kernel

1 files changed, 20 insertions, 11 deletions

diff --git a/backend/src/driver/cl_gen_kernel.cpp b/backend/src/driver/cl_gen_kernel.cpp
index cfde43e5..c9ad3a39 100644
--- a/backend/src/driver/cl_gen_kernel.cpp
+++ b/backend/src/driver/cl_gen_kernel.cpp
@@ -79,19 +79,12 @@ static size_t genGetKernelMaxWorkGroupSize(cl_kernel kernel, Kernel* ker, const
 }
 
 extern "C"
-cl_int GenCreateKernel(cl_kernel kernel, const cl_device_id device)
+cl_int GenGetWorkgroupInfo(cl_kernel kernel, const cl_device_id device, cl_kernel_workgroup_info wgInfo)
 {
-  GBE_ASSERT(getGenKernelPrivate(kernel, device) == NULL);
-
-  Program* p = reinterpret_cast<Program*>(getGenProgramPrivate(kernel->program, device));
-  if (p == NULL)
-    return CL_INVALID_PROGRAM;
-
-  Kernel* ker = p->getKernel(kernel->name);
+  Kernel* ker = reinterpret_cast<Kernel*>(getGenKernelPrivate(kernel, device));
   if (ker == NULL)
-    return CL_INVALID_KERNEL_NAME;
+    return CL_INVALID_VALUE;
 
-  cl_kernel_workgroup_info wgInfo = &kernel->wg_info[findIndexByDevice(kernel->program->ctx, device)];
   ker->getCompileWorkGroupSize(wgInfo->compile_wg_sz);
   wgInfo->local_mem_sz = ker->getSLMSize();
   wgInfo->private_mem_sz = ker->getStackSize();
@@ -108,6 +101,22 @@ cl_int GenCreateKernel(cl_kernel kernel, const cl_device_id device)
     memcpy(wgInfo->global_work_sz, device->max_3d_global_work_sizes, sizeof(device->max_3d_global_work_sizes));
   }
 
+  return CL_SUCCESS;
+}
+
+extern "C"
+cl_int GenCreateKernel(cl_kernel kernel, const cl_device_id device)
+{
+  GBE_ASSERT(getGenKernelPrivate(kernel, device) == NULL);
+
+  Program* p = reinterpret_cast<Program*>(getGenProgramPrivate(kernel->program, device));
+  if (p == NULL)
+    return CL_INVALID_PROGRAM;
+
+  Kernel* ker = p->getKernel(kernel->name);
+  if (ker == NULL)
+    return CL_INVALID_KERNEL_NAME;
+
   setGenKernelPrivate(kernel, device, ker);
   return CL_SUCCESS;
 }
@@ -750,7 +759,7 @@ cl_int GenEnqueueNDRangeKernel(cl_command_queue queue, cl_kernel kernel, const u
   }
 
   GenGPUWorkItemNDRange* ndRange = GBE_NEW(GenGPUWorkItemNDRange, gpuQueue->bufmgr, gpuQueue->ctx,
-                                   gpuDev->device_id, 
+                                   gpuDev->device_id,
                                    event_ret, events, num_events);
   if (ndRange->gpuState->stateInit(
         queue->device->max_compute_unit * gpuDev->max_thread_per_unit, cst_sz / 32) != true) {