
+
+## definitions
+float64 = DType.new("float64")
+float32 = DType.new("float32")
+float16 = DType.new("float16")
+
+int64 = DType.new("int64")
+int32 = DType.new("int32")
+int16 = DType.new("int16")
+int8 = DType.new("int8")
+uint64 = DType.new("uint64")
+uint32 = DType.new("uint32")
+uint16 = DType.new("uint16")
+uint8 = DType.new("uint8")
+
+## dtype aliases
+index = int64
+default_float = float32
+
+floats = [float64, float32, float16]
+integers = [int64, int32, int16, int8, uint64, uint32, uint16, uint8]

caten/helpers.py

@@ -0,0 +1,20 @@
+from __future__ import annotations
+
+import functools
+import operator
+from typing import Any, Iterable, TypeVar
+
+T = TypeVar("T")
+
+def prod(x:Iterable[T]) -> T|int: return functools.reduce(operator.mul, x, 1)
+
+def argfix(*x: Any) -> tuple[Any, ...]:
+  if x and x[0].__class__ in (tuple, list):
+    if len(x) != 1: raise ValueError(f"bad arg {x}")
+    return tuple(x[0])
+  return x
+
+def align_left(*shapes: tuple[Any, ...]) -> tuple[tuple[Any, ...], ...]:
+  # unsqueeze left to make every shape same length
+  max_dim = max(len(shape) for shape in shapes)
+  return tuple((1,) * (max_dim - len(shape)) + shape for shape in shapes)