rsmi_init: Do not complain loudly when no driver is found

When librocm-smi is pulled through a dependency, we may end up on a system
without actual hardware supported by ROCM, and rsmi_init() failing is
actually expected, we do want to frighten the user in such a case.


[ROCm/rocm_smi_lib commit: 8ca4207d5c]

projects/rocm-smi-lib/src/rocm_smi.cc

@@ -485,6 +485,20 @@ rsmi_init(uint64_t flags) {
   if (smi.ref_count() == 1) {
     try {
       smi.Initialize(flags);
+    } catch(const amd::smi::rsmi_exception& e) {
+      smi.Cleanup();
+      if (e.error_code() == RSMI_INITIALIZATION_ERROR &&
+          !strcmp(e.what(),
+               "Failed to initialize rocm_smi library (KFD node discovery).")) {
+        // This system does not actually have ROCM drivers set up
+        // We were probably just called through dependency, just report the
+        // error and log without complaining loudly.
+        std::ostringstream ss;
+        ss << "Exception caught: " << e.what() << ".";
+        LOG_INFO(ss);
+        return RSMI_STATUS_NOT_SUPPORTED;
+      }
+      throw amd::smi::rsmi_exception(RSMI_STATUS_INIT_ERROR, __FUNCTION__);
     } catch(...) {
       smi.Cleanup();
       throw amd::smi::rsmi_exception(RSMI_STATUS_INIT_ERROR, __FUNCTION__);