Update RESHAPE and TILE conformance testing

Signed-off-by: Won Jeon <won.jeon@arm.com>
Change-Id: Iaf59472cb32e03a92bad87ae9dba1b7548f20268

11 files changed, 213 insertions, 48 deletions

diff --git a/reference_model/CMakeLists.txt b/reference_model/CMakeLists.txt
index 178594b..d2dce3c 100644
--- a/reference_model/CMakeLists.txt
+++ b/reference_model/CMakeLists.txt
@@ -73,6 +73,7 @@ set(CXX_SOURCE
     src/generate/generate_dot_product_states.cc
     src/generate/generate_dot_product.cc
     src/generate/generate_pseudo_random.cc
+    src/generate/generate_fixed_data.cc
     src/generate/generate_entry.cc
     src/generate/generate_utils.cc
     src/verify/verify_abs_error.cc
@@ -173,6 +174,7 @@ add_library(tosa_reference_generate_lib SHARED
   src/generate/generate_dot_product_states.cc
   src/generate/generate_dot_product.cc
   src/generate/generate_pseudo_random.cc
+  src/generate/generate_fixed_data.cc
   src/generate/generate_entry.cc
   src/generate/generate_utils.cc
   src/generate/generate_config.cc
diff --git a/reference_model/src/generate/generate_entry.cc b/reference_model/src/generate/generate_entry.cc
index 741cd79..91b2fc7 100644
--- a/reference_model/src/generate/generate_entry.cc
+++ b/reference_model/src/generate/generate_entry.cc
@@ -1,4 +1,4 @@
-// Copyright (c) 2023, ARM Limited.
+// Copyright (c) 2023-2024, ARM Limited.
 //
 //    Licensed under the Apache License, Version 2.0 (the "License");
 //    you may not use this file except in compliance with the License.
@@ -15,6 +15,7 @@
 #include "generate.h"
 
 #include "generate_dot_product.h"
+#include "generate_fixed_data.h"
 #include "generate_pseudo_random.h"
 #include "generate_utils.h"
 
@@ -36,6 +37,10 @@ bool generate(const GenerateConfig& cfg, void* data, size_t size)
             return generatePseudoRandom(cfg, data, size);
             break;
         }
+        case GeneratorType::FixedData: {
+            return generateFixedData(cfg, data, size);
+            break;
+        }
         default: {
             WARNING("[Generator] Unsupported generation mode.");
             break;
diff --git a/reference_model/src/generate/generate_fixed_data.cc b/reference_model/src/generate/generate_fixed_data.cc
new file mode 100644
index 0000000..d83ee58
--- /dev/null
+++ b/reference_model/src/generate/generate_fixed_data.cc
@@ -0,0 +1,56 @@
+// Copyright (c) 2024, ARM Limited.
+//
+//    Licensed under the Apache License, Version 2.0 (the "License");
+//    you may not use this file except in compliance with the License.
+//    You may obtain a copy of the License at
+//
+//         http://www.apache.org/licenses/LICENSE-2.0
+//
+//    Unless required by applicable law or agreed to in writing, software
+//    distributed under the License is distributed on an "AS IS" BASIS,
+//    WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+//    See the License for the specific language governing permissions and
+//    limitations under the License.
+#include "generate.h"
+#include "generate_utils.h"
+
+#include <algorithm>
+#include <array>
+#include <iterator>
+#include <type_traits>
+#include <vector>
+
+namespace TosaReference
+{
+bool generateFixedData(const GenerateConfig& cfg, void* data, size_t size)
+{
+    // Check we support the operator
+    if (cfg.opType == Op::Op_UNKNOWN)
+    {
+        WARNING("[Generator][FD] Unknown operator.");
+        return false;
+    }
+
+    switch (cfg.dataType)
+    {
+        case DType::DType_SHAPE: {
+            int32_t* outData            = reinterpret_cast<int32_t*>(data);
+            std::vector<int32_t> inData = cfg.fixedDataInfo.data;
+            const auto T                = TosaReference::numElementsFromShape(cfg.shape);
+            if (T != inData.size())
+            {
+                WARNING("[Generator][FD] Size does not match.");
+                return false;
+            }
+            for (auto t = 0; t < T; t++)
+            {
+                outData[t] = inData[t];
+            }
+            return true;
+        }
+        default:
+            WARNING("[Generator][FD] Unsupported type.");
+            return false;
+    }
+}
+}    // namespace TosaReference
diff --git a/reference_model/src/generate/generate_fixed_data.h b/reference_model/src/generate/generate_fixed_data.h
new file mode 100644
index 0000000..50371c8
--- /dev/null
+++ b/reference_model/src/generate/generate_fixed_data.h
@@ -0,0 +1,34 @@
+// Copyright (c) 2024, ARM Limited.
+//
+//    Licensed under the Apache License, Version 2.0 (the "License");
+//    you may not use this file except in compliance with the License.
+//    You may obtain a copy of the License at
+//
+//         http://www.apache.org/licenses/LICENSE-2.0
+//
+//    Unless required by applicable law or agreed to in writing, software
+//    distributed under the License is distributed on an "AS IS" BASIS,
+//    WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+//    See the License for the specific language governing permissions and
+//    limitations under the License.
+
+#ifndef GENERATE_FIXED_DATA_H_
+#define GENERATE_FIXED_DATA_H_
+
+#include "generate_utils.h"
+
+namespace TosaReference
+{
+
+/// \brief Perform fixed data generation
+///
+/// \param cfg Generator related meta-data
+/// \param data Buffer to generate the data to
+/// \param size Size of the buffer
+///
+/// \return True on successful generation
+bool generateFixedData(const GenerateConfig& cfg, void* data, size_t size);
+
+};    // namespace TosaReference
+
+#endif    // GENERATE_FIXED_DATA_H_
diff --git a/reference_model/src/generate/generate_utils.cc b/reference_model/src/generate/generate_utils.cc
index c16d1c6..9eda0b6 100644
--- a/reference_model/src/generate/generate_utils.cc
+++ b/reference_model/src/generate/generate_utils.cc
@@ -33,6 +33,7 @@ NLOHMANN_JSON_SERIALIZE_ENUM(DType,
                                  { DType::DType_FP16, "FP16" },
                                  { DType::DType_BF16, "BF16" },
                                  { DType::DType_FP32, "FP32" },
+                                 { DType::DType_SHAPE, "SHAPE" },
                              })
 
 NLOHMANN_JSON_SERIALIZE_ENUM(Op,
@@ -93,6 +94,7 @@ NLOHMANN_JSON_SERIALIZE_ENUM(GeneratorType,
                                  { GeneratorType::OpFullRange, "OP_FULL_RANGE" },
                                  { GeneratorType::OpBoundary, "OP_BOUNDARY" },
                                  { GeneratorType::OpSpecial, "OP_SPECIAL" },
+                                 { GeneratorType::FixedData, "FIXED_DATA" },
                              })
 
 // NOTE: This assumes it's VARIABLE if the InputType is not recognized
@@ -130,6 +132,11 @@ void from_json(const nlohmann::json& j, PseudoRandomInfo& pseudoRandomInfo)
     }
 }
 
+void from_json(const nlohmann::json& j, FixedDataInfo& fixedDataInfo)
+{
+    j.at("data").get_to(fixedDataInfo.data);
+}
+
 void from_json(const nlohmann::json& j, GenerateConfig& cfg)
 {
     j.at("data_type").get_to(cfg.dataType);
@@ -158,6 +165,13 @@ void from_json(const nlohmann::json& j, GenerateConfig& cfg)
     {
         j.at("pseudo_random_info").get_to(cfg.pseudoRandomInfo);
     }
+
+    // Set up defaults for fixedDataInfo
+    cfg.fixedDataInfo.data = std::vector<int32_t>();
+    if (j.contains("fixed_data_info"))
+    {
+        j.at("fixed_data_info").get_to(cfg.fixedDataInfo);
+    }
 }
 
 std::optional<GenerateConfig> parseGenerateConfig(const char* json, const char* tensorName)
@@ -209,6 +223,7 @@ size_t elementSizeFromType(DType type)
             return 2;
         case DType::DType_INT32:
         case DType::DType_FP32:
+        case DType::DType_SHAPE:
             return 4;
         default:
             return 0;
diff --git a/reference_model/src/generate/generate_utils.h b/reference_model/src/generate/generate_utils.h
index f9ec713..697b404 100644
--- a/reference_model/src/generate/generate_utils.h
+++ b/reference_model/src/generate/generate_utils.h
@@ -1,4 +1,4 @@
-// Copyright (c) 2023, ARM Limited.
+// Copyright (c) 2023-2024, ARM Limited.
 //
 //    Licensed under the Apache License, Version 2.0 (the "License");
 //    you may not use this file except in compliance with the License.
@@ -34,6 +34,7 @@ enum class GeneratorType
     OpFullRange,
     OpBoundary,
     OpSpecial,
+    FixedData,
 };
 
 /// \brief Supported input types
@@ -65,6 +66,14 @@ struct PseudoRandomInfo
     bool round;
 };
 
+/// \brief Fixed data generator meta-data
+struct FixedDataInfo
+{
+    FixedDataInfo() = default;
+
+    std::vector<int32_t> data;
+};
+
 /// \brief Generator configuration
 struct GenerateConfig
 {
@@ -76,6 +85,7 @@ struct GenerateConfig
     tosa::Op opType;
     DotProductInfo dotProductInfo;
     PseudoRandomInfo pseudoRandomInfo;
+    FixedDataInfo fixedDataInfo;
 };
 
 /// \brief Parse the generator config when given in JSON form
diff --git a/scripts/schemavalidation/datagen-config.schema.json b/scripts/schemavalidation/datagen-config.schema.json
index 08d564b..a74d79f 100644
--- a/scripts/schemavalidation/datagen-config.schema.json
+++ b/scripts/schemavalidation/datagen-config.schema.json
@@ -1,5 +1,5 @@
  {
-    "$comment": "Copyright (c) 2023, ARM Limited.",
+    "$comment": "Copyright (c) 2023-2024, ARM Limited.",
     "$comment": "SPDX-License-Identifier: Apache-2.0",
     "$id": "datagen-config.schema.json",
     "$schema": "https://json-schema.org/draft/2020-12/schema",
@@ -147,6 +147,23 @@
                             },
                             "additionalProperties": false,
                             "required": [ "sub_generator" ]
+                        },
+                        "fixed_data_info": {
+                            "description": "info required for FIXED_DATA generator",
+                            "type": "object",
+                            "properties":
+                            {
+                                "data": {
+                                    "description": "flattened fixed data",
+                                    "type": "array",
+                                    "items": {
+                                        "description": "integer tensor data",
+                                        "type": "integer"
+                                    }
+                                }
+                            },
+                            "additionalProperties": false,
+                            "required": [ "data" ]
                         }
                     },
                     "additionalProperties": false,
diff --git a/verif/generator/datagenerator.py b/verif/generator/datagenerator.py
index b5ef35d..743475c 100644
--- a/verif/generator/datagenerator.py
+++ b/verif/generator/datagenerator.py
@@ -78,7 +78,7 @@ class GenerateLibrary:
             size_bytes = size * 2
             # Create buffer of bytes and initialize to zero
             buffer = (ct.c_ubyte * size_bytes)(0)
-        elif dtype == "INT32":
+        elif dtype == "INT32" or dtype == "SHAPE":
             # Create buffer and initialize to zero
             buffer = (ct.c_int32 * size)(0)
             size_bytes = size * 4
diff --git a/verif/generator/tosa_arg_gen.py b/verif/generator/tosa_arg_gen.py
index 00490fa..a655a50 100644
--- a/verif/generator/tosa_arg_gen.py
+++ b/verif/generator/tosa_arg_gen.py
@@ -731,7 +731,11 @@ class TosaTensorValuesGen:
                     # Change from inclusive to exclusive range
                     data_range = (data_range[0], data_range[1] + 1)
                 # Ignore lazy data gen option and create data array using any range limits
-                arr = testGen.getRandTensor(shape, dtype, data_range)
+
+                if "fixed_data" in argsDict and argsDict["fixed_data"][idx] is not None:
+                    arr = np.int64(argsDict["fixed_data"][idx])
+                else:
+                    arr = testGen.getRandTensor(shape, dtype, data_range)
                 if roundMode:
                     arr = np.round(arr)
                 if idx < pCount:
@@ -751,7 +755,13 @@ class TosaTensorValuesGen:
         for idx, shape in enumerate(shapeList):
 
             tens_meta = {}
-            tens_meta["generator"] = gtu.DataGenType(dg_type).name
+            if "fixed_data" in argsDict and argsDict["fixed_data"][idx] is not None:
+                tens_meta["generator"] = gtu.DataGenType(
+                    gtu.DataGenType.FIXED_DATA
+                ).name
+            else:
+                tens_meta["generator"] = gtu.DataGenType(dg_type).name
+
             tens_meta["data_type"] = gtu.DTYPE_ATTRIBUTES[dtypeList[idx]]["json"]
             tens_meta["shape"] = [int(i) for i in shape]
             tens_meta["input_pos"] = idx
@@ -764,23 +774,30 @@ class TosaTensorValuesGen:
 
             if dg_type == gtu.DataGenType.PSEUDO_RANDOM:
                 info = {}
-                # TODO - generate seed for this generator based on test
-                info["rng_seed"] = 42
-
-                data_range = None
-                if "data_range_list" in argsDict:
-                    data_range = argsDict["data_range_list"][idx]["range"]
-                    if "round" in argsDict["data_range_list"][idx]:
-                        info["round"] = argsDict["data_range_list"][idx]["round"]
-                elif "data_range" in argsDict:
-                    data_range = argsDict["data_range"]
-
-                if data_range is None:
-                    data_range = testGen.getDTypeRange(
-                        dtypeList[idx], high_inclusive=True
-                    )
-                info["range"] = [str(v) for v in data_range]
-                tens_meta["pseudo_random_info"] = info
+                if (
+                    tens_meta["generator"]
+                    == gtu.DataGenType(gtu.DataGenType.FIXED_DATA).name
+                ):
+                    info["data"] = [int(i) for i in argsDict["fixed_data"][idx]]
+                    tens_meta["fixed_data_info"] = info
+                else:
+                    # TODO - generate seed for this generator based on test
+                    info["rng_seed"] = 42
+
+                    data_range = None
+                    if "data_range_list" in argsDict:
+                        data_range = argsDict["data_range_list"][idx]["range"]
+                        if "round" in argsDict["data_range_list"][idx]:
+                            info["round"] = argsDict["data_range_list"][idx]["round"]
+                    elif "data_range" in argsDict:
+                        data_range = argsDict["data_range"]
+
+                    if data_range is None:
+                        data_range = testGen.getDTypeRange(
+                            dtypeList[idx], high_inclusive=True
+                        )
+                    info["range"] = [str(v) for v in data_range]
+                    tens_meta["pseudo_random_info"] = info
             elif dg_type == gtu.DataGenType.DOT_PRODUCT:
                 info = {}
                 info["s"] = argsDict["s"]
@@ -812,6 +829,9 @@ class TosaTensorValuesGen:
                 dg_tens_meta[temp_name] = tens_meta
                 # Create data now using the temporary name to access meta details
                 data = testGen.dgl.get_tensor_data(temp_name, tens_data)
+                if tens_meta["data_type"] == "SHAPE":
+                    # Tensor type SHAPE and Numpy file type must be the same
+                    data = np.int64(data)
                 # Remove the item as we will give it the correct name later
                 del dg_tens_meta[temp_name]
 
@@ -1014,6 +1034,27 @@ class TosaTensorValuesGen:
         return placeholders
 
     @staticmethod
+    def tvgReshape(testGen, op, dtypeList, shapeList, argsDict, error_name=None):
+        dtypeList[1] = DType.SHAPE
+        shapeList[1] = [len(argsDict["new_shape"])]
+        # Create a new list for the pre-generated data in argsDict["fixed_data"]
+        argsDict["fixed_data"] = [None, argsDict["new_shape"]]
+
+        return TosaTensorValuesGen.tvgLazyGenDefault(
+            testGen, op, dtypeList, shapeList, argsDict, error_name
+        )
+
+    @staticmethod
+    def tvgTile(testGen, op, dtypeList, shapeList, argsDict, error_name=None):
+        dtypeList[1] = DType.SHAPE
+        shapeList[1] = [len(argsDict["multiples"])]
+        argsDict["fixed_data"] = [None, argsDict["multiples"]]
+
+        return TosaTensorValuesGen.tvgLazyGenDefault(
+            testGen, op, dtypeList, shapeList, argsDict, error_name
+        )
+
+    @staticmethod
     def tvgSelect(testGen, opName, dtypeList, shapeList, argsDict, error_name=None):
         # Set datatype of condition tensor to boolean
         dtypeList[0] = DType.BOOL
diff --git a/verif/generator/tosa_test_gen.py b/verif/generator/tosa_test_gen.py
index 67ac367..159ee83 100644
--- a/verif/generator/tosa_test_gen.py
+++ b/verif/generator/tosa_test_gen.py
@@ -1558,22 +1558,14 @@ class TosaTestGen:
     ):
         assert len(inputs) == 2
         a = inputs[0]
-        # second input is not properly generated yet
-        # new_shape = inputs[1]
-
-        # modify inputs[1] by a shape tensor from new_shape arg value
-        new_shape_attr = args_dict["new_shape"]
-        shape_array = np.array(new_shape_attr)
-        shape = shape_array.shape
-        new_shape = self.ser.addPlaceholder(shape, DType.SHAPE, shape_array)
-        inputs[1] = new_shape
-
+        shape = inputs[1]
+        shape_attr = args_dict["new_shape"]
         result_tensor = OutputShaper.reshapeOp(
-            self.ser, self.rng, a, new_shape_attr, error_name
+            self.ser, self.rng, a, shape_attr, error_name
         )
 
         # Invalidate Input/Output list for error if checks.
-        input_list = [a.name, new_shape.name]
+        input_list = [a.name, shape.name]
         output_list = [result_tensor.name]
         pCount, cCount = op["operands"]
         num_operands = pCount + cCount
@@ -1725,16 +1717,8 @@ class TosaTestGen:
     ):
         assert len(inputs) == 2
         a = inputs[0]
-        # second input is not properly generated yet
-        # multiples = inputs[1]
-
-        # modify inputs[1] by a shape tensor from multiples arg value
+        multiples = inputs[1]
         multiples_attr = args_dict["multiples"]
-        shape_array = np.int64(np.array(multiples_attr))
-        shape = shape_array.shape
-        multiples = self.ser.addPlaceholder(shape, DType.SHAPE, shape_array)
-        inputs[1] = multiples
-
         result_tensor = OutputShaper.tileOp(
             self.ser, self.rng, a, multiples_attr, error_name
         )
@@ -4236,7 +4220,7 @@ class TosaTestGen:
             "build_fcn": (
                 build_reshape,
                 TosaTensorGen.tgBasic,
-                TosaTensorValuesGen.tvgLazyGenDefault,
+                TosaTensorValuesGen.tvgReshape,
                 TosaArgGen.agReshape,
             ),
             "types": TYPE_FIB,
@@ -4302,7 +4286,7 @@ class TosaTestGen:
             "build_fcn": (
                 build_tile,
                 TosaTensorGen.tgBasic,
-                TosaTensorValuesGen.tvgLazyGenDefault,
+                TosaTensorValuesGen.tvgTile,
                 TosaArgGen.agTile,
             ),
             "types": TYPE_FIB,
diff --git a/verif/generator/tosa_utils.py b/verif/generator/tosa_utils.py
index 3d733f4..33db95f 100644
--- a/verif/generator/tosa_utils.py
+++ b/verif/generator/tosa_utils.py
@@ -1,4 +1,4 @@
-# Copyright (c) 2021-2023, ARM Limited.
+# Copyright (c) 2021-2024, ARM Limited.
 # SPDX-License-Identifier: Apache-2.0
 import struct
 import sys
@@ -49,13 +49,14 @@ class DataGenType(IntEnum):
     OP_BOUNDARY = 2
     OP_FULLSET = 3
     OP_SPECIAL = 4
+    FIXED_DATA = 5
 
 
 def dtypeIsSupportedByCompliance(dtype):
     """Types supported by the new data generation and compliance flow."""
     if isinstance(dtype, list) or isinstance(dtype, tuple):
         dtype = dtype[0]
-    return dtype in (DType.FP32, DType.FP16)
+    return dtype in (DType.FP32, DType.FP16, DType.SHAPE)
 
 
 def getOpNameFromOpListName(opName):