enable fp32 cat slice

Differential Revision: D70216001

Pull Request resolved: pytorch/executorch#8759

Author: zonglinpeng

backends/cadence/fusion_g3/operators/op_cat.cpp

@@ -115,7 +115,8 @@ Tensor& cat_out(
       (out.scalar_type() == ScalarType::Char) ||
       (out.scalar_type() == ScalarType::UInt32) ||
       (out.scalar_type() == ScalarType::UInt16) ||
-      (out.scalar_type() == ScalarType::Byte)) {
+      (out.scalar_type() == ScalarType::Byte) ||
+      (out.scalar_type() == ScalarType::Float)) {
     XT_KERNEL_CHECK(
         ctx,
         out,

backends/cadence/fusion_g3/operators/op_slice_copy.cpp

@@ -101,7 +101,8 @@ Tensor& slice_copy_Tensor_out(
        (out.scalar_type() == ScalarType::Char) ||
        (out.scalar_type() == ScalarType::UInt32) ||
        (out.scalar_type() == ScalarType::UInt16) ||
-       (out.scalar_type() == ScalarType::Byte))) {
+       (out.scalar_type() == ScalarType::Byte) ||
+       (out.scalar_type() == ScalarType::Float))) {
     XT_KERNEL_CHECK(
         ctx,
         out,
@@ -132,4 +133,4 @@ Tensor& slice_copy_Tensor_out(
 } // namespace native
 } // namespace G3
 } // namespace impl
-} // namespace cadence
+} // namespace cadence

backends/cadence/fusion_g3/operators/xt_utils.h

@@ -19,6 +19,8 @@ inline int get_element_size(ScalarType dtype) {
     return sizeof(short);
   } else if ((dtype == ScalarType::Char) || (dtype == ScalarType::Byte)) {
     return sizeof(char);
+  } else if (dtype == ScalarType::Float) {
+    return sizeof(float);
   }
   return 0;
 }

backends/cadence/utils/facto_util.py

@@ -18,6 +18,7 @@
 
 # seed to generate identical cases every run to reproduce from bisect
 random_manager.seed(1729)
+MAX_CASES = 50
 
 
 def apply_tensor_contraints(op_name: str, tensor_constraints: list[object]) -> None:
@@ -46,6 +47,14 @@ def apply_tensor_contraints(op_name: str, tensor_constraints: list[object]) -> N
                     cp.Value.Le(lambda deps, dtype, struct: 2**2),
                 ]
             )
+        case "slice_copy.Tensor":
+            tensor_constraints.extend(
+                [
+                    cp.Rank.Le(lambda deps: 2),
+                    cp.Value.Ge(lambda deps, dtype, struct: 1),
+                    cp.Value.Le(lambda deps, dtype, struct: 2),
+                ]
+            )
         case _:
             tensor_constraints.extend(
                 [
@@ -124,4 +133,4 @@ def facto_testcase_gen(op_name: str) -> List[Tuple[List[str], OrderedDict[str, s
     return [
         (posargs, inkwargs)
         for posargs, inkwargs, _ in ArgumentTupleGenerator(spec).gen()
-    ]
+    ][:MAX_CASES]