Sync-up with master

CHANGELOG.md

@@ -1,12 +1,19 @@
-# Change Log for ROCm Validation Suite (RVS)
+# Changelog for ROCm Validation Suite (RVS)
 
 Full documentation for RVS is available at [ROCmValidationSuite.Readme](https://github.com/ROCm/ROCmValidationSuite).
 
-## RVS 1.0.0 for ROCm 6.2
+## RVS 1.1.0 for ROCm 6.3.0
+
+### Added
+
+- Support for hipBLASLT blas library and option to select blas library in conf. file.
+- Added Babel, thermal and performance benchmark test for MI308X.
 
 ### Changed
 
-- Updated GST performance benchmark test for better numbers.
+- Babel parameters made runtime configurable.
+
+## RVS 1.0.0 for ROCm 6.2.0
 
 ### Added
 
@@ -15,18 +22,22 @@ Full documentation for RVS is available at [ROCmValidationSuite.Readme](https://
 - IET (power) stress test for MI300A.
 - IET (power transition) test for MI300X.
 
-## RVS 1.0.0 for ROCm 6.1
-
 ### Changed
 
-- Updated pebb & pbqt logs to include PCI BDF.
+- Updated GST performance benchmark test for better numbers.
+
+## RVS 1.0.0 for ROCm 6.1.0
 
 ### Added
 
 - Support data types (BF16 and FP8) based GEMM operations in GPU Stress Test (GST) module.
 - Babel test for MI300X.
 
-## RVS 1.0.0 for ROCm 6.0
+### Changed
+
+- Updated pebb & pbqt logs to include PCI BDF.
+
+## RVS 1.0.0 for ROCm 6.0.0
 
 ### Added
 
@@ -35,22 +46,22 @@ Full documentation for RVS is available at [ROCmValidationSuite.Readme](https://
 - Navi31 and Navi32 specific configurations
 - GST stress test support for MI300X
 
-## RVS 1.0.0 for ROCm 5.7
+## RVS 1.0.0 for ROCm 5.7.0
 
 ### Added
 
 - Introduced new RVS interface APIs enabling test execution from external components.
-- Added new "device_index" property for conf. file.
+- Added new `device_index` property for conf. file.
 
 ### Changed
 
 - Moved all static internal libraries to a single public shared library (rvslib).
 - Use HIP stream callback mechanism for gemm operations completion (instead of polling).
 
-### Optimizations
-
-- In GST and IET modules, use of callback mechanism instead of polling for HIP stream reduced the CPU utilization %.
-
 ### Removed
 
 - yaml-cpp source download and build removed from RVS cmake build.
+
+### Optimized
+
+- In GST and IET modules, use of callback mechanism instead of polling for HIP stream reduced the CPU utilization %.

CMakeLists.txt

@@ -26,7 +26,9 @@
 cmake_minimum_required ( VERSION 3.5.0 )
 set(CMAKE_SHARED_LINKER_FLAGS_INIT "-Wl,--enable-new-dtags,--rpath,$ORIGIN:$ORIGIN/.." CACHE STRING "RUNPATH for libraries")
 set(CMAKE_EXE_LINKER_FLAGS_INIT "-Wl,--enable-new-dtags,--rpath,$ORIGIN/../lib" CACHE STRING "RUNPATH for executables")
-project ("rocm-validation-suite")
+
+project ("rocm-validation-suite"
+          VERSION 1.1.0)
 
 # Default libdir to "lib", this skips GNUInstallDirs from trying to take a guess if it's unset:
 set(CMAKE_INSTALL_LIBDIR "lib" CACHE STRING "Library install directory")
@@ -148,25 +150,25 @@ endif()
 ## Include common cmake modules
 include ( utils )
 
-## Setup the package version default .
-set ( RVS_VERSION "1.0.0" )
+## Update RVS version
+set ( RVS_VERSION "${PROJECT_VERSION}" )
 get_version ( ${RVS_VERSION} )
 
 # Package Generator  #######################################################
 set(CPACK_PACKAGE_NAME "rocm-validation-suite")
 set(CPACK_PACKAGE_DESCRIPTION "ROCm Validation Suite")
 set(CPACK_PACKAGE_DESCRIPTION_SUMMARY "The ROCm Validation Suite (RVS) is a system validation and diagnostics tool for monitoring, stress testing, detecting and troubleshooting issues that affects the functionality and performance of AMD GPU(s) operating in a high-performance/AI/ML computing environment.")
-set(CPACK_PACKAGE_VERSION_MAJOR "${VERSION_MAJOR}")
-set(CPACK_PACKAGE_VERSION_MINOR "${VERSION_MINOR}")
-set(CPACK_PACKAGE_VERSION_PATCH "${VERSION_PATCH}")
+set(CPACK_PACKAGE_VERSION_MAJOR "${PROJECT_VERSION_MAJOR}")
+set(CPACK_PACKAGE_VERSION_MINOR "${PROJECT_VERSION_MINOR}")
+set(CPACK_PACKAGE_VERSION_PATCH "${PROJECT_VERSION_PATCH}")
 set(CPACK_PACKAGE_VENDOR "Advanced Micro Devices Inc.")
 set(CPACK_PACKAGE_CONTACT "ROCM Validation Suite Support <[email protected]>")
 
 # Package dependencies
 if(FETCH_ROCMPATH_FROM_ROCMCORE)
-  set(ROCM_DEPENDENT_PACKAGES "hip-runtime-amd, comgr, hsa-rocr, rocblas, rocm-smi-lib, ${ROCM_CORE}")
+  set(ROCM_DEPENDENT_PACKAGES "hip-runtime-amd, comgr, hsa-rocr, rocblas, rocm-smi-lib, hiprand, hipblaslt, ${ROCM_CORE}")
 else()
-  set(ROCM_DEPENDENT_PACKAGES "hip-runtime-amd, comgr, hsa-rocr, rocblas, rocm-smi-lib")
+  set(ROCM_DEPENDENT_PACKAGES "hip-runtime-amd, comgr, hsa-rocr, rocblas, rocm-smi-lib, hiprand, hipblaslt")
 endif()
 if (${RVS_OS_TYPE} STREQUAL "ubuntu")
 set(CPACK_DEBIAN_PACKAGE_DEPENDS "${ROCM_DEPENDENT_PACKAGES}, libpci3, libyaml-cpp-dev")
@@ -180,9 +182,10 @@ endif()
 set(CPACK_ARCHIVE_COMPONENT_INSTALL ON)
 set(CPACK_COMPONENTS_ALL applications rvsmodule)
 
-# Package naming change
+# Set default ROCm version for package naming
 set(ROCM_VERSION_FOR_PACKAGE "99999")
-# update version accordingly
+
+# Update ROCm version for package naming
 if(DEFINED ENV{ROCM_LIBPATCH_VERSION})
   set(ROCM_VERSION_FOR_PACKAGE $ENV{ROCM_LIBPATCH_VERSION})
 endif()
@@ -207,6 +210,7 @@ if(CPACK_RPM_PACKAGE_RELEASE)
   set(CPACK_RPM_PACKAGE_RELEASE_DIST ON)
 endif()
 
+# Complete RVS package version naming
 set(CPACK_PACKAGE_VERSION "${CPACK_PACKAGE_VERSION_MAJOR}.${CPACK_PACKAGE_VERSION_MINOR}.${CPACK_PACKAGE_VERSION_PATCH}.${ROCM_VERSION_FOR_PACKAGE}")
 
 set(CPACK_DEBIAN_FILE_NAME "DEB-DEFAULT")
@@ -490,7 +494,7 @@ if(BUILD_ADDRESS_SANITIZER)
   set(ASAN_LD_FLAGS "-fuse-ld=lld")
 endif()
 
-set(HCC_CXX_FLAGS "-fno-gpu-rdc --offload-arch=gfx90a --offload-arch=gfx1030 --offload-arch=gfx803 --offload-arch=gfx900 --offload-arch=gfx906 --offload-arch=gfx908 --offload-arch=gfx1101 --offload-arch=gfx940 --offload-arch=gfx941 --offload-arch=gfx942 --offload-arch=gfx1200 --offload-arch=gfx1201")
+set(HCC_CXX_FLAGS "-fno-gpu-rdc --offload-arch=gfx90a --offload-arch=gfx1030 --offload-arch=gfx803 --offload-arch=gfx900 --offload-arch=gfx906 --offload-arch=gfx908 --offload-arch=gfx1101 --offload-arch=gfx940 --offload-arch=gfx941 --offload-arch=gfx942 --offload-arch=gfx1200 --offload-arch=gfx1201 --offload-arch=gfx1100")
 
 add_subdirectory(rvslib)
 add_subdirectory(rvs)
@@ -603,13 +607,10 @@ configure_file(${CMAKE_SOURCE_DIR}/rvs/conf/deviceid.sh.in
   @ONLY)
 install(FILES ${CMAKE_SOURCE_DIR}/rvs/conf/deviceid.sh
 	DESTINATION ${CPACK_PACKAGING_INSTALL_PREFIX}/${CMAKE_INSTALL_DATADIR}/${CPACK_PACKAGE_NAME}/conf/
-  PERMISSIONS OWNER_EXECUTE OWNER_WRITE OWNER_READ
-                    GROUP_EXECUTE GROUP_READ
-  COMPONENT rvsmodule
+        COMPONENT rvsmodule
 )
 install(DIRECTORY ${CMAKE_SOURCE_DIR}/testscripts/
 	DESTINATION ${CPACK_PACKAGING_INSTALL_PREFIX}/${CMAKE_INSTALL_DATADIR}/${CPACK_PACKAGE_NAME}/testscripts
-  USE_SOURCE_PERMISSIONS
   COMPONENT rvsmodule
 )
 

FEATURES.md

@@ -4,10 +4,10 @@
 ## GPU Properties – GPUP
 The GPU Properties module queries the configuration of a target device and returns the device’s static characteristics. These static values can be used to debug issues such as device support, performance and firmware problems.
 
-## GPU Monitor – GM module
+## GPU Monitor – GM module [deprecated]
 The GPU monitor tool is capable of running on one, some or all of the GPU(s) installed and will report various information at regular intervals. The module can be configured to halt another RVS modules execution if one of the quantities exceeds a specified boundary value.
 
-## PCI Express State Monitor – PESM module
+## PCI Express State Monitor – PESM module [deprecated]
 The PCIe State Monitor tool is used to actively monitor the PCIe interconnect between the host platform and the GPU. The module will register a “listener” on a target GPU’s PCIe interconnect, and log a message whenever it detects a state change. The PESM will be able to detect the following state changes:
 
 1.	PCIe link speed changes
@@ -24,7 +24,7 @@ The PCIe Qualification Tool is used to qualify the PCIe bus on which the GPU is
 3.	PCIe link speed
 4.	PCIe link width
 
-## SBIOS Mapping Qualification Tool – SMQT module
+## SBIOS Mapping Qualification Tool – SMQT module [deprecated]
 The GPU SBIOS mapping qualification tool is designed to verify that a platform’s SBIOS has satisfied the BAR mapping requirements for VDI and Radeon Instinct products for ROCm support.
 
 ## P2P Benchmark and Qualification Tool – PBQT module
@@ -34,7 +34,7 @@ The P2P Benchmark and Qualification Tool is designed to provide the list of all
 The PCIe Bandwidth Benchmark attempts to saturate the PCIe bus with DMA transfers between system memory and a target GPU card’s memory. The maximum bandwidth obtained is reported to help debug low bandwidth issues. The benchmark should be capable of targeting one, some or all of the GPUs installed in a platform, reporting individual benchmark statistics for each.
 
 ## GPU Stress Test - GST module
-The GPU Stress Test runs various GEMM operations as workloads to stress the GPU FLOPS performance. GEMM operations include SGEMM, DGEMM and HGEMM (Single/Double/Half-precision General Matrix Multiplication) operations based on configured parameters. The duration of the test is configurable, both in terms of time (how long to run) and iterations (how many times to run).
+The GPU Stress Test runs various GEMM computations as workloads to stress the GPU FLOPS performance and check whether it meets the configured target GFLOPS. GEMM workloads shall be configured as either operation type or data type. GEMM based on operation types include SGEMM, DGEMM and HGEMM (Single/Double/Half-precision General Matrix Multiplication) - configured using operation parameter. GEMM based on data types include `fp8`, `i8`, `fp16`, `bf16`, `fp32` and `tf32` (`xf32`) - configured using data type parameter. The duration of the test is configurable, both in terms of time (how long to run) and iterations (how many times to run).
 
 ## Input EDPp Test - IET module
 The Input EDPp Test runs GEMM workloads to stress the GPU power (i.e. TGP). This test is used to verify if the GPU is capable of handling max. power stress for a sustained period of time. Also checks whether GPU power reaches a set target power.

LICENSE

@@ -1,6 +1,6 @@
 MIT License
 
-Copyright (c) 2018-2022 Advanced Micro Devices, Inc. All rights reserved.
+Copyright (c) 2018-2024 Advanced Micro Devices, Inc. All rights reserved.
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

README.md

@@ -20,19 +20,11 @@ CentOS :
 
 RHEL :
 
-    sudo yum install -y cmake3 doxygen rpm rpm-build git gcc-c++ yaml-cpp-devel
-
-    wget http://mirror.centos.org/centos/7/os/x86_64/Packages/pciutils-devel-3.5.1-3.el7.x86_64.rpm
-
-    sudo rpm -ivh pciutils-devel-3.5.1-3.el7.x86_64.rpm
+    sudo yum install -y cmake3 doxygen rpm rpm-build git gcc-c++ yaml-cpp-devel pciutils-devel
 
 SLES :
 
-    sudo SUSEConnect -p sle-module-desktop-applications/15.1/x86_64
-
-    sudo SUSEConnect --product sle-module-development-tools/15.1/x86_64
-
-    sudo zypper  install -y cmake doxygen pciutils-devel libpci3 rpm git rpm-build gcc-c++ yaml-cpp-devel
+    sudo zypper install -y cmake doxygen pciutils-devel libpci3 rpm git rpm-build gcc-c++ yaml-cpp-devel
 
 ## Install ROCm stack, rocblas and rocm-smi-lib
 Install ROCm stack for Ubuntu/CentOS/SLES/RHEL. Refer to
@@ -136,7 +128,6 @@ SUSE :
     Command examples
     ./rvs --help ; Lists all options to run RVS test suite
     ./rvs -g ; Lists supported GPUs available in the machine
-    ./rvs -d 3 ; Run set of RVS default sanity tests (in rvs.conf) with verbose level 3
     ./rvs -c conf/gst_single.conf ; Run GST module default test configuration
 
 ### Run version pre-compiled and packaged with ROCm release
@@ -146,7 +137,6 @@ SUSE :
     Command examples
     ./rvs --help ; Lists all options to run RVS test suite
     ./rvs -g ; Lists supported GPUs available in the machine
-    ./rvs -d 3 ; Run set of RVS sanity tests (in rvs.conf) with verbose level 3
     ./rvs -c ../share/rocm-validation-suite/conf/gst_single.conf ; Run GST default test configuration
 
 To run GPU specific test configuration, use configuration files from GPU folders in "/opt/rocm/share/rocm-validation-suite/conf"
@@ -159,5 +149,5 @@ Note: If present, always use GPU specific configurations instead of default test
 ## Reporting
 
 Test results, errors and verbose logs are printed as terminal output. To enable json logging use "-j" command line option.
-The json output file is stored in /var/tmp folder and the name of the file will be printed.
+The json output file path can be specified after "-j" option. If not specified, a file is created in /var/tmp folder and the name of the file will be printed to stdout.
 

babel.so/CMakeLists.txt

@@ -61,7 +61,7 @@ set(HIP_HCC_BUILD_FLAGS "${HIP_HCC_BUILD_FLAGS} -DHIP_VERSION_MAJOR=${HIP_VERSIO
 set(HIP_HCC_BUILD_FLAGS)
 set(HIP_HCC_BUILD_FLAGS "${HIP_HCC_BUILD_FLAGS} -fPIC ${HCC_CXX_FLAGS} -I${HSA_PATH}/include ${ASAN_CXX_FLAGS}")
 
-set(HIP_STREAM_BUILD_FLAGS "-DNONTEMPORAL=1 -DDWORDS_PER_LANE=4 -DTBSIZE=1024 -DCHUNKS_PER_BLOCK=2 -O3 -std=c++17")
+set(HIP_STREAM_BUILD_FLAGS "-DNONTEMPORAL=1 -DTBSIZE=1024 -O3 -std=c++17")
 
 # Set compiler and compiler flags
 set(CMAKE_CXX_COMPILER "${HIPCC_PATH}/bin/hipcc")
@@ -177,3 +177,4 @@ if (RVS_BUILD_TESTS)
   )
   include(${CMAKE_CURRENT_SOURCE_DIR}/tests.cmake)
 endif()
+

babel.so/include/HIPStream.h

@@ -23,40 +23,16 @@
 template <class T>
 class HIPStream : public Stream<T>
 {
-#ifdef __HIP_PLATFORM_NVCC__
-  #ifndef DWORDS_PER_LANE
-  #define DWORDS_PER_LANE 1
-  #endif
-  #ifndef CHUNKS_PER_BLOCK
-  #define CHUNKS_PER_BLOCK 8
-  #endif
-#else
-  #ifndef DWORDS_PER_LANE
-  #define DWORDS_PER_LANE 4
-  #endif
-  #ifndef CHUNKS_PER_BLOCK
-  #define CHUNKS_PER_BLOCK 1
-  #endif
-#endif
-  // make sure that either:
-  //    DWORDS_PER_LANE is less than sizeof(T), in which case we default to 1 element
-  //    or
-  //    DWORDS_PER_LANE is divisible by sizeof(T)
-  static_assert((DWORDS_PER_LANE * sizeof(unsigned int) < sizeof(T)) ||
-                (DWORDS_PER_LANE * sizeof(unsigned int) % sizeof(T) == 0),
-                "DWORDS_PER_LANE not divisible by sizeof(element_type)");
 
-  static constexpr unsigned int chunks_per_block{CHUNKS_PER_BLOCK};
-  // take into account the datatype size
-  // that is, if we specify 4 DWORDS_PER_LANE, this is 2 FP64 elements
-  // and 4 FP32 elements
-  static constexpr unsigned int elements_per_lane{
-    (DWORDS_PER_LANE * sizeof(unsigned int)) < sizeof(T) ? 1 : (
-     DWORDS_PER_LANE * sizeof(unsigned int) / sizeof(T))};
   protected:
+
+    unsigned int dwords_per_lane;
+    unsigned int chunks_per_block;
+    unsigned int elements_per_lane;
+
     // Size of arrays
     const unsigned int array_size;
-    const unsigned int block_cnt;
+    unsigned int block_cnt;
     const bool evt_timing;
     hipEvent_t start_ev;
     hipEvent_t stop_ev;
@@ -71,7 +47,8 @@ class HIPStream : public Stream<T>
     T *d_c;
 
   public:
-    HIPStream(const unsigned int, const bool, const int);
+    HIPStream(const unsigned int, const bool, const int,
+        const unsigned int, const unsigned int);
     ~HIPStream();
 
     virtual float read() override;
@@ -85,3 +62,4 @@ class HIPStream : public Stream<T>
     virtual void init_arrays(T initA, T initB, T initC) override;
     virtual void read_arrays(std::vector<T>& a, std::vector<T>& b, std::vector<T>& c) override;
 };
+

babel.so/include/action.h

@@ -1,6 +1,6 @@
 /********************************************************************************
  *
- * Copyright (c) 2018-2022 Advanced Micro Devices, Inc. All rights reserved.
+ * Copyright (c) 2018-2024 Advanced Micro Devices, Inc. All rights reserved.
  *
  * MIT LICENSE:
  * Permission is hereby granted, free of charge, to any person obtaining a copy of
@@ -53,20 +53,22 @@ using std::map;
 #define RVS_CONF_MEM_MIBIBYTE           "mibibytes"
 #define RVS_CONF_OP_CSV                 "o/p_csv"
 #define RVS_CONF_SUBTEST                "subtest"
+#define RVS_CONF_DWORDS_PER_LANE        "dwords_per_lane"
+#define RVS_CONF_CHUNKS_PER_BLOCK       "chunks_per_block"
 
 #define MEM_DEFAULT_ARRAY_SIZE          33554432   // 32 MB
 #define MEM_DEFAULT_NUM_ITER            100
 #define MEM_DEFAULT_TEST_TYPE           1
 #define MEM_DEFAULT_MEM_MIBIBYTE        false
 #define MEM_DEFAULT_OP_CSV              false
 #define MEM_DEFAULT_SUBTEST             5
+#define MEM_DEFAULT_DWORDS_PER_LANE     4
+#define MEM_DEFAULT_CHUNKS_PER_BLOCK    2
 
 #define MEM_NO_COMPATIBLE_GPUS          "No AMD compatible GPU found!"
 #define FLOATING_POINT_REGEX            "^[0-9]*\\.?[0-9]+$"
 #define JSON_CREATE_NODE_ERROR          "JSON cannot create node"
 
-
-
 /**
  * @class mem_action
  * @ingroup MEM
@@ -84,12 +86,10 @@ class mem_action: public rvs::actionbase {
     virtual ~mem_action();
 
     virtual int run(void);
-
+  
     std::string mem_ops_type;
 
  protected:
-    //! TRUE if JSON output is required
-    bool bjson;
     //! Memory in bytes
     bool mibibytes;
     //! output in csv 
@@ -102,7 +102,10 @@ class mem_action: public rvs::actionbase {
     uint64_t num_iterations;
     //! number of iterations
     uint64_t array_size;
-
+    //! number of dwords per lane
+    uint16_t dwords_per_lane;
+    //! number of chunks per block
+    uint16_t chunks_per_block;
 
     // configuration properties getters
     bool get_all_mem_config_keys(void);
@@ -113,7 +116,6 @@ class mem_action: public rvs::actionbase {
   */
   int get_num_amd_gpu_devices(void);
   int get_all_selected_gpus(void);
-
   bool do_mem_stress_test(map<int, uint16_t> mem_gpus_device_index);
 };