[flang-commits] [flang] [flang][cuda] Add entry point for alloc/free and simple copy (PR #109867)

Slava Zakharin via flang-commits flang-commits at lists.llvm.org
Tue Sep 24 14:30:08 PDT 2024


================
@@ -17,12 +17,24 @@ namespace Fortran::runtime::cuda {
 
 extern "C" {
 
+/// Allocate memory on the device.
+void *RTDECL(CUFMemAlloc)(
----------------
vzakhari wrote:

Do you plan to have different APIs for different CUF attributes or should we pass the device/managed/etc. attribute to the APIs to use different allocation/free methods?

https://github.com/llvm/llvm-project/pull/109867


More information about the flang-commits mailing list