Feature/moderncpp

common/CudaWorker/CudaWorkerThread.hpp

@@ -18,6 +18,13 @@
 
 #include <chrono>
 
+
+inline int format_as(cudaError_enum type) {
+  return static_cast<int>(type);
+}
+
+
+
 #include "FieldWorkers.hpp"
 #include <DcgmTaskRunner.h>
 

common/CudaWorker/FieldWorkers.hpp

@@ -15,16 +15,21 @@
  */
 #pragma once
 
-#include <cublas_proxy.hpp>
+//#include <cublas_proxy.hpp>
+#include "../cublas_proxy/cublas_proxy.hpp"
 #include <cuda.h>
 
 #if (CUDA_VERSION_USED >= 11)
 #include "DcgmDgemm.hpp"
 #endif
 
 #include <DcgmLogging.h>
+
 #include <fmt/format.h>
 #include <timelib.h>
+inline int format_as(cublasStatus_t type) {
+  return static_cast<int>(type);
+  }
 
 using namespace Dcgm;
 

common/DcgmLogging.h

@@ -31,7 +31,7 @@
 #include <mutex>
 #include <string>
 #include <syslog.h>
-
+#include "format.hpp" // now define formatting for classes in this module
 #define DCGM_LOGGING_SEVERITY_OPTIONS "NONE, FATAL, ERROR, WARN, INFO, DEBUG, VERB"
 
 #define DCGM_LOGGING_SEVERITY_STRING_VERBOSE "VERB"

common/DcgmMutex.cpp

@@ -21,6 +21,8 @@
 #include <chrono>
 #include <ratio>
 
+inline int format_as(dcgmMutexSt type) {  return static_cast<int>(type);}
+
 /*****************************************************************************/
 DcgmMutex::DcgmMutex(int timeoutMs)
     // Cast to long long to avoid overflowing before widening to a long long

common/DcgmWatchTable.cpp

@@ -18,7 +18,7 @@
 #include "DcgmLogging.h"
 #include "DcgmUtilities.h"
 #include "DcgmWatchTable.h"
-
+inline int format_as(DcgmWatcherType_t type) {  return static_cast<int>(type);}
 /*****************************************************************************/
 DcgmWatchTable::DcgmWatchTable()
     : m_entityWatchHashTable()

common/tests/TaskRunnerTests.cpp

@@ -288,7 +288,7 @@ TEST_CASE("TaskRunner: Limited Queue")
 
     tr.Stop();
     fmt::print("Wait iterations elapsed: {}\n", cWaitIterations - waitIterations);
-    fmt::print("Iterations: {}\nExecutions: {}\nFailed to add: {}\n", iterations, executed, failedToAdd);
+    fmt::print("Iterations: {}\nExecutions: {}\nFailed to add: {}\n", (int)iterations, (int)executed, (int)failedToAdd);
     REQUIRE(executed >= cTaskRunnerCapacity);
     REQUIRE((failedToAdd + executed) == iterations);
 }
@@ -310,4 +310,4 @@ TEST_CASE("TaskRunner: Task with attempts")
     REQUIRE(fut.has_value());
     REQUIRE_THROWS_AS((*fut).get(), std::future_error);
     tr.Stop();
-}
+}

dcgmlib/format.hpp

@@ -0,0 +1,30 @@
+#ifndef DCGM_FORMAT_H
+#define DCGM_FORMAT_H
+#ifdef DcgmWatcherType_t
+inline int format_as(DcgmWatcherType_t type) {  return static_cast<int>(type);}
+#endif
+inline int format_as(dcgmChipArchitecture_enum type) {  return static_cast<int>(type);}
+inline int format_as(dcgmConfigType_enum type) {  return static_cast<int>(type);}
+inline int format_as(dcgmDiagnosticLevel_t type) {  return static_cast<int>(type);}
+
+inline int format_as(dcgmGroupType_enum type) {  return static_cast<int>(type);}
+inline int format_as(dcgmHealthSystems_enum type) {  return static_cast<int>(type);}
+inline int format_as(dcgmModuleId_t type) {  return static_cast<int>(type);}
+inline int format_as(dcgmModuleStatus_t type) {  return static_cast<int>(type);}
+//inline int format_as(dcgmMutexSt type) {  return static_cast<int>(type);}
+
+#ifdef dcgmNvLinkLinkState
+inline int format_as(dcgmNvLinkLinkState type) {  return static_cast<int>(type);}
+#endif
+inline int format_as(dcgmOrder_enum type) {  return static_cast<int>(type);}
+inline int format_as(dcgmPolicyCondition_enum type) {  return static_cast<int>(type);}
+inline int format_as(dcgmPolicyValidation_enum type) {  return static_cast<int>(type);}
+inline int format_as(dcgmPolicyAction_enum type) {  return static_cast<int>(type);}
+
+inline int format_as(dcgmReturn_enum type) {  return static_cast<int>(type);}
+inline int format_as(dcgm_field_entity_group_t type) {  return static_cast<int>(type);}
+
+#ifdef nvmlReturn_enum
+inline int format_as(nvmlReturn_enum type) {  return static_cast<int>(type);}
+#endif
+#endif

dcgmlib/src/DcgmCacheManager.cpp

@@ -27,10 +27,8 @@
 #include <TimeLib.hpp>
 #include <dcgm_agent.h>
 #include <dcgm_nvswitch_structs.h>
-
 #include <fmt/chrono.h>
 #include <fmt/format.h>
-
 #include <algorithm>
 #include <cmath>
 #include <cstring>
@@ -43,6 +41,10 @@
 #include <sys/types.h>
 #include <unistd.h>
 
+inline int format_as(nvmlReturn_enum type) {  return static_cast<int>(type);}
+inline int format_as(dcgmNvLinkLinkState_t type) {  return static_cast<int>(type);}
+inline int format_as(DcgmWatcherType_t type) {  return static_cast<int>(type);}
+inline int format_as(dcgmEntityStatusType_enum type) {  return static_cast<int>(type);}
 
 #define DRIVER_VERSION_510 510
 

dcgmlib/src/DcgmGroupManager.cpp

@@ -25,7 +25,7 @@
 #include "DcgmSettings.h"
 #include <fmt/format.h>
 #include <stdexcept>
-
+inline int format_as(dcgmEntityStatusType_enum type) {  return static_cast<int>(type);}
 /*****************************************************************************
  * Implementation for Group Manager Class
  *****************************************************************************/

dcgmlib/src/DcgmHostEngineHandler.cpp

@@ -45,7 +45,7 @@
 #include <nvml_injection.h>
 #include <ranges>
 #endif
-
+inline int format_as(DcgmWatcherType_t type) {  return static_cast<int>(type);}
 DcgmHostEngineHandler *DcgmHostEngineHandler::mpHostEngineHandlerInstance = nullptr;
 DcgmModuleCore DcgmHostEngineHandler::mModuleCoreObj;
 

dcgmlib/src/DcgmVgpu.cpp

@@ -21,7 +21,7 @@
 #include "dcgm_structs.h"
 #include "dcgm_structs_internal.h"
 #include <DcgmStringHelpers.h>
-
+inline int format_as(nvmlReturn_enum type) {  return static_cast<int>(type);}
 /*****************************************************************************/
 static std::string_view ConvertNvmlGridLicenseStateToString(unsigned int licenseState)
 {

dcgmproftester/DcgmProfTester.cpp

@@ -26,7 +26,7 @@
 #include "dcgm_structs.h"
 #include "timelib.h"
 #include "vector_types.h"
-#include <cublas_proxy.hpp>
+#include "../cublas_proxy/cublas_proxy.hpp"
 #include <cuda.h>
 #include <dcgm_agent.h>
 

dcgmproftester/PhysicalGpu.cpp

@@ -25,7 +25,8 @@
 #include "dcgm_fields_internal.hpp"
 #include "timelib.h"
 #include "vector_types.h"
-#include <cublas_proxy.hpp>
+//#include <cublas_proxy.hpp>
+#include "../cublas_proxy/cublas_proxy.hpp"
 #include <cuda.h>
 #include <dcgm_agent.h>
 

modules/nvswitch/DcgmNvSwitchManager.cpp

@@ -17,16 +17,15 @@
 #include <optional>
 #include <string>
 #include <tuple>
-
 #include <DcgmLogging.h>
 #include <DcgmSettings.h>
-
 #include "FieldIds.h"
 #include "NvSwitchData.h"
 #include "UpdateFunctions.h"
-
 #include "DcgmNvSwitchManager.h"
 
+inline int format_as(dcgmNvLinkLinkState_t type) {  return static_cast<int>(type);}
+
 namespace DcgmNs
 {
 using phys_id_t      = uint32_t;

modules/policy/DcgmPolicyManager.h

@@ -167,4 +167,6 @@ class DcgmPolicyManager
     dcgmReturn_t WatchFields(dcgm_connection_id_t connectionId);
 };
 
+inline int format_as(DcgmViolationPolicyAlert_enum type) {  return static_cast<int>(type);}
+
 #endif // DCGMPOLICYMANAGER_H

nvml-injection/include/InjectionArgument.h

@@ -2039,12 +2039,18 @@ class InjectionArgument
         return m_value.processInfo_v1Ptr;
     }
 
+  /*
+nvml-injection/include/InjectionArgument.h:2042:5: error: ‘InjectionArgument::InjectionArgument(nvmlProcessInfo_v2_t*)’ cannot be overloaded with ‘InjectionArgument::InjectionArgument(nvmlProcessInfo_t*)’
+ 2042 |     InjectionArgument(nvmlProcessInfo_v2_t *processInfo_v2Ptr)
+      |     ^~~~~~~~~~~~~~~~~
+
     InjectionArgument(nvmlProcessInfo_v2_t *processInfo_v2Ptr)
         : m_type(INJECTION_PROCESSINFO_V2_PTR)
     {
         memset(&m_value, 0, sizeof(m_value));
         m_value.processInfo_v2Ptr = processInfo_v2Ptr;
     }
+  */
     nvmlProcessInfo_v2_t *AsProcessInfo_v2Ptr() const
     {
         return m_value.processInfo_v2Ptr;

nvvs/include/JsonResult.hpp

@@ -28,6 +28,10 @@
 #include <string>
 #include <vector>
 
+inline int format_as(nvvsReturn_t type) {
+  return static_cast<int>(type);
+}
+
 /*
  * This file contains the JSON serialization and deserialization logic and structures for the NVVS JSON result format.
  * Here is an example of the JSON format:
@@ -69,6 +73,7 @@
 namespace DcgmNs::Nvvs::Json
 {
 
+
 struct Info
 {
     std::vector<std::string> messages;

nvvs/plugin_src/memory/l1tag.cu

@@ -15,7 +15,7 @@
  */
 
 #include "L1TagCuda.h"
-#include "newrandom.h"
+//#include "newrandom.h"
 
 __device__ void ReportError
 (
@@ -66,11 +66,11 @@ extern "C" __global__ void InitL1Data(const L1TagParams params)
 
     // Init RNG (each SM data region will have the same data)
     unsigned64 s[2];
-    InitRand<2>(s, params.randSeed + threadIdx.x);
+    //InitRand<2>(s, params.randSeed + threadIdx.x);
 
     for (uint32_t i = threadIdx.x; i < smidDataBytes / sizeof(*buf); i += blockDim.x)
     {
-        const uint16_t rnd = static_cast<uint16_t>(FastRand(s) >> 48);
+        const uint16_t rnd = 2;//static_cast<uint16_t>(FastRand(s) >> 48);
         buf[i] = EncodeOffset(i, rnd);
     }
 }
@@ -92,8 +92,8 @@ extern "C" __global__ void L1TagTest(const L1TagParams params)
 
     // Init RNG (each SM will use the same seed, for equivalent data accesses)
     unsigned64 s[2];
-    InitRand<2>(s, params.randSeed + hwtid);
-    uint32_t rnd = static_cast<uint32_t>(FastRand(s));
+    //InitRand<2>(s, params.randSeed + hwtid);
+    uint32_t rnd = 1;//static_cast<uint32_t>(FastRand(s));
 
     // Run the test for the specified iterations
     for (uint64_t iter = 0; iter < params.iterations; iter++)
@@ -168,7 +168,7 @@ extern "C" __global__ void L1TagTest(const L1TagParams params)
             }
 
             // Always use a new random offset
-            rnd = static_cast<uint32_t>(FastRand(s));
+            rnd = 3;//static_cast<uint32_t>(FastRand(s));
         }
     }
 }

nvvs/plugin_src/memtest/Memtest.cpp

@@ -49,6 +49,11 @@
 #include <random>
 #include <span>
 
+//inline int format_as(dcgmDiagnosticLevel_t type) {  return static_cast<int>(type);}
+inline int format_as(cudaError_enum type) {  return static_cast<int>(type);}
+inline int format_as(cudaError type) {  return static_cast<int>(type);}
+
+
 const unsigned int NUM_ITERATIONS = 1000;
 
 static __thread unsigned long *err_addr;

nvvs/plugin_src/pcie/Pcie.h

@@ -21,7 +21,8 @@
 #include "PluginDevice.h"
 
 
-#include <cublas_proxy.hpp>
+//#include <cublas_proxy.hpp>
+#include "../cublas_proxy/cublas_proxy.hpp"
 #include <dcgm_structs.h>
 #include <iostream>
 #include <string>

nvvs/plugin_src/targetedpower/TargetedPower_wrapper.h

@@ -29,7 +29,8 @@
 #include <CudaCommon.h>
 
 #include <NvvsStructs.h>
-#include <cublas_proxy.hpp>
+//#include <cublas_proxy.hpp>
+#include "../cublas_proxy/cublas_proxy.hpp"
 #include <cuda.h>
 
 #define TP_MAX_DIMENSION 8192 /* Maximum single dimension */

nvvs/src/NvidiaValidationSuite.cpp

@@ -38,8 +38,12 @@
 #include <time.h>
 #include <vector>
 
-using namespace DcgmNs::Nvvs;
 
+
+using namespace DcgmNs::Nvvs;
+inline int format_as(dcgmEntityStatusType_enum type) {  return static_cast<int>(type);}
+//inline int format_as(dcgmDiagnosticLevel_t type) {  return static_cast<int>(type);}
+//inline int format_as(cudaError_enum type) {  return static_cast<int>(type);}
 DcgmHandle dcgmHandle;
 DcgmSystem dcgmSystem;
 NvvsCommon nvvsCommon __attribute__((visibility("default")));

nvvs/src/Plugin.cpp

@@ -16,6 +16,11 @@
 #include "Plugin.h"
 #include "PluginStrings.h"
 
+
+inline int format_as(nvvsPluginResult_enum type) {
+  return static_cast<int>(type);
+}
+
 const double DUMMY_TEMPERATURE_VALUE = 30.0;
 
 /*************************************************************************/

nvvs/src/PluginLib.cpp

@@ -19,6 +19,10 @@
 
 #include <dlfcn.h>
 
+inline int format_as(dcgmDiagAuxDataType type) {
+  return static_cast<int>(type);
+}
+
 
 /*****************************************************************************/
 PluginLib::PluginLib()

testing/TestDiagResponseWrapper.cpp

@@ -22,6 +22,11 @@
 #include "NvvsJsonStrings.h"
 #include "TestDiagResponseWrapper.h"
 
+
+inline int format_as(dcgmDiagResult_enum type) {
+  return static_cast<int>(type);
+}
+
 TestDiagResponseWrapper::TestDiagResponseWrapper() = default;
 
 TestDiagResponseWrapper::~TestDiagResponseWrapper() = default;