Update Paddle to 0116 with metax lock_runtime (#2355)

* [Metax] lock_runtime.c (#216)

* update Paddle to 0116

* fix yml

* update image

* remote test_fused_rms_norm_op.py

---------

Co-authored-by: duqimeng <[email protected]>

Author: YqGe585

.github/workflows/_IXUCA.yml

@@ -31,7 +31,7 @@ jobs:
     if:  ${{ needs.check-bypass.outputs.can-skip != 'true' && github.repository_owner == 'PaddlePaddle' }}
     runs-on: iluvatar-gpu-2
     container:
-      image: ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-ixuca:latest
+      image: ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-ixuca:3.3.0
       env:
         LD_LIBRARY_PATH: /usr/local/corex/lib
         LIBRARY_PATH: /usr/local/corex/lib
@@ -125,6 +125,7 @@ jobs:
           export FLAG_SKIP_FLOAT64=1
           cd PaddleCustomDevice/backends/iluvatar_gpu/tests
           # Run tests
+          ixsmi
           bash run_test.sh
 
       - name: Terminate and delete the container

Paddle

@@ -559,3 +559,4 @@ test_tensor.py
 test_rms_norm.py
 test_api_compatibility.py
 test_celu_op.py
+test_fused_rms_norm_op.py

backends/iluvatar_gpu/tests/disabled_test.txt

@@ -13,11 +13,13 @@
 // limitations under the License.
 
 #include "common/utils.h"
+#define _GNU_SOURCE
+#include <dlfcn.h>
+#include <stdio.h>
 
 #include "glog/logging.h"
 #include "paddle/phi/backends/context_pool.h"
 #include "paddle/phi/backends/custom/custom_context.h"
-
 namespace phi {
 namespace {
 C_Status AsyncMemCpyH2D(const C_Device device,
@@ -295,3 +297,14 @@ std::ostream& operator<<(std::ostream& os, const phi::DenseTensor& t) {
 #undef FOR_EACH_DATA_TYPE_TO_PRINT
 #undef CALL_PRINT_TENSOR
 }  // namespace phi
+
+// lock_mcruntime.c
+
+__attribute__((constructor)) void lock_mcruntime() {
+  void* handle =
+      dlopen("/opt/maca/lib/libmcruntime.so", RTLD_LAZY | RTLD_NODELETE);
+  if (!handle) {
+    fprintf(stderr, "Failed to lock libmcruntime.so: %s\n", dlerror());
+  } else {
+  }
+}