You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by mo...@apache.org on 2021/01/09 02:00:45 UTC

[tvm] branch main updated: [RELAY, TOPI] Threefry PRNG: splittable and stateless (#7083)

This is an automated email from the ASF dual-hosted git repository.

moreau pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git


The following commit(s) were added to refs/heads/main by this push:
     new 701bcc2  [RELAY,TOPI] Threefry PRNG: splittable and stateless (#7083)
701bcc2 is described below

commit 701bcc2c5faf0d8f589478000a92d3ad577e8df6
Author: Tristan Konolige <tr...@gmail.com>
AuthorDate: Fri Jan 8 18:00:24 2021 -0800

    [RELAY,TOPI] Threefry PRNG: splittable and stateless (#7083)
    
    * [RELAY,TOPI] Threefry PRNG: splittable and stateless
    
    * Fix sphinx?
    
    * Lint fixes
    
    * sphinx fixes round 2
    
    * fix inputs for tests
    
    * reorganize to random, fix uninitialized memory bug
    
    * silence linter
    
    * silence linter even further
    
    * s
    
    * strengthen Threefry key type checking, add tests
    
    * replace static variable with function for Threefry key type
    
    * lint fix
    
    * Remove old todos, improve assert messages
    
    * describe how random number is generated
    
    * add tests for incorrect output size. also vary test sizes
    
    Co-authored-by: Altan Haan <ah...@octoml.ai>
---
 include/tvm/relay/attrs/random.h           |  42 +++
 python/tvm/relay/__init__.py               |   1 +
 python/tvm/relay/op/__init__.py            |   1 +
 python/tvm/relay/op/algorithm.py           |   2 +-
 python/tvm/relay/op/op_attrs.py            |   5 +
 python/tvm/relay/op/random/__init__.py     |  20 ++
 python/tvm/relay/op/random/_kernel.py      |  29 ++
 python/tvm/relay/op/random/_make.py        |  20 ++
 python/tvm/relay/op/random/kernel.py       | 134 ++++++++++
 python/tvm/relay/op/strategy/generic.py    |  44 ++++
 python/tvm/topi/__init__.py                |   1 +
 python/tvm/topi/random/__init__.py         |  22 ++
 python/tvm/topi/random/kernel.py           | 410 +++++++++++++++++++++++++++++
 src/relay/op/random/kernel.cc              |  89 +++++++
 tests/python/relay/test_prng.py            | 142 ++++++++++
 tests/python/topi/python/test_topi_prng.py | 116 ++++++++
 16 files changed, 1077 insertions(+), 1 deletion(-)

diff --git a/include/tvm/relay/attrs/random.h b/include/tvm/relay/attrs/random.h
new file mode 100644
index 0000000..8238f10
--- /dev/null
+++ b/include/tvm/relay/attrs/random.h
@@ -0,0 +1,42 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied.  See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+/*!
+ * \file tvm/relay/attrs/vision.h
+ * \brief Auxiliary attributes for random operators.
+ */
+#ifndef TVM_RELAY_ATTRS_RANDOM_H_
+#define TVM_RELAY_ATTRS_RANDOM_H_
+
+#include <tvm/ir/attrs.h>
+
+namespace tvm {
+namespace relay {
+
+struct ThreefryGenerateAttrs : public tvm::AttrsNode<ThreefryGenerateAttrs> {
+  Array<Integer> out_shape;
+
+  TVM_DECLARE_ATTRS(ThreefryGenerateAttrs, "relay.attrs.ThreefryGenerateAttrs") {
+    TVM_ATTR_FIELD(out_shape).describe("Shape of random numbers to generate");
+  }
+};
+
+}  // namespace relay
+}  // namespace tvm
+#endif  // TVM_RELAY_ATTRS_RANDOM_H_
diff --git a/python/tvm/relay/__init__.py b/python/tvm/relay/__init__.py
index cd96ecc..97f6d1c 100644
--- a/python/tvm/relay/__init__.py
+++ b/python/tvm/relay/__init__.py
@@ -45,6 +45,7 @@ from .op import annotation
 from .op import vision
 from .op import contrib
 from .op import dyn
+from .op import random
 from .op.reduce import *
 from .op.tensor import *
 from .op.transform import *
diff --git a/python/tvm/relay/op/__init__.py b/python/tvm/relay/op/__init__.py
index f6afa44..1f267ab 100644
--- a/python/tvm/relay/op/__init__.py
+++ b/python/tvm/relay/op/__init__.py
@@ -43,6 +43,7 @@ from . import memory
 from . import image
 from . import vision
 from . import op_attrs
+from . import random
 
 
 # operator registry
diff --git a/python/tvm/relay/op/algorithm.py b/python/tvm/relay/op/algorithm.py
index 99140fc..6fd5c06 100644
--- a/python/tvm/relay/op/algorithm.py
+++ b/python/tvm/relay/op/algorithm.py
@@ -17,9 +17,9 @@
 """Classic algorithm operation"""
 from __future__ import absolute_import as _abs
 
+from ..expr import Constant, Expr, TupleWrapper
 from . import _make
 from .dyn import _make as _dyn_make
-from ..expr import TupleWrapper, Expr, Constant
 
 
 def sort(data, axis=-1, is_ascend=1):
diff --git a/python/tvm/relay/op/op_attrs.py b/python/tvm/relay/op/op_attrs.py
index cb837b1..4107681 100644
--- a/python/tvm/relay/op/op_attrs.py
+++ b/python/tvm/relay/op/op_attrs.py
@@ -552,3 +552,8 @@ class SpaceToBatchNDAttrs(Attrs):
 @tvm._ffi.register_object("relay.attrs.BatchToSpaceNDAttrs")
 class BatchToSpaceNDAttrs(Attrs):
     """Attributes used in BatchToSpaceNDAttrs operators"""
+
+
+@tvm._ffi.register_object("relay.attrs.ThreefryGenerateAttrs")
+class ThreefryGenerateAttrs(Attrs):
+    """Attributes used in ThreefryGenerateAttrs operators"""
diff --git a/python/tvm/relay/op/random/__init__.py b/python/tvm/relay/op/random/__init__.py
new file mode 100644
index 0000000..8366f4a
--- /dev/null
+++ b/python/tvm/relay/op/random/__init__.py
@@ -0,0 +1,20 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+# pylint: disable=wildcard-import
+"""PRNG related operators."""
+from .kernel import *
+from . import _kernel
diff --git a/python/tvm/relay/op/random/_kernel.py b/python/tvm/relay/op/random/_kernel.py
new file mode 100644
index 0000000..8be3397
--- /dev/null
+++ b/python/tvm/relay/op/random/_kernel.py
@@ -0,0 +1,29 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+"""Splittable and parallelizable PRNG kernels."""
+# pylint: disable=invalid-name,unused-argument
+from __future__ import absolute_import
+
+from .. import strategy
+from ..op import register_strategy, register_pattern, OpPattern
+
+
+# Threefry
+register_strategy("random.threefry_generate", strategy.threefry_generate_strategy)
+register_pattern("random.threefry_generate", OpPattern.OPAQUE)
+register_strategy("random.threefry_split", strategy.threefry_split_strategy)
+register_pattern("random.threefry_split", OpPattern.OPAQUE)
diff --git a/python/tvm/relay/op/random/_make.py b/python/tvm/relay/op/random/_make.py
new file mode 100644
index 0000000..51a8a6a
--- /dev/null
+++ b/python/tvm/relay/op/random/_make.py
@@ -0,0 +1,20 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+"""Constructor APIs"""
+import tvm._ffi
+
+tvm._ffi._init_api("relay.op.random._make", __name__)
diff --git a/python/tvm/relay/op/random/kernel.py b/python/tvm/relay/op/random/kernel.py
new file mode 100644
index 0000000..9663494
--- /dev/null
+++ b/python/tvm/relay/op/random/kernel.py
@@ -0,0 +1,134 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+"""Splittable and parallelizable PRNG kernels."""
+# pylint: disable=invalid-name,unused-argument
+from __future__ import absolute_import
+
+import sys
+import numpy as np
+
+from ...expr import Constant
+from .... import nd
+from . import _make
+
+
+def threefry_key(seed):
+    """Create a new Threefry random number generator key.
+
+    Example
+    -------
+
+    .. code-block:: python
+
+        gen = threefry_key(0)
+        _, random_number = threefry_generate(gen, (4,))
+
+    Parameters
+    ----------
+    seed : int
+        Starting seed for the key
+
+    Returns
+    -------
+    key : relay.Expr
+        New key to pass to future uses of :py:func:`threefry_split` or
+        :py:func:`threefry_generate`.
+    """
+    s = np.frombuffer(seed.to_bytes(32, sys.byteorder), dtype="uint64")
+    a = np.concatenate((s, np.array([0, 0, 0, 0, 1 << 63, 0], dtype="uint64")))
+    return Constant(nd.array(a))
+
+
+def threefry_generate(key, shape):
+    """Generate an array of random bits (`uint64`) using the Threefry algorithm
+
+    Example
+    -------
+
+    .. code-block:: python
+
+        key = threefry_key(0)
+        new_key, random1 = threefry_generate(key, (4,))
+        _, random2 = threefry_generate(new_key, (4,))
+        # random1 and random2 are different random numbers
+
+    Parameters
+    ----------
+    key : relay.Expr
+        key that uniquely determines the random values. Multiple uses with the
+        same key will generate the same random values. This key should be
+        treated as an opaque pointer. You can create one from calling
+        :py:func:`threefry_key`, :py:func:`threefry_split`, or
+        :py:func:`threefry_generate`. **Do not use this key again after calling
+        this function.**
+
+    shape : Sequence[int]
+        Desired outputs shape of random numbers. **Currently the total
+        number of elements must be a multiple of 4.**
+
+    Returns
+    -------
+    new_key : relay.Expr
+        New key to pass to future uses of :py:func:`threefry_split` or
+        :py:func:`threefry_generate`.
+
+    random_array : relay.Expr
+        Array of random numbers. Has shape `shape`.
+    """
+    return _make.threefry_generate(key, shape)
+
+
+def threefry_split(key):
+    """Split an existing Threefry key into two new ones.
+
+    This is useful if you have to subsequent calls which each need their own
+    independent random number generation.
+
+    Example
+    -------
+
+    .. code-block:: python
+
+        def foo(key):
+            new_key, num = threefry_generate(key, (4,))
+            return num
+
+        key = threefry_key(0)
+        key1, key2 = threefry_split(key)
+        assert foo(key1) != foo(key2)
+
+    Parameters
+    ----------
+    key : relay.Expr
+        key that uniquely determines the random values. Multiple uses with the
+        same generator will generate the same random values. This generator should be
+        treated as an opaque pointer. You can create one from calling
+        :py:func:`threefry_key`, :py:func:`threefry_split`, or
+        :py:func:`threefry_generate`. **Do not use this generator again after calling
+        this function.**
+
+    Returns
+    -------
+    new_key_1 : relay.Expr
+        New key to pass to future uses of :py:func:`threefry_split` or
+        :py:func:`threefry_generate`.
+
+    new_key_2 : relay.Expr
+        New key to pass to future uses of :py:func:`threefry_split` or
+        :py:func:`threefry_generate`.
+    """
+    return _make.threefry_split(key)
diff --git a/python/tvm/relay/op/strategy/generic.py b/python/tvm/relay/op/strategy/generic.py
index 9599e72..363832e 100644
--- a/python/tvm/relay/op/strategy/generic.py
+++ b/python/tvm/relay/op/strategy/generic.py
@@ -1317,3 +1317,47 @@ def argwhere_strategy(attrs, inputs, out_type, target):
         name="argwhere.generic",
     )
     return strategy
+
+
+# threefry_generate
+def wrap_compute_threefry_generate(topi_compute):
+    """Wrap threefry_generate topi compute"""
+
+    def _compute_threefry_generate(attrs, inputs, _):
+        return topi_compute(inputs[0], attrs.out_shape)
+
+    return _compute_threefry_generate
+
+
+@override_native_generic_func("threefry_generate_strategy")
+def threefry_generate_strategy(attrs, inputs, out_type, target):
+    """threefry_generate generic strategy"""
+    strategy = _op.OpStrategy()
+    strategy.add_implementation(
+        wrap_compute_threefry_generate(topi.random.threefry_generate),
+        wrap_topi_schedule(topi.generic.schedule_extern),
+        name="threefry_generate.generic",
+    )
+    return strategy
+
+
+# threefry_split
+def wrap_compute_threefry_split(topi_compute):
+    """Wrap threefry_split topi compute"""
+
+    def _compute_threefry_split(attrs, inputs, _):
+        return topi_compute(inputs[0])
+
+    return _compute_threefry_split
+
+
+@override_native_generic_func("threefry_split_strategy")
+def threefry_split_strategy(attrs, inputs, out_type, target):
+    """threefry_split generic strategy"""
+    strategy = _op.OpStrategy()
+    strategy.add_implementation(
+        wrap_compute_threefry_split(topi.random.threefry_split),
+        wrap_topi_schedule(topi.generic.schedule_extern),
+        name="threefry_split.generic",
+    )
+    return strategy
diff --git a/python/tvm/topi/__init__.py b/python/tvm/topi/__init__.py
index 97951d9..cb94b5b 100644
--- a/python/tvm/topi/__init__.py
+++ b/python/tvm/topi/__init__.py
@@ -54,6 +54,7 @@ from . import vision
 from . import image
 from . import sparse
 from . import hls
+from . import random
 
 # error reporting
 from .utils import InvalidShapeError
diff --git a/python/tvm/topi/random/__init__.py b/python/tvm/topi/random/__init__.py
new file mode 100644
index 0000000..ee8d1d6
--- /dev/null
+++ b/python/tvm/topi/random/__init__.py
@@ -0,0 +1,22 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+# pylint: disable=wildcard-import
+"""Pseudorandom generator kernels and operators."""
+from __future__ import absolute_import
+
+from .kernel import *
diff --git a/python/tvm/topi/random/kernel.py b/python/tvm/topi/random/kernel.py
new file mode 100644
index 0000000..576fd92
--- /dev/null
+++ b/python/tvm/topi/random/kernel.py
@@ -0,0 +1,410 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+"""Pseudorandom number kernels."""
+import tvm
+import tvm.topi
+from ... import tir
+from ...tir import ir_builder
+
+
+# Threefry PRNG with splitting based on
+# - J. K. Salmon, M. A. Moraes, R. O. Dror and D. E. Shaw, "Parallel random numbers: As easy as 1,
+#   2, 3," SC '11: Proceedings of 2011 International Conference for High Performance Computing,
+#   Networking, Storage and Analysis, Seattle, WA, 2011, pp. 1-12, doi: 10.1145/2063384.2063405.
+# - Claessen, K. ; Palka, M. (2013) "Splittable Pseudorandom Number Generators using Cryptographic
+#   Hashing". Proceedings of Haskell Symposium 2013 pp. 47-58.  MLA
+# - Ferguson, Niels, et al. "The Skein hash function family." Submission to NIST (round 3) 7.7.5
+#   (2010): 3.
+
+
+# Threefry is a counter based PRNG: given a unique input, it generates a unique random number. As
+# there is no state to maintain, we can apply it to a sequence of numbers (0..N) to generate a
+# sequence of random numbers in parallel. In order to make the PRNG splittable (that is we can
+# generate a sequence of random numbers in one place, and another sequence in another), we add a
+# path and key in addition to the counter. The path allows us to encode a sequence of splits (a 0 in
+# the path indicates the left result of a split, a 1 indicates the right). To avoid continuously
+# growing the path, we can compress an existing path into the key portion of the generator by
+# hashing the current key, path, and counter to create the new key (this same technique is used if
+# we run out of room for the counter). They key is initialized with a unique initial state.
+#
+# Random numbers are generated by applying the Threefry hash to the current key, path, and counter.
+
+# This module use encoding e4 from the appendix of "Splittable Pseudorandom Number Generators using
+# Cryptographic Hashing" (confusingly, the definition in the paper uses e3 to define the encoding
+# function). This encoding uses a 10 element uint64 tensor where each byte means the following:
+
+# .. code-block:
+
+#     gen:
+#     words: 0 1 2 3 | 4 5  | 6 7     | 8 9
+#     usage: key     | path | counter | position of next step in path encoded in binary
+#                                       ex: 0b00010 -> next path entry goes one from the right
+
+# Right now, counter only uses the rightmost word.
+
+# Threefry rotation constants from the Skein paper ("The Skein Hash Function Family"
+# https://www.schneier.com/wp-content/uploads/2015/01/skein.pdf)
+_ROTATIONS = {
+    4: [[14, 16], [52, 57], [23, 40], [5, 37], [25, 33], [46, 12], [58, 22], [32, 32]],
+    8: [
+        [46, 36, 19, 37],
+        [33, 27, 14, 42],
+        [17, 49, 36, 39],
+        [44, 9, 54, 56],
+        [39, 30, 34, 24],
+        [13, 50, 10, 17],
+        [25, 29, 39, 43],
+        [8, 35, 56, 22],
+    ],
+    16: [
+        [24, 13, 8, 47, 8, 17, 22, 37],
+        [38, 19, 10, 55, 49, 18, 23, 52],
+        [33, 4, 51, 13, 34, 41, 59, 17],
+        [5, 20, 48, 41, 47, 28, 16, 25],
+        [41, 9, 37, 31, 12, 47, 44, 30],
+        [16, 34, 56, 51, 4, 53, 42, 41],
+        [31, 44, 47, 46, 19, 42, 44, 25],
+        [9, 48, 35, 52, 23, 31, 37, 20],
+    ],
+}
+
+# Threefry permutation constants from the Skein paper ("The Skein Hash Function Family"
+# https://www.schneier.com/wp-content/uploads/2015/01/skein.pdf)
+_PERMUTATIONS = {
+    4: [0, 3, 2, 1],
+    8: [2, 1, 4, 7, 6, 5, 0, 3],
+    16: [0, 9, 2, 13, 6, 11, 4, 15, 10, 7, 12, 3, 14, 5, 8, 1],
+}
+
+
+def _threefry(
+    irb, key_buf, key_offset, counter_buf, counter_offset, out_buf, out_offset, out_shape
+):
+    """IRBuilder code for running Threefry
+
+    Parameters
+    ----------
+    irb: IRBuilder
+        IRBuilder that this code will be generated for.
+
+    key_buf: BufferVar
+        Buffer to read the key from.
+
+    key_offset: number
+        Threefry will write to :code:`key_buf[key_offset:key_offset+4]`
+
+    counter_buf: BufferVar
+        Buffer to read the counter from.
+
+    counter_offset: number
+        Threefry will write to :code:`counter_buf[counter_offset:counter_offset+4]`
+
+    out_buf: BufferVar
+        Buffer to read the counter from.
+
+    out_offset: number
+        Threefry will write to :code:`out_buf[out_offset:out_offset+4*product(out_shape)]`
+
+    out_shape: number
+        Determines the number of ouput states to generate. :code:`state[i]` will correspond to
+        counter+i.
+    """
+    nrounds = 20
+    nwords = 4
+    iwidth = 64
+    assert nrounds % 4 == 0
+    assert nwords in [4, 8, 16]
+
+    # The paper has constants for 32 bit threefry, but we keep the implementation simple by only
+    # using 64-bit words.
+    assert key_buf.dtype == "uint64", "threefry only supports 64-bit keys"
+    assert key_buf.dtype == counter_buf.dtype, "threefry key and counter must be the same dtype"
+
+    def mix(a, b, rotation):
+        x = a + b  # TODO should be wrapping
+        y = x ^ ((b << rotation) | (b >> (iwidth - rotation)))
+        return [x, y]
+
+    # temporary buffer for holding the results of _PERMUTATIONS
+    tmp = irb.allocate(out_buf.dtype, out_shape, name="tmp", scope="global")
+    tmp_offset = 0
+
+    # Initialize entire key. It is composed of the original key with one
+    # element appended. The appended element is the xor of all key words plus a
+    # constant.
+    full_key = irb.allocate("uint64", nwords + 1, name="full_key", scope="global")
+    for i in range(nwords):
+        full_key[i] = key_buf[key_offset + i]
+    # initial key constant, full_key[nwords] is equivalent to k_{N_W} in the Skein paper.
+    full_key[nwords] = tvm.tir.const(0x1BD11BDAA9FC1A22, dtype="uint64")
+    for i in range(nwords):
+        full_key[nwords] ^= key_buf[key_offset + i]
+
+    with irb.for_range(0, out_shape, dtype="uint64", name="i") as i:
+        for j in range(nwords):
+            out_buf[out_offset + i * nwords + j] = counter_buf[counter_offset + j] + i
+
+    def key_schedule(s, i):
+        # Threefry uses no tweak, so the key schedule is simple
+        if i == nwords - 1:
+            return full_key[(s + i) % (nwords + 1)] + tvm.tir.const(s, dtype="uint64")
+        return full_key[(s + i) % (nwords + 1)]
+
+    with irb.for_range(0, out_shape, name="l") as l:  # pylint: disable=invalid-name
+        for i in range(nrounds // 4):
+            for j in range(nwords):
+                out_buf[out_offset + l * nwords + j] += key_schedule(i, j)  # TODO wrapping
+            for k in range(4):
+                for j in range(nwords // 2):
+                    (
+                        out_buf[out_offset + l * nwords + j * 2 + 0],
+                        out_buf[out_offset + l * nwords + j * 2 + 1],
+                    ) = mix(
+                        out_buf[out_offset + l * nwords + j * 2 + 0],
+                        out_buf[out_offset + l * nwords + j * 2 + 1],
+                        _ROTATIONS[nwords][(i * 4 + k) % 8][j],
+                    )
+                for j in range(nwords):
+                    tmp[tmp_offset + l * nwords + j] = out_buf[
+                        out_offset + l * nwords + _PERMUTATIONS[nwords][j]
+                    ]
+                # number of rounds is even, so out always contains the result
+                (out_buf, tmp) = (tmp, out_buf)
+                (out_offset, tmp_offset) = (tmp_offset, out_offset)
+
+
+def threefry_generate(gen, out_shape):
+    """Generate a series of random values
+
+    Notes
+    -----
+    This function uses the counter portion of the generator state to generate a series of random
+    numbers in parallel. Random number `i` is generated by applying Threefry to the current
+    generator state with the counter portion incremented by `i`. This means that each random number
+    is generated independently from each other random number, so we can compute them in parallel.
+
+    If there is not enough room left in the counter to generate the desired shape of random values,
+    then a new generator is created by applying Threefry to the current key, path, and counter.
+    This new generator will have a reset counter.
+
+    Parameters
+    ----------
+    gen : Tensor[10, uint64]
+        Generator state. Can be create with :py:func:`tvm.relay.threefry_key`. This should not be
+        reused in another function, otherwise random numbers will be repeated.
+
+    out_shape : Sequence[int]
+        Output shape of the random numbers. Product of all dimensions must be a multiple of 4.
+
+    Returns
+    -------
+    new_gen : Tensor[10, uint64]
+        The new generator state to be used in subsequent calls.
+
+    rand : Tensor[out_shape, uint64]
+        Tensor of random numbers with shape `out_shape`.
+    """
+    out_len = tir.const(1)
+    for s in out_shape:
+        out_len *= s
+    assert (
+        out_len.value % 4 == 0
+    ), f"Threefry can only generate arrays who's size is a multiple of 4 ({out_len} was provided)."
+    assert (
+        out_len.value <= 2 ** 64 - 1
+    ), f"Can only generate up to 2^64 random numbers, but {out_len} were requested."
+
+    def gen_ir(gen_ptr, out_gen_ptr, out_array_ptr):
+        irb = ir_builder.create()
+        gen = irb.buffer_ptr(gen_ptr)
+        out_gen = irb.buffer_ptr(out_gen_ptr)
+        out_array = irb.buffer_ptr(out_array_ptr)
+
+        # Create a temporary array to hold the generator state we will use to create the random
+        # numbers. We cannot use gen because we may need to update the key + path if there is not
+        # enough room in the counter.
+        tmp = irb.allocate(gen.dtype, 10, name="tmp", scope="global")
+
+        # TODO(tkonolige): for now we only use the last word of the counter for counting. It is too
+        # much work to figure out how to do 128 bit addition.
+
+        # Max value for counter should be 2**64-2 because we need to reserve a special value to
+        # indicate the counter is used up.
+        with irb.if_scope(gen[7] < tir.const(2 ** 64 - 1, dtype=gen.dtype) - out_len):
+            for i in range(10):
+                tmp[i] = gen[i]
+        with irb.else_scope():
+            # no room left in the counter, we have to change the path or key
+            with irb.if_scope(gen[8] == 0 and gen[9] == 0):
+                # out of room in the path, have to generate new key
+
+                # The paper says the counter that we will be hashing should be a special value of
+                # all ones. We need to allocate some space for it because we cannot overwrite gen.
+                tmp_counter = irb.allocate(gen.dtype, 2, name="tmp_counter", scope="global")
+                tmp_counter[0] = tir.const(0xFFFFFFFFFFFFFFFF, dtype=gen.dtype)
+                tmp_counter[1] = tir.const(0xFFFFFFFFFFFFFFFF, dtype=gen.dtype)
+                _threefry(irb, gen, 0, tmp_counter, 0, tmp, 0, 1)
+                tmp[4] = tir.const(0, dtype=gen.dtype)  # zero path, i.e. no path
+                tmp[5] = tir.const(0, dtype=gen.dtype)
+                tmp[6] = tir.const(0, dtype=gen.dtype)  # zero counter
+                tmp[7] = tir.const(0, dtype=gen.dtype)
+                tmp[8] = tir.const(1 << 63, dtype=gen.dtype)  # one in the leftmost position
+                tmp[9] = tir.const(0, dtype=gen.dtype)
+            with irb.else_scope():
+                tmp[0] = gen[0]
+                tmp[1] = gen[1]
+                tmp[2] = gen[2]
+                tmp[3] = gen[3]
+                tmp[4] = gen[4] | gen[8]  # add a 1 to the path
+                tmp[5] = gen[5] | gen[9]
+                tmp[6] = tir.const(0, dtype=gen.dtype)  # zero counter
+                tmp[7] = tir.const(0, dtype=gen.dtype)
+                _shift_right(irb, gen[8], gen[9], tmp, 8, tmp, 9)
+
+        # Compute random values
+        _threefry(irb, tmp, 0, tmp, 4, out_array, 0, out_len // 4)
+
+        # Update generator state
+        out_gen[0] = tmp[0]  # key stays the same
+        out_gen[1] = tmp[1]
+        out_gen[2] = tmp[2]
+        out_gen[3] = tmp[3]
+        out_gen[4] = tmp[4]  # path stays the same
+        out_gen[5] = tmp[5]
+        out_gen[6] = tir.const(0, dtype=gen.dtype)  # unused, leave it as 0
+        out_gen[7] = tmp[7] + tir.Cast(gen.dtype, out_len)  # increment counter
+        out_gen[8] = tmp[8]  # path unchanged, so no update here
+        out_gen[9] = tmp[9]
+
+        return irb.get()
+
+    out_gen = tvm.tir.decl_buffer((10,), name="out_gen", dtype="uint64")
+    out_array = tvm.tir.decl_buffer(out_shape, name="out_array", dtype="uint64")
+    return tvm.te.extern(
+        [out_gen.shape, out_array.shape],
+        [gen],
+        lambda ins, outs: gen_ir(ins[0], outs[0], outs[1]),
+        out_buffers=[out_gen, out_array],
+        name="threefry_generate",
+        tag="threefry_generate",
+    )
+
+
+def _shift_right(irb, a, b, out_a, a_off, out_b, b_off):
+    """Binary shift a 128bit number composed of two 64 bit words right by one."""
+    with irb.if_scope(a == 1):
+        out_a[a_off] = tir.const(0, dtype=a.dtype)
+        out_b[b_off] = tir.const(0x8000000000000000, dtype=a.dtype)
+    with irb.else_scope():
+        with irb.if_scope(a == 0):
+            out_a[a_off] = tir.const(0, dtype=a.dtype)
+            out_b[b_off] = b >> 1
+        with irb.else_scope():
+            out_a[a_off] = a >> 1
+            out_b[b_off] = tir.const(0, dtype=a.dtype)
+
+
+def threefry_split(gen):
+    """Split a single generator state into two new ones
+
+    Notes
+    -----
+    The new generator is created by appending a one (for the right output) or a zero (for the left
+    output) to the end of the path portion of the generator If there is no longer and room in the
+    path, then we create a new key portion of the generator by applying Threefry to the old state,
+    path, and counter. i.e. :code:`new_key = threefry(old_key, [old_path, old_counter])`. This
+    resets the path portion of the new generator.
+
+    Parameters
+    ----------
+    gen : Tensor[10, uint64]
+        Generator state. Can be create with :py:func:`tvm.relay.threefry_key`. This should not be
+        reused in another function, otherwise random numbers will be repeated.
+
+    Returns
+    -------
+    out_gen_left : Tensor[10, uint64]
+        New generator state that is distinct from `out_gen_right`.
+
+    out_gen_right : Tensor[10, uint64]
+        New generator state that is distinct from `out_gen_left`.
+    """
+
+    def gen_ir(gen_ptr, out_left_ptr, out_right_ptr):
+        irb = ir_builder.create()
+        gen = irb.buffer_ptr(gen_ptr)
+        out_left = irb.buffer_ptr(out_left_ptr)
+        out_right = irb.buffer_ptr(out_right_ptr)
+
+        with irb.if_scope(gen[8] == 0 and gen[9] == 0):
+            # Generate new key because we have run out of room to extend the path
+            _threefry(irb, gen, 0, gen, 4, out_left, 0, 1)
+            out_left[4] = tir.const(0, dtype=gen.dtype)
+            out_left[5] = tir.const(0, dtype=gen.dtype)
+            out_left[6] = tir.const(0, dtype=gen.dtype)  # counter gets zeroed
+            out_left[7] = tir.const(0, dtype=gen.dtype)  # counter gets zeroed
+            out_left[8] = tir.const(
+                1 << 62, dtype=gen.dtype
+            )  # one in the second from the leftmost position
+            out_left[9] = tir.const(0, dtype=gen.dtype)
+
+            out_right[0] = out_left[0]
+            out_right[1] = out_left[1]
+            out_right[2] = out_left[2]
+            out_right[3] = out_left[3]
+            out_right[4] = tir.const(1 << 63, dtype=gen.dtype)  # one in the leftmost position
+            out_right[5] = tir.const(0, dtype=gen.dtype)
+            out_right[6] = tir.const(0, dtype=gen.dtype)
+            out_right[7] = tir.const(0, dtype=gen.dtype)
+            out_right[8] = tir.const(
+                1 << 62, dtype=gen.dtype
+            )  # one in the second from the leftmost position
+            out_right[9] = tir.const(0, dtype=gen.dtype)
+        with irb.else_scope():
+            out_left[0] = gen[0]
+            out_left[1] = gen[1]
+            out_left[2] = gen[2]
+            out_left[3] = gen[3]
+            out_left[4] = gen[4]  # adding a zero here, but its already zero padded
+            out_left[5] = gen[5]
+            out_left[6] = gen[6]
+            out_left[7] = gen[7]
+            # move path position over one bit
+            _shift_right(irb, gen[8], gen[9], out_left, 8, out_left, 9)
+
+            out_right[0] = gen[0]
+            out_right[1] = gen[1]
+            out_right[2] = gen[2]
+            out_right[3] = gen[3]
+            out_right[4] = gen[4] | gen[8]  # add a one to the path
+            out_right[5] = gen[5] | gen[9]
+            out_right[6] = gen[6]
+            out_right[7] = gen[7]
+            _shift_right(irb, gen[8], gen[9], out_right, 8, out_right, 9)
+
+        return irb.get()
+
+    out_left = tvm.tir.decl_buffer((10,), name="out_left", dtype="uint64")
+    out_right = tvm.tir.decl_buffer((10,), name="out_right", dtype="uint64")
+    return tvm.te.extern(
+        [out_left.shape, out_right.shape],
+        [gen],
+        lambda ins, outs: gen_ir(ins[0], outs[0], outs[1]),
+        out_buffers=[out_left, out_right],
+        name="threefry_split",
+        tag="threefry_split",
+    )
diff --git a/src/relay/op/random/kernel.cc b/src/relay/op/random/kernel.cc
new file mode 100644
index 0000000..ec092a7
--- /dev/null
+++ b/src/relay/op/random/kernel.cc
@@ -0,0 +1,89 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied.  See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+#include <tvm/relay/attrs/random.h>
+#include <tvm/relay/op.h>
+
+namespace tvm {
+namespace relay {
+
+TVM_REGISTER_NODE_TYPE(ThreefryGenerateAttrs);
+
+static TensorType ThreefryKeyType() { return TensorType({10}, tvm::DataType::UInt(64)); }
+
+bool ThreefryGenerateRel(const Array<Type>& types, int num_inputs, const Attrs& attrs,
+                         const TypeReporter& reporter) {
+  const ThreefryGenerateAttrs* param = attrs.as<ThreefryGenerateAttrs>();
+  ICHECK_EQ(types.size(), 2) << "ThreefryGenerate should have one input and one output";
+
+  reporter->Assign(types[0], ThreefryKeyType());
+
+  std::vector<IndexExpr> oshape;
+  for (auto& x : param->out_shape) {
+    oshape.push_back(x);
+  }
+  // generate returns the next key and an array of random values
+  // TODO(@tkonolige, @altanh): support other output dtypes?
+  reporter->Assign(types[1],
+                   TupleType({ThreefryKeyType(), TensorType(oshape, tvm::DataType::UInt(64))}));
+  return true;
+}
+
+Expr MakeThreefryGenerate(Expr key, Array<Integer> out_shape) {
+  auto attrs = make_object<ThreefryGenerateAttrs>();
+  attrs->out_shape = out_shape;
+  static const Op& op = Op::Get("random.threefry_generate");
+  return Call(op, {key}, Attrs(attrs), {});
+}
+
+TVM_REGISTER_GLOBAL("relay.op.random._make.threefry_generate").set_body_typed(MakeThreefryGenerate);
+
+RELAY_REGISTER_OP("random.threefry_generate")
+    .describe(
+        R"doc(Generate an array of random numbers using the Threefry algorithm.)doc" TVM_ADD_FILELINE)
+    .set_num_inputs(1)
+    .set_attrs_type<ThreefryGenerateAttrs>()
+    .add_argument("key", "Tensor", "Input Threefry key")
+    .add_type_rel("ThreefryGenerate", ThreefryGenerateRel);
+
+bool ThreefrySplitRel(const Array<Type>& types, int num_inputs, const Attrs& attrs,
+                      const TypeReporter& reporter) {
+  ICHECK_EQ(types.size(), 2) << "ThreefrySplit should have one input and one output";
+
+  reporter->Assign(types[0], ThreefryKeyType());
+  reporter->Assign(types[1], TupleType({ThreefryKeyType(), ThreefryKeyType()}));
+
+  return true;
+}
+
+Expr MakeThreefrySplit(Expr key) {
+  static const Op& op = Op::Get("random.threefry_split");
+  return Call(op, {key}, Attrs(), {});
+}
+
+TVM_REGISTER_GLOBAL("relay.op.random._make.threefry_split").set_body_typed(MakeThreefrySplit);
+
+RELAY_REGISTER_OP("random.threefry_split")
+    .describe(R"doc(Split the input Threefry key into two new ones.)doc" TVM_ADD_FILELINE)
+    .set_num_inputs(1)
+    .add_argument("key", "Tensor", "Input Threefry key")
+    .add_type_rel("ThreefrySplit", ThreefrySplitRel);
+
+}  // namespace relay
+}  // namespace tvm
diff --git a/tests/python/relay/test_prng.py b/tests/python/relay/test_prng.py
new file mode 100644
index 0000000..2109d3b
--- /dev/null
+++ b/tests/python/relay/test_prng.py
@@ -0,0 +1,142 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+import pytest
+import tvm
+import tvm.relay
+import tvm.testing
+from tvm.relay.testing import run_infer_type
+
+
+@tvm.testing.parametrize_targets
+def test_threefry_repeatability(target, ctx):
+    target, ctx = "llvm", tvm.cpu(0)
+    key1 = tvm.relay.random.threefry_key(1)
+    rand1 = tvm.relay.random.threefry_generate(key1, (12,))
+    out_key1, out1 = tvm.relay.create_executor(
+        "vm", tvm.IRModule.from_expr(tvm.relay.Function([], rand1)), target=target, ctx=ctx
+    ).evaluate()()
+
+    key2 = tvm.relay.random.threefry_key(1)
+    rand2 = tvm.relay.random.threefry_generate(key2, (12,))
+    out_key2, out2 = tvm.relay.create_executor(
+        "vm", tvm.IRModule.from_expr(tvm.relay.Function([], rand2)), target=target, ctx=ctx
+    ).evaluate()()
+
+    assert (
+        out1.asnumpy() == out2.asnumpy()
+    ).all(), "Generate on same seed should have the same output random numbers"
+
+    assert (
+        out_key1.asnumpy() == out_key2.asnumpy()
+    ).all(), "Generate on same seed should have the same next keys"
+
+
+@tvm.testing.parametrize_targets
+def test_threefry_split(target, ctx):
+    key = tvm.relay.random.threefry_key(1)
+    left, right = tvm.relay.TupleWrapper(tvm.relay.random.threefry_split(key), 2)
+    _, rand1 = tvm.relay.TupleWrapper(tvm.relay.random.threefry_generate(left, (16,)), 2)
+    _, rand2 = tvm.relay.TupleWrapper(tvm.relay.random.threefry_generate(right, (16,)), 2)
+    out1, out2 = tvm.relay.create_executor(
+        "vm",
+        tvm.IRModule.from_expr(tvm.relay.Function([], tvm.relay.Tuple((rand1, rand2)))),
+        target=target,
+        ctx=ctx,
+    ).evaluate()()
+
+    assert (
+        out1.asnumpy() != out2.asnumpy()
+    ).any(), "Generate after split should not have the same output"
+
+
+@tvm.testing.parametrize_targets
+def test_threefry_sequential_generate(target, ctx):
+    key = tvm.relay.random.threefry_key(1)
+    key, rand1 = tvm.relay.TupleWrapper(tvm.relay.random.threefry_generate(key, (4,)), 2)
+    _, rand2 = tvm.relay.TupleWrapper(tvm.relay.random.threefry_generate(key, (4,)), 2)
+    out1, out2 = tvm.relay.create_executor(
+        "vm",
+        tvm.IRModule.from_expr(tvm.relay.Function([], tvm.relay.Tuple((rand1, rand2)))),
+        target=target,
+        ctx=ctx,
+    ).evaluate()()
+
+    assert (
+        out1.asnumpy() != out2.asnumpy()
+    ).any(), "Sequential generates should not have the same output"
+
+
+def test_threefry_generate_infer():
+    oshape = (12,)
+    key_type = tvm.relay.TensorType([10], dtype="uint64")
+    gen_type = tvm.relay.TensorType(oshape, dtype="uint64")
+    expected_type = tvm.relay.TupleType([key_type, gen_type])
+
+    key = tvm.relay.random.threefry_key(1)
+    rand1 = tvm.relay.random.threefry_generate(key, oshape)
+    f = tvm.relay.Function([], rand1)
+    f = run_infer_type(f)
+    assert tvm.ir.structural_equal(f.ret_type, expected_type)
+
+
+def test_threefry_split_infer():
+    key_type = tvm.relay.TensorType([10], dtype="uint64")
+    expected_type = tvm.relay.TupleType([key_type, key_type])
+
+    key = tvm.relay.random.threefry_key(1)
+    out_keys = tvm.relay.random.threefry_split(key)
+    f = tvm.relay.Function([], out_keys)
+    f = run_infer_type(f)
+    assert tvm.ir.structural_equal(f.ret_type, expected_type)
+
+
+@pytest.mark.xfail(raises=tvm.error.TVMError)
+def test_threefry_generate_infer_fail():
+    # xfail: key size should be 10
+    fake_key = tvm.relay.const([1, 2, 3, 4, 5, 6, 7, 8, 9], dtype="uint64")
+    rand1 = tvm.relay.random.threefry_generate(fake_key, (12,))
+    f = tvm.relay.Function([], rand1)
+    f = run_infer_type(f)
+
+
+@pytest.mark.xfail(raises=tvm.error.TVMError)
+def test_threefry_split_infer_fail():
+    # xfail: key size should be 10
+    fake_key = tvm.relay.const([1, 2, 3, 4, 5, 6, 7, 8, 9], dtype="uint64")
+    out_keys = tvm.relay.random.threefry_split(fake_key)
+    f = tvm.relay.Function([], out_keys)
+    f = run_infer_type(f)
+
+
+@tvm.testing.requires_llvm
+@pytest.mark.xfail(raises=tvm.error.TVMError)
+def test_threefry_generate_incorrect_out_size():
+    key = tvm.relay.random.threefry_key(1)
+    # xfail: output size should be multiple of 4
+    key, rand1 = tvm.relay.TupleWrapper(tvm.relay.random.threefry_generate(key, (5,)), 2)
+    out1, out2 = tvm.relay.create_executor(
+        "vm",
+        tvm.IRModule.from_expr(tvm.relay.Function([], rand1)),
+        target=tvm.target.Target("llvm"),
+        ctx=tvm.context("cpu"),
+    ).evaluate()()
+
+
+if __name__ == "__main__":
+    test_threefry_repeatability(tvm.target.Target("llvm"), tvm.context("cpu"))
+    test_threefry_split(tvm.target.Target("llvm"), tvm.context("cpu"))
+    test_threefry_sequential_generate(tvm.target.Target("llvm"), tvm.context("cpu"))
diff --git a/tests/python/topi/python/test_topi_prng.py b/tests/python/topi/python/test_topi_prng.py
new file mode 100644
index 0000000..43b0494
--- /dev/null
+++ b/tests/python/topi/python/test_topi_prng.py
@@ -0,0 +1,116 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+import tvm
+import tvm.relay
+import tvm.testing
+import tvm.topi
+import numpy as np
+
+
+def threefry_split(target, ctx, gen):
+    gen_placeholder = tvm.te.placeholder(gen.shape, name="gen", dtype="uint64")
+    left_placeholder, right_placeholder = tvm.topi.random.threefry_split(gen_placeholder)
+    s = tvm.topi.generic.schedule_extern([left_placeholder, right_placeholder])
+    f = tvm.build(s, [gen_placeholder, left_placeholder, right_placeholder])
+    left = tvm.nd.array(np.zeros(gen.shape, dtype="uint64"))
+    right = tvm.nd.array(np.zeros(gen.shape, dtype="uint64"))
+    f(tvm.nd.array(gen), left, right)
+    return left.asnumpy(), right.asnumpy()
+
+
+def threefry_generate(target, ctx, gen, size):
+    gen_placeholder = tvm.te.placeholder(gen.shape, name="gen", dtype="uint64")
+    left_placeholder, right_placeholder = tvm.topi.random.threefry_generate(gen_placeholder, size)
+    s = tvm.topi.generic.schedule_extern([left_placeholder, right_placeholder])
+    f = tvm.build(s, [gen_placeholder, left_placeholder, right_placeholder])
+    out_gen = tvm.nd.array(np.zeros(gen.shape, dtype="uint64"))
+    rands = tvm.nd.array(np.zeros(size, dtype="uint64"))
+    f(tvm.nd.array(gen), out_gen, rands)
+    return out_gen.asnumpy(), rands.asnumpy()
+
+
+@tvm.testing.parametrize_targets
+def test_threefry_split(target, ctx):
+    # test that results of split do not equal eachother or the input
+    gen = tvm.relay.random.threefry_key(0).data.asnumpy()
+    a, b = threefry_split(target, ctx, gen)
+    assert (a != b).any() and (
+        a != gen
+    ).any(), "Splitting a gen should result in different output gens"
+    # unittest some split inputs
+    assert (a == np.array([0, 0, 0, 0, 0, 0, 0, 0, 1 << 62, 0], dtype="uint64")).all()
+    assert (b == np.array([0, 0, 0, 0, 1 << 63, 0, 0, 0, 1 << 62, 0], dtype="uint64")).all()
+
+    # test enough splits to go over path length
+    for i in range(129):
+        a, b = threefry_split(target, ctx, b)
+    assert (a[0:4] == b[0:4]).all(), "State part of split should be the same"
+    assert (b[0:4] != np.zeros(4, dtype="uint64")).any()
+
+    # check that split then generate does not generate the same for both sides
+    a, a_rands = threefry_generate(target, ctx, a, (100,))
+    b, b_rands = threefry_generate(target, ctx, b, (100,))
+    assert (
+        a_rands != b_rands
+    ).all(), "Numbers generated from different initial states should be different"
+
+    # check repeatability
+    _, rands1 = threefry_generate(target, ctx, a, (100,))
+    _, rands2 = threefry_generate(target, ctx, a, (100,))
+    assert (
+        rands1 == rands2
+    ).all(), "Numbers generated from the same initial state should be the same"
+
+    a1, b1 = threefry_split(target, ctx, a)
+    a2, b2 = threefry_split(target, ctx, a)
+    assert (a1 == a2).all() and (
+        b1 == b2
+    ).all(), "Split called on the same input should return the same result"
+
+
+@tvm.testing.parametrize_targets
+def test_threefry_generate(target, ctx):
+    gen = tvm.relay.random.threefry_key(0).data.asnumpy()
+
+    # check that we can generate some data
+    a, rands = threefry_generate(target, ctx, gen, (100,))
+    assert (
+        rands.shape[0] == 100 and len(rands.shape) == 1
+    ), "Output shape should match requested shape"
+
+    # check that gen out does not equal input
+    assert (a != gen).any(), "Output generator should be different from input generator"
+
+    # test enough generates to go over generate limit
+    gen = np.array(
+        [0, 0, 0, 0, 0, 0, 0, 2 ** 64 - 2, 1 << 63, 0], dtype="uint64"
+    )  # make counter large
+    a, rands = threefry_generate(target, ctx, gen, (100,))
+    assert gen[4] != a[4], "Overflow of counter should trigger path change"
+    assert a[7] == 100, "Overflow of counter should still update counter"
+
+    # check generate with path at length limit
+    gen = np.array([0, 0, 0, 0, 0, 0, 0, 2 ** 64 - 2, 0, 0], dtype="uint64")  # make counter large
+    a, rands = threefry_generate(target, ctx, gen, (100,))
+    assert (
+        gen[0:4] != a[0:4]
+    ).any(), "Overflowing counter with no space left in path should change state"
+
+
+if __name__ == "__main__":
+    test_threefry_split(tvm.target.Target("llvm"), tvm.context("cpu"))
+    test_threefry_generate(tvm.target.Target("llvm"), tvm.context("cpu"))