virtual int2 split_kernel_global_size(device_memory& kg, device_memory& data, DeviceTask * /*task*/) { cl_device_type type = OpenCLInfo::get_device_type(device->cdDevice); /* Use small global size on CPU devices as it seems to be much faster. */ if(type == CL_DEVICE_TYPE_CPU) { VLOG(1) << "Global size: (64, 64)."; return make_int2(64, 64); } cl_ulong max_buffer_size; clGetDeviceInfo(device->cdDevice, CL_DEVICE_MAX_MEM_ALLOC_SIZE, sizeof(cl_ulong), &max_buffer_size, NULL); if(DebugFlags().opencl.mem_limit) { max_buffer_size = min(max_buffer_size, cl_ulong(DebugFlags().opencl.mem_limit - device->stats.mem_used)); } VLOG(1) << "Maximum device allocation size: " << string_human_readable_number(max_buffer_size) << " bytes. (" << string_human_readable_size(max_buffer_size) << ")."; /* Limit to 2gb, as we shouldn't need more than that and some devices may support much more. */ max_buffer_size = min(max_buffer_size / 2, (cl_ulong)2l*1024*1024*1024); size_t num_elements = max_elements_for_max_buffer_size(kg, data, max_buffer_size); int2 global_size = make_int2(max(round_down((int)sqrt(num_elements), 64), 64), (int)sqrt(num_elements)); VLOG(1) << "Global size: " << global_size << "."; return global_size; }
virtual int2 split_kernel_global_size(device_memory& kg, device_memory& data, DeviceTask * /*task*/) { cl_device_type type = OpenCLInfo::get_device_type(device->cdDevice); /* Use small global size on CPU devices as it seems to be much faster. */ if(type == CL_DEVICE_TYPE_CPU) { VLOG(1) << "Global size: (64, 64)."; return make_int2(64, 64); } cl_ulong max_buffer_size; clGetDeviceInfo(device->cdDevice, CL_DEVICE_MAX_MEM_ALLOC_SIZE, sizeof(cl_ulong), &max_buffer_size, NULL); VLOG(1) << "Maximum device allocation size: " << string_human_readable_number(max_buffer_size) << " bytes. (" << string_human_readable_size(max_buffer_size) << ")."; size_t num_elements = max_elements_for_max_buffer_size(kg, data, max_buffer_size / 2); int2 global_size = make_int2(round_down((int)sqrt(num_elements), 64), (int)sqrt(num_elements)); VLOG(1) << "Global size: " << global_size << "."; return global_size; }