Minor comment on how memory is allocated for scalar in py_divide_by_scalar

ndgrigorian · ndgrigorian · commit a44fb2f4c0d5 · 2024-09-04T10:17:10.000-07:00
diff --git a/dpctl/tensor/libtensor/source/elementwise_functions/true_divide.cpp b/dpctl/tensor/libtensor/source/elementwise_functions/true_divide.cpp
@@ -306,6 +306,7 @@ py_divide_by_scalar(const dpctl::tensor::usm_ndarray &src,
     alignas(double) char scalar_alloc[sizeof(double)] = {0};
 
     divide_by_scalar_fn_ptr_t fn;
+    // placement new into stack memory means no call to delete is necessary
     switch (src_typeid) {
     case float16_typeid:
     {

Original file line number	Diff line number	Diff line change
`@@ -306,6 +306,7 @@ py_divide_by_scalar(const dpctl::tensor::usm_ndarray &src,`
`306`	`306`	`alignas(double) char scalar_alloc[sizeof(double)] = {0};`
`307`	`307`
`308`	`308`	`divide_by_scalar_fn_ptr_t fn;`
	`309`	`+ // placement new into stack memory means no call to delete is necessary`
`309`	`310`	`switch (src_typeid) {`
`310`	`311`	`case float16_typeid:`
`311`	`312`	`{`