Abstract HPCG and add NVIDIA-HPCG

var/ramble/repos/builtin/applications/hpcg/application.py

@@ -6,20 +6,26 @@
 # option. This file may not be copied, modified, or distributed
 # except according to those terms.
 
-import os
 from ramble.appkit import *
-from ramble.expander import Expander
+from ramble.base_app.builtin.hpcg import Hpcg as BaseHpcg
 
 
-class Hpcg(ExecutableApplication):
-    """Define HPCG application"""
+class Hpcg(BaseHpcg):
+    """The High Performance Conjugate Gradients (HPCG) Benchmark project is an
+    effort to create a new metric for ranking HPC systems. HPCG is intended as
+    a complement to the High Performance LINPACK (HPL) benchmark, currently
+    used to rank the TOP500 computing systems. The computational and data
+    access patterns of HPL are still representative of some important scalable
+    applications, but not all. HPCG is designed to exercise computational and
+    data access patterns that more closely match a different and broad set of
+    important applications, and to give incentive to computer system designers
+    to invest in capabilities that will have impact on the collective
+    performance of these applications."""
 
     name = "hpcg"
 
     maintainers("douglasjacobsen")
 
-    tags("benchmark-app", "mini-app", "benchmark")
-
     define_compiler("gcc9", pkg_spec="[email protected]", package_manager="spack*")
 
     software_spec(
@@ -35,110 +41,6 @@ class Hpcg(ExecutableApplication):
 
     required_package("hpcg", package_manager="spack*")
 
-    executable("execute", "xhpcg", use_mpi=True)
-
-    executable("move-log", "mv HPCG-Benchmark*.txt {out_file}", use_mpi=False)
-
     workload("standard", executables=["execute", "move-log"])
 
-    workload_variable(
-        "matrix_size",
-        default="104 104 104",
-        description="Dimensions of the matrix to use",
-        workloads=["standard"],
-    )
-
-    workload_variable(
-        "iterations",
-        default="60",
-        description="Number of iterations to perform",
-        workloads=["standard"],
-    )
-
-    workload_variable(
-        "out_file",
-        default="{experiment_run_dir}/hpcg_result.out",
-        description="Output file for results",
-        workloads=["standard"],
-    )
-
-    log_str = Expander.expansion_str("out_file")
-
-    figure_of_merit(
-        "Status",
-        log_file=log_str,
-        fom_regex=r"Final Summary::HPCG result is (?P<status>[a-zA-Z]+) with a GFLOP/s rating of=(?P<gflops>[0-9]+\.[0-9]+)",
-        group_name="status",
-        units="",
-    )
-
-    figure_of_merit(
-        "Gflops",
-        log_file=log_str,
-        fom_regex=r"Final Summary::HPCG result is (?P<status>[a-zA-Z]+) with a GFLOP/s rating of=(?P<gflops>[0-9]+\.[0-9]+)",
-        group_name="gflops",
-        units="GFLOP/s",
-    )
-
-    figure_of_merit(
-        "Time",
-        log_file=log_str,
-        fom_regex=r"Final Summary::Results are.* execution time.*is=(?P<exec_time>[0-9]+\.[0-9]*)",
-        group_name="exec_time",
-        units="s",
-    )
-
-    figure_of_merit(
-        "ComputeDotProductMsg",
-        log_file=log_str,
-        fom_regex=r"Final Summary::Reference version of ComputeDotProduct used.*=(?P<msg>.*)",
-        group_name="msg",
-        units="",
-    )
-
-    figure_of_merit(
-        "ComputeSPMVMsg",
-        log_file=log_str,
-        fom_regex=r"Final Summary::Reference version of ComputeSPMV used.*=(?P<msg>.*)",
-        group_name="msg",
-        units="",
-    )
-
-    figure_of_merit(
-        "ComputeMGMsg",
-        log_file=log_str,
-        fom_regex=r"Final Summary::Reference version of ComputeMG used.*=(?P<msg>.*)",
-        group_name="msg",
-        units="",
-    )
-
-    figure_of_merit(
-        "ComputeWAXPBYMsg",
-        log_file=log_str,
-        fom_regex=r"Final Summary::Reference version of ComputeWAXPBY used.*=(?P<msg>.*)",
-        group_name="msg",
-        units="",
-    )
-
-    figure_of_merit(
-        "HPCG 2.4 Rating",
-        log_file=log_str,
-        fom_regex=r"Final Summary::HPCG 2\.4 rating.*=(?P<rating>[0-9]+\.*[0-9]*)",
-        group_name="rating",
-        units="",
-    )
-
-    def _make_experiments(self, workspace, app_inst=None):
-        super()._make_experiments(workspace)
-
-        input_path = os.path.join(
-            self.expander.expand_var_name("experiment_run_dir"), "hpcg.dat"
-        )
-
-        with open(input_path, "w+") as f:
-            f.write("HPCG benchmark input file\n")
-            f.write(
-                "Sandia National Laboratories; University of Tennessee, Knoxville\n"
-            )
-            f.write(self.expander.expand_var_name("matrix_size") + "\n")
-            f.write(self.expander.expand_var_name("iterations") + "\n")
+    workload_group("all_workloads", workloads=["standard"], mode="append")

var/ramble/repos/builtin/applications/nvidia-hpcg/application.py

@@ -0,0 +1,149 @@
+# Copyright 2022-2025 The Ramble Authors
+#
+# Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
+# https://www.apache.org/licenses/LICENSE-2.0> or the MIT license
+# <LICENSE-MIT or https://opensource.org/licenses/MIT>, at your
+# option. This file may not be copied, modified, or distributed
+# except according to those terms.
+
+from ramble.appkit import *
+from ramble.base_app.builtin.hpcg import Hpcg as BaseHpcg
+
+
+class NvidiaHpcg(BaseHpcg):
+    """NVIDIA's HPCG benchmark accelerates the High Performance Conjugate
+    Gradients (HPCG) Benchmark. HPCG is a software package that performs a
+    fixed number of multigrid preconditioned (using a symmetric Gauss-Seidel
+    smoother) conjugate gradient (PCG) iterations using double precision
+    (64-bit) floating point values."""
+
+    name = "nvidia-hpcg"
+
+    maintainers("douglasjacobsen")
+
+    executable(
+        "execute",
+        "./hpcg.sh --dat {experiment_run_dir}/hpcg.dat",
+        use_mpi=True,
+    )
+
+    workload("standard", executables=["execute"])
+
+    workload_group("all_workloads", workloads=["standard"], mode="append")
+
+    workload_variable(
+        "nvshmem_disable_cuda_vmm",
+        default="1",
+        description="",
+        workload_group="all_workloads",
+    )
+    environment_variable(
+        "NVSHMEM_DISABLE_CUDA_VMM",
+        "{nvshmem_disable_cuda_vmm}",
+        description="",
+        workload_group="all_workloads",
+    )
+
+    workload_variable(
+        "hpl_fct_comm_policy",
+        default="1",
+        description="",
+        workload_group="all_workloads",
+    )
+    environment_variable(
+        "HPL_FCT_COMM_POLICY",
+        "{hpl_fct_comm_policy}",
+        description="",
+        workload_group="all_workloads",
+    )
+
+    workload_variable(
+        "hpl_use_nvshmem",
+        default="0",
+        description="Whether to use NVSHMEM or not",
+        workload_group="all_workloads",
+    )
+    environment_variable(
+        "HPL_USE_NVSHMEM",
+        "{hpl_use_nvshmem}",
+        description="Whether or not to use NVSHMEM",
+        workload_group="all_workloads",
+    )
+
+    workload_variable(
+        "hpl_p2p_as_bcast",
+        default="0",
+        description="0 = ncclBcast, 1 = ncclSend/Recv",
+        workload_group="all_workloads",
+    )
+    environment_variable(
+        "HPL_P2P_AS_BCAST",
+        "{hpl_p2p_as_bcast}",
+        description="Whether or not to use P2P for BCAST",
+        workload_group="all_workloads",
+    )
+
+    workload_variable(
+        "pmix_mca_gds",
+        default="^ds12",
+        description="",
+        workload_group="all_workloads",
+    )
+    environment_variable(
+        "PMIX_MCA_gds",
+        "{pmix_mca_gds}",
+        description="PMIX MCA gds",
+        workload_group="all_workloads",
+    )
+
+    workload_variable(
+        "ompi_mca_btl",
+        default="^vader,tcp,openib,uct",
+        description="",
+        workload_group="all_workloads",
+    )
+    environment_variable(
+        "OMPI_MCA_btl",
+        "{ompi_mca_btl}",
+        description="OpenMPI MCA btl",
+        workload_group="all_workloads",
+    )
+
+    workload_variable(
+        "ompi_mca_pml",
+        default="ucx",
+        description="",
+        workload_group="all_workloads",
+    )
+    environment_variable(
+        "OMPI_MCA_pml",
+        "{ompi_mca_pml}",
+        description="OpenMPI MCA pml",
+        workload_group="all_workloads",
+    )
+
+    workload_variable(
+        "ucx_net_devices",
+        default="enp6s0,enp12s0,enp134s0,enp140s0",
+        description="",
+        workload_group="all_workloads",
+    )
+    environment_variable(
+        "UCX_NET_DEVICES",
+        "{ucx_net_devices}",
+        description="UCX Net Devices",
+        workload_group="all_workloads",
+    )
+
+    workload_variable(
+        "ucx_max_rndv_rails",
+        default="4",
+        description="",
+        workload_group="all_workloads",
+    )
+    environment_variable(
+        "UCX_MAX_RNDV_RAILS",
+        "{ucx_max_rndv_rails}",
+        description="UCX MAximum RNDV Rails",
+        workload_group="all_workloads",
+    )