重置上传,新增了多个标准库,开始 TransPyV 自举实验

This commit is contained in:
2026-07-19 11:38:15 +08:00
parent 796222a300
commit 4e66207ba1
1041 changed files with 6597 additions and 27814 deletions

5
includes/hello.py Normal file
View File

@@ -0,0 +1,5 @@
import stdio
def helloworld():
stdio.printf("Hello, world!\n")

View File

@@ -22,7 +22,11 @@ from .__function import Function, BasicBlock, block_append_text, function_move_b
# builder.build_store(ConstInt(42), a)
# r = builder.build_load(i32_ty, a)
# builder.build_ret(r)
#
# 注意: 加 @t.NoVTable 以与硬编码 BUILDER_SIZE=32 保持一致
# (所有方法均为顶层函数,无需 vtable
# ============================================================
@t.NoVTable
class IRBuilder:
Pool: memhub.MemBuddy | t.CPtr # 内存池
Func: Function | t.CPtr # 所属函数
@@ -897,7 +901,10 @@ def build_gep_struct(builder: IRBuilder | t.CPtr, struct_ty: LLVMType | t.CPtr,
# ============================================================
# PhiIncoming: phi 节点的入边 (value, block) 对
# 继承 GSListNode[PhiIncoming] 获取强类型 Next
#
# 注意: 必须加 @t.NoVTable理由见 __function.py 的 Line 注释)。
# ============================================================
@t.NoVTable
class PhiIncoming(GSListNode[PhiIncoming]):
Val: Value | t.CPtr
Block: BasicBlock | t.CPtr
@@ -965,7 +972,10 @@ def build_phi(builder: IRBuilder | t.CPtr, ty: LLVMType | t.CPtr,
# ============================================================
# SwitchCase: switch 指令的 case 分支
#
# 注意: 必须加 @t.NoVTable理由见 __function.py 的 Line 注释)。
# ============================================================
@t.NoVTable
class SwitchCase(GSListNode[SwitchCase]):
CaseVal: Value | t.CPtr
Block: BasicBlock | t.CPtr

View File

@@ -13,7 +13,12 @@ from .__types import LLVMType, ParamNode, TypePrint
#
# 每条 LLVM IR 指令序列化为一行文本,通过链表组织。
# 继承 GSListNode[Line] 获取强类型 Next: Line|CPtr
#
# 注意: 必须加 @t.NoVTable。否则 TransPyC 为子类添加 vtable 指针(T*)
# 但完全丢失父类 GSListNode[T] 的 Next 字段,导致 struct 布局错位,
# 所有字段访问读取垃圾值(见续 XV 会话根因分析)。
# ============================================================
@t.NoVTable
class Line(GSListNode[Line]):
Buf: t.CChar | t.CPtr # 行文本NUL 结尾)
Len: t.CSizeT # 文本长度(不含 NUL
@@ -25,7 +30,10 @@ class Line(GSListNode[Line]):
# 持有名称 + 指令行链表 + 终止指令标志。
# 继承 GSListNode[BasicBlock] 获取强类型 Next: BasicBlock|CPtr
# Lines 为 GSList[Line] 指针,提供 O(1) 追加。
#
# 注意: 必须加 @t.NoVTable理由同 Line
# ============================================================
@t.NoVTable
class BasicBlock(GSListNode[BasicBlock]):
Name: t.CChar | t.CPtr # 块名(如 "entry"/"if.then"
Lines: GSList[Line] | t.CPtr # 指令行链表容器Head/Tail/Count
@@ -37,7 +45,10 @@ class BasicBlock(GSListNode[BasicBlock]):
#
# 继承 GSListNode[Param] 获取强类型 Next: Param|CPtr
# Attrs 存储参数属性文本(如 "noundef"/"nocapture"/"readonly"None=无属性
#
# 注意: 必须加 @t.NoVTable理由同 Line
# ============================================================
@t.NoVTable
class Param(GSListNode[Param]):
Ty: LLVMType | t.CPtr # 参数类型
Name: t.CChar | t.CPtr # 参数名(如 "%0"/"%a"
@@ -51,7 +62,10 @@ class Param(GSListNode[Param]):
# 继承 GSListNode[Function] 获取强类型 Next: Function|CPtr
# Params/Blocks 为 GSList 指针,提供 O(1) 追加。
# Attrs 存储函数级属性文本(如 "nounwind"/"noinline"None=无属性
#
# 注意: 必须加 @t.NoVTable理由同 Line
# ============================================================
@t.NoVTable
class Function(GSListNode[Function]):
Name: t.CChar | t.CPtr # 函数名
RetTy: LLVMType | t.CPtr # 返回类型
@@ -161,9 +175,12 @@ def function_move_block_to_end(func: Function | t.CPtr, block: BasicBlock | t.CP
用于解决 SSA 名非单调问题BB 按创建顺序排列,但指令按控制流顺序生成,
导致 SSA 编号在文本输出时非单调递增llc 报错。
在 position_at_end 时调用此函数,使 BB 顺序与 position 顺序一致。
注意:必须使用 GSList[BasicBlock] 带类型参数,否则 TransPyC 创建不完整
struct 类型,导致 blocks.Head/Tail/Count 字段访问读取垃圾值。
"""
if func is None or block is None: return
blocks: GSList | t.CPtr = func.Blocks
blocks: GSList[BasicBlock] | t.CPtr = func.Blocks
if blocks is None: return
# 已经是 Tail无需移动

View File

@@ -10,7 +10,10 @@ from linkedlist import GSListNode
# ParamNode: 函数参数类型链表节点
# 继承 GSListNode[ParamNode] 获取强类型 Next: ParamNode|CPtr
# Ty 用 t.CVoid | t.CPtr 前向引用,使用时转型为 (LLVMType | t.CPtr)
#
# 注意: 必须加 @t.NoVTable理由见 __function.py 的 Line 注释)。
# ============================================================
@t.NoVTable
class ParamNode(GSListNode[ParamNode]):
Ty: t.CVoid | t.CPtr # LLVMType*(前向引用,使用时转型)

View File

@@ -13,7 +13,10 @@ from .__types import LLVMType, TypePrint
# 表示一个 LLVM IR 值SSA 临时值(%0/%result或常量字面量42/0
# 每个值有类型指针、名字、是否常量标志。
# 继承 GSListNode[Value] 获取强类型 Next: Value|CPtr值池链表
#
# 注意: 必须加 @t.NoVTable理由见 __function.py 的 Line 注释)。
# ============================================================
@t.NoVTable
class Value(GSListNode[Value]):
Ty: LLVMType | t.CPtr # 值的类型
Name: t.CChar | t.CPtr # SSA 名("%0"/"%result")或常量文本("42"/"1.5e+00"

View File

@@ -28,7 +28,10 @@ VERIFY_ERR_PHI_NOT_FIRST: t.CDefine = 4
# ============================================================
# NameEntry: 已定义 SSA 名条目(链表节点)
#
# 注意: 必须加 @t.NoVTable理由见 __function.py 的 Line 注释)。
# ============================================================
@t.NoVTable
class NameEntry(GSListNode[NameEntry]):
Name: t.CChar | t.CPtr # SSA 名(如 "%0"/"%result"

View File

@@ -84,7 +84,7 @@ class MemManager:
string.memset(ptr, 0, total)
return ptr
def realloc(self, ptr: t.CVoid | t.CPtr, old_size: t.CSizeT, new_size: t.CSizeT) -> t.CVoid | t.CPtr:
def realloc(self, ptr: t.CVoid | t.CPtr, new_size: t.CSizeT) -> t.CVoid | t.CPtr:
if ptr is None:
return self.alloc(new_size)
if new_size == 0:
@@ -93,10 +93,7 @@ class MemManager:
new_ptr: t.CVoid | t.CPtr = self.alloc(new_size)
if new_ptr is None:
return ptr
copy_size: t.CSizeT = old_size
if new_size < old_size:
copy_size = new_size
string.memcpy(new_ptr, ptr, copy_size)
# 基类无法追踪旧分配大小, 不复制数据; 子类可覆盖以保留数据
self.free(ptr)
return new_ptr
@@ -453,7 +450,7 @@ class MemBuddy(MemManager):
return 1
# realloc 覆盖: 伙伴系统从块头读取旧阶数, 无需 old_size
def realloc(self, ptr: t.CVoid | t.CPtr, old_size: t.CSizeT, new_size: t.CSizeT) -> t.CVoid | t.CPtr:
def realloc(self, ptr: t.CVoid | t.CPtr, new_size: t.CSizeT) -> t.CVoid | t.CPtr:
if ptr is None:
return self.alloc(new_size)
if new_size == 0:
@@ -477,3 +474,48 @@ class MemBuddy(MemManager):
string.memcpy(new_ptr, ptr, old_data_size)
self.free(ptr)
return new_ptr
# === 状态查询 ===
@property
def mem_size(self) -> t.CSizeT:
# 可用区总大小2 的幂),分配上限为 mem_size - MEMBUDDY_HEADER_SIZE
return self.usable_size
def stats(self) -> t.CSizeT:
# 返回可用区总大小,用于状态统计
return self.usable_size
def free_count(self) -> t.CSizeT:
# 统计所有阶数空闲链中的块总数
count: t.CSizeT = 0
i: t.CInt
for i in range(MEMBUDDY_MAX_ORDERS + 1):
head_val: t.CUInt64T = self.free_lists[i]
cur: t.CVoid | t.CPtr = t.CVoid(head_val, t.CPtr)
while cur is not None:
count += 1
cur = t.CVoid(c.Deref(t.CUInt64T(cur, t.CPtr)), t.CPtr)
return count
def self_check(self) -> t.CInt:
# 验证伙伴分配器内部一致性,返回 0=OK非 0=损坏
if self.usable is None:
return 0
i: t.CInt
for i in range(MEMBUDDY_MAX_ORDERS + 1):
head_val: t.CUInt64T = self.free_lists[i]
cur: t.CVoid | t.CPtr = t.CVoid(head_val, t.CPtr)
while cur is not None:
# 检查指针在可用区内
if t.CUInt64T(cur) < t.CUInt64T(self.usable):
return 1
if t.CUInt64T(cur) >= t.CUInt64T(self.usable) + self.usable_size:
return 2
# 检查对齐order i 的块必须按 _block_size_at_order(i) 对齐
offset: t.CSizeT = t.CUInt64T(cur) - t.CUInt64T(self.usable)
bs: t.CSizeT = _block_size_at_order(i)
if offset % bs != 0:
return 3
cur = t.CVoid(c.Deref(t.CUInt64T(cur, t.CPtr)), t.CPtr)
return 0

View File

@@ -39,10 +39,9 @@ def strncpy(dest: str, src: str, n: t.CSizeT) -> str:
dest += 1
i += 1
# 填充剩余空间为 '\0'
while i < n:
for i in range(n):
dest[0] = 0
dest += 1
i += 1
return original_dest
@@ -351,4 +350,16 @@ def split(s: str, delim: str, result: t.CArray[str]) -> int:
start[0] = '\0'
start += 1
result[count] = None # 结尾标记
return count
return count
ascii_lowercase: str = "abcdefghijklmnopqrstuvwxyz"
ascii_uppercase: str = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
ascii_letters: str = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ"
digits: str = "0123456789"
hexdigits: str = "0123456789abcdefABCDEF"
octdigits: str = "01234567"
punctuation: str = "!\"#$%&'()*+,-./:;<=>?@[]^_`{|}~"
whitespace: str = " \t\n\r\f\v"
printable: str = "01234567890123456789abcdefABCDEF01234567!\"#$%&'()*+,-./:;<=>?@[]^_`{|}~ \t\n\r\f\v"

View File

@@ -20,11 +20,13 @@ def begin(name: str):
SetConsoleOutputCP(CP_UTF8)
SetConsoleCP(CP_UTF8)
stdio.printf("=== %s ===\n\n", name)
stdio.fflush(None)
def section(name: str):
"""打印分节标题"""
stdio.printf("--- %s ---\n", name)
stdio.fflush(None)
def ok(msg: str):
@@ -32,6 +34,7 @@ def ok(msg: str):
global _pass_count
_pass_count += 1
stdio.printf("PASS: %s\n", msg)
stdio.fflush(None)
def fail(msg: str):
@@ -39,6 +42,7 @@ def fail(msg: str):
global _fail_count
_fail_count += 1
stdio.printf("FAIL: %s\n", msg)
stdio.fflush(None)
def check(cond: t.CInt, ok_msg: str, fail_msg: str):
@@ -52,6 +56,7 @@ def check(cond: t.CInt, ok_msg: str, fail_msg: str):
def info(msg: str):
"""打印信息消息(不计入 PASS/FAIL"""
stdio.printf("%s\n", msg)
stdio.fflush(None)
def end() -> t.CInt:
@@ -61,6 +66,7 @@ def end() -> t.CInt:
_total_fail += _fail_count
stdio.printf("\n--- Summary ---\n")
stdio.printf("PASS: %d, FAIL: %d\n", _pass_count, _fail_count)
stdio.fflush(None)
return _fail_count
@@ -68,4 +74,5 @@ def summary() -> t.CInt:
"""打印所有套件的总计汇总,返回总失败数"""
stdio.printf("\n=== Total Summary ===\n")
stdio.printf("Total PASS: %d, Total FAIL: %d\n", _total_pass, _total_fail)
stdio.fflush(None)
return _total_fail

View File

@@ -31,9 +31,8 @@ class Vector[T]():
if self.pool == None: return
new_capacity: t.CSizeT = self.capacity * 2
if new_capacity == 0: new_capacity = 4
old_size: t.CSizeT = self.capacity * self.elem_size
new_size: t.CSizeT = new_capacity * self.elem_size
new_data: t.CVoid | t.CPtr = self.pool.realloc(self.data, old_size, new_size)
new_data: t.CVoid | t.CPtr = self.pool.realloc(self.data, new_size)
if new_data == None: return
self.data = new_data
self.capacity = new_capacity