[GitHub] [tvm] masahi commented on a change in pull request #7572: [Vulkan] Support passing 64 bit scalar

[GitBox <gi...@apache.org>]

masahi commented on a change in pull request #7572:
URL: https://github.com/apache/tvm/pull/7572#discussion_r586022330



##########
File path: src/runtime/pack_args.h
##########
@@ -177,25 +188,28 @@ template <int N, typename F>
 inline PackedFunc PackFuncNonBufferArg_(F f, int base, const std::vector<ArgConvertCode>& codes) {
   int num_args = static_cast<int>(codes.size());
   auto ret = [f, codes, base, num_args](TVMArgs args, TVMRetValue* ret) {
-    TempArray<ArgUnion, N> holder_(num_args);
-    ArgUnion* holder = holder_.data();
+    TempArray<ArgUnion64, N> holder_(num_args);
+    ArgUnion64* holder = holder_.data();
     for (int i = 0; i < num_args; ++i) {
       switch (codes[i]) {
-        case INT64_TO_INT64:
+        case INT64_TO_INT64: {
+          holder[i].v_int64 = args.values[base + i].v_int64;
+          break;
+        }
         case FLOAT64_TO_FLOAT64: {
-          LOG(FATAL) << "Do not support 64bit argument to device function";
+          holder[i].v_float64 = args.values[base + i].v_float64;

Review comment:
       yes, previously both int64 and fp64 had an issue for the same reason, they are both supported now.




----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org