Data Definition

Description

Defines a Tensor variable.

Prototype

Tensor(dtype, shape, scope, name, enable_buffer_reuse=False, no_reuse_list=None, reuse_list=None, is_workspace=False, is_atomic_add=False)

Parameters

Restrictions

1. When the total size of the tensors exceeds the total size of the corresponding buffer type, a build error is reported.

   In the following example, the size of data_a is 1025 x 1024 bytes, which is greater than the total size of L1 Buffer by 1 MB.

   import numpy as np
   import sys
   from te import tik
   import tvm
   
   def buffer_allocate_test6():
       tik_instance = tik.Tik()
       data_a = tik_instance.Tensor("int8", (1025 * 1024,), name="data_a", scope=tik.scope_cbuf)
       tik_instance.BuildCCE(kernel_name="buffer_allocate_test",inputs=[],outputs=[])
       return tik_instance
   
   if __name__ == "__main__":
       tik_instance = buffer_allocate_test6()

   Build error:

   RuntimeError: Appiled buffer size(1049600B) more than avaiable buffer size(1048576B).

2. If a tensor is access beyond its defined scope, a build error will be reported.
   In the following example, data_a_l1 is defined only in new_stmt_scope. Beyond the defined scope, an error will be reported when the data_move API is called to access data_a_l1 again.

   import numpy as np
   import sys
   from te import tik
   import tvm
   
   def tensor_outrange_examine_test6():
       tik_instance = tik.Tik()
       data_a = tik_instance.Tensor("float16", (128,), name="data_a", scope=tik.scope_gm)
       data_b = tik_instance.Tensor("float16", (128,), name="data_b", scope=tik.scope_gm)
       with tik_instance.new_stmt_scope():
           data_a_ub = tik_instance.Tensor("float16", (128,), name="data_a_ub", scope=tik.scope_ubuf)
           data_a_l1 = tik_instance.Tensor("float16", (128,), name="data_a_l1", scope=tik.scope_cbuf)
       tik_instance.data_move(data_a_l1, data_a, 0, 1, 128 // 16, 0, 0)
       tik_instance.data_move(data_a_ub, data_a_l1, 0, 1, 128 // 16, 0, 0)
       tik_instance.data_move(data_b, data_a_ub, 0, 1, 128 // 16, 0, 0)
       tik_instance.BuildCCE(kernel_name="tensor_outrange_examine", inputs=[data_a], outputs=[data_b])
       return tik_instance
   
   

   Build error:

   RuntimeError: This tensor is not defined in this scope.

3. If a tensor is beyond its defined scope, the buffer can be reused.
   In the following example, as data_a_ub1 and data_a_ub2 are beyond the defined scopes, the occupied buffer of size 126,976 bytes (62 x 2 x 1024 bytes) can be reused by data_b_ub.

   import numpy as np
   import sys
   from te import tik
   import tvm
   
   def double_buffer_test6():
       tik_instance = tik.Tik()
       data_a = tik_instance.Tensor("int8", (124 *1024,), name="data_a", scope=tik.scope_ubuf)
       with tik_instance.for_range(0, 2):
           data_a_ub1 = tik_instance.Tensor("int8", (62 * 1024,), name="data_a_ub1", scope=tik.scope_ubuf)
           data_a_ub2 = tik_instance.Tensor("int8", (62 * 1024,), name="data_a_ub2", scope=tik.scope_ubuf)
       data_b_ub = tik_instance.Tensor("int8", (125 * 1024,), name="data_b_ub", scope=tik.scope_ubuf)
       tik_instance.BuildCCE(kernel_name="tbe_double_buffer_no_loop", inputs=[ ], outputs=[ ])
       return tik_instance
   if __name__ == "__main__":
       tik_instance = double_buffer_test6()

   If data_b_ub exceeds the Unified Buffer size, the following error is reported during the build:

   RuntimeError: Tensor data_b_ub appiles buffer size(128000B) more than avaiable buffer size(126976B).

4. shape does not support scalar variables. It supports only immediates or Python variables.
5. For user-defined tensors, the starting address of the allocated buffer scope will be aligned according to General Restrictions.

   If the total size of a buffer type is exceeded due to address alignment, a build error is reported.

Returns

Tensor instance

Example

from te import tik
tik_instance = tik.Tik()
data_A = tik_instance.Tensor("float16", (128,), name="data_A", scope=tik.scope_gm)

Description

Defines a Scalar variable.

Prototype

Scalar(dtype="int64", name="reg_buf", init_value=None)

Parameters

Restrictions

When the initial value is an Expr, its immediate can only be of integer type instead of float, for example:

from te import tik
tik_instance = tik.Tik()
index_reg = tik_instance.Scalar(dtype="float32")
# Immediate: float
index_reg.set_as(10.2)

# Assign an initial value to the scalar using init_value.
index_reg1 = tik_instance.Scalar(dtype="float32", init_value=10.2)
 
index_reg2 = tik_instance.Scalar(dtype="float32")
# Expr. The immediate is of float type. An error occurs with the CCE compiler (CCEC), because the hardware does not support this data type.
#index_reg2.set_as(index_reg+2.2); 

Returns

Scalar instance

Example

from te import tik
tik_instance = tik.Tik()
# Immediate: int
index_reg = tik_instance.Scalar(dtype = "int32")
index_reg.set_as(10)   

# Immediate: float
index_reg2 = tik_instance.Scalar(dtype = "float16")
index_reg2.set_as(10.2) 

# A Scalar variable
index_reg3 = tik_instance.Scalar(dtype = "float16")
index_reg3.set_as(index_reg2)  

# Tensor value
data_A = tik_instance.Tensor("float16", (128,), name="data_A", scope=tik.scope_gm)
index_reg3.set_as(data_A[0])// tensor value

#An Expr
index_reg4 = tik_instance.Scalar(dtype = "int32")
index_reg4.set_as(index_reg+20) 

Description

Defines an InputScalar variable. An InputScalar serves as an inputs argument passed to the BuildCCE call. It supports a range of basic data types including int, uint, and float.

Prototype

InputScalar(dtype="int64", name="input_scalar")

Parameters

Restrictions

1. Currently, InputScalar can be used in scenarios where a variable is an Expr.

   For example, if repeat_times in vec_abs can be an Expr, the code can be as follows:
       from te import tik
       tik_instance = tik.Tik()
       data_A = tik_instance.Tensor("float16", (128,), name="data_A", scope=tik.scope_gm)
       data_B = tik_instance.Tensor("float16", (128,), name="data_B", scope=tik.scope_gm)
       src_ub = tik_instance.Tensor("float16", (128,), name="src_ub", scope=tik.scope_ubuf)
       dst_ub = tik_instance.Tensor("float16", (128,), name="dst_ub", scope=tik.scope_ubuf)
       inputscalar = tik_instance.InputScalar(dtype="int16", name="inputscalar")
       tik_instance.vec_abs(128, dst_ub, src_ub, inputscalar, 8, 8)
       tik_instance.BuildCCE(kernel_name="simple_add",inputs=[data_A,data_B,inputscalar],outputs=[])

2. Ensure that each InputScalar object has a unique name.

Example

from te import tik
tik_instance = tik.Tik()
data_A = tik_instance.Tensor("float16", (128,), name="data_A", scope=tik.scope_gm)
data_B = tik_instance.Tensor("float16", (128,), name="data_B", scope=tik.scope_gm)
abc = tik_instance.InputScalar(dtype="int16", name="abc")
src_ub = tik_instance.Tensor("float16", (128,), name="src_ub", scope=tik.scope_ubuf)
dst_ub = tik_instance.Tensor("float16", (128,), name="dst_ub", scope=tik.scope_ubuf)
tik_instance.vec_abs(128, dst_ub, src_ub, abc, 8, 8)
tik_instance.BuildCCE(kernel_name="simple_add",inputs=[data_A,data_B,abc],outputs=[])