Fix cooperative launch APIs to set hipGetLastError (#1935)

* Fix cooperative launch APIs to set hipGetLastError

Previously, the cooperative launch APIs did not properly log their
errors in the global hipGetLastError variable before returning back
to the user. As such, the APIs would leave hipSuccess in the
last error, which would break some use cases.

This fixes that problem by making a trampoline function that does
the HIP_INIT_API and ihipLogStatus.

* Add missing flag to the log of multi-GPU launch

include/hip/hcc_detail/functional_grid_launch.hpp

@@ -37,14 +37,15 @@ THE SOFTWARE.
 hipError_t ihipExtLaunchMultiKernelMultiDevice(hipLaunchParams* launchParamsList, int numDevices,
                                                unsigned int flags, hip_impl::program_state& ps);
 
-hipError_t ihipLaunchCooperativeKernel(const void* f, dim3 gridDim, dim3 blockDimX, void** kernelParams,
-                unsigned int sharedMemBytes, hipStream_t stream, hip_impl::program_state& ps);
-
-hipError_t ihipLaunchCooperativeKernelMultiDevice(hipLaunchParams* launchParamsList, int  numDevices,
-                unsigned int  flags, hip_impl::program_state& ps);
-
-
+hipError_t hipLaunchCooperativeKernel(const void* f, dim3 gridDim,
+                                    dim3 blockDim, void** args,
+                                    size_t sharedMem, hipStream_t stream,
+                                    hip_impl::program_state& ps);
 
+hipError_t hipLaunchCooperativeKernelMultiDevice(hipLaunchParams* launchParamsList,
+                                                 int  numDevices,
+                                                 unsigned int flags,
+                                                 hip_impl::program_state& ps);
 
 #pragma GCC visibility push(hidden)
 
@@ -202,22 +203,24 @@ hipError_t hipExtLaunchMultiKernelMultiDevice(hipLaunchParams* launchParamsList,
 template <typename F>
 inline
 __attribute__((visibility("hidden")))
-hipError_t hipLaunchCooperativeKernel(F f, dim3 gridDim, dim3 blockDimX, void** kernelParams,
-                unsigned int sharedMemBytes, hipStream_t stream) {
-
+hipError_t hipLaunchCooperativeKernel(F f, dim3 gridDim, dim3 blockDim,
+                                      void** args, size_t sharedMem,
+                                      hipStream_t stream) {
     hip_impl::hip_init();
     auto& ps = hip_impl::get_program_state();
-    return ihipLaunchCooperativeKernel(reinterpret_cast<void*>(f), gridDim, blockDimX, kernelParams, sharedMemBytes, stream, ps);
+    return hipLaunchCooperativeKernel(reinterpret_cast<void*>(f), gridDim,
+                                      blockDim, args, sharedMem, stream, ps);
 }
 
 inline
 __attribute__((visibility("hidden")))
-hipError_t hipLaunchCooperativeKernelMultiDevice(hipLaunchParams* launchParamsList, int  numDevices,
-                unsigned int  flags) {
+hipError_t hipLaunchCooperativeKernelMultiDevice(hipLaunchParams* launchParamsList,
+                                                 int  numDevices,
+                                                 unsigned int  flags) {
 
     hip_impl::hip_init();
     auto& ps = hip_impl::get_program_state();
-    return ihipLaunchCooperativeKernelMultiDevice(launchParamsList, numDevices, flags, ps);
+    return hipLaunchCooperativeKernelMultiDevice(launchParamsList, numDevices, flags, ps);
 }
 
 #pragma GCC visibility pop