IntenseWebs/openvino
3
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[GPU] Detect dpas support (#7643)

Browse Source 
* Detect dpas support
* New debug flag DISABLE_ONEDNN is added
This commit is contained in:
Mingyu Kim 2021-09-27 11:38:57 +09:00 committed by GitHub
parent 8262aed30c
commit 95f8544aa6
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
4 changed files with 7 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
inference-engine/thirdparty/clDNN/api/cldnn/runtime/debug_configuration.hpp vendored
View File

@ -27,6 +27,7 @@ public:
int verbose; // Verbose execution
int print_multi_kernel_perf; // Print execution time of each kernel in multi-kernel primitimive
int disable_usm; // Disable usm usage
int disable_onednn; // Disable onednn for discrete GPU (no effect for integrated GPU)
std::string dump_graphs; // Dump optimized graph
std::string dump_sources; // Dump opencl sources
std::string dump_layers_path; // Enable dumping intermediate buffers and set the dest path

5
inference-engine/thirdparty/clDNN/runtime/debug_configuration.cpp vendored
View File

@ -102,7 +102,8 @@ debug_configuration::debug_configuration()
, dump_sources(std::string())
, dump_layers_path(std::string())
, dump_layers(std::string())
, dump_layers_dst_only(0) {
, dump_layers_dst_only(0)
, disable_onednn(0) {
#ifdef GPU_DEBUG_CONFIG
get_common_debug_env_var("Verbose", verbose);
get_gpu_debug_env_var("PrintMultiKernelPerf", print_multi_kernel_perf);
@ -112,6 +113,8 @@ debug_configuration::debug_configuration()
get_gpu_debug_env_var("DumpLayersPath", dump_layers_path);
get_gpu_debug_env_var("DumpLayers", dump_layers);
get_gpu_debug_env_var("DumpLayersDstOnly", dump_layers_dst_only);
get_gpu_debug_env_var("DisableOnednn", disable_onednn);
if (dump_layers_path.length() > 0 && !disable_usm) {
disable_usm = 1;
GPU_DEBUG_COUT << "DisableUsm=1 because of DumpLayersPath" << std::endl;

1
inference-engine/thirdparty/clDNN/runtime/ocl/ocl_device.cpp vendored
View File

@ -246,6 +246,7 @@ device_info init_device_info(const cl::Device& device) {
auto features = device.getInfo<CL_DEVICE_FEATURE_CAPABILITIES_INTEL>();
info.supports_imad = info.supports_imad || (features & CL_DEVICE_FEATURE_FLAG_DP4A_INTEL);
info.supports_immad = info.supports_immad || (features & CL_DEVICE_FEATURE_FLAG_DPAS_INTEL);
} else {
info.gfx_ver = {0, 0, 0};
info.device_id = driver_dev_id();

1
inference-engine/thirdparty/clDNN/runtime/ocl/ocl_ext.hpp vendored
View File

@ -33,6 +33,7 @@ typedef cl_bitfield cl_device_feature_capabilities_intel;
/* For GPU devices, version 1.0.0: */
#define CL_DEVICE_FEATURE_FLAG_DP4A_INTEL (1 << 0)
#define CL_DEVICE_FEATURE_FLAG_DPAS_INTEL (1 << 1)
namespace cl {