diff --git a/.github/workflows/run.yml b/.github/workflows/run.yml
index c1e8a2f76..0dd660f01 100644
--- a/.github/workflows/run.yml
+++ b/.github/workflows/run.yml
@@ -222,6 +222,32 @@ jobs:
             --disable-logger \
             workspace setup --dry-run
 
+      - name: Dry run dynamic saxpy/openmp with dynamic Venado
+        run: |
+          ./bin/benchpark system init --dest=venado-system1 lanl-venado cuda=12.5 compiler=cce +gtl
+          system_id=$(./bin/benchpark system id ./venado-system1)
+          ./bin/benchpark experiment init --dest=saxpy-openmp-venado saxpy+openmp
+          ./bin/benchpark setup ./saxpy-openmp-venado ./venado-system1 workspace/
+          . workspace/setup.sh
+          ramble \
+            --workspace-dir "workspace/saxpy-openmp-venado/$system_id/workspace" \
+            --disable-progress-bar \
+            --disable-logger \
+            workspace setup --dry-run
+
+      - name: Dry run dynamic saxpy/cuda with dynamic Venado
+        run: |
+          ./bin/benchpark system init --dest=venado-system2 lanl-venado cuda=12.5 compiler=cce +gtl
+          system_id=$(./bin/benchpark system id ./venado-system2)
+          ./bin/benchpark experiment init --dest=saxpy-cuda-venado saxpy+cuda
+          ./bin/benchpark setup ./saxpy-cuda-venado ./venado-system2 workspace/
+          . workspace/setup.sh
+          ramble \
+            --workspace-dir "workspace/saxpy-cuda-venado/$system_id/workspace" \
+            --disable-progress-bar \
+            --disable-logger \
+            workspace setup --dry-run
+
       - name: Dry run dynamic gromacs/openmp with dynamic Ruby
         run: |
           ./bin/benchpark system init --dest=cluster-system llnl-cluster compiler=gcc 
diff --git a/legacy/systems/LLNL-Tioga-HPECray-zen3-MI250X-Slingshot/auxiliary_software_files/packages.yaml b/legacy/systems/LLNL-Tioga-HPECray-zen3-MI250X-Slingshot/auxiliary_software_files/packages.yaml
index 1c050e729..2528cb68e 100644
--- a/legacy/systems/LLNL-Tioga-HPECray-zen3-MI250X-Slingshot/auxiliary_software_files/packages.yaml
+++ b/legacy/systems/LLNL-Tioga-HPECray-zen3-MI250X-Slingshot/auxiliary_software_files/packages.yaml
@@ -302,6 +302,7 @@ packages:
         gtl_cutoff_size: 4096
         fi_cxi_ats: 0
         gtl_lib_path: /opt/cray/pe/mpich/8.1.26/gtl/lib
+        gtl_libs: ["libmpi_gtl_hsa"]
         ldflags: "-L/opt/cray/pe/mpich/8.1.26/ofi/crayclang/16.0/lib -lmpi -L/opt/cray/pe/mpich/8.1.26/gtl/lib -Wl,-rpath=/opt/cray/pe/mpich/8.1.26/gtl/lib -lmpi_gtl_hsa"
     - spec: cray-mpich@8.1.26%cce@16.0.0 ~gtl +wrappers
       prefix: /opt/cray/pe/mpich/8.1.26/ofi/crayclang/16.0
diff --git a/repo/amg2023/package.py b/repo/amg2023/package.py
index c7e2ac0c0..5cc54af78 100644
--- a/repo/amg2023/package.py
+++ b/repo/amg2023/package.py
@@ -36,7 +36,7 @@ class Amg2023(CMakePackage, CudaPackage, ROCmPackage):
 
     depends_on("hypre+cuda", when="+cuda")
     requires("+cuda", when="^hypre+cuda")
-    for arch in ("none", "50", "60", "70", "80"):
+    for arch in ("none", "50", "60", "70", "80", "90"):
         depends_on(f"hypre cuda_arch={arch}", when=f"cuda_arch={arch}")
 
     depends_on("hypre+rocm", when="+rocm")
@@ -44,6 +44,10 @@ class Amg2023(CMakePackage, CudaPackage, ROCmPackage):
     for target in ("none", "gfx803", "gfx900", "gfx906", "gfx908", "gfx90a", "gfx942"):
         depends_on(f"hypre amdgpu_target={target}", when=f"amdgpu_target={target}")
 
+    def setup_build_environment(self, env):
+        if "+cuda" in self.spec:
+            env.set("NVCC_APPEND_FLAGS", "-allow-unsupported-compiler")
+
     def cmake_args(self):
         cmake_options = []
         cmake_options.append(self.define_from_variant("AMG_WITH_CALIPER", "caliper"))
diff --git a/repo/caliper/for_aarch64.patch b/repo/caliper/for_aarch64.patch
new file mode 100644
index 000000000..d3fed9a09
--- /dev/null
+++ b/repo/caliper/for_aarch64.patch
@@ -0,0 +1,11 @@
+--- spack-src/src/services/callpath/Callpath.cpp.bak	2020-10-28 14:38:19.668122844 +0900
++++ spack-src/src/services/callpath/Callpath.cpp	2020-10-28 15:03:12.258061188 +0900
+@@ -63,7 +63,7 @@
+         unw_context_t unw_ctx;
+         unw_cursor_t  unw_cursor;
+ 
+-        unw_getcontext(&unw_ctx);
++        unw_getcontext(unw_ctx);
+ 
+         if (unw_init_local(&unw_cursor, &unw_ctx) < 0) {
+             Log(0).stream() << "callpath: unable to init libunwind cursor" << endl;
diff --git a/repo/caliper/package.py b/repo/caliper/package.py
new file mode 100644
index 000000000..1b13781bd
--- /dev/null
+++ b/repo/caliper/package.py
@@ -0,0 +1,291 @@
+# Copyright 2023 Lawrence Livermore National Security, LLC and other
+# Benchpark Project Developers. See the top-level COPYRIGHT file for details.
+#
+# SPDX-License-Identifier: Apache-2.0
+
+import os
+import sys
+
+from spack.package import *
+
+
+class Caliper(CachedCMakePackage, CudaPackage, ROCmPackage):
+    """Caliper is a program instrumentation and performance measurement
+    framework. It is designed as a performance analysis toolbox in a
+    library, allowing one to bake performance analysis capabilities
+    directly into applications and activate them at runtime.
+    """
+
+    homepage = "https://github.com/LLNL/Caliper"
+    git = "https://github.com/LLNL/Caliper.git"
+    url = "https://github.com/LLNL/Caliper/archive/v2.12.1.tar.gz"
+    tags = ["e4s", "radiuss"]
+
+    maintainers("daboehme", "adrienbernede")
+
+    test_requires_compiler = True
+
+    license("BSD-3-Clause")
+
+    version("master", branch="master")
+    version("2.12.1", sha256="2b5a8f98382c94dc75cc3f4517c758eaf9a3f9cea0a8dbdc7b38506060d6955c")
+    version("2.11.0", sha256="b86b733cbb73495d5f3fe06e6a9885ec77365c8aa9195e7654581180adc2217c")
+    version("2.10.0", sha256="14c4fb5edd5e67808d581523b4f8f05ace8549698c0e90d84b53171a77f58565")
+    version("2.9.1", sha256="4771d630de505eff9227e0ec498d0da33ae6f9c34df23cb201b56181b8759e9e")
+    version("2.9.0", sha256="507ea74be64a2dfd111b292c24c4f55f459257528ba51a5242313fa50978371f")
+    version(
+        "2.8.0",
+        sha256="17807b364b5ac4b05997ead41bd173e773f9a26ff573ff2fe61e0e70eab496e4",
+        deprecated=True,
+    )
+    version(
+        "2.7.0",
+        sha256="b3bf290ec2692284c6b4f54cc0c507b5700c536571d3e1a66e56626618024b2b",
+        deprecated=True,
+    )
+    version(
+        "2.6.0",
+        sha256="6efcd3e4845cc9a6169e0d934840766b12182c6d09aa3ceca4ae776e23b6360f",
+        deprecated=True,
+    )
+    version(
+        "2.5.0",
+        sha256="d553e60697d61c53de369b9ca464eb30710bda90fba9671201543b64eeac943c",
+        deprecated=True,
+    )
+    version(
+        "2.4.0", tag="v2.4.0", commit="30577b4b8beae104b2b35ed487fec52590a99b3d", deprecated=True
+    )
+    version(
+        "2.3.0", tag="v2.3.0", commit="9fd89bb0120750d1f9dfe37bd963e24e478a2a20", deprecated=True
+    )
+    version(
+        "2.2.0", tag="v2.2.0", commit="c408e9b3642c7aa80eff37b0826d819c57e7bc04", deprecated=True
+    )
+    version(
+        "2.1.1", tag="v2.1.1", commit="0593b0e01c1d8d3e50c990399cc0fee403485599", deprecated=True
+    )
+    version(
+        "2.0.1", tag="v2.0.1", commit="4d7ff46381c53a461e62edd949e2d9dea9db7b08", deprecated=True
+    )
+    version(
+        "1.9.1", tag="v1.9.1", commit="cfc1defbbee20b50dd3e3477badd09a92b1df970", deprecated=True
+    )
+    version(
+        "1.9.0", tag="v1.9.0", commit="8356e747349b285aa621c5b74e71559f0babc4a1", deprecated=True
+    )
+    version(
+        "1.8.0", tag="v1.8.0", commit="117c1ef596b617dc71407b8b67eebef094a654f8", deprecated=True
+    )
+    version(
+        "1.7.0", tag="v1.7.0", commit="898277c93d884d4e7ca1ffcf3bbea81d22364f26", deprecated=True
+    )
+
+    #depends_on("c", type="build")  # generated
+    #depends_on("cxx", type="build")  # generated
+    #depends_on("fortran", type="build")  # generated
+
+    is_linux = sys.platform.startswith("linux")
+    variant("shared", default=True, description="Build shared libraries")
+    variant("adiak", default=True, description="Enable Adiak support")
+    variant("mpi", default=True, description="Enable MPI support")
+    # libunwind has some issues on Mac
+    variant(
+        "libunwind", default=sys.platform != "darwin", description="Enable stack unwind support"
+    )
+    variant("libdw", default=is_linux, description="Enable DWARF symbol lookup")
+    # pthread_self() signature is incompatible with PAPI_thread_init() on Mac
+    variant("papi", default=sys.platform != "darwin", description="Enable PAPI service")
+    variant("libpfm", default=False, description="Enable libpfm (perf_events) service")
+    # Gotcha is Linux-only
+    variant("gotcha", default=is_linux, description="Enable GOTCHA support")
+    variant("sampler", default=is_linux, description="Enable sampling support on Linux")
+    variant("sosflow", default=False, description="Enable SOSflow support")
+    variant("fortran", default=False, description="Enable Fortran support")
+    variant("variorum", default=False, description="Enable Variorum support")
+    variant("vtune", default=False, description="Enable Intel Vtune support")
+    variant("kokkos", default=True, when="@2.3.0:", description="Enable Kokkos profiling support")
+    variant("tests", default=False, description="Enable tests")
+    variant("tools", default=True, description="Enable tools")
+    variant("python", default=False, when="@v2.12:", description="Build Python bindings")
+
+    depends_on("adiak@0.1:0", when="@2.2:2.10 +adiak")
+    depends_on("adiak@0.4:0", when="@2.11: +adiak")
+
+    depends_on("papi@5.3:5", when="@:2.2 +papi")
+    depends_on("papi@5.3:", when="@2.3: +papi")
+
+    depends_on("libpfm4@4.8:4", when="+libpfm")
+
+    depends_on("mpi", when="+mpi")
+    depends_on("unwind@1.2:1", when="+libunwind")
+    depends_on("elfutils", when="+libdw")
+    depends_on("variorum", when="+variorum")
+    depends_on("intel-oneapi-vtune", when="+vtune")
+
+    depends_on("sosflow@spack", when="@1.0:1+sosflow")
+
+    depends_on("cmake", type="build")
+    depends_on("python", type="build")
+
+    depends_on("python@3", when="+python", type=("build", "link", "run"))
+    depends_on("py-pybind11", when="+python", type=("build", "link", "run"))
+
+    # sosflow support not yet in 2.0
+    conflicts("+sosflow", "@2:")
+    conflicts("+adiak", "@:2.1")
+    conflicts("+libdw", "@:2.4")
+    conflicts("+rocm", "@:2.7")
+    conflicts("+rocm+cuda")
+
+    patch("for_aarch64.patch", when="@:2.11 target=aarch64:")
+    patch(
+        "sampler-service-missing-libunwind-include-dir.patch",
+        when="@2.9.0:2.9.1 +libunwind +sampler",
+    )
+
+    def _get_sys_type(self, spec):
+        sys_type = spec.architecture
+        if "SYS_TYPE" in env:
+            sys_type = env["SYS_TYPE"]
+        return sys_type
+
+    def initconfig_compiler_entries(self):
+        spec = self.spec
+        entries = super().initconfig_compiler_entries()
+
+        if spec.satisfies("+rocm"):
+            entries.insert(0, cmake_cache_path("CMAKE_CXX_COMPILER", spec["hip"].hipcc))
+
+        entries.append(cmake_cache_option("WITH_FORTRAN", spec.satisfies("+fortran")))
+
+        entries.append(cmake_cache_option("BUILD_SHARED_LIBS", spec.satisfies("+shared")))
+        entries.append(cmake_cache_option("BUILD_TESTING", spec.satisfies("+tests")))
+        entries.append(cmake_cache_option("WITH_TOOLS", spec.satisfies("+tools")))
+        entries.append(cmake_cache_option("BUILD_DOCS", False))
+        entries.append(cmake_cache_path("PYTHON_EXECUTABLE", spec["python"].command.path))
+
+        return entries
+
+    def initconfig_hardware_entries(self):
+        spec = self.spec
+        entries = super().initconfig_hardware_entries()
+
+        if spec.satisfies("+cuda"):
+            entries.append(cmake_cache_option("WITH_CUPTI", True))
+            entries.append(cmake_cache_option("WITH_NVTX", True))
+            entries.append(cmake_cache_path("CUDA_TOOLKIT_ROOT_DIR", spec["cuda"].prefix))
+            entries.append(cmake_cache_path("CUPTI_PREFIX", spec["cuda"].prefix))
+        else:
+            entries.append(cmake_cache_option("WITH_CUPTI", False))
+            entries.append(cmake_cache_option("WITH_NVTX", False))
+
+        if spec.satisfies("+rocm"):
+            entries.append(cmake_cache_option("WITH_ROCTRACER", True))
+            entries.append(cmake_cache_option("WITH_ROCTX", True))
+        else:
+            entries.append(cmake_cache_option("WITH_ROCTRACER", False))
+            entries.append(cmake_cache_option("WITH_ROCTX", False))
+
+        return entries
+
+    def initconfig_mpi_entries(self):
+        spec = self.spec
+        entries = super().initconfig_mpi_entries()
+
+        entries.append(cmake_cache_option("WITH_MPI", spec.satisfies("+mpi")))
+
+        return entries
+
+    def initconfig_package_entries(self):
+        spec = self.spec
+        entries = []
+
+        # TPL locations
+        entries.append("#------------------{0}".format("-" * 60))
+        entries.append("# TPLs")
+        entries.append("#------------------{0}\n".format("-" * 60))
+
+        if spec.satisfies("+adiak"):
+            entries.append(cmake_cache_path("adiak_DIR", spec["adiak"].prefix))
+        if spec.satisfies("+papi"):
+            entries.append(cmake_cache_path("PAPI_PREFIX", spec["papi"].prefix))
+        if spec.satisfies("+libdw"):
+            entries.append(cmake_cache_path("LIBDW_PREFIX", spec["elfutils"].prefix))
+        if spec.satisfies("+libpfm"):
+            entries.append(cmake_cache_path("LIBPFM_INSTALL", spec["libpfm4"].prefix))
+        if spec.satisfies("+sosflow"):
+            entries.append(cmake_cache_path("SOS_PREFIX", spec["sosflow"].prefix))
+        if spec.satisfies("+variorum"):
+            entries.append(cmake_cache_path("VARIORUM_PREFIX", spec["variorum"].prefix))
+        if spec.satisfies("+vtune"):
+            itt_dir = join_path(spec["intel-oneapi-vtune"].prefix, "vtune", "latest")
+            entries.append(cmake_cache_path("ITT_PREFIX", itt_dir))
+        if spec.satisfies("+libunwind"):
+            entries.append(cmake_cache_path("LIBUNWIND_PREFIX", spec["unwind"].prefix))
+
+        # Build options
+        entries.append("#------------------{0}".format("-" * 60))
+        entries.append("# Build Options")
+        entries.append("#------------------{0}\n".format("-" * 60))
+
+        entries.append(cmake_cache_option("WITH_ADIAK", spec.satisfies("+adiak")))
+        entries.append(cmake_cache_option("WITH_GOTCHA", spec.satisfies("+gotcha")))
+        entries.append(cmake_cache_option("WITH_SAMPLER", spec.satisfies("+sampler")))
+        entries.append(cmake_cache_option("WITH_PAPI", spec.satisfies("+papi")))
+        entries.append(cmake_cache_option("WITH_LIBDW", spec.satisfies("+libdw")))
+        entries.append(cmake_cache_option("WITH_LIBPFM", spec.satisfies("+libpfm")))
+        entries.append(cmake_cache_option("WITH_SOSFLOW", spec.satisfies("+sosflow")))
+        entries.append(cmake_cache_option("WITH_KOKKOS", spec.satisfies("+kokkos")))
+        entries.append(cmake_cache_option("WITH_VARIORUM", spec.satisfies("+variorum")))
+        entries.append(cmake_cache_option("WITH_VTUNE", spec.satisfies("+vtune")))
+        entries.append(cmake_cache_option("WITH_PYTHON_BINDINGS", spec.satisfies("+python")))
+
+        # -DWITH_CALLPATH was renamed -DWITH_LIBUNWIND in 2.5
+        callpath_flag = "LIBUNWIND" if spec.satisfies("@2.5:") else "CALLPATH"
+        entries.append(cmake_cache_option("WITH_%s" % callpath_flag, spec.satisfies("+libunwind")))
+
+        return entries
+
+    def cmake_args(self):
+        return []
+
+    def setup_run_environment(self, env):
+        if self.spec.satisfies("+python"):
+            env.prepend_path("PYTHONPATH", self.spec.prefix.join(python_platlib))
+            env.prepend_path("PYTHONPATH", self.spec.prefix.join(python_purelib))
+
+    @run_after("install")
+    def cache_test_sources(self):
+        """Copy the example source files after the package is installed to an
+        install test subdirectory for use during `spack test run`."""
+        cache_extra_test_sources(self, [join_path("examples", "apps")])
+
+    def test_cxx_example(self):
+        """build and run cxx-example"""
+
+        exe = "cxx-example"
+        source_file = "{0}.cpp".format(exe)
+
+        source_path = find_required_file(
+            self.test_suite.current_test_cache_dir, source_file, expected=1, recursive=True
+        )
+
+        lib_dir = self.prefix.lib if os.path.exists(self.prefix.lib) else self.prefix.lib64
+
+        cxx = which(os.environ["CXX"])
+        test_dir = os.path.dirname(source_path)
+        with working_dir(test_dir):
+            cxx(
+                "-L{0}".format(lib_dir),
+                "-I{0}".format(self.prefix.include),
+                source_path,
+                "-o",
+                exe,
+                "-std=c++11",
+                "-lcaliper",
+                "-lstdc++",
+            )
+
+            cxx_example = which(exe)
+            cxx_example()
diff --git a/repo/caliper/sampler-service-missing-libunwind-include-dir.patch b/repo/caliper/sampler-service-missing-libunwind-include-dir.patch
new file mode 100644
index 000000000..4acd660f6
--- /dev/null
+++ b/repo/caliper/sampler-service-missing-libunwind-include-dir.patch
@@ -0,0 +1,14 @@
+diff -ruN spack-src/src/services/sampler/CMakeLists.txt spack-src-patched/src/services/sampler/CMakeLists.txt
+--- spack-src/src/services/sampler/CMakeLists.txt	2022-11-30 13:52:42.000000000 -0500
++++ spack-src-patched/src/services/sampler/CMakeLists.txt	2023-05-04 20:43:47.240310306 -0400
+@@ -17,6 +17,10 @@
+ 
+ include_directories(${CMAKE_CURRENT_BINARY_DIR} ${CMAKE_CURRENT_SOURCE_DIR})
+ 
++if (CALIPER_HAVE_LIBUNWIND)
++	  include_directories(${LIBUNWIND_INCLUDE_DIRS})
++endif()
++
+ add_library(caliper-sampler OBJECT ${CALIPER_SAMPLER_SOURCES})
+ 
+ add_service_objlib("caliper-sampler")
diff --git a/repo/cray-mpich/package.py b/repo/cray-mpich/package.py
index b0f2e1561..36d80d64b 100644
--- a/repo/cray-mpich/package.py
+++ b/repo/cray-mpich/package.py
@@ -17,8 +17,8 @@ def libs(self):
 
         if self.spec.satisfies("+gtl"):
             gtl_lib_prefix = self.spec.extra_attributes["gtl_lib_path"]
-            libraries = ["libmpi_gtl_hsa"]
-            libs += find_libraries(libraries, root=gtl_lib_prefix, recursive=True)
+            gtl_libs = self.spec.extra_attributes["gtl_libs"]
+            libs += find_libraries(gtl_libs, root=gtl_lib_prefix, recursive=True)
 
         return libs
 
@@ -29,8 +29,6 @@ def setup_run_environment(self, env):
         if self.spec.satisfies("+gtl"):
             env.set("MPICH_GPU_SUPPORT_ENABLED", "1")
             env.prepend_path("LD_LIBRARY_PATH", self.spec.extra_attributes["gtl_lib_path"])
-            env.set("GTL_HSA_VSMSG_CUTOFF_SIZE", str(self.spec.extra_attributes["gtl_cutoff_size"]))
-            env.set("FI_CXI_ATS", str(self.spec.extra_attributes["fi_cxi_ats"]))
         else:
             env.set("MPICH_GPU_SUPPORT_ENABLED", "0")
             gtl_path = self.spec.extra_attributes.get("gtl_lib_path", "")
diff --git a/repo/hypre/package.py b/repo/hypre/package.py
index 31613bcd3..8a12e268d 100644
--- a/repo/hypre/package.py
+++ b/repo/hypre/package.py
@@ -54,3 +54,5 @@ def setup_build_environment(self, env):
             if spec.satisfies("+openmp"):
                 libsci_name += "_mp"
             env.append_flags("LDFLAGS", f"-L{spec['lapack'].prefix}/lib -l{libsci_name}")
+        if "+cuda" in self.spec:
+            env.set("NVCC_APPEND_FLAGS", "-allow-unsupported-compiler")
diff --git a/repo/saxpy/CMakeLists.txt b/repo/saxpy/CMakeLists.txt
index fc0c5945b..9369aa4d2 100644
--- a/repo/saxpy/CMakeLists.txt
+++ b/repo/saxpy/CMakeLists.txt
@@ -43,6 +43,7 @@ if(USE_CUDA)
     set_source_files_properties(saxpy.cc PROPERTIES LANGUAGE "CUDA")
     set(CMAKE_CUDA_STANDARD_REQUIRED TRUE)
     set(CMAKE_CXX_STANDARD_REQUIRED TRUE)
+    set(CMAKE_CXX_EXTENSIONS OFF)
     set(CMAKE_CUDA_FLAGS "-Xcompiler -DUSE_CUDA")
     set_target_properties(${PROJECT_NAME} PROPERTIES CUDA_SEPARABLE_COMPILATION ON)
 endif()
diff --git a/repo/saxpy/package.py b/repo/saxpy/package.py
index 56f0c211e..25dc76ce0 100644
--- a/repo/saxpy/package.py
+++ b/repo/saxpy/package.py
@@ -42,6 +42,7 @@ def stage_source(self):
     def setup_build_environment(self, env):
         if "+cuda" in self.spec:
             env.set("CUDAHOSTCXX", self.spec["mpi"].mpicxx)
+            env.set("NVCC_APPEND_FLAGS", "-std=c++14 -allow-unsupported-compiler")
 
     def cmake_args(self):
         spec = self.spec
diff --git a/systems/all_system_definitions/HPECray-neoverse-H100-Slingshot/system_definition.yaml b/systems/all_system_definitions/HPECray-neoverse-H100-Slingshot/system_definition.yaml
new file mode 100644
index 000000000..982de3828
--- /dev/null
+++ b/systems/all_system_definitions/HPECray-neoverse-H100-Slingshot/system_definition.yaml
@@ -0,0 +1,32 @@
+# Copyright 2023 Lawrence Livermore National Security, LLC and other
+# Benchpark Project Developers. See the top-level COPYRIGHT file for details.
+#
+# SPDX-License-Identifier: Apache-2.0
+
+system_definition:
+  name: HPECray-neoverse-H100-Slingshot
+  integrator:
+    vendor: HPECray
+    name: EX254n
+  processor:
+    vendor: NVIDIA
+    name: Grace
+    ISA: Armv9
+    uArch: neoverse
+  accelerator:
+    vendor: NVIDIA
+    name: H100
+    ISA: PTX
+    uArch: sm_90
+  interconnect:
+    vendor: HPECray
+    name: Slingshot11
+  systems-tested:
+    lanl-venado:
+      os: HPECrayOS
+      scheduler: slurm
+      compiler: cce
+      runtime: cuda
+      mpi: cray-mpich
+  top500-system-instances:
+    - Venado (LANL) <https://www.top500.org/system/180246>
diff --git a/systems/lanl-venado/compilers/cce/00-cce-18-compilers.yaml b/systems/lanl-venado/compilers/cce/00-cce-18-compilers.yaml
new file mode 100644
index 000000000..7ce4aeb35
--- /dev/null
+++ b/systems/lanl-venado/compilers/cce/00-cce-18-compilers.yaml
@@ -0,0 +1,25 @@
+compilers:
+- compiler:
+    spec: cce@18.0.0
+    paths:
+      cc: /opt/cray/pe/cce/18.0.0/bin/craycc
+      cxx: /opt/cray/pe/cce/18.0.0/bin/crayCC
+      f77: /opt/cray/pe/cce/18.0.0/bin/crayftn
+      fc: /opt/cray/pe/cce/18.0.0/bin/crayftn
+    flags:
+      cflags: -g -O2 --gcc-toolchain=/usr/projects/hpcsoft/tce/24-07/cos3-aarch64-cc90/compilers/gcc/12.3.0
+      cxxflags: -g -O2 --gcc-toolchain=/usr/projects/hpcsoft/tce/24-07/cos3-aarch64-cc90/compilers/gcc/12.3.0
+      fflags: -g -O2 -hnopattern
+      ldflags: -ldl
+    operating_system: sles15
+    target: aarch64
+    modules: []
+    environment:
+      prepend_path:
+        LD_LIBRARY_PATH: "/opt/cray/pe/cce/18.0.0/cce/aarch64/lib:/opt/cray/libfabric/1.20.1/lib64:/usr/projects/hpcsoft/tce/24-07/cos3-aarch64-cc90/compilers/gcc/12.3.0/lib:/usr/projects/hpcsoft/tce/24-07/cos3-aarch64-cc90/compilers/gcc/12.3.0/lib64:/opt/cray/pe/gcc-libs"
+    extra_rpaths:
+    - /opt/cray/pe/gcc-libs
+    - /opt/cray/pe/cce/18.0.0/cce/aarch64/lib
+    - /opt/cray/libfabric/1.20.1/lib64
+    - /usr/projects/hpcsoft/tce/24-07/cos3-aarch64-cc90/compilers/gcc/12.3.0/lib
+    - /usr/projects/hpcsoft/tce/24-07/cos3-aarch64-cc90/compilers/gcc/12.3.0/lib64
diff --git a/systems/lanl-venado/compilers/gcc/00-gcc-12-compilers.yaml b/systems/lanl-venado/compilers/gcc/00-gcc-12-compilers.yaml
new file mode 100644
index 000000000..f28815e8f
--- /dev/null
+++ b/systems/lanl-venado/compilers/gcc/00-gcc-12-compilers.yaml
@@ -0,0 +1,14 @@
+compilers:
+- compiler:
+    spec: gcc@12.3.0
+    paths:
+      cc: /usr/projects/hpcsoft/tce/24-07/cos3-aarch64-cc90/compilers/gcc/12.3.0/bin/gcc
+      cxx: /usr/projects/hpcsoft/tce/24-07/cos3-aarch64-cc90/compilers/gcc/12.3.0/bin/g++
+      f77: /usr/projects/hpcsoft/tce/24-07/cos3-aarch64-cc90/compilers/gcc/12.3.0/bin/gfortran
+      fc: /usr/projects/hpcsoft/tce/24-07/cos3-aarch64-cc90/compilers/gcc/12.3.0/bin/gfortran
+    flags: {}
+    operating_system: sles15
+    target: aarch64
+    modules: []
+    environment: {}
+    extra_rpaths: []
diff --git a/systems/lanl-venado/externals/base/00-packages.yaml b/systems/lanl-venado/externals/base/00-packages.yaml
new file mode 100644
index 000000000..63682360b
--- /dev/null
+++ b/systems/lanl-venado/externals/base/00-packages.yaml
@@ -0,0 +1,43 @@
+# Copyright 2023 Lawrence Livermore National Security, LLC and other
+# Benchpark Project Developers. See the top-level COPYRIGHT file for details.
+#
+# SPDX-License-Identifier: Apache-2.0
+
+packages:
+  tar:
+    externals:
+    - spec: tar@1.34
+      prefix: /usr
+    buildable: false
+  cmake:
+    externals:
+    - spec: cmake@3.29.6
+      prefix: /usr/projects/hpcsoft/tce/24-07/cos3-aarch64-cc90/packages/cmake/cmake-3.29.6
+    buildable: false
+  gmake:
+    externals:
+    - spec: gmake@4.2.1
+      prefix: /usr
+    buildable: false
+  automake:
+    externals:
+    - spec: automake@1.15.1
+      prefix: /usr
+    buildable: false
+  autoconf:
+    externals:
+    - spec: autoconf@2.69
+      prefix: /usr
+    buildable: false
+  fftw:
+    externals:
+    - spec: fftw@3.3.10.8
+      prefix: /opt/cray/pe/fftw/3.3.10.8/arm_grace
+    buildable: false
+  python:
+    externals:
+    - spec: python@3.10.9
+      prefix: /usr/projects/hpcsoft/common/aarch64/anaconda/2023.03-python-3.10
+    buildable: false
+  mpi:
+    buildable: false
diff --git a/systems/lanl-venado/externals/libsci/00-gcc-packages.yaml b/systems/lanl-venado/externals/libsci/00-gcc-packages.yaml
new file mode 100644
index 000000000..f58d5a59a
--- /dev/null
+++ b/systems/lanl-venado/externals/libsci/00-gcc-packages.yaml
@@ -0,0 +1,5 @@
+packages:
+  cray-libsci:
+    externals:
+    - spec: cray-libsci@24.07.0%gcc
+      prefix: /opt/cray/pe/libsci/24.07.0/gnu/12.3/aarch64
diff --git a/systems/lanl-venado/externals/libsci/01-cce-packages.yaml b/systems/lanl-venado/externals/libsci/01-cce-packages.yaml
new file mode 100644
index 000000000..8c1608b6b
--- /dev/null
+++ b/systems/lanl-venado/externals/libsci/01-cce-packages.yaml
@@ -0,0 +1,5 @@
+packages:
+  cray-libsci:
+    externals:
+    - spec: cray-libsci@24.07.0%cce
+      prefix: /opt/cray/pe/libsci/24.07.0/cray/17.0/aarch64
diff --git a/systems/lanl-venado/system.py b/systems/lanl-venado/system.py
new file mode 100644
index 000000000..67e7d2038
--- /dev/null
+++ b/systems/lanl-venado/system.py
@@ -0,0 +1,220 @@
+# Copyright 2023 Lawrence Livermore National Security, LLC and other
+# Benchpark Project Developers. See the top-level COPYRIGHT file for details.
+#
+# SPDX-License-Identifier: Apache-2.0
+
+import pathlib
+
+from benchpark.directives import variant
+from benchpark.system import System
+
+id_to_resources = {
+    "grace-hopper": {
+        "sys_cores_per_node": 144,
+        "sys_gpus_per_node": 4,
+    },
+    "grace-grace": {
+        "sys_cores_per_node": 144,
+    },
+}
+
+
+class LanlVenado(System):
+    variant(
+        "cluster",
+        default="grace-hopper",
+        values=("grace-hopper", "grace-grace"),
+        description="Which cluster to run on",
+    )
+
+    variant(
+        "cuda",
+        default="12-5",
+        values=("11.8", "12.5"),
+        description="CUDA version",
+    )
+
+    variant(
+        "compiler",
+        default="cce",
+        values=("gcc", "cce"),
+        description="Which compiler to use",
+    )
+
+    variant(
+        "gtl",
+        default=False,
+        values=(True, False),
+        description="Use GTL-enabled MPI",
+    )
+
+    variant(
+        "lapack",
+        default="cusolver",
+        values=("cusolver", "cray-libsci"),
+        description="Which lapack to use",
+    )
+
+    variant(
+        "blas",
+        default="cublas",
+        values=("cublas", "cray-libsci"),
+        description="Which blas to use",
+    )
+
+    def initialize(self):
+        super().initialize()
+
+        self.scheduler = "slurm"
+        attrs = id_to_resources.get(self.spec.variants["cluster"][0])
+        for k, v in attrs.items():
+            setattr(self, k, v)
+
+    def generate_description(self, output_dir):
+        super().generate_description(output_dir)
+
+        sw_description = pathlib.Path(output_dir) / "software.yaml"
+
+        with open(sw_description, "w") as f:
+            f.write(self.sw_description())
+
+    def system_specific_variables(self):
+        return {
+            "cuda_arch": "90",
+            "default_cuda_version": self.spec.variants["cuda"][0],
+            "extra_batch_opts": '"-A llnl_ai_g -pgpu"',
+        }
+
+    def external_pkg_configs(self):
+        externals = LanlVenado.resource_location / "externals"
+
+        selections = [externals / "base" / "00-packages.yaml"]
+
+        cuda_cfg_path = self.next_adhoc_cfg()
+        with open(cuda_cfg_path, "w") as f:
+            f.write(self.cuda_config(self.spec.variants["cuda"][0]))
+        selections.append(cuda_cfg_path)
+
+        mpi_cfg_path = self.next_adhoc_cfg()
+        with open(mpi_cfg_path, "w") as f:
+            f.write(self.mpi_config())
+        selections.append(mpi_cfg_path)
+
+        if self.spec.satisfies("compiler=cce"):
+            selections.append(externals / "libsci" / "01-cce-packages.yaml")
+        elif self.spec.satisfies("compiler=gcc"):
+            selections.append(externals / "libsci" / "00-gcc-packages.yaml")
+
+        return selections
+
+    def compiler_configs(self):
+        compilers = LanlVenado.resource_location / "compilers"
+
+        selections = []
+        # TODO: Construct/extract/customize compiler information from the working set
+        if self.spec.satisfies("compiler=cce"):
+            selections.append(compilers / "cce" / "00-cce-18-compilers.yaml")
+        selections.append(compilers / "gcc" / "00-gcc-12-compilers.yaml")
+
+        return selections
+
+    def mpi_config(self):
+        mpi_version = "8.1.30"
+        gtl = (
+            "+gtl"
+            if self.spec.satisfies("compiler=cce") and self.spec.satisfies("+gtl")
+            else "~gtl"
+        )
+
+        # TODO: Construct/extract this information from the working set
+        if self.spec.satisfies("compiler=cce"):
+            compiler = "cce@18.0.0"
+            mpi_compiler_suffix = "crayclang/17.0"
+        elif self.spec.satisfies("compiler=gcc"):
+            compiler = "gcc@12.3.0"
+            mpi_compiler_suffix = "gnu/12.3"
+
+        return f"""\
+packages:
+  cray-mpich:
+    externals:
+    - spec: cray-mpich@{mpi_version}%{compiler} {gtl} +wrappers
+      prefix: /opt/cray/pe/mpich/{mpi_version}/ofi/{mpi_compiler_suffix}
+      extra_attributes:
+        gtl_lib_path: /opt/cray/pe/mpich/{mpi_version}/gtl/lib
+        gtl_libs: ["libmpi_gtl_cuda"]
+        ldflags: "-L/opt/cray/pe/mpich/{mpi_version}/ofi/{mpi_compiler_suffix}/lib -lmpi -L/opt/cray/pe/mpich/{mpi_version}/gtl/lib -Wl,-rpath=/opt/cray/pe/mpich/{mpi_version}/gtl/lib -lmpi_gtl_cuda"
+"""
+
+    def cuda_config(self, cuda_version):
+        template = """\
+packages:
+  blas:
+    require:
+      - {blas}
+  lapack:
+    require:
+      - {lapack}
+  curand:
+    externals:
+    - spec: curand@{x}
+      prefix: /opt/nvidia/hpc_sdk/Linux_aarch64/24.7/cuda/{x}
+    buildable: false
+  cusparse:
+    externals:
+    - spec: cusparse@{x}
+      prefix: /opt/nvidia/hpc_sdk/Linux_aarch64/24.7/cuda/{x}
+    buildable: false
+  cuda:
+    externals:
+    - spec: cuda@{x}
+      prefix: /opt/nvidia/hpc_sdk/Linux_aarch64/24.7/cuda/{x}
+    buildable: false
+  cub:
+    externals:
+    - spec: cub@{x}
+      prefix: /opt/nvidia/hpc_sdk/Linux_aarch64/24.7/cuda/{x}
+    buildable: false
+  cublas:
+    externals:
+    - spec: cublas@{x}
+      prefix: /opt/nvidia/hpc_sdk/Linux_aarch64/24.7/math_libs/{x}
+    buildable: false
+  cusolver:
+    externals:
+    - spec: cusolver@{x}
+      prefix: /opt/nvidia/hpc_sdk/Linux_aarch64/24.7/math_libs/{x}
+    buildable: false
+  cufft:
+    externals:
+    - spec: cufft@{x}
+      prefix: /opt/nvidia/hpc_sdk/Linux_aarch64/24.7/math_libs/{x}
+    buildable: false
+"""
+        return template.format(
+            x=cuda_version,
+            blas=self.spec.variants["blas"][0],
+            lapack=self.spec.variants["lapack"][0],
+        )
+
+    def sw_description(self):
+        """This is somewhat vestigial: for the Tioga config that is committed
+        to the repo, multiple instances of mpi/compilers are stored and
+        and these variables were used to choose consistent dependencies.
+        The configs generated by this class should only ever have one
+        instance of MPI etc., so there is no need for that. The experiments
+        will fail if these variables are not defined though, so for now
+        they are still generated (but with more-generic values).
+        """
+        return f"""\
+software:
+  packages:
+    default-compiler:
+      pkg_spec: {self.spec.variants["compiler"][0]}
+    default-mpi:
+      pkg_spec: cray-mpich
+    default-lapack:
+      pkg_spec: {self.spec.variants["lapack"][0]}
+    default-blas:
+      pkg_spec: {self.spec.variants["blas"][0]}
+"""
diff --git a/systems/llnl-elcapitan/system.py b/systems/llnl-elcapitan/system.py
index 1f4b4fb30..335e39390 100644
--- a/systems/llnl-elcapitan/system.py
+++ b/systems/llnl-elcapitan/system.py
@@ -163,6 +163,7 @@ def mpi_config(self, cce_version):
         gtl_cutoff_size: 4096
         fi_cxi_ats: 0
         gtl_lib_path: /opt/cray/pe/mpich/{mpi_version}/gtl/lib
+        gtl_libs: ["libmpi_gtl_hsa"]
         ldflags: "-L/opt/cray/pe/mpich/{mpi_version}/ofi/crayclang/{short_cce_version}/lib -lmpi -L/opt/cray/pe/mpich/{mpi_version}/gtl/lib -Wl,-rpath=/opt/cray/pe/mpich/{mpi_version}/gtl/lib -lmpi_gtl_hsa"
 """