Disabled native dp4a in Intel CPU Runtime for OpenCL because it is slower than emulated dp4a

Author: ProjectPhysX

src/opencl.hpp

@@ -154,12 +154,13 @@ struct Device_Info {
 			cores_per_cu = is_gpu ? (intel_16_cores_per_cu ? 16.0f : 8.0f) : 0.5f; // Intel GPUs have 16 cores/CU (PVC) or 8 cores/CU (integrated/Arc), Intel CPUs (with HT) have 1/2 core/CU
 			if(is_gpu&&!uses_ram) { // fix wrong global memory capacity reporting for Intel dGPUs
 #if defined(_WIN32)
-				memory = (uint)((cl_device.getInfo<CL_DEVICE_GLOBAL_MEM_SIZE>()*50ull/49ull)/1048576ull); // 98% on Windows https://github.com/intel/compute-runtime/blob/master/shared/source/os_interface/windows/wddm_memory_manager.cpp#L964
+				memory = (uint)((cl_device.getInfo<CL_DEVICE_GLOBAL_MEM_SIZE>()*50ull/49ull)/1048576ull); // 98% on Windows https://github.com/intel/compute-runtime/blob/master/shared/source/os_interface/windows/wddm_memory_manager.cpp#L958
 #elif defined(__linux__)
-				memory = (uint)((cl_device.getInfo<CL_DEVICE_GLOBAL_MEM_SIZE>()*20ull/19ull)/1048576ull); // 95% on Linux   https://github.com/intel/compute-runtime/blob/master/shared/source/os_interface/linux/drm_memory_manager.cpp#L1424
+				memory = (uint)((cl_device.getInfo<CL_DEVICE_GLOBAL_MEM_SIZE>()*20ull/19ull)/1048576ull); // 95% on Linux   https://github.com/intel/compute-runtime/blob/master/shared/source/os_interface/linux/drm_memory_manager.cpp#L1521
 #endif // Linux
 			}
 			patch_intel_gpu_above_4gb = patch_intel_gpu_above_4gb||(is_gpu&&memory>4096u); // enable memory allocations greater than 4GB for Intel GPUs with >4GB VRAM
+			if(is_cpu) is_dp4a_capable = 0u; // native dp4a in Intel CPU Runtime for OpenCL is slower than emulated dp4a
 		} else if(vendor_id==0x10DE||vendor_id==0x13B5) { // Nvidia GPU/CPU
 			nvidia_compute_capability = 10u*(uint)cl_device.getInfo<CL_DEVICE_COMPUTE_CAPABILITY_MAJOR_NV>()+(uint)cl_device.getInfo<CL_DEVICE_COMPUTE_CAPABILITY_MINOR_NV>();
 			const bool nvidia__32_cores_per_cu = (nvidia_compute_capability <30); // identify Fermi GPUs