Add 16A8W support and test for mul operation

Add 16A8W quantization support and test for the mul operation in ExecutorTorch ARM backend.

This follows the pattern established for linear operations, extending int16 support to mul operations.

Changes:
- Add INT16 dtype validation support in op_mul.py
- Add test_mul_tensor_16a8w_tosa_INT test function
- Enable test_mul.py in test targets configuration

The 16A8W configuration uses 16-bit activations with 8-bit weights, enabling higher precision for activations while maintaining weight efficiency.

Differential Revision: [D80510628](https://our.internmc.facebook.com/intern/diff/D80510628/)

[ghstack-poisoned]

Author: Ninja91

backends/arm/operators/op_mul.py

@@ -51,7 +51,7 @@ def define_node(
         validate_valid_dtype(
             self.target,
             [*inputs, output],
-            [ts.DType.INT8, ts.DType.INT32],
+            [ts.DType.INT8, ts.DType.INT16, ts.DType.INT32],
             output.tosa_spec,
         )
 

backends/arm/test/ops/test_mul.py

@@ -8,16 +8,23 @@
 
 from typing import Tuple
 
+import pytest
 import torch
+from executorch.backends.arm.quantizer.arm_quantizer import (
+    get_symmetric_a16w8_quantization_config,
+    TOSAQuantizer,
+)
 
-from executorch.backends.arm.test import common
+from executorch.backends.arm.test import common, conftest
 from executorch.backends.arm.test.tester.test_pipeline import (
     EthosU55PipelineINT,
     EthosU85PipelineINT,
     TosaPipelineFP,
     TosaPipelineINT,
     VgfPipeline,
 )
+from executorch.backends.arm.tosa_specification import TosaSpecification
+from executorch.backends.xnnpack.test.tester import Quantize
 
 input_t1 = Tuple[torch.Tensor, torch.Tensor]  # Input x
 aten_op = "torch.ops.aten.mul.Tensor"
@@ -268,3 +275,102 @@ def test_mul_tensor_vgf_INT_int32(test_data: torch.Tensor):
     )
     pipeline.pop_stage("check.quant_nodes")
     pipeline.run()
+
+
+def get_symmetric_a16w8_mul_quantizer(per_channel_quantization=False):
+    tosa_version = conftest.get_option("tosa_version")
+    tosa_profiles = {
+        "1.0": TosaSpecification.create_from_string("TOSA-1.0+INT+int16"),
+    }
+
+    quantizer = TOSAQuantizer(tosa_profiles[tosa_version])
+    quantizer.set_global(
+        get_symmetric_a16w8_quantization_config(is_per_channel=per_channel_quantization)
+    )
+
+    return Quantize(
+        quantizer,
+        get_symmetric_a16w8_quantization_config(
+            is_per_channel=per_channel_quantization
+        ),
+    )
+
+
+@common.parametrize("test_data", test_data_suite)
+def test_mul_tensor_16a8w_tosa_INT(test_data: input_t1):
+    """Test mul operation with 16A8W quantization (16-bit activations, 8-bit weights)"""
+    per_channel_quantization = False
+
+    pipeline = TosaPipelineINT[input_t1](
+        Mul(),
+        test_data(),
+        aten_op,
+        exir_op=[],
+        per_channel_quantization=per_channel_quantization,
+        use_to_edge_transform_and_lower=True,
+        tosa_extensions=["int16"],
+    )
+
+    pipeline.change_args(
+        "quantize",
+        get_symmetric_a16w8_mul_quantizer(
+            per_channel_quantization=per_channel_quantization
+        ),
+    )
+    pipeline.run()
+
+
+@common.parametrize("test_data", test_data_suite)
+@common.XfailIfNoCorstone300
+@pytest.mark.xfail(
+    reason="Vela compilation fails with 'Invalid arguments' for int16 mul operations"
+)
+def test_mul_tensor_16a8w_u55_INT16(test_data: input_t1):
+    """Test mul operation with 16A8W quantization on U55 (16-bit activations, 8-bit weights)"""
+    per_channel_quantization = False
+
+    pipeline = EthosU55PipelineINT[input_t1](
+        Mul(),
+        test_data(),
+        aten_op,
+        exir_ops=[],
+        per_channel_quantization=per_channel_quantization,
+        use_to_edge_transform_and_lower=True,
+        run_on_fvp=True,
+    )
+
+    pipeline.change_args(
+        "quantize",
+        get_symmetric_a16w8_mul_quantizer(
+            per_channel_quantization=per_channel_quantization
+        ),
+    )
+    pipeline.run()
+
+
+@common.parametrize("test_data", test_data_suite)
+@common.XfailIfNoCorstone320
+@pytest.mark.xfail(
+    reason="Vela compilation fails with 'Invalid arguments' for int16 mul operations"
+)
+def test_mul_tensor_16a8w_u85_INT16(test_data: input_t1):
+    """Test mul operation with 16A8W quantization on U85 (16-bit activations, 8-bit weights)"""
+    per_channel_quantization = False
+
+    pipeline = EthosU85PipelineINT[input_t1](
+        Mul(),
+        test_data(),
+        aten_op,
+        exir_ops=[],
+        per_channel_quantization=per_channel_quantization,
+        use_to_edge_transform_and_lower=True,
+        run_on_fvp=True,
+    )
+
+    pipeline.change_args(
+        "quantize",
+        get_symmetric_a16w8_mul_quantizer(
+            per_channel_quantization=per_channel_quantization
+        ),
+    )
+    pipeline.run()

backends/arm/test/targets.bzl

@@ -16,6 +16,7 @@ def define_arm_tests():
         "ops/test_add.py",
         "ops/test_avg_pool2d.py",
         "ops/test_linear.py", 
+        "ops/test_mul.py",
         "ops/test_slice.py",
         "ops/test_sigmoid.py",
         "ops/test_tanh.py",