added flag --rdma-mpi

Author: Malmahrouqi3

toolchain/mfc/args.py

@@ -76,14 +76,15 @@ def add_common_arguments(p, mask = None):
     test.add_argument("-l", "--list",         action="store_true", help="List all available tests.")
     test.add_argument("-f", "--from",         default=test_cases[0].get_uuid(), type=str, help="First test UUID to run.")
     test.add_argument("-t", "--to",           default=test_cases[-1].get_uuid(), type=str, help="Last test UUID to run.")
-    test.add_argument("-o", "--only",         nargs="+", type=str, default=[], metavar="L", help="Only run tests with specified properties.")
-    test.add_argument("-a", "--test-all",     action="store_true", default=False, help="Run the Post Process Tests too.")
-    test.add_argument("-%", "--percent",      type=int, default=100, help="Percentage of tests to run.")
-    test.add_argument("-m", "--max-attempts", type=int, default=1, help="Maximum number of attempts to run a test.")
-    test.add_argument(      "--no-build",     action="store_true",                    default=False,      help="(Testing) Do not rebuild MFC.")
-    test.add_argument(      "--no-examples",  action="store_true",                    default=False,      help="Do not test example cases." )
-    test.add_argument("--case-optimization",  action="store_true", default=False, help="(GPU Optimization) Compile MFC targets with some case parameters hard-coded.")
-    test.add_argument(      "--dry-run",      action="store_true",                    default=False,      help="Build and generate case files but do not run tests.")
+    test.add_argument("-o", "--only",         nargs="+", type=str,     default=[], metavar="L", help="Only run tests with specified properties.")
+    test.add_argument("-a", "--test-all",     action="store_true",     default=False,     help="Run the Post Process Tests too.")
+    test.add_argument("-%", "--percent",      type=int,                default=100,       help="Percentage of tests to run.")
+    test.add_argument("-m", "--max-attempts", type=int,                default=1,         help="Maximum number of attempts to run a test.")
+    test.add_argument("-r", "--rdma-mpi",     action="store_true",     default=False,     help="Enable RDMA MPI for tests")
+    test.add_argument(      "--no-build",     action="store_true",     default=False,     help="(Testing) Do not rebuild MFC.")
+    test.add_argument(      "--no-examples",  action="store_true",     default=False,     help="Do not test example cases." )
+    test.add_argument("--case-optimization",  action="store_true",     default=False,     help="(GPU Optimization) Compile MFC targets with some case parameters hard-coded.")
+    test.add_argument(      "--dry-run",      action="store_true",     default=False,     help="Build and generate case files but do not run tests.")
 
     test_meg = test.add_mutually_exclusive_group()
     test_meg.add_argument("--generate",          action="store_true", default=False, help="(Test Generation) Generate golden files.")

toolchain/mfc/test/case.py

@@ -133,7 +133,9 @@ def run(self, targets: List[Union[str, MFCTarget]], gpus: Set[int]) -> subproces
         filepath          = f'{self.get_dirpath()}/case.py'
         tasks             = ["-n", str(self.ppn)]
         jobs              = ["-j", str(ARG("jobs"))] if ARG("case_optimization") else []
-        case_optimization = ["--case-optimization"] if ARG("case_optimization") else []
+        case_optimization = ["--case-optimization"]  if ARG("case_optimization") else []
+        rdma_mpi_args     = ["--rdma-mpi"]           if ARG("rdma_mpi")          else []
+
 
         if self.params.get("bubbles_lagrange", 'F') == 'T':
             input_bubbles_lagrange(self)
@@ -144,7 +146,7 @@ def run(self, targets: List[Union[str, MFCTarget]], gpus: Set[int]) -> subproces
 
         command = [
             mfc_script, "run", filepath, "--no-build", *tasks, *case_optimization,
-            *jobs, "-t", *target_names, *gpus_select, *ARG("--")
+            *jobs, "-t", *target_names, *gpus_select, *rdma_mpi_args, *ARG("--")
         ]
 
         return common.system(command, print_cmd=False, text=True, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)

toolchain/mfc/test/test.py

@@ -64,6 +64,10 @@ def __filter(cases_) -> typing.List[TestCase]:
             if any(label in case.trace for label in skip):
                 cases.remove(case)
 
+    for case in cases[:]:
+        if ARG("rdma_mpi") and case.ppn <= 1:
+            cases.remove(case)
+            skipped_cases.append(case)
 
     if ARG("no_examples"):
         cases = [case for case in cases if not "Example" in case.trace]
@@ -180,7 +184,10 @@ def _handle_case(case: TestCase, devices: typing.Set[int]):
         cons.print(f"  [bold magenta]{case.get_uuid()}[/bold magenta]     SKIP     {case.trace}")
         return
 
-    cmd = case.run([PRE_PROCESS, SIMULATION], gpus=devices)
+    if ARG("rdma_mpi"):
+        cmd = case.run([PRE_PROCESS, SIMULATION], gpus=devices, rdma_mpi=True)
+    else:
+        cmd = case.run([PRE_PROCESS, SIMULATION], gpus=devices)
     out_filepath = os.path.join(case.get_dirpath(), "out_pre_sim.txt")
 
     common.file_write(out_filepath, cmd.stdout)
@@ -224,6 +231,10 @@ def _handle_case(case: TestCase, devices: typing.Set[int]):
     if ARG("test_all"):
         case.delete_output()
         cmd = case.run([PRE_PROCESS, SIMULATION, POST_PROCESS], gpus=devices)
+        if ARG("rdma_mpi"):
+            cmd = case.run([PRE_PROCESS, SIMULATION, POST_PROCESS], gpus=devices, rdma_mpi=True)
+        else:
+            cmd = case.run([PRE_PROCESS, SIMULATION, POST_PROCESS], gpus=devices)
         out_filepath = os.path.join(case.get_dirpath(), "out_post.txt")
         common.file_write(out_filepath, cmd.stdout)