Rename transformOutputFromAllocationToLogical (#3341)

csrc/runtime/allocations.cpp

@@ -671,12 +671,11 @@ class BackwardTraverseFromAllocToLogical {
 // Another example, if the logical domain is [I1*I2] and the allocation domain
 // is [I1, I2], then we will allocate as [I1, I2] and do a tensor.view(I1*I2) to
 // get a tensor whose semantics is [I1*I2] but memory is [I1,I2]
-at::Tensor transformOutputFromAllocationToLogical(
+at::Tensor transformFromAllocationToLogical(
     at::Tensor tensor,
     TensorView* tv,
     ExpressionEvaluator& ee) {
-  FUSER_PERF_SCOPE(
-      "fusion_executor::allocations::transformOutputFromAllocationToLogical");
+  FUSER_PERF_SCOPE("allocations::transformFromAllocationToLogical");
   // Ignore reductions because reductions does not exist in tensor's definition
   auto logical = TensorDomain::noReductions(tv->getLogicalDomain());
   auto alloc = TensorDomain::noReductions(tv->getMaybeAllocationDomain());
@@ -751,9 +750,8 @@ std::pair<std::vector<int64_t>, std::vector<int64_t>> inferShapeOfOutput(
       at::empty_strided(size_stride.first, size_stride.second, options);
   // TODO(jiej): we should refactor it here, there's no need to use
   // meta_tensor at all, size + stride should be used directly in the
-  // `transformOutputFromAllocationToLogical`
-  meta_tensor =
-      transformOutputFromAllocationToLogical(meta_tensor, tv, expr_eval);
+  // `transformFromAllocationToLogical`
+  meta_tensor = transformFromAllocationToLogical(meta_tensor, tv, expr_eval);
   return {meta_tensor.sizes().vec(), meta_tensor.strides().vec()};
 }