Fix C-style hipLaunchKernel (#1835)

* Fix bug in LaunchKernel test
Instead of passing the address of the gpu buffer, pass the address
of the pointer that holds the address of the gpu buffer

* Fix hipLaunchKernel's kernarg buffer construction.
The hipLaunchKernel implementation should rely on ihipModuleLaunchKernel
to construct the kernarg buffer correctly based on kernel metadata.

* Fix a bug in get_functions where the Kernel_descriptor wasn't constructed with the correct kernarg layout information.

* Fix a bug in kernarg layout parsing dealing with kernel without any arg

* teach ihipModuleLaunchKernel to handle kernel without any arg

* Add a more interesting test

hipamd/src/hip_module.cpp

@@ -188,7 +188,8 @@ hipError_t ihipModuleLaunchKernel(TlsData *tls, hipFunction_t f, uint32_t gridSi
                 return hipErrorNotInitialized;
             }
 
-        } else {
+        } 
+        else if (f->_kernarg_layout.size() != 0) {
             return hipErrorInvalidValue;
         }
 
@@ -1470,21 +1471,7 @@ hipError_t hipLaunchKernel(
    hipFunction_t kd = hip_impl::get_program_state().kernel_descriptor((std::uintptr_t)func_addr,
                                                            hip_impl::target_agent(stream));
 
-   if(kd == nullptr || kd->_header == nullptr)
-       return ihipLogStatus(hipErrorInvalidValue);
-
-   size_t szKernArg = kd->_header->kernarg_segment_byte_size;
-
-   if(args == NULL && szKernArg != 0)
-      return ihipLogStatus(hipErrorInvalidValue);
-
-   void* config[]{
-        HIP_LAUNCH_PARAM_BUFFER_POINTER,
-        args,
-        HIP_LAUNCH_PARAM_BUFFER_SIZE,
-	    &szKernArg,
-        HIP_LAUNCH_PARAM_END};
-
-   return ihipLogStatus(ihipModuleLaunchKernel(tls, kd, numBlocks.x, numBlocks.y, numBlocks.z,
-                          dimBlocks.x, dimBlocks.y, dimBlocks.z, sharedMemBytes, stream, nullptr, (void**)&config, nullptr, nullptr, 0));
+   return hipModuleLaunchKernel(kd, numBlocks.x, numBlocks.y, numBlocks.z,
+                          dimBlocks.x, dimBlocks.y, dimBlocks.z, sharedMemBytes,
+                          stream, args, nullptr);
 }

hipamd/src/program_state.inl

@@ -613,7 +613,8 @@ public:
                 for (auto&& kernel_symbol : it->second) {
                     functions[aa].second.emplace(
                         function.first,
-                        Kernel_descriptor{kernel_object(kernel_symbol), it->first});
+                        Kernel_descriptor{kernel_object(kernel_symbol), it->first,
+                                          kernargs_size_align(function.first)});
                 }
             }
         }, agent);
@@ -672,11 +673,12 @@ public:
             auto dx1 = kernels_md.find("CodeProps", dx);
             dx = kernels_md.find("Args:", dx);
 
-            if (dx1 < dx) {
+            if (dx1 < dx || dx == std::string::npos) {
                 dx = dx1;
+                // create an empty kernarg laybout vector for kernels without any arg 
+                kernargs[fn];
                 continue;
             }
-            if (dx == std::string::npos) break;
 
             static constexpr decltype(kernels_md.size()) args_sz{5};
             dx = parse_args_v2(kernels_md, dx + args_sz, dx1, kernargs[fn]);