File size: 11,034 Bytes
66c9c8a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
import ctypes
import math
from typing import Any

import warp
from warp.types import *


class fabricbucket_t(ctypes.Structure):
    _fields_ = [
        ("index_start", ctypes.c_size_t),
        ("index_end", ctypes.c_size_t),
        ("ptr", ctypes.c_void_p),
        ("lengths", ctypes.c_void_p),
    ]

    def __init__(self, index_start=0, index_end=0, ptr=None, lengths=None):
        self.index_start = index_start
        self.index_end = index_end
        self.ptr = ctypes.c_void_p(ptr)
        self.lengths = ctypes.c_void_p(lengths)


class fabricarray_t(ctypes.Structure):
    _fields_ = [
        ("buckets", ctypes.c_void_p),  # array of fabricbucket_t on the correct device
        ("nbuckets", ctypes.c_size_t),
        ("size", ctypes.c_size_t),
    ]

    def __init__(self, buckets=None, nbuckets=0, size=0):
        self.buckets = ctypes.c_void_p(buckets)
        self.nbuckets = nbuckets
        self.size = size


class indexedfabricarray_t(ctypes.Structure):
    _fields_ = [
        ("fa", fabricarray_t),
        ("indices", ctypes.c_void_p),
        ("size", ctypes.c_size_t),
    ]

    def __init__(self, fa=None, indices=None):
        if fa is None:
            self.fa = fabricarray_t()
        else:
            self.fa = fa.__ctype__()

        if indices is None:
            self.indices = ctypes.c_void_p(None)
            self.size = 0
        else:
            self.indices = ctypes.c_void_p(indices.ptr)
            self.size = indices.size


def fabric_to_warp_dtype(type_info, attrib_name):
    if not type_info[0]:
        raise RuntimeError(f"Attribute '{attrib_name}' cannot be used in Warp")

    base_type_dict = {
        "b": warp.bool,  # boolean
        "i1": warp.int8,
        "i2": warp.int16,
        "i4": warp.int32,
        "i8": warp.int64,
        "u1": warp.uint8,
        "u2": warp.uint16,
        "u4": warp.uint32,
        "u8": warp.uint64,
        "f2": warp.float16,
        "f4": warp.float32,
        "f8": warp.float64,
    }

    base_dtype = base_type_dict.get(type_info[1])
    if base_dtype is None:
        raise RuntimeError(f"Attribute '{attrib_name}' base data type '{type_info[1]}' is not supported in Warp")

    elem_count = type_info[2]
    role = type_info[4]

    if role in ("text", "path"):
        raise RuntimeError(f"Attribute '{attrib_name}' role '{role}' is not supported in Warp")

    if elem_count > 1:
        # vector or matrix type
        if role == "quat" and elem_count == 4:
            return quaternion(base_dtype)
        elif role in ("matrix", "transform", "frame"):
            # only square matrices are currently supported
            mat_size = int(math.sqrt(elem_count))
            assert mat_size * mat_size == elem_count
            return matrix((mat_size, mat_size), base_dtype)
        else:
            return vector(elem_count, base_dtype)
    else:
        # scalar type
        return base_dtype


class fabricarray(noncontiguous_array_base[T]):
    # member attributes available during code-gen (e.g.: d = arr.shape[0])
    # (initialized when needed)
    _vars = None

    def __init__(self, data=None, attrib=None, dtype=Any, ndim=None):
        super().__init__(ARRAY_TYPE_FABRIC)

        if data is not None:
            from .context import runtime

            # ensure the attribute name was also specified
            if not isinstance(attrib, str):
                raise ValueError(f"Invalid attribute name: {attrib}")

            # get the fabric interface dictionary
            if isinstance(data, dict):
                iface = data
            elif hasattr(data, "__fabric_arrays_interface__"):
                iface = data.__fabric_arrays_interface__
            else:
                raise ValueError(
                    "Invalid data argument for fabricarray: expected dict or object with __fabric_arrays_interface__"
                )

            version = iface.get("version")
            if version != 1:
                raise ValueError(f"Unsupported Fabric interface version: {version}")

            device = iface.get("device")
            if not isinstance(device, str):
                raise ValueError(f"Invalid Fabric interface device: {device}")

            self.device = runtime.get_device(device)

            attribs = iface.get("attribs")
            if not isinstance(attribs, dict):
                raise ValueError("Failed to get Fabric interface attributes")

            # look up attribute info by name
            attrib_info = attribs.get(attrib)
            if not isinstance(attrib_info, dict):
                raise ValueError(f"Failed to get attribute '{attrib}'")

            type_info = attrib_info["type"]
            assert len(type_info) == 5

            self.dtype = fabric_to_warp_dtype(type_info, attrib)

            self.access = attrib_info["access"]

            pointers = attrib_info["pointers"]
            counts = attrib_info["counts"]

            if not (hasattr(pointers, "__len__") and hasattr(counts, "__len__") and len(pointers) == len(counts)):
                raise RuntimeError("Attribute pointers and counts must be lists of the same size")

            # check whether it's an array
            array_depth = type_info[3]
            if array_depth == 0:
                self.ndim = 1
                array_lengths = None
            elif array_depth == 1:
                self.ndim = 2
                array_lengths = attrib_info["array_lengths"]
                if not hasattr(array_lengths, "__len__") or len(array_lengths) != len(pointers):
                    raise RuntimeError(
                        "Attribute `array_lengths` must be a list of the same size as `pointers` and `counts`"
                    )
            else:
                raise ValueError(f"Invalid attribute array depth: {array_depth}")

            num_buckets = len(pointers)
            size = 0

            buckets = (fabricbucket_t * num_buckets)()
            for i in range(num_buckets):
                buckets[i].index_start = size
                buckets[i].index_end = size + counts[i]
                buckets[i].ptr = pointers[i]
                if array_lengths:
                    buckets[i].lengths = array_lengths[i]
                size += counts[i]

            if self.device.is_cuda:
                # copy bucket info to device
                with warp.ScopedStream(self.device.null_stream):
                    buckets_size = ctypes.sizeof(buckets)
                    buckets_ptr = self.device.allocator.alloc(buckets_size)
                    runtime.core.memcpy_h2d(self.device.context, buckets_ptr, ctypes.addressof(buckets), buckets_size)
            else:
                buckets_ptr = ctypes.addressof(buckets)

            self.buckets = buckets
            self.size = size
            self.shape = (size,)

            self.ctype = fabricarray_t(buckets_ptr, num_buckets, size)

        else:
            # empty array or type annotation
            self.dtype = dtype
            self.ndim = ndim or 1
            self.device = None
            self.access = None
            self.buckets = None
            self.size = 0
            self.shape = (0,)
            self.ctype = fabricarray_t()

    def __del__(self):
        # release the GPU copy of bucket info
        if self.buckets is not None and self.device.is_cuda:
            buckets_size = ctypes.sizeof(self.buckets)
            with self.device.context_guard:
                self.device.allocator.free(self.ctype.buckets, buckets_size)

    def __ctype__(self):
        return self.ctype

    def __len__(self):
        return self.size

    def __str__(self):
        if self.device is None:
            # type annotation
            return f"fabricarray{self.dtype}"
        else:
            return str(self.numpy())

    def __getitem__(self, key):
        if isinstance(key, array):
            return indexedfabricarray(fa=self, indices=key)
        else:
            raise ValueError(f"Fabric arrays only support indexing using index arrays, got key of type {type(key)}")

    @property
    def vars(self):
        # member attributes available during code-gen (e.g.: d = arr.shape[0])
        # Note: we use a shared dict for all fabricarray instances
        if fabricarray._vars is None:
            fabricarray._vars = {"size": warp.codegen.Var("size", uint64)}
        return fabricarray._vars

    def fill_(self, value):
        # TODO?
        # filling Fabric arrays of arrays is not supported, because they are jagged arrays of arbitrary lengths
        if self.ndim > 1:
            raise RuntimeError("Filling Fabric arrays of arrays is not supported")

        super().fill_(value)


# special case for fabric array of arrays
# equivalent to calling fabricarray(..., ndim=2)
def fabricarrayarray(**kwargs):
    kwargs["ndim"] = 2
    return fabricarray(**kwargs)


class indexedfabricarray(noncontiguous_array_base[T]):
    # member attributes available during code-gen (e.g.: d = arr.shape[0])
    # (initialized when needed)
    _vars = None

    def __init__(self, fa=None, indices=None, dtype=None, ndim=None):
        super().__init__(ARRAY_TYPE_FABRIC_INDEXED)

        if fa is not None:
            check_index_array(indices, fa.device)
            self.fa = fa
            self.indices = indices
            self.dtype = fa.dtype
            self.ndim = fa.ndim
            self.device = fa.device
            self.size = indices.size
            self.shape = (indices.size,)
            self.ctype = indexedfabricarray_t(fa, indices)
        else:
            # allow empty indexedarrays in type annotations
            self.fa = None
            self.indices = None
            self.dtype = dtype
            self.ndim = ndim or 1
            self.device = None
            self.size = 0
            self.shape = (0,)
            self.ctype = indexedfabricarray_t()

    def __ctype__(self):
        return self.ctype

    def __len__(self):
        return self.size

    def __str__(self):
        if self.device is None:
            # type annotation
            return f"indexedfabricarray{self.dtype}"
        else:
            return str(self.numpy())

    @property
    def vars(self):
        # member attributes available during code-gen (e.g.: d = arr.shape[0])
        # Note: we use a shared dict for all indexedfabricarray instances
        if indexedfabricarray._vars is None:
            indexedfabricarray._vars = {"size": warp.codegen.Var("size", uint64)}
        return indexedfabricarray._vars

    def fill_(self, value):
        # TODO?
        # filling Fabric arrays of arrays is not supported, because they are jagged arrays of arbitrary lengths
        if self.ndim > 1:
            raise RuntimeError("Filling indexed Fabric arrays of arrays is not supported")

        super().fill_(value)


# special case for indexed fabric array of arrays
# equivalent to calling fabricarray(..., ndim=2)
def indexedfabricarrayarray(**kwargs):
    kwargs["ndim"] = 2
    return indexedfabricarray(**kwargs)