Formalise the description of operator argument types

- Standardise the terminology for operator arguments. Argument,
  Operand, and Parameter were used interchangeably.
- Introduce a templatized tensor_t<> type for tensor arguments.
  Scalars are represented by rank-0 tensors.
- Types can be checked with the XSD schema.

Signed-off-by: Kevin Petit <kevin.petit@arm.com>
Change-Id: Ic57b9387950824e994c5e7f9ec1489c29159b974

7 files changed, 356 insertions, 268 deletions

diff --git a/chapters/custom.adoc b/chapters/custom.adoc
index 4804e25..e748f38 100644
--- a/chapters/custom.adoc
+++ b/chapters/custom.adoc
@@ -12,9 +12,9 @@
 Hardware implementing TOSA may choose to add additional custom operators that are not expressed in the existing TOSA operations. These operators are not expected to be portable across TOSA implementations. The input and output signatures must be expressed in the corresponding TOSA node.
 
 ==== CUSTOM
-Input Operands:
+Input arguments:
 
-* Num input operands – Scalar number of input operands
-* Num output operands – Scalar number of output operands
+* Num input arguments – Scalar number of input arguments
+* Num output arguments – Scalar number of output arguments
 * Operator code – untyped data consisting of the operator data
 * Affine transform description for each tensor
diff --git a/chapters/introduction.adoc b/chapters/introduction.adoc
index 9cfccb7..b369070 100644
--- a/chapters/introduction.adoc
+++ b/chapters/introduction.adoc
@@ -79,7 +79,7 @@ The following principles govern the selection of operators within TOSA.
 |Consistent sub-operation definition reduces the operator implementation cost.
 
 |P4
-|The valid input and output ranges for all operands shall be specified.
+|The valid input and output ranges for all arguments shall be specified.
 |Ranges are required to make consistent (numerically agreeing) implementations possible.
 
 |P5
@@ -108,11 +108,11 @@ The following table summarizes the three profiles:
 
 === Levels
 
-A TOSA level defines operator parameter ranges that an implementation shall support.
+A TOSA level defines operator argument ranges that an implementation shall support.
 This is distinct from a profile that defines the operations and data-types supported.
 This version of the specification defines two TOSA levels:
 
-* No level : allows the full range of parameters specified by the operations according to the operation data types.
+* No level : allows the full range of arguments specified by the operations according to the operation data types.
 * Level 8K : ranges are expected to be sufficient for applications with frame sizes up to 8K.
 
 Later versions of the specification may define additional levels.
@@ -522,7 +522,7 @@ To convert a network containing quantized tensors to TOSA, generate explicit RES
 This reduces quantized operations to purely integer operations.
 
 As an example, an ADD between two quantized tensors requires the integer values represent the same range.
-The scale parameters for RESCALE can be calculated to ensure that the resulting tensors represent the same range.
+The scale arguments for RESCALE can be calculated to ensure that the resulting tensors represent the same range.
 Then the ADD is performed, and a RESCALE can be used to ensure that the result is scaled properly.
 
 RESCALE provides support for per-tensor and per-channel scaling values to ensure compatibility with a range of possible quantization implementations.
diff --git a/chapters/operators.adoc b/chapters/operators.adoc
index 897ff17..3a4c831 100644
--- a/chapters/operators.adoc
+++ b/chapters/operators.adoc
@@ -9,14 +9,23 @@
 
 == Operators
 
-=== Operator Parameters
+=== Operator Arguments
 
-An operator processes input operands to produce output operands. An operator can have three types of parameters:
+Operators process input arguments to produce output arguments.
+Their behavior can be configured using attribute arguments.
+Arguments may have one of the following types:
 
-* An input operand. This must be a tensor or a list of tensors and data is read by the operation.
-* An output operand. This must be a tensor or a list of tensors and data is written by the operation.
-* An attribute. This is a parameter that is constant for a particular instance of the operator. It may have any data type supported by TOSA. It is expected to be set at compile time.
+* `tensor_t<element_type>`, abbreviated `T<element_type>`, represents a tensor whose elements are of type `element_type` where `element_type` can be any of the data types supported in TOSA.
+* `tensor_list_t` represents a list of tensors. When lists are homogeneous, i.e. contain tensors of the same type, their type is further qualified as follows: `tensor_list_t<T<element_type>>`.
+* `tosa_graph_t` represents a TOSA graph (see <<operator-graphs>>).
 
+Arguments belong to one of three categories: Input, Output, or Attribute. The category to which an argument belongs further constrains its type:
+
+* An Input argument must be a tensor or a list of tensors used to provide the data read by the operation.
+* An Output argument must be a tensor or a list of tensors into which the data produced by the operation is written.
+* An Attribute argument is constant, i.e. its value is known at compilation time. It may have any data type supported by TOSA.
+
+[[operator-graphs]]
 === Operator Graphs
 
 A TOSA graph is a collection of TOSA operators where:
@@ -27,7 +36,7 @@ A TOSA graph is a collection of TOSA operators where:
 * The attributes must be in the valid range permitted for the operator
 * The tensor dimensions must be in the valid range permitted for the operator
 
-Some operators, such as control flow operators, take a graph of other operators as an attribute. The type tosa_graph_t will denote a graph of operators and the following functions define the tensor shape list for the graph input and outputs:
+Some operators, such as control flow operators, take a graph of other operators as an attribute. The type `tosa_graph_t` will denote a graph of operators and the following functions define the tensor shape list for the graph input and outputs:
 
 [source,c++]
 ----
diff --git a/tools/genspec.py b/tools/genspec.py
index c64f05b..11a3e72 100755
--- a/tools/genspec.py
+++ b/tools/genspec.py
@@ -18,10 +18,11 @@ class TOSASpecAsciidocGenerator:
 
     def generate_operator(self, op, file):
         file.write("\n*Arguments:*\n")
-        file.write("[cols='2,1,1,1,2,4']")
+        file.write("[cols='3,3,2,2,4,8']")
         file.write("\n|===\n")
         file.write("|Argument|Type|Name|Shape|Rank|Description\n\n")
         for arg in op.arguments:
+            # Argument
             cats = arg.categories
             if len(cats) > 1:
                 cattext = ""
@@ -33,6 +34,19 @@ class TOSASpecAsciidocGenerator:
                     sep = " "
             else:
                 cattext = cats[0].name.title()
+
+            # Type
+            if arg.type == 'tensor_t':
+                argtype = 'T<{}>'.format(arg.tensor_element_type)
+            elif arg.type == 'tensor_list_t':
+                if arg.tensor_element_type == '-':
+                    argtype = 'tensor_list_t'
+                else:
+                    argtype = 'tensor_list_t<T<{}>>'.format(arg.tensor_element_type)
+            else:
+                argtype = arg.type
+
+            # Rank
             if len(arg.rank) > 0:
                 if (arg.rank[0] == arg.rank[1]):
                     rank = f'{arg.rank[0]}'
@@ -40,9 +54,12 @@ class TOSASpecAsciidocGenerator:
                     rank = f'{arg.rank[0]} to {arg.rank[1]}'
             else:
                 rank = ""
+
+            # Format and write line
             file.write(
-                f"|{cattext}|{arg.type}|{arg.name}|{arg.shape}|{rank}|{arg.description}\n"
+                f"|{cattext}|{argtype}|{arg.name}|{arg.shape}|{rank}|{arg.description}\n"
             )
+
         file.write("|===\n")
         if op.typesupports:
             file.write("\n*Supported Data Types:*\n\n")
diff --git a/tools/tosa.py b/tools/tosa.py
index 218412f..d01d9c2 100644
--- a/tools/tosa.py
+++ b/tools/tosa.py
@@ -36,11 +36,12 @@ class TOSALevel:
         self.maximums = maximums
 
 class TOSAOperatorArgument:
-    def __init__(self, name, description, categories, ty, shape, levellimits, rank):
+    def __init__(self, name, description, categories, ty, elty, shape, levellimits, rank):
         self.name = name
         self.description = description
         self.categories = categories
         self.type = ty
+        self.tensor_element_type = elty
         self.shape = shape
         self.levellimits = levellimits
         self.rank = rank
@@ -144,14 +145,15 @@ class TOSASpec:
         desc = arg.find("description").text.strip()
         argcats = []
         argtype = arg.get("type")
+        argtelty = arg.get("tensor-element-type")
         shape = arg.get("shape")
         levellimits = []
         rank = []
         r = arg.find("rank")
         if r != None:
-            if shape == "-":
-                raise RuntimeError(f"rank is not empty, but shape is '-' for {op_name}: {name}")
             rank = [r.get('min'),r.get('max')]
+            if shape == "-" and (rank[0] != '0' or rank[1] != '0'):
+                raise RuntimeError(f"rank is not empty or non-zero, but shape is '-' for {op_name}: {name}")
             # validate rank against the shape argument
             (shape_check, shape_rank) = get_rank_from_shape(shape)
             if shape_check and (shape_rank < int(rank[0]) or shape_rank > int(rank[1])):
@@ -170,7 +172,7 @@ class TOSASpec:
         for cat in cats:
             argcats.append(TOSAOperatorArgumentCategory(cat[0], cat[1].split(",")))
 
-        return TOSAOperatorArgument(name, desc, argcats, argtype, shape, levellimits, rank)
+        return TOSAOperatorArgument(name, desc, argcats, argtype, argtelty, shape, levellimits, rank)
 
     def __load_enum(self, arg):
         name = arg.get("name")
diff --git a/tosa.xml b/tosa.xml
index 430f5fd..678c6ac 100644
--- a/tosa.xml
+++ b/tosa.xml
@@ -16,15 +16,16 @@
       <operator>
         <name>ARGMAX</name>
         <arguments>
-          <argument category="input" name="input" type="in_t*" shape="shape1">
+          <argument category="input" name="input" type="tensor_t" shape="shape1" tensor-element-type="in_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape1)" limit="MAX_RANK"/>
             <rank min="1" max="MAX_RANK"/>
           </argument>
-          <argument category="attribute" name="axis" type="int32_t" shape="-">
+          <argument category="attribute" name="axis" type="tensor_t" shape="-" tensor-element-type="int32_t">
             <description>Axis in range from 0 to rank(shape1) - 1</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="out_t">
             <description>Output tensor, with rank = rank(shape1) - 1</description>
             <rank min="0" max="MAX_RANK - 1"/>
           </argument>
@@ -51,23 +52,23 @@
       <operator>
         <name>AVG_POOL2D</name>
         <arguments>
-          <argument category="input" name="input" type="in_out_t*" shape="[N,IH,IW,C]">
+          <argument category="input" name="input" type="tensor_t" shape="[N,IH,IW,C]" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="4" max="4"/>
           </argument>
-          <argument category="attribute" name="kernel" type="int32_t*" shape="[2]">
+          <argument category="attribute" name="kernel" type="tensor_t" shape="[2]" tensor-element-type="int32_t">
             <description>[kernel_y, kernel_x]</description>
             <levellimit value="kernel_y" limit="MAX_KERNEL"/>
             <levellimit value="kernel_x" limit="MAX_KERNEL"/>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="stride" type="int32_t*" shape="[2]">
+          <argument category="attribute" name="stride" type="tensor_t" shape="[2]" tensor-element-type="int32_t">
             <description>[stride_y, stride_x]</description>
             <levellimit value="stride_y" limit="MAX_STRIDE"/>
             <levellimit value="stride_x" limit="MAX_STRIDE"/>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="pad" type="int32_t*" shape="[4]">
+          <argument category="attribute" name="pad" type="tensor_t" shape="[4]" tensor-element-type="int32_t">
             <description>[pad_top, pad_bottom, pad_left, pad_right]</description>
             <levellimit value="pad_top"    limit="MAX_KERNEL"/>
             <levellimit value="pad_bottom" limit="MAX_KERNEL"/>
@@ -75,16 +76,19 @@
             <levellimit value="pad_right"  limit="MAX_KERNEL"/>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="acc_size" type="acc_size_t" shape="-">
+          <argument category="attribute" name="acc_size" type="tensor_t" shape="-" tensor-element-type="acc_size_t">
             <description>Enumerated type, must be one of INT32, FP16, FP32, as defined in the Supported Data Types table for this operation</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="attribute" name="input_zp" type="in_out_t" shape="-">
+          <argument category="attribute" name="input_zp" type="tensor_t" shape="-" tensor-element-type="in_out_t">
             <description>Input tensor zero point. Must be zero for non-int8 types.</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="attribute" name="output_zp" type="in_out_t" shape="-">
+          <argument category="attribute" name="output_zp" type="tensor_t" shape="-" tensor-element-type="in_out_t">
             <description>Output tensor zero point. Must be zero for non-int8 types.</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="[N,OH,OW,C]">
+          <argument category="output" name="output" type="tensor_t" shape="[N,OH,OW,C]" tensor-element-type="in_out_t">
             <description>Output tensor 4D</description>
             <rank min="4" max="4"/>
           </argument>
@@ -115,21 +119,21 @@
       <operator>
         <name>CONV2D</name>
         <arguments>
-          <argument category="input" name="input" type="in_t*" shape="[N,IH,IW,IC]">
+          <argument category="input" name="input" type="tensor_t" shape="[N,IH,IW,IC]" tensor-element-type="in_t">
             <description>Input tensor</description>
             <rank min="4" max="4"/>
           </argument>
-          <argument category="input" name="weight" type="weight_t*" shape="[OC,KH,KW,IC]">
+          <argument category="input" name="weight" type="tensor_t" shape="[OC,KH,KW,IC]" tensor-element-type="weight_t">
             <description>Weight kernel size KH x KW</description>
             <levellimit value="dilation_y * KH" limit="MAX_KERNEL"/>
             <levellimit value="dilation_x * KW" limit="MAX_KERNEL"/>
             <rank min="4" max="4"/>
           </argument>
-          <argument category="input" name="bias" type="out_t*" shape="[OC]">
+          <argument category="input" name="bias" type="tensor_t" shape="[OC]" tensor-element-type="out_t">
             <description>Per output channel bias data.</description>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="pad" type="int32_t*" shape="[4]">
+          <argument category="attribute" name="pad" type="tensor_t" shape="[4]" tensor-element-type="int32_t">
             <description>[pad_top, pad_bottom, pad_left, pad_right]</description>
             <levellimit value="pad_top"    limit="MAX_KERNEL"/>
             <levellimit value="pad_bottom" limit="MAX_KERNEL"/>
@@ -137,23 +141,25 @@
             <levellimit value="pad_right"  limit="MAX_KERNEL"/>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="stride" type="int32_t*" shape="[2]">
+          <argument category="attribute" name="stride" type="tensor_t" shape="[2]" tensor-element-type="int32_t">
             <description>[stride_y, stride_x]</description>
             <levellimit value="stride_y" limit="MAX_STRIDE"/>
             <levellimit value="stride_x" limit="MAX_STRIDE"/>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="dilation" type="int32_t*" shape="[2]">
+          <argument category="attribute" name="dilation" type="tensor_t" shape="[2]" tensor-element-type="int32_t">
             <description>[dilation_y, dilation_x]</description>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="input_zp" type="in_t" shape="-">
+          <argument category="attribute" name="input_zp" type="tensor_t" shape="-" tensor-element-type="in_t">
             <description>Input tensor zero point. Must be zero for non-int8 types.</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="attribute" name="weight_zp" type="weight_t" shape="-">
+          <argument category="attribute" name="weight_zp" type="tensor_t" shape="-" tensor-element-type="weight_t">
             <description>Weight zero point. Must be zero for non-int8 types.</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="out_t*" shape="[N,OH,OW,OC]">
+          <argument category="output" name="output" type="tensor_t" shape="[N,OH,OW,OC]" tensor-element-type="out_t">
             <description>Output tensor</description>
             <rank min="4" max="4"/>
           </argument>
@@ -186,22 +192,22 @@
       <operator>
         <name>CONV3D</name>
         <arguments>
-          <argument category="input" name="input" type="in_t*" shape="[N,ID,IH,IW,IC]">
+          <argument category="input" name="input" type="tensor_t" shape="[N,ID,IH,IW,IC]" tensor-element-type="in_t">
             <description>Input tensor</description>
             <rank min="5" max="5"/>
           </argument>
-          <argument category="input" name="weight" type="weight_t*" shape="[OC,KD,KH,KW,IC]">
+          <argument category="input" name="weight" type="tensor_t" shape="[OC,KD,KH,KW,IC]" tensor-element-type="weight_t">
             <description>Weight kernel size KDxKHxKW</description>
             <levellimit value="dilation_d * KD" limit="MAX_KERNEL"/>
             <levellimit value="dilation_y * KH" limit="MAX_KERNEL"/>
             <levellimit value="dilation_x * KW" limit="MAX_KERNEL"/>
             <rank min="5" max="5"/>
           </argument>
-          <argument category="input" name="bias" type="out_t*" shape="[OC]">
+          <argument category="input" name="bias" type="tensor_t" shape="[OC]" tensor-element-type="out_t">
             <description>Per output channel bias data.</description>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="pad" type="int32_t*" shape="[6]">
+          <argument category="attribute" name="pad" type="tensor_t" shape="[6]" tensor-element-type="int32_t">
             <description>[pad_d0, pad_d1, pad_top, pad_bottom, pad_left, pad_right]</description>
             <levellimit value="pad_d0"     limit="MAX_KERNEL"/>
             <levellimit value="pad_d1"     limit="MAX_KERNEL"/>
@@ -211,24 +217,26 @@
             <levellimit value="pad_right"  limit="MAX_KERNEL"/>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="stride" type="int32_t*" shape="[3]">
+          <argument category="attribute" name="stride" type="tensor_t" shape="[3]" tensor-element-type="int32_t">
             <description>[stride_d, stride_y, stride_x]</description>
             <levellimit value="stride_y" limit="MAX_STRIDE"/>
             <levellimit value="stride_x" limit="MAX_STRIDE"/>
             <levellimit value="stride_d" limit="MAX_STRIDE"/>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="dilation" type="int32_t*" shape="[3]">
+          <argument category="attribute" name="dilation" type="tensor_t" shape="[3]" tensor-element-type="int32_t">
             <description>[dilation_d, dilation_y, dilation_x]</description>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="input_zp" type="in_t" shape="-">
+          <argument category="attribute" name="input_zp" type="tensor_t" shape="-" tensor-element-type="in_t">
             <description>Input tensor zero point. Must be zero for non-int8 types.</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="attribute" name="weight_zp" type="weight_t" shape="-">
+          <argument category="attribute" name="weight_zp" type="tensor_t" shape="-" tensor-element-type="weight_t">
             <description>Weight zero point. Must be zero for non-int8 types.</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="out_t*" shape="[N,OD,OH,OW,OC]">
+          <argument category="output" name="output" type="tensor_t" shape="[N,OD,OH,OW,OC]" tensor-element-type="out_t">
             <description>Output tensor</description>
             <rank min="5" max="5"/>
           </argument>
@@ -261,21 +269,21 @@
       <operator>
         <name>DEPTHWISE_CONV2D</name>
         <arguments>
-          <argument category="input" name="input" type="in_t*" shape="[N,H,W,C]">
+          <argument category="input" name="input" type="tensor_t" shape="[N,H,W,C]" tensor-element-type="in_t">
             <description>Input tensor</description>
             <rank min="4" max="4"/>
           </argument>
-          <argument category="input" name="weight" type="weight_t*" shape="[KH,KW,C,M]">
+          <argument category="input" name="weight" type="tensor_t" shape="[KH,KW,C,M]" tensor-element-type="weight_t">
             <description>Weight kernel size KH x KW</description>
             <levellimit value="dilation_y * KH" limit="MAX_KERNEL"/>
             <levellimit value="dilation_x * KW" limit="MAX_KERNEL"/>
             <rank min="4" max="4"/>
           </argument>
-          <argument category="input" name="bias" type="out_t*" shape="[C*M]">
+          <argument category="input" name="bias" type="tensor_t" shape="[C*M]" tensor-element-type="out_t">
             <description>Per output channel bias data.</description>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="pad" type="int32_t*" shape="[4]">
+          <argument category="attribute" name="pad" type="tensor_t" shape="[4]" tensor-element-type="int32_t">
             <description>[pad_top, pad_bottom, pad_left, pad_right]</description>
             <levellimit value="pad_top"    limit="MAX_KERNEL"/>
             <levellimit value="pad_bottom" limit="MAX_KERNEL"/>
@@ -283,23 +291,25 @@
             <levellimit value="pad_right"  limit="MAX_KERNEL"/>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="stride" type="int32_t*" shape="[2]">
+          <argument category="attribute" name="stride" type="tensor_t" shape="[2]" tensor-element-type="int32_t">
             <description>[stride_y, stride_x]</description>
             <levellimit value="stride_y" limit="MAX_STRIDE"/>
             <levellimit value="stride_x" limit="MAX_STRIDE"/>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="dilation" type="int32_t*" shape="[2]">
+          <argument category="attribute" name="dilation" type="tensor_t" shape="[2]" tensor-element-type="int32_t">
             <description>[dilation_y, dilation_x]</description>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="input_zp" type="in_t" shape="-">
+          <argument category="attribute" name="input_zp" type="tensor_t" shape="-" tensor-element-type="in_t">
             <description>Input tensor zero point. Must be zero for non-int8 types.</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="attribute" name="weight_zp" type="weight_t" shape="-">
+          <argument category="attribute" name="weight_zp" type="tensor_t" shape="-" tensor-element-type="weight_t">
             <description>Weight zero point. Must be zero for non-int8 types.</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="out_t*" shape="[N,OH,OW,C*M]">
+          <argument category="output" name="output" type="tensor_t" shape="[N,OH,OW,C*M]" tensor-element-type="out_t">
             <description>Output tensor</description>
             <rank min="4" max="4"/>
           </argument>
@@ -332,24 +342,25 @@
       <operator>
         <name>FFT2D</name>
         <arguments>
-          <argument category="input" name="input_real" type="in_out_t*" shape="[N,H,W]">
+          <argument category="input" name="input_real" type="tensor_t" shape="[N,H,W]" tensor-element-type="in_out_t">
             <description>Real part of the complex input. H,W must be powers of two.</description>
             <levellimit value="H" limit="MAX_KERNEL"/>
             <levellimit value="W" limit="MAX_KERNEL"/>
             <rank min="3" max="3"/>
           </argument>
-          <argument category="input" name="input_imag" type="in_out_t*" shape="[N,H,W]">
+          <argument category="input" name="input_imag" type="tensor_t" shape="[N,H,W]" tensor-element-type="in_out_t">
             <description>Imaginary part of the complex input. H,W must be powers of two.</description>
             <rank min="3" max="3"/>
           </argument>
-          <argument category="attribute" name="inverse" type="bool_t" shape="-">
+          <argument category="attribute" name="inverse" type="tensor_t" shape="-" tensor-element-type="bool_t">
             <description>false for forward FFT, true for inverse FFT</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output_real" type="in_out_t*" shape="[N,H,W]">
+          <argument category="output" name="output_real" type="tensor_t" shape="[N,H,W]" tensor-element-type="in_out_t">
             <description>Real part of the complex output.</description>
             <rank min="3" max="3"/>
           </argument>
-          <argument category="output" name="output_imag" type="in_out_t*" shape="[N,H,W]">
+          <argument category="output" name="output_imag" type="tensor_t" shape="[N,H,W]" tensor-element-type="in_out_t">
             <description>Imaginary part of the complex output.</description>
             <rank min="3" max="3"/>
           </argument>
@@ -365,25 +376,27 @@
       <operator>
         <name>FULLY_CONNECTED</name>
         <arguments>
-          <argument category="input" name="input" type="in_t*" shape="[N,IC]">
+          <argument category="input" name="input" type="tensor_t" shape="[N,IC]" tensor-element-type="in_t">
             <description>Input tensor</description>
             <rank min="2" max="2"/>
           </argument>
-          <argument category="attribute" name="weight" type="weight_t*" shape="[OC,IC]">
+          <argument category="attribute" name="weight" type="tensor_t" shape="[OC,IC]" tensor-element-type="weight_t">
             <description>Weights</description>
             <rank min="2" max="2"/>
           </argument>
-          <argument category="attribute" name="bias" type="out_t*" shape="[OC]">
+          <argument category="attribute" name="bias" type="tensor_t" shape="[OC]" tensor-element-type="out_t">
             <description>Per output channel bias data.</description>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="input_zp" type="in_t" shape="-">
+          <argument category="attribute" name="input_zp" type="tensor_t" shape="-" tensor-element-type="in_t">
             <description>Input tensor zero point. Must be zero for non-int8 types.</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="attribute" name="weight_zp" type="weight_t" shape="-">
+          <argument category="attribute" name="weight_zp" type="tensor_t" shape="-" tensor-element-type="weight_t">
             <description>Weight zero point. Must be zero for non-int8 types.</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="out_t*" shape="[N,OC]">
+          <argument category="output" name="output" type="tensor_t" shape="[N,OC]" tensor-element-type="out_t">
             <description>Output tensor</description>
             <rank min="2" max="2"/>
           </argument>
@@ -416,21 +429,23 @@
       <operator>
         <name>MATMUL</name>
         <arguments>
-          <argument category="input" name="A" type="in_t*" shape="[N,H,C]">
+          <argument category="input" name="A" type="tensor_t" shape="[N,H,C]" tensor-element-type="in_t">
             <description>Input tensor A, N matrices of size HxC</description>
             <rank min="3" max="3"/>
           </argument>
-          <argument category="input" name="B" type="in_t*" shape="[N,C,W]">
+          <argument category="input" name="B" type="tensor_t" shape="[N,C,W]" tensor-element-type="in_t">
             <description>Input tensor B, N matrices of size CxW</description>
             <rank min="3" max="3"/>
           </argument>
-          <argument category="attribute" name="A_zp" type="in_t" shape="-">
+          <argument category="attribute" name="A_zp" type="tensor_t" shape="-" tensor-element-type="in_t">
             <description>Input tensor A zero point. Must be zero for non-int8 types.</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="attribute" name="B_zp" type="in_t" shape="-">
+          <argument category="attribute" name="B_zp" type="tensor_t" shape="-" tensor-element-type="in_t">
             <description>Input tensor B zero point. Must be zero for non-int8 types.</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="out_t*" shape="[N,H,W]">
+          <argument category="output" name="output" type="tensor_t" shape="[N,H,W]" tensor-element-type="out_t">
             <description>Output tensor, N matrices of size HxW</description>
             <rank min="3" max="3"/>
           </argument>
@@ -461,23 +476,23 @@
       <operator>
         <name>MAX_POOL2D</name>
         <arguments>
-          <argument category="input" name="input" type="in_out_t*" shape="[N,IH,IW,C]">
+          <argument category="input" name="input" type="tensor_t" shape="[N,IH,IW,C]" tensor-element-type="in_out_t">
             <description>Input tensor 4D</description>
             <rank min="4" max="4"/>
           </argument>
-          <argument category="attribute" name="kernel" type="int32_t*" shape="[2]">
+          <argument category="attribute" name="kernel" type="tensor_t" shape="[2]" tensor-element-type="int32_t">
             <description>[kernel_y, kernel_x]</description>
             <levellimit value="kernel_y" limit="MAX_KERNEL"/>
             <levellimit value="kernel_x" limit="MAX_KERNEL"/>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="stride" type="int32_t*" shape="[2]">
+          <argument category="attribute" name="stride" type="tensor_t" shape="[2]" tensor-element-type="int32_t">
             <description>[stride_y, stride_x]</description>
             <levellimit value="stride_y" limit="MAX_STRIDE"/>
             <levellimit value="stride_x" limit="MAX_STRIDE"/>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="pad" type="int32_t*" shape="[4]">
+          <argument category="attribute" name="pad" type="tensor_t" shape="[4]" tensor-element-type="int32_t">
             <description>[pad_top, pad_bottom, pad_left, pad_right]</description>
             <levellimit value="pad_top"    limit="MAX_KERNEL"/>
             <levellimit value="pad_bottom" limit="MAX_KERNEL"/>
@@ -485,7 +500,7 @@
             <levellimit value="pad_right"  limit="MAX_KERNEL"/>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="[N,OH,OW,C]">
+          <argument category="output" name="output" type="tensor_t" shape="[N,OH,OW,C]" tensor-element-type="in_out_t">
             <description>Output tensor 4D</description>
             <rank min="4" max="4"/>
           </argument>
@@ -511,17 +526,17 @@
       <operator>
         <name>RFFT2D</name>
         <arguments>
-          <argument category="input" name="input" type="in_out_t*" shape="[N,H,W]">
+          <argument category="input" name="input" type="tensor_t" shape="[N,H,W]" tensor-element-type="in_out_t">
             <description>Real input. H,W must be powers of two.</description>
             <levellimit value="H" limit="MAX_KERNEL"/>
             <levellimit value="W" limit="MAX_KERNEL"/>
             <rank min="3" max="3"/>
           </argument>
-          <argument category="output" name="output_real" type="in_out_t*" shape="[N,H,W/2 + 1]">
+          <argument category="output" name="output_real" type="tensor_t" shape="[N,H,W/2 + 1]" tensor-element-type="in_out_t">
             <description>Real part of the complex output</description>
             <rank min="3" max="3"/>
           </argument>
-          <argument category="output" name="output_imag" type="in_out_t*" shape="[N,H,W/2 + 1]">
+          <argument category="output" name="output_imag" type="tensor_t" shape="[N,H,W/2 + 1]" tensor-element-type="in_out_t">
             <description>Imaginary part of the complex output.</description>
             <rank min="3" max="3"/>
           </argument>
@@ -537,21 +552,21 @@
       <operator>
         <name>TRANSPOSE_CONV2D</name>
         <arguments>
-          <argument category="input" name="input" type="in_t*" shape="[N,IH,IW,IC]">
+          <argument category="input" name="input" type="tensor_t" shape="[N,IH,IW,IC]" tensor-element-type="in_t">
             <description>Input tensor</description>
             <rank min="4" max="4"/>
           </argument>
-          <argument category="input" name="weight" type="weight_t*" shape="[OC,KH,KW,IC]">
+          <argument category="input" name="weight" type="tensor_t" shape="[OC,KH,KW,IC]" tensor-element-type="weight_t">
             <description>Weight kernel size KH x KW</description>
             <levellimit value="KH" limit="MAX_KERNEL"/>
             <levellimit value="KW" limit="MAX_KERNEL"/>
             <rank min="4" max="4"/>
           </argument>
-          <argument category="input" name="bias" type="out_t*" shape="[OC]">
+          <argument category="input" name="bias" type="tensor_t" shape="[OC]" tensor-element-type="out_t">
             <description>Per output channel bias data.</description>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="out_pad" type="int32_t*" shape="[4]">
+          <argument category="attribute" name="out_pad" type="tensor_t" shape="[4]" tensor-element-type="int32_t">
             <description>[out_pad_top, out_pad_bottom, out_pad_left, out_pad_right]</description>
             <levellimit value="out_pad_top"    limit="MAX_KERNEL"/>
             <levellimit value="out_pad_bottom" limit="MAX_KERNEL"/>
@@ -559,23 +574,25 @@
             <levellimit value="out_pad_right"  limit="MAX_KERNEL"/>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="stride" type="int32_t*" shape="[2]">
+          <argument category="attribute" name="stride" type="tensor_t" shape="[2]" tensor-element-type="int32_t">
             <description>[stride_y, stride_x]</description>
             <levellimit value="stride_y" limit="MAX_STRIDE"/>
             <levellimit value="stride_x" limit="MAX_STRIDE"/>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="out_shape" type="int32_t*" shape="[4]">
+          <argument category="attribute" name="out_shape" type="tensor_t" shape="[4]" tensor-element-type="int32_t">
             <description>[N,OH,OW,OC]</description>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="input_zp" type="in_t" shape="-">
+          <argument category="attribute" name="input_zp" type="tensor_t" shape="-" tensor-element-type="in_t">
             <description>Input tensor zero point. Must be zero for non-int8 types.</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="attribute" name="weight_zp" type="weight_t" shape="-">
+          <argument category="attribute" name="weight_zp" type="tensor_t" shape="-" tensor-element-type="weight_t">
             <description>Weight zero point. Must be zero for non-int8 types.</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="out_t*" shape="[N,OH,OW,OC]">
+          <argument category="output" name="output" type="tensor_t" shape="[N,OH,OW,OC]" tensor-element-type="out_t">
             <description>Output tensor</description>
             <rank min="4" max="4"/>
           </argument>
@@ -610,18 +627,20 @@
       <operator>
         <name>CLAMP</name>
         <arguments>
-          <argument category="input" name="input" type="in_out_t*" shape="shape">
+          <argument category="input" name="input" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="attribute" name="min_val" type="in_out_t" shape="-">
+          <argument category="attribute" name="min_val" type="tensor_t" shape="-" tensor-element-type="in_out_t">
             <description>Minimum clip value</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="attribute" name="max_val" type="in_out_t" shape="-">
+          <argument category="attribute" name="max_val" type="tensor_t" shape="-" tensor-element-type="in_out_t">
             <description>Maximum clip value</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type and shape as input</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
@@ -647,12 +666,12 @@
       <operator>
         <name>SIGMOID</name>
         <arguments>
-          <argument category="input" name="input" type="in_out_t*" shape="shape">
+          <argument category="input" name="input" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type and shape as input</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
@@ -676,12 +695,12 @@
       <operator>
         <name>TANH</name>
         <arguments>
-          <argument category="input" name="input" type="in_out_t*" shape="shape">
+          <argument category="input" name="input" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type and shape as input</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
@@ -707,15 +726,15 @@
       <operator>
         <name>ADD</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_out_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_out_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -741,18 +760,19 @@
       <operator>
         <name>ARITHMETIC_RIGHT_SHIFT</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_out_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_out_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="attribute" name="round" type="bool_t" shape="-">
+          <argument category="attribute" name="round" type="tensor_t" shape="-" tensor-element-type="bool_t">
             <description>If true then the shift is rounded</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -768,15 +788,15 @@
       <operator>
         <name>BITWISE_AND</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_out_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_out_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -792,15 +812,15 @@
       <operator>
         <name>BITWISE_OR</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_out_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_out_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -816,15 +836,15 @@
       <operator>
         <name>BITWISE_XOR</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_out_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_out_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -840,15 +860,15 @@
       <operator>
         <name>INTDIV</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_out_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_out_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -862,15 +882,15 @@
       <operator>
         <name>LOGICAL_AND</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_out_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_out_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -884,15 +904,15 @@
       <operator>
         <name>LOGICAL_LEFT_SHIFT</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_out_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_out_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -908,15 +928,15 @@
       <operator>
         <name>LOGICAL_RIGHT_SHIFT</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_out_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_out_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -932,15 +952,15 @@
       <operator>
         <name>LOGICAL_OR</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_out_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_out_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -954,15 +974,15 @@
       <operator>
         <name>LOGICAL_XOR</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_out_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_out_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -976,15 +996,15 @@
       <operator>
         <name>MAXIMUM</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_out_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_out_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -1010,15 +1030,15 @@
       <operator>
         <name>MINIMUM</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_out_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_out_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -1044,18 +1064,19 @@
       <operator>
         <name>MUL</name>
         <arguments>
-          <argument category="input" name="input1" type="in_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input(MT)|attribute(BI,MI)" name="shift" type="int8_t" shape="-">
+          <argument category="input(MT)|attribute(BI,MI)" name="shift" type="tensor_t" shape="-" tensor-element-type="int8_t">
             <description>Result right shift (int32_t data type only)</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -1084,15 +1105,15 @@
       <operator>
         <name>POW</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_out_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_out_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -1117,15 +1138,15 @@
       <operator>
         <name>SUB</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_out_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_out_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -1151,15 +1172,15 @@
       <operator>
         <name>TABLE</name>
         <arguments>
-          <argument category="input" name="input" type="in_t*" shape="shape">
+          <argument category="input" name="input" type="tensor_t" shape="shape" tensor-element-type="in_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input(MT)|attribute(BI,MI)" name="table" type="table_t*" shape="[TABLE_SIZE]">
+          <argument category="input(MT)|attribute(BI,MI)" name="table" type="tensor_t" shape="[TABLE_SIZE]" tensor-element-type="table_t">
             <description>Lookup table tensor</description>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="output" name="output" type="out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="out_t">
             <description>Output tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -1179,12 +1200,12 @@
       <operator>
         <name>ABS</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape">
+          <argument category="input" name="input1" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type, size as the input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
@@ -1209,12 +1230,12 @@
       <operator>
         <name>BITWISE_NOT</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape">
+          <argument category="input" name="input1" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type, size as the input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
@@ -1229,12 +1250,12 @@
       <operator>
         <name>CEIL</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape">
+          <argument category="input" name="input1" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type, size as the input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
@@ -1258,12 +1279,12 @@
       <operator>
         <name>CLZ</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape">
+          <argument category="input" name="input1" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type, size as the input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
@@ -1276,12 +1297,12 @@
       <operator>
         <name>EXP</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape">
+          <argument category="input" name="input1" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type, size as the input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
@@ -1305,12 +1326,12 @@
       <operator>
         <name>FLOOR</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape">
+          <argument category="input" name="input1" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type, size as the input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
@@ -1334,12 +1355,12 @@
       <operator>
         <name>LOG</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape">
+          <argument category="input" name="input1" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type, size as the input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
@@ -1363,12 +1384,12 @@
       <operator>
         <name>LOGICAL_NOT</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape">
+          <argument category="input" name="input1" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type, size as the input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
@@ -1381,18 +1402,20 @@
       <operator>
         <name>NEGATE</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape">
+          <argument category="input" name="input1" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="attribute" name="input1_zp" type="in_out_t" shape="-">
+          <argument category="attribute" name="input1_zp" type="tensor_t" shape="-" tensor-element-type="in_out_t">
             <description>Input 1 zero point. Must be zero for non-int8 types.</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="attribute" name="output_zp" type="in_out_t" shape="-">
+          <argument category="attribute" name="output_zp" type="tensor_t" shape="-" tensor-element-type="in_out_t">
             <description>Output zero point. Must be zero for non-int8 types.</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type, size as the input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
@@ -1420,12 +1443,12 @@
       <operator>
         <name>RECIPROCAL</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape">
+          <argument category="input" name="input1" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type, size as the input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
@@ -1449,12 +1472,12 @@
       <operator>
         <name>RSQRT</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape">
+          <argument category="input" name="input1" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type, size as the input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
@@ -1480,19 +1503,19 @@
       <operator>
         <name>SELECT</name>
         <arguments>
-          <argument category="input" name="input1" type="bool_t" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="bool_t">
             <description>Input selector tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_out_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_out_t">
             <description>Input value tensor if input1 is True</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input3" type="in_out_t*" shape="shape3">
+          <argument category="input" name="input3" type="tensor_t" shape="shape3" tensor-element-type="in_out_t">
             <description>Input value tensor if input1 is False</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type as input2 and input3, with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -1523,15 +1546,15 @@
       <operator>
         <name>EQUAL</name>
         <arguments>
-          <argument category="input" name="input1" type="in_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -1558,15 +1581,15 @@
       <operator>
         <name>GREATER</name>
         <arguments>
-          <argument category="input" name="input1" type="in_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -1593,15 +1616,15 @@
       <operator>
         <name>GREATER_EQUAL</name>
         <arguments>
-          <argument category="input" name="input1" type="in_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input2" type="in_t*" shape="shape2">
+          <argument category="input" name="input2" type="tensor_t" shape="shape2" tensor-element-type="in_t">
             <description>Input tensor with the same rank as input1</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="out_t">
             <description>Output tensor with broadcast shape if necessary</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
@@ -1630,14 +1653,15 @@
       <operator>
         <name>REDUCE_ALL</name>
         <arguments>
-          <argument category="input" name="input" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="1" max="4"/>
           </argument>
-          <argument category="attribute" name="axis" type="int32_t" shape="-">
+          <argument category="attribute" name="axis" type="tensor_t" shape="-" tensor-element-type="int32_t">
             <description>Axis to reduce, in range from 0 to rank(shape1)-1</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor. Same rank as the input tensor.</description>
             <rank min="1" max="4"/>
           </argument>
@@ -1650,14 +1674,15 @@
       <operator>
         <name>REDUCE_ANY</name>
         <arguments>
-          <argument category="input" name="input" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="1" max="4"/>
           </argument>
-          <argument category="attribute" name="axis" type="int32_t" shape="-">
+          <argument category="attribute" name="axis" type="tensor_t" shape="-" tensor-element-type="int32_t">
             <description>Axis to reduce, in range from 0 to rank(shape1)-1</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor. Same rank as the input tensor.</description>
             <rank min="1" max="4"/>
           </argument>
@@ -1670,14 +1695,15 @@
       <operator>
         <name>REDUCE_MAX</name>
         <arguments>
-          <argument category="input" name="input" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="1" max="4"/>
           </argument>
-          <argument category="attribute" name="axis" type="int32_t" shape="-">
+          <argument category="attribute" name="axis" type="tensor_t" shape="-" tensor-element-type="int32_t">
             <description>Axis to reduce, in range from 0 to rank(shape1)-1</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor. Same rank as the input tensor.</description>
             <rank min="1" max="4"/>
           </argument>
@@ -1704,14 +1730,15 @@
       <operator>
         <name>REDUCE_MIN</name>
         <arguments>
-          <argument category="input" name="input" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="1" max="4"/>
           </argument>
-          <argument category="attribute" name="axis" type="int32_t" shape="-">
+          <argument category="attribute" name="axis" type="tensor_t" shape="-" tensor-element-type="int32_t">
             <description>Axis to reduce, in range from 0 to rank(shape1)-1</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor. Same rank as the input tensor.</description>
             <rank min="1" max="4"/>
           </argument>
@@ -1738,14 +1765,15 @@
       <operator>
         <name>REDUCE_PRODUCT</name>
         <arguments>
-          <argument category="input" name="input" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="1" max="4"/>
           </argument>
-          <argument category="attribute" name="axis" type="int32_t" shape="-">
+          <argument category="attribute" name="axis" type="tensor_t" shape="-" tensor-element-type="int32_t">
             <description>Axis to reduce, in range from 0 to rank(shape1)-1</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor. Same rank as the input tensor.</description>
             <rank min="1" max="4"/>
           </argument>
@@ -1769,14 +1797,15 @@
       <operator>
         <name>REDUCE_SUM</name>
         <arguments>
-          <argument category="input" name="input" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor with rank from 1 to 4</description>
             <rank min="1" max="4"/>
           </argument>
-          <argument category="attribute" name="axis" type="int32_t" shape="-">
+          <argument category="attribute" name="axis" type="tensor_t" shape="-" tensor-element-type="int32_t">
             <description>Axis to reduce, in range from 0 to rank(shape1)-1</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor. Same rank as the input tensor.</description>
             <rank min="1" max="4"/>
           </argument>
@@ -1803,15 +1832,15 @@
       <operator>
         <name>CONCAT</name>
         <arguments>
-          <!-- FIXME express list of tensors better -->
-          <argument category="input" name="input1" type="in_out_t*" shape="shapes1[]">
+          <argument category="input" name="input1" type="tensor_list_t" shape="shapes1" tensor-element-type="in_out_t">
             <description>List of input tensors. All inputs must have the same rank and data type</description>
             <rank min="1" max="MAX_RANK"/>
           </argument>
-          <argument category="attribute" name="axis" type="int32_t" shape="-">
+          <argument category="attribute" name="axis" type="tensor_t" shape="-" tensor-element-type="int32_t">
             <description>Axis along which concatenation is to occur, in range from 0 to rank(shape)-1</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="1" max="MAX_RANK"/>
@@ -1840,18 +1869,19 @@
       <operator>
         <name>PAD</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="1" max="MAX_RANK"/>
           </argument>
-          <argument category="attribute" name="padding" type="int32_t" shape="[rank(shape1),2]">
+          <argument category="attribute" name="padding" type="tensor_t" shape="[rank(shape1),2]" tensor-element-type="int32_t">
             <description>Number of pad elements at the start and end of each dimension</description>
             <rank min="2" max="2"/>
           </argument>
-          <argument category="attribute" name="pad_const" type="in_out_t" shape="-">
+          <argument category="attribute" name="pad_const" type="tensor_t" shape="-" tensor-element-type="in_out_t">
             <description>Constant value to be used as padding</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type as the input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="1" max="MAX_RANK"/>
@@ -1880,16 +1910,16 @@
       <operator>
         <name>RESHAPE</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape1)" limit="MAX_RANK"/>
             <rank min="1" max="MAX_RANK"/>
           </argument>
-          <argument category="attribute" name="new_shape" type="int32_t" shape="[rank(shape)]">
+          <argument category="attribute" name="new_shape" type="tensor_t" shape="[rank(shape)]" tensor-element-type="int32_t">
             <description>List of values, with each element giving the size of the result tensor for the given dimension. At most one dimension may be given as -1 to automatically calculate the dimension size.</description>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type, size as the input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="1" max="MAX_RANK"/>
@@ -1918,15 +1948,16 @@
       <operator>
         <name>REVERSE</name>
         <arguments>
-          <argument category="input" name="input" type="in_out_t*" shape="shape">
+          <argument category="input" name="input" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="1" max="MAX_RANK"/>
           </argument>
-          <argument category="attribute" name="axis" type="int32_t" shape="-">
+          <argument category="attribute" name="axis" type="tensor_t" shape="-" tensor-element-type="int32_t">
             <description>Axis to reverse, in range from 0 to rank(shape)-1</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor. Same shape as input tensor</description>
             <rank min="1" max="MAX_RANK"/>
           </argument>
@@ -1954,21 +1985,21 @@
       <operator>
         <name>SLICE</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="1" max="MAX_RANK"/>
           </argument>
-          <argument category="attribute" name="start" type="index_t" shape="[rank(shape1)]">
+          <argument category="attribute" name="start" type="tensor_t" shape="[rank(shape1)]" tensor-element-type="index_t">
             <description>List of integer coordinates, of length equal to the rank of input1. Start coordinate for slicing.</description>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="size" type="index_t" shape="[rank(shape1)]">
+          <argument category="attribute" name="size" type="tensor_t" shape="[rank(shape1)]" tensor-element-type="index_t">
             <description>List of integer size values, of length equal to the rank of input1. Size of the input to be
 used.</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type as the input tensor</description>
             <rank min="1" max="MAX_RANK"/>
           </argument>
@@ -1996,15 +2027,15 @@ used.</description>
       <operator>
         <name>TILE</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="1" max="MAX_RANK"/>
           </argument>
-          <argument category="attribute" name="multiples" type="int32_t" shape="[rank(shape1)]">
+          <argument category="attribute" name="multiples" type="tensor_t" shape="[rank(shape1)]" tensor-element-type="int32_t">
             <description>Number of times to replicate input1 in each dimension</description>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type, rank as the input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="1" max="MAX_RANK"/>
@@ -2033,15 +2064,15 @@ used.</description>
       <operator>
         <name>TRANSPOSE</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape1">
+          <argument category="input" name="input1" type="tensor_t" shape="shape1" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="1" max="MAX_RANK"/>
           </argument>
-          <argument category="attribute" name="perms" type="int32_t" shape="[rank(shape1)]">
+          <argument category="attribute" name="perms" type="tensor_t" shape="[rank(shape1)]" tensor-element-type="int32_t">
             <description>List of integers of length equal to the rank of input1. Values must be valid dimensions within shape1, and may not be repeated.</description>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of same type, rank as the input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="1" max="MAX_RANK"/>
@@ -2072,15 +2103,15 @@ used.</description>
       <operator>
         <name>GATHER</name>
         <arguments>
-          <argument category="input" name="values" type="in_out_t*" shape="[N,K,C]">
+          <argument category="input" name="values" type="tensor_t" shape="[N,K,C]" tensor-element-type="in_out_t">
             <description>3D value tensor</description>
             <rank min="3" max="3"/>
           </argument>
-          <argument category="input" name="indices" type="index_t*" shape="[N,W]">
+          <argument category="input" name="indices" type="tensor_t" shape="[N,W]" tensor-element-type="index_t">
             <description>2D index tensor</description>
             <rank min="2" max="2"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="[N,W,C]">
+          <argument category="output" name="output" type="tensor_t" shape="[N,W,C]" tensor-element-type="in_out_t">
             <description>3D output tensor</description>
             <rank min="3" max="3"/>
           </argument>
@@ -2107,19 +2138,19 @@ used.</description>
       <operator>
         <name>SCATTER</name>
         <arguments>
-          <argument category="input" name="values_in" type="in_out_t*" shape="[N,K,C]">
+          <argument category="input" name="values_in" type="tensor_t" shape="[N,K,C]" tensor-element-type="in_out_t">
             <description>3D values in tensor</description>
             <rank min="3" max="3"/>
           </argument>
-          <argument category="input" name="indices" type="index_t*" shape="[N,W]">
+          <argument category="input" name="indices" type="tensor_t" shape="[N,W]" tensor-element-type="index_t">
             <description>2D index tensor</description>
             <rank min="2" max="2"/>
           </argument>
-          <argument category="input" name="input" type="in_out_t*" shape="[N,W,C]">
+          <argument category="input" name="input" type="tensor_t" shape="[N,W,C]" tensor-element-type="in_out_t">
             <description>3D input tensor</description>
             <rank min="3" max="3"/>
           </argument>
-          <argument category="output" name="values_out" type="in_out_t*" shape="[N,K,C]">
+          <argument category="output" name="values_out" type="tensor_t" shape="[N,K,C]" tensor-element-type="in_out_t">
             <description>3D output tensor</description>
             <rank min="3" max="3"/>
           </argument>
@@ -2148,28 +2179,29 @@ used.</description>
       <operator>
         <name>RESIZE</name>
         <arguments>
-          <argument category="input" name="input" type="in_t*" shape="[N,IH,IW,C]">
+          <argument category="input" name="input" type="tensor_t" shape="[N,IH,IW,C]" tensor-element-type="in_t">
             <description>Input tensor</description>
             <rank min="4" max="4"/>
           </argument>
-          <argument category="attribute" name="scale" type="int16_t*" shape="[4]">
+          <argument category="attribute" name="scale" type="tensor_t" shape="[4]" tensor-element-type="int16_t">
             <description>[scale_y_n, scale_y_d, scale_x_n, scale_x_d]</description>
             <levellimit value="scale_y_n/scale_y_d" limit="MAX_SCALE"/>
             <levellimit value="scale_x_n/scale_x_d" limit="MAX_SCALE"/>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="offset" type="int16_t*" shape="[2]">
+          <argument category="attribute" name="offset" type="tensor_t" shape="[2]" tensor-element-type="int16_t">
             <description>[offset_y, offset_x]</description>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="border" type="int16_t*" shape="[2]">
+          <argument category="attribute" name="border" type="tensor_t" shape="[2]" tensor-element-type="int16_t">
             <description>[border_y, border_x]</description>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="mode" type="resize_mode_t" shape="-">
+          <argument category="attribute" name="mode" type="tensor_t" shape="-" tensor-element-type="resize_mode_t">
             <description>BILINEAR or NEAREST</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="output" name="output" type="out_t*" shape="[N,OH,OW,C]">
+          <argument category="output" name="output" type="tensor_t" shape="[N,OH,OW,C]" tensor-element-type="out_t">
             <description>Output tensor</description>
             <rank min="4" max="4"/>
           </argument>
@@ -2201,12 +2233,12 @@ used.</description>
       <operator>
         <name>CAST</name>
         <arguments>
-          <argument category="input" name="input" type="in_t" shape="shape">
+          <argument category="input" name="input" type="tensor_t" shape="shape" tensor-element-type="in_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="out_t" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="out_t">
             <description>Output tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
@@ -2319,37 +2351,42 @@ used.</description>
       <operator>
         <name>RESCALE</name>
         <arguments>
-          <argument category="input" name="input" type="in_t" shape="shape">
+          <argument category="input" name="input" type="tensor_t" shape="shape" tensor-element-type="in_t">
             <description>Input tensor</description>
             <levellimit value="rank(shape)" limit="MAX_RANK"/>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="out_t" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="out_t">
             <description>Output tensor with the same shape as input</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="attribute" name="input_zp" type="in_t" shape="-">
+          <argument category="attribute" name="input_zp" type="tensor_t" shape="-" tensor-element-type="in_t">
             <description>Input tensor zero point. Must be zero for non-int8 types.</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="attribute" name="output_zp" type="out_t" shape="-">
+          <argument category="attribute" name="output_zp" type="tensor_t" shape="-" tensor-element-type="out_t">
             <description>Output tensor zero point. Must be zero for non-int8 types.</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="input(MT)|attribute(BI,MI)" name="multiplier" type="mul_t*" shape="[NC]">
+          <argument category="input(MT)|attribute(BI,MI)" name="multiplier" type="tensor_t" shape="[NC]" tensor-element-type="mul_t">
             <description>Scaling multiplier array</description>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="input(MT)|attribute(BI,MI)" name="shift" type="int8_t*" shape="[NC]">
+          <argument category="input(MT)|attribute(BI,MI)" name="shift" type="tensor_t" shape="[NC]" tensor-element-type="int8_t">
             <description>Scaling shift array</description>
             <rank min="1" max="1"/>
           </argument>
-          <argument category="attribute" name="scale32" type="bool_t" shape="-">
+          <argument category="attribute" name="scale32" type="tensor_t" shape="-" tensor-element-type="bool_t">
             <description>if (scale32) mul_t=int32_t else mul_t=int16_t</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="attribute" name="double_round" type="bool_t" shape="-">
+          <argument category="attribute" name="double_round" type="tensor_t" shape="-" tensor-element-type="bool_t">
             <description>Select double round mode</description>
+            <rank min="0" max="0"/>
           </argument>
-          <argument category="attribute" name="per_channel" type="bool_t" shape="-">
+          <argument category="attribute" name="per_channel" type="tensor_t" shape="-" tensor-element-type="bool_t">
             <description>if (per_channel) NC=shape[rank(shape)-1] else NC=1</description>
+            <rank min="0" max="0"/>
           </argument>
         </arguments>
         <types>
@@ -2380,11 +2417,11 @@ used.</description>
       <operator>
         <name>CONST</name>
         <arguments>
-          <argument category="attribute" name="values" type="out_t*" shape="shape">
+          <argument category="attribute" name="values" type="tensor_t" shape="shape" tensor-element-type="out_t">
             <description>Constant values</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="out_t">
             <description>Output tensor of the same type, size as the input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
@@ -2413,11 +2450,11 @@ used.</description>
       <operator>
         <name>IDENTITY</name>
         <arguments>
-          <argument category="input" name="input1" type="in_out_t*" shape="shape">
+          <argument category="input" name="input1" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
-          <argument category="output" name="output" type="in_out_t*" shape="shape">
+          <argument category="output" name="output" type="tensor_t" shape="shape" tensor-element-type="in_out_t">
             <description>Output tensor of the same type, size as the input tensor</description>
             <rank min="0" max="MAX_RANK"/>
           </argument>
@@ -2447,20 +2484,20 @@ used.</description>
       <operator>
         <name>COND_IF</name>
         <arguments>
-          <argument category="input" name="condition" type="bool_t*" shape="shape">
+          <argument category="input" name="condition" type="tensor_t" shape="shape" tensor-element-type="bool_t">
             <description>Input condition as a size 1 tensor</description>
             <rank min="1" max="MAX_RANK"/>
           </argument>
-          <argument category="input" name="input_list" type="tensor_list_t" shape="-">
+          <argument category="input" name="input_list" type="tensor_list_t" shape="-" tensor-element-type="-">
             <description>List of input tensors</description>
           </argument>
-          <argument category="attribute" name="then_graph" type="tosa_graph_t" shape="-">
+          <argument category="attribute" name="then_graph" type="tosa_graph_t" shape="-" tensor-element-type="-">
             <description>TOSA graph to execute if condition is true</description>
           </argument>
-          <argument category="attribute" name="else_graph" type="tosa_graph_t" shape="-">
+          <argument category="attribute" name="else_graph" type="tosa_graph_t" shape="-" tensor-element-type="-">
             <description>TOSA graph to execute if condition is false</description>
           </argument>
-          <argument category="output" name="output_list" type="tensor_list_t" shape="-">
+          <argument category="output" name="output_list" type="tensor_list_t" shape="-" tensor-element-type="-">
             <description>List of output tensors</description>
           </argument>
         </arguments>
@@ -2468,16 +2505,16 @@ used.</description>
       <operator>
         <name>WHILE_LOOP</name>
         <arguments>
-          <argument category="input" name="input_list" type="tensor_list_t" shape="-">
+          <argument category="input" name="input_list" type="tensor_list_t" shape="-" tensor-element-type="-">
             <description>List of input tensors</description>
           </argument>
-          <argument category="attribute" name="cond_graph" type="tosa_graph_t" shape="-">
+          <argument category="attribute" name="cond_graph" type="tosa_graph_t" shape="-" tensor-element-type="-">
             <description>TOSA graph to evaluate the condition</description>
           </argument>
-          <argument category="attribute" name="body_graph" type="tosa_graph_t" shape="-">
+          <argument category="attribute" name="body_graph" type="tosa_graph_t" shape="-" tensor-element-type="-">
             <description>TOSA graph to execute the loop body</description>
           </argument>
-          <argument category="output" name="output_list" type="tensor_list_t" shape="-">
+          <argument category="output" name="output_list" type="tensor_list_t" shape="-" tensor-element-type="-">
             <description>List of output tensors</description>
           </argument>
         </arguments>
diff --git a/tosa.xsd b/tosa.xsd
index fe08885..40fd613 100644
--- a/tosa.xsd
+++ b/tosa.xsd
@@ -30,6 +30,7 @@
 
 <xs:simpleType name="datatype">
   <xs:restriction base="xs:string">
+    <xs:enumeration value="-"/>
     <xs:enumeration value="bool_t"/>
     <xs:enumeration value="int4_t"/>
     <xs:enumeration value="int8_t"/>
@@ -44,6 +45,13 @@
   </xs:restriction>
 </xs:simpleType>
 
+<xs:simpleType name="enumtypename">
+  <xs:restriction base="xs:string">
+    <xs:enumeration value="resize_mode_t"/>
+    <xs:enumeration value="acc_size_t"/>
+  </xs:restriction>
+</xs:simpleType>
+
 <xs:simpleType name="typename">
   <xs:restriction base="xs:string">
     <xs:enumeration value="in_t"/>
@@ -53,6 +61,8 @@
     <xs:enumeration value="weight_t"/>
     <xs:enumeration value="resize_t"/>
     <xs:enumeration value="table_t"/>
+    <xs:enumeration value="index_t"/>
+    <xs:enumeration value="mul_t"/>
     <xs:enumeration value="TABLE_SIZE"/>
   </xs:restriction>
 </xs:simpleType>
@@ -66,6 +76,18 @@
   </xs:restriction>
 </xs:simpleType>
 
+<xs:simpleType name="argument-type">
+  <xs:restriction base="xs:string">
+    <xs:enumeration value="tensor_t"/>
+    <xs:enumeration value="tensor_list_t"/>
+    <xs:enumeration value="tosa_graph_t"/>
+  </xs:restriction>
+</xs:simpleType>
+
+<xs:simpleType name="argument-tensor-element-type">
+  <xs:union memberTypes="datatype typename enumtypename"/>
+</xs:simpleType>
+
 <!-- Element definitions -->
 
 <xs:element name="version">
@@ -179,7 +201,8 @@
     </xs:sequence>
     <xs:attribute name="category" type="argumentcategory" use="required"/>
     <xs:attribute name="name" type="xs:string" use="required"/>
-    <xs:attribute name="type" type="xs:string" use="required"/>
+    <xs:attribute name="type" type="argument-type" use="required"/>
+    <xs:attribute name="tensor-element-type" type="argument-tensor-element-type" use="required"/>
     <xs:attribute name="shape" type="xs:string" use="required"/>
   </xs:complexType>
 </xs:element>