mawei/StarPilot
1
0
Fork 0
mirror of https://github.com/firestar5683/StarPilot.git synced 2026-11-08 05:23:41 +08:00
Code Issues Packages Projects Releases Wiki Activity
Files
63243d7e2d768514b61b5cd5d357200ecab7e208
StarPilot/tinygrad_repo/tinygrad/uop/ops.py
T
Daniel Koepping 68984ff07b bump tinygrad (#39085)
2026-10-20 03:55:45 +00:00

1915 lines
107 KiB
Python
Raw Blame History

from __future__ import annotations
from typing import Any, Callable, cast, TYPE_CHECKING, Type, Sequence, Iterable, Final, Iterator
import sys, time, functools, itertools, math, operator, hashlib, os, types, pickle, pathlib, inspect, weakref, collections, struct
from dataclasses import dataclass, replace
from enum import Enum, auto
from tinygrad.uop import Ops, GroupOp
from tinygrad.dtype import ConstType, dtypes, DType, DTypeLike, truncate, least_upper_dtype, least_upper_float, Invalid, AddrSpace, strong_dtype
from tinygrad.dtype import PyConst, InvalidType, bitcast
from tinygrad.device import Buffer, BufferSpec, MultiBuffer, canonicalize_device, is_disk_device, TinyELF
from tinygrad.helpers import ContextVar, all_int, prod, getenv, all_same, Context, partition, temp, unwrap, T, argfix, Metadata, flatten, TRACEMETA
from tinygrad.helpers import PROFILE, dedup, cdiv, cmod, floordiv, floormod, diskcache_put, to_function_name, cpu_profile, TracingKey
from tinygrad.helpers import VIZ, SPEC, CAPTURE_PROCESS_REPLAY, DISALLOW_BROADCAST, get_shape, fully_flatten, to_tuple
from tinygrad.helpers import colored, ansilen, printable, Target, is_image_shape, strides_for_shape
if TYPE_CHECKING:
from tinygrad.renderer import Estimates
class AxisType(Enum):
def __repr__(self): return str(self)
def __lt__(self, other:AxisType): return self.value < other.value
# Nesting order: RANGE args sort by axis type, then axis id.
DEVICE = auto(); GLOBAL = auto(); LOCAL = auto(); WARP = auto(); WEAK = auto(); LOOP = auto() # noqa: E702
UPCAST = auto(); PLACEHOLDER = auto() # noqa: E702
@dataclass(frozen=True, order=True)
class ParamArg:
slot: int
dtype: DType
vmin_vmax: tuple[PyConst, PyConst]|None = None
multiple_of: int|None = None
name: str|None = None
addrspace: AddrSpace|None = AddrSpace.GLOBAL
device: str|tuple[str, ...]|None = None
volatile: bool = False
# (h, w) if this is an image2d buffer, then the size CONST is h*w*4
image: tuple[int, int]|None = None
# the device Buffer for a realized BUFFER. the UOp is the owner of the Buffer: they live and die together (1:1)
buffer: Buffer|MultiBuffer|None = None
bind_on_realize: bool = False
# the bound value of a Variable (an ALU PARAM with a value range); None means unbound
val: PyConst|None = None
# the memory the linker gives an ALLOC
spec: BufferSpec|None = None
def __repr__(self):
fields = (("vmin_vmax", None), ("multiple_of", None), ("name", None), ("addrspace", AddrSpace.GLOBAL), ("device", None),
("volatile", False), ("image", None), ("bind_on_realize", False), ("val", None), ("spec", None))
args = [repr(self.slot), repr(self.dtype)] + [f"{k}={v!r}" for k,default in fields if (v:=getattr(self, k)) != default]
if self.buffer is not None:
args.append(f"buffer=UOp.new_buffer({self.device!r}, {self.buffer.nbytes // self.dtype.itemsize}, {self.dtype!r}, {self.slot}).buffer")
return f"ParamArg({', '.join(args)})"
axis_letters = {AxisType.DEVICE: "d", AxisType.GLOBAL: "g", AxisType.LOCAL: "l", AxisType.WARP: "w", AxisType.WEAK: "L",
AxisType.LOOP: "L", AxisType.UPCAST: "u"}
axis_colors = {AxisType.DEVICE: "green", AxisType.GLOBAL: "blue", AxisType.LOCAL: "cyan", AxisType.WARP: "CYAN",
AxisType.WEAK: "red", AxisType.LOOP: "red", AxisType.UPCAST: "yellow"}
range_start = {Ops.STAGE: 1, Ops.REDUCE: 1, Ops.END: 1, Ops.CALL: 1, Ops.LINEAR: 0}
# https://en.wikipedia.org/wiki/Identity_element
def identity_element(op:Ops, dt:DType) -> PyConst: return dt.const({Ops.ADD:0, Ops.MUL:1, Ops.MAX:dt.min}[op])
# With True as the default, this matches the old symbolic behavior
def resolve(x:UOp|bool, default:bool=True):
if isinstance(x, bool): return x
assert x.dtype == dtypes.bool, "UOp in resolve must be bool"
# NOTE: generating the text for the exception is expensive, so we do this
return bool(sx.vmin) if (sx:=x.simplify()).vmin == sx.vmax else default
# smax/smin are replacements for max/min that preserve symbolic
def _suop(lst, uop_fxn, python_fxn):
uops, nums = partition(lst, lambda x: isinstance(x, UOp))
return ssimplify(functools.reduce(uop_fxn, uops + ([python_fxn(nums)] if nums else [])))
def smax(*lst) -> sint: return _suop(argfix(*lst), UOp.maximum, max)
def smin(*lst) -> sint: return _suop(argfix(*lst), UOp.minimum, min)
def srender(x:sint) -> str: return x.render() if isinstance(x, UOp) else str(x)
def _align_left(*shapes:tuple[sint, ...]) -> tuple[tuple[sint, ...], ...]:
max_dim = max(len(s) for s in shapes)
return tuple((1,)*(max_dim-len(s))+s for s in shapes)
def _broadcast_shape(*shapes:tuple[sint, ...]) -> tuple[sint, ...]:
if all_same(shapes): return shapes[0]
# per right-aligned dim: sizes of 1 broadcast to the others, which must all agree
ret = []
for sizes in zip(*_align_left(*shapes)):
if len(rest:=dedup([s for s in sizes if isinstance(s, UOp) or s != 1])) > 1:
raise IndexError(f"shape mismatch: objects cannot be broadcast to a single shape {shapes}")
ret.append(rest[0] if rest else 1)
return tuple(ret)
def broadcast_axes(src_shape:tuple[sint, ...], out_shape:tuple[sint, ...]) -> tuple[int, ...]:
# out axes that are added or expanded
if (nleft:=len(out_shape)-len(src_shape)) < 0: raise RuntimeError(f"cannot broadcast {src_shape} into {out_shape}")
return tuple(range(nleft)) + tuple(nleft+i for i,s in enumerate(src_shape) if resolve(s == 1, default=False) and resolve(out_shape[nleft+i] != 1))
def ssimplify(uop:sint): return uop.ssimplify() if isinstance(uop, UOp) else uop
def sym_infer(uop: UOp|int, var_vals: dict[str, int]) -> int: return uop.sym_infer(var_vals) if isinstance(uop, UOp) else uop
def range_str(u:UOp, color=False) -> str:
ret = '_'.join([str(x) if x >= 0 else "m"+str(-x) for x in u.axis_id])
return colored(ret, axis_colors[u.axis_type]) if color else ret
def multirange_str(rngs:Iterable[UOp], color=False, pad=None) -> str:
ret = ','.join([range_str(x, color=color) for x in sorted(rngs, key=lambda x: x.arg)])
if pad is not None: ret += " " * (pad-ansilen(ret))
return ret
def shape_to_shape_arg(arg:tuple[sint, ...]) -> UOp:
src = tuple(x if isinstance(x, UOp) else UOp.const(x) for x in arg)
for x in src:
if not dtypes.is_int(x.dtype): raise RuntimeError(f"shape must be int, got {x.dtype} in {arg}")
return src[0] if len(src) == 1 else UOp(Ops.STACK, src=src)
def promo_dtype(src:tuple[UOp,...]) -> DType:
dts = [x.dtype for x in src]
return dts[0] if all_same(dts) else least_upper_dtype(*dts)
def dtype_from_uop(op:Ops, src:tuple[UOp,...], arg:Any) -> DType:
# here are the dtype production rules, total over all Ops
match op:
case Ops.STORE | Ops.LINEAR | Ops.SINK | Ops.PROGRAM | Ops.SOURCE | \
Ops.BACKEDGE | Ops.BARRIER | Ops.IF | Ops.ENDIF | Ops.NOOP | \
Ops.REWRITE_ERROR | Ops.PYLITERAL:
# always void
return dtypes.void
case Ops.CALL:
# a call has the dtype of its body, void for opaque bodies
return src[0].dtype
case Ops.CUSTOM_FUNCTION:
# an external function states its return dtype in the arg
return arg.dtype
case Ops.CUSTOM | Ops.CUSTOMI:
assert isinstance(arg, tuple) and len(arg) == 2 and isinstance(arg[1], DType), f"CUSTOM/CUSTOMI arg must be (str, DType), got {arg}"
return arg[1]
case Ops.INS:
# arg is (instruction, dtype), a queue command or an asm line is void
assert isinstance(arg, tuple) and len(arg) == 2 and isinstance(arg[1], DType), f"INS arg must be (instruction, DType), got {arg}"
return arg[1]
case Ops.INDEX:
# an image access is always float, no matter the storage dtype
# TODO: should there be a CAST so src[0].dtype just work?
if (b:=src[0]).op is Ops.PARAM and is_image_shape(b.shape): return dtypes.float
return b.dtype
case Ops.LOAD | Ops.UNSHARD | Ops.REDUCE | Ops.AFTER | Ops.RANGE | \
Ops.CONTIGUOUS_BACKWARD | Ops.COPY | Ops.STAGE | Ops.DETACH | \
Ops.MSTACK | Ops.MSELECT | Ops.ALLREDUCE | Ops.SPECIAL | Ops.END:
# pass through first
return src[0].dtype
case Ops.CMPLT | Ops.CMPNE | Ops.CMPEQ:
return dtypes.bool
case Ops.SIN | Ops.LOG2 | Ops.EXP2 | Ops.SQRT | Ops.RECIPROCAL:
return dtypes.bool if src[0].base.is_invalid else least_upper_float(src[0].dtype)
case Ops.WHERE:
if src[0].dtype != dtypes.bool: raise RuntimeError(f"where cond must be bool, got {src[0].dtype}")
return promo_dtype(src[1:])
case Ops.STACK:
if len(src) == 0: return dtypes.void
return promo_dtype(src)
case Ops.WMMA:
# WMMA output dtype is the accumulator dtype (src[2])
return src[2].dtype
case Ops.GETADDR:
return dtypes.uint64
case Ops.THREEFRY:
return dtypes.uint64
case Ops.FDIV:
return least_upper_float(promo_dtype(src))
case Ops.SHL | Ops.SHR:
if not all(dtypes.is_int(x.dtype) or x.base.is_invalid for x in src):
raise RuntimeError(f"shift operands must be int, got {[x.dtype for x in src]}")
return src[0].dtype
case Ops.BUFFER | Ops.ALLOC | Ops.PARAM:
assert isinstance(arg, ParamArg), f"{op} must have ParamArg"
return arg.dtype
case Ops.BINARY:
return dtypes.uint8
case Ops.CAST | Ops.BITCAST:
assert isinstance(arg, DType), f"CAST/BITCAST arg must be DType, got {arg}"
return arg
case Ops.CONST:
# derived from the value. order matters: bool is an int subclass, ConstFloat is a float subclass
if isinstance(arg, InvalidType): return dtypes.bool # Invalid is always bool, the promo lattice bottom
if isinstance(arg, bool): return dtypes.bool
if isinstance(arg, int): return dtypes.weakint
if isinstance(arg, float): return dtypes.weakfloat
raise TypeError(f"no dtype for CONST with arg {arg}")
if op in GroupOp.Unary: return src[0].dtype
# NOTE: CMPLT, CMPNE, CMPEQ, WHERE, SHL, SHR are handled above
if op in GroupOp.Broadcastable: return promo_dtype(src)
if op in GroupOp.Movement: return src[0].dtype
raise RuntimeError(f"no dtype for {op} with arg {arg}")
class UOpMetaClass(type):
ucache:dict[tuple, weakref.ReferenceType[UOp]] = {}
def __call__(cls, op:Ops, src:tuple[UOp,...]=tuple(), arg:Any=None, tag:Any=None, metadata:tuple[Metadata,...]|None=None):
# NOTE: the key must separate nodes of different dtype: a CONST's dtype is the type of its arg, and True == 1 as dict keys
if (wret:=UOpMetaClass.ucache.get(key:=(op, src, arg, tag, type(arg)), None)) is not None and (ret:=wret()) is not None: return ret
UOpMetaClass.ucache[key] = weakref.ref(created:=super().__call__(op, src, arg, tag))
if metadata is not None: all_metadata[created] = metadata
if SPEC > 1:
from tinygrad.uop.spec import spec_full
if SPEC > 2:
# SPEC=3 checks the shape
_ = created._shape
with Context(CHECK_OOB=0): fret = cast(bool|None, spec_full.rewrite(created))
if fret is not True: raise RuntimeError(f"SPEC ISSUE {fret}: {created}")
return created
# some uops map to other stuff
all_metadata:weakref.WeakKeyDictionary[UOp, tuple[Metadata, ...]] = weakref.WeakKeyDictionary() # TODO: should this be here?
# recursive_property replaces functools.cached_property in recursive UOp functions to prevent RecursionError
class recursive_property:
def __init__(self, fxn):
self.fxn = fxn
self.nm = fxn.__name__
self.__doc__ = fxn.__doc__
def __get__(self, x:UOp|None, owner=None):
if x is None: return self
for node in x.toposort(gate=lambda node: self.nm not in node.__dict__, enter_calls=True): node.__dict__[self.nm] = self.fxn(node)
return x.__dict__[self.nm]
class _UOpTuple(tuple):
# UOps are hash-consed, so structurally identical UOps are the same object and equality can be identity.
# This keeps tuple < from rescanning deep equal prefixes with O(n) value equality at every level of the walk.
__hash__ = tuple.__hash__
def __eq__(self, other): return self is other
def __ne__(self, other): return self is not other
# we import this late so we can use resolve/smax in mixins
from tinygrad.mixin.rand import RandMixin
# NOTE: this should be frozen, but frozen is slower
@dataclass(eq=False, slots=True)
class UOp(RandMixin, metaclass=UOpMetaClass):
op:Ops
src:tuple[UOp, ...] = tuple()
arg:Any = None
tag:Any = None
@recursive_property
def dtype(self) -> DType: return dtype_from_uop(self.op, self.src, self.arg)
def __del__(self):
# NOTE: getattr because this object may be partially constructed (e.g. if __init__ raised, like the BEAM timeout SIGALRM)
try: del UOpMetaClass.ucache[(self.op, self.src, self.arg, self.tag, type(self.arg))]
except (AttributeError, KeyError): pass
def __reduce__(self): return UOp, (self.op, self.src, self.arg, self.tag, self.metadata)
def replace(self, **kwargs) -> UOp:
new_args = (kwargs.pop("op", self.op), kwargs.pop("src", self.src), kwargs.pop("arg", self.arg), kwargs.pop("tag", self.tag))
assert len(kwargs) == 0, f"unused kwargs in replace {list(kwargs)}"
if (self.op, self.src, self.arg, self.tag) == new_args: return self
return UOp(*new_args)
def rtag(self, tag=True): return self.replace(tag=tag)
@property
def val(self):
if self.op is Ops.CONST: return self.arg
if self.op is Ops.GETADDR: return cast("Buffer", self.src[0].buffer).get_buf(to_tuple(self.arg)[0])
# a casted const CAST(dt, CONST(v)) is one const: .val reads the value through the CAST
assert self.op is Ops.CAST and self.src[0].op is Ops.CONST, f"val is only for consts, got {self.op}"
return self.src[0].val
@property
def is_invalid(self) -> bool: return self.op is Ops.CONST and self.val is Invalid
@recursive_property
def key(self) -> bytes:
return hashlib.sha256(str((self.op, self.dtype, self.arg)).encode() + b"".join([s.key for s in self.src])).digest()
def __repr__(self):
from tinygrad.uop.render import pretty_print
return pretty_print(self)
def argstr(self):
if self.op is Ops.REDUCE: return f'({", ".join(map(str, self.arg))})'
return repr(self.arg)
def tagstr(self): return f", tag={self.tag}" if self.tag is not None else ""
@functools.cached_property
def backward_slice(self:UOp) -> dict[UOp, None]:
res: dict[UOp, None] = self.toposort()
res.pop(self)
return res
@property
def backward_slice_with_self(self:UOp) -> dict[UOp, None]: return {self:None, **self.backward_slice}
def op_in_backward_slice_with_self(self, *ops:Ops) -> bool:
# Check self first, then iterate backward_slice (avoids creating intermediate dict)
return self.op in ops or any(x.op in ops for x in self.backward_slice)
@recursive_property
def _bool_slice(self) -> frozenset[UOp]: return frozenset().union(*[s.bool_slice for s in self.src])
# NOTE: self is added outside the cache, a cached self-reference is a cycle the refcounter can't free
@property
def bool_slice(self) -> frozenset[UOp]: return self._bool_slice | {self} if self.dtype is dtypes.bool else self._bool_slice
def toposort(self, gate:Callable|None=None, enter_calls=False) -> dict[UOp, None]:
cache: dict[UOp, None] = {}
stack: list[tuple[UOp, bool]] = [(self, False)] # each stack entry is (node, visited_flag)
while stack:
node, visited = stack.pop()
if node in cache: continue
if not visited:
if gate is None or gate(node):
stack.append((node, True)) # push node back on stack to process after its srcs
for s in reversed(node.src if enter_calls else node.src_without_body):
stack.append((s, False)) # push srcs on the stack
else: cache[node] = None # second time i'm seeing this node, add it to returned toposort
return cache
def topovisit(self, visitor:Callable[[UOp], T], cache:dict[UOp, T]) -> T:
# NOTE: this shares a lot of code with toposort
stack: list[tuple[UOp, bool]] = [(self, False)]
while stack:
node, visited = stack.pop()
if node in cache: continue
if not visited:
stack.append((node, True))
for s in reversed(node.src): stack.append((s, False))
else: cache[node] = visitor(node)
return cache[self]
@functools.cached_property
def tuplize(self) -> _UOpTuple:
# arg goes through repr: args of different types (None, str, tuple) must stay mutually comparable for the sort
return _UOpTuple((self.op.value, repr(self.arg), self.dtype,)+tuple([x.tuplize for x in self.src]))
# *** uop shape stuff ***
@recursive_property
def _shape(self) -> tuple[sint, ...]|None:
match self.op:
# late ops don't have shape
case Ops.IF | Ops.BARRIER | Ops.SINK | Ops.REWRITE_ERROR | Ops.ENDIF | Ops.BACKEDGE | \
Ops.LINEAR | Ops.PROGRAM | Ops.SOURCE:
return None
# INS shape is always scalar, vector width is in the instruction encoding
case Ops.INS:
if self.dtype is dtypes.void: return None
return ()
# special (terrible) case for RESHAPE on NOOP
case Ops.RESHAPE:
if self.src[0].op is Ops.NOOP: return self.marg
# hacks for NOOP
case Ops.NOOP:
return self.src[0]._shape if len(self.src) >= 1 else None
case Ops.INDEX:
shp:list[sint] = []
for s in self.src[1:]: shp.extend(list(s.shape))
return tuple(shp) + self.src[0].shape[len(self.src[1:]):]
case Ops.STACK:
if len(self.src) == 0: return ()
return (len(self.src),) + self.src[0].shape
case Ops.CONST:
return ()
# some ops init the shape
case Ops.GETADDR: return ()
case Ops.RANGE | Ops.SPECIAL: return ()
case Ops.BINARY: return (len(self.arg),)
case Ops.BUFFER | Ops.ALLOC | Ops.PARAM:
assert len(self.src[0].as_shape) <= 1
if (img:=self.arg.image) is not None: return (img[0], img[1], 4)
return self.src[0].as_shape
case Ops.CUSTOM | Ops.CUSTOMI:
if self.dtype is dtypes.void: return None
input_shapes = [x._shape for x in self.src if x._shape is not None]
return _broadcast_shape(*input_shapes) if input_shapes else None
case Ops.CUSTOM_FUNCTION: return None if self.dtype is dtypes.void else ()
case Ops.PYLITERAL: return None
case Ops.STAGE:
# STAGE adds the existing shape to the front, opposite of INDEX
return tuple([int(r.vmax+1) for r in self.src[1:]])+self.src[0].shape
# wmma output shape = accumulator shape (src[2])
case Ops.WMMA:
wmma_b = _broadcast_shape(self.src[0].shape[:-1], self.src[1].shape[:-1], self.src[2].shape[:-1])
return wmma_b + (self.src[2].shape[-1],)
# passthrough ops
case Ops.MSTACK | Ops.MSELECT | Ops.DETACH | Ops.CONTIGUOUS_BACKWARD | Ops.AFTER | Ops.LOAD | \
Ops.COPY | Ops.ALLREDUCE | Ops.STORE | Ops.END | Ops.CALL:
return self.src[0]._shape
case Ops.BITCAST:
ps = self.src[0]._shape
if ps is None: return None
if (output_sz:=self.dtype.itemsize) != (input_sz:=self.src[0].dtype.itemsize) and len(ps) > 0:
if isinstance(ps[-1], int) and (ps[-1]*input_sz) % output_sz: raise RuntimeError("unsupported size in bitcast")
return ps[:-1]+(ssimplify((ps[-1]*input_sz) // output_sz),)
return ps
# movement ops change the shape
if self.op in GroupOp.Movement.union({Ops.UNSHARD, Ops.REDUCE}):
ps = self.src[0]._shape
if ps is None: raise RuntimeError(f"movement op {self.op} requires shape, {self.src[0].op} doesn't have one")
match self.op:
case Ops.RESHAPE:
if not all(x >= 0 for x in self.marg): raise ValueError(f"shape can't contain negative numbers {self.marg}")
# with symbolic views prod equality can be true at runtime but unprovable, only reject provably unequal products
if resolve(prod(ps) != prod(self.marg), False): raise ValueError(f"bad reshape: {ps} -> {self.marg}")
return self.marg
case Ops.EXPAND:
return tuple(self.marg) + ps
case Ops.PERMUTE:
if sorted(self.marg) != list(range(len(ps))): raise ValueError(f"invalid permutation {self.marg} of len {len(ps)}")
return tuple(ps[i] for i in self.marg)
case Ops.PAD:
# TODO: why do i need resolve here?
if len(ps) != len(self.marg) or not all(resolve(sz>=0) and resolve(0<=o) and resolve(o+s<=sz) for s,(o,sz) in zip(ps, self.marg)):
raise ValueError(f"invalid pad {self.marg} for {ps}")
return tuple(sz for _,sz in self.marg)
case Ops.SHRINK:
# TODO: why do i need resolve here?
if len(ps) != len(self.marg) or not all(resolve(0<=o) and resolve(sz>=0) and resolve(o+sz<=s) for s,(o,sz) in zip(ps, self.marg)):
raise ValueError(f"invalid shrink {self.marg} for {ps}")
return tuple(sz for _,sz in self.marg)
case Ops.UNSHARD: return tuple(s*(int(self.src[1:][self.arg.index(a)].vmax)+1) if a in self.arg else s for a,s in enumerate(ps))
case Ops.REDUCE:
num_axes = self.arg[1]
if not isinstance(num_axes, int) or num_axes < 0 or num_axes > len(ps):
raise ValueError(f"invalid type for axis: {num_axes}")
return ps[num_axes:]
if self.op in GroupOp.Unary.union({Ops.CAST}):
assert len(self.src) == 1, "unary ops must have 1 src"
return self.src[0]._shape
# elementwise ops keep the shape the same. all inputs with shape must match
if self.op in GroupOp.Broadcastable:
input_shapes = [x._shape for x in self.src]
assert len(self.src) > 0 and all(x is not None for x in input_shapes), f"None input shape not supported for {self.op}"
if DISALLOW_BROADCAST and not all_same(input_shapes):
raise RuntimeError(f"shape mismatch at {self.op}: {input_shapes} {[x.op for x in self.src]}")
# broadcasting lives in _shape property now
return _broadcast_shape(*input_shapes)
# all Ops must be explicitly handled
raise NotImplementedError(f"no shape handling for {self.op} with {self.dtype}")
@property
def shape(self) -> tuple[sint, ...]:
if (ret:=self._shape) is None: raise RuntimeError(f"shape requested, but {self.op} doesn't have a shape")
return ret
@property
def shard_shape(self) -> tuple[sint, ...]:
if not isinstance(self.device, tuple) or self.axis is None: return self.shape
dcount = int(self.src[1].vmax)+1 if self.op is Ops.UNSHARD else len(self.device)
return tuple(x//dcount if i == self.axis else x for i,x in enumerate(self.shape))
@property
def max_shard_shape(self) -> tuple[int, ...]: return to_max_shape(self.shard_shape)
@functools.cached_property
def ended_ranges(self) -> tuple[UOp, ...]:
if self.op is Ops.CALL and self.body.op is Ops.CUSTOM_FUNCTION: return ()
if self.op is Ops.END: return tuple(r for r in self.src[1:] if r.op is Ops.RANGE)
if self.op is Ops.BACKEDGE: return self.src[1:2] # the condition's other ranges remain live
if self.op in range_start: return self.src[range_start[self.op]:]
if self.op is Ops.AFTER: return tuple(flatten([x.ended_ranges for x in self.src[1:]]))
if self.op is Ops.BARRIER: return tuple(flatten([x.ended_ranges for x in self.src]))
# UNSHARD ends the DEVICE range: its src is per-device index math, the device axis is carried by the axis metadata
if self.op is Ops.UNSHARD: return self.src[1:]
return ()
# determine what ranges this is in
@recursive_property
def _ranges(self) -> dict[UOp, None]:
ret: dict[UOp, None] = {}
for s in self.src: ret.update(s.ranges)
for er in self.ended_ranges:
if er.op is Ops.RANGE:
# if it's a single RANGE, we don't flow through it.
ret.pop(er, None)
else:
# if it's not a RANGE, we include all ranges in srcs.
# technically we shouldn't flow through these ranges either, but this is pre pm_add_control_flow so it's the same.
for s in er.ranges: ret.pop(s, None)
return ret
@property
def ranges(self) -> dict[UOp, None]:
if self.op is Ops.RANGE: return {self:None} | self._ranges
return self._ranges
@property
def axis_id(self) -> tuple[int, ...]:
assert self.op is Ops.RANGE, f"axis_id is only for RANGE, not {self.op}"
return self.arg[1:]
@property
def axis_type(self) -> AxisType:
assert self.op is Ops.RANGE, f"axis_type is only for RANGE, not {self.op}"
return self.arg[0]
# *** uop evaluation ***
def simplify(self, tracked=False):
if self.op is Ops.CONST: return self
if self.op is Ops.SINK and all(s.op is Ops.CONST or (s.op is Ops.STACK and len(s.src) == 0) for s in self.src): return self
# late import!
from tinygrad.uop.symbolic import symbolic
with Context(TRACK_MATCH_STATS=0 if not tracked else TRACK_MATCH_STATS.value):
return graph_rewrite(self, symbolic, name="simplify")
def ssimplify(self) -> UOp|ConstType:
if (ret := self.simplify()).op is Ops.CAST and ret.src[0].op is Ops.CONST: return ret.dtype.const(ret.src[0].val)
return ret.val if ret.op is Ops.CONST else ret
def _eval(self, dtype, expected_type:Type[T]) -> T:
assert self.dtype in dtype, f"eval with wrong dtype {self}"
vmin, vmax = (simple_self:=self.simplify())._min_max
if vmin != vmax: raise ValueError(f"eval failed to be a single number, range is {vmin} to {vmax} in {simple_self.render()}")
assert isinstance(vmin, expected_type), f"vmin is wrong dtype {type(vmin)} != {expected_type}"
return vmin
def __bool__(self): return self._eval((dtypes.bool,), bool)
def __int__(self): return self._eval(dtypes.ints+(dtypes.weakint,), int)
def __float__(self): return float(self._eval(dtypes.floats+(dtypes.weakfloat,), float))
def substitute(self, dvars:dict[UOp, UOp], name:str|None=None, extra_pm:PatternMatcher|None=None, walk:bool=False, enter_calls:bool=False):
dvars = {k:v for k,v in dvars.items() if k is not v}
if len(dvars) == 0: return self
with Context(TRACK_MATCH_STATS=(0 if name is None else TRACK_MATCH_STATS.value)):
return graph_rewrite(self, (extra_pm+_substitute) if extra_pm is not None else _substitute, dvars,
bottom_up=True, walk=walk, enter_calls=enter_calls, name=name)
# NOTE: this is not called by Tensor slice (Tensor handles UOps directly), but satisfies SupportsIndex for type checking
def __index__(self): return self.__int__()
# *** uop tracing stuff ***
@recursive_property
def trace_num(self):
num = next(ucount)
# tags can contain UOps (callify tags nodes with their originals): store them as trace_nums, same as srcs
tag = tuple(t.trace_num if isinstance(t, UOp) else t for t in self.tag) if isinstance(self.tag, tuple) else self.tag
# the trace must not retain the device Buffer: store a placeholder instead (the real one would pin memory and fail pickling)
arg = replace(self.arg, buffer=cast("Buffer", object())) if isinstance(self.arg, ParamArg) and self.arg.buffer is not None else self.arg
uop_fields[num] = (self.op, tuple(s.trace_num for s in self.src), arg, tag)+((self.metadata,) if TRACEMETA>=2 else ())
return num
# *** uop syntactic sugar ***
def sink(*srcs:UOp|None, **kwargs): # pylint: disable=no-self-argument
return UOp(Ops.SINK, src=tuple([x for x in srcs if x is not None]), **kwargs)
def group(*srcs:UOp|None, **kwargs): # pylint: disable=no-self-argument
if len(srcs) == 1 and isinstance(srcs[0], UOp): return srcs[0]
return UOp(Ops.STACK).after(*[x for x in srcs if x is not None], **kwargs)
@property
def body(self) -> UOp:
"""the body of a CALL: the program, copy or function reference being called (its first src)"""
if self.op is not Ops.CALL: raise RuntimeError(f"body requested, but {self.op} is not a CALL")
return self.src[0]
@property
def is_inline_call(self) -> bool:
return self.op is Ops.CALL and self.body.op is Ops.SINK and self.body.arg is None and not self.arg.precompile
@property
def has_unbound_outputs(self) -> bool:
"""does this call still have unresolved outputs: ALLOCs among its inputs (minted by call_with_outputs,
resolved when the call is inlined or the outputs are materialized). a lifecycle query, not a call type"""
return self.op is Ops.CALL and any((b:=x.unsharded_base).op is Ops.ALLOC and not b.arg.bind_on_realize for x in self.src[1:])
@property
def unbound_outputs(self) -> tuple[UOp, ...]:
"""the unresolved outputs of this call: an AFTER on each ALLOC input, usable like a normal buffer"""
return tuple(x.after(self) for x in self.src[1:] if (b:=x.unsharded_base).op is Ops.ALLOC and not b.arg.bind_on_realize)
def index(self, *srcs:UOp|int|None, **kwargs):
new_srcs: list[UOp] = [UOp.const(x) if isinstance(x, int) else x for x in srcs if x is not None]
if len(new_srcs) == 1 and new_srcs[0].op is Ops.CONST and self.op is Ops.STACK: return self.src[new_srcs[0].val]
return UOp(Ops.INDEX, src=(self,)+tuple(new_srcs), **kwargs)
def __getitem__(self, idx):
# buffers index into INDEX UOps (scalar lookup); everything else uses the shared mixin view path
if self.addrspace in (None, AddrSpace.ALU) or self.device is not None: return super(UOp, self).__getitem__(idx)
idx = self._normalize_indices(list(argfix(idx)))
if len(slice_idx:=[i for i,x in enumerate(idx) if isinstance(x, slice)]):
# apply SHRINK for slices that aren't the full range
bounds = tuple((s.start or 0, s.stop if s.stop is not None else self.shape[i]) if isinstance(s, slice) else (0, self.shape[i])
for i, s in enumerate(idx))
src = self.shrink(bounds)
non_slice_args = [x for x in idx if not isinstance(x, slice)]
if not non_slice_args: return src # all dims are slices, no indexing needed
perm = src.permute(tuple([i for i in range(src.ndim) if i not in slice_idx] + slice_idx))
return perm.index(*non_slice_args)
return self.index(*idx)
@property
def _uop(self) -> UOp: return self
@classmethod
def _wrap_uop(cls, u:UOp) -> UOp: return u
def const_like(self, b:ConstLike, dtype:DType|None=None):
ret = UOp.const(b, dtype or self.dtype)
return ret._mop(Ops.EXPAND, arg=self._shape) if self._shape and ret._shape != self._shape else ret
def vconst_like(self, b:ConstLike):
# for use after movement ops have been removed
return UOp.const(b, self.dtype).broadcast(self.max_numel())
def ufix(self, x):
if isinstance(x, UOp): return x
return UOp.const(x)
def broadcast(self, count:int):
if count == 1: return self
return UOp(Ops.STACK, src=(self,)*count)
def load(self, *src:UOp, **kwargs): return UOp(Ops.LOAD, src=(self,)+src, **kwargs)
def store(self, src:UOp|ConstType, gate:UOp|None=None, **kwargs):
srcs = (self, self.const_like(src) if not isinstance(src, UOp) else src) + ((gate,) if gate is not None else ())
return UOp(Ops.STORE, src=srcs, **kwargs)
def end(self, *src:UOp): return UOp(Ops.END, src=(self,)+src) if len(src) else self
def backedge(self, loop:UOp, cond:UOp): return UOp(Ops.BACKEDGE, src=(self, loop, cond))
def after(self, *src:UOp, **kwargs): return UOp(Ops.AFTER, src=(self,)+src, **kwargs) if len(src) else self
@property
def without_after(self) -> UOp: return self.src[0].without_after if self.op is Ops.AFTER else self
def barrier(self, *src:UOp): return UOp(Ops.BARRIER, src=(self,)+src)
def ins(self, arg, **kwargs): return UOp(Ops.INS, kwargs.pop("src", self.src), (arg, kwargs.pop("dtype", self.dtype)), kwargs.pop("tag", self.tag))
def contract(self, *rngs:UOp):
assert all(x.axis_type == AxisType.UPCAST for x in rngs), "all contract ranges must be upcast"
return UOp.stack(*[self.substitute(dict(zip(rngs, [r.const_like(i) for r,i in zip(rngs, idx)])))
for idx in itertools.product(*[range(int(r.vmax)+1) for r in rngs])])
def alu(self, op, *src:UOp, **kwargs): return UOp(op, src=(self, *src), **kwargs)
@staticmethod
def const(b:ConstLike, dtype:DType|None=None):
if dtype is None or b is Invalid: dtype = dtypes.from_py(b)
if isinstance(b, UOp): return b.cast(dtype)
# NOTE: it always has to be STACK now, even if they are all the same
if isinstance(b, tuple): return UOp.stack(*[UOp.const(c, dtype) for c in b])
# .cast folds away at exactly the dtypes a CONST derives (bool/weakint/weakfloat): bare there, the pair everywhere else
return UOp(Ops.CONST, arg=dtype.const(b), src=()).cast(dtype)
# cast, except for CONST, in which case rebuild a new CONST at the dtype
def ccast(self, dtype:DType): return UOp.const(self.val, dtype) if self.op is Ops.CONST else self.cast(dtype)
# a forced CAST for bool: .cast(bool) folds, so UOp.const cannot state the width
@staticmethod
def cconst(b:ConstLike, dtype:DType): return UOp(Ops.CAST, src=(UOp.const(b),), arg=dtype)
@staticmethod
def range(end:sint, axis_id, axis_type=AxisType.WEAK, *, dtype=dtypes.weakint, **kwargs):
return UOp(Ops.RANGE, src=(sint_to_uop(end, dtype),), arg=(axis_type, axis_id), **kwargs)
@staticmethod
def loop(axis_id:int): return UOp(Ops.RANGE, src=(UOp(Ops.NOOP),), arg=(AxisType.WEAK, axis_id))
@staticmethod
def special(end:sint, name:str): return UOp(Ops.SPECIAL, src=(sint_to_uop(end),), arg=name)
@staticmethod
def wmma(a:UOp, b:UOp, acc:UOp, dims:tuple[int, int, int], threads:int, tc_upcast_axes=None):
# dtype_in is stored in the arg (not derived from src[0].dtype) because bitcast rewrites change src dtypes
return UOp(Ops.WMMA, src=(a, b, acc), arg=(dims, a.dtype, threads, tc_upcast_axes))
def _rop(self, op:Ops, axis:tuple[int, ...]):
# NOTE: we don't allow reduce on 1s axis
axis = tuple(sorted(axis))
reduce_axis = tuple(x for x in axis if resolve(self.shape[x] != 1))
if not len(reduce_axis):
return self.reshape(tuple(s for i,s in enumerate(self.shape) if i not in axis))
# permute so reduced axes are at the front
perm = reduce_axis + tuple(i for i in range(len(self.shape)) if i not in reduce_axis)
ret = UOp(Ops.REDUCE, src=(self.permute(perm),), arg=(op, len(reduce_axis)))
return ret.reshape(tuple(s for i,s in enumerate(self.shape) if i not in axis)) if axis != reduce_axis else ret
@staticmethod
def invalid() -> UOp: return UOp.const(Invalid)
def valid(self, cond) -> UOp: return cond.where(self, self.const_like(Invalid))
def get_idx(self) -> UOp:
if self.op is Ops.STACK: return UOp.stack(*(x.get_idx() for x in self.src))
return self.src[1] if self.op is Ops.WHERE and self.src[2].is_invalid else self
def get_valid(self) -> UOp:
if self.op is Ops.STACK: return UOp.stack(*(x.get_valid() for x in self.src))
return self.src[0] if self.op is Ops.WHERE and self.src[2].is_invalid else UOp.const(not self.is_invalid)
def reduce(self, *src:UOp, **kwargs):
arg = kwargs.pop('arg', None)
if isinstance(arg, Ops): arg = (arg, 0)
return UOp(Ops.REDUCE, src=(self,)+src, arg=arg, **kwargs)
def bufferize(self, *args, **kwargs): return UOp(Ops.STAGE, src=(self,)+args, **kwargs)
def allreduce(self, op, device:str|tuple[str, ...]):
assert isinstance(self.device, tuple), f"allreduce must be on tuple {self.device} isn't"
return UOp(Ops.ALLREDUCE, src=(self,), arg=(op, device))
def overflows(self, dtype:DType) -> bool: return self.vmin < dtype.min or dtype.max < self.vmax
def split_uop(self:UOp, sep:Ops) -> Iterator[UOp]:
if self.op is sep:
for s in self.src: yield from s.split_uop(sep)
else: yield self
# *** multi-device helpers ***
def unshard(self, axis:int|tuple[int, ...]|None, device_range:UOp|tuple[UOp, ...]|None=None):
assert axis is not None, "multi None is no longer supported"
# an UNSHARD carries the value and one sharding range per sharded axis (arg is the tuple of sharded axes,
# sorted). the single-axis axis form defaults the range to a DEVICE range over the devices; a range need not
# be DEVICE, e.g. a LOCAL range shards a kernel tile into per-thread fragments
if isinstance(axis, int): axis = (axis,)
if device_range is None:
assert isinstance(self.device, tuple), f"multi device must be tuple, {self.device} isn't"
device_range = (UOp.range(len(self.device), 0, AxisType.DEVICE),)
if isinstance(device_range, UOp): device_range = (device_range,)
assert isinstance(device_range, tuple) and len(axis) == len(device_range) and len(set(axis)) == len(axis)
axis, device_range = map(tuple, zip(*sorted(zip(axis, device_range))))
return UOp(Ops.UNSHARD, src=(self, *device_range), arg=axis)
@property
def sharding(self) -> tuple[tuple[int, UOp], ...]:
"""(axis, RANGE) pairs this value is sharded over (the source of truth for shard bounds/counts)."""
return tuple(zip(self.arg, self.src[1:])) if self.op is Ops.UNSHARD else ()
@property
def bounds(self):
if self.axis is None: raise RuntimeError("bounds is not defined when axis is None")
dcount = int(self.src[1].vmax)+1 if self.op is Ops.UNSHARD else len(self.device)
return tuple(itertools.pairwise(itertools.accumulate([self.src[0].shape[self.axis] for _ in range(dcount)], initial=0)))
@functools.cached_property
def axis(self) -> int|None:
# COPY removes axis. TODO: add more tests for this, and consider MSELECT/MSTACK
if self.op is Ops.COPY: return None
if self.op is Ops.UNSHARD:
if len(self.arg) != 1: raise RuntimeError(f"UOp is sharded on multiple axes {self.arg}, use .sharding")
return self.arg[0]
# NOTE: they all have to share an axis, we always choose [-1]. src axes are right-aligned into the output shape
if self.op in GroupOp.ALU.union({Ops.STACK}):
return axes[-1] if (axes := dedup([x.axis+len(self.shape)-len(x.shape) for x in self.src if x.axis is not None])) else None
if len(self.src) == 0: return None
src_axis = self.src[0].axis
if self.op is Ops.SHRINK and src_axis is not None and self.marg[src_axis] != (0, self.src[0].shape[src_axis]):
return None # SHRINK will remove the sharding if it's on axis
if self.op is Ops.REDUCE:
if src_axis is None: return None
if src_axis < self.arg[1]: return None
return src_axis - self.arg[1]
if self.op is Ops.RESHAPE:
if src_axis is None: return None
arg_acc:list[sint] = [ssimplify(x) for x in itertools.accumulate(self.marg, operator.mul, initial=1)]
# new_axis is the last one that preserves prod(prior to new_axis) and must not move items between shards
target = ssimplify(prod(self.src[0].shape[:src_axis]))
if target not in arg_acc: raise RuntimeError(f"reshape {self.src[0].shape} -> {self.shape} moved items between shards")
new_axis = len(arg_acc) - arg_acc[::-1].index(target) - 1
dcount = len(self.device) if isinstance(self.device, tuple) else \
int(next(u.src[1] for u in self.src[0].toposort() if u.op is Ops.UNSHARD).vmax)+1
if self.shape[new_axis] % dcount != 0: raise RuntimeError(f"reshape {self.src[0].shape} -> {self.shape} moved items between shards")
return new_axis
if self.op is Ops.PERMUTE: return self.marg.index(src_axis) if src_axis is not None else None
if self.op is Ops.EXPAND: return src_axis + len(self.marg) if src_axis is not None else None
return src_axis
def _shard(self, axis:int, rng:UOp) -> UOp:
if len(self.shape) == 0: return self # scalars broadcast, no sharding needed
dcount = int(rng.vmax)+1
if self.shape[axis] % dcount != 0: raise RuntimeError(f"multi axis uneven: {self.shape[axis]=} {axis=} {dcount=}")
sz = self.shape[axis] // dcount
return self.shrink(tuple((0,s) if i != axis else (rng*sz,rng*sz+sz) for i,s in enumerate(self.shape)))
def shard(self, devices:tuple[str, ...], axis:int|None=None) -> UOp:
copied = self.copy_to_device(devices)
return copied if axis is None else copied._shard(axis, UOp.range(len(devices), 0, AxisType.DEVICE)).unshard(axis)
def copy_to_device(self, device:str|tuple[str, ...], arg=None):
if is_disk_device(device):
raise RuntimeError("COPY to DISK is not allowed; use STORE to an explicit disk buffer")
assert arg is None or isinstance(self.device, tuple)
inp = self if arg is None else UOp(Ops.MSELECT, src=(self,), arg=arg)
if inp.dtype in dtypes.weaks: raise RuntimeError(f"cannot create storage for weak dtype {inp.dtype}")
# multi-device COPYs carry the DEVICE range as src[1] (like UNSHARD's sharding ranges)
return UOp(Ops.COPY, src=(inp, *UOp.device_range_src(device)), arg=device)
def store_call(self, src:UOp) -> UOp:
"""Executable bulk transfer into this buffer."""
return self.param_like(0).store(src.param_like(1)).call(self, src)
def mselect(self, arg:int) -> UOp: return UOp(Ops.MSELECT, src=(self,), arg=arg)
def mstack(self, *srcs: UOp) -> UOp: return UOp(Ops.MSTACK, src=(self,)+srcs) if len(srcs) else self
@property
def metadata(self) -> tuple[Metadata, ...]|None: return all_metadata.get(self, None)
# little helpers
def on_disk(self:UOp): return isinstance(self.device, str) and self.device.startswith("DISK")
def on_creation_device(self:UOp): return isinstance(self.device, str) and self.device.startswith(("DISK", "NPY", "PYTHON"))
def needs_storage(self:UOp) -> bool: return not self.is_virtual and (self.storage_base.op is Ops.ALLOC or not self.has_buffer_identity())
# *** uop movement ops ***
@property
def base(self) -> UOp:
if self.op in GroupOp.Movement: return self.src[0].base
if self.op is Ops.DETACH: return self.src[0].base # DETACH can't change base
return self
# base with UNSHARD
@property
def unsharded_base(self) -> UOp:
if self.op in GroupOp.Movement: return self.src[0].base
if self.op is Ops.DETACH: return self.src[0].base # DETACH can't change base
# TODO: why can't this be in normal base?
if self.op is Ops.UNSHARD: return self.src[0].base
return self
# the storage this uop ultimately targets: base with UNSHARD, BITCAST and AFTER stripped
@property
def storage_base(self) -> UOp:
b = self.unsharded_base
while b.op in {Ops.BITCAST, Ops.AFTER, Ops.UNSHARD}: b = b.src[0].unsharded_base
return b
# cached property here makes external_uop_gc fail, why?
@property
def as_shape(self) -> tuple[sint, ...]:
if self.op is Ops.CONST: return (self.val,)
if self.op is not Ops.STACK: return (ssimplify(self),)
return tuple(s.val if s.op is Ops.CONST else ssimplify(s) for s in self.src)
@functools.cached_property
def marg(self):
match self.op:
case Ops.RESHAPE | Ops.EXPAND: return self.src[1].as_shape
case Ops.PAD | Ops.SHRINK: return tuple(zip(self.src[1].as_shape, self.src[2].as_shape))
case Ops.PERMUTE: return self.arg
case _: raise RuntimeError(f"{self.op} is not a MovementOp")
def _mop(self, op:Ops, arg) -> UOp:
# early NOOP
if op is Ops.EXPAND and len(arg) == 0: return self
if op in {Ops.SHRINK, Ops.PAD} and len(arg) == 0:
assert len(self.shape) == 0, "0 len arg only valid on zero length shape"
return self
match op:
case Ops.RESHAPE | Ops.EXPAND: src_args = [arg]
case Ops.PAD | Ops.SHRINK: src_args = list(zip(*arg))
case Ops.PERMUTE: src_args = []
case Ops.STACK:
srcs = (self,)+tuple(arg)
dtype = dtype_from_uop(Ops.STACK, srcs, None)
return UOp(Ops.STACK, src=tuple(u if u.base.is_invalid else u.ccast(dtype) for u in srcs))
case _: raise RuntimeError(f"{op} is not a MovementOp")
usrcs = [shape_to_shape_arg(arg) for arg in src_args]
if len(usrcs) == 0: return UOp(op, src=(self,), arg=arg)
return UOp(op, src=(self,)+UOp.sink(*usrcs).simplify().src)
# *** uop Buffer stuff ***
unique_num = itertools.count(0)
def getaddr(self, device=None) -> UOp:
if self.without_after.op not in {Ops.BUFFER, Ops.ALLOC, Ops.SHRINK, Ops.BITCAST, Ops.BINARY,
Ops.MSTACK, Ops.MSELECT, Ops.PARAM, Ops.LINEAR}: return self
return UOp(Ops.GETADDR, src=(self,), arg=device or to_tuple(self.device)[0])
@staticmethod
def device_range_src(device:str|tuple[str, ...]|None) -> tuple[UOp, ...]:
# BUFFER/ALLOC/COPY carry a DEVICE range when targeting multiple devices
return (UOp.range(len(device), 0, AxisType.DEVICE),) if isinstance(device, tuple) else ()
@staticmethod
def _buffer(size:int, dtype:DType, slot:int, addrspace=AddrSpace.GLOBAL, device=None, buffer:Buffer|MultiBuffer|None=None) -> UOp:
return UOp(Ops.BUFFER, src=(shape_to_shape_arg((size,)),)+UOp.device_range_src(device),
arg=ParamArg(slot, dtype, addrspace=addrspace, device=device, buffer=buffer))
@staticmethod
def new_buffer(device:str|tuple[str, ...], size:int, dtype:DType, num=None):
if dtype in dtypes.weaks: raise RuntimeError(f"cannot create storage for weak dtype {dtype}")
assert dtype.itemsize > 0, f"not a storage dtype: {dtype}"
assert isinstance(size, int), f"new_buffer size must be a concrete int, got {size}"
slot = next(UOp.unique_num) if num is None else num
buf = MultiBuffer(device, size * dtype.itemsize) if isinstance(device, tuple) else Buffer(device, size * dtype.itemsize)
return UOp._buffer(size, dtype, slot, device=device, buffer=buf)
@staticmethod
def from_buffer(opaque:Buffer|MultiBuffer, dtype:DType, device:str|tuple[str, ...]|None=None):
if dtype in dtypes.weaks: raise RuntimeError(f"cannot create storage for weak dtype {dtype}")
assert dtype.itemsize > 0, f"not a storage dtype: {dtype}"
assert opaque.nbytes % dtype.itemsize == 0, f"{opaque.nbytes} bytes is not divisible by {dtype.itemsize} for {dtype}"
# Both storage identity and interpretation are part of the UOp key.
return UOp._buffer(opaque.nbytes // dtype.itemsize, dtype, -id(opaque), device=device or opaque.device, buffer=opaque)
def empty_like(self, dtype:DTypeLike|None=None, device:str|tuple[str, ...]|None=None) -> UOp:
device = canonicalize_device(self.device if device is None else device)
axis = self.axis if isinstance(device, tuple) else None
ret = UOp.empty(self.shard_shape if axis is not None else self.shape, dtype=self.commit_dtype() if dtype is None else dtype, device=device)
return ret.unshard(axis) if axis is not None else ret
@staticmethod
def _frompy(x:list|tuple|bytes, dtype:DType) -> UOp:
if isinstance(x, bytes): ret, data = UOp.new_buffer("PYTHON", len(x)//dtype.itemsize, dtype), x
else:
# bfloat16 and fp8 have no struct format, so pack a float32 buffer and cast
bdtype = dtypes.float32 if dtype in [dtypes.bfloat16, *dtypes.fp8s] else dtype
assert bdtype.fmt is not None, f"{bdtype=} has None fmt"
ret = UOp.new_buffer("PYTHON", prod(shape:=get_shape(x)), bdtype).reshape(shape)
data = struct.pack(f"{prod(shape)}{bdtype.fmt}", *[truncate[bdtype](bdtype.const(xi)) for xi in fully_flatten(x)])
if not data: ret.buffer.allocate(memoryview(bytearray()))
else: (buf:=ret.buffer.ensure_allocated()).allocator._copyin(buf._buf, memoryview(data))
if ret.dtype != dtype: ret = ret.cast(dtype)
return ret
def clone(self, device=None) -> UOp:
device = canonicalize_device(device or self.device)
if is_disk_device(device):
raise RuntimeError("cannot clone DISK storage; use STORE to an explicit disk buffer")
ret = self.empty_like(device=device)
src = self if self.device is None or self.device == device else self.copy_to_device(device)
return ret.after(ret.store(src.cast(ret.dtype)))
@recursive_property
def device(self) -> str|tuple[str, ...]|None:
if self.op is Ops.PARAM: return self.arg.device
if self.op is Ops.STAGE: return self.src[0].device if self.arg is None else self.arg.device
if self.op is Ops.AFTER: return self.src[0].device
if self.op is Ops.MSELECT:
assert isinstance(self.src[0].device, tuple), f"mselect must be on tuple device, getting {self.src[0].device}"
return self.src[0].device[self.arg]
if self.op is Ops.MSTACK: return tuple(cast(str, x.device) for x in self.src)
if self.op in {Ops.BUFFER, Ops.ALLOC}: return self.arg.device
if self.op is Ops.COPY: return self.arg
if self.op is Ops.ALLREDUCE: return self.arg[1]
if self.op is Ops.CALL and self.body.op is Ops.STORE: return self.src[2].device # a copy pushes from its src
for x in self.src:
if x.device is not None: return x.device
return None
@property
def is_virtual(self) -> bool:
# NOTE: no device means no place to store, weak means no width to store. neither can back a buffer as-is
# TODO: unify with has_buffer_identity
return self.device is None or self.dtype in dtypes.weaks
@recursive_property
def addrspace(self) -> AddrSpace|None:
if self.op is Ops.PARAM: return self.arg.addrspace
if self.op in {Ops.BUFFER, Ops.ALLOC}: return self.arg.addrspace
if self.op in {Ops.SPECIAL, Ops.RANGE, Ops.CONST}: return AddrSpace.ALU
if self.op is Ops.BINARY: return AddrSpace.GLOBAL
if self.op is Ops.LOAD: return AddrSpace.ALU # LOAD brings things into the ALU
if self.op in {Ops.INDEX, Ops.CAST, Ops.AFTER, Ops.REDUCE, Ops.STORE, Ops.MSTACK, Ops.MSELECT, Ops.END, Ops.UNSHARD}:
return self.src[0].addrspace
if self.op in GroupOp.Movement: return self.src[0].addrspace
if self.op in {Ops.STACK, Ops.WMMA} or self.op in GroupOp.Elementwise:
ad = [x.addrspace for x in self.src if x.addrspace is not None]
if not len(ad) or not all_same(ad): return None
return ad[0]
return None
@property
def buf_uop(self) -> UOp:
if self.op in GroupOp.Defines: return self
if self.op is Ops.MSELECT: return self.src[0].buf_uop.mselect(self.arg)
if self.op is Ops.MSTACK: return UOp(Ops.MSTACK, src=tuple(x.buf_uop for x in self.src))
if self.base.op is Ops.AFTER: return self.base.src[0].buf_uop.base
s = self
while len(s.src) and s.op not in {Ops.BUFFER, Ops.ALLOC, Ops.PARAM, Ops.STAGE, Ops.MSTACK}: s = s.src[0]
return s
def contiguous_view(self) -> tuple[UOp, int]|None:
from tinygrad.schedule.prepare import pm_mops
from tinygrad.uop.symbolic import symbolic
# WEBGPU and CL do not support views.
# WEBGPU requires that minUniformBufferOffsetAlignment be at least 32 bytes: https://gpuweb.github.io/gpuweb/#adapter-capability-guarantees
# CL 1.1 provides the clCreateSubBuffer API, but at the time of writing, relevant CL runtimes (rusticl, adreno, nvidia, amd) do not provide
# reasonable values for CL_DEVICE_MEM_BASE_ADDR_ALIGN. cl_ext_buffer_device_address could potentially help, but this extension is not provided
# by relevant CL runtimes at time of writing.
if (dev:=self.device) is not None and any(d.startswith(("WEBGPU", "CL")) for d in to_tuple(dev)): return None
rng = UOp.range(self.numel(), 0)
out = graph_rewrite(self.flatten().index(rng), pm_mops+symbolic, name="contiguous_view_offset")
if out.op is not Ops.INDEX or len(out.src)-1 != len((b:=out.src[0]).shape): return None
offset = (sum(i*s for i,s in zip(out.src[1:], strides_for_shape(b.shape))) - rng).ssimplify()
if not isinstance(offset, int): return None
if b.op is not Ops.BITCAST: return b, offset
osz, isz = b.element_size(), b.src[0].element_size()
if (offset*osz) % isz or (self.numel()*osz) % isz: return b, offset
return b.src[0].flatten()[offset*osz//isz:(offset+self.numel())*osz//isz].contiguous_view()
def contiguous_view_offset(self) -> int|None: return None if (view := self.contiguous_view()) is None else view[1]
def has_buffer_identity(self, after_ok=False):
"""Check if this UOp has a storage identity in the graph, whether or not its buffer is bound."""
# TODO: this is confusing because UOp.variable('v', 0, 1, dtypes.weakfloat) is True for jit to work, but it doesn't have a buffer
if self.op in {Ops.RESHAPE, Ops.UNSHARD, Ops.MSELECT}: return self.src[0].has_buffer_identity(after_ok)
if after_ok and self.op == Ops.AFTER: return self.src[0].has_buffer_identity(after_ok)
return self.op in GroupOp.Defines
def _base_buffer_is_realized(self) -> bool:
"""Walk through AFTER chain to find if the underlying buffer is realized (has allocated memory)."""
u = self.base
while u.op is Ops.AFTER: u = u.src[0]
return u.is_realized
@functools.cached_property
def _buffer_view(self) -> tuple[UOp, int]:
# Cache only the UOp and byte offset, never an allocated Buffer view.
if (cv := self.contiguous_view()) is None: raise RuntimeError(f"non-contiguous view is not supported for {self.device} buffer")
return cv[0], cv[1]*cv[0].dtype.itemsize
@property
def buffer(self) -> Buffer|MultiBuffer:
# a bare STAGE (same-device materialization) keeps the source's buffer
if self.op is Ops.STAGE and self.arg is None: return self.src[0].buffer
if self.op in {Ops.CONTIGUOUS_BACKWARD, Ops.RESHAPE, Ops.UNSHARD, Ops.DETACH, Ops.AFTER, Ops.BITCAST}: return self.src[0].buffer
# this buffer can process disk tensors and simple movement ops.
# NOTE: the view Buffer returned here is transient (short-lived), it only wraps an offset into the base BUFFER's storage
if self is not self.base:
src, offset = self._buffer_view
if isinstance(buf:=src.buffer, MultiBuffer):
mbuf = MultiBuffer.__new__(MultiBuffer)
mbuf.bufs = [x.view(prod(self.max_shape) * self.dtype.itemsize, offset) for x in buf.bufs]
return mbuf
return buf.view(prod(self.max_shape) * self.dtype.itemsize, offset)
if self.op is Ops.MSELECT:
ret = self.src[0].buffer
assert isinstance(ret, MultiBuffer)
return ret.bufs[self.arg]
if self.op is Ops.MSTACK:
ret = MultiBuffer.__new__(MultiBuffer)
ret.bufs = [cast(Buffer, x.buffer) for x in self.src]
assert all_same([x.nbytes for x in ret.bufs]) and all_same([x.dtype for x in self.src]), "multibuffers mismatch buffers"
return ret
assert self.op is Ops.BUFFER and self.arg.buffer is not None, f"must be a realized BUFFER {self}"
return self.arg.buffer
@property
def realized(self) -> Buffer|MultiBuffer|None:
if self.op is Ops.UNSHARD: return self.src[0].realized
# only these can be realized
if self.op not in (Ops.BUFFER, Ops.MSTACK): return None
# LOCAL/REG scratch buffers are never realized
if self.op is Ops.BUFFER and self.addrspace in (AddrSpace.LOCAL, AddrSpace.REG): return None
# an ALLOC (directly or as an MSTACK source) is not realized
if self.op_in_backward_slice_with_self(Ops.ALLOC): return None
# NOTE: this is used by the JIT to determine which inputs we capture
return self.buffer if self.buffer.is_allocated() else None
@property
def is_realized(self) -> bool: return self.base.realized is not None
# *** uop Variable stuff ***
@staticmethod
def variable(name:str, min_val:PyConst, max_val:PyConst, dtype:DType=dtypes.weakint, multiple_of:int=1) -> UOp:
# a Variable is a scalar ALU PARAM with a name and a value range; binding it sets the val payload on the arg
return UOp(Ops.PARAM, src=(UOp(Ops.STACK),), arg=ParamArg(-1, dtype, name=name, vmin_vmax=(min_val, max_val),
multiple_of=multiple_of, addrspace=AddrSpace.ALU))
@property
def is_variable(self) -> bool:
# a Variable is a scalar ALU PARAM that carries a value range
return self.op is Ops.PARAM and self.arg.vmin_vmax is not None and self.arg.addrspace is AddrSpace.ALU and self._shape == ()
@property
def is_bound_var(self) -> bool:
# a bound Variable is a Variable with the val payload set
return self.is_variable and self.arg.val is not None
@property
def expr(self) -> str:
assert self.op in {Ops.PARAM, Ops.BUFFER}
return unwrap(self.arg.name)
def bind(self, val:int|UOp):
assert self.is_variable and not self.is_bound_var, f"op is {self.op}, need an unbound Variable"
uval = UOp.const(val) if isinstance(val, int) else val
assert uval.op is Ops.CONST, f"bind value must be a CONST, not {uval.op}"
assert self.vmin <= uval.vmin and uval.vmax <= self.vmax, f"bind {val} not in range [{self.vmin}, {self.vmax}]"
assert uval.divides(self.arg.multiple_of) is not None, f"bind {val} not divisible by {self.arg.multiple_of}"
return self.replace(arg=replace(self.arg, val=uval.val))
def unbound(self) -> Variable:
assert self.is_variable, f"op is {self.op}, need Variable"
# strip the tag too: tags are kernel-graph processing state, the unbound Variable is the canonical node
return self.replace(arg=replace(self.arg, val=None), tag=None)
def unbind(self) -> tuple[Variable, int]:
assert self.is_bound_var, f"can't unbind {self}"
return self.unbound(), self.arg.val
def unbind_all(self) -> tuple[UOp, dict[Variable, int]]:
bound = {x: x.unbound() for x in self.backward_slice_with_self if x.is_bound_var}
return self.substitute(bound, walk=True), {v: cast(int, x.arg.val) for x, v in bound.items()}
def variables(self) -> list[Variable]:
ret = set()
for x in self.backward_slice_with_self:
if x.op is Ops.PARAM and x.addrspace is AddrSpace.ALU:
ret.add(x.unbound() if x.is_variable else x)
elif x.op is Ops.RANGE and x.axis_type is AxisType.DEVICE:
ret.add(UOp.variable("_device_num", 0, x.vmax, dtype=x.dtype))
return sorted(ret, key=lambda v: (v.arg.name or "", v.arg.slot))
# *** uop symbolic stuff ***
def const_factor(self) -> int:
"""largest known int that divides self"""
# TODO: for negatives it's not the largest
if self.op is Ops.CONST: return self.val
if self.op is Ops.STACK: return math.gcd(*[x.const_factor() for x in self.src])
if self.op is Ops.ADD: return math.gcd(self.src[0].const_factor(), self.src[1].const_factor())
if self.op is Ops.MUL: return self.src[0].val if self.src[0].op is Ops.CONST else self.src[1].val if self.src[1].op is Ops.CONST else 1
if self.op in GroupOp.Defines and self.arg.multiple_of is not None: return self.arg.multiple_of
return 1
def divides(self, v:int) -> UOp|None:
if v==1: return self
if self.op is Ops.CONST: return self.const_like(self.val//v) if self.val%v == 0 else None
if self.op is Ops.STACK:
srcs = tuple(s.divides(v) for s in self.src)
return None if any(s is None for s in srcs) else UOp(Ops.STACK, src=cast(tuple[UOp, ...], srcs))
if self.op is Ops.ADD: return d0+d1 if (d0:=self.src[0].divides(v)) is not None and (d1:=self.src[1].divides(v)) is not None else None
if self.op is Ops.MUL:
if (d0:=self.src[0].divides(v)) is not None: return d0 * self.src[1]
if (d1:=self.src[1].divides(v)) is not None: return self.src[0] * d1
if self.op in GroupOp.Defines and self.arg.multiple_of is not None:
return self // v if self.arg.multiple_of%v == 0 else None
return None # generic None if we aren't sure
def pop_const(self, op=Ops.ADD) -> tuple[UOp, PyConst]: # NOTE: assume Invalid ALU is resolved
return (self.src[0], self.src[1].val) if self.op is op and self.src[1].op is Ops.CONST else (self, identity_element(op, self.dtype))
@staticmethod
def gcd(*uops: UOp) -> UOp:
# Strip explicit coefficients, not multiple_of, so symbolic factors stay recognizable by divide_exact.
terms, factors = zip(*[u.pop_const(Ops.MUL) for u in uops])
count = functools.reduce(operator.and_, [collections.Counter(term.split_uop(Ops.MUL)) for term in terms])
if not count: factors = tuple(u.const_factor() for u in uops)
return math.prod(count.elements(), start=uops[0].const_like(math.gcd(*factors)))
def divide_exact(self, v:UOp) -> UOp|None:
if self is v: return self.const_like(1)
if v.op is Ops.CONST: return self.divides(v.val)
if self.op is Ops.ADD: return None if (s0:=self.src[0].divide_exact(v)) is None or (s1:=self.src[1].divide_exact(v)) is None else s0+s1
if self.op is Ops.MUL:
(fac, const), (div_fac, div_const) = self.pop_const(Ops.MUL), v.pop_const(Ops.MUL)
new_count = collections.Counter(fac.split_uop(Ops.MUL))
new_count.subtract(div_fac.split_uop(Ops.MUL))
if const%div_const==0 and all(v>=0 for v in new_count.values()): return math.prod(new_count.elements(), start=self.const_like(const//div_const))
return None # generic None if we aren't sure
@property
def vmin(self) -> PyConst: return self._min_max[0]
@property
def vmax(self) -> PyConst: return self._min_max[1]
@functools.cached_property
def _min_max(self) -> tuple[PyConst, PyConst]:
if self.op in GroupOp.Binary and not dtypes.is_float(self.dtype):
(s0_vmin, s0_vmax), (s1_vmin, s1_vmax) = self.src[0]._min_max, self.src[1]._min_max
if self.op is Ops.ADD: return s0_vmin+s1_vmin, s0_vmax+s1_vmax
if self.op is Ops.SUB: return s0_vmin-s1_vmax, s0_vmax-s1_vmin
if self.op is Ops.AND and dtypes.is_int(self.dtype) and s1_vmin == s1_vmax >= 0:
return 0, s1_vmax if s0_vmin < 0 else min(s0_vmax, int(s1_vmax) & ((1 << int(s0_vmax).bit_length()) - 1))
if self.op is Ops.MUL: return min(vals:=(s0_vmin*s1_vmin, s0_vmin*s1_vmax, s0_vmax*s1_vmin, s0_vmax*s1_vmax)), max(vals)
# SHL/SHR on consts only
if self.op is Ops.SHL and s1_vmin == s1_vmax and all_int(t:=(s0_vmin, s0_vmax, s1_vmin)): return t[0] << t[2], t[1] << t[2]
if self.op is Ops.SHR and s1_vmin == s1_vmax and all_int(t:=(s0_vmin, s0_vmax, s1_vmin)): return t[0] >> t[2], t[1] >> t[2]
if self.op is Ops.CMOD:
if (c:=s1_vmin) == s1_vmax > 0:
return (0 if s0_vmin > 0 else s0_vmin if 0 >= s0_vmin > -c else -(s1_vmax-1), 0 if s0_vmax < 0 else s0_vmax if 0 <= s0_vmax < c else c-1)
if s1_vmin > 0: return (0, s1_vmax-1) if s0_vmin >= 0 else (-(s1_vmax-1), 0) if s0_vmax <= 0 else (-(s1_vmax-1), s1_vmax-1)
if s1_vmax < 0: return (0, -s1_vmin-1) if s0_vmin >= 0 else (-(-s1_vmin-1), 0) if s0_vmax <= 0 else (-(-s1_vmin-1), -s1_vmin-1)
if self.op is Ops.CDIV:
assert isinstance(s0_vmin, int) and isinstance(s0_vmax, int) and isinstance(s1_vmin, int) and isinstance(s1_vmax, int)
if s1_vmin*s1_vmax>0:
return min(vals:=(cdiv(s0_vmin, s1_vmin), cdiv(s0_vmin, s1_vmax), cdiv(s0_vmax, s1_vmin), cdiv(s0_vmax, s1_vmax))), max(vals)
if self.op is Ops.FLOORDIV:
assert isinstance(s0_vmin, int) and isinstance(s0_vmax, int) and isinstance(s1_vmin, int) and isinstance(s1_vmax, int)
if s0_vmin > s0_vmax: return 0, 0 # numerator range is empty (e.g. RANGE with end=0)
if s1_vmin*s1_vmax>0: return min(vals:=(s0_vmin//s1_vmin, s0_vmin//s1_vmax, s0_vmax//s1_vmin, s0_vmax//s1_vmax)), max(vals)
if self.op is Ops.FLOORMOD:
assert isinstance(s0_vmin, int) and isinstance(s0_vmax, int) and isinstance(s1_vmin, int) and isinstance(s1_vmax, int)
if s0_vmin > s0_vmax: return 0, 0 # numerator range is empty (e.g. RANGE with end=0)
if (c:=s1_vmin) == s1_vmax > 0: return (s0_vmin%c, s0_vmax%c) if s0_vmin//c == s0_vmax//c else (0, c-1)
if (c:=s1_vmin) == s1_vmax < 0: return (s0_vmin%c, s0_vmax%c) if s0_vmin//c == s0_vmax//c else (c+1, 0)
if s1_vmin > 0: return (0, s1_vmax-1)
if s1_vmax < 0: return (s1_vmin+1, 0)
if self.op is Ops.XOR and s1_vmin == s1_vmax == -1 and isinstance(s0_vmin, int) and isinstance(s0_vmax, int):
return ~int(s0_vmax), ~int(s0_vmin)
if self.op is Ops.XOR and dtypes.is_int(self.dtype) and s0_vmin >= 0 and s1_vmin >= 0:
return 0, (1 << int(max(s0_vmax, s1_vmax)).bit_length()) - 1
if self.op is Ops.MAX: return max(s0_vmin, s1_vmin), max(s0_vmax, s1_vmax)
if self.op is Ops.CMPLT: return (s0_vmax<s1_vmin, s0_vmin<s1_vmax)
if self.op is Ops.CMPNE: return ((s0_vmax < s1_vmin) or (s1_vmax < s0_vmin), not (s0_vmin == s0_vmax == s1_vmin == s1_vmax))
if self.op is Ops.OR and self.dtype == dtypes.bool: return s0_vmin or s1_vmin, s0_vmax or s1_vmax
if self.op is Ops.AND and self.dtype == dtypes.bool: return s0_vmin and s1_vmin, s0_vmax and s1_vmax
if self.op is Ops.WHERE: return min(self.src[1].vmin, self.src[2].vmin), max(self.src[1].vmax, self.src[2].vmax)
# NOTE: returned UOp is assumed to be CONST
if self.op in GroupOp.Defines and self.arg.vmin_vmax is not None: return self.arg.vmin_vmax
if self.op in (Ops.RANGE, Ops.SPECIAL) and self.dtype is not dtypes.void: return 0, (self.src[0]-1).vmax
if self.op is Ops.STACK: return min(x.vmin for x in self.src), max(x.vmax for x in self.src)
# a load from a constant table is one of its values
if self.op is Ops.LOAD and (b:=self.src[0].buf_uop).op is Ops.BINARY:
return min(e:=memoryview(b.arg).cast(unwrap(self.dtype.fmt)).tolist()), max(e)
# a NAN is outside every interval
if self.op is Ops.CONST and self.val is not Invalid and not (isinstance(self.val, float) and math.isnan(self.val)): return self.val, self.val
if self.op is Ops.PAD: return min(self.src[0].vmin, 0), max(self.src[0].vmax, 0) # PAD adds zeros
if self.op in GroupOp.Movement|{Ops.INDEX, Ops.STAGE, Ops.AFTER, Ops.DETACH, Ops.COPY, Ops.CONTIGUOUS_BACKWARD}: return self.src[0]._min_max
if self.op is Ops.CAST:
# rounding is monotone (truncation toward zero into an int, to-nearest onto the value grid into a float)
smin, smax = self.src[0]._min_max
trunc = truncate.get(self.dtype) if dtypes.is_float(self.dtype) else math.trunc if dtypes.is_int(self.dtype) else None
if trunc is not None: smin, smax = (trunc(v) if math.isfinite(v) else v for v in (smin, smax))
if dtypes.is_unsigned(self.dtype) and 0 <= smin and smax <= self.dtype.max: return smin, smax
# a signed or float destination holds the part of the source that overlaps it: overflow is undefined, a nan bound overlaps nothing
if self.dtype in dtypes.floats+dtypes.sints+dtypes.weaks and smin <= self.dtype.max and self.dtype.min <= smax:
return max(self.dtype.min, smin), min(smax, self.dtype.max)
return self.dtype.min, self.dtype.max
@functools.cached_property
def _sym_fxn(self):
from tinygrad.uop.render import _render_with_splits, renderer_infer
sself = self.simplify()
varnames = tuple(dedup(x.expr for x in sself.toposort() if x.op is Ops.PARAM and x.arg.addrspace == AddrSpace.ALU))
# TODO: sanitize varnames, or don't use naked eval while staying fast
ret = _render_with_splits(list(sself.toposort()), renderer_infer, {sself})
lines = [f" {k}={v}" for k,v in ret.items() if k != "ast"] + [f" return {ret['ast']}"]
ns: dict[str, Any] = {"max": max, "cdiv": cdiv, "cmod": cmod, "floordiv": floordiv, "floormod": floormod, "bitcast": bitcast, "dtypes": dtypes}
exec(f"def _f({','.join(varnames)}):\n"+'\n'.join(lines), ns) # pylint: disable=exec-used
return ns["_f"], varnames
def sym_infer(self, var_vals:dict[str, int]):
fxn, varnames = self._sym_fxn
return fxn(**{k:v for k,v in var_vals.items() if k in varnames})
def render(self, simplify=True, pm:PatternMatcher|None=None) -> str:
ctx: dict[UOp, str] = {}
from tinygrad.uop.render import renderer
pm = renderer if pm is None else pm
for u in (s:=self.simplify() if simplify else self).toposort():
ctx[u] = cast(str, pm.rewrite(u, ctx=ctx))
return ctx[s]
def render_uir(self) -> str:
from tinygrad.uop.render import render_uir
return render_uir(self)
# *** uop high level syntactic sugar ***
@staticmethod
def alloc(shape:tuple[sint, ...], dtype:DType, slot:int|None=None, addrspace=AddrSpace.GLOBAL, device=None, axis:int|None=None,
spec:BufferSpec|None=None, bind_on_realize:bool=False):
ret = UOp(Ops.ALLOC, src=(shape_to_shape_arg((prod(to_max_shape(shape)),)),)+UOp.device_range_src(device),
arg=ParamArg(next(UOp.unique_num) if slot is None else slot, strong_dtype(dtype),
addrspace=addrspace, device=device, spec=spec, bind_on_realize=bind_on_realize))
return ret.reshape(()) if not shape else ret.view_as(shape, axis)
def alloc_like(self, slot:int|None=None, addrspace=AddrSpace.GLOBAL): return UOp.alloc(self.max_shard_shape, self.dtype, slot, addrspace)
@staticmethod
def placeholder(shape:tuple[int, ...], dtype:DType, slot:int|None=None, addrspace=AddrSpace.GLOBAL, device=None):
dtype = strong_dtype(dtype) # storage is never weak: a placeholder commits the width of what's put in it
if slot is None: slot = next(UOp.unique_num)
if addrspace is AddrSpace.GLOBAL:
ret = UOp.param(slot, dtype, prod(shape), device=device)
else:
assert addrspace in (AddrSpace.LOCAL, AddrSpace.REG)
assert device is None, "LOCAL and REG placeholders cannot have a device"
ret = UOp._buffer(prod(shape), dtype, slot, addrspace=addrspace)
if len(shape) > 1: ret = ret.reshape(shape)
return ret
def placeholder_like(self, slot:int, addrspace=AddrSpace.GLOBAL):
assert all_int(self.shape), "no placeholder-like on symbolic shape"
return UOp.placeholder(self.max_shard_shape, self.dtype, slot, addrspace)
# set is store+end+after
def set(self:UOp, val:UOp|ConstType, end:UOp|tuple[UOp, ...]|list[UOp]=()) -> UOp:
return self.src[0].after(self.store(val).end(*argfix(end)))
# TODO: this should replace placeholder
@staticmethod
def param(slot:int, dtype:DType, shape:tuple[sint, ...]|sint|None=None, device=None, vmin_vmax:tuple[PyConst, PyConst]|None=None,
multiple_of:int|None=None, name=None, addrspace=AddrSpace.GLOBAL, volatile:bool=False):
"""create a PARAM: a single sint or 1-d shape gives a flat param of that size, a None shape gives a scalar param.
src[0] stores the concrete max size (never symbolic): a multi-dim shape is a RESHAPE on top of the flat param,
a symbolic shape is a max-size param shrunk to the real shape"""
if dtype in dtypes.weaks: raise RuntimeError(f"cannot create param for weak dtype {dtype}")
shape = (shape,) if isinstance(shape, (int, UOp)) else shape or ()
ret = UOp(Ops.PARAM, src=(shape_to_shape_arg((prod(to_max_shape(shape)),) if shape else ()),),
arg=ParamArg(slot, dtype, vmin_vmax, multiple_of, name, addrspace, device, volatile))
return ret.view_as(shape)
def param_like(self, slot:int, name:str|None=None):
# Scalar arguments bind by slot; names and values stay at the call site, not in schedule cache keys.
if self.op is Ops.PARAM and self.addrspace is AddrSpace.ALU:
return self.replace(arg=replace(self.arg, slot=slot, name=name, val=None))
# multi-device values become a per-shard sized param wrapped in UNSHARD: the sharding lives in the graph, not the arg
if self.axis is not None and isinstance(self.device, tuple):
return UOp.param(slot, self.dtype, prod(to_max_shape(self.shard_shape)), self.device, name=name).view_as(self.shard_shape, self.axis)
return UOp.param(slot, self.dtype, self._shape, self.device, name=name)
def view_as(self:UOp, shape:tuple[sint, ...], axis:int|None=None) -> UOp:
"""view flat storage as the given (possibly symbolic) shape, optionally sharded on axis, the UNSHARD gives back the multiplied shape"""
max_shape = to_max_shape(shape)
ret = self.reshape(max_shape) if shape else self
if tuple(max_shape) != tuple(shape): ret = ret.shrink_to(shape)
return ret if axis is None else ret.unshard(axis)
@staticmethod
def custom_function(name:str, *src:UOp, dtype:DType=dtypes.void) -> UOp:
return UOp(Ops.CUSTOM_FUNCTION, src=src, arg=CustomFunction(name, dtype))
def call(self, *srcs:UOp, grad_fxn:Callable|None=None,
name:str|None=None, precompile:bool=False, precompile_backward:bool=False, aux:Any=None) -> UOp:
"""call a body with the given args: a plain CallInfo CALL. all inputs must be ready (buffers/params), this never
creates ALLOCs: use call_with_outputs for calls that produce values"""
assert self.op in OPAQUE_CALL_BODIES, f"cannot call a {self.op} body, use call_with_outputs for value-producing bodies"
# calls are launched per device, so an open DEVICE range is allowed to cross the call boundary
assert all(r.axis_type is AxisType.DEVICE for r in self.ranges), \
f"ranges {self.ranges} are leaking out of the call in {self.render_uir()}"
# an external C call is a CALL on a CUSTOM_FUNCTION body stating the (possibly void) return dtype, the callee
# (a function pointer) in source, rendered as an indirect call
return UOp(Ops.CALL, src=(self,)+srcs, arg=CallInfo(grad_fxn, name, precompile, precompile_backward, aux))
@staticmethod
def call_with_outputs(values:tuple[UOp, ...], *srcs:UOp, grad_fxn:Callable|None=None,
name:str|None=None, precompile:bool=False, precompile_backward:bool=False, aux:Any=None,
output_pos:tuple[int, ...]|None=None) -> tuple[UOp, ...]:
"""call a body producing the given values, returning the outputs. the body stores into output PARAMs, and the
outputs are ALLOCs passed as extra inputs to the call (you AFTER on them like normal buffers).
the buffers are bound to the output PARAMs positionally wherever the call is resolved, just like the args.
output_pos gives the position of each output in the arg list (default: a block after the inputs), the inputs take
the remaining positions in order; when it's given, input params must already be slotted at their final positions.
output_pos must be strictly ascending: the body's stores and the call args pair positionally by values order"""
# the device defaults to the first device in the values or args, like srcs-based device resolution
default_dev = next((x.device for x in itertools.chain(values, srcs) if x.device is not None), None)
pos = tuple(range(len(srcs), len(srcs)+len(values))) if output_pos is None else output_pos
assert len(pos) == len(values) and len(set(pos)) == len(pos), "output_pos must be one distinct position per output"
assert all(a < b for a, b in zip(pos, pos[1:])), f"output_pos {output_pos} must be strictly ascending"
assert all(0 <= p < len(srcs)+len(values) for p in pos), f"output_pos {output_pos} must be within the arg list"
# the inputs take the slots not in pos, in order: symbolic output shapes resolve against the final argument slots
param_map: list[UOp|None] = [None] * (len(srcs) + len(values))
it = iter(srcs)
for i in range(len(param_map)):
if i not in pos: param_map[i] = next(it)
def mint(o:UOp, p:int) -> tuple[UOp, UOp]:
# Declare output storage in the callee's shape, then bind its shape in the caller's scope.
dev = o.device if o.device is not None else default_dev
axis = o.axis if isinstance(o.device, tuple) else None
buf = UOp.alloc(o.shard_shape, o.dtype, device=dev, axis=axis)
shp = tuple(graph_rewrite(s, _pm_resolve_params, param_map, walk=True) if isinstance(s, UOp) else s for s in o.shard_shape)
return UOp.alloc(shp, o.dtype, slot=buf.buf_uop.arg.slot, device=dev, axis=axis), buf.param_like(p)
outputs = tuple(mint(o, p) for o, p in zip(values, pos))
rets = tuple(r for r, _ in outputs)
body = UOp.sink(*[p.store(v) for v, (_, p) in zip(values, outputs)])
args: list[UOp|None] = [None] * (len(srcs) + len(values))
for p, r in zip(pos, rets): args[p] = r
it = iter(x.contiguous() if precompile else x for x in srcs)
call = body.call(*[r if r is not None else next(it) for r in args], grad_fxn=grad_fxn, name=name, precompile=precompile,
precompile_backward=precompile_backward, aux=aux)
return tuple(r.after(call) for r in rets)
# one-line convenience for the single-output case: self is the value
def call_with_output(self, *srcs:UOp, **kwargs) -> UOp: return UOp.call_with_outputs((self,), *srcs, **kwargs)[0]
def custom_kernel(*srcs:UOp, fxn:Callable, grad_fxn:Callable|None=None) -> list[UOp]:
placeholders = [UOp.placeholder_like(s, slot=i) for i,s in enumerate(srcs)]
kernel = fxn(*placeholders).call(*srcs, grad_fxn=grad_fxn)
return [s.after(kernel) for s in srcs]
def to_elf(self) -> TinyELF:
assert self.op is Ops.PROGRAM and isinstance(self.arg, ProgramInfo), "to_elf should only be called on a PROGRAM ast"
params = tuple(u for u in self.src[1].src if u.op is Ops.PARAM and u.addrspace != AddrSpace.ALU)
# sig slots are compact: buffers in globals order (runtimes launch buffers in that order), then vars. raw call-arg
# positions skip buffers for kernels using a sparse subset of the call's buffers (CL binds bufs[slot])
gmap = {s:j for j, s in enumerate(self.arg.globals)}
sig = tuple((u.arg.name, gmap[u.arg.slot], u.dtype, u._shape) for u in params) + \
tuple((v.arg.name, len(self.arg.globals)+j, v.dtype, v._shape) for j, v in enumerate(self.arg.vars))
return TinyELF(self.src[3].arg, self.src[0].arg.function_name, self.arg.target, sig, self.key)
@property
def src_without_body(self) -> tuple[UOp, ...]: return self.src[1:] if self.op is Ops.CALL else self.src
def uopfunc(fn:Callable[..., UOp]) -> Callable[..., UOp]: # sugar for body.call(*args): uop args become params
def param(i:int, n:str, a:UOp) -> UOp:
if a.addrspace in (None, AddrSpace.ALU): return UOp.param(i, a.commit_dtype(dtypes.int), name=n, addrspace=AddrSpace.ALU)
return UOp.param(i, a.dtype, 1 if a.op is Ops.INDEX else a.max_numel(), a.device, name=n, addrspace=a.addrspace)
def outlined(*args, **kwargs) -> UOp:
bound = inspect.signature(fn).bind(*args, **kwargs).arguments
ins = {n: a for n, a in bound.items() if isinstance(a, UOp)}
f = graph_rewrite(fn(**(bound|{n: param(i, n, a) for i, (n, a) in enumerate(ins.items())})), pm_renumber_slots, ctx=itertools.count(), walk=True)
with Context(TRACK_MATCH_STATS=0): return f.call(*ins.values(), name=fn.__name__)
return functools.wraps(fn)(outlined)
@dataclass(frozen=True)
class KernelInfo:
name: str = "test" # name of the kernel
applied_opts: tuple = tuple()
opts_to_apply: tuple|None = None
estimates: Estimates|None = None
beam: int = 0
@property
def function_name(self): return to_function_name(self.name)
@dataclass(frozen=True)
class ProgramInfo:
global_size: tuple[int|float, ...] = (1, 1, 1)
local_size: tuple[int, ...] = (1, 1, 1)
vars: tuple[UOp, ...] = ()
globals: tuple[int, ...] = ()
outs: tuple[int, ...] = ()
ins: tuple[int, ...] = ()
target: Target = Target()
def launch_dims(self, var_vals:dict[str, int]) -> tuple[tuple[int, ...], tuple[int, ...]]:
global_size = tuple([sym_infer(sz, var_vals) for sz in self.global_size]) # type: ignore[arg-type]
local_size = tuple([sym_infer(sz, var_vals) for sz in self.local_size])
return global_size, local_size
def vals(self, var_vals:dict[str, int]) -> tuple[int, ...]:
try: return tuple(var_vals[k.expr] for k in self.vars)
except KeyError as e: raise RuntimeError(f"unbound Variable {e}") from None
@staticmethod
def from_sink(sink:UOp, target:Target=Target()) -> ProgramInfo:
_vars: list[UOp] = []
_globals: list[int] = []
outs: list[int] = []
ins: list[int] = []
global_size: list[int] = [1, 1, 1]
local_size: list[int] = [1, 1, 1]
for u in sink.toposort():
if u.op is Ops.PARAM and u.addrspace == AddrSpace.ALU: _vars.append(u)
if u.op is Ops.PARAM and u.addrspace != AddrSpace.ALU: _globals.append(u.arg.slot)
if u.op in (Ops.STORE, Ops.LOAD):
if (idx:=u.src[0]).op in (Ops.INDEX, Ops.SHRINK) or (u.src[0].op is Ops.CAST and (idx:=u.src[0].src[0]).op is Ops.INDEX):
if (buf:=idx.src[0].buf_uop).op is Ops.PARAM: (outs if u.op is Ops.STORE else ins).append(buf.arg.slot)
if u.op is Ops.SPECIAL: (local_size if u.arg[0] == 'l' else global_size)[int(u.arg[-1])] = cast(int, u.src[0].ssimplify())
if not outs and not ins: outs = ins = _globals # if neither is inferred, default to all buffers
return ProgramInfo(tuple(global_size), tuple(local_size),
tuple(sorted(_vars, key=lambda v: v.arg.slot)), tuple(sorted(dedup(_globals))), tuple(sorted(dedup(outs))),
tuple(sorted(dedup(ins))), target)
# the body of a CALL is always one of these: programs (SINK/PROGRAM/LINEAR), bulk stores, and function references
OPAQUE_CALL_BODIES = {Ops.SINK, Ops.PROGRAM, Ops.LINEAR, Ops.STORE, Ops.CUSTOM_FUNCTION}
# the arg of CUSTOM_FUNCTION: an external function symbol and the dtype of the value it returns
@dataclass(frozen=True)
class CustomFunction:
name: str
dtype: DType = dtypes.void
@dataclass(frozen=True)
class CallInfo:
grad_fxn: Callable|None = None
name: str|None = None
precompile: bool = False
precompile_backward: bool = False
aux: Any = None
# grad_fxn can't be pickled
def __reduce__(self): return (CallInfo, (None, self.name, self.precompile, self.precompile_backward, self.aux))
def __repr__(self):
gf = id(self.grad_fxn) if self.grad_fxn else None
return f"CallInfo({gf}, {repr(self.name)}, {self.precompile}, {self.precompile_backward})"
# ******** ops in python ********
def safe_exp2(x):
try: return 2 ** x
except OverflowError: return math.inf
def safe_pow(x, y):
if isinstance(x, int) and isinstance(y, int) and y < 0: return x**(y%2) if abs(x) == 1 else 0
try: return math.nan if isinstance(p:=pow(x, y), complex) else p
except ZeroDivisionError: return math.inf
except ValueError: return math.inf if x > 0 else -math.inf
python_alu: dict[Ops, Callable] = {
Ops.LOG2: lambda x: math.log2(x) if x > 0 else -math.inf if x == 0 else math.nan, Ops.EXP2: safe_exp2,
Ops.SQRT: lambda x: math.sqrt(x) if x >= 0 else math.nan, Ops.RECIPROCAL: lambda x: 1/x if x != 0 else math.copysign(math.inf, x),
Ops.SIN: lambda x: math.sin(x) if not math.isinf(x) else math.nan, Ops.POW: safe_pow,
Ops.TRUNC: lambda x: math.trunc(x) if math.isfinite(x) else x,
Ops.NEG: operator.neg, Ops.ADD: operator.add, Ops.SUB: operator.sub, Ops.MUL: operator.mul, Ops.CMPNE: operator.ne, Ops.CMPLT: operator.lt,
Ops.XOR: operator.xor, Ops.OR: operator.or_, Ops.AND: operator.and_, Ops.SHR: operator.rshift, Ops.SHL: operator.lshift, Ops.MAX: max,
Ops.CMOD: cmod, Ops.CDIV: cdiv, Ops.FLOORDIV: floordiv, Ops.FLOORMOD: floormod,
Ops.MULACC: lambda x,y,z: (x*y)+z, Ops.WHERE: lambda x,y,z: y if x else z, Ops.CMPEQ: operator.eq}
def exec_alu(op:Ops, dtype:DType, operands, truncate_output=True):
if any(isinstance(x, tuple) for x in operands):
count = max(len(x) for x in operands if isinstance(x, tuple))
return tuple([exec_alu(op, dtype, [x[i] if isinstance(x, tuple) else x for x in operands]) for i in range(count)])
if op in GroupOp.Binary and Invalid in operands: return Invalid
alu = python_alu[op](*operands)
if truncate_output and (truncate_fxn:=truncate.get(dtype)) is not None: return truncate_fxn(alu)
return alu
# ***** pattern matcher *****
def get_location() -> tuple[str, int]:
frm = sys._getframe(1)
# skip over ops.py and anything in mixin
while frm.f_back is not None and not frm.f_back.f_code.co_filename.startswith("<frozen"):
fn = frm.f_code.co_filename.replace("\\", "/")
if not (fn.endswith("/ops.py") or "/mixin/" in fn): break
frm = frm.f_back
return frm.f_code.co_filename, frm.f_lineno
class UPat(RandMixin):
__slots__ = ("op", "match_dtype", "match_tag", "arg", "name", "src", "is_any")
def __init__(self, op:Ops|tuple[Ops, ...]|set[Ops]|None=None, dtype:DType|tuple[DType, ...]|set[DType]|None=None,
src:tuple[UPat, ...]|list[UPat]|UPat|None=None, arg:Any=None,
name:str|None=None, allow_any_len:bool=False, custom_early_reject:set[Ops]|None=None, location=None, is_any:bool=False,
tag:Any=None):
assert op is None or isinstance(op, (Ops, tuple, set)), f"op must be Ops or tuple of Ops, not {op!r}"
self.op: tuple[Ops, ...]|None = (op,) if isinstance(op, Ops) else (tuple(op) if isinstance(op, set) else op)
self.match_dtype: tuple[DType, ...]|None = (dtype,) if isinstance(dtype, DType) else (tuple(dtype) if isinstance(dtype, set) else dtype)
self.match_tag: tuple[Any, ...]|None = (tag,) if isinstance(tag, str) else (tuple(tag) if isinstance(tag, set) else tag)
self.arg, self.name, self._in_src, self.custom_early_reject = arg, name, src, custom_early_reject
self.src: Any = None
self.is_any = is_any
assert self.name != "ctx", "UPat can't be named ctx"
assert dtype is None or isinstance(dtype, DType) or all(isinstance(x, DType) for x in dtype), f"invalid dtype {dtype}"
# try all permutations if it's a list
if isinstance(src, list): self.src = list(itertools.permutations(src)) if not all_same(src) else [tuple(src)]
# only one if it's a tuple
elif isinstance(src, tuple): self.src = [src]
# repeat if it's a UPat
elif isinstance(src, UPat): self.src = [itertools.repeat(src)]
self.strict_length = not (allow_any_len or isinstance(src, UPat) or src is None)
self.required_len: int = 0 if isinstance(src, UPat) or src is None else len(src)
self.location = location or get_location()
if custom_early_reject is not None: self.early_reject = custom_early_reject
else:
upat_match = [src] if isinstance(src, UPat) else ([] if src is None else self.src[0])
self.early_reject = {pp.op[0] for pp in upat_match if pp.op is not None and len(pp.op) == 1}
@property
def dtype(self) -> DType: return self.match_dtype[0] if self.match_dtype is not None else dtypes.void
def __reduce__(self):
return UPat, (self.op, self.match_dtype, self._in_src, self.arg, self.name, not self.strict_length, self.custom_early_reject, self.location,
self.is_any, self.match_tag)
def named(self, name:str):
return UPat(self.op, self.match_dtype, self._in_src, self.arg, name, not self.strict_length, self.custom_early_reject, tag=self.match_tag)
@staticmethod
def any(*src): return UPat(src=src, is_any=True)
def or_casted(self, name:str|None=None): return UPat.any(self if name is None else self.named(name), UPat(Ops.CAST, name=name, src=(self,)))
def or_bitcasted(self, name:str|None=None): return UPat.any(self if name is None else self.named(name), UPat(Ops.BITCAST, name=name, src=(self,)))
def or_after(self, name:str|None=None):
return UPat.any(self if name is None else self.named(name), UPat(Ops.AFTER, name=name, src=(self,), allow_any_len=True))
@staticmethod
@functools.cache
def var(name:str|None=None, dtype:DType|tuple[DType, ...]|None=None): return UPat(dtype=dtype, name=name)
@staticmethod
@functools.cache
def cvar(name:str|None=None, dtype:DType|tuple[DType, ...]|None=None, arg=None): return UPat(Ops.CONST, dtype, name=name, arg=arg)
@staticmethod
def const(b:ConstType, dtype:DType|tuple[DType, ...]|None=None): return UPat(Ops.CONST, dtype=dtype, arg=b)
@staticmethod
def custom_function(fn:str, **kwargs): return UPat(Ops.CUSTOM_FUNCTION, arg=CustomFunction(fn), **kwargs)
# lil helper
def f(self, op, **kwargs): return UPat(op, src=(self,), **kwargs)
# copied from UOp
def sink(*srcs:UPat|None, **kwargs): # pylint: disable=no-self-argument
return UPat(Ops.SINK, src=tuple([x for x in srcs if x is not None]), **kwargs)
def index(self, *srcs:UPat|None, **kwargs):
return UPat(Ops.INDEX, src=(self,)+tuple(x for x in srcs if x is not None), **kwargs)
def cast(self, dtype=None, **kwargs):
if dtype is not None and self.match_dtype == (dtype,): return self
return UPat(Ops.CAST, dtype, (self,), **kwargs)
def bitcast(self, dtype=None): return UPat(Ops.BITCAST, dtype, (self,))
def load(self, *src:UPat, **kwargs): return UPat(Ops.LOAD, src=(self,)+src, **kwargs)
def store(self, *src:UPat, **kwargs): return UPat(Ops.STORE, src=(self,)+src, **kwargs)
def reduce(self, *src:UPat, **kwargs):
arg = kwargs.pop('arg', None)
if isinstance(arg, Ops): arg = (arg, 0)
return UPat(Ops.REDUCE, self.match_dtype, src=(self,)+src, arg=arg, **kwargs)
def broadcast(self, **kwargs): return UPat(Ops.STACK, self.match_dtype, src=self, **kwargs)
def after(self, *src:UPat, **kwargs): return UPat(Ops.AFTER, self.match_dtype, (self,)+src, **kwargs)
def end(self, *src:UPat, **kwargs): return UPat(Ops.END, src=(self,)+src, **kwargs)
def backedge(self, loop:UPat, cond:UPat, **kwargs): return UPat(Ops.BACKEDGE, src=(self, loop, cond), **kwargs)
def _broadcasted(self, y, reverse=False) -> tuple[UPat, UPat]:
y = self.ufix(y)
return (y, self) if reverse else (self, y)
def ufix(self, x): return UPat.cvar(arg=x) if not isinstance(x, UPat) else x
def __floordiv__(self, x): return self._binop(Ops.FLOORDIV, x, False)
def __rfloordiv__(self, x): return self._binop(Ops.FLOORDIV, x, True)
def mod(self, x, reverse=False): return self._binop(Ops.FLOORMOD, x, reverse)
def alu(self, op:Ops, *src:UPat):
asrc = (self,)+src
return UPat(op, dtypes.bool if op in {Ops.CMPLT, Ops.CMPNE} else asrc[-1].match_dtype, list(asrc) if op in GroupOp.Commutative else asrc)
def match(self:UPat, uop:UOp, store:dict[str, UOp]) -> list[dict[str, UOp]]:
if self.is_any: return flatten([x.match(uop, store.copy()) for x in self.src[0]])
if (self.op is not None and uop.op not in self.op) or \
(self.name is not None and store.setdefault(self.name, uop) is not uop) or \
(self.match_dtype is not None and uop.dtype not in self.match_dtype) or \
(self.arg is not None and self.arg != uop.arg) or \
(self.match_tag is not None and uop.tag not in self.match_tag) or \
(len(uop.src) < self.required_len) or \
(self.strict_length and len(uop.src) != self.required_len): return []
if self.src is None: return [store]
res: list[dict[str, UOp]] = []
for vp in self.src:
stores, new_stores = [store.copy()], []
for uu, vv in zip(uop.src, vp):
for s in stores: new_stores.extend(vv.match(uu, s))
stores, new_stores = new_stores, []
res.extend(stores)
return res
def deconstruct_function(fxn:Callable) -> tuple:
# globals can be referenced from arbitrarily nested code objects (comprehensions/lambdas, pre PEP 709)
def names(co:types.CodeType) -> set: return set(co.co_names).union(*(names(c) for c in co.co_consts if isinstance(c, types.CodeType)))
new_globals = {k:v for k,v in fxn.__globals__.items() if k in names(fxn.__code__)}
# NOTE: optional round trip through pickle!
assert fxn.__closure__ is None, "closures are not supported in pattern matchers"
ret = fxn.__code__, new_globals, fxn.__name__, fxn.__defaults__
return pickle.loads(pickle.dumps(ret)) if getenv("TEST_PICKLE") else ret
@functools.cache
def upat_interpret(p:UPat, fxn:Callable) -> Callable:
real_fxn = types.FunctionType(*deconstruct_function(fxn))
if 'ctx' in inspect.signature(real_fxn).parameters:
def universal_match(uop, ctx):
for match in p.match(uop, {}):
if (ret:=real_fxn(ctx=ctx, **match)) is not None: return ret # pylint: disable=not-callable
return None
else:
def universal_match(uop, _):
for match in p.match(uop, {}):
if (ret:=real_fxn(**match)) is not None: return ret # pylint: disable=not-callable
return None
return universal_match
def upat_deferred_compile(p:UPat, fxn:Callable, entry:list) -> Callable:
def lazy_compile(uop, ctx):
from tinygrad.uop.upat import upat_compile
entry[1] = upat_compile(p, fxn)
return entry[1](uop, ctx)
return lazy_compile
class PatternMatcher:
def __init__(self, patterns:Sequence[tuple[UPat, Callable|tuple]], compiled=bool(getenv("UPAT_COMPILE", 1))):
# if this comes from a pickle, we reconstruct the lambda functions here
self.patterns:list[tuple[UPat, Callable]] = [(p,types.FunctionType(*fxn) if isinstance(fxn, tuple) else fxn) for p,fxn in patterns]
# NOTE: use of DefaultDict here is very dangerous! all keys will live for the lifetime of the PatternMatcher!
self.pdict: dict[Ops, list[list]] = {}
# uop is required, arg is optional
for p,fxn in self.patterns:
assert p.op is not None
entry: list = [p, None, p.early_reject]
entry[1] = upat_deferred_compile(p, fxn, entry) if compiled else upat_interpret(p, fxn)
for uop in p.op: self.pdict.setdefault(uop, []).append(entry)
def __reduce__(self): return PatternMatcher, ([(x,deconstruct_function(fxn) if fxn.__name__ == "<lambda>" else fxn) for x,fxn in self.patterns],)
@functools.cache # pylint: disable=method-cache-max-size-none
def __add__(self, more:PatternMatcher) -> PatternMatcher: return PatternMatcher(self.patterns+more.patterns)
def rewrite(self, uop:UOp, ctx=None):
if len(pats:=self.pdict.get(uop.op, [])):
if (ler:=uop.__dict__.get('_src_ops')) is None: uop.__dict__['_src_ops'] = ler = {u.op for u in uop.src}
for _,match,early_reject in pats:
if not early_reject.issubset(ler): continue
if (ret:=match(uop, ctx)) is not None and ret is not uop: return ret
return None
# *** tracking pattern matcher ***
TRACK_MATCH_STATS = ContextVar("TRACK_MATCH_STATS", 2 if VIZ else 0)
REWRITE_STACK_LIMIT = ContextVar("REWRITE_STACK_LIMIT", 250000)
match_stats:dict[UPat, list[int|float]] = dict()
# TRACK_MATCH_STATS>=2 or VIZ=1 saves all matches
ucount = itertools.count()
uop_fields:dict[int, tuple] = {}
@dataclass(frozen=True)
class TrackedGraphRewrite:
loc:tuple[str, int] # location that called graph_rewrite
sink:int # the sink input to graph_rewrite
matches:list[tuple[int, int, tuple, float]] # before/after UOp, UPat location and time
name:str # name of the rewrite
depth:int # depth if it's a subrewrite
bottom_up:bool
walk:bool
enter_calls:bool
tracked_keys:list[TracingKey] = []
tracked_ctxs:list[list[TrackedGraphRewrite]] = []
_name_cnt:dict[str, itertools.count] = {}
# find the unittest frame we're capturing in
PROCESS_REPLAY_LOC = ContextVar("PROCESS_REPLAY_LOC", "")
def get_process_replay_loc() -> str:
if PROCESS_REPLAY_LOC.value: return PROCESS_REPLAY_LOC.value
frm = sys._getframe(1)
while (f_back:=frm.f_back) is not None and f_back.f_globals.get("__name__", "").split(".")[0] not in ("unittest", "_pytest"):
frm = f_back
return f"{frm.f_code.co_filename.split('/')[-1]}:{frm.f_lineno} {frm.f_code.co_name}"
if CAPTURE_PROCESS_REPLAY:
replay_capture: list[bytes] = []
import atexit, uuid
@atexit.register
def save_to_diskcache():
uid = uuid.uuid4() # one id per process
for i,v in enumerate(replay_capture): diskcache_put("process_replay", f"{uid}_{i}", v, prepickled=True)
def add_trace_group(kt:TracingKey) -> None:
tracked_keys.append(kt)
tracked_ctxs.append([])
active_group:list[int] = []
active_rewrites:list[TrackedGraphRewrite] = []
def rewrite_group(name:Callable[..., str|TracingKey]|bool=True, replay:bool=False, new_ctx:bool=True):
if not new_ctx: assert not callable(name) and not replay, "name fxn and replay are only supported for new_ctx groups"
def _decorator(func):
def __wrapper(*args, **kwargs):
# without tracking, we just call the function (unless top-level, which always profiles)
if TRACK_MATCH_STATS < 2 and not new_ctx: return func(*args, **kwargs)
fn = key = func.__name__
idx = -1
if TRACK_MATCH_STATS >= 2:
if new_ctx:
add_trace_group(key:=TracingKey(n:=f"{fn} n{next(_name_cnt.setdefault(fn, itertools.count(1)))}", (n,)))
active_group.append(idx:=len(tracked_keys)-1)
else:
rewrite_name = str(kwargs.get("name", None) or fn)
assert args and isinstance(args[0], UOp), f"invalid match tracing inputs for {rewrite_name} with {args}"
loc = ((frm:=sys._getframe(1)).f_code.co_filename, frm.f_lineno)
depth = len(active_rewrites)
if not tracked_ctxs: add_trace_group(TracingKey(f"default {fn}"))
dest_group = active_group[-1] if active_group else len(tracked_ctxs)-1
tracked_ctxs[dest_group].append(ctx:=TrackedGraphRewrite(loc, args[0].trace_num, [], rewrite_name, depth, kwargs.get("bottom_up", False),
kwargs.get("walk", False), kwargs.get("enter_calls", False)))
active_rewrites.append(ctx)
key = rewrite_name # profile spans are named after the rewrite step
with cpu_profile(key, "TINY") as e:
ret = func(*args, **kwargs)
if TRACK_MATCH_STATS >= 2:
if new_ctx: active_group.pop()
else: active_rewrites.pop()
if callable(name):
name_ret = name(*args, **kwargs, ret=ret)
assert isinstance(name_ret, (TracingKey, str)), f"name function returned {type(name_ret)}"
tracked_keys[idx] = k = TracingKey(n:=tracked_keys[idx].display_name.replace(fn, name_ret), (n,)) if isinstance(name_ret, str) else name_ret
e.name = TracingKey(k.display_name if isinstance(name_ret, str) else f"{fn} for {k.display_name}", k.keys)
if CAPTURE_PROCESS_REPLAY and replay:
# capture global context vars and all the args passed in
with Context(PROCESS_REPLAY_LOC=get_process_replay_loc()):
replay_capture.append(pickle.dumps((fn, args, kwargs, ContextVar._cache, ret)))
return ret
return __wrapper
return _decorator
class TrackedPatternMatcher(PatternMatcher):
def rewrite(self, uop:UOp, ctx=None):
if len(pats:=self.pdict.get(uop.op, [])):
ret = None
ler = {u.op for u in uop.src}
for p,match,early_reject in pats:
if p not in match_stats: match_stats[p] = [0,0,0.0,0.0]
st = time.perf_counter()
if not early_reject.issubset(ler):
match_stats[p][2] += time.perf_counter()-st
continue
match_stats[p][1] += 1
try: ret = match(uop, ctx)
except Exception as e:
if TRACK_MATCH_STATS >= 2 and active_rewrites:
err_str = f"{type(e).__name__}\n{sys.exc_info()[1]}"
active_rewrites[-1].matches.append((uop.trace_num, UOp(Ops.REWRITE_ERROR, src=uop.src, arg=err_str).trace_num, p.location, 0))
raise
if ret is not None and ret is not uop:
match_stats[p][0] += 1
match_stats[p][3] += (et:=time.perf_counter()-st)
if TRACK_MATCH_STATS >= 3: print(f"{et*1e6:7.2f} us -- ", printable(p.location))
if TRACK_MATCH_STATS >= 2 and isinstance(ret, UOp) and active_rewrites:
active_rewrites[-1].matches.append((uop.trace_num, ret.trace_num, p.location, et))
return ret
match_stats[p][2] += time.perf_counter()-st
return None
@dataclass(frozen=True)
class RewriteTrace: keys:list[TracingKey]; rewrites:list[list[TrackedGraphRewrite]]; uop_fields:dict[int, tuple] # noqa: E702
if TRACK_MATCH_STATS or PROFILE:
PatternMatcher = TrackedPatternMatcher # type: ignore
import atexit
@atexit.register
def print_match_stats():
if TRACK_MATCH_STATS >= 2:
with open(fn:=temp("rewrites.pkl", append_user=True), "wb") as f:
print(f"rewrote {len(tracked_ctxs)} graphs and matched {sum(len(r.matches) for x in tracked_ctxs for r in x)} times, saved to {fn}")
pickle.dump(RewriteTrace(tracked_keys, tracked_ctxs, uop_fields), f)
if getenv("PRINT_MATCH_STATS", int(TRACK_MATCH_STATS.value and not VIZ)):
ret = [0,0,0.0,0.0]
for k,v in sorted(list(match_stats.items()), key=lambda x: x[1][2]+x[1][3]):
loc_str = f"{k.location[0].split('/')[-1]}:{k.location[1]}"
if v[1] != 0: print(f"{v[0]:6d} / {v[1]:7d} -- {v[3]*1000.:9.2f} / {(v[2]+v[3])*1000.:9.2f} ms -- {loc_str:20s}", printable(k.location))
ret = [x+y for x,y in zip(ret, v)]
print(f"{ret[0]:6d} / {ret[1]:7d} -- {ret[3]*1000.:9.2f} / {(ret[2]+ret[3])*1000.:9.2f} ms -- TOTAL")
print(f"{len(match_stats)} rules, {sum(v[0] > 0 for v in match_stats.values())} matched once")
TRACK_MATCH_STATS.value = 0
launch_viz("REWRITE_DATA", temp("rewrites.pkl", append_user=True))
def launch_viz(env_str:str, data:str):
os.environ[f"{env_str}_DATA"] = data
if not TRACK_MATCH_STATS and not PROFILE:
os.environ["VIZ"], os.environ["PROFILE"], os.environ["TRACK_MATCH_STATS"] = "0", "0", "0"
args = ['--rewrites-path', os.getenv("REWRITE_DATA", "")] if os.getenv("REWRITE_DATA", "") else []
args += ['--profile-path', os.getenv("PROFILE_DATA", "")] if os.getenv("PROFILE_DATA", "") else []
viz_path = pathlib.Path(__file__).resolve().parent.parent / "viz" / "serve.py"
if VIZ > 0 and sys.stdout.isatty(): os.execv(sys.executable, [sys.executable, viz_path.as_posix()] + args)
if VIZ: print("saved viz files, view using: python -m tinygrad.viz.cli")
VIZ.value = 0
# *** simple graph rewrite engine ***
# A pure Python sentinel, but *typed* as UOp so it fits all the dict annotations
SENTINEL: Final[UOp] = cast(UOp, object())
class BottomUpGate(Exception): pass
class RewriteContext:
def __init__(self, pm, bpm, ctx=None, enter_calls=False):
self.pm: PatternMatcher|None = pm
self.bpm: PatternMatcher|None = bpm
self.bpm_cache: dict[UOp, UOp|None] = {}
self.ctx = ctx
self.replace: dict[UOp, UOp] = {}
self.enter_calls = enter_calls
# no cache needed: pm_rewrite is called at most once per UOp due to the replace dict check in unified_rewrite
def pm_rewrite(self, x:UOp) -> UOp|None: return unwrap(self.pm).rewrite(x, self.ctx)
def cached_bpm_rewrite(self, x:UOp) -> UOp|None:
if (ret:=self.bpm_cache.get(x,SENTINEL)) is not SENTINEL: return ret
ret = self.bpm_cache[x] = unwrap(self.bpm).rewrite(x, self.ctx)
return ret
def walk_rewrite(self, root:UOp) -> UOp:
"""MLIR-style Walk Pattern Rewrite Driver: single-pass, no re-traversal into rewritten subtrees."""
stack: list[tuple[UOp, bool]] = [(root, False)]
while stack:
n, processed = stack.pop()
if n in self.replace: continue
if not processed:
# bottom-up: try bpm on original node first, if it rewrites, use result as-is (no traversal into replacement)
if self.bpm is not None and (rewritten:=self.cached_bpm_rewrite(n)) is not None:
self.replace[n] = rewritten
continue
# no rewrite, process children then come back to rebuild
stack.append((n, True))
# CALL bodies are never rewritten separately, rewrites that need them pass enter_calls=True
for x in reversed(n.src[1:] if n.op is Ops.CALL and not self.enter_calls else n.src):
if x not in self.replace: stack.append((x, False))
else:
# rebuild node with rewritten srcs
skip = int(n.op is Ops.CALL and not self.enter_calls)
new_src = n.src[:skip] + tuple(self.replace.get(x, x) for x in n.src[skip:])
new_n = UOp(n.op, new_src, n.arg, n.tag) if new_src != n.src else n
# top-down: try pm on rebuilt node, use result as-is (no re-traversal)
if self.pm is not None and (rewritten:=self.pm_rewrite(new_n)) is not None: new_n = rewritten
self.replace[n] = new_n
return self.replace.get(root, root)
def unified_rewrite(self, root:UOp) -> UOp:
stack: collections.deque[tuple[UOp, int, UOp]] = collections.deque([(root, 0, root)])
on_stack = {root} # all UOps either on the stack or in self.replace, i.e. dont have to be placed again
waitlist: dict[UOp, list[tuple[UOp, int, UOp]]] = {} # UOps waiting on a dependency to be in self.replace
while stack:
if len(stack) > REWRITE_STACK_LIMIT: raise RuntimeError("infinite loop in graph_rewrite (stack too big)")
n, stage, new_n = stack.pop()
if n in self.replace: continue # skip any nodes we have seen
if stage == 0:
# if bottom up, we rewrite this node early. in both cases, we add its srcs to the stack
if self.bpm is not None:
# apply rewrite rules until a fixed point is reached. may return `uop` itself if PatternMatcher doesn't match
test_n: UOp|None = n
seen = set()
try:
while test_n is not None:
if test_n in seen: raise RuntimeError("infinite loop in fixed_point_rewrite")
seen.add(test_n)
new_n, test_n = test_n, self.cached_bpm_rewrite(test_n)
except BottomUpGate:
# if the bpm matching raised a gate, we are done with this node and dont continue down the srcs
self.replace[n] = unwrap(test_n)
if n in waitlist: stack.extend(waitlist.pop(n))
continue
stack.append((n, 1, new_n))
# NOTE: CALLs are handled as a special case: their bodies are not included in the graph_rewrite,
# rewrites that need them pass enter_calls=True
for x in reversed(new_n.src[1:] if new_n.op is Ops.CALL and not self.enter_calls else new_n.src):
if x in on_stack: continue
stack.append((x, 0, x))
on_stack.add(x)
elif stage == 1:
skip = int(new_n.op is Ops.CALL and not self.enter_calls)
tmp = list(new_n.src[:skip])
for x in new_n.src[skip:]:
if (rx:=self.replace.get(x, SENTINEL)) is SENTINEL:
# source not ready: register in waitlist instead of spinning
waitlist.setdefault(x, []).append((n, 1, new_n))
break
tmp.append(rx)
else:
# in stage 1, once all srcs are rewritten, rebuild (if changed) or run top-down rewrite
if (new_src:=tuple(tmp)) == new_n.src:
# if top down, do the rewrite. if no rewrite or bottom up, we are done rewriting this node so we add it to the dict
if self.pm is None or (new_src_n:=self.pm_rewrite(new_n)) is None:
self.replace[n] = new_n
if n in waitlist: stack.extend(waitlist.pop(n))
continue
else:
# if srcs changed from rewrites, construct a new UOp with the new srcs
new_src_n = UOp(new_n.op, new_src, new_n.arg, new_n.tag)
# trigger a rewrite of new_src_n, then after that rewrite is done, link it back to n
stack.append((n, 2, new_src_n))
stack.append((new_src_n, 0, new_src_n))
else:
# in stage 2, we link the result of new_n to the result of n
if (replaced_new_n:=self.replace.get(new_n, SENTINEL)) is SENTINEL:
# not ready: register in waitlist instead of spinning
waitlist.setdefault(new_n, []).append((n, 2, new_n))
else:
# otherwise we are done
self.replace[n] = replaced_new_n
if n in waitlist: stack.extend(waitlist.pop(n))
if root not in self.replace:
def label(u:UOp) -> str: return f"{u.op.name}@{id(u):x}"
details = [f" {label(n)} -> {label(new_n)} waits for {label(dep)}"
for dep, waiters in itertools.islice(waitlist.items(), 5) for n, _, new_n in waiters[:1]]
raise RuntimeError("graph_rewrite stalled: unresolved rewrite dependencies (possible cycle). "
"A replacement may depend on the node being rewritten.\n" + "\n".join(details))
return self.replace[root]
@rewrite_group(new_ctx=False)
def graph_rewrite(sink:UOp, pm:PatternMatcher, ctx=None, bottom_up=False, name=None, bpm=None, walk=False, enter_calls=False) -> UOp:
rewrite_ctx = RewriteContext(pm if not bottom_up else None, pm if bottom_up else bpm, ctx, enter_calls)
return rewrite_ctx.walk_rewrite(sink) if walk else rewrite_ctx.unified_rewrite(sink)
def sint_to_uop(x:sint, dtype=dtypes.weakint) -> UOp: return UOp.const(x, dtype)
def to_max_shape(shape:tuple[sint, ...]) -> tuple[int, ...]: return tuple(int(x.vmax) if isinstance(x, UOp) else x for x in shape)
_substitute = PatternMatcher([(UPat(tuple(Ops), name="x"), lambda ctx,x: ctx.get(x,None))])
_pm_resolve_params = PatternMatcher([(UPat(Ops.PARAM, name="p"), lambda ctx,p: ctx[p.arg.slot] if p.arg.slot >= 0 else None)])
def resolve_returned_after(r:UOp, t:UOp) -> UOp|None:
"""Extract a call output's matching store, preserving writes to the enclosing scope's output PARAMs."""
stores = [st for st in t.src if st.op is Ops.STORE and st.src[0].unsharded_base is r.unsharded_base]
if len(stores) != 1: return None
return r.after(stores[0]) if r.unsharded_base.op is Ops.PARAM else stores[0].src[1]
remove_all_tags = PatternMatcher([(UPat(GroupOp.All, name="x"), lambda x: x.replace(tag=None) if x.tag is not None else None)])
# a store's storage keeps the views and drops AFTERs (they only sequence stores)
pm_drop_after = PatternMatcher([(UPat(Ops.AFTER, name="a"), lambda a: a.src[0])])
pm_renumber_slots = PatternMatcher([
(UPat(Ops.RANGE, name="u"), lambda ctx, u: u.replace(arg=(u.axis_type, next(ctx))+u.axis_id[1:])),
(UPat(Ops.BUFFER, name="u"), lambda ctx, u: u.replace(arg=replace(u.arg, slot=next(ctx))) if u.addrspace is AddrSpace.REG else None),
])
def gate_kernel_sink(x:UOp) -> bool:
if x.op is Ops.LINEAR: return False
if x.op is Ops.SINK and isinstance(x.arg, KernelInfo): return False
return True
# *** what was symbolic.py ***
sint = int|UOp
Variable = UOp
ConstLike = ConstType|Variable|tuple[ConstType, ...]
Reference in New Issue View Git Blame Copy Permalink