Fix BSSN build config selection

bssnEScalar_class::Initialize() already calls setup_transfer_caches(),
but bssnEM_class::Initialize() did not. When USE_TRANSFER_CACHE=1,
the sync_cache pointers remain NULL, causing SIGSEGV in wrapper
methods that dereference sync_cache_*[lev].

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

AMSS_NCKU_GPUCheck.py

@@ -1,559 +0,0 @@
-#!/usr/bin/env python3
-#
-# Current most stable GPU-branch baseline:
-#   GPU_Calculation="yes"
-#   Equation_Class="BSSN"
-#   Initial_Data_Method="Ansorg-TwoPuncture"
-#   puncture_data_set="Manually"
-#   basic_grid_set="Patch"
-#   grid_center_set="Cell"
-#   Symmetry="equatorial-symmetry"
-#   Time_Evolution_Method="runge-kutta-45"
-#   Finite_Diffenence_Method="4th-order"
-#   boundary_choice="BAM-choice"
-#   gauge_choice=0
-#   tetrad_type=2
-#   AHF_Find="no"
-#   devide_factor=2.0
-#   static_grid_type="Linear"
-#   moving_grid_type="Linear"
-#   AMSS_Z4C_MRBD=0
-#   Do not enable AMSS_CUDA_BH_INTERP_RESIDENT unless a dedicated
-#   CPU/GPU trajectory comparison has been run for that configuration.
-"""
-Check whether AMSS_NCKU_Input.py is suitable for the current GPU branch.
-
-Usage:
-  python3 AMSS_NCKU_GPUCheck.py
-  python3 AMSS_NCKU_GPUCheck.py -f /path/to/AMSS_NCKU_Input.py
-"""
-
-from __future__ import annotations
-
-import argparse
-import importlib.util
-import os
-from dataclasses import dataclass, field
-from pathlib import Path
-from typing import Any, Iterable, List, Sequence
-
-
-SUPPORTED_EQUATIONS = {"BSSN", "BSSN-EScalar", "BSSN-EM", "Z4C"}
-SUPPORTED_INITIAL_DATA = {
-    "Ansorg-TwoPuncture",
-    "Lousto-Analytical",
-    "Cao-Analytical",
-    "KerrSchild-Analytical",
-}
-SUPPORTED_SYMMETRIES = {
-    "no-symmetry",
-    "equatorial-symmetry",
-    "octant-symmetry",
-}
-SUPPORTED_GRIDS = {"Patch", "Shell-Patch"}
-SUPPORTED_CENTERS = {"Cell", "Vertex"}
-SUPPORTED_FD = {"2nd-order", "4th-order", "6th-order", "8th-order"}
-SUPPORTED_GAUGES = {0, 1, 2, 3, 4, 5, 6, 7}
-SUPPORTED_TETRADS = {0, 1, 2}
-SUPPORTED_AHF = {"yes", "no"}
-SUPPORTED_BOUNDARIES = {"BAM-choice", "Shibata-choice"}
-SUPPORTED_PUNCTURE_DATA = {"Manually", "Automatically-BBH"}
-
-STABLE_BASELINE = {
-    "GPU_Calculation": "yes",
-    "Equation_Class": "BSSN",
-    "Initial_Data_Method": "Ansorg-TwoPuncture",
-    "puncture_data_set": "Manually",
-    "basic_grid_set": "Patch",
-    "grid_center_set": "Cell",
-    "Symmetry": "equatorial-symmetry",
-    "Time_Evolution_Method": "runge-kutta-45",
-    "Finite_Diffenence_Method": "4th-order",
-    "boundary_choice": "BAM-choice",
-    "gauge_choice": 0,
-    "tetrad_type": 2,
-    "AHF_Find": "no",
-    "devide_factor": 2.0,
-    "static_grid_type": "Linear",
-    "moving_grid_type": "Linear",
-    "AMSS_Z4C_MRBD": 0,
-}
-
-
-@dataclass
-class CheckResult:
-    ok: bool = True
-    warnings: List[str] = field(default_factory=list)
-    risks: List[str] = field(default_factory=list)
-    notes: List[str] = field(default_factory=list)
-
-    def add_warning(self, msg: str) -> None:
-        self.warnings.append(msg)
-
-    def add_risk(self, msg: str) -> None:
-        self.ok = False
-        self.risks.append(msg)
-
-    def add_note(self, msg: str) -> None:
-        self.notes.append(msg)
-
-    def extend_notes(self, messages: Iterable[str]) -> None:
-        self.notes.extend(messages)
-
-
-def load_input_module(path: Path):
-    spec = importlib.util.spec_from_file_location("amss_ncku_input", str(path))
-    if spec is None or spec.loader is None:
-        raise RuntimeError(f"cannot load input module from {path}")
-    module = importlib.util.module_from_spec(spec)
-    spec.loader.exec_module(module)  # type: ignore[union-attr]
-    return module
-
-
-def get_attr(mod: Any, name: str, default: Any = None) -> Any:
-    return getattr(mod, name, default)
-
-
-def as_text(value: Any) -> str:
-    if isinstance(value, str):
-        return value.strip()
-    return str(value).strip()
-
-
-def as_lower_text(value: Any) -> str:
-    return as_text(value).lower()
-
-
-def as_float(value: Any, default: float | None = None) -> float | None:
-    try:
-        return float(value)
-    except (TypeError, ValueError):
-        return default
-
-
-def as_int(value: Any, default: int | None = None) -> int | None:
-    try:
-        return int(value)
-    except (TypeError, ValueError):
-        return default
-
-
-def sequence_len(value: Any) -> int | None:
-    try:
-        return len(value)
-    except TypeError:
-        return None
-
-
-def sequence_values(value: Any) -> List[float] | None:
-    try:
-        return [float(v) for v in value]
-    except (TypeError, ValueError):
-        return None
-
-
-def approx_equal(a: Any, b: float, tol: float = 1.0e-12) -> bool:
-    value = as_float(a)
-    return value is not None and abs(value - b) <= tol
-
-
-def env_truthy(name: str) -> bool:
-    value = os.environ.get(name)
-    return value is not None and value.strip().lower() in {
-        "1",
-        "yes",
-        "y",
-        "true",
-        "on",
-        "enable",
-        "enabled",
-    }
-
-
-def stable_baseline_differences(mod: Any) -> List[str]:
-    diffs = []
-    for name, expected in STABLE_BASELINE.items():
-        if not hasattr(mod, name):
-            continue
-        actual = get_attr(mod, name, None)
-        if isinstance(expected, float):
-            if not approx_equal(actual, expected):
-                diffs.append(f"{name}={actual!r} (stable baseline: {expected!r})")
-        elif actual != expected:
-            diffs.append(f"{name}={actual!r} (stable baseline: {expected!r})")
-    return diffs
-
-
-def add_membership_check(
-    r: CheckResult,
-    name: str,
-    value: Any,
-    supported: Sequence[Any] | set[Any],
-    *,
-    risk_message: str | None = None,
-    note_message: str | None = None,
-) -> None:
-    if value not in supported:
-        r.add_risk(risk_message or f"Unsupported {name}: {value!r}")
-    elif note_message:
-        r.add_note(note_message)
-
-
-def check_positive_int(r: CheckResult, name: str, value: Any) -> None:
-    parsed = as_int(value)
-    if parsed is None or parsed <= 0:
-        r.add_risk(f"{name} must be a positive integer; got {value!r}")
-
-
-def check_nonnegative_number(r: CheckResult, name: str, value: Any) -> None:
-    parsed = as_float(value)
-    if parsed is None or parsed < 0.0:
-        r.add_risk(f"{name} must be a non-negative number; got {value!r}")
-
-
-def check_grid_geometry(r: CheckResult, mod: Any, grid: str) -> None:
-    grid_level = as_int(get_attr(mod, "grid_level", None))
-    static_grid_level = as_int(get_attr(mod, "static_grid_level", None))
-    moving_grid_level = as_int(get_attr(mod, "moving_grid_level", None))
-    refinement_level = as_int(get_attr(mod, "refinement_level", None))
-    analysis_level = as_int(get_attr(mod, "analysis_level", 0))
-
-    for name in (
-        "grid_level",
-        "static_grid_level",
-        "moving_grid_level",
-        "static_grid_number",
-        "moving_grid_number",
-        "quarter_sphere_number",
-    ):
-        check_positive_int(r, name, get_attr(mod, name, None))
-
-    if grid_level is not None and static_grid_level is not None:
-        if static_grid_level > grid_level:
-            r.add_risk("static_grid_level cannot exceed grid_level.")
-        if moving_grid_level is not None and moving_grid_level != grid_level - static_grid_level:
-            r.add_risk(
-                "moving_grid_level should equal grid_level - static_grid_level; "
-                f"got {moving_grid_level}, expected {grid_level - static_grid_level}."
-            )
-    if grid_level is not None:
-        if refinement_level is None or refinement_level < 0 or refinement_level > grid_level:
-            r.add_risk(f"refinement_level must be in [0, grid_level]; got {refinement_level!r}")
-        if analysis_level is None or analysis_level < 0 or analysis_level >= grid_level:
-            r.add_risk(f"analysis_level must be in [0, grid_level); got {analysis_level!r}")
-
-    largest_max = sequence_values(get_attr(mod, "largest_box_xyz_max", None))
-    largest_min = sequence_values(get_attr(mod, "largest_box_xyz_min", None))
-    if largest_max is None or len(largest_max) != 3:
-        r.add_risk("largest_box_xyz_max must contain three numeric values.")
-    elif any(v <= 0.0 for v in largest_max):
-        r.add_risk(f"largest_box_xyz_max values must be positive; got {largest_max!r}")
-    if largest_min is None or len(largest_min) != 3:
-        r.add_risk("largest_box_xyz_min must contain three numeric values.")
-    elif largest_max is not None and len(largest_max) == 3:
-        for idx, (lo, hi) in enumerate(zip(largest_min, largest_max)):
-            if lo >= hi:
-                r.add_risk(
-                    f"largest_box_xyz_min[{idx}] must be smaller than largest_box_xyz_max[{idx}]."
-                )
-
-    if grid == "Shell-Patch" and largest_max is not None and len(largest_max) == 3:
-        if max(largest_max) - min(largest_max) > 1.0e-12:
-            r.add_risk("Shell-Patch requires a cubic largest_box_xyz_max.")
-
-    if not approx_equal(get_attr(mod, "devide_factor", None), 2.0):
-        r.add_risk("devide_factor must remain 2.0; the AMR code documents only this ratio as supported.")
-    if as_text(get_attr(mod, "static_grid_type", "")) != "Linear":
-        r.add_risk("static_grid_type must remain 'Linear'.")
-    if as_text(get_attr(mod, "moving_grid_type", "")) != "Linear":
-        r.add_risk("moving_grid_type must remain 'Linear'.")
-
-    shell_shape = sequence_values(get_attr(mod, "shell_grid_number", None))
-    if grid == "Shell-Patch":
-        if shell_shape is None or len(shell_shape) != 3:
-            r.add_risk("Shell-Patch requires shell_grid_number with three numeric values.")
-        elif any(int(v) <= 0 for v in shell_shape):
-            r.add_risk(f"shell_grid_number values must be positive; got {shell_shape!r}")
-
-
-def check_punctures(r: CheckResult, mod: Any, init: str, puncture_data: str) -> None:
-    puncture_number = as_int(get_attr(mod, "puncture_number", None))
-    if puncture_number is None or puncture_number <= 0:
-        r.add_risk(f"puncture_number must be a positive integer; got {puncture_number!r}")
-        return
-
-    if init == "Ansorg-TwoPuncture" and puncture_number != 2:
-        r.add_warning(
-            "Ansorg-TwoPuncture is validated on the GPU branch mainly for puncture_number=2."
-        )
-    if puncture_data == "Automatically-BBH":
-        r.add_risk("puncture_data_set='Automatically-BBH' is documented as still developing.")
-
-    for name in ("position_BH", "parameter_BH", "dimensionless_spin_BH", "momentum_BH"):
-        value = get_attr(mod, name, None)
-        outer = sequence_len(value)
-        if outer != puncture_number:
-            r.add_risk(f"{name} must have puncture_number rows; got {outer!r}.")
-            continue
-        for idx in range(puncture_number):
-            if sequence_len(value[idx]) != 3:
-                r.add_risk(f"{name}[{idx}] must contain three values.")
-                break
-
-    if init == "Ansorg-TwoPuncture":
-        for name in ("parameter_BH", "position_BH", "momentum_BH"):
-            if get_attr(mod, name, None) is None:
-                r.add_risk(f"Ansorg-TwoPuncture requires {name}.")
-
-
-def check_output_and_time(r: CheckResult, mod: Any) -> None:
-    for name in (
-        "Final_Evolution_Time",
-        "Check_Time",
-        "Dump_Time",
-        "D2_Dump_Time",
-        "Analysis_Time",
-        "Courant_Factor",
-        "Dissipation",
-    ):
-        check_nonnegative_number(r, name, get_attr(mod, name, None))
-    check_positive_int(r, "Evolution_Step_Number", get_attr(mod, "Evolution_Step_Number", None))
-
-    start_time = as_float(get_attr(mod, "Start_Evolution_Time", None))
-    final_time = as_float(get_attr(mod, "Final_Evolution_Time", None))
-    if start_time is None:
-        r.add_risk("Start_Evolution_Time must be numeric.")
-    elif final_time is not None and final_time <= start_time:
-        r.add_risk("Final_Evolution_Time must be greater than Start_Evolution_Time.")
-
-    for name in ("GW_L_max", "GW_M_max", "Detector_Number"):
-        check_positive_int(r, name, get_attr(mod, name, None))
-    detector_min = as_float(get_attr(mod, "Detector_Rmin", None))
-    detector_max = as_float(get_attr(mod, "Detector_Rmax", None))
-    if detector_min is None or detector_min <= 0.0:
-        r.add_risk(f"Detector_Rmin must be positive; got {detector_min!r}")
-    if detector_max is None or detector_max <= 0.0:
-        r.add_risk(f"Detector_Rmax must be positive; got {detector_max!r}")
-    if detector_min is not None and detector_max is not None and detector_max <= detector_min:
-        r.add_risk("Detector_Rmax must be greater than Detector_Rmin.")
-
-
-def check_equation_specific(r: CheckResult, mod: Any, eq: str, grid: str, fd: str) -> None:
-    if eq == "BSSN":
-        r.add_note("Equation_Class=BSSN is the current validated GPU baseline.")
-    elif eq == "BSSN-EScalar":
-        r.add_warning("BSSN-EScalar has a CUDA path, but it is less broadly validated than BSSN.")
-        fr_choice = as_int(get_attr(mod, "FR_Choice", None))
-        if fr_choice not in {1, 2, 3, 4, 5}:
-            r.add_risk(f"FR_Choice must be one of 1..5 for BSSN-EScalar; got {fr_choice!r}")
-        if approx_equal(get_attr(mod, "FR_a2", None), 0.0):
-            r.add_risk("CUDA BSSN-EScalar requires nonzero FR_a2.")
-        elif not approx_equal(get_attr(mod, "FR_a2", None), 3.0):
-            r.add_warning("CUDA BSSN-EScalar now passes FR_a2 to the kernel, but non-3.0 values need CPU/GPU regression.")
-        for name in ("FR_l2", "FR_phi0", "FR_r0", "FR_sigma0"):
-            check_nonnegative_number(r, name, get_attr(mod, name, None))
-    elif eq == "BSSN-EM":
-        r.add_warning(
-            "BSSN-EM is accepted by the build, but this checker cannot certify its physics/output "
-            "without a CPU/GPU regression run."
-        )
-        if fd == "8th-order":
-            r.add_note("BSSN-EM with 8th-order enables extra CUDA AMR batching defaults.")
-    elif eq == "Z4C":
-        r.add_warning(
-            "Z4C has CUDA support, but the resident path and Shell/CPBC combinations are more constrained."
-        )
-        if grid == "Patch":
-            r.add_warning("Z4C+Patch avoids Shell CPBC, but still needs a dedicated regression test.")
-        else:
-            r.add_warning("Z4C+Shell-Patch uses CPBC/Shell logic and is not the stable BSSN baseline.")
-
-
-def check_runtime_environment(r: CheckResult, mod: Any, eq: str, grid: str, fd: str) -> None:
-    if env_truthy("AMSS_CUDA_BH_INTERP_RESIDENT"):
-        r.add_risk(
-            "AMSS_CUDA_BH_INTERP_RESIDENT is enabled in the environment; this option previously caused "
-            "late-time trajectory drift and should stay off unless explicitly revalidated."
-        )
-    else:
-        r.add_note("AMSS_CUDA_BH_INTERP_RESIDENT is not enabled; this matches the fixed stable default.")
-
-    if eq in {"BSSN", "BSSN-EScalar", "Z4C"}:
-        r.add_note("makefile_and_run.py will default AMSS_CUDA_AMR_RESTRICT_DEVICE=1 for this equation.")
-    if fd in {"2nd-order", "8th-order"}:
-        r.add_warning(
-            f"{fd} disables some interpolation/CUDA-aware MPI fast paths by default; validate performance and output."
-        )
-    if grid == "Shell-Patch":
-        r.add_warning(
-            "Shell-Patch changes runtime defaults and MPI process handling; use at least the script-adjusted 4 MPI ranks."
-        )
-
-    z4c_mrbd = as_int(get_attr(mod, "AMSS_Z4C_MRBD", 0), 0)
-    if z4c_mrbd not in {0, 1, 2}:
-        r.add_risk(f"AMSS_Z4C_MRBD must be 0, 1, or 2; got {z4c_mrbd!r}")
-    elif eq == "Z4C" and z4c_mrbd == 2:
-        r.add_risk("Z4C GPU resident path does not support AMSS_Z4C_MRBD=2.")
-    elif eq == "Z4C" and z4c_mrbd in {0, 1}:
-        r.add_note(f"Z4C will build with AMSS_Z4C_MRBD={z4c_mrbd}.")
-
-
-def check_stable_profile(r: CheckResult, mod: Any) -> None:
-    diffs = stable_baseline_differences(mod)
-    if not diffs:
-        r.add_note("This input matches the documented most stable GPU baseline.")
-        return
-    r.add_warning(
-        "This input differs from the documented most stable GPU baseline: " + "; ".join(diffs)
-    )
-
-
-def check_input(mod: Any) -> CheckResult:
-    r = CheckResult()
-
-    gpu_text = as_lower_text(get_attr(mod, "GPU_Calculation", "no"))
-    gpu = gpu_text == "yes"
-    eq = as_text(get_attr(mod, "Equation_Class", ""))
-    init = as_text(get_attr(mod, "Initial_Data_Method", ""))
-    symmetry = as_text(get_attr(mod, "Symmetry", ""))
-    time_method = as_text(get_attr(mod, "Time_Evolution_Method", ""))
-    grid = as_text(get_attr(mod, "basic_grid_set", ""))
-    center = as_text(get_attr(mod, "grid_center_set", ""))
-    fd = as_text(get_attr(mod, "Finite_Diffenence_Method", ""))
-    gauge = get_attr(mod, "gauge_choice", None)
-    tetrad = get_attr(mod, "tetrad_type", None)
-    ahf = as_text(get_attr(mod, "AHF_Find", "no")).lower()
-    boundary = as_text(get_attr(mod, "boundary_choice", ""))
-    puncture_data = as_text(get_attr(mod, "puncture_data_set", ""))
-    cpu_part = get_attr(mod, "CPU_Part", None)
-    gpu_part = get_attr(mod, "GPU_Part", None)
-
-    if gpu_text not in {"yes", "no"}:
-        r.add_risk(f"GPU_Calculation must be 'yes' or 'no'; got {get_attr(mod, 'GPU_Calculation', None)!r}")
-    if not gpu:
-        r.add_note("GPU_Calculation=no; this check only targets the GPU branch.")
-        return r
-
-    r.add_note("GPU_Calculation=yes detected.")
-
-    add_membership_check(r, "Equation_Class", eq, SUPPORTED_EQUATIONS)
-    add_membership_check(r, "Symmetry", symmetry, SUPPORTED_SYMMETRIES)
-    add_membership_check(r, "Initial_Data_Method", init, SUPPORTED_INITIAL_DATA)
-    add_membership_check(r, "basic_grid_set", grid, SUPPORTED_GRIDS)
-    add_membership_check(r, "grid_center_set", center, SUPPORTED_CENTERS)
-    add_membership_check(r, "Finite_Diffenence_Method", fd, SUPPORTED_FD)
-    add_membership_check(r, "gauge_choice", gauge, SUPPORTED_GAUGES)
-    add_membership_check(r, "tetrad_type", tetrad, SUPPORTED_TETRADS)
-    add_membership_check(r, "AHF_Find", ahf, SUPPORTED_AHF)
-    add_membership_check(r, "boundary_choice", boundary, SUPPORTED_BOUNDARIES)
-    add_membership_check(r, "puncture_data_set", puncture_data, SUPPORTED_PUNCTURE_DATA)
-
-    if init != "Ansorg-TwoPuncture":
-        r.add_risk(
-            f"Initial_Data_Method={init!r} is not validated as safe on this GPU branch; "
-            "the stable path is Ansorg-TwoPuncture."
-        )
-    else:
-        r.add_note("Initial_Data_Method=Ansorg-TwoPuncture is supported.")
-
-    if time_method != "runge-kutta-45":
-        r.add_risk(f"Only Time_Evolution_Method='runge-kutta-45' is supported; got {time_method!r}.")
-    if grid == "Patch":
-        r.add_note("basic_grid_set=Patch is the current stable GPU grid path.")
-    elif grid == "Shell-Patch":
-        r.add_warning("basic_grid_set=Shell-Patch has GPU support but is outside the stable BSSN baseline.")
-    if center == "Vertex":
-        r.add_warning("grid_center_set=Vertex is compiled by macros, but the stable GPU baseline is Cell.")
-    if symmetry != "equatorial-symmetry":
-        r.add_warning("The stable validation case uses equatorial-symmetry; other symmetries need regression tests.")
-    if fd != "4th-order":
-        r.add_warning("The stable validation case uses 4th-order finite differences.")
-    if gauge not in {0, 1}:
-        r.add_warning("Input comments recommend gauge_choice 0 or 1; other gauges need dedicated validation.")
-    if tetrad != 2:
-        r.add_warning("Input comments recommend tetrad_type=2; other tetrads affect wave extraction conventions.")
-
-    if ahf == "yes":
-        r.add_warning("AHF_Find=yes is supported by macros, but it is outside the current stable GPU baseline.")
-
-    if boundary == "Shibata-choice":
-        r.add_risk("Shibata-choice is not faithfully distinguished in the current macro generator; it maps to the BAM branch.")
-    elif boundary == "BAM-choice":
-        r.add_note("boundary_choice=BAM-choice is supported.")
-
-    if cpu_part is not None or gpu_part is not None:
-        r.add_warning("CPU_Part/GPU_Part are printed and propagated, but they do not control a real mixed CPU/GPU split in this branch.")
-
-    check_output_and_time(r, mod)
-    check_grid_geometry(r, mod, grid)
-    check_punctures(r, mod, init, puncture_data)
-    check_equation_specific(r, mod, eq, grid, fd)
-    check_runtime_environment(r, mod, eq, grid, fd)
-    check_stable_profile(r, mod)
-
-    return r
-
-
-def main() -> int:
-    parser = argparse.ArgumentParser()
-    parser.add_argument(
-        "-f",
-        "--file",
-        "--input",
-        dest="input_file",
-        default="AMSS_NCKU_Input.py",
-        help="path to AMSS_NCKU_Input.py",
-    )
-    args = parser.parse_args()
-
-    path = Path(args.input_file).resolve()
-    if not path.exists():
-        print(f"ERROR: input file not found: {path}")
-        return 2
-
-    try:
-        mod = load_input_module(path)
-    except Exception as exc:
-        print(f"ERROR: failed to load input file: {exc}")
-        return 2
-
-    result = check_input(mod)
-
-    print(f"Input: {path}")
-    print(f"GPU_Calculation: {get_attr(mod, 'GPU_Calculation', 'no')}")
-    print(f"Symmetry: {get_attr(mod, 'Symmetry', '')}")
-    print(f"Equation_Class: {get_attr(mod, 'Equation_Class', '')}")
-    print(f"Initial_Data_Method: {get_attr(mod, 'Initial_Data_Method', '')}")
-    print(f"puncture_data_set: {get_attr(mod, 'puncture_data_set', '')}")
-    print(f"basic_grid_set: {get_attr(mod, 'basic_grid_set', '')}")
-    print(f"grid_center_set: {get_attr(mod, 'grid_center_set', '')}")
-    print(f"Finite_Diffenence_Method: {get_attr(mod, 'Finite_Diffenence_Method', '')}")
-    print(f"gauge_choice: {get_attr(mod, 'gauge_choice', '')}")
-    print(f"tetrad_type: {get_attr(mod, 'tetrad_type', '')}")
-    print(f"boundary_choice: {get_attr(mod, 'boundary_choice', '')}")
-    print(f"AHF_Find: {get_attr(mod, 'AHF_Find', '')}")
-    print(f"AMSS_Z4C_MRBD: {get_attr(mod, 'AMSS_Z4C_MRBD', 0)}")
-    print("")
-
-    for msg in result.notes:
-        print(f"NOTE: {msg}")
-    for msg in result.warnings:
-        print(f"WARNING: {msg}")
-    for msg in result.risks:
-        print(f"RISK: {msg}")
-
-    print("")
-    if result.risks:
-        print("Verdict: review the risks above before running.")
-        return 1
-
-    if result.warnings:
-        print("Verdict: runnable on the current GPU branch, but keep the warnings in mind.")
-        return 0
-
-    print("Verdict: OK to run on the current GPU branch.")
-    return 0
-
-
-if __name__ == "__main__":
-    raise SystemExit(main())

AMSS_NCKU_Input.py

@@ -13,31 +13,15 @@ import numpy
 
 ## Setting MPI processes and the output file directory
 
-File_directory   = "case3"                    ## output file directory
+File_directory   = "GW150914"                    ## output file directory
 Output_directory = "binary_output"               ## binary data file directory
                                                  ## The file directory name should not be too long
-MPI_processes    = 2                             ## number of mpi processes used in the simulation
-
-GPU_Calculation  = "yes"                          ## Use GPU or not 
-                                                 ## (prefer "no" in the current version, because the GPU part may have bugs when integrated in this Python interface)
-CPU_Part         = 1.0
-GPU_Part         = 0.0
-
-## Aggressive runtime overrides for fastest low-accuracy GPU runs.
-AMSS_EVOLVE_TIMING                 = 0
-AMSS_ANALYSIS_MAP_EVERY            = 1000000000
-AMSS_INTERP_FAST                   = 1
-AMSS_INTERP_GPU                    = 1
-AMSS_CUDA_AWARE_MPI                = 1
-AMSS_CUDA_RESIDENT_SYNC            = 1
-AMSS_CUDA_BSSN_RESIDENT_SYNC       = 1
-AMSS_CUDA_KEEP_RESIDENT_AFTER_STEP = 1
-AMSS_CUDA_KEEP_ALL_LEVELS          = 1
-AMSS_CUDA_AMR_RESTRICT_DEVICE      = 1
-AMSS_CUDA_AMR_RESTRICT_BATCH       = 1
-AMSS_CUDA_DEVICE_SEGMENT_BATCH     = 1
-AMSS_CUDA_UNCACHED_DEVICE_BUFFERS  = 1
-AMSS_CUDA_AMR_HOST_STAGED          = 1
+MPI_processes    = 64                             ## number of mpi processes used in the simulation
+
+GPU_Calculation  = "no"                          ## Use GPU or not 
+                                                 ## (prefer "no" in the current version, because the GPU part may have bugs when integrated in this Python interface)
+CPU_Part         = 1.0
+GPU_Part         = 0.0
 
 #################################################
 
@@ -61,14 +45,14 @@ Finite_Diffenence_Method = "4th-order"             ## finite-difference method:
 ## Setting the time evolutionary information
 
 Start_Evolution_Time     = 0.0                    ## start evolution time t0
-Final_Evolution_Time     = 200.0                 ## final evolution time t1
-Check_Time               = 1000000000.0
-Dump_Time                = 1000000000.0           ## time inteval dT for dumping binary data
-D2_Dump_Time             = 1000000000.0           ## dump the ascii data for 2d surface after dT'
-Analysis_Time            = 1000000000.0           ## dump the puncture position and GW psi4 after dT"
-Evolution_Step_Number    = 1000000                ## stop the calculation after the maximal step number
-Courant_Factor           = 0.8                    ## Courant Factor
-Dissipation              = 0.15                   ## Kreiss-Oliger Dissipation Strength
+Final_Evolution_Time     = 1000.0                 ## final evolution time t1
+Check_Time               = 100.0
+Dump_Time                = 100.0                  ## time inteval dT for dumping binary data
+D2_Dump_Time             = 100.0                  ## dump the ascii data for 2d surface after dT'
+Analysis_Time            = 0.1                    ## dump the puncture position and GW psi4 after dT"
+Evolution_Step_Number    = 10000000               ## stop the calculation after the maximal step number
+Courant_Factor           = 0.5                    ## Courant Factor
+Dissipation              = 0.15                   ## Kreiss-Oliger Dissipation Strength
 
 #################################################
 
@@ -80,22 +64,22 @@ Dissipation              = 0.15                   ## Kreiss-Oliger Dissipation S
 basic_grid_set    = "Patch"                          ## grid structure: choose "Patch" or "Shell-Patch"
 grid_center_set   = "Cell"                           ## grid center: chose "Cell" or "Vertex"
 
-grid_level        = 7                                ## total number of AMR grid levels
-static_grid_level = 4                                ## number of AMR static grid levels
-moving_grid_level = grid_level - static_grid_level   ## number of AMR moving grid levels
-
-analysis_level    = 0
-refinement_level  = 2                                ## time refinement start from this grid level
+grid_level        = 9                                ## total number of AMR grid levels
+static_grid_level = 5                                ## number of AMR static grid levels
+moving_grid_level = grid_level - static_grid_level   ## number of AMR moving grid levels
+
+analysis_level    = 0
+refinement_level  = 3                                ## time refinement start from this grid level
 
 largest_box_xyz_max = [320.0, 320.0, 320.0]          ## scale of the largest box
                                                      ## not ne cess ary to be cubic for "Patch" grid s tructure
                                                      ## need to be a cubic box for "Shell-Patch" grid structure
 largest_box_xyz_min = - numpy.array(largest_box_xyz_max)  
 
-static_grid_number = 64                               ## grid points of each static AMR grid (in x direction)
-                                                     ## (grid points in y and z directions are automatically adjusted)
-moving_grid_number = 32                               ## grid points of each moving AMR grid
-shell_grid_number  = [32, 32, 100]                   ## grid points of Shell-Patch grid
+static_grid_number = 96                              ## grid points of each static AMR grid (in x direction)
+                                                     ## (grid points in y and z directions are automatically adjusted)
+moving_grid_number = 48                              ## grid points of each moving AMR grid
+shell_grid_number  = [32, 32, 100]                   ## grid points of Shell-Patch grid
                                                      ## in (phi, theta, r) direction
 devide_factor      = 2.0                             ## resolution between different grid levels dh0/dh1, only support 2.0 now
                                                      
@@ -103,7 +87,7 @@ devide_factor      = 2.0                             ## resolution between diffe
 static_grid_type   = 'Linear'                        ## AMR static grid structure , only supports "Linear"
 moving_grid_type   = 'Linear'                        ## AMR moving grid structure , only supports "Linear"
 
-quarter_sphere_number = 16                           ## grid number of 1/4 s pher ical surface
+quarter_sphere_number = 96                           ## grid number of 1/4 s pher ical surface
                                                      ## (which is needed for evaluating the spherical surface integral)
 
 #################################################
@@ -126,15 +110,15 @@ puncture_data_set     = "Manually"                       ## Method to give Punct
 
 ## initial orbital distance and ellipticity for BBHs system
 ## ( needed for "Automatically-BBH" case , not affect the "Manually" case )
-Distance = 12.0
+Distance = 10.0
 e0       = 0.0
 
 ## black hole parameter (M Q* a*)
-parameter_BH[0] = [ 0.5,  0.0,  0.0 ]   
-parameter_BH[1] = [ 0.5,  0.0,  0.0 ]  
+parameter_BH[0] = [ 36.0/(36.0+29.0),  0.0,  +0.31 ]   
+parameter_BH[1] = [ 29.0/(36.0+29.0),  0.0,  -0.46 ]  
 ## dimensionless spin in each direction
-dimensionless_spin_BH[0] = [ 0.0,  0.0,  0.0 ]   
-dimensionless_spin_BH[1] = [ 0.0,  0.0,  0.0 ]  
+dimensionless_spin_BH[0] = [ 0.0,  0.0,  +0.31 ]   
+dimensionless_spin_BH[1] = [ 0.0,  0.0,  -0.46 ]  
 
 ## use Brugmann's convention
 ##  -----0-----> y
@@ -145,13 +129,13 @@ dimensionless_spin_BH[1] = [ 0.0,  0.0,  0.0 ]
 ## If puncture_data_set is chosen to be "Manually", it is necessary to set the position and momentum of each puncture manually
 
 ## initial position for each puncture
-position_BH[0]  = [  0.0,  6.0, 0.0 ]  
-position_BH[1]  = [  0.0, -6.0, 0.0 ] 
+position_BH[0]  = [  0.0,  10.0*29.0/(36.0+29.0), 0.0 ]  
+position_BH[1]  = [  0.0, -10.0*36.0/(36.0+29.0), 0.0 ] 
 
 ## initial mumentum for each puncture
 ## (needed for "Manually" case, does not affect the "Automatically-BBH" case)
-momentum_BH[0]  = [ -0.06,  -0.01,   0.0 ]
-momentum_BH[1]  = [ +0.06,  +0.01,   0.0 ]
+momentum_BH[0]  = [ -0.09530152296974252,  -0.00084541526517121,   0.0 ]
+momentum_BH[1]  = [ +0.09530152296974252,  +0.00084541526517121,   0.0 ]
 
 
 #################################################
@@ -161,11 +145,11 @@ momentum_BH[1]  = [ +0.06,  +0.01,   0.0 ]
 
 ## Setting the gravitational wave information
 
-GW_L_max        = 2                      ## maximal L number in gravitational wave
-GW_M_max        = 2                      ## maximal M number in gravitational wave
-Detector_Number = 2                     ## number of dector
+GW_L_max        = 4                      ## maximal L number in gravitational wave
+GW_M_max        = 4                      ## maximal M number in gravitational wave
+Detector_Number = 12                     ## number of dector
 Detector_Rmin   = 50.0                   ## nearest dector distance
-Detector_Rmax   = 100.0                  ## farest dector distance
+Detector_Rmax   = 160.0                  ## farest dector distance
 
 #################################################
 
@@ -176,8 +160,8 @@ Detector_Rmax   = 100.0                  ## farest dector distance
 
 AHF_Find       = "no"                    ## whether to find the apparent horizon: choose "yes" or "no"
 
-AHF_Find_Every = 1000000000
-AHF_Dump_Time  = 1000000000.0
+AHF_Find_Every = 24
+AHF_Dump_Time  = 20.0
 
 #################################################
 

AMSS_NCKU_Program.py

@@ -58,36 +58,31 @@ File_directory = os.path.join(input_data.File_directory)
 
 ## If the specified output directory exists, ask the user whether to continue
 if os.path.exists(File_directory):
-    auto_overwrite = str(getattr(input_data, "Auto_Overwrite_Output", "yes")).strip().lower()
-    if auto_overwrite in ("1", "yes", "y", "true", "on", "continue"):
-        print( " Output dictionary has been existed; Auto_Overwrite_Output=yes, continue the calculation. " )
-        print(                                                                                              )
-    else:
-        print( " Output dictionary has been existed !!!  "                                                              )
-        print( " If you want to overwrite the existing file directory, please input 'continue' in the terminal !! "     ) 
-        print( " If you want to retain the existing file directory, please input 'stop' in the terminal to stop the "   ) 
-        print( " simulation. Then you can reset the output dictionary in the input script file AMSS_NCKU_Input.py !!! " )
-        print(                                                                                                          )
-        ## Prompt whether to overwrite the existing directory
-        while True:
-            try:
-                inputvalue = input()
-                ## If the user agrees to overwrite, proceed and remove the existing directory
-                if ( inputvalue == "continue" ):
-                    print( " Continue the calculation !!! " )
-                    print(                                  )
-                    break  
-                ## If the user chooses not to overwrite, exit and keep the existing directory
-                elif ( inputvalue == "stop" ):
-                    print( " Stop the calculation !!! "    )
-                    sys.exit() 
-                ## If the user input is invalid, prompt again
-                else:
-                    print( " Please input your choice !!! "                   )
-                    print( " Input 'continue' or 'stop' in the terminal !!! " )
-            except ValueError:
+    print( " Output dictionary has been existed !!!  "                                                              )
+    print( " If you want to overwrite the existing file directory, please input 'continue' in the terminal !! "     ) 
+    print( " If you want to retain the existing file directory, please input 'stop' in the terminal to stop the "   ) 
+    print( " simulation. Then you can reset the output dictionary in the input script file AMSS_NCKU_Input.py !!! " )
+    print(                                                                                                          )
+    ## Prompt whether to overwrite the existing directory
+    while True:
+        try:
+            inputvalue = input()
+            ## If the user agrees to overwrite, proceed and remove the existing directory
+            if ( inputvalue == "continue" ):
+                print( " Continue the calculation !!! " )
+                print(                                  )
+                break  
+            ## If the user chooses not to overwrite, exit and keep the existing directory
+            elif ( inputvalue == "stop" ):
+                print( " Stop the calculation !!! "    )
+                sys.exit() 
+            ## If the user input is invalid, prompt again
+            else:
                 print( " Please input your choice !!! "                   )
                 print( " Input 'continue' or 'stop' in the terminal !!! " )
+        except ValueError:
+            print( " Please input your choice !!! "                   )
+            print( " Input 'continue' or 'stop' in the terminal !!! " )
         
 ## Remove the existing output directory if present
 shutil.rmtree(File_directory, ignore_errors=True)
@@ -179,11 +174,14 @@ import generate_macrodef
 generate_macrodef.generate_macrodef_h()
 print( " AMSS-NCKU macro file macrodef.h has been generated. " )
      
-generate_macrodef.generate_macrodef_fh()
-print( " AMSS-NCKU macro file macrodef.fh has been generated. " )
-
-
-##################################################################
+generate_macrodef.generate_macrodef_fh()
+print( " AMSS-NCKU macro file macrodef.fh has been generated. " )
+
+generate_macrodef.generate_build_config()
+print( " AMSS-NCKU build config AMSS_NCKU_build.mk has been generated. " )
+
+
+##################################################################
 
 # Compile the AMSS-NCKU program according to user requirements
 
@@ -222,11 +220,13 @@ shutil.copytree(AMSS_NCKU_source_path, AMSS_NCKU_source_copy)
 
 # Copy the generated macro files into the AMSS_NCKU source folder
 
-macrodef_h_path  = os.path.join(File_directory, "macrodef.h") 
-macrodef_fh_path = os.path.join(File_directory, "macrodef.fh") 
-
-shutil.copy2(macrodef_h_path,  AMSS_NCKU_source_copy)
-shutil.copy2(macrodef_fh_path, AMSS_NCKU_source_copy)
+macrodef_h_path  = os.path.join(File_directory, "macrodef.h")
+macrodef_fh_path = os.path.join(File_directory, "macrodef.fh")
+build_config_path = os.path.join(File_directory, "AMSS_NCKU_build.mk")
+
+shutil.copy2(macrodef_h_path,  AMSS_NCKU_source_copy)
+shutil.copy2(macrodef_fh_path, AMSS_NCKU_source_copy)
+shutil.copy2(build_config_path, AMSS_NCKU_source_copy)
 
 # Notes on copying files:
 # shutil.copy2 preserves file metadata such as modification time.
@@ -263,7 +263,7 @@ print()
 if (input_data.GPU_Calculation == "no"):
     ABE_file = os.path.join(AMSS_NCKU_source_copy, "ABE")
 elif (input_data.GPU_Calculation == "yes"):
-    ABE_file = os.path.join(AMSS_NCKU_source_copy, "ABE_CUDA")
+    ABE_file = os.path.join(AMSS_NCKU_source_copy, "ABEGPU")
 
 if not os.path.exists( ABE_file ):
     print(                                                                                                  )

AMSS_NCKU_Program_Plot.py

@@ -1,100 +0,0 @@
-##################################################################
-##
-## AMSS-NCKU Plot-Only Restart Script
-## Author: Xiaoqu / Claude
-## 2026/05/12
-##
-## This script checks for existing output data from AMSS_NCKU_Program.py.
-## If data exists, it skips all computation and goes directly to plotting,
-## saving time when plotting was interrupted.
-## If no data is found, it exits with a message.
-##
-##################################################################
-
-## Guard against re-execution by multiprocessing child processes.
-if __name__ != '__main__':
-    import sys as _sys
-    _sys.exit(0)
-
-
-import os
-import sys
-
-import AMSS_NCKU_Input as input_data
-
-##################################################################
-
-## Construct paths from input configuration
-File_directory         = os.path.join(input_data.File_directory)
-output_directory       = os.path.join(File_directory, "AMSS_NCKU_output")
-binary_results_directory = os.path.join(output_directory, input_data.Output_directory)
-figure_directory       = os.path.join(File_directory, "figure")
-
-##################################################################
-
-## Check whether the required output data files exist
-
-required_files = [
-    os.path.join(binary_results_directory, "bssn_BH.dat"),
-    os.path.join(binary_results_directory, "bssn_ADMQs.dat"),
-    os.path.join(binary_results_directory, "bssn_psi4.dat"),
-    os.path.join(binary_results_directory, "bssn_constraint.dat"),
-]
-
-missing_files = [f for f in required_files if not os.path.exists(f)]
-
-if missing_files:
-    print(" No existing AMSS_NCKU_Program.py output data found. ")
-    print(" The following required files are missing: ")
-    for f in missing_files:
-        print(f"   {f}")
-    print()
-    print(" Please run AMSS_NCKU_Program.py first to generate the simulation data. ")
-    print(" Exiting. ")
-    sys.exit(1)
-
-print(" Found existing AMSS_NCKU_Program.py output data. "          )
-print(" Skipping all computation and going directly to plotting. " )
-print()
-
-## Ensure the figure directory exists (it should, but be safe)
-os.makedirs(figure_directory, exist_ok=True)
-
-##################################################################
-
-## Plot the AMSS-NCKU program results
-
-import plot_xiaoqu
-import plot_GW_strain_amplitude_xiaoqu
-from parallel_plot_helper import run_plot_tasks_parallel
-
-plot_tasks = []
-
-## Plot black hole trajectory
-plot_tasks.append((plot_xiaoqu.generate_puncture_orbit_plot,   (binary_results_directory, figure_directory)))
-plot_tasks.append((plot_xiaoqu.generate_puncture_orbit_plot3D, (binary_results_directory, figure_directory)))
-
-## Plot black hole separation vs. time
-plot_tasks.append((plot_xiaoqu.generate_puncture_distence_plot, (binary_results_directory, figure_directory)))
-
-## Plot gravitational waveforms (psi4 and strain amplitude)
-for i in range(input_data.Detector_Number):
-    plot_tasks.append((plot_xiaoqu.generate_gravitational_wave_psi4_plot, (binary_results_directory, figure_directory, i)))
-    plot_tasks.append((plot_GW_strain_amplitude_xiaoqu.generate_gravitational_wave_amplitude_plot, (binary_results_directory, figure_directory, i)))
-
-## Plot ADM mass evolution
-for i in range(input_data.Detector_Number):
-    plot_tasks.append((plot_xiaoqu.generate_ADMmass_plot, (binary_results_directory, figure_directory, i)))
-
-## Plot Hamiltonian constraint violation over time
-for i in range(input_data.grid_level):
-    plot_tasks.append((plot_xiaoqu.generate_constraint_check_plot, (binary_results_directory, figure_directory, i)))
-
-run_plot_tasks_parallel(plot_tasks)
-
-## Plot stored binary data (runs serially, not in the parallel pool)
-plot_xiaoqu.generate_binary_data_plot(binary_results_directory, figure_directory)
-
-print()
-print(" Plotting completed successfully. ")
-print()

AMSS_NCKU_source/ABE.C

@@ -198,16 +198,16 @@ int main(int argc, char *argv[])
       if (myrank == 0)
       {
             string out_dir;
-            string filename;
-            map<string, string>::iterator iter;
-            iter = parameters::str_par.find("output dir");
-            if (iter != parameters::str_par.end())
-            {
-                  out_dir = iter->second;
-            }
-            filename = out_dir + "/setting.par";
-            ofstream setfile;
-            setfile.open(filename.c_str(), ios::trunc);
+            char filename[50];
+            map<string, string>::iterator iter;
+            iter = parameters::str_par.find("output dir");
+            if (iter != parameters::str_par.end())
+            {
+                  out_dir = iter->second;
+            }
+            sprintf(filename, "%s/setting.par", out_dir.c_str());
+            ofstream setfile;
+            setfile.open(filename, ios::trunc);
 
             if (!setfile.good())
             {
@@ -484,11 +484,7 @@ int main(int argc, char *argv[])
             cout << endl;
       }
 
-      // Let the process teardown reclaim the simulation object.  Some derived
-      // equation classes keep MPI/CUDA-backed state whose destructor ordering
-      // is fragile at program shutdown.
-      if (getenv("AMSS_DELETE_ADM_ON_EXIT"))
-            delete ADM;
+      delete ADM;
 
       //=======================caculation done=============================================================
 

AMSS_NCKU_source/Block.C

@@ -6,68 +6,14 @@
 #include <cstdio>
 #include <string>
 #include <cmath>
-#include <new>
-using namespace std;
-
-#include "Block.h"
-#include "misc.h"
-
-#if USE_CUDA_BSSN || USE_CUDA_Z4C
-#include <cuda_runtime_api.h>
-#endif
-
-namespace {
-
-bool cuda_pin_gridfuncs_enabled()
-{
-  static int enabled = -1;
-  if (enabled < 0)
-  {
-    const char *env = getenv("AMSS_CUDA_PIN_GRIDFUNCS");
-    enabled = (env && atoi(env) != 0) ? 1 : 0;
-  }
-  return enabled != 0;
-}
-
-double *alloc_gridfunc(size_t count, unsigned char &pinned)
-{
-  pinned = 0;
-#if USE_CUDA_BSSN || USE_CUDA_Z4C
-  if (cuda_pin_gridfuncs_enabled())
-  {
-    double *ptr = 0;
-    cudaError_t err = cudaMallocHost((void **)&ptr, count * sizeof(double));
-    if (err == cudaSuccess)
-    {
-      pinned = 1;
-      return ptr;
-    }
-    cudaGetLastError();
-  }
-#endif
-  return (double *)malloc(sizeof(double) * count);
-}
-
-void free_gridfunc(double *ptr, unsigned char pinned)
-{
-  if (!ptr)
-    return;
-#if USE_CUDA_BSSN || USE_CUDA_Z4C
-  if (pinned)
-  {
-    cudaFreeHost(ptr);
-    return;
-  }
-#else
-  (void)pinned;
-#endif
-  free(ptr);
-}
-
-}
-
-Block::Block(int DIM, int *shapei, double *bboxi, int ranki, int ingfsi, int fngfsi, int levi, const int cgpui) : rank(ranki), lev(levi), cgpu(cgpui), ingfs(ingfsi), fngfs(fngfsi), igfs(0), fgfs(0), fgfs_pinned(0)
-{
+#include <new>
+using namespace std;
+
+#include "Block.h"
+#include "misc.h"
+
+Block::Block(int DIM, int *shapei, double *bboxi, int ranki, int ingfsi, int fngfsi, int levi, const int cgpui) : rank(ranki), ingfs(ingfsi), fngfs(fngfsi), lev(levi), cgpu(cgpui)
+{
   for (int i = 0; i < dim; i++)
     X[i] = 0;
 
@@ -122,15 +68,14 @@ Block::Block(int DIM, int *shapei, double *bboxi, int ranki, int ingfsi, int fng
 #endif
     }
 
-    int nn = shape[0] * shape[1] * shape[2];
-    fgfs = new double *[fngfs];
-    fgfs_pinned = new unsigned char[fngfs];
-    for (int i = 0; i < fngfs; i++)
-    {
-      fgfs[i] = alloc_gridfunc((size_t)nn, fgfs_pinned[i]);
-      if (!(fgfs[i]))
-      {
-        cout << "on node#" << rank << ", out of memory when constructing Block." << endl;
+    int nn = shape[0] * shape[1] * shape[2];
+    fgfs = new double *[fngfs];
+    for (int i = 0; i < fngfs; i++)
+    {
+      fgfs[i] = (double *)malloc(sizeof(double) * nn);
+      if (!(fgfs[i]))
+      {
+        cout << "on node#" << rank << ", out of memory when constructing Block." << endl;
         MPI_Abort(MPI_COMM_WORLD, 1);
       }
       memset(fgfs[i], 0, sizeof(double) * nn);
@@ -158,19 +103,17 @@ Block::~Block()
   {
     for (int i = 0; i < dim; i++)
       delete[] X[i];
-    for (int i = 0; i < ingfs; i++)
-      free(igfs[i]);
-    delete[] igfs;
-    for (int i = 0; i < fngfs; i++)
-      free_gridfunc(fgfs[i], fgfs_pinned ? fgfs_pinned[i] : 0);
-    delete[] fgfs;
-    delete[] fgfs_pinned;
-    X[0] = X[1] = X[2] = 0;
-    igfs = 0;
-    fgfs = 0;
-    fgfs_pinned = 0;
-  }
-}
+    for (int i = 0; i < ingfs; i++)
+      free(igfs[i]);
+    delete[] igfs;
+    for (int i = 0; i < fngfs; i++)
+      free(fgfs[i]);
+    delete[] fgfs;
+    X[0] = X[1] = X[2] = 0;
+    igfs = 0;
+    fgfs = 0;
+  }
+}
 void Block::checkBlock()
 {
   int myrank;
@@ -241,14 +184,12 @@ void Block::swapList(MyList<var> *VarList1, MyList<var> *VarList2, int myrank)
   if (rank == myrank)
   {
     MyList<var> *varl1 = VarList1, *varl2 = VarList2;
-    while (varl1 && varl2)
-    {
-      misc::swap<double *>(fgfs[varl1->data->sgfn], fgfs[varl2->data->sgfn]);
-      if (fgfs_pinned)
-        misc::swap<unsigned char>(fgfs_pinned[varl1->data->sgfn], fgfs_pinned[varl2->data->sgfn]);
-      varl1 = varl1->next;
-      varl2 = varl2->next;
-    }
+    while (varl1 && varl2)
+    {
+      misc::swap<double *>(fgfs[varl1->data->sgfn], fgfs[varl2->data->sgfn]);
+      varl1 = varl1->next;
+      varl2 = varl2->next;
+    }
     if (varl1 || varl2)
     {
       cout << "error in Block::swaplist, var lists does not match." << endl;

AMSS_NCKU_source/Block.h

@@ -13,15 +13,14 @@ public:
    int shape[dim];
    double bbox[2 * dim];
    double *X[dim];
-   int rank; // where the real data locate in
-   int lev, cgpu;
-   int ingfs, fngfs;
-   int *(*igfs);
-   double *(*fgfs);
-   unsigned char *fgfs_pinned;
+   int rank; // where the real data locate in
+   int lev, cgpu;
+   int ingfs, fngfs;
+   int *(*igfs);
+   double *(*fgfs);
 
 public:
-   Block() : rank(0), lev(0), cgpu(0), ingfs(0), fngfs(0), igfs(0), fgfs(0), fgfs_pinned(0) {};
+   Block() {};
    Block(int DIM, int *shapei, double *bboxi, int ranki, int ingfsi, int fngfs, int levi, const int cgpui = 0);
 
    ~Block();

AMSS_NCKU_source/MPatch.C

@@ -11,15 +11,12 @@
 using namespace std;
 
 #include "misc.h"
-#include "MPatch.h"
-#include "Parallel.h"
-#include "fmisc.h"
-#if USE_CUDA_BSSN
-#include "bssn_rhs_cuda.h"
-#endif
-#ifdef INTERP_LB_PROFILE
-#include "interp_lb_profile.h"
-#endif
+#include "MPatch.h"
+#include "Parallel.h"
+#include "fmisc.h"
+#ifdef INTERP_LB_PROFILE
+#include "interp_lb_profile.h"
+#endif
 
 namespace
 {
@@ -157,8 +154,8 @@ void build_block_bin_index(Patch *patch, const double *DH, BlockBinIndex &index)
   index.valid = true;
 }
 
-int find_block_index_for_point(const BlockBinIndex &index, const double *pox, const double *DH)
-{
+int find_block_index_for_point(const BlockBinIndex &index, const double *pox, const double *DH)
+{
   if (!index.valid)
     return -1;
 
@@ -178,448 +175,10 @@ int find_block_index_for_point(const BlockBinIndex &index, const double *pox, co
   for (size_t bi = 0; bi < index.views.size(); bi++)
     if (point_in_block_view(index.views[bi], pox, DH))
       return int(bi);
-
-  return -1;
-}
-
-inline int fortran_idint_local(double x)
-{
-  return int(x);
-}
-
-bool interp_fast_enabled()
-{
-  static int enabled = -1;
-  if (enabled < 0)
-  {
-    const char *env = getenv("AMSS_INTERP_FAST");
-    enabled = (!env || atoi(env) != 0) ? 1 : 0;
-  }
-  return enabled != 0;
-}
-
-bool interp_gpu_enabled()
-{
-  static int enabled = -1;
-  if (enabled < 0)
-  {
-    const char *env = getenv("AMSS_INTERP_GPU");
-    enabled = (env && atoi(env) != 0) ? 1 : 0;
-  }
-  return enabled != 0;
-}
-
-bool interp_fast_compare_enabled()
-{
-  static int enabled = -1;
-  if (enabled < 0)
-  {
-    const char *env = getenv("AMSS_INTERP_FAST_COMPARE");
-    enabled = (env && atoi(env) != 0) ? 1 : 0;
-  }
-  return enabled != 0;
-}
-
-double interp_fast_compare_tol()
-{
-  static double tol = -1.0;
-  if (tol < 0.0)
-  {
-    const char *env = getenv("AMSS_INTERP_FAST_COMPARE_TOL");
-    tol = (env && atof(env) > 0.0) ? atof(env) : 1.0e-11;
-  }
-  return tol;
-}
-
-long long interp_fast_compare_limit()
-{
-  static long long limit = -1;
-  if (limit < 0)
-  {
-    const char *env = getenv("AMSS_INTERP_FAST_COMPARE_LIMIT");
-    limit = (env && atoll(env) > 0) ? atoll(env) : 4096;
-  }
-  return limit;
-}
-
-struct FastInterpStencil
-{
-  int cxB[dim];
-  double cx[dim];
-  double wx[8];
-  double wy[8];
-  double wz[8];
-  int nsamples;
-  int loc[512];
-  unsigned char sign_mask[512];
-  double weight[512];
-};
-
-inline void lagrange_unit_weights(double x, int ordn, double *w)
-{
-  for (int i = 0; i < ordn; i++)
-  {
-    double num = 1.0;
-    double den = 1.0;
-    for (int j = 0; j < ordn; j++)
-    {
-      if (j == i)
-        continue;
-      num *= (x - double(j));
-      den *= double(i - j);
-    }
-    w[i] = num / den;
-  }
-}
-
-inline void z_unit_weights(double x, int ordn, double *w)
-{
-  if (ordn == 6)
-  {
-    static const double c_uniform[6] = {-1.0, 5.0, -10.0, 10.0, -5.0, 1.0};
-    for (int i = 0; i < 6; i++)
-    {
-      if (x == double(i))
-      {
-        for (int j = 0; j < 6; j++)
-          w[j] = (j == i) ? 1.0 : 0.0;
-        return;
-      }
-    }
-    double den = 0.0;
-    for (int i = 0; i < 6; i++)
-    {
-      w[i] = c_uniform[i] / (x - double(i));
-      den += w[i];
-    }
-    for (int i = 0; i < 6; i++)
-      w[i] /= den;
-    return;
-  }
-  lagrange_unit_weights(x, ordn, w);
-}
-
-inline bool fast_interp_map_index(int idx, int extent, int d,
-                                  int &mapped, unsigned char &mask)
-{
-  if (idx > 0)
-    mapped = idx;
-  else
-  {
-    mask |= (unsigned char)(1u << d);
-#ifdef Vertex
-#ifdef Cell
-#error Both Cell and Vertex are defined
-#endif
-    mapped = 2 - idx;
-#else
-#ifdef Cell
-    mapped = 1 - idx;
-#else
-#error Not define Vertex nor Cell
-#endif
-#endif
-  }
-  return mapped >= 1 && mapped <= extent;
-}
-
-bool prepare_fast_interp_stencil(Block *BP, const double *pox, int ordn,
-                                 int Symmetry, FastInterpStencil &st)
-{
-  if (!BP || ordn <= 0 || ordn > 8)
-    return false;
-
-  st.nsamples = 0;
-
-  const int NO_SYMM = 0;
-  const int OCTANT = 2;
-  int cmin[dim], cmax[dim], cxT[dim];
-  for (int d = 0; d < dim; d++)
-  {
-    const double *X = BP->X[d];
-    const double dX = X[1] - X[0];
-    const int cxI = fortran_idint_local((pox[d] - X[0]) / dX + 0.4) + 1;
-    st.cxB[d] = cxI - ordn / 2 + 1;
-    cxT[d] = st.cxB[d] + ordn - 1;
-    cmin[d] = 1;
-    cmax[d] = BP->shape[d];
-
-#ifdef Vertex
-#ifdef Cell
-#error Both Cell and Vertex are defined
-#endif
-    if (Symmetry == OCTANT && d < 2 && fabs(X[0]) < dX)
-      cmin[d] = -ordn / 2 + 2;
-    if (Symmetry != NO_SYMM && d == 2 && fabs(X[0]) < dX)
-      cmin[d] = -ordn / 2 + 2;
-#else
-#ifdef Cell
-    if (Symmetry == OCTANT && d < 2 && fabs(X[0]) < dX)
-      cmin[d] = -ordn / 2 + 1;
-    if (Symmetry != NO_SYMM && d == 2 && fabs(X[0]) < dX)
-      cmin[d] = -ordn / 2 + 1;
-#else
-#error Not define Vertex nor Cell
-#endif
-#endif
-
-    if (st.cxB[d] < cmin[d])
-    {
-      st.cxB[d] = cmin[d];
-      cxT[d] = st.cxB[d] + ordn - 1;
-    }
-    if (cxT[d] > cmax[d])
-    {
-      cxT[d] = cmax[d];
-      st.cxB[d] = cxT[d] + 1 - ordn;
-    }
-
-    if (st.cxB[d] > 0)
-      st.cx[d] = (pox[d] - X[st.cxB[d] - 1]) / dX;
-    else
-    {
-#ifdef Vertex
-#ifdef Cell
-#error Both Cell and Vertex are defined
-#endif
-      st.cx[d] = (pox[d] + X[1 - st.cxB[d]]) / dX;
-#else
-#ifdef Cell
-      st.cx[d] = (pox[d] + X[-st.cxB[d]]) / dX;
-#else
-#error Not define Vertex nor Cell
-#endif
-#endif
-    }
-  }
-
-  lagrange_unit_weights(st.cx[0], ordn, st.wx);
-  lagrange_unit_weights(st.cx[1], ordn, st.wy);
-  z_unit_weights(st.cx[2], ordn, st.wz);
-
-  for (int kk = 0; kk < ordn; kk++)
-  {
-    for (int jj = 0; jj < ordn; jj++)
-    {
-      for (int ii = 0; ii < ordn; ii++)
-      {
-        unsigned char mask = 0;
-        int ix, iy, iz;
-        if (!fast_interp_map_index(st.cxB[0] + ii, BP->shape[0], 0, ix, mask) ||
-            !fast_interp_map_index(st.cxB[1] + jj, BP->shape[1], 1, iy, mask) ||
-            !fast_interp_map_index(st.cxB[2] + kk, BP->shape[2], 2, iz, mask))
-          return false;
-        const int s = st.nsamples++;
-        st.loc[s] = (ix - 1) + (iy - 1) * BP->shape[0] +
-                    (iz - 1) * BP->shape[0] * BP->shape[1];
-        st.sign_mask[s] = mask;
-        st.weight[s] = st.wx[ii] * st.wy[jj] * st.wz[kk];
-      }
-    }
-  }
-  return true;
-}
-
-bool interpolate_var_list_with_stencil(Block *BP, MyList<var> *VarList,
-                                       int num_var, const double *pox,
-                                       int ordn, int Symmetry,
-                                       const FastInterpStencil &st,
-                                       double *out)
-{
-  if (num_var <= 0 || num_var > 128)
-    return false;
-
-  double *data_ptrs[128];
-  double *soa_ptrs[128];
-  var *vars[128];
-  MyList<var> *varl = VarList;
-  int k = 0;
-  while (varl)
-  {
-    if (k >= num_var)
-      return false;
-    vars[k] = varl->data;
-    data_ptrs[k] = BP->fgfs[vars[k]->sgfn];
-    soa_ptrs[k] = vars[k]->SoA;
-    out[k] = 0.0;
-    varl = varl->next;
-    k++;
-  }
-
-  if (k != num_var)
-    return false;
-
-  for (int s = 0; s < st.nsamples; s++)
-  {
-    const int loc = st.loc[s];
-    const double w = st.weight[s];
-    const unsigned char mask = st.sign_mask[s];
-    if (mask == 0)
-    {
-      for (int v = 0; v < num_var; v++)
-        out[v] += w * data_ptrs[v][loc];
-    }
-    else
-    {
-      for (int v = 0; v < num_var; v++)
-      {
-        const double *SoA = soa_ptrs[v];
-        double sgn = 1.0;
-        if (mask & 1u)
-          sgn *= SoA[0];
-        if (mask & 2u)
-          sgn *= SoA[1];
-        if (mask & 4u)
-          sgn *= SoA[2];
-        out[v] += w * sgn * data_ptrs[v][loc];
-      }
-    }
-  }
-
-  if (interp_fast_compare_enabled())
-  {
-    static int report_count = 0;
-    static long long compare_calls = 0;
-    if (compare_calls++ >= interp_fast_compare_limit())
-      return true;
-    const double tol = interp_fast_compare_tol();
-    varl = VarList;
-    k = 0;
-    while (varl)
-    {
-      var *vp = vars[k];
-      double ref = 0.0;
-      double x = pox[0], y = pox[1], z = pox[2];
-      f_global_interp(BP->shape, BP->X[0], BP->X[1], BP->X[2],
-                      BP->fgfs[vp->sgfn], ref,
-                      x, y, z, ordn, vp->SoA, Symmetry);
-      const double diff = fabs(ref - out[k]);
-      const double scale = 1.0 + fabs(ref);
-      if (diff > tol * scale && report_count < 32)
-      {
-        int rank = 0;
-        MPI_Comm_rank(MPI_COMM_WORLD, &rank);
-        fprintf(stderr,
-                "[AMSS-INTERP-CMP][rank %d] var=%s diff=%.17e ref=%.17e fast=%.17e p=(%.17e,%.17e,%.17e)\n",
-                rank, vp->name, diff, ref, out[k], pox[0], pox[1], pox[2]);
-        report_count++;
-      }
-      varl = varl->next;
-      k++;
-    }
-  }
-
-  return true;
-}
-
-bool interpolate_var_list_fast(Block *BP, MyList<var> *VarList, int num_var,
-                               const double *pox, int ordn, int Symmetry,
-                               double *out)
-{
-  if (!interp_fast_enabled())
-    return false;
-
-  FastInterpStencil st;
-  if (!prepare_fast_interp_stencil(BP, pox, ordn, Symmetry, st))
-    return false;
-
-  return interpolate_var_list_with_stencil(BP, VarList, num_var, pox,
-                                           ordn, Symmetry, st, out);
-}
-
-struct CachedInterpPoint
-{
-  Block *bp;
-  int owner_rank;
-  FastInterpStencil stencil;
-};
-
-struct SurfaceInterpCache
-{
-  Patch *patch;
-  int NN;
-  int symmetry;
-  double key[9];
-  vector<CachedInterpPoint> points;
-
-  SurfaceInterpCache() : patch(0), NN(0), symmetry(-1) {}
-};
-
-bool surface_cache_key_matches(const SurfaceInterpCache &cache, Patch *patch,
-                               int NN, double **XX, int Symmetry)
-{
-  if (cache.patch != patch || cache.NN != NN || cache.symmetry != Symmetry ||
-      int(cache.points.size()) != NN || NN <= 0)
-    return false;
-  const int mid = NN / 2;
-  const int last = NN - 1;
-  const int ids[3] = {0, mid, last};
-  int p = 0;
-  for (int q = 0; q < 3; q++)
-    for (int d = 0; d < dim; d++)
-      if (cache.key[p++] != XX[d][ids[q]])
-        return false;
-  return true;
-}
-
-SurfaceInterpCache *find_surface_cache(Patch *patch, int NN, double **XX,
-                                       int Symmetry)
-{
-  static vector<SurfaceInterpCache> caches;
-  for (size_t i = 0; i < caches.size(); i++)
-    if (surface_cache_key_matches(caches[i], patch, NN, XX, Symmetry))
-      return &caches[i];
-  if (caches.size() >= 24)
-    caches.erase(caches.begin());
-  caches.push_back(SurfaceInterpCache());
-  return &caches.back();
-}
-
-bool build_surface_cache(SurfaceInterpCache &cache, Patch *patch, int NN,
-                         double **XX, int Symmetry, const double *DH,
-                         const BlockBinIndex &block_index, int ordn)
-{
-  int myrank = 0;
-  MPI_Comm_rank(MPI_COMM_WORLD, &myrank);
-  cache.patch = patch;
-  cache.NN = NN;
-  cache.symmetry = Symmetry;
-  cache.points.clear();
-  cache.points.resize(NN);
-  const int mid = NN / 2;
-  const int last = NN - 1;
-  const int ids[3] = {0, mid, last};
-  int p = 0;
-  for (int q = 0; q < 3; q++)
-    for (int d = 0; d < dim; d++)
-      cache.key[p++] = XX[d][ids[q]];
-
-  for (int j = 0; j < NN; j++)
-  {
-    double pox[dim];
-    for (int d = 0; d < dim; d++)
-      pox[d] = XX[d][j];
-    const int block_i = find_block_index_for_point(block_index, pox, DH);
-    if (block_i < 0)
-    {
-      cache.points[j].bp = 0;
-      cache.points[j].owner_rank = -1;
-      continue;
-    }
-    Block *BP = block_index.views[block_i].bp;
-    cache.points[j].bp = BP;
-    cache.points[j].owner_rank = BP->rank;
-    cache.points[j].stencil.nsamples = 0;
-    if (BP->rank == myrank)
-    {
-      if (!prepare_fast_interp_stencil(BP, pox, ordn, Symmetry,
-                                       cache.points[j].stencil))
-        return false;
-    }
-  }
-  return true;
-}
-} // namespace
+
+  return -1;
+}
+} // namespace
 
 Patch::Patch(int DIM, int *shapei, double *bboxi, int levi, bool buflog, int Symmetry) : lev(levi)
 {
@@ -1002,26 +561,22 @@ void Patch::Interp_Points(MyList<var> *VarList,
     if (block_i >= 0)
     {
       Block *BP = block_index.views[block_i].bp;
-      owner_rank[j] = BP->rank;
-      if (myrank == BP->rank)
-      {
-        //---> interpolation
-        if (!interpolate_var_list_fast(BP, VarList, num_var, pox, ordn,
-                                       Symmetry, Shellf + j * num_var))
-        {
-          varl = VarList;
-          int k = 0;
-          while (varl) // run along variables
-          {
-            f_global_interp(BP->shape, BP->X[0], BP->X[1], BP->X[2], BP->fgfs[varl->data->sgfn], Shellf[j * num_var + k],
-                            pox[0], pox[1], pox[2], ordn, varl->data->SoA, Symmetry);
-            varl = varl->next;
-            k++;
-          }
-        }
-      }
-    }
-  }
+      owner_rank[j] = BP->rank;
+      if (myrank == BP->rank)
+      {
+        //---> interpolation
+        varl = VarList;
+        int k = 0;
+        while (varl) // run along variables
+        {
+          f_global_interp(BP->shape, BP->X[0], BP->X[1], BP->X[2], BP->fgfs[varl->data->sgfn], Shellf[j * num_var + k],
+                          pox[0], pox[1], pox[2], ordn, varl->data->SoA, Symmetry);
+          varl = varl->next;
+          k++;
+        }
+      }
+    }
+  }
 
   // Replace MPI_Allreduce with per-owner MPI_Bcast:
   // Group consecutive points by owner rank and broadcast each group.
@@ -1104,8 +659,10 @@ void Patch::Interp_Points(MyList<var> *VarList,
     varl = varl->next;
   }
 
-  // owner_rank[j] records which MPI rank owns point j
-  int *owner_rank;
+  memset(Shellf, 0, sizeof(double) * NN * num_var);
+
+  // owner_rank[j] records which MPI rank owns point j
+  int *owner_rank;
   owner_rank = new int[NN];
   for (int j = 0; j < NN; j++)
     owner_rank[j] = -1;
@@ -1113,117 +670,12 @@ void Patch::Interp_Points(MyList<var> *VarList,
   double DH[dim];
   for (int i = 0; i < dim; i++)
     DH[i] = getdX(i);
-  BlockBinIndex block_index;
-  build_block_bin_index(this, DH, block_index);
-  SurfaceInterpCache *surface_cache = 0;
-  bool use_surface_cache = false;
-  if (interp_fast_enabled())
-  {
-    surface_cache = find_surface_cache(this, NN, XX, Symmetry);
-    use_surface_cache = surface_cache_key_matches(*surface_cache, this, NN, XX, Symmetry);
-    if (!use_surface_cache)
-      use_surface_cache = build_surface_cache(*surface_cache, this, NN, XX,
-                                              Symmetry, DH, block_index, ordn);
-  }
-
-  // --- Interpolation phase (identical to original) ---
-#if USE_CUDA_BSSN
-  const bool use_gpu_interp = interp_gpu_enabled() && use_surface_cache && num_var == 2 &&
-                              VarList && VarList->next && !VarList->next->next;
-#else
-  const bool use_gpu_interp = false;
-#endif
-  if (use_gpu_interp)
-  {
-#if USE_CUDA_BSSN
-    vector<vector<int> > local_points(block_index.views.size());
-    for (int j = 0; j < NN; j++)
-    {
-      for (int i = 0; i < dim; i++)
-      {
-        if (myrank == 0 && (XX[i][j] < bbox[i] + lli[i] * DH[i] || XX[i][j] > bbox[dim + i] - uui[i] * DH[i]))
-        {
-          cout << "Patch::Interp_Points: point (";
-          for (int k = 0; k < dim; k++)
-          {
-            cout << XX[k][j];
-            if (k < dim - 1)
-              cout << ",";
-            else
-              cout << ") is out of current Patch." << endl;
-          }
-          MPI_Abort(MPI_COMM_WORLD, 1);
-        }
-      }
-
-      CachedInterpPoint &cp = surface_cache->points[j];
-      Block *BP = cp.bp;
-      owner_rank[j] = cp.owner_rank;
-      if (BP && myrank == BP->rank)
-      {
-        for (size_t bi = 0; bi < block_index.views.size(); bi++)
-        {
-          if (block_index.views[bi].bp == BP)
-          {
-            local_points[bi].push_back(j);
-            break;
-          }
-        }
-      }
-    }
-
-    var *v0 = VarList->data;
-    var *v1 = VarList->next->data;
-    double soa6[6] = {
-        v0->SoA[0], v0->SoA[1], v0->SoA[2],
-        v1->SoA[0], v1->SoA[1], v1->SoA[2]};
-
-    for (size_t bi = 0; bi < local_points.size(); bi++)
-    {
-      const int count = int(local_points[bi].size());
-      if (count <= 0)
-        continue;
-
-      Block *BP = block_index.views[bi].bp;
-      vector<double> px(count), py(count), pz(count), out(2 * count);
-      for (int q = 0; q < count; q++)
-      {
-        const int j = local_points[bi][q];
-        px[q] = XX[0][j];
-        py[q] = XX[1][j];
-        pz[q] = XX[2][j];
-      }
-
-      const double dx = BP->X[0][1] - BP->X[0][0];
-      const double dy = BP->X[1][1] - BP->X[1][0];
-      const double dz = BP->X[2][1] - BP->X[2][0];
-      const int ok = bssn_cuda_interp_host_two_fields(
-          BP, BP->shape,
-          BP->fgfs[v0->sgfn], BP->fgfs[v1->sgfn],
-          BP->X[0][0], BP->X[1][0], BP->X[2][0],
-          dx, dy, dz,
-          &px[0], &py[0], &pz[0], count,
-          ordn, Symmetry, soa6, &out[0]);
-      if (ok != 0)
-      {
-        if (myrank == 0)
-          cout << "Patch::Interp_Points: CUDA two-field interpolation failed" << endl;
-        MPI_Abort(MPI_COMM_WORLD, 1);
-      }
-
-      for (int q = 0; q < count; q++)
-      {
-        const int j = local_points[bi][q];
-        Shellf[j * num_var] = out[2 * q];
-        Shellf[j * num_var + 1] = out[2 * q + 1];
-      }
-    }
-#endif
-  }
-  else
-  {
-  for (int j = 0; j < NN; j++)
-  {
+  BlockBinIndex block_index;
+  build_block_bin_index(this, DH, block_index);
+
+  // --- Interpolation phase (identical to original) ---
+  for (int j = 0; j < NN; j++)
+  {
     double pox[dim];
     for (int i = 0; i < dim; i++)
     {
@@ -1240,59 +692,28 @@ void Patch::Interp_Points(MyList<var> *VarList,
             cout << ") is out of current Patch." << endl;
         }
         MPI_Abort(MPI_COMM_WORLD, 1);
-      }
-    }
-
-    if (use_surface_cache)
-    {
-      CachedInterpPoint &cp = surface_cache->points[j];
-      Block *BP = cp.bp;
-      owner_rank[j] = cp.owner_rank;
-      if (BP && myrank == BP->rank)
-      {
-        if (!interpolate_var_list_with_stencil(BP, VarList, num_var, pox,
-                                               ordn, Symmetry, cp.stencil,
-                                               Shellf + j * num_var))
-        {
-          MyList<var> *varl_fallback = VarList;
-          int k = 0;
-          while (varl_fallback)
-          {
-            f_global_interp(BP->shape, BP->X[0], BP->X[1], BP->X[2], BP->fgfs[varl_fallback->data->sgfn], Shellf[j * num_var + k],
-                            pox[0], pox[1], pox[2], ordn, varl_fallback->data->SoA, Symmetry);
-            varl_fallback = varl_fallback->next;
-            k++;
-          }
-        }
-      }
-    }
-    else
-    {
-      const int block_i = find_block_index_for_point(block_index, pox, DH);
-      if (block_i >= 0)
-      {
-        Block *BP = block_index.views[block_i].bp;
-        owner_rank[j] = BP->rank;
-      if (myrank == BP->rank)
-      {
-        if (!interpolate_var_list_fast(BP, VarList, num_var, pox, ordn,
-                                       Symmetry, Shellf + j * num_var))
-        {
-          MyList<var> *varl_fallback = VarList;
-          int k = 0;
-          while (varl_fallback)
-          {
-            f_global_interp(BP->shape, BP->X[0], BP->X[1], BP->X[2], BP->fgfs[varl_fallback->data->sgfn], Shellf[j * num_var + k],
-                            pox[0], pox[1], pox[2], ordn, varl_fallback->data->SoA, Symmetry);
-            varl_fallback = varl_fallback->next;
-            k++;
-          }
-        }
-      }
-      }
-    }
-  }
-  }
+      }
+    }
+
+    const int block_i = find_block_index_for_point(block_index, pox, DH);
+    if (block_i >= 0)
+    {
+      Block *BP = block_index.views[block_i].bp;
+      owner_rank[j] = BP->rank;
+      if (myrank == BP->rank)
+      {
+        varl = VarList;
+        int k = 0;
+        while (varl)
+        {
+          f_global_interp(BP->shape, BP->X[0], BP->X[1], BP->X[2], BP->fgfs[varl->data->sgfn], Shellf[j * num_var + k],
+                          pox[0], pox[1], pox[2], ordn, varl->data->SoA, Symmetry);
+          varl = varl->next;
+          k++;
+        }
+      }
+    }
+  }
 
 #ifdef INTERP_LB_PROFILE
   double t_interp_end = MPI_Wtime();
@@ -1544,26 +965,22 @@ void Patch::Interp_Points(MyList<var> *VarList,
     if (block_i >= 0)
     {
       Block *BP = block_index.views[block_i].bp;
-      owner_rank[j] = BP->rank;
-      if (myrank == BP->rank)
-      {
-        //---> interpolation
-        if (!interpolate_var_list_fast(BP, VarList, num_var, pox, ordn,
-                                       Symmetry, Shellf + j * num_var))
-        {
-          varl = VarList;
-          int k = 0;
-          while (varl) // run along variables
-          {
-            f_global_interp(BP->shape, BP->X[0], BP->X[1], BP->X[2], BP->fgfs[varl->data->sgfn], Shellf[j * num_var + k],
-                            pox[0], pox[1], pox[2], ordn, varl->data->SoA, Symmetry);
-            varl = varl->next;
-            k++;
-          }
-        }
-      }
-    }
-  }
+      owner_rank[j] = BP->rank;
+      if (myrank == BP->rank)
+      {
+        //---> interpolation
+        varl = VarList;
+        int k = 0;
+        while (varl) // run along variables
+        {
+          f_global_interp(BP->shape, BP->X[0], BP->X[1], BP->X[2], BP->fgfs[varl->data->sgfn], Shellf[j * num_var + k],
+                          pox[0], pox[1], pox[2], ordn, varl->data->SoA, Symmetry);
+          varl = varl->next;
+          k++;
+        }
+      }
+    }
+  }
 
   // Collect unique global owner ranks and translate to local ranks in Comm_here
   // Then broadcast each owner's points via MPI_Bcast on Comm_here

AMSS_NCKU_source/Parallel.h

@@ -104,14 +104,6 @@ namespace Parallel
     double **recv_bufs;
     int *send_buf_caps;
     int *recv_buf_caps;
-    unsigned char *send_buf_pinned;
-    unsigned char *recv_buf_pinned;
-    unsigned char *send_buf_is_dev;
-    unsigned char *recv_buf_is_dev;
-    int *send_buf_caps_dev;
-    int *recv_buf_caps_dev;
-    double **send_bufs_dev;
-    double **recv_bufs_dev;
     MPI_Request *reqs;
     MPI_Status *stats;
     int max_reqs;
@@ -119,14 +111,12 @@ namespace Parallel
     int *tc_req_node;
     int *tc_req_is_recv;
     int *tc_completed;
-    bool cuda_aware_mode;
     SyncCache();
     void invalidate();
     void destroy();
   };
 
   void Sync_cached(MyList<Patch> *PatL, MyList<var> *VarList, int Symmetry, SyncCache &cache);
-  void Sync_ensure_cache(MyList<Patch> *PatL, int Symmetry, SyncCache &cache);
   void transfer_cached(MyList<gridseg> **src, MyList<gridseg> **dst,
                        MyList<var> *VarList1, MyList<var> *VarList2,
                        int Symmetry, SyncCache &cache);
@@ -189,13 +179,13 @@ namespace Parallel
   MyList<Parallel::gridseg> *clone_gsl(MyList<Parallel::gridseg> *p, bool first_only);
   MyList<Parallel::gridseg> *build_bulk_gsl(Patch *Pat); // similar to build_owned_gsl0 but does not care rank issue
   MyList<Parallel::gridseg> *build_bulk_gsl(Block *bp, Patch *Pat);
-  void build_PhysBD_gstl(Patch *Pat, MyList<Parallel::gridseg> *srci, MyList<Parallel::gridseg> *dsti,
-                         MyList<Parallel::gridseg> **out_src, MyList<Parallel::gridseg> **out_dst);
-  void PeriodicBD(Patch *Pat, MyList<var> *VarList, int Symmetry);
-  double L2Norm(Patch *Pat, var *vf);
-  void L2Norm7(Patch *Pat, var **vf, double *norms);
-  void checkgsl(MyList<Parallel::gridseg> *pp, bool first_only);
-  void checkvarl(MyList<var> *pp, bool first_only);
+  void build_PhysBD_gstl(Patch *Pat, MyList<Parallel::gridseg> *srci, MyList<Parallel::gridseg> *dsti,
+                         MyList<Parallel::gridseg> **out_src, MyList<Parallel::gridseg> **out_dst);
+  void PeriodicBD(Patch *Pat, MyList<var> *VarList, int Symmetry);
+  double L2Norm(Patch *Pat, var *vf);
+  void L2Norm7(Patch *Pat, var **vf, double *norms);
+  void checkgsl(MyList<Parallel::gridseg> *pp, bool first_only);
+  void checkvarl(MyList<var> *pp, bool first_only);
   MyList<Parallel::gridseg> *divide_gsl(MyList<Parallel::gridseg> *p, Patch *Pat);
   MyList<Parallel::gridseg> *divide_gs(MyList<Parallel::gridseg> *p, Patch *Pat);
   void prepare_inter_time_level(Patch *Pat,
@@ -227,12 +217,12 @@ namespace Parallel
   void aligncheck(double *bbox0, double *bboxl, int lev, double *DH0, int *shape);
   bool point_locat_gsl(double *pox, MyList<Parallel::gridseg> *gsl);
   void checkpatchlist(MyList<Patch> *PatL, bool buflog);
-
-  double L2Norm(Patch *Pat, var *vf, MPI_Comm Comm_here);
-  void L2Norm7(Patch *Pat, var **vf, double *norms, MPI_Comm Comm_here);
-  bool PatList_Interp_Points(MyList<Patch> *PatL, MyList<var> *VarList,
-                             int NN, double **XX,
-                             double *Shellf, int Symmetry, MPI_Comm Comm_here);
+
+  double L2Norm(Patch *Pat, var *vf, MPI_Comm Comm_here);
+  void L2Norm7(Patch *Pat, var **vf, double *norms, MPI_Comm Comm_here);
+  bool PatList_Interp_Points(MyList<Patch> *PatL, MyList<var> *VarList,
+                             int NN, double **XX,
+                             double *Shellf, int Symmetry, MPI_Comm Comm_here);
 #if (PSTR == 1 || PSTR == 2 || PSTR == 3)
   MyList<Block> *distribute(MyList<Patch> *PatchLIST, int cpusize, int ingfsi, int fngfsi,
                             bool periodic, int start_rank, int end_rank, int nodes = 0);

AMSS_NCKU_source/ShellPatch.h

@@ -102,16 +102,6 @@ public:
                  //-1: means no dumy dimension at all; 0: means rho; 1: means sigma
    };
 
-   // Thread-safe search result (no pointers to shared mutable state)
-   struct PointSearchResult
-   {
-      bool found;
-      Block *Bg;
-      double gx, gy, gz; // global Cartesian coordinates
-      double lx, ly, lz; // local coordinates within the found block
-      int ssst;          // source shell-patch type (-1 = Cartesian)
-   };
-
    int myrank;
    int shape[dim];   // for (rho, sigma, R), for rho and sigma means number of points for every pi/2
    double Rrange[2]; // for Rmin and Rmax
@@ -185,12 +175,6 @@ public:
                          MyList<Patch> *Pp, double CDH[dim], MyList<pointstru> *pss);
    bool prolongpointstru(MyList<pointstru> *&psul, bool ssyn, int tsst, MyList<ss_patch> *sPp, double DH[dim],
                          MyList<Patch> *Pp, double CDH[dim], double x, double y, double z, int Symmetry, int rank_in);
-   // Read-only point search — thread-safe (no shared mutable state modified)
-   PointSearchResult prolongpointstru_search(bool ssyn, int tsst, MyList<ss_patch> *sPp, double DH[dim],
-                                             MyList<Patch> *Pp, double CDH[dim], double x, double y, double z,
-                                             int Symmetry, int rank_in);
-   // Append a search result to a linked list — use inside omp critical section
-   void prolongpointstru_append(MyList<pointstru> *&psul, const PointSearchResult &sr, int tsst);
    void setupintintstuff(int cpusize, MyList<Patch> *CPatL, int Symmetry);
    void intertransfer(MyList<pointstru> **src, MyList<pointstru> **dst,
                       MyList<var> *VarList1 /* source */, MyList<var> *VarList2 /*target */,
@@ -211,11 +195,11 @@ public:
    bool Interp_One_Point(MyList<var> *VarList,
                          double *XX, /*input global Cartesian coordinate*/
                          double *Shellf, int Symmetry);
-   void write_Pablo_file_ss(int *ext, double xmin, double xmax, double ymin, double ymax, double zmin, double zmax,
-                            char *filename, int sst);
-   double L2Norm(var *vf);
-   void L2Norm7(var **vf, double *norms);
-   void Find_Maximum(MyList<var> *VarList, double *XX, double *Shellf);
-};
+   void write_Pablo_file_ss(int *ext, double xmin, double xmax, double ymin, double ymax, double zmin, double zmax,
+                            char *filename, int sst);
+   double L2Norm(var *vf);
+   void L2Norm7(var **vf, double *norms);
+   void Find_Maximum(MyList<var> *VarList, double *XX, double *Shellf);
+};
 
 #endif /* SHELLPATCH_H */

AMSS_NCKU_source/Z4c_class.C

@@ -3,7 +3,6 @@
 #include <sstream>
 #include <cstdio>
 #include <map>
-#include <string>
 using namespace std;
 #else
 #include <stdio.h>
@@ -29,20 +28,6 @@ using namespace std;
 #include "kodiss.h"
 #include "parameters.h"
 
-#ifndef USE_CUDA_Z4C
-#define USE_CUDA_Z4C 0
-#endif
-
-#if USE_CUDA_Z4C && (ABEtype == 2)
-#include "z4c_rhs_cuda.h"
-#endif
-#if USE_CUDA_BSSN
-#include "bssn_rhs_cuda.h"
-#ifdef WithShell
-#include "bssn_gpu.h"
-#endif
-#endif
-
 #ifdef With_AHF
 #include "derivatives.h"
 #include "myglobal.h"
@@ -52,81 +37,6 @@ using namespace std;
 
 // Define Z4c_class
 
-#if USE_CUDA_Z4C && (ABEtype == 2) && defined(WithShell)
-// GPU-accelerated Z4C shell RHS: same parameter signature as f_compute_rhs_Z4c_ss.
-// Internally calls gpu_rhs_z4c_ss which modifies trK→trKd before upload,
-// runs BSSN algebraic kernels, then applies Z4C post-processing (TZ_rhs, damping).
-extern "C" {
-static int cuda_compute_rhs_z4c_ss(
-    int *ex, double &T, double *crho, double *sigma, double *R,
-    double *X, double *Y, double *Z,
-    double *drhodx, double *drhody, double *drhodz,
-    double *dsigmadx, double *dsigmady, double *dsigmadz,
-    double *dRdx, double *dRdy, double *dRdz,
-    double *drhodxx, double *drhodxy, double *drhodxz, double *drhodyy, double *drhodyz, double *drhodzz,
-    double *dsigmadxx, double *dsigmadxy, double *dsigmadxz, double *dsigmadyy, double *dsigmadyz, double *dsigmadzz,
-    double *dRdxx, double *dRdxy, double *dRdxz, double *dRdyy, double *dRdyz, double *dRdzz,
-    double *chi, double *trK,
-    double *gxx, double *gxy, double *gxz, double *gyy, double *gyz, double *gzz,
-    double *Axx, double *Axy, double *Axz, double *Ayy, double *Ayz, double *Azz,
-    double *Gamx, double *Gamy, double *Gamz,
-    double *Lap, double *betax, double *betay, double *betaz,
-    double *dtSfx, double *dtSfy, double *dtSfz,
-    double *TZ,
-    double *chi_rhs, double *trK_rhs,
-    double *gxx_rhs, double *gxy_rhs, double *gxz_rhs, double *gyy_rhs, double *gyz_rhs, double *gzz_rhs,
-    double *Axx_rhs, double *Axy_rhs, double *Axz_rhs, double *Ayy_rhs, double *Ayz_rhs, double *Azz_rhs,
-    double *Gamx_rhs, double *Gamy_rhs, double *Gamz_rhs,
-    double *Lap_rhs, double *betax_rhs, double *betay_rhs, double *betaz_rhs,
-    double *dtSfx_rhs, double *dtSfy_rhs, double *dtSfz_rhs,
-    double *TZ_rhs,
-    double *rho_mat, double *Sx, double *Sy, double *Sz,
-    double *Sxx, double *Sxy, double *Sxz, double *Syy, double *Syz, double *Szz,
-    double *Gamxxx, double *Gamxxy, double *Gamxxz, double *Gamxyy, double *Gamxyz, double *Gamxzz,
-    double *Gamyxx, double *Gamyxy, double *Gamyxz, double *Gamyyy, double *Gamyyz, double *Gamyzz,
-    double *Gamzxx, double *Gamzxy, double *Gamzxz, double *Gamzyy, double *Gamzyz, double *Gamzzz,
-    double *Rxx, double *Rxy, double *Rxz, double *Ryy, double *Ryz, double *Rzz,
-    double *ham_Res, double *movx_Res, double *movy_Res, double *movz_Res,
-    double *Gmx_Res, double *Gmy_Res, double *Gmz_Res,
-    int &Symmetry, int &Lev, double &eps, int &sst, int &co)
-{
-    return gpu_rhs_z4c_ss(0, 0,  // calledby=ABE_main, mpi_rank=device_0
-                     ex, T, crho, sigma, R, X, Y, Z,
-                     drhodx, drhody, drhodz,
-                     dsigmadx, dsigmady, dsigmadz,
-                     dRdx, dRdy, dRdz,
-                     drhodxx, drhodxy, drhodxz, drhodyy, drhodyz, drhodzz,
-                     dsigmadxx, dsigmadxy, dsigmadxz, dsigmadyy, dsigmadyz, dsigmadzz,
-                     dRdxx, dRdxy, dRdxz, dRdyy, dRdyz, dRdzz,
-                     chi, trK,
-                     gxx, gxy, gxz, gyy, gyz, gzz,
-                     Axx, Axy, Axz, Ayy, Ayz, Azz,
-                     Gamx, Gamy, Gamz,
-                     Lap, betax, betay, betaz,
-                     dtSfx, dtSfy, dtSfz,
-                     TZ,
-                     chi_rhs, trK_rhs,
-                     gxx_rhs, gxy_rhs, gxz_rhs, gyy_rhs, gyz_rhs, gzz_rhs,
-                     Axx_rhs, Axy_rhs, Axz_rhs, Ayy_rhs, Ayz_rhs, Azz_rhs,
-                     Gamx_rhs, Gamy_rhs, Gamz_rhs,
-                     Lap_rhs, betax_rhs, betay_rhs, betaz_rhs,
-                     dtSfx_rhs, dtSfy_rhs, dtSfz_rhs,
-                     TZ_rhs,
-                     rho_mat, Sx, Sy, Sz,
-                     Sxx, Sxy, Sxz, Syy, Syz, Szz,
-                     Gamxxx, Gamxxy, Gamxxz, Gamxyy, Gamxyz, Gamxzz,
-                     Gamyxx, Gamyxy, Gamyxz, Gamyyy, Gamyyz, Gamyzz,
-                     Gamzxx, Gamzxy, Gamzxz, Gamzyy, Gamzyz, Gamzzz,
-                     Rxx, Rxy, Rxz, Ryy, Ryz, Rzz,
-                     ham_Res, movx_Res, movy_Res, movz_Res,
-                     Gmx_Res, Gmy_Res, Gmz_Res,
-                     Symmetry, Lev, eps, sst, co);
-}
-}
-// Redirect all Z4C shell RHS calls in Step/SHStep to GPU
-#define f_compute_rhs_Z4c_ss cuda_compute_rhs_z4c_ss
-#endif
-
 // This class inherits some members and methods from the parent `bssn_class` and modifies others.
 // The modified members and methods are defined below (and in the header Z4c_class.h).
 // The remaining members/methods are inherited from `bssn_class` (declared in bssn_class.h).
@@ -222,13 +132,6 @@ void Z4c_class::Initialize()
     PhysTime = StartTime;
     Setup_Black_Hole_position();
   }
-
-  sync_cache_pre = new Parallel::SyncCache[GH->levels];
-  sync_cache_cor = new Parallel::SyncCache[GH->levels];
-  sync_cache_rp_coarse = new Parallel::SyncCache[GH->levels];
-  sync_cache_rp_fine = new Parallel::SyncCache[GH->levels];
-  sync_cache_restrict = new Parallel::SyncCache[GH->levels];
-  sync_cache_outbd = new Parallel::SyncCache[GH->levels];
 }
 
 //================================================================================================
@@ -262,581 +165,13 @@ Z4c_class::~Z4c_class()
 
 //================================================================================================
 
-#ifndef AMSS_Z4C_MRBD
-#define AMSS_Z4C_MRBD 0
-#endif
-#define MRBD AMSS_Z4C_MRBD // 0: fix BD for meshrefinement level; 1: sommerfeld_bam for them; 2: sommerfeld_yo for them
+#define MRBD 0 // 0: fix BD for meshrefinement level; 1: sommerfeld_bam for them; 2: sommerfeld_yo for them
 
 #ifndef CPBC
 // for sommerfeld boundary
 
-#if USE_CUDA_Z4C && (ABEtype == 2)
-#if (MRBD == 2)
-#error "USE_CUDA_Z4C resident path does not support MRBD == 2"
-#endif
-
-namespace {
-
-static const int k_z4c_cuda_bh_state_indices[3] = {18, 19, 20};
-
-bool fill_z4c_cuda_views(Block *cg, MyList<var> *vars,
-                         double **host_views,
-                         double *propspeeds = 0,
-                         double *soa_flat = 0)
-{
-  int idx = 0;
-  while (vars && idx < Z4C_CUDA_STATE_COUNT)
-  {
-    host_views[idx] = cg->fgfs[vars->data->sgfn];
-    if (propspeeds)
-      propspeeds[idx] = vars->data->propspeed;
-    if (soa_flat)
-    {
-      soa_flat[3 * idx + 0] = vars->data->SoA[0];
-      soa_flat[3 * idx + 1] = vars->data->SoA[1];
-      soa_flat[3 * idx + 2] = vars->data->SoA[2];
-    }
-    vars = vars->next;
-    ++idx;
-  }
-  return idx == Z4C_CUDA_STATE_COUNT && vars == 0;
-}
-
-void z4c_cuda_download_level_state(MyList<Patch> *PatL, MyList<var> *vars, int myrank, bool release_ctx)
-{
-  MyList<Patch> *Pp = PatL;
-  while (Pp)
-  {
-    MyList<Block> *BP = Pp->data->blb;
-    while (BP)
-    {
-      Block *cg = BP->data;
-      if (myrank == cg->rank && z4c_cuda_has_resident_state(cg))
-      {
-        double *state_out[Z4C_CUDA_STATE_COUNT];
-        if (!fill_z4c_cuda_views(cg, vars, state_out))
-        {
-          cout << "CUDA Z4C state list mismatch on resident state download" << endl;
-          MPI_Abort(MPI_COMM_WORLD, 1);
-        }
-        if (z4c_cuda_download_resident_state(cg, cg->shape, state_out))
-        {
-          cout << "CUDA Z4C resident state download failed" << endl;
-          MPI_Abort(MPI_COMM_WORLD, 1);
-        }
-        if (release_ctx)
-          z4c_cuda_release_step_ctx(cg);
-      }
-      if (BP == Pp->data->ble)
-        break;
-      BP = BP->next;
-    }
-    Pp = Pp->next;
-  }
-}
-
-bool z4c_cuda_patch_contains_point(Patch *patch, const double *point)
-{
-  if (!patch)
-    return false;
-  for (int d = 0; d < dim; d++)
-  {
-    const double h = patch->getdX(d);
-    const double lo = patch->bbox[d] + patch->lli[d] * h;
-    const double hi = patch->bbox[dim + d] - patch->uui[d] * h;
-    if (point[d] < lo || point[d] > hi)
-      return false;
-  }
-  return true;
-}
-
-bool z4c_cuda_point_in_block(Patch *patch, Block *block,
-                             const double *point, const double *DH)
-{
-  if (!patch || !block)
-    return false;
-  for (int d = 0; d < dim; d++)
-  {
-    double llb;
-    double uub;
-#ifdef Vertex
-#ifdef Cell
-#error Both Cell and Vertex are defined
-#endif
-    llb = (feq(block->bbox[d], patch->bbox[d], DH[d] / 2))
-              ? block->bbox[d] + patch->lli[d] * DH[d]
-              : block->bbox[d] + (ghost_width - 0.5) * DH[d];
-    uub = (feq(block->bbox[dim + d], patch->bbox[dim + d], DH[d] / 2))
-              ? block->bbox[dim + d] - patch->uui[d] * DH[d]
-              : block->bbox[dim + d] - (ghost_width - 0.5) * DH[d];
-#else
-#ifdef Cell
-    llb = (feq(block->bbox[d], patch->bbox[d], DH[d] / 2))
-              ? block->bbox[d] + patch->lli[d] * DH[d]
-              : block->bbox[d] + ghost_width * DH[d];
-    uub = (feq(block->bbox[dim + d], patch->bbox[dim + d], DH[d] / 2))
-              ? block->bbox[dim + d] - patch->uui[d] * DH[d]
-              : block->bbox[dim + d] - ghost_width * DH[d];
-#else
-#error Not define Vertex nor Cell
-#endif
-#endif
-    if (point[d] - llb < -DH[d] / 2 || point[d] - uub > DH[d] / 2)
-      return false;
-  }
-  return true;
-}
-
-int z4c_cuda_interp_tile_start(const double *coords, int n, double x, double dx, int ordn)
-{
-  if (!coords || n <= ordn)
-    return 0;
-  int cxi = int((x - coords[0]) / dx + 0.4) + 1;
-  int start = cxi - ordn / 2;
-  if (start < 0)
-    start = 0;
-  const int max_start = n - ordn;
-  if (start > max_start)
-    start = max_start;
-  return start;
-}
-
-bool z4c_cuda_interp_bh_point_resident(MyList<Patch> *PatL,
-                                       int myrank,
-                                       const double *point,
-                                       var *forx, var *fory, var *forz,
-                                       int Symmetry,
-                                       double *shellf)
-{
-  const int ordn = 2 * ghost_width;
-  int owner_rank = -1;
-
-  shellf[0] = shellf[1] = shellf[2] = 0.0;
-
-  MyList<Patch> *PL = PatL;
-  while (PL)
-  {
-    Patch *patch = PL->data;
-    if (!z4c_cuda_patch_contains_point(patch, point))
-    {
-      PL = PL->next;
-      continue;
-    }
-
-    double DH[dim];
-    for (int d = 0; d < dim; d++)
-      DH[d] = patch->getdX(d);
-
-    MyList<Block> *BP = patch->blb;
-    while (BP)
-    {
-      Block *block = BP->data;
-      if (z4c_cuda_point_in_block(patch, block, point, DH))
-      {
-        owner_rank = block->rank;
-        if (myrank == owner_rank)
-        {
-          int interp_ordn = ordn;
-          int interp_sym = Symmetry;
-          double x = point[0];
-          double y = point[1];
-          double z = point[2];
-
-          if (z4c_cuda_has_resident_state(block) &&
-              block->shape[0] >= ordn && block->shape[1] >= ordn && block->shape[2] >= ordn)
-          {
-            const int sx = ordn;
-            const int sy = ordn;
-            const int sz = ordn;
-            const int region_all = sx * sy * sz;
-            const int i0 = z4c_cuda_interp_tile_start(block->X[0], block->shape[0], x, DH[0], ordn);
-            const int j0 = z4c_cuda_interp_tile_start(block->X[1], block->shape[1], y, DH[1], ordn);
-            const int k0 = z4c_cuda_interp_tile_start(block->X[2], block->shape[2], z, DH[2], ordn);
-            double *packed_fields = new double[3 * region_all];
-            var *vars[3] = {forx, fory, forz};
-            for (int f = 0; f < 3; f++)
-            {
-              if (z4c_cuda_pack_state_region_to_host_buffer(block,
-                                                            k_z4c_cuda_bh_state_indices[f],
-                                                            packed_fields + f * region_all,
-                                                            block->shape,
-                                                            i0, j0, k0,
-                                                            sx, sy, sz) != 0)
-              {
-                delete[] packed_fields;
-                cout << "CUDA Z4C BH tile download failed" << endl;
-                MPI_Abort(MPI_COMM_WORLD, 1);
-              }
-              int tile_shape[3] = {sx, sy, sz};
-              f_global_interp(tile_shape,
-                              block->X[0] + i0,
-                              block->X[1] + j0,
-                              block->X[2] + k0,
-                              packed_fields + f * region_all,
-                              shellf[f],
-                              x, y, z,
-                              interp_ordn,
-                              vars[f]->SoA,
-                              interp_sym);
-            }
-            delete[] packed_fields;
-          }
-          else
-          {
-            f_global_interp(block->shape, block->X[0], block->X[1], block->X[2],
-                            block->fgfs[forx->sgfn], shellf[0],
-                            x, y, z, interp_ordn, forx->SoA, interp_sym);
-            f_global_interp(block->shape, block->X[0], block->X[1], block->X[2],
-                            block->fgfs[fory->sgfn], shellf[1],
-                            x, y, z, interp_ordn, fory->SoA, interp_sym);
-            f_global_interp(block->shape, block->X[0], block->X[1], block->X[2],
-                            block->fgfs[forz->sgfn], shellf[2],
-                            x, y, z, interp_ordn, forz->SoA, interp_sym);
-          }
-        }
-        break;
-      }
-      if (BP == patch->ble)
-        break;
-      BP = BP->next;
-    }
-
-    if (owner_rank >= 0)
-      break;
-    PL = PL->next;
-  }
-
-  if (owner_rank < 0)
-    return false;
-
-  MPI_Bcast(shellf, 3, MPI_DOUBLE, owner_rank, MPI_COMM_WORLD);
-  return true;
-}
-
-bool z4c_cuda_compute_porg_rhs_resident(cgh *GH,
-                                        int ilev,
-                                        int myrank,
-                                        int BH_num,
-                                        double **BH_PS,
-                                        double **BH_RHS,
-                                        var *forx, var *fory, var *forz,
-                                        int Symmetry)
-{
-  for (int n = 0; n < BH_num; n++)
-  {
-    double shellf[3] = {0.0, 0.0, 0.0};
-    int lev = ilev;
-    while (lev >= 0 &&
-           !z4c_cuda_interp_bh_point_resident(GH->PatL[lev], myrank, BH_PS[n],
-                                              forx, fory, forz, Symmetry, shellf))
-    {
-      --lev;
-    }
-    if (lev < 0)
-      return false;
-    BH_RHS[n][0] = -shellf[0];
-    BH_RHS[n][1] = -shellf[1];
-    BH_RHS[n][2] = -shellf[2];
-  }
-  return true;
-}
-
-bool z4c_cuda_resident_step_enabled()
-{
-  static int enabled = -1;
-  if (enabled < 0)
-  {
-    const char *env = getenv("AMSS_Z4C_CUDA_RESIDENT");
-    enabled = (env && atoi(env) != 0) ? 1 : 0;
-  }
-  return enabled != 0;
-}
-
-} // namespace
-#endif
-
 void Z4c_class::Step(int lev, int YN)
 {
-#if USE_CUDA_Z4C && (ABEtype == 2)
-  double dT_lev = dT * pow(0.5, Mymax(lev, trfls));
-#ifdef With_AHF
-  AH_Step_Find(lev, dT_lev);
-#endif
-  bool BB = fgt(PhysTime, StartTime, dT_lev / 2);
-  double ndeps = numepss;
-  if (lev < GH->movls)
-    ndeps = numepsb;
-  double TRK4 = PhysTime;
-  int iter_count = 0;
-  int pre = 0, cor = 1;
-  int ERROR = 0;
-
-#ifdef WithShell
-  if (bssn_cuda_use_resident_sync(lev))
-  {
-    for (int dl = 0; dl < GH->levels; dl++)
-      bssn_cuda_download_level_state_if_present(GH->PatL[dl], StateList, myrank);
-  }
-#endif
-
-  MyList<Patch> *Pp = GH->PatL[lev];
-  while (Pp)
-  {
-    MyList<Block> *BP = Pp->data->blb;
-    while (BP)
-    {
-      Block *cg = BP->data;
-      if (myrank == cg->rank)
-      {
-        double *state_in[Z4C_CUDA_STATE_COUNT];
-        double *state_out[Z4C_CUDA_STATE_COUNT];
-        double propspeed[Z4C_CUDA_STATE_COUNT];
-        double soa_flat[3 * Z4C_CUDA_STATE_COUNT];
-        if (!fill_z4c_cuda_views(cg, StateList, state_in, propspeed, soa_flat) ||
-            !fill_z4c_cuda_views(cg, SynchList_pre, state_out))
-        {
-          cout << "CUDA Z4C state list mismatch on predictor step" << endl;
-          MPI_Abort(MPI_COMM_WORLD, 1);
-        }
-        int apply_bam_bc = 0;
-#if (MRBD == 0)
-#if (SommerType == 0)
-        apply_bam_bc = (lev == 0) ? 1 : 0;
-#endif
-#elif (MRBD == 1)
-        apply_bam_bc = 1;
-#endif
-        int keep_resident_state = z4c_cuda_resident_step_enabled() ? 1 : 0;
-        int apply_enforce_ga = 0;
-#if (AGM == 0)
-        apply_enforce_ga = 1;
-#endif
-        if (z4c_cuda_rk4_substep(cg,
-                                 cg->shape, cg->X[0], cg->X[1], cg->X[2],
-                                 state_in, state_out,
-                                 propspeed, soa_flat, Pp->data->bbox,
-                                 dT_lev, TRK4, iter_count, apply_bam_bc,
-                                 Symmetry, lev, ndeps, pre,
-                                 keep_resident_state, apply_enforce_ga, chitiny))
-        {
-          cout << "CUDA Z4C predictor substep failed in domain: ("
-               << cg->bbox[0] << ":" << cg->bbox[3] << ","
-               << cg->bbox[1] << ":" << cg->bbox[4] << ","
-               << cg->bbox[2] << ":" << cg->bbox[5] << ")" << endl;
-          ERROR = 1;
-        }
-      }
-      if (BP == Pp->data->ble)
-        break;
-      BP = BP->next;
-    }
-    Pp = Pp->next;
-  }
-
-  {
-    int erh = ERROR;
-    MPI_Allreduce(&erh, &ERROR, 1, MPI_INT, MPI_SUM, MPI_COMM_WORLD);
-  }
-  if (ERROR)
-  {
-    if (myrank == 0 && ErrorMonitor->outfile)
-      ErrorMonitor->outfile << "CUDA Z4C failed in predictor at t = " << PhysTime
-                            << ", lev = " << lev << endl;
-    MPI_Abort(MPI_COMM_WORLD, 1);
-  }
-
-  Parallel::Sync_cached(GH->PatL[lev], SynchList_pre, Symmetry, sync_cache_pre[lev]);
-
-  if (BH_num > 0 && lev == GH->levels - 1)
-  {
-    compute_Porg_rhs(Porg0, Porg_rhs, Sfx0, Sfy0, Sfz0, lev);
-    for (int ithBH = 0; ithBH < BH_num; ithBH++)
-    {
-      f_rungekutta4_scalar(dT_lev, Porg0[ithBH][0], Porg[ithBH][0], Porg_rhs[ithBH][0], iter_count);
-      f_rungekutta4_scalar(dT_lev, Porg0[ithBH][1], Porg[ithBH][1], Porg_rhs[ithBH][1], iter_count);
-      f_rungekutta4_scalar(dT_lev, Porg0[ithBH][2], Porg[ithBH][2], Porg_rhs[ithBH][2], iter_count);
-      if (Symmetry > 0)
-        Porg[ithBH][2] = fabs(Porg[ithBH][2]);
-      if (Symmetry == 2)
-      {
-        Porg[ithBH][0] = fabs(Porg[ithBH][0]);
-        Porg[ithBH][1] = fabs(Porg[ithBH][1]);
-      }
-    }
-  }
-
-  if ((lev == a_lev) && (LastAnas + dT_lev >= AnasTime))
-    z4c_cuda_download_level_state(GH->PatL[lev], SynchList_pre, myrank, false);
-  if (lev == a_lev)
-    AnalysisStuff(lev, dT_lev);
-
-  for (iter_count = 1; iter_count < 4; iter_count++)
-  {
-    if (iter_count == 1 || iter_count == 3)
-      TRK4 += dT_lev / 2;
-    Pp = GH->PatL[lev];
-    while (Pp)
-    {
-      MyList<Block> *BP = Pp->data->blb;
-      while (BP)
-      {
-        Block *cg = BP->data;
-        if (myrank == cg->rank)
-        {
-          double *state_in[Z4C_CUDA_STATE_COUNT];
-          double *state_out[Z4C_CUDA_STATE_COUNT];
-          double propspeed[Z4C_CUDA_STATE_COUNT];
-          double soa_flat[3 * Z4C_CUDA_STATE_COUNT];
-          if (!fill_z4c_cuda_views(cg, SynchList_pre, state_in, propspeed, soa_flat) ||
-              !fill_z4c_cuda_views(cg, SynchList_cor, state_out))
-          {
-            cout << "CUDA Z4C state list mismatch on corrector step" << endl;
-            MPI_Abort(MPI_COMM_WORLD, 1);
-          }
-          int apply_bam_bc = 0;
-#if (MRBD == 0)
-#if (SommerType == 0)
-          apply_bam_bc = (lev == 0) ? 1 : 0;
-#endif
-#elif (MRBD == 1)
-          apply_bam_bc = 1;
-#endif
-          int keep_resident_state = z4c_cuda_resident_step_enabled() ? 1 : 0;
-          int apply_enforce_ga = 0;
-#if (AGM == 0)
-          apply_enforce_ga = 1;
-#elif (AGM == 1)
-          apply_enforce_ga = (iter_count == 3) ? 1 : 0;
-#endif
-          if (z4c_cuda_rk4_substep(cg,
-                                   cg->shape, cg->X[0], cg->X[1], cg->X[2],
-                                   state_in, state_out,
-                                   propspeed, soa_flat, Pp->data->bbox,
-                                   dT_lev, TRK4, iter_count, apply_bam_bc,
-                                   Symmetry, lev, ndeps, cor,
-                                   keep_resident_state, apply_enforce_ga, chitiny))
-          {
-            cout << "CUDA Z4C corrector substep failed in domain: ("
-                 << cg->bbox[0] << ":" << cg->bbox[3] << ","
-                 << cg->bbox[1] << ":" << cg->bbox[4] << ","
-                 << cg->bbox[2] << ":" << cg->bbox[5] << ")" << endl;
-            ERROR = 1;
-          }
-        }
-        if (BP == Pp->data->ble)
-          break;
-        BP = BP->next;
-      }
-      Pp = Pp->next;
-    }
-
-    {
-      int erh = ERROR;
-      MPI_Allreduce(&erh, &ERROR, 1, MPI_INT, MPI_SUM, MPI_COMM_WORLD);
-    }
-    if (ERROR)
-    {
-      if (myrank == 0 && ErrorMonitor->outfile)
-        ErrorMonitor->outfile << "CUDA Z4C failed in RK4 substep#" << iter_count
-                              << " at t = " << PhysTime
-                              << ", lev = " << lev << endl;
-      MPI_Abort(MPI_COMM_WORLD, 1);
-    }
-
-    Parallel::Sync_cached(GH->PatL[lev], SynchList_cor, Symmetry, sync_cache_cor[lev]);
-
-    if (BH_num > 0 && lev == GH->levels - 1)
-    {
-      if (z4c_cuda_resident_step_enabled())
-      {
-        if (!z4c_cuda_compute_porg_rhs_resident(GH, lev, myrank, BH_num,
-                                                Porg, Porg1,
-                                                Sfx, Sfy, Sfz, Symmetry))
-        {
-          if (myrank == 0 && ErrorMonitor->outfile)
-            ErrorMonitor->outfile << "CUDA Z4C failed to interpolate black-hole shift at t = "
-                                  << PhysTime << endl;
-          MPI_Abort(MPI_COMM_WORLD, 1);
-        }
-      }
-      else
-      {
-        compute_Porg_rhs(Porg, Porg1, Sfx, Sfy, Sfz, lev);
-      }
-      for (int ithBH = 0; ithBH < BH_num; ithBH++)
-      {
-        f_rungekutta4_scalar(dT_lev, Porg0[ithBH][0], Porg1[ithBH][0], Porg_rhs[ithBH][0], iter_count);
-        f_rungekutta4_scalar(dT_lev, Porg0[ithBH][1], Porg1[ithBH][1], Porg_rhs[ithBH][1], iter_count);
-        f_rungekutta4_scalar(dT_lev, Porg0[ithBH][2], Porg1[ithBH][2], Porg_rhs[ithBH][2], iter_count);
-        if (Symmetry > 0)
-          Porg1[ithBH][2] = fabs(Porg1[ithBH][2]);
-        if (Symmetry == 2)
-        {
-          Porg1[ithBH][0] = fabs(Porg1[ithBH][0]);
-          Porg1[ithBH][1] = fabs(Porg1[ithBH][1]);
-        }
-      }
-    }
-
-    if (iter_count < 3)
-    {
-      Pp = GH->PatL[lev];
-      while (Pp)
-      {
-        MyList<Block> *BP = Pp->data->blb;
-        while (BP)
-        {
-          Block *cg = BP->data;
-          cg->swapList(SynchList_pre, SynchList_cor, myrank);
-          if (BP == Pp->data->ble)
-            break;
-          BP = BP->next;
-        }
-        Pp = Pp->next;
-      }
-      if (BH_num > 0 && lev == GH->levels - 1)
-      {
-        for (int ithBH = 0; ithBH < BH_num; ithBH++)
-        {
-          Porg[ithBH][0] = Porg1[ithBH][0];
-          Porg[ithBH][1] = Porg1[ithBH][1];
-          Porg[ithBH][2] = Porg1[ithBH][2];
-        }
-      }
-    }
-  }
-
-  z4c_cuda_download_level_state(GH->PatL[lev], SynchList_cor, myrank, false);
-
-#if (RPS == 0)
-  RestrictProlong(lev, YN, BB);
-#endif
-
-  Pp = GH->PatL[lev];
-  while (Pp)
-  {
-    MyList<Block> *BP = Pp->data->blb;
-    while (BP)
-    {
-      Block *cg = BP->data;
-      cg->swapList(StateList, SynchList_cor, myrank);
-      cg->swapList(OldStateList, SynchList_cor, myrank);
-      if (BP == Pp->data->ble)
-        break;
-      BP = BP->next;
-    }
-    Pp = Pp->next;
-  }
-  if (BH_num > 0 && lev == GH->levels - 1)
-  {
-    for (int ithBH = 0; ithBH < BH_num; ithBH++)
-    {
-      Porg0[ithBH][0] = Porg1[ithBH][0];
-      Porg0[ithBH][1] = Porg1[ithBH][1];
-      Porg0[ithBH][2] = Porg1[ithBH][2];
-    }
-  }
-#else
   double dT_lev = dT * pow(0.5, Mymax(lev, trfls));
 #ifdef With_AHF
   AH_Step_Find(lev, dT_lev);
@@ -1004,13 +339,6 @@ void Z4c_class::Step(int lev, int YN)
   }
 
 #ifdef WithShell
-#if USE_CUDA_Z4C
-  if (bssn_cuda_use_resident_sync(lev))
-  {
-    for (int dl = 0; dl < GH->levels; dl++)
-      bssn_cuda_download_level_state_if_present(GH->PatL[dl], StateList, myrank);
-  }
-#endif
   // evolve Shell Patches
   if (lev == 0)
   {
@@ -1714,11 +1042,9 @@ void Z4c_class::Step(int lev, int YN)
       Porg0[ithBH][2] = Porg1[ithBH][2];
     }
   }
-#endif
 }
 #else
 // for constraint preserving boundary (CPBC)
-// Note: CPBC path uses CPU Fortran RHS; GPU resident sync is a no-op here.
 #ifndef WithShell
 #error "CPBC only supports Shell"
 #endif
@@ -1748,14 +1074,6 @@ void Z4c_class::Step(int lev, int YN)
   int pre = 0, cor = 1;
   int ERROR = 0;
 
-#if USE_CUDA_Z4C && defined(WithShell)
-  if (bssn_cuda_use_resident_sync(lev))
-  {
-    for (int dl = 0; dl < GH->levels; dl++)
-      bssn_cuda_download_level_state_if_present(GH->PatL[dl], StateList, myrank);
-  }
-#endif
-
   MyList<ss_patch> *sPp;
   // Predictor
   MyList<Patch> *Pp = GH->PatL[lev];
@@ -3086,11 +2404,6 @@ void Z4c_class::Check_extrop()
 
 //================================================================================================
 
-#if USE_CUDA_Z4C && (ABEtype == 2) && defined(WithShell)
-#undef f_compute_rhs_Z4c_ss
-#define f_compute_rhs_Z4c_ss compute_rhs_z4c_ss_
-#endif
-
 // this member function is used to compute and output constraint violation
 
 //================================================================================================
@@ -3366,12 +2679,11 @@ void Z4c_class::Interp_Constraint()
   }
 
   ofstream outfile;
-  char suffix[64];
-  sprintf(suffix, "/interp_constraint_%05d.dat", int(PhysTime / dT + 0.5));
-  string filename = ErrorMonitor->out_dir + suffix;
+  char filename[50];
+  sprintf(filename, "%s/interp_constraint_%05d.dat", ErrorMonitor->out_dir.c_str(), int(PhysTime / dT + 0.5)); 
   // 0.5 for round off
   
-  outfile.open(filename.c_str());
+  outfile.open(filename);
   outfile << "#  corrdinate, H_Res, Px_Res, Py_Res, Pz_Res, Gx_Res, Gy_Res, Gz_Res, ...." << endl;
   for (int i = 0; i < n; i++)
   {

AMSS_NCKU_source/Z4c_rhs.f90

@@ -94,31 +94,29 @@
                Hcon,Mxcon,Mycon,Mzcon,Gmxcon,Gmycon,Gmzcon,                    &            
                Symmetry,Lev,eps,co)
 
-  if (co == 0) then
-#if (ABV == 0)  
-    call ricci_gamma(ex, X, Y, Z,                                      &
-                 chi,                                                  &
-                 dxx    ,   gxy    ,   gxz    ,   dyy    ,   gyz    ,   dzz,&
-                 Gamx   ,  Gamy    ,  Gamz    , &
-                 Gamxxx,Gamxxy,Gamxxz,Gamxyy,Gamxyz,Gamxzz,&
-                 Gamyxx,Gamyxy,Gamyxz,Gamyyy,Gamyyz,Gamyzz,&
-                 Gamzxx,Gamzxy,Gamzxz,Gamzyy,Gamzyz,Gamzzz,&
-                 Rxx,Rxy,Rxz,Ryy,Ryz,Rzz,&
-                 Symmetry)
-#endif
-    call constraint_bssn(ex, X, Y, Z,&
-                 chi,trK, &
-                 dxx,gxy,gxz,dyy,gyz,dzz, &
-                 Axx,Axy,Axz,Ayy,Ayz,Azz, &
-                 Gamx,Gamy,Gamz,&
-                 Lap,betax,betay,betaz,rho,Sx,Sy,Sz,&
-                 Gamxxx, Gamxxy, Gamxxz,Gamxyy, Gamxyz, Gamxzz, &
-                 Gamyxx, Gamyxy, Gamyxz,Gamyyy, Gamyyz, Gamyzz, &
-                 Gamzxx, Gamzxy, Gamzxz,Gamzyy, Gamzyz, Gamzzz, &
-                 Rxx,Rxy,Rxz,Ryy,Ryz,Rzz, &
-                 Hcon,Mxcon,Mycon,Mzcon,Gmxcon,Gmycon,Gmzcon, &
-                 Symmetry)
-  endif
+#if (ABV == 0)  
+  call ricci_gamma(ex, X, Y, Z,                                      &
+               chi,                                                  &
+               dxx    ,   gxy    ,   gxz    ,   dyy    ,   gyz    ,   dzz,&
+               Gamx   ,  Gamy    ,  Gamz    , &
+               Gamxxx,Gamxxy,Gamxxz,Gamxyy,Gamxyz,Gamxzz,&
+               Gamyxx,Gamyxy,Gamyxz,Gamyyy,Gamyyz,Gamyzz,&
+               Gamzxx,Gamzxy,Gamzxz,Gamzyy,Gamzyz,Gamzzz,&
+               Rxx,Rxy,Rxz,Ryy,Ryz,Rzz,&
+               Symmetry)
+#endif
+  call constraint_bssn(ex, X, Y, Z,&
+               chi,trK, &
+               dxx,gxy,gxz,dyy,gyz,dzz, &
+               Axx,Axy,Axz,Ayy,Ayz,Azz, &
+               Gamx,Gamy,Gamz,&
+               Lap,betax,betay,betaz,rho,Sx,Sy,Sz,&
+               Gamxxx, Gamxxy, Gamxxz,Gamxyy, Gamxyz, Gamxzz, &
+               Gamyxx, Gamyxy, Gamyxz,Gamyyy, Gamyyz, Gamyzz, &
+               Gamzxx, Gamzxy, Gamzxz,Gamzyy, Gamzyz, Gamzzz, &
+               Rxx,Rxy,Rxz,Ryy,Ryz,Rzz, &
+               Hcon,Mxcon,Mycon,Mzcon,Gmxcon,Gmycon,Gmzcon, &
+               Symmetry)
 
   return
 
@@ -228,12 +226,11 @@
 
   call get_Z4cparameters(kappa1,kappa2,kappa3,FF,eta)
 
-!!! sanity check
-#ifdef DEBUG
-  dX = sum(chi)+sum(trK)+sum(dxx)+sum(gxy)+sum(gxz)+sum(dyy)+sum(gyz)+sum(dzz) &
-      +sum(Axx)+sum(Axy)+sum(Axz)+sum(Ayy)+sum(Ayz)+sum(Azz)                   &
-      +sum(Gamx)+sum(Gamy)+sum(Gamz)                                           &
-      +sum(Lap)+sum(betax)+sum(betay)+sum(betaz)+sum(dtSfx)+sum(dtSfy)+sum(dtSfz) &
+!!! sanity check
+  dX = sum(chi)+sum(trK)+sum(dxx)+sum(gxy)+sum(gxz)+sum(dyy)+sum(gyz)+sum(dzz) &
+      +sum(Axx)+sum(Axy)+sum(Axz)+sum(Ayy)+sum(Ayz)+sum(Azz)                   &
+      +sum(Gamx)+sum(Gamy)+sum(Gamz)                                           &
+      +sum(Lap)+sum(betax)+sum(betay)+sum(betaz)+sum(dtSfx)+sum(dtSfy)+sum(dtSfz) &
       +sum(TZ)
   if(dX.ne.dX) then
      if(sum(chi).ne.sum(chi))write(*,*)"Z4c_rhs.f90: find NaN in chi"
@@ -260,11 +257,10 @@
      if(sum(dtSfx).ne.sum(dtSfx))write(*,*)"Z4c_rhs.f90: find NaN in dtSfx"
      if(sum(dtSfy).ne.sum(dtSfy))write(*,*)"Z4c_rhs.f90: find NaN in dtSfy"
      if(sum(dtSfz).ne.sum(dtSfz))write(*,*)"Z4c_rhs.f90: find NaN in dtSfz"
-     if(sum(TZ).ne.sum(Tz))write(*,*)"Z4c_rhs.f90: find NaN in TZ"
-     gont = 1
-     return
-  endif
-#endif
+     if(sum(TZ).ne.sum(Tz))write(*,*)"Z4c_rhs.f90: find NaN in TZ"
+     gont = 1
+     return
+  endif
 
   PI = dacos(-ONE)
 
@@ -1267,32 +1263,30 @@
 
   endif
 
-  if (co == 0) then
-#if (ABV == 0)  
-    call ricci_gamma(ex, X, Y, Z,                                      &
-                 chi,                                                  &
-                 dxx    ,   gxy    ,   gxz    ,   dyy    ,   gyz    ,   dzz,&
-                 Gamx   ,  Gamy    ,  Gamz    , &
-                 Gamxxx,Gamxxy,Gamxxz,Gamxyy,Gamxyz,Gamxzz,&
-                 Gamyxx,Gamyxy,Gamyxz,Gamyyy,Gamyyz,Gamyzz,&
-                 Gamzxx,Gamzxy,Gamzxz,Gamzyy,Gamzyz,Gamzzz,&
-                 Rxx,Rxy,Rxz,Ryy,Ryz,Rzz,&
-                 Symmetry)
-#endif
-
-    call constraint_bssn(ex, X, Y, Z,&
-                 chi,trK, &
-                 dxx,gxy,gxz,dyy,gyz,dzz, &
-                 Axx,Axy,Axz,Ayy,Ayz,Azz, &
-                 Gamx,Gamy,Gamz,&
-                 Lap,betax,betay,betaz,rho,Sx,Sy,Sz,&
-                 Gamxxx, Gamxxy, Gamxxz,Gamxyy, Gamxyz, Gamxzz, &
-                 Gamyxx, Gamyxy, Gamyxz,Gamyyy, Gamyyz, Gamyzz, &
-                 Gamzxx, Gamzxy, Gamzxz,Gamzyy, Gamzyz, Gamzzz, &
-                 Rxx,Rxy,Rxz,Ryy,Ryz,Rzz, &
-                 Hcon,Mxcon,Mycon,Mzcon,Gmxcon,Gmycon,Gmzcon, &
-                 Symmetry)
-  endif
+#if (ABV == 0)  
+  call ricci_gamma(ex, X, Y, Z,                                      &
+               chi,                                                  &
+               dxx    ,   gxy    ,   gxz    ,   dyy    ,   gyz    ,   dzz,&
+               Gamx   ,  Gamy    ,  Gamz    , &
+               Gamxxx,Gamxxy,Gamxxz,Gamxyy,Gamxyz,Gamxzz,&
+               Gamyxx,Gamyxy,Gamyxz,Gamyyy,Gamyyz,Gamyzz,&
+               Gamzxx,Gamzxy,Gamzxz,Gamzyy,Gamzyz,Gamzzz,&
+               Rxx,Rxy,Rxz,Ryy,Ryz,Rzz,&
+               Symmetry)
+#endif
+
+  call constraint_bssn(ex, X, Y, Z,&
+               chi,trK, &
+               dxx,gxy,gxz,dyy,gyz,dzz, &
+               Axx,Axy,Axz,Ayy,Ayz,Azz, &
+               Gamx,Gamy,Gamz,&
+               Lap,betax,betay,betaz,rho,Sx,Sy,Sz,&
+               Gamxxx, Gamxxy, Gamxxz,Gamxyy, Gamxyz, Gamxzz, &
+               Gamyxx, Gamyxy, Gamyxz,Gamyyy, Gamyyz, Gamyzz, &
+               Gamzxx, Gamzxy, Gamzxz,Gamzyy, Gamzyz, Gamzzz, &
+               Rxx,Rxy,Rxz,Ryy,Ryz,Rzz, &
+               Hcon,Mxcon,Mycon,Mzcon,Gmxcon,Gmycon,Gmzcon, &
+               Symmetry)
 
   gont = 0
 

AMSS_NCKU_source/Z4c_rhs_ss.f90

@@ -121,12 +121,11 @@
 
   call get_Z4cparameters(kappa1,kappa2,kappa3,FF,eta)
 
-!!! sanity check
-#ifdef DEBUG
-  dX = sum(chi)+sum(trK)+sum(dxx)+sum(gxy)+sum(gxz)+sum(dyy)+sum(gyz)+sum(dzz) &
-      +sum(Axx)+sum(Axy)+sum(Axz)+sum(Ayy)+sum(Ayz)+sum(Azz)                   &
-      +sum(Gamx)+sum(Gamy)+sum(Gamz)                                           &
-      +sum(Lap)+sum(betax)+sum(betay)+sum(betaz)+sum(dtSfx)+sum(dtSfy)+sum(dtSfz) &
+!!! sanity check
+  dX = sum(chi)+sum(trK)+sum(dxx)+sum(gxy)+sum(gxz)+sum(dyy)+sum(gyz)+sum(dzz) &
+      +sum(Axx)+sum(Axy)+sum(Axz)+sum(Ayy)+sum(Ayz)+sum(Azz)                   &
+      +sum(Gamx)+sum(Gamy)+sum(Gamz)                                           &
+      +sum(Lap)+sum(betax)+sum(betay)+sum(betaz)+sum(dtSfx)+sum(dtSfy)+sum(dtSfz) &
       +sum(TZ)
   if(dX.ne.dX) then
      if(sum(chi).ne.sum(chi))write(*,*)"Z4c_rhs_ss.f90: find NaN in chi"
@@ -153,11 +152,10 @@
      if(sum(dtSfx).ne.sum(dtSfx))write(*,*)"Z4c_rhs_ss.f90: find NaN in dtSfx"
      if(sum(dtSfy).ne.sum(dtSfy))write(*,*)"Z4c_rhs_ss.f90: find NaN in dtSfy"
      if(sum(dtSfz).ne.sum(dtSfz))write(*,*)"Z4c_rhs_ss.f90: find NaN in dtSfz"
-     if(sum(TZ).ne.sum(Tz))write(*,*)"Z4c_rhs_ss.f90: find NaN in TZ"
-     gont = 1
-     return
-  endif
-#endif
+     if(sum(TZ).ne.sum(Tz))write(*,*)"Z4c_rhs_ss.f90: find NaN in TZ"
+     gont = 1
+     return
+  endif
 
   PI = dacos(-ONE)
 
@@ -1390,43 +1388,41 @@
   call kodis_sh(ex,crho,sigma,R,TZ,TZ_rhs,SSS,Symmetry,eps,sst)
   endif
 
-  if (co == 0) then
-#if (ABV == 1)  
-    call ricci_gamma_ss(ex,crho,sigma,R,X, Y, Z,                               &
-                 drhodx, drhody, drhodz,                                       &
-                 dsigmadx,dsigmady,dsigmadz,                                   &
-                 dRdx,dRdy,dRdz,                                               &
-                 drhodxx,drhodxy,drhodxz,drhodyy,drhodyz,drhodzz,              &
-                 dsigmadxx,dsigmadxy,dsigmadxz,dsigmadyy,dsigmadyz,dsigmadzz,  &
-                 dRdxx,dRdxy,dRdxz,dRdyy,dRdyz,dRdzz,                          &
-                 chi,                                                          &
-                 dxx    ,   gxy    ,   gxz    ,   dyy    ,   gyz    ,   dzz,&
-                 Gamx   ,  Gamy    ,  Gamz    , &
-                 Gamxxx,Gamxxy,Gamxxz,Gamxyy,Gamxyz,Gamxzz,&
-                 Gamyxx,Gamyxy,Gamyxz,Gamyyy,Gamyyz,Gamyzz,&
-                 Gamzxx,Gamzxy,Gamzxz,Gamzyy,Gamzyz,Gamzzz,&
-                 Rxx,Rxy,Rxz,Ryy,Ryz,Rzz,&
-                 Symmetry,Lev,sst)
-#endif
-    call constraint_bssn_ss(ex,crho,sigma,R,X, Y, Z,  &
-                 drhodx, drhody, drhodz,                                       &
-                 dsigmadx,dsigmady,dsigmadz,                                   &
-                 dRdx,dRdy,dRdz,                                               &
-                 drhodxx,drhodxy,drhodxz,drhodyy,drhodyz,drhodzz,              &
-                 dsigmadxx,dsigmadxy,dsigmadxz,dsigmadyy,dsigmadyz,dsigmadzz,  &
-                 dRdxx,dRdxy,dRdxz,dRdyy,dRdyz,dRdzz,                          &
-                 chi,trK, &
-                 dxx,gxy,gxz,dyy,gyz,dzz, &
-                 Axx,Axy,Axz,Ayy,Ayz,Azz, &
-                 Gamx,Gamy,Gamz,&
-                 Lap,betax,betay,betaz,rho,Sx,Sy,Sz,&
-                 Gamxxx, Gamxxy, Gamxxz,Gamxyy, Gamxyz, Gamxzz, &
-                 Gamyxx, Gamyxy, Gamyxz,Gamyyy, Gamyyz, Gamyzz, &
-                 Gamzxx, Gamzxy, Gamzxz,Gamzyy, Gamzyz, Gamzzz, &
-                 Rxx,Rxy,Rxz,Ryy,Ryz,Rzz, &
-                 Hcon,Mxcon,Mycon,Mzcon,Gmxcon,Gmycon,Gmzcon, &
-                 Symmetry,Lev,sst)
-  endif
+#if (ABV == 1)  
+  call ricci_gamma_ss(ex,crho,sigma,R,X, Y, Z,                                 &
+               drhodx, drhody, drhodz,                                         &
+               dsigmadx,dsigmady,dsigmadz,                                     &
+               dRdx,dRdy,dRdz,                                                 &
+               drhodxx,drhodxy,drhodxz,drhodyy,drhodyz,drhodzz,                &
+               dsigmadxx,dsigmadxy,dsigmadxz,dsigmadyy,dsigmadyz,dsigmadzz,    &
+               dRdxx,dRdxy,dRdxz,dRdyy,dRdyz,dRdzz,                            &
+               chi,                                                  &
+               dxx    ,   gxy    ,   gxz    ,   dyy    ,   gyz    ,   dzz,&
+               Gamx   ,  Gamy    ,  Gamz    , &
+               Gamxxx,Gamxxy,Gamxxz,Gamxyy,Gamxyz,Gamxzz,&
+               Gamyxx,Gamyxy,Gamyxz,Gamyyy,Gamyyz,Gamyzz,&
+               Gamzxx,Gamzxy,Gamzxz,Gamzyy,Gamzyz,Gamzzz,&
+               Rxx,Rxy,Rxz,Ryy,Ryz,Rzz,&
+               Symmetry,Lev,sst)
+  call constraint_bssn_ss(ex,crho,sigma,R,X, Y, Z,  &
+               drhodx, drhody, drhodz,                                         &
+               dsigmadx,dsigmady,dsigmadz,                                     &
+               dRdx,dRdy,dRdz,                                                 &
+               drhodxx,drhodxy,drhodxz,drhodyy,drhodyz,drhodzz,                &
+               dsigmadxx,dsigmadxy,dsigmadxz,dsigmadyy,dsigmadyz,dsigmadzz,    &
+               dRdxx,dRdxy,dRdxz,dRdyy,dRdyz,dRdzz,                            &
+               chi,trK, &
+               dxx,gxy,gxz,dyy,gyz,dzz, &
+               Axx,Axy,Axz,Ayy,Ayz,Azz, &
+               Gamx,Gamy,Gamz,&
+               Lap,betax,betay,betaz,rho,Sx,Sy,Sz,&
+               Gamxxx, Gamxxy, Gamxxz,Gamxyy, Gamxyz, Gamxzz, &
+               Gamyxx, Gamyxy, Gamyxz,Gamyyy, Gamyyz, Gamyzz, &
+               Gamzxx, Gamzxy, Gamzxz,Gamzyy, Gamzyz, Gamzzz, &
+               Rxx,Rxy,Rxz,Ryy,Ryz,Rzz, &
+               Hcon,Mxcon,Mycon,Mzcon,Gmxcon,Gmycon,Gmzcon, &
+               Symmetry,Lev,sst)
+#endif
 
   gont = 0
 

AMSS_NCKU_source/bssnEScalar_class.h

@@ -51,7 +51,7 @@ public:
      void Compute_Psi4(int lev);
      void Step(int lev, int YN);
      void AnalysisStuff_EScalar(int lev, double dT_lev);
-     void Interp_Constraint();
+     void Interp_Constraint(bool infg);
      void Constraint_Out(); 
 
 protected:

AMSS_NCKU_source/bssn_class.h

@@ -31,11 +31,19 @@ using namespace std;
 #include "surface_integral.h"
 #include "checkpoint.h"
 
-extern void setpbh(int iBHN, double **iPBH, double *iMass, int rBHN);
-
-class bssn_class
-{
-public:
+extern void setpbh(int iBHN, double **iPBH, double *iMass, int rBHN);
+
+#ifndef BSSN_USE_TRANSFER_CACHE
+#define BSSN_USE_TRANSFER_CACHE 1
+#endif
+
+#ifndef BSSN_USE_ESCALAR_C_KERNEL
+#define BSSN_USE_ESCALAR_C_KERNEL 1
+#endif
+
+class bssn_class
+{
+public:
        int ngfs;
        int nprocs, myrank;
        cgh *GH;
@@ -48,7 +56,6 @@ public:
        double StartTime, TotalTime;
        double AnasTime, DumpTime, d2DumpTime, CheckTime;
        double LastAnas, LastConsOut;
-       bool cuda_level0_constraint_cache_valid;
        int *ConstraintRefreshLevels;
        double Courant;
        double numepss, numepsb, numepsh;
@@ -144,7 +151,7 @@ public:
        bssn_class(double Couranti, double StartTimei, double TotalTimei, double DumpTimei, double d2DumpTimei, double CheckTimei, double AnasTimei,
                   int Symmetryi, int checkruni, char *checkfilenamei, double numepssi, double numepsbi, double numepshi,
                   int a_levi, int maxli, int decni, double maxrexi, double drexi);
-       virtual ~bssn_class();
+       ~bssn_class();
 
        void Evolve(int Steps);
        void RecursiveStep(int lev);
@@ -168,14 +175,25 @@ public:
        void Setup_KerrSchild();
        void Enforce_algcon(int lev, int fg);
 
-       void testRestrict();
-       void testOutBd();
-       
-       bool check_Stdin_Abort(); 
-
-       virtual void Setup_Initial_Data_Cao();
-       virtual void Setup_Initial_Data_Lousto();
-       virtual void Initialize();
+       void testRestrict();
+       void testOutBd();
+       
+       bool check_Stdin_Abort(); 
+       bool use_transfer_cache() const;
+       void setup_transfer_caches();
+       void invalidate_transfer_caches();
+       void destroy_transfer_caches();
+       void sync_predictor_start(int lev, MyList<var> *VarList, Parallel::AsyncSyncState &async_state);
+       void sync_predictor_finish(int lev, Parallel::AsyncSyncState &async_state, MyList<var> *VarList);
+       void sync_corrector_start(int lev, MyList<var> *VarList, Parallel::AsyncSyncState &async_state);
+       void sync_corrector_finish(int lev, Parallel::AsyncSyncState &async_state, MyList<var> *VarList);
+       void sync_evolution(int lev, MyList<var> *VarList, Parallel::SyncCache *cache_array = 0);
+       void restrict_evolution(int lev, MyList<var> *src_var_list, MyList<var> *dst_var_list);
+       void outbdlow2hi_evolution(int lev, MyList<var> *src_var_list, MyList<var> *dst_var_list);
+
+       virtual void Setup_Initial_Data_Cao();
+       virtual void Setup_Initial_Data_Lousto();
+       virtual void Initialize();
        virtual void Read_Ansorg();
        virtual void Read_Pablo() {};
        virtual void Compute_Psi4(int lev);

AMSS_NCKU_source/bssn_escalar_rhs_c.C

@@ -0,0 +1,169 @@
+#include "macrodef.h"
+#include "bssn_rhs.h"
+#include "share_func.h"
+#include "tool.h"
+#include <vector>
+
+namespace
+{
+    // Reuse the temporary workspace across block calls to avoid repeated heap churn
+    // in the EScalar wrapper. MPI ranks execute this path sequentially, so a single
+    // process-local buffer is sufficient here.
+    std::vector<double> g_escalar_tmp_store;
+}
+
+#ifdef fortran1
+#define f_frpotential frpotential
+#endif
+#ifdef fortran2
+#define f_frpotential FRPOTENTIAL
+#endif
+#ifdef fortran3
+#define f_frpotential frpotential_
+#endif
+
+extern "C"
+{
+    void f_frpotential(int *, double *, double *, double *);
+}
+
+int f_compute_rhs_bssn_escalar_c(int *ex, double &T,
+                                 double *X, double *Y, double *Z,
+                                 double *chi, double *trK,
+                                 double *dxx, double *gxy, double *gxz, double *dyy, double *gyz, double *dzz,
+                                 double *Axx, double *Axy, double *Axz, double *Ayy, double *Ayz, double *Azz,
+                                 double *Gamx, double *Gamy, double *Gamz,
+                                 double *Lap, double *betax, double *betay, double *betaz,
+                                 double *dtSfx, double *dtSfy, double *dtSfz,
+                                 double *Sphi, double *Spi,
+                                 double *chi_rhs, double *trK_rhs,
+                                 double *gxx_rhs, double *gxy_rhs, double *gxz_rhs, double *gyy_rhs, double *gyz_rhs, double *gzz_rhs,
+                                 double *Axx_rhs, double *Axy_rhs, double *Axz_rhs, double *Ayy_rhs, double *Ayz_rhs, double *Azz_rhs,
+                                 double *Gamx_rhs, double *Gamy_rhs, double *Gamz_rhs,
+                                 double *Lap_rhs, double *betax_rhs, double *betay_rhs, double *betaz_rhs,
+                                 double *dtSfx_rhs, double *dtSfy_rhs, double *dtSfz_rhs,
+                                 double *Sphi_rhs, double *Spi_rhs,
+                                 double *rho, double *Sx, double *Sy, double *Sz,
+                                 double *Sxx, double *Sxy, double *Sxz, double *Syy, double *Syz, double *Szz,
+                                 double *Gamxxx, double *Gamxxy, double *Gamxxz, double *Gamxyy, double *Gamxyz, double *Gamxzz,
+                                 double *Gamyxx, double *Gamyxy, double *Gamyxz, double *Gamyyy, double *Gamyyz, double *Gamyzz,
+                                 double *Gamzxx, double *Gamzxy, double *Gamzxz, double *Gamzyy, double *Gamzyz, double *Gamzzz,
+                                 double *Rxx, double *Rxy, double *Rxz, double *Ryy, double *Ryz, double *Rzz,
+                                 double *ham_Res, double *movx_Res, double *movy_Res, double *movz_Res,
+                                 double *Gmx_Res, double *Gmy_Res, double *Gmz_Res,
+                                 int &Symmetry, int &Lev, double &eps, int &co)
+{
+    const int nx = ex[0], ny = ex[1], nz = ex[2];
+    const int all = nx * ny * nz;
+
+    const size_t workspace_size = size_t(all) * 17;
+    if (g_escalar_tmp_store.size() < workspace_size)
+        g_escalar_tmp_store.resize(workspace_size);
+
+    double *tmp_ptr = g_escalar_tmp_store.data();
+    auto alloc_tmp = [&](int n = 1) -> double *
+    {
+        double *ptr = tmp_ptr;
+        tmp_ptr += size_t(all) * n;
+        return ptr;
+    };
+
+    double *chix = alloc_tmp(), *chiy = alloc_tmp(), *chiz = alloc_tmp();
+    double *Kx = alloc_tmp(), *Ky = alloc_tmp(), *Kz = alloc_tmp();
+    double *fxx = alloc_tmp(), *fxy = alloc_tmp(), *fxz = alloc_tmp();
+    double *fyy = alloc_tmp(), *fyz = alloc_tmp(), *fzz = alloc_tmp();
+    double *Lapx = alloc_tmp(), *Lapy = alloc_tmp(), *Lapz = alloc_tmp();
+    double *V = alloc_tmp(), *dVdSphi = alloc_tmp();
+
+    const double ZEO = 0.0, ONE = 1.0, TWO = 2.0, HALF = 0.5;
+    const double SSS[3] = {1.0, 1.0, 1.0};
+
+    fderivs(ex, chi, chix, chiy, chiz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
+    fderivs(ex, Lap, Lapx, Lapy, Lapz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
+    fderivs(ex, Sphi, Kx, Ky, Kz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
+    fdderivs(ex, Sphi, fxx, fxy, fxz, fyy, fyz, fzz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
+
+    f_frpotential(ex, Sphi, V, dVdSphi);
+
+    for (int i = 0; i < all; ++i)
+    {
+        const double alpn1 = Lap[i] + ONE;
+        const double chin1 = chi[i] + ONE;
+        const double gxx = dxx[i] + ONE;
+        const double gyy = dyy[i] + ONE;
+        const double gzz = dzz[i] + ONE;
+        const double det = gxx * gyy * gzz + gxy[i] * gyz[i] * gxz[i] + gxz[i] * gxy[i] * gyz[i]
+                         - gxz[i] * gyy * gxz[i] - gxy[i] * gxy[i] * gzz - gxx * gyz[i] * gyz[i];
+        const double gupxx = (gyy * gzz - gyz[i] * gyz[i]) / det;
+        const double gupxy = -(gxy[i] * gzz - gyz[i] * gxz[i]) / det;
+        const double gupxz = (gxy[i] * gyz[i] - gyy * gxz[i]) / det;
+        const double gupyy = (gxx * gzz - gxz[i] * gxz[i]) / det;
+        const double gupyz = -(gxx * gyz[i] - gxy[i] * gxz[i]) / det;
+        const double gupzz = (gxx * gyy - gxy[i] * gxy[i]) / det;
+
+        Sphi_rhs[i] = alpn1 * Spi[i];
+
+        Spi_rhs[i] = gupxx * fxx[i] + gupyy * fyy[i] + gupzz * fzz[i]
+                   + TWO * (gupxy * fxy[i] + gupxz * fxz[i] + gupyz * fyz[i])
+                   - ((Gamx[i] + (gupxx * chix[i] + gupxy * chiy[i] + gupxz * chiz[i]) / TWO / chin1) * Kx[i]
+                   +  (Gamy[i] + (gupxy * chix[i] + gupyy * chiy[i] + gupyz * chiz[i]) / TWO / chin1) * Ky[i]
+                   +  (Gamz[i] + (gupxz * chix[i] + gupyz * chiy[i] + gupzz * chiz[i]) / TWO / chin1) * Kz[i]);
+
+        Spi_rhs[i] = Spi_rhs[i] * alpn1
+                   + gupxx * Lapx[i] * Kx[i] + gupxy * Lapx[i] * Ky[i] + gupxz * Lapx[i] * Kz[i]
+                   + gupxy * Lapy[i] * Kx[i] + gupyy * Lapy[i] * Ky[i] + gupyz * Lapy[i] * Kz[i]
+                   + gupxz * Lapz[i] * Kx[i] + gupyz * Lapz[i] * Ky[i] + gupzz * Lapz[i] * Kz[i];
+
+        Spi_rhs[i] = Spi_rhs[i] * chin1 + alpn1 * (trK[i] * Spi[i] - dVdSphi[i]);
+
+        rho[i] = chin1 * ((gupxx * Kx[i] * Kx[i] + gupyy * Ky[i] * Ky[i] + gupzz * Kz[i] * Kz[i]) * HALF
+               + gupxy * Kx[i] * Ky[i] + gupxz * Kx[i] * Kz[i] + gupyz * Ky[i] * Kz[i])
+               + Spi[i] * Spi[i] * HALF + V[i];
+        Sx[i] = -Spi[i] * Kx[i];
+        Sy[i] = -Spi[i] * Ky[i];
+        Sz[i] = -Spi[i] * Kz[i];
+
+        const double pressure = (rho[i] - Spi[i] * Spi[i]) / chin1;
+        Sxx[i] = Kx[i] * Kx[i] - pressure * gxx;
+        Sxy[i] = Kx[i] * Ky[i] - pressure * gxy[i];
+        Sxz[i] = Kx[i] * Kz[i] - pressure * gxz[i];
+        Syy[i] = Ky[i] * Ky[i] - pressure * gyy;
+        Syz[i] = Ky[i] * Kz[i] - pressure * gyz[i];
+        Szz[i] = Kz[i] * Kz[i] - pressure * gzz;
+    }
+
+    if (f_compute_rhs_bssn(ex, T, X, Y, Z,
+                           chi, trK,
+                           dxx, gxy, gxz, dyy, gyz, dzz,
+                           Axx, Axy, Axz, Ayy, Ayz, Azz,
+                           Gamx, Gamy, Gamz,
+                           Lap, betax, betay, betaz,
+                           dtSfx, dtSfy, dtSfz,
+                           chi_rhs, trK_rhs,
+                           gxx_rhs, gxy_rhs, gxz_rhs, gyy_rhs, gyz_rhs, gzz_rhs,
+                           Axx_rhs, Axy_rhs, Axz_rhs, Ayy_rhs, Ayz_rhs, Azz_rhs,
+                           Gamx_rhs, Gamy_rhs, Gamz_rhs,
+                           Lap_rhs, betax_rhs, betay_rhs, betaz_rhs,
+                           dtSfx_rhs, dtSfy_rhs, dtSfz_rhs,
+                           rho, Sx, Sy, Sz,
+                           Sxx, Sxy, Sxz, Syy, Syz, Szz,
+                           Gamxxx, Gamxxy, Gamxxz, Gamxyy, Gamxyz, Gamxzz,
+                           Gamyxx, Gamyxy, Gamyxz, Gamyyy, Gamyyz, Gamyzz,
+                           Gamzxx, Gamzxy, Gamzxz, Gamzyy, Gamzyz, Gamzzz,
+                           Rxx, Rxy, Rxz, Ryy, Ryz, Rzz,
+                           ham_Res, movx_Res, movy_Res, movz_Res,
+                           Gmx_Res, Gmy_Res, Gmz_Res,
+                           Symmetry, Lev, eps, co))
+        return 1;
+
+    lopsided_kodis(ex, X, Y, Z, Sphi, Sphi_rhs, betax, betay, betaz, Symmetry, SSS, eps);
+    lopsided_kodis(ex, X, Y, Z, Spi, Spi_rhs, betax, betay, betaz, Symmetry, SSS, eps);
+
+    for (int i = 0; i < all; ++i)
+    {
+        if (Sphi_rhs[i] != Sphi_rhs[i] || Spi_rhs[i] != Spi_rhs[i] || rho[i] != rho[i])
+            return 1;
+    }
+
+    return 0;
+}

AMSS_NCKU_source/bssn_gpu.h

@@ -1,56 +1,73 @@
-
-#ifndef BSSN_GPU_H_
-#define BSSN_GPU_H_
-#include "bssn_macro.h"
-#include "macrodef.h"
-
-#define DEVICE_ID 0
-// #define DEVICE_ID_BY_MPI_RANK
-#define GRID_DIM 256
-#define BLOCK_DIM 128
-
-#define _FH2_(i, j, k) fh[(i) + (j) * _1D_SIZE[2] + (k) * _2D_SIZE[2]]
-#define _FH3_(i, j, k) fh[(i) + (j) * _1D_SIZE[3] + (k) * _2D_SIZE[3]]
-#define pow2(x) ((x) * (x))
-#define TimeBetween(a, b) ((b.tv_sec - a.tv_sec) + (b.tv_usec - a.tv_usec) / 1000000.0f)
-#define M_ metac.
-#define Mh_ meta->
-#define Ms_ metassc.
-#define Msh_ metass->
-
-// #define TIMING
-
-#define RHS_SS_PARA int calledby, int mpi_rank, int *ex, double &T, double *crho, double *sigma, double *R, double *X, double *Y, double *Z, double *drhodx, double *drhody, double *drhodz, double *dsigmadx, double *dsigmady, double *dsigmadz, double *dRdx, double *dRdy, double *dRdz, double *drhodxx, double *drhodxy, double *drhodxz, double *drhodyy, double *drhodyz, double *drhodzz, double *dsigmadxx, double *dsigmadxy, double *dsigmadxz, double *dsigmadyy, double *dsigmadyz, double *dsigmadzz, double *dRdxx, double *dRdxy, double *dRdxz, double *dRdyy, double *dRdyz, double *dRdzz, double *chi, double *trK, double *dxx, double *gxy, double *gxz, double *dyy, double *gyz, double *dzz, double *Axx, double *Axy, double *Axz, double *Ayy, double *Ayz, double *Azz, double *Gamx, double *Gamy, double *Gamz, double *Lap, double *betax, double *betay, double *betaz, double *dtSfx, double *dtSfy, double *dtSfz, double *chi_rhs, double *trK_rhs, double *gxx_rhs, double *gxy_rhs, double *gxz_rhs, double *gyy_rhs, double *gyz_rhs, double *gzz_rhs, double *Axx_rhs, double *Axy_rhs, double *Axz_rhs, double *Ayy_rhs, double *Ayz_rhs, double *Azz_rhs, double *Gamx_rhs, double *Gamy_rhs, double *Gamz_rhs, double *Lap_rhs, double *betax_rhs, double *betay_rhs, double *betaz_rhs, double *dtSfx_rhs, double *dtSfy_rhs, double *dtSfz_rhs, double *rho, double *Sx, double *Sy, double *Sz, double *Sxx, double *Sxy, double *Sxz, double *Syy, double *Syz, double *Szz, double *Gamxxx, double *Gamxxy, double *Gamxxz, double *Gamxyy, double *Gamxyz, double *Gamxzz, double *Gamyxx, double *Gamyxy, double *Gamyxz, double *Gamyyy, double *Gamyyz, double *Gamyzz, double *Gamzxx, double *Gamzxy, double *Gamzxz, double *Gamzyy, double *Gamzyz, double *Gamzzz, double *Rxx, double *Rxy, double *Rxz, double *Ryy, double *Ryz, double *Rzz, double *ham_Res, double *movx_Res, double *movy_Res, double *movz_Res, double *Gmx_Res, double *Gmy_Res, double *Gmz_Res, int &Symmetry, int &Lev, double &eps, int &sst, int &co
-
-/**  main function */
-int gpu_rhs(int calledby, int mpi_rank, int *ex, double &T,
-            double *X, double *Y, double *Z,
-            double *chi, double *trK,
-            double *dxx, double *gxy, double *gxz, double *dyy, double *gyz, double *dzz,
-            double *Axx, double *Axy, double *Axz, double *Ayy, double *Ayz, double *Azz,
-            double *Gamx, double *Gamy, double *Gamz,
-            double *Lap, double *betax, double *betay, double *betaz,
-            double *dtSfx, double *dtSfy, double *dtSfz,
-            double *chi_rhs, double *trK_rhs,
-            double *gxx_rhs, double *gxy_rhs, double *gxz_rhs, double *gyy_rhs, double *gyz_rhs, double *gzz_rhs,
-            double *Axx_rhs, double *Axy_rhs, double *Axz_rhs, double *Ayy_rhs, double *Ayz_rhs, double *Azz_rhs,
-            double *Gamx_rhs, double *Gamy_rhs, double *Gamz_rhs,
-            double *Lap_rhs, double *betax_rhs, double *betay_rhs, double *betaz_rhs,
-            double *dtSfx_rhs, double *dtSfy_rhs, double *dtSfz_rhs,
-            double *rho, double *Sx, double *Sy, double *Sz, double *Sxx,
-            double *Sxy, double *Sxz, double *Syy, double *Syz, double *Szz,
-            double *Gamxxx, double *Gamxxy, double *Gamxxz, double *Gamxyy, double *Gamxyz, double *Gamxzz,
-            double *Gamyxx, double *Gamyxy, double *Gamyxz, double *Gamyyy, double *Gamyyz, double *Gamyzz,
-            double *Gamzxx, double *Gamzxy, double *Gamzxz, double *Gamzyy, double *Gamzyz, double *Gamzzz,
-            double *Rxx, double *Rxy, double *Rxz, double *Ryy, double *Ryz, double *Rzz,
-            double *ham_Res, double *movx_Res, double *movy_Res, double *movz_Res,
-            double *Gmx_Res, double *Gmy_Res, double *Gmz_Res,
-            int &Symmetry, int &Lev, double &eps, int &co);
-
-int gpu_rhs_ss(RHS_SS_PARA);
-
-#define Z4C_SS_PARA int calledby, int mpi_rank, int *ex, double &T, double *crho, double *sigma, double *R, double *X, double *Y, double *Z, double *drhodx, double *drhody, double *drhodz, double *dsigmadx, double *dsigmady, double *dsigmadz, double *dRdx, double *dRdy, double *dRdz, double *drhodxx, double *drhodxy, double *drhodxz, double *drhodyy, double *drhodyz, double *drhodzz, double *dsigmadxx, double *dsigmadxy, double *dsigmadxz, double *dsigmadyy, double *dsigmadyz, double *dsigmadzz, double *dRdxx, double *dRdxy, double *dRdxz, double *dRdyy, double *dRdyz, double *dRdzz, double *chi, double *trK, double *dxx, double *gxy, double *gxz, double *dyy, double *gyz, double *dzz, double *Axx, double *Axy, double *Axz, double *Ayy, double *Ayz, double *Azz, double *Gamx, double *Gamy, double *Gamz, double *Lap, double *betax, double *betay, double *betaz, double *dtSfx, double *dtSfy, double *dtSfz, double *TZ, double *chi_rhs, double *trK_rhs, double *gxx_rhs, double *gxy_rhs, double *gxz_rhs, double *gyy_rhs, double *gyz_rhs, double *gzz_rhs, double *Axx_rhs, double *Axy_rhs, double *Axz_rhs, double *Ayy_rhs, double *Ayz_rhs, double *Azz_rhs, double *Gamx_rhs, double *Gamy_rhs, double *Gamz_rhs, double *Lap_rhs, double *betax_rhs, double *betay_rhs, double *betaz_rhs, double *dtSfx_rhs, double *dtSfy_rhs, double *dtSfz_rhs, double *TZ_rhs, double *rho, double *Sx, double *Sy, double *Sz, double *Sxx, double *Sxy, double *Sxz, double *Syy, double *Syz, double *Szz, double *Gamxxx, double *Gamxxy, double *Gamxxz, double *Gamxyy, double *Gamxyz, double *Gamxzz, double *Gamyxx, double *Gamyxy, double *Gamyxz, double *Gamyyy, double *Gamyyz, double *Gamyzz, double *Gamzxx, double *Gamzxy, double *Gamzxz, double *Gamzyy, double *Gamzyz, double *Gamzzz, double *Rxx, double *Rxy, double *Rxz, double *Ryy, double *Ryz, double *Rzz, double *ham_Res, double *movx_Res, double *movy_Res, double *movz_Res, double *Gmx_Res, double *Gmy_Res, double *Gmz_Res, int &Symmetry, int &Lev, double &eps, int &sst, int &co
-
-int gpu_rhs_z4c_ss(Z4C_SS_PARA);
-
-#endif
+
+#ifndef BSSN_GPU_H_
+#define BSSN_GPU_H_
+#include "bssn_macro.h"
+#include "macrodef.fh"
+
+#define DEVICE_ID 0
+// #define DEVICE_ID_BY_MPI_RANK
+#define GRID_DIM 256
+#define BLOCK_DIM 128
+
+#define _FH2_(i, j, k) fh[(i) + (j) * _1D_SIZE[2] + (k) * _2D_SIZE[2]]
+#define _FH3_(i, j, k) fh[(i) + (j) * _1D_SIZE[3] + (k) * _2D_SIZE[3]]
+#define pow2(x) ((x) * (x))
+#define TimeBetween(a, b) ((b.tv_sec - a.tv_sec) + (b.tv_usec - a.tv_usec) / 1000000.0f)
+#define M_ metac.
+#define Mh_ meta->
+#define Ms_ metassc.
+#define Msh_ metass->
+
+// #define TIMING
+
+#define RHS_SS_PARA int calledby, int mpi_rank, int *ex, double &T, double *crho, double *sigma, double *R, double *X, double *Y, double *Z, double *drhodx, double *drhody, double *drhodz, double *dsigmadx, double *dsigmady, double *dsigmadz, double *dRdx, double *dRdy, double *dRdz, double *drhodxx, double *drhodxy, double *drhodxz, double *drhodyy, double *drhodyz, double *drhodzz, double *dsigmadxx, double *dsigmadxy, double *dsigmadxz, double *dsigmadyy, double *dsigmadyz, double *dsigmadzz, double *dRdxx, double *dRdxy, double *dRdxz, double *dRdyy, double *dRdyz, double *dRdzz, double *chi, double *trK, double *dxx, double *gxy, double *gxz, double *dyy, double *gyz, double *dzz, double *Axx, double *Axy, double *Axz, double *Ayy, double *Ayz, double *Azz, double *Gamx, double *Gamy, double *Gamz, double *Lap, double *betax, double *betay, double *betaz, double *dtSfx, double *dtSfy, double *dtSfz, double *chi_rhs, double *trK_rhs, double *gxx_rhs, double *gxy_rhs, double *gxz_rhs, double *gyy_rhs, double *gyz_rhs, double *gzz_rhs, double *Axx_rhs, double *Axy_rhs, double *Axz_rhs, double *Ayy_rhs, double *Ayz_rhs, double *Azz_rhs, double *Gamx_rhs, double *Gamy_rhs, double *Gamz_rhs, double *Lap_rhs, double *betax_rhs, double *betay_rhs, double *betaz_rhs, double *dtSfx_rhs, double *dtSfy_rhs, double *dtSfz_rhs, double *rho, double *Sx, double *Sy, double *Sz, double *Sxx, double *Sxy, double *Sxz, double *Syy, double *Syz, double *Szz, double *Gamxxx, double *Gamxxy, double *Gamxxz, double *Gamxyy, double *Gamxyz, double *Gamxzz, double *Gamyxx, double *Gamyxy, double *Gamyxz, double *Gamyyy, double *Gamyyz, double *Gamyzz, double *Gamzxx, double *Gamzxy, double *Gamzxz, double *Gamzyy, double *Gamzyz, double *Gamzzz, double *Rxx, double *Rxy, double *Rxz, double *Ryy, double *Ryz, double *Rzz, double *ham_Res, double *movx_Res, double *movy_Res, double *movz_Res, double *Gmx_Res, double *Gmy_Res, double *Gmz_Res, int &Symmetry, int &Lev, double &eps, int &sst, int &co
+
+/**  main function */
+int gpu_rhs(int calledby, int mpi_rank, int *ex, double &T,
+            double *X, double *Y, double *Z,
+
+            double *chi, double *trK,
+
+            double *dxx, double *gxy, double *gxz, double *dyy, double *gyz, double *dzz,
+
+            double *Axx, double *Axy, double *Axz, double *Ayy, double *Ayz, double *Azz,
+
+            double *Gamx, double *Gamy, double *Gamz,
+
+            double *Lap, double *betax, double *betay, double *betaz,
+
+            double *dtSfx, double *dtSfy, double *dtSfz,
+
+            double *chi_rhs, double *trK_rhs,
+
+            double *gxx_rhs, double *gxy_rhs, double *gxz_rhs, double *gyy_rhs, double *gyz_rhs, double *gzz_rhs,
+
+            double *Axx_rhs, double *Axy_rhs, double *Axz_rhs, double *Ayy_rhs, double *Ayz_rhs, double *Azz_rhs,
+
+            double *Gamx_rhs, double *Gamy_rhs, double *Gamz_rhs,
+
+            double *Lap_rhs, double *betax_rhs, double *betay_rhs, double *betaz_rhs,
+
+            double *dtSfx_rhs, double *dtSfy_rhs, double *dtSfz_rhs,
+
+            double *rho, double *Sx, double *Sy, double *Sz, double *Sxx,
+            double *Sxy, double *Sxz, double *Syy, double *Syz, double *Szz,
+
+            double *Gamxxx, double *Gamxxy, double *Gamxxz, double *Gamxyy, double *Gamxyz, double *Gamxzz,
+
+            double *Gamyxx, double *Gamyxy, double *Gamyxz, double *Gamyyy, double *Gamyyz, double *Gamyzz,
+
+            double *Gamzxx, double *Gamzxy, double *Gamzxz, double *Gamzyy, double *Gamzyz, double *Gamzzz,
+
+            double *Rxx, double *Rxy, double *Rxz, double *Ryy, double *Ryz, double *Rzz,
+
+            double *ham_Res, double *movx_Res, double *movy_Res, double *movz_Res,
+            double *Gmx_Res, double *Gmy_Res, double *Gmz_Res,
+            int &Symmetry, int &Lev, double &eps, int &co);
+
+int gpu_rhs_ss(RHS_SS_PARA);
+
+/** Init GPU side data in GPUMeta. */
+// void init_fluid_meta_gpu(GPUMeta *gpu_meta);
+
+#endif

AMSS_NCKU_source/bssn_gpu_class.h

@@ -0,0 +1,210 @@
+
+#ifndef BSSN_GPU_CLASS_H
+#define BSSN_GPU_CLASS_H
+
+#ifdef newc
+#include <iostream>
+#include <iomanip>
+#include <fstream>
+#include <cstdlib>
+#include <string>
+#include <cmath>
+using namespace std;
+#else
+#include <iostream.h>
+#include <iomanip.h>
+#include <fstream.h>
+#include <stdlib.h>
+#include <string.h>
+#include <math.h>
+#endif
+
+#include <mpi.h>
+
+#include "macrodef.h"
+#include "cgh.h"
+#include "ShellPatch.h"
+#include "misc.h"
+#include "var.h"
+#include "MyList.h"
+#include "monitor.h"
+#include "surface_integral.h"
+#include "checkpoint.h"
+
+// added by yangquan
+#include "bssn_macro.h"
+
+extern void setpbh(int iBHN, double **iPBH, double *iMass, int rBHN);
+
+class bssn_class
+{
+public:
+       // added by yangquan
+       //----------------------
+       int gpu_num_mynode;
+       int cpu_core_num_mynode;
+       int mpi_process_num_mynode;
+       int my_sequence_mynode;
+       int mynode_id;
+       int use_gpu;
+
+       virtual void Step_GPU(int lev, int YN);
+       virtual void Get_runtime_envirment();
+       // virtual void Step_OPENMP(int lev,int YN);
+       //----------------------
+
+       int ngfs;
+       int nprocs, myrank;
+       cgh *GH;
+       ShellPatch *SH;
+       double PhysTime;
+
+       int checkrun;
+       char checkfilename[50];
+       int Steps;
+       double StartTime, TotalTime;
+       double AnasTime, DumpTime, d2DumpTime, CheckTime;
+       double LastAnas, LastConsOut;
+       double Courant;
+       double numepss, numepsb, numepsh;
+       int Symmetry;
+       int maxl, decn;
+       double maxrex, drex;
+       int trfls, a_lev;
+
+       double dT;
+       double chitiny;
+
+       double **Porg0, **Porgbr, **Porg, **Porg1, **Porg_rhs;
+       int BH_num, BH_num_input;
+       double *Mass, *Pmom, *Spin;
+       double ADMMass;
+
+       var *phio, *trKo;
+       var *gxxo, *gxyo, *gxzo, *gyyo, *gyzo, *gzzo;
+       var *Axxo, *Axyo, *Axzo, *Ayyo, *Ayzo, *Azzo;
+       var *Gmxo, *Gmyo, *Gmzo;
+       var *Lapo, *Sfxo, *Sfyo, *Sfzo;
+       var *dtSfxo, *dtSfyo, *dtSfzo;
+
+       var *phi0, *trK0;
+       var *gxx0, *gxy0, *gxz0, *gyy0, *gyz0, *gzz0;
+       var *Axx0, *Axy0, *Axz0, *Ayy0, *Ayz0, *Azz0;
+       var *Gmx0, *Gmy0, *Gmz0;
+       var *Lap0, *Sfx0, *Sfy0, *Sfz0;
+       var *dtSfx0, *dtSfy0, *dtSfz0;
+
+       var *phi, *trK;
+       var *gxx, *gxy, *gxz, *gyy, *gyz, *gzz;
+       var *Axx, *Axy, *Axz, *Ayy, *Ayz, *Azz;
+       var *Gmx, *Gmy, *Gmz;
+       var *Lap, *Sfx, *Sfy, *Sfz;
+       var *dtSfx, *dtSfy, *dtSfz;
+
+       var *phi1, *trK1;
+       var *gxx1, *gxy1, *gxz1, *gyy1, *gyz1, *gzz1;
+       var *Axx1, *Axy1, *Axz1, *Ayy1, *Ayz1, *Azz1;
+       var *Gmx1, *Gmy1, *Gmz1;
+       var *Lap1, *Sfx1, *Sfy1, *Sfz1;
+       var *dtSfx1, *dtSfy1, *dtSfz1;
+
+       var *phi_rhs, *trK_rhs;
+       var *gxx_rhs, *gxy_rhs, *gxz_rhs, *gyy_rhs, *gyz_rhs, *gzz_rhs;
+       var *Axx_rhs, *Axy_rhs, *Axz_rhs, *Ayy_rhs, *Ayz_rhs, *Azz_rhs;
+       var *Gmx_rhs, *Gmy_rhs, *Gmz_rhs;
+       var *Lap_rhs, *Sfx_rhs, *Sfy_rhs, *Sfz_rhs;
+       var *dtSfx_rhs, *dtSfy_rhs, *dtSfz_rhs;
+
+       var *rho, *Sx, *Sy, *Sz, *Sxx, *Sxy, *Sxz, *Syy, *Syz, *Szz;
+
+       var *Gamxxx, *Gamxxy, *Gamxxz, *Gamxyy, *Gamxyz, *Gamxzz;
+       var *Gamyxx, *Gamyxy, *Gamyxz, *Gamyyy, *Gamyyz, *Gamyzz;
+       var *Gamzxx, *Gamzxy, *Gamzxz, *Gamzyy, *Gamzyz, *Gamzzz;
+
+       var *Rxx, *Rxy, *Rxz, *Ryy, *Ryz, *Rzz;
+
+       var *Rpsi4, *Ipsi4;
+       var *t1Rpsi4, *t1Ipsi4, *t2Rpsi4, *t2Ipsi4;
+
+       var *Cons_Ham, *Cons_Px, *Cons_Py, *Cons_Pz, *Cons_Gx, *Cons_Gy, *Cons_Gz;
+
+#ifdef Point_Psi4
+       var *phix, *phiy, *phiz;
+       var *trKx, *trKy, *trKz;
+       var *Axxx, *Axxy, *Axxz;
+       var *Axyx, *Axyy, *Axyz;
+       var *Axzx, *Axzy, *Axzz;
+       var *Ayyx, *Ayyy, *Ayyz;
+       var *Ayzx, *Ayzy, *Ayzz;
+       var *Azzx, *Azzy, *Azzz;
+#endif
+       // FIXME: uc = StateList, up = OldStateList, upp = SynchList_cor; so never touch these three data
+       MyList<var> *StateList, *SynchList_pre, *SynchList_cor, *RHSList;
+       MyList<var> *OldStateList, *DumpList;
+       MyList<var> *ConstraintList;
+
+       monitor *ErrorMonitor, *Psi4Monitor, *BHMonitor, *MAPMonitor;
+       monitor *ConVMonitor;
+       surface_integral *Waveshell;
+       checkpoint *CheckPoint;
+
+public:
+       bssn_class(double Couranti, double StartTimei, double TotalTimei, double DumpTimei, double d2DumpTimei, double CheckTimei, double AnasTimei,
+                  int Symmetryi, int checkruni, char *checkfilenamei, double numepssi, double numepsbi, double numepshi,
+                  int a_levi, int maxli, int decni, double maxrexi, double drexi);
+       ~bssn_class();
+
+       void Evolve(int Steps);
+       void RecursiveStep(int lev);
+#if (PSTR == 1)
+       void ParallelStep();
+       void SHStep();
+#endif
+       void RestrictProlong(int lev, int YN, bool BB, MyList<var> *SL, MyList<var> *OL, MyList<var> *corL);
+       void RestrictProlong_aux(int lev, int YN, bool BB, MyList<var> *SL, MyList<var> *OL, MyList<var> *corL);
+       void RestrictProlong(int lev, int YN, bool BB);
+       void ProlongRestrict(int lev, int YN, bool BB);
+       void Setup_Black_Hole_position();
+       void compute_Porg_rhs(double **BH_PS, double **BH_RHS, var *forx, var *fory, var *forz, int lev);
+       bool read_Pablo_file(int *ext, double *datain, char *filename);
+       void write_Pablo_file(int *ext, double xmin, double xmax, double ymin, double ymax, double zmin, double zmax,
+                             char *filename);
+       void AnalysisStuff(int lev, double dT_lev);
+       void Setup_KerrSchild();
+       void Enforce_algcon(int lev, int fg);
+
+       void testRestrict();
+       void testOutBd();
+
+       virtual void Setup_Initial_Data_Lousto();
+       virtual void Setup_Initial_Data_Cao();
+       virtual void Initialize();
+       virtual void Read_Ansorg();
+       virtual void Read_Pablo() {};
+       virtual void Compute_Psi4(int lev);
+       virtual void Step(int lev, int YN);
+       virtual void Interp_Constraint(bool infg);
+       virtual void Constraint_Out();
+       virtual void Compute_Constraint();
+
+#ifdef With_AHF
+protected:
+       MyList<var> *AHList, *AHDList, *GaugeList;
+       int AHfindevery;
+       double AHdumptime;
+       int *lastahdumpid, HN_num; // number of possible horizons
+       int *findeveryl;
+       double *xc, *yc, *zc, *xr, *yr, *zr;
+       bool *trigger;
+       double *dTT;
+       int *dumpid;
+
+public:
+       void AH_Prepare_derivatives();
+       bool AH_Interp_Points(MyList<var> *VarList,
+                             int NN, double **XX,
+                             double *Shellf, int Symmetryi);
+       void AH_Step_Find(int lev, double dT_lev);
+#endif
+};
+#endif /* BSSN_GPU_CLASS_H */

AMSS_NCKU_source/bssn_gpu_rhs_ss.cu

@@ -20,14 +20,12 @@ using namespace std;
 
 __device__ volatile unsigned int global_count = 0;
 
-#ifdef RESULT_CHECK
 void compare_result_gpu(int ftag1,double * datac,int data_num){
 	double * data = (double*)malloc(sizeof(double)*data_num);
 	cudaMemcpy(data, datac, data_num * sizeof(double), cudaMemcpyDeviceToHost);
 	compare_result(ftag1,data,data_num);
 	free(data);
 }
-#endif
 
 __global__ void sub_symmetry_bd_ss_partF(int ord, double * func, double *funcc)
 {
@@ -155,11 +153,11 @@ __global__ void sub_symmetry_bd_ss_partJ(int ord,double * func, double * funcc,d
 
 inline void sub_symmetry_bd_ss(int ord,double * func, double * funcc,double * SoA){
 	sub_symmetry_bd_ss_partF<<<GRID_DIM,BLOCK_DIM>>>(ord,func,funcc);
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 	sub_symmetry_bd_ss_partI<<<GRID_DIM,BLOCK_DIM>>>(ord,func,funcc,SoA[0]);
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 	sub_symmetry_bd_ss_partJ<<<GRID_DIM,BLOCK_DIM>>>(ord,func,funcc,SoA[1]);
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 }
 
 __global__ void sub_fderivs_shc_part1(double *fx,double *fy,double *fz){
@@ -249,13 +247,13 @@ inline void sub_fderivs_shc(int& sst,double * f,double * fh,double *fx,double *f
 	//cudaMemset(Msh_ gy,0,h_3D_SIZE[0] * sizeof(double));
 	//cudaMemset(Msh_ gz,0,h_3D_SIZE[0] * sizeof(double));
 	sub_symmetry_bd_ss(2,f,fh,SoA1);
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 			//compare_result_gpu(0,fh,h_3D_SIZE[2]);	
 	sub_fderivs_sh<<<GRID_DIM,BLOCK_DIM>>>(fh,Msh_ gx,Msh_ gy,Msh_ gz);
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 	
 	sub_fderivs_shc_part1<<<GRID_DIM,BLOCK_DIM>>>(fx,fy,fz);
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 			//compare_result_gpu(1,fx,h_3D_SIZE[0]);
 			//compare_result_gpu(2,fy,h_3D_SIZE[0]);
 			//compare_result_gpu(3,fz,h_3D_SIZE[0]);
@@ -453,17 +451,17 @@ inline void sub_fdderivs_shc(int& sst,double * f,double * fh,
 	
 	//fderivs_sh
 	sub_symmetry_bd_ss(2,f,fh,SoA1);
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 			//compare_result_gpu(1,fh,h_3D_SIZE[2]);	
 	sub_fderivs_sh<<<GRID_DIM,BLOCK_DIM>>>(fh,Msh_ gx,Msh_ gy,Msh_ gz);
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 	
 	//fdderivs_sh
 	sub_symmetry_bd_ss(2,f,fh,SoA1);
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 			//compare_result_gpu(21,fh,h_3D_SIZE[2]);
 	sub_fdderivs_sh<<<GRID_DIM,BLOCK_DIM>>>(fh,Msh_ gxx,Msh_ gxy,Msh_ gxz,Msh_ gyy,Msh_ gyz,Msh_ gzz);
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 			/*compare_result_gpu(11,Msh_ gx,h_3D_SIZE[0]);
 			compare_result_gpu(12,Msh_ gy,h_3D_SIZE[0]);
 			compare_result_gpu(13,Msh_ gz,h_3D_SIZE[0]);
@@ -474,7 +472,7 @@ inline void sub_fdderivs_shc(int& sst,double * f,double * fh,
 			compare_result_gpu(5,Msh_ gyz,h_3D_SIZE[0]);
 			compare_result_gpu(6,Msh_ gzz,h_3D_SIZE[0]);*/
 	sub_fdderivs_shc_part1<<<GRID_DIM,BLOCK_DIM>>>(fxx,fxy,fxz,fyy,fyz,fzz);
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 			/*compare_result_gpu(1,fxx,h_3D_SIZE[0]);
 			compare_result_gpu(2,fxy,h_3D_SIZE[0]);
 			compare_result_gpu(3,fxz,h_3D_SIZE[0]);
@@ -498,9 +496,9 @@ __global__ void computeRicci_ss_part1(double * dst)
  inline void computeRicci_ss(int &sst,double * src,double* dst,double * SoA, Meta* meta)
 {
 	sub_fdderivs_shc(sst,src,Mh_ fh,Mh_ fxx,Mh_ fxy,Mh_ fxz,Mh_ fyy,Mh_ fyz,Mh_ fzz,SoA);
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 	computeRicci_ss_part1<<<GRID_DIM,BLOCK_DIM>>>(dst);
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 	
 }
 __global__ void sub_lopsided_ss_part1(double * dst)
@@ -518,9 +516,9 @@ __global__ void sub_lopsided_ss_part1(double * dst)
 inline void sub_lopsided_ss(int& sst,double *src,double* dst,double *SoA)
 {
 		sub_fderivs_shc(sst,src,Mh_ fh,Mh_ fxx,Mh_ fxy,Mh_ fxz,SoA);
-		cudaDeviceSynchronize();
+		cudaThreadSynchronize();
 		sub_lopsided_ss_part1<<<GRID_DIM,BLOCK_DIM>>>(dst);
-		cudaDeviceSynchronize();
+		cudaThreadSynchronize();
 }
 
 __global__ void sub_kodis_sh_part1(double *f,double *fh,double *f_rhs)
@@ -592,11 +590,11 @@ inline void sub_kodis_ss(int &sst,double *f,double *fh,double *f_rhs,double *SoA
 	}
 			//compare_result_gpu(10,f,h_3D_SIZE[0]);
 	sub_symmetry_bd_ss(3,f,fh,SoA1);
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 			//compare_result_gpu(0,fh,h_3D_SIZE[3]);
 			
 	sub_kodis_sh_part1<<<GRID_DIM,BLOCK_DIM>>>(f,fh,f_rhs);
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 			//compare_result_gpu(1,f_rhs,h_3D_SIZE[0]);
 }
 
@@ -1701,7 +1699,7 @@ void destroy_meta(Meta *meta,Metass *metass)
 	if(Msh_ gzz) cudaFree(Msh_ gzz);
 	
 #if (GAUGE == 2 || GAUGE == 3 || GAUGE == 4 || GAUGE == 5 || GAUGE == 6 || GAUGE == 7)
-	if(Mh_ reta) cudaFree(Mh_ reta);
+	if(Mh_ reta) CUDA_SAFE_CALL(cudaFree(Mh_ reta));
 
 #endif	
 	
@@ -1897,7 +1895,7 @@ int gpu_rhs_ss(RHS_SS_PARA)
 
 	//1.2 local Data
 	cudaMalloc((void**)&(Mh_ gxx), matrix_size * sizeof(double));
-	cudaMalloc((void**)&(Mh_ gyy), matrix_size * sizeof(double));
+	CUDA_SAFE_CALL( cudaMalloc((void**)&(Mh_ gyy), matrix_size * sizeof(double)));
 	cudaMalloc((void**)&(Mh_ gzz), matrix_size * sizeof(double));
 	cudaMalloc((void**)&(Mh_ chix), matrix_size * sizeof(double));
 	cudaMalloc((void**)&(Mh_ chiy), matrix_size * sizeof(double));
@@ -2162,7 +2160,7 @@ int gpu_rhs_ss(RHS_SS_PARA)
 	
 	double tmp_con2 = 1/Mass[0] - tmp_con;
 	cudaMemcpyToSymbol(C1, &tmp_con2, sizeof(double));
-	tmp_con2 = 1/Mass[1] - tmp_con;
+	double tmp_con2 = 1/Mass[1] - tmp_con;
 	cudaMemcpyToSymbol(C2, &tmp_con2, sizeof(double));
 		
 
@@ -2235,7 +2233,7 @@ int gpu_rhs_ss(RHS_SS_PARA)
   		if((sst == 2 || sst == 4) && abs[1] < dYh) 
   		{
   			ijkmin_h[1] = -2; 
-  			ijkmin3_h[1] = -3;
+  			ijkmin_h[1] = -3;
   		}
   		if((sst == 3 || sst == 5) && abs_Y_ex2 < dYh)
   		{
@@ -2289,13 +2287,13 @@ int gpu_rhs_ss(RHS_SS_PARA)
 
 
 #ifdef TIMING1
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 	gettimeofday(&tv2, NULL);
    	cout<<"TIME USED"<<TimeBetween(tv1, tv2)<<endl; 
 #endif	
 	//cout<<"GPU meta data ready.\n";
 	
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 
 
 //-------------get device info-------------------------------------
@@ -2308,7 +2306,7 @@ int gpu_rhs_ss(RHS_SS_PARA)
 	//sub_enforce_ga(matrix_size);
 	//4.1-----compute rhs---------
 	compute_rhs_ss_part1<<<GRID_DIM,BLOCK_DIM>>>();
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 
 	sub_fderivs_shc(sst,Mh_ betax,Mh_ fh,Mh_ betaxx,Mh_ betaxy,Mh_ betaxz,ass);
 	sub_fderivs_shc(sst,Mh_ betay,Mh_ fh,Mh_ betayx,Mh_ betayy,Mh_ betayz,sas);
@@ -2324,7 +2322,7 @@ int gpu_rhs_ss(RHS_SS_PARA)
 	sub_fderivs_shc(sst,Mh_ gyz,Mh_ fh,Mh_ gyzx,Mh_ gyzy,Mh_ gyzz, saa);
 	
 	compute_rhs_ss_part2<<<GRID_DIM,BLOCK_DIM>>>();
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 	
 	sub_fdderivs_shc(sst,Mh_ betax,Mh_ fh,Mh_ gxxx,Mh_ gxyx,Mh_ gxzx,Mh_ gyyx,Mh_ gyzx,Mh_ gzzx,ass);
 	sub_fdderivs_shc(sst,Mh_ betay,Mh_ fh,Mh_ gxxy,Mh_ gxyy,Mh_ gxzy,Mh_ gyyy,Mh_ gyzy,Mh_ gzzy,sas);
@@ -2334,7 +2332,7 @@ int gpu_rhs_ss(RHS_SS_PARA)
 	sub_fderivs_shc( sst,Mh_ Gamz, Mh_ fh,Mh_ Gamzx, Mh_ Gamzy, Mh_ Gamzz,ssa);
 	
 	compute_rhs_ss_part3<<<GRID_DIM,BLOCK_DIM>>>();
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 	
 	computeRicci_ss(sst,Mh_ dxx,Mh_ Rxx,sss, meta);
 	computeRicci_ss(sst,Mh_ dyy,Mh_ Ryy,sss, meta);
@@ -2342,25 +2340,25 @@ int gpu_rhs_ss(RHS_SS_PARA)
 	computeRicci_ss(sst,Mh_ gxy,Mh_ Rxy,aas, meta);
 	computeRicci_ss(sst,Mh_ gxz,Mh_ Rxz,asa, meta);
 	computeRicci_ss(sst,Mh_ gyz,Mh_ Ryz,saa, meta);
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 	
 	compute_rhs_ss_part4<<<GRID_DIM,BLOCK_DIM>>>();
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 	
 	sub_fdderivs_shc(sst,Mh_ chi,Mh_ fh,Mh_ fxx,Mh_ fxy,Mh_ fxz,Mh_ fyy,Mh_ fyz,Mh_ fzz,sss);
 	
-	//cudaDeviceSynchronize();
+	//cudaThreadSynchronize();
 	//compare_result_gpu(0,Mh_ chi,h_3D_SIZE[0]);
 	//compare_result_gpu(1,Mh_ chi,h_3D_SIZE[0]);
 	//compare_result_gpu(2,Mh_ fyz,h_3D_SIZE[0]);
 	
 	compute_rhs_ss_part5<<<GRID_DIM,BLOCK_DIM>>>();
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 	
 	sub_fdderivs_shc(sst,Mh_ Lap,Mh_ fh,Mh_ fxx,Mh_ fxy,Mh_ fxz,Mh_ fyy,Mh_ fyz,Mh_ fzz,sss);
 	
 	compute_rhs_ss_part6<<<GRID_DIM,BLOCK_DIM>>>();
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 	
 #if (GAUGE == 2 || GAUGE == 3 || GAUGE == 4 || GAUGE == 5)
 	sub_fderivs_shc(sst,Mh_ chi,Mh_ fh, Mh_ dtSfx_rhs, Mh_ dtSfy_rhs, Mh_ dtSfz_rhs,sss);
@@ -2425,7 +2423,7 @@ int gpu_rhs_ss(RHS_SS_PARA)
 	}
 	if(co == 0){
 		compute_rhs_ss_part7<<<GRID_DIM,BLOCK_DIM>>>();
-		cudaDeviceSynchronize();
+		cudaThreadSynchronize();
 
 		sub_fderivs_shc(sst,Mh_ Axx,Mh_ fh,Mh_ gxxx,Mh_ gxxy,Mh_ gxxz,sss);
 		sub_fderivs_shc(sst,Mh_ Axy,Mh_ fh,Mh_ gxyx,Mh_ gxyy,Mh_ gxyz,aas);
@@ -2434,7 +2432,7 @@ int gpu_rhs_ss(RHS_SS_PARA)
 		sub_fderivs_shc(sst,Mh_ Ayz,Mh_ fh,Mh_ gyzx,Mh_ gyzy,Mh_ gyzz,saa);
 		sub_fderivs_shc(sst,Mh_ Azz,Mh_ fh,Mh_ gzzx,Mh_ gzzy,Mh_ gzzz,sss);
 		compute_rhs_ss_part8<<<GRID_DIM,BLOCK_DIM>>>();
-		cudaDeviceSynchronize();
+		cudaThreadSynchronize();
 	}
 	
 #if (ABV == 1)
@@ -2514,7 +2512,7 @@ int gpu_rhs_ss(RHS_SS_PARA)
 	//test kodis
 	//sub_kodis_sh(sst,Msh_ drhodx,Mh_ fh2,Msh_ drhody,sss);
 #ifdef TIMING
-	cudaDeviceSynchronize();
+	cudaThreadSynchronize();
 	gettimeofday(&tv2, NULL);
    	cout<<"MPI rank is: "<<mpi_rank<<" GPU TIME is"<<TimeBetween(tv1, tv2)<<" (s)."<<endl; 
 #endif
@@ -2524,55 +2522,4 @@ int gpu_rhs_ss(RHS_SS_PARA)
 	return 0;//TODO return
 }
 
-#if (ABEtype == 2)
-// Z4C Shell GPU: calls BSSN gpu_rhs_ss with trKd=trK+2*TZ, then applies
-// TZ_rhs = alpn1*Hcon/2 and constraint damping on CPU.
-int gpu_rhs_z4c_ss(Z4C_SS_PARA)
-{
-    int matrix_size = ex[0] * ex[1] * ex[2];
-    double k1 = 0.02, k2 = 0.0;
-
-    double *trKd_host = new double[matrix_size];
-    for (int _i = 0; _i < matrix_size; _i++)
-        trKd_host[_i] = trK[_i] + 2.0 * TZ[_i];
-
-    int result = gpu_rhs_ss(calledby, mpi_rank,
-        ex, T, crho, sigma, R, X, Y, Z,
-        drhodx, drhody, drhodz, dsigmadx, dsigmady, dsigmadz,
-        dRdx, dRdy, dRdz,
-        drhodxx, drhodxy, drhodxz, drhodyy, drhodyz, drhodzz,
-        dsigmadxx, dsigmadxy, dsigmadxz, dsigmadyy, dsigmadyz, dsigmadzz,
-        dRdxx, dRdxy, dRdxz, dRdyy, dRdyz, dRdzz,
-        chi, trKd_host, dxx, gxy, gxz, dyy, gyz, dzz,
-        Axx, Axy, Axz, Ayy, Ayz, Azz,
-        Gamx, Gamy, Gamz,
-        Lap, betax, betay, betaz,
-        dtSfx, dtSfy, dtSfz,
-        chi_rhs, trK_rhs,
-        gxx_rhs, gxy_rhs, gxz_rhs, gyy_rhs, gyz_rhs, gzz_rhs,
-        Axx_rhs, Axy_rhs, Axz_rhs, Ayy_rhs, Ayz_rhs, Azz_rhs,
-        Gamx_rhs, Gamy_rhs, Gamz_rhs,
-        Lap_rhs, betax_rhs, betay_rhs, betaz_rhs,
-        dtSfx_rhs, dtSfy_rhs, dtSfz_rhs,
-        rho, Sx, Sy, Sz, Sxx, Sxy, Sxz, Syy, Syz, Szz,
-        Gamxxx, Gamxxy, Gamxxz, Gamxyy, Gamxyz, Gamxzz,
-        Gamyxx, Gamyxy, Gamyxz, Gamyyy, Gamyyz, Gamyzz,
-        Gamzxx, Gamzxy, Gamzxz, Gamzyy, Gamzyz, Gamzzz,
-        Rxx, Rxy, Rxz, Ryy, Ryz, Rzz,
-        ham_Res, movx_Res, movy_Res, movz_Res,
-        Gmx_Res, Gmy_Res, Gmz_Res,
-        Symmetry, Lev, eps, sst, co);
-    delete[] trKd_host;
-    if (result != 0) return result;
-
-    for (int _i = 0; _i < matrix_size; _i++) {
-        double alp = Lap[_i] + 1.0;
-        TZ_rhs[_i] = alp * ham_Res[_i] * 0.5;
-        TZ_rhs[_i]  -= alp * (2.0 + k2) * k1 * TZ[_i];
-        trK_rhs[_i] += alp * k1 * (1.0 - k2) * TZ[_i];
-    }
-    return 0;
-}
-#endif // ABEtype == 2
-
 #endif //WithShell

AMSS_NCKU_source/bssn_rhs.h

@@ -63,13 +63,34 @@ extern "C"
                                double *, double *, double *, double *, double *, double *,                                         // Christoffel
                                double *, double *, double *, double *, double *, double *,                                         // Christoffel
                                double *, double *, double *, double *, double *, double *,                                         // Ricci
-                               double *, double *, double *, double *, double *, double *, double *,                               // constraint violation
-                               int &, int &, double &, int &);
-}
-
-extern "C"
-{
-        int f_compute_rhs_bssn_ss(int *, double &, double *, double *, double *,                                                      // ex,T,rho,sigma,R
+                               double *, double *, double *, double *, double *, double *, double *,                               // constraint violation
+                               int &, int &, double &, int &);
+}
+
+int f_compute_rhs_bssn_escalar_c(int *, double &, double *, double *, double *,                                                      // ex,T,X,Y,Z
+                                 double *, double *,                                                                                 // chi, trK
+                                 double *, double *, double *, double *, double *, double *,                                         // gij
+                                 double *, double *, double *, double *, double *, double *,                                         // Aij
+                                 double *, double *, double *,                                                                       // Gam
+                                 double *, double *, double *, double *, double *, double *, double *,                               // Gauge
+                                 double *, double *,                                                                                 // Sphi, Spi
+                                 double *, double *,                                                                                 // chi, trK
+                                 double *, double *, double *, double *, double *, double *,                                         // gij
+                                 double *, double *, double *, double *, double *, double *,                                         // Aij
+                                 double *, double *, double *,                                                                       // Gam
+                                 double *, double *, double *, double *, double *, double *, double *,                               // Gauge
+                                 double *, double *,                                                                                 // Sphi, Spi
+                                 double *, double *, double *, double *, double *, double *, double *, double *, double *, double *, // stress-energy
+                                 double *, double *, double *, double *, double *, double *,                                         // Christoffel
+                                 double *, double *, double *, double *, double *, double *,                                         // Christoffel
+                                 double *, double *, double *, double *, double *, double *,                                         // Christoffel
+                                 double *, double *, double *, double *, double *, double *,                                         // Ricci
+                                 double *, double *, double *, double *, double *, double *, double *,                               // constraint violation
+                                 int &, int &, double &, int &);
+
+extern "C"
+{
+        int f_compute_rhs_bssn_ss(int *, double &, double *, double *, double *,                                                      // ex,T,rho,sigma,R
                                   double *, double *, double *,                                                                       // X,Y,Z
                                   double *, double *, double *,                                                                       // drhodx,drhody,drhodz
                                   double *, double *, double *,                                                                       // dsigmadx,dsigmady,dsigmadz
@@ -96,10 +117,10 @@ extern "C"
                                   int &, int &, double &, int &, int &);
 }
 
-extern "C"
-{
-        int f_compute_rhs_bssn_escalar(int *, double &, double *, double *, double *,                                                      // ex,T,X,Y,Z
-                                       double *, double *,                                                                                 // chi, trK
+extern "C"
+{
+        int f_compute_rhs_bssn_escalar(int *, double &, double *, double *, double *,                                                      // ex,T,X,Y,Z
+                                       double *, double *,                                                                                 // chi, trK
                                        double *, double *, double *, double *, double *, double *,                                         // gij
                                        double *, double *, double *, double *, double *, double *,                                         // Aij
                                        double *, double *, double *,                                                                       // Gam
@@ -116,14 +137,14 @@ extern "C"
                                        double *, double *, double *, double *, double *, double *,                                         // Christoffel
                                        double *, double *, double *, double *, double *, double *,                                         // Christoffel
                                        double *, double *, double *, double *, double *, double *,                                         // Ricci
-                                       double *, double *, double *, double *, double *, double *, double *,                               // constraint violation
-                                       int &, int &, double &, int &);
-}
-
-extern "C"
-{
-        int f_compute_rhs_bssn_escalar_ss(int *, double &, double *, double *, double *,                                                      // ex,T,rho,sigma,R
-                                          double *, double *, double *,                                                                       // X,Y,Z
+                                       double *, double *, double *, double *, double *, double *, double *,                               // constraint violation
+                                       int &, int &, double &, int &);
+}
+
+extern "C"
+{
+        int f_compute_rhs_bssn_escalar_ss(int *, double &, double *, double *, double *,                                                      // ex,T,rho,sigma,R
+                                          double *, double *, double *,                                                                       // X,Y,Z
                                           double *, double *, double *,                                                                       // drhodx,drhody,drhodz
                                           double *, double *, double *,                                                                       // dsigmadx,dsigmady,dsigmadz
                                           double *, double *, double *,                                                                       // dRdx,dRdy,dRdz

AMSS_NCKU_source/bssn_rhs_c.C

@@ -1098,12 +1098,12 @@ int f_compute_rhs_bssn(int *ex, double &T,
             betaz_rhs[i] = FF * dtSfz[i];
 
             reta[i] =
-                gupxx[i] * chix[i] * chix[i]
-                + gupyy[i] * chiy[i] * chiy[i]
-                + gupzz[i] * chiz[i] * chiz[i]
-                + TWO * ( gupxy[i] * chix[i] * chiy[i]
-                        + gupxz[i] * chix[i] * chiz[i]
-                        + gupyz[i] * chiy[i] * chiz[i] );
+                gupxx[i] * dtSfx_rhs[i] * dtSfx_rhs[i]
+                + gupyy[i] * dtSfy_rhs[i] * dtSfy_rhs[i]
+                + gupzz[i] * dtSfz_rhs[i] * dtSfz_rhs[i]
+                + TWO * ( gupxy[i] * dtSfx_rhs[i] * dtSfy_rhs[i]
+                        + gupxz[i] * dtSfx_rhs[i] * dtSfz_rhs[i]
+                        + gupyz[i] * dtSfy_rhs[i] * dtSfz_rhs[i] );
 
             #if (GAUGE == 2)
             reta[i] = 1.31 / 2.0 * sqrt( reta[i] / chin1[i] ) / pow( (ONE - sqrt(chin1[i])), 2.0 );
@@ -1116,12 +1116,12 @@ int f_compute_rhs_bssn(int *ex, double &T,
             dtSfz_rhs[i] = Gamz_rhs[i] - reta[i] * dtSfz[i];
             #elif (GAUGE == 4 || GAUGE == 5)
             reta[i] =
-                gupxx[i] * chix[i] * chix[i]
-                + gupyy[i] * chiy[i] * chiy[i]
-                + gupzz[i] * chiz[i] * chiz[i]
-                + TWO * ( gupxy[i] * chix[i] * chiy[i]
-                        + gupxz[i] * chix[i] * chiz[i]
-                        + gupyz[i] * chiy[i] * chiz[i] );
+                gupxx[i] * dtSfx_rhs[i] * dtSfx_rhs[i]
+                + gupyy[i] * dtSfy_rhs[i] * dtSfy_rhs[i]
+                + gupzz[i] * dtSfz_rhs[i] * dtSfz_rhs[i]
+                + TWO * ( gupxy[i] * dtSfx_rhs[i] * dtSfy_rhs[i]
+                        + gupxz[i] * dtSfx_rhs[i] * dtSfz_rhs[i]
+                        + gupyz[i] * dtSfy_rhs[i] * dtSfz_rhs[i] );
 
             #if (GAUGE == 4)
             reta[i] = 1.31 / 2.0 * sqrt( reta[i] / chin1[i] ) / pow( (ONE - sqrt(chin1[i])), 2.0 );

AMSS_NCKU_source/bssn_rhs_cuda.h

@@ -1,413 +0,0 @@
-#ifndef BSSN_RHS_CUDA_H
-#define BSSN_RHS_CUDA_H
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-enum {
-    BSSN_CUDA_STATE_COUNT = 24,
-    BSSN_ESCALAR_CUDA_STATE_COUNT = 26,
-    BSSN_EM_CUDA_STATE_COUNT = 32,
-    BSSN_EM_CUDA_SOURCE_COUNT = 4,
-    BSSN_CUDA_MATTER_COUNT = 10
-};
-
-int f_compute_rhs_bssn(int *ex, double &T,
-                       double *X, double *Y, double *Z,
-                       double *chi, double *trK,
-                       double *dxx, double *gxy, double *gxz, double *dyy, double *gyz, double *dzz,
-                       double *Axx, double *Axy, double *Axz, double *Ayy, double *Ayz, double *Azz,
-                       double *Gamx, double *Gamy, double *Gamz,
-                       double *Lap, double *betax, double *betay, double *betaz,
-                       double *dtSfx, double *dtSfy, double *dtSfz,
-                       double *chi_rhs, double *trK_rhs,
-                       double *gxx_rhs, double *gxy_rhs, double *gxz_rhs, double *gyy_rhs, double *gyz_rhs, double *gzz_rhs,
-                       double *Axx_rhs, double *Axy_rhs, double *Axz_rhs, double *Ayy_rhs, double *Ayz_rhs, double *Azz_rhs,
-                       double *Gamx_rhs, double *Gamy_rhs, double *Gamz_rhs,
-                       double *Lap_rhs, double *betax_rhs, double *betay_rhs, double *betaz_rhs,
-                       double *dtSfx_rhs, double *dtSfy_rhs, double *dtSfz_rhs,
-                       double *rho, double *Sx, double *Sy, double *Sz,
-                       double *Sxx, double *Sxy, double *Sxz, double *Syy, double *Syz, double *Szz,
-                       double *Gamxxx, double *Gamxxy, double *Gamxxz, double *Gamxyy, double *Gamxyz, double *Gamxzz,
-                       double *Gamyxx, double *Gamyxy, double *Gamyxz, double *Gamyyy, double *Gamyyz, double *Gamyzz,
-                       double *Gamzxx, double *Gamzxy, double *Gamzxz, double *Gamzyy, double *Gamzyz, double *Gamzzz,
-                       double *Rxx, double *Rxy, double *Rxz, double *Ryy, double *Ryz, double *Rzz,
-                       double *ham_Res, double *movx_Res, double *movy_Res, double *movz_Res,
-                       double *Gmx_Res, double *Gmy_Res, double *Gmz_Res,
-                       int &Symmetry, int &Lev, double &eps, int &co);
-
-int bssn_cuda_rk4_substep(void *block_tag,
-                          int *ex, double *X, double *Y, double *Z,
-                          double **state_host_in,
-                          double **state_host_out,
-                          double **matter_host,
-                          const double *propspeed,
-                          const double *soa_flat,
-                          const double *bbox,
-                          double &dT,
-                          double &T,
-                          int &RK4,
-                          int &apply_bam_bc,
-                          int &Symmetry,
-                          int &Lev,
-                          double &eps,
-                          int &co,
-                          int &use_zero_matter,
-                          int &keep_resident_state,
-                          int &apply_enforce_ga,
-                          double &chitiny);
-
-int bssn_escalar_cuda_rk4_substep(void *block_tag,
-                                  int *ex, double *X, double *Y, double *Z,
-                                  double **state_host_in,
-                                  double **state_host_out,
-                                  const double *propspeed,
-                                  const double *soa_flat,
-                                  const double *bbox,
-                                  double &dT,
-                                  double &T,
-                                  int &RK4,
-                                  int &apply_bam_bc,
-                                  int &Symmetry,
-	                                  int &Lev,
-	                                  double &eps,
-	                                  int &co,
-	                                  int &keep_resident_state,
-	                                  int &apply_enforce_ga,
-	                                  double &chitiny);
-
-int bssn_escalar_cuda_compute_constraints(int *ex, double *X, double *Y, double *Z,
-                                          double **state_host_in,
-                                          double **constraint_host_out,
-                                          int &Symmetry,
-                                          int &Lev,
-                                          double &eps);
-
-int bssn_em_cuda_rk4_substep(void *block_tag,
-                             int *ex, double *X, double *Y, double *Z,
-                             double **state_host_in,
-                             double **state_host_out,
-                             double **source_host,
-                             const double *propspeed,
-                             const double *soa_flat,
-                             const double *bbox,
-                             double &dT,
-                             double &T,
-                             int &RK4,
-                             int &apply_bam_bc,
-                             int &Symmetry,
-                             int &Lev,
-                             double &eps,
-                             int &co,
-                             int &keep_resident_state,
-                             int &apply_enforce_ga,
-                             double &chitiny);
-
-int bssn_em_cuda_resident_zero_fast_state(void *block_tag);
-
-int bssn_cuda_copy_state_region_to_host(void *block_tag,
-                                        int state_index,
-                                        double *host_state,
-                                        int *ex,
-                                        int i0, int j0, int k0,
-                                        int sx, int sy, int sz);
-
-int bssn_cuda_copy_state_region_from_host(void *block_tag,
-                                          int state_index,
-                                          double *host_state,
-                                          int *ex,
-                                          int i0, int j0, int k0,
-                                          int sx, int sy, int sz);
-
-int bssn_cuda_download_resident_state(void *block_tag,
-                                      int *ex,
-                                      double **state_host_out);
-
-int bssn_escalar_cuda_download_resident_state(void *block_tag,
-                                              int *ex,
-                                              double **state_host_out);
-
-int bssn_cuda_upload_resident_state_count(void *block_tag,
-                                          int *ex,
-                                          double **state_host_in,
-                                          int state_count);
-
-int bssn_escalar_cuda_upload_resident_state(void *block_tag,
-                                            int *ex,
-                                            double **state_host_in);
-
-int bssn_cuda_keep_only_resident_state_count(void *block_tag,
-                                             int *ex,
-                                             double **state_host_key,
-                                             int state_count);
-
-int bssn_escalar_cuda_keep_only_resident_state(void *block_tag,
-                                               int *ex,
-                                               double **state_host_key);
-
-int bssn_cuda_download_resident_state_count_if_present(void *block_tag,
-                                                       int *ex,
-                                                       double **state_host_out,
-                                                       int state_count);
-
-int bssn_cuda_download_resident_state_if_present(void *block_tag,
-                                                int *ex,
-                                                double **state_host_out);
-
-int bssn_cuda_download_constraint_outputs(int *ex,
-                                          double **constraint_host_out);
-
-int bssn_cuda_pack_state_region_to_host_buffer(void *block_tag,
-                                               int state_index,
-                                               double *host_buffer,
-                                               int *ex,
-                                               int i0, int j0, int k0,
-                                               int sx, int sy, int sz);
-
-int bssn_cuda_interp_state_point3(void *block_tag,
-                                  int *ex,
-                                  int state0,
-                                  int state1,
-                                  int state2,
-                                  double x0,
-                                  double y0,
-                                  double z0,
-                                  double dx,
-                                  double dy,
-                                  double dz,
-                                  double px,
-                                  double py,
-                                  double pz,
-                                  int ordn,
-                                  int symmetry,
-                                  double **state_host_key,
-                                  const double *soa3,
-                                  double *out3);
-
-int bssn_cuda_interp_host_two_fields(void *block_tag,
-                                     int *ex,
-                                     double *field0,
-                                     double *field1,
-                                     double x0,
-                                     double y0,
-                                     double z0,
-                                     double dx,
-                                     double dy,
-                                     double dz,
-                                     const double *px,
-                                     const double *py,
-                                     const double *pz,
-                                     int npoints,
-                                     int ordn,
-                                     int symmetry,
-                                     const double *soa6,
-                                     double *out_interleaved);
-
-int bssn_cuda_unpack_state_region_from_host_buffer(void *block_tag,
-                                                   int state_index,
-                                                   double *host_buffer,
-                                                   int *ex,
-                                                   int i0, int j0, int k0,
-                                                   int sx, int sy, int sz);
-
-int bssn_cuda_unpack_state_region_from_host_buffer_for_host_views(void *block_tag,
-                                                                  double **state_host_key,
-                                                                  int state_count,
-                                                                  int state_index,
-                                                                  double *host_buffer,
-                                                                  int *ex,
-                                                                  int i0, int j0, int k0,
-                                                                  int sx, int sy, int sz);
-
-int bssn_cuda_pack_state_batch_to_host_buffer(void *block_tag,
-                                              int state_count,
-                                              double *host_buffer,
-                                              int *ex,
-                                              int i0, int j0, int k0,
-                                              int sx, int sy, int sz);
-
-int bssn_cuda_pack_state_batch_to_host_buffer_for_host_views(void *block_tag,
-                                                             double **state_host_key,
-                                                             int state_count,
-                                                             double *host_buffer,
-                                                             int *ex,
-                                                             int i0, int j0, int k0,
-                                                             int sx, int sy, int sz);
-
-int bssn_cuda_unpack_state_batch_from_host_buffer(void *block_tag,
-                                                  int state_count,
-                                                  double *host_buffer,
-                                                  int *ex,
-                                                  int i0, int j0, int k0,
-                                                  int sx, int sy, int sz);
-
-int bssn_cuda_unpack_state_batch_from_host_buffer_for_host_views(void *block_tag,
-                                                                 double **state_host_key,
-                                                                 int state_count,
-                                                                 double *host_buffer,
-                                                                 int *ex,
-                                                                 int i0, int j0, int k0,
-                                                                 int sx, int sy, int sz);
-
-int bssn_cuda_pack_state_batch_to_device_buffer(void *block_tag,
-                                                int state_count,
-                                                double *device_buffer,
-                                                int *ex,
-                                                int i0, int j0, int k0,
-                                                int sx, int sy, int sz);
-
-int bssn_cuda_pack_state_batch_to_device_buffer_for_host_views(void *block_tag,
-                                                               double **state_host_key,
-                                                               int state_count,
-                                                               double *device_buffer,
-                                                               int *ex,
-                                                               int i0, int j0, int k0,
-                                                               int sx, int sy, int sz);
-
-int bssn_cuda_unpack_state_batch_from_device_buffer(void *block_tag,
-                                                    int state_count,
-                                                    double *device_buffer,
-                                                    int *ex,
-                                                    int i0, int j0, int k0,
-                                                    int sx, int sy, int sz);
-
-int bssn_cuda_unpack_state_batch_from_device_buffer_for_host_views(void *block_tag,
-                                                                   double **state_host_key,
-                                                                   int state_count,
-                                                                   double *device_buffer,
-                                                                   int *ex,
-                                                                   int i0, int j0, int k0,
-                                                                   int sx, int sy, int sz);
-
-int bssn_cuda_pack_state_segments_to_device_buffer(void *block_tag,
-                                                   int state_count,
-                                                   double *device_buffer,
-                                                   int *ex,
-                                                   int segment_count,
-                                                   const int *segment_meta);
-
-int bssn_cuda_pack_state_segments_to_device_buffer_for_host_views(void *block_tag,
-                                                                  double **state_host_key,
-                                                                  int state_count,
-                                                                  double *device_buffer,
-                                                                  int *ex,
-                                                                  int segment_count,
-                                                                  const int *segment_meta);
-
-int bssn_cuda_unpack_state_segments_from_device_buffer(void *block_tag,
-                                                       int state_count,
-                                                       double *device_buffer,
-                                                       int *ex,
-                                                       int segment_count,
-                                                       const int *segment_meta);
-
-int bssn_cuda_unpack_state_segments_from_device_buffer_for_host_views(void *block_tag,
-                                                                      double **state_host_key,
-                                                                      int state_count,
-                                                                      double *device_buffer,
-                                                                      int *ex,
-                                                                      int segment_count,
-                                                                      const int *segment_meta);
-
-int bssn_cuda_restrict_state_segments_to_device_buffer(void *block_tag,
-                                                       int state_count,
-                                                       double *device_buffer,
-                                                       int *ex,
-                                                       int segment_count,
-                                                       const int *segment_meta);
-
-int bssn_cuda_restrict_state_segments_to_device_buffer_for_host_views(void *block_tag,
-                                                                      double **state_host_key,
-                                                                      int state_count,
-                                                                      double *device_buffer,
-                                                                      int *ex,
-                                                                      int segment_count,
-                                                                      const int *segment_meta,
-                                                                      const double *state_soa);
-
-int bssn_cuda_prolong_state_segments_to_device_buffer(void *block_tag,
-                                                      int state_count,
-                                                      double *device_buffer,
-                                                      int *ex,
-                                                      int segment_count,
-                                                      const int *segment_meta);
-
-int bssn_cuda_prolong_state_segments_to_device_buffer_for_host_views(void *block_tag,
-                                                                     double **state_host_key,
-                                                                     int state_count,
-                                                                     double *device_buffer,
-                                                                     int *ex,
-                                                                     int segment_count,
-                                                                     const int *segment_meta,
-                                                                     const double *state_soa);
-
-int bssn_cuda_restrict_state_batch_to_device_buffer(void *block_tag,
-                                                    int state_count,
-                                                    double *device_buffer,
-                                                    int *ex,
-                                                    int sx, int sy, int sz,
-                                                    int fi0, int fj0, int fk0);
-
-int bssn_cuda_restrict_state_batch_to_device_buffer_for_host_views(void *block_tag,
-                                                                   double **state_host_key,
-                                                                   int state_count,
-                                                                   double *device_buffer,
-                                                                   int *ex,
-                                                                   int sx, int sy, int sz,
-                                                                   int fi0, int fj0, int fk0,
-                                                                   const double *state_soa);
-
-int bssn_cuda_prolong_state_batch_to_device_buffer(void *block_tag,
-                                                   int state_count,
-                                                   double *device_buffer,
-                                                   int *ex,
-                                                   int sx, int sy, int sz,
-                                                   int ii0, int jj0, int kk0,
-                                                   int lbc_i, int lbc_j, int lbc_k);
-
-int bssn_cuda_prolong_state_batch_to_device_buffer_for_host_views(void *block_tag,
-                                                                  double **state_host_key,
-                                                                  int state_count,
-                                                                  double *device_buffer,
-                                                                  int *ex,
-                                                                  int sx, int sy, int sz,
-                                                                  int ii0, int jj0, int kk0,
-                                                                  int lbc_i, int lbc_j, int lbc_k,
-                                                                  const double *state_soa);
-
-int bssn_cuda_download_state_subset(void *block_tag,
-                                    int *ex,
-                                    int subset_count,
-                                    const int *state_indices,
-                                    double **state_host_out);
-
-int bssn_cuda_upload_state_subset(void *block_tag,
-                                  int *ex,
-                                  int subset_count,
-                                  const int *state_indices,
-                                  double **state_host_in);
-
-int bssn_cuda_prepare_inter_time_level(void *block_tag,
-                                       int *ex,
-                                       int state_count,
-                                       double **src1_host_key,
-                                       double **src2_host_key,
-                                       double **src3_host_key,
-                                       double **dst_host_key,
-                                       int source_count,
-                                       int tindex);
-
-int bssn_cuda_has_resident_state(void *block_tag);
-
-void bssn_cuda_release_step_ctx(void *block_tag);
-
-#ifdef __cplusplus
-}
-// C++-only helpers declared for derived equation classes (Z4C, etc.)
-// Defined in bssn_class.C. Requires MyList, Patch, var from including TU.
-bool bssn_cuda_use_resident_sync(int lev);
-void bssn_cuda_download_level_state_if_present(MyList<Patch> *PatL, MyList<var> *vars, int myrank);
-#endif
-
-#endif

AMSS_NCKU_source/checkpoint.C

@@ -76,11 +76,8 @@ checkpoint::checkpoint(bool checked, const char fname[], int myrank) : filename(
 
 	I_Print = (myrank == 0);
 
-	size_t filename_len = out_dir.size() + strlen(fname) + 32;
-#ifdef CHECKDETAIL
-	filename_len += 32;
-#endif
-	filename = new char[filename_len];
+	int i = strlen(fname);
+	filename = new char[i+30];
 	// cout << filename << endl;
 	// cout << i << endl;
 
@@ -103,12 +100,12 @@ checkpoint::checkpoint(bool checked, const char fname[], int myrank) : filename(
                 cout << " checkpoint class created " << endl;
         }
 }
-checkpoint::~checkpoint()
-{
-	CheckList->clearList();
-	if (filename)
-		delete[] filename;
-}
+checkpoint::~checkpoint()
+{
+	CheckList->clearList();
+	if (I_Print)
+		delete[] filename;
+}
 
 void checkpoint::addvariable(var *VV)
 {
@@ -139,7 +136,7 @@ void checkpoint::writecheck_cgh(double time, cgh *GH)
 	if (I_Print)
 	{
 		// char fname[50];
-		char fname[4096]; 
+		char fname[50+50]; 
 		sprintf(fname, "%s_cgh.CHK", filename);
 
 		outfile.open(fname, ios::out | ios::trunc);
@@ -198,7 +195,7 @@ void checkpoint::readcheck_cgh(double &time, cgh *GH, int myrank, int nprocs, in
 	int DIM = dim;
 	ifstream infile;
 	// char fname[50];
-	char fname[4096]; 
+	char fname[50+50]; 
 	sprintf(fname, "%s_cgh.CHK", filename);
 
 	infile.open(fname);
@@ -300,7 +297,7 @@ void checkpoint::writecheck_sh(double time, ShellPatch *SH)
 
 	if (I_Print)
 	{
-		char fname[4096];
+		char fname[50];
 		sprintf(fname, "%s_sh.CHK", filename);
 
 		outfile.open(fname, ios::out | ios::trunc);
@@ -338,7 +335,7 @@ void checkpoint::readcheck_sh(ShellPatch *SH, int myrank)
 	int DIM = dim;
 	ifstream infile;
 	// char fname[50];
-	char fname[4096]; 
+	char fname[50+50]; 
 	sprintf(fname, "%s_sh.CHK", filename);
 
 	infile.open(fname);
@@ -393,7 +390,7 @@ void checkpoint::write_Black_Hole_position(int BH_num_input, int BH_num, double
 
 	if (I_Print)
 	{
-		char fname[4096];
+		char fname[50];
 		sprintf(fname, "%s_BHp.CHK", filename);
 
 		outfile.open(fname, ios::out | ios::trunc);
@@ -420,7 +417,7 @@ void checkpoint::read_Black_Hole_position(int &BH_num_input, int &BH_num, double
 {
 	ifstream infile;
 	// char fname[50];
-	char fname[4096]; 
+	char fname[50+50]; 
 	sprintf(fname, "%s_BHp.CHK", filename);
 
 	infile.open(fname);
@@ -464,7 +461,7 @@ void checkpoint::write_bssn(double LastDump, double Last2dDump, double LastAnas)
 
 	if (I_Print)
 	{
-		char fname[4096];
+		char fname[50];
 		sprintf(fname, "%s_bssn.CHK", filename);
 
 		outfile.open(fname, ios::out | ios::trunc);
@@ -484,7 +481,7 @@ void checkpoint::read_bssn(double &LastDump, double &Last2dDump, double &LastAna
 {
 	ifstream infile;
 	// char fname[50];
-	char fname[4096]; 
+	char fname[50+50]; 
 	sprintf(fname, "%s_bssn.CHK", filename);
 
 	infile.open(fname);
@@ -509,7 +506,7 @@ void checkpoint::write_bssn(double LastDump, double Last2dDump, double LastAnas)
 	ofstream outfile;
 
 	// char fname[50];
-	char fname[4096]; 
+	char fname[50+50]; 
 	sprintf(fname, "%s_bssn.CHK", filename);
 
 	outfile.open(fname, ios::out | ios::trunc);
@@ -530,7 +527,7 @@ void checkpoint::read_bssn(double &LastDump, double &Last2dDump, double &LastAna
 {
 	ifstream infile;
 	// char fname[50];
-	char fname[4096]; 
+	char fname[50+50]; 
 	sprintf(fname, "%s_bssn.CHK", filename);
 
 	infile.open(fname);
@@ -554,7 +551,7 @@ void checkpoint::write_Black_Hole_position(int BH_num_input, int BH_num, double
 	ofstream outfile;
 
 	// char fname[50];
-	char fname[4096]; 
+	char fname[50+50]; 
 	sprintf(fname, "%s_BHp.CHK", filename);
 
 	outfile.open(fname, ios::out | ios::trunc);
@@ -584,7 +581,7 @@ void checkpoint::read_Black_Hole_position(int &BH_num_input, int &BH_num, double
 {
 	ifstream infile;
 	// char fname[50];
-	char fname[4096]; 
+	char fname[50+50]; 
 	sprintf(fname, "%s_BHp.CHK", filename);
 
 	infile.open(fname);
@@ -631,7 +628,7 @@ void checkpoint::writecheck_cgh(double time, cgh *GH)
 	ofstream outfile;
 
 	// char fname[50];
-	char fname[4096]; 
+	char fname[50+50]; 
 	sprintf(fname, "%s_cgh.CHK", filename);
 
 	outfile.open(fname, ios::out | ios::trunc);
@@ -741,7 +738,7 @@ void checkpoint::readcheck_cgh(double &time, cgh *GH, int myrank, int nprocs, in
 	int DIM = dim;
 	ifstream infile;
 	// char fname[50];
-	char fname[4096]; 
+	char fname[50+50]; 
 	sprintf(fname, "%s_cgh.CHK", filename);
 
 	infile.open(fname);

AMSS_NCKU_source/fd_cuda_helpers.cuh

@@ -1,412 +0,0 @@
-#ifndef AMSS_NCKU_FD_CUDA_HELPERS_CUH
-#define AMSS_NCKU_FD_CUDA_HELPERS_CUH
-
-#ifndef ghost_width
-#error "ghost_width must be defined before including fd_cuda_helpers.cuh"
-#endif
-
-#if ghost_width < 2 || ghost_width > 5
-#error "CUDA finite-difference helpers support ghost_width 2..5"
-#endif
-
-#define AMSS_FD_CENTER_RADIUS (ghost_width - 1)
-#define AMSS_FD_LK_RADIUS     (ghost_width)
-
-__device__ __forceinline__ int fd_axis_radius(int qF, int qminF, int qmaxF)
-{
-#if AMSS_FD_CENTER_RADIUS >= 4
-    if (qF - 4 >= qminF && qF + 4 <= qmaxF) return 4;
-#endif
-#if AMSS_FD_CENTER_RADIUS >= 3
-    if (qF - 3 >= qminF && qF + 3 <= qmaxF) return 3;
-#endif
-#if AMSS_FD_CENTER_RADIUS >= 2
-    if (qF - 2 >= qminF && qF + 2 <= qmaxF) return 2;
-#endif
-    if (qF - 1 >= qminF && qF + 1 <= qmaxF) return 1;
-    return 0;
-}
-
-__device__ __forceinline__ int fd_common_radius(int iF, int jF, int kF,
-                                                int iminF, int jminF, int kminF,
-                                                int imaxF, int jmaxF, int kmaxF)
-{
-    int r = fd_axis_radius(iF, iminF, imaxF);
-    const int ry = fd_axis_radius(jF, jminF, jmaxF);
-    const int rz = fd_axis_radius(kF, kminF, kmaxF);
-    if (ry < r) r = ry;
-    if (rz < r) r = rz;
-    return r;
-}
-
-__device__ __forceinline__ double fd_first_coef(int r, int off)
-{
-    switch (r) {
-    case 1:
-        if (off == -1) return -1.0;
-        if (off ==  1) return  1.0;
-        return 0.0;
-    case 2:
-        if (off == -2) return  1.0;
-        if (off == -1) return -8.0;
-        if (off ==  1) return  8.0;
-        if (off ==  2) return -1.0;
-        return 0.0;
-    case 3:
-        if (off == -3) return  -1.0;
-        if (off == -2) return   9.0;
-        if (off == -1) return -45.0;
-        if (off ==  1) return  45.0;
-        if (off ==  2) return  -9.0;
-        if (off ==  3) return   1.0;
-        return 0.0;
-    case 4:
-        if (off == -4) return    3.0;
-        if (off == -3) return  -32.0;
-        if (off == -2) return  168.0;
-        if (off == -1) return -672.0;
-        if (off ==  1) return  672.0;
-        if (off ==  2) return -168.0;
-        if (off ==  3) return   32.0;
-        if (off ==  4) return   -3.0;
-        return 0.0;
-    default:
-        return 0.0;
-    }
-}
-
-__device__ __forceinline__ double fd_second_coef(int r, int off)
-{
-    switch (r) {
-    case 1:
-        if (off == -1) return  1.0;
-        if (off ==  0) return -2.0;
-        if (off ==  1) return  1.0;
-        return 0.0;
-    case 2:
-        if (off == -2) return  -1.0;
-        if (off == -1) return  16.0;
-        if (off ==  0) return -30.0;
-        if (off ==  1) return  16.0;
-        if (off ==  2) return  -1.0;
-        return 0.0;
-    case 3:
-        if (off == -3) return    2.0;
-        if (off == -2) return  -27.0;
-        if (off == -1) return  270.0;
-        if (off ==  0) return -490.0;
-        if (off ==  1) return  270.0;
-        if (off ==  2) return  -27.0;
-        if (off ==  3) return    2.0;
-        return 0.0;
-    case 4:
-        if (off == -4) return     -9.0;
-        if (off == -3) return    128.0;
-        if (off == -2) return  -1008.0;
-        if (off == -1) return   8064.0;
-        if (off ==  0) return -14350.0;
-        if (off ==  1) return   8064.0;
-        if (off ==  2) return  -1008.0;
-        if (off ==  3) return    128.0;
-        if (off ==  4) return     -9.0;
-        return 0.0;
-    default:
-        return 0.0;
-    }
-}
-
-__device__ __forceinline__ double fd_first_denom(int r)
-{
-    return (r == 4) ? 840.0 : ((r == 3) ? 60.0 : ((r == 2) ? 12.0 : 2.0));
-}
-
-__device__ __forceinline__ double fd_second_denom(int r)
-{
-    return (r == 4) ? 5040.0 : ((r == 3) ? 180.0 : ((r == 2) ? 12.0 : 1.0));
-}
-
-__device__ __forceinline__ double fd_fetch_axis(const double *src,
-                                                int iF, int jF, int kF,
-                                                int axis, int off,
-                                                int SoA0, int SoA1, int SoA2)
-{
-    if (axis == 0) iF += off;
-    else if (axis == 1) jF += off;
-    else kF += off;
-    return fetch_sym_ord2_direct(src, iF, jF, kF, SoA0, SoA1, SoA2);
-}
-
-__device__ __forceinline__ double fd_fetch_axis2(const double *src,
-                                                 int iF, int jF, int kF,
-                                                 int axis_a, int off_a,
-                                                 int axis_b, int off_b,
-                                                 int SoA0, int SoA1, int SoA2)
-{
-    if (axis_a == 0) iF += off_a;
-    else if (axis_a == 1) jF += off_a;
-    else kF += off_a;
-    if (axis_b == 0) iF += off_b;
-    else if (axis_b == 1) jF += off_b;
-    else kF += off_b;
-    return fetch_sym_ord2_direct(src, iF, jF, kF, SoA0, SoA1, SoA2);
-}
-
-__device__ __forceinline__ double fd_first_axis_radius(const double *src,
-                                                       int iF, int jF, int kF,
-                                                       int axis, int r, double h,
-                                                       int SoA0, int SoA1, int SoA2)
-{
-    if (r <= 0) return 0.0;
-    double s = 0.0;
-#pragma unroll
-    for (int off = -4; off <= 4; ++off) {
-        const double c = fd_first_coef(r, off);
-        if (c != 0.0) {
-            s += c * fd_fetch_axis(src, iF, jF, kF, axis, off, SoA0, SoA1, SoA2);
-        }
-    }
-    return s / (fd_first_denom(r) * h);
-}
-
-__device__ __forceinline__ double fd_second_axis_radius(const double *src,
-                                                        int iF, int jF, int kF,
-                                                        int axis, int r, double h,
-                                                        int SoA0, int SoA1, int SoA2)
-{
-    if (r <= 0) return 0.0;
-    double s = 0.0;
-#pragma unroll
-    for (int off = -4; off <= 4; ++off) {
-        const double c = fd_second_coef(r, off);
-        if (c != 0.0) {
-            s += c * fd_fetch_axis(src, iF, jF, kF, axis, off, SoA0, SoA1, SoA2);
-        }
-    }
-    return s / (fd_second_denom(r) * h * h);
-}
-
-__device__ __forceinline__ double fd_mixed_axis_radius(const double *src,
-                                                       int iF, int jF, int kF,
-                                                       int axis_a, int r_a, double h_a,
-                                                       int axis_b, int r_b, double h_b,
-                                                       int SoA0, int SoA1, int SoA2)
-{
-    if (r_a <= 0 || r_b <= 0) return 0.0;
-    double s = 0.0;
-#pragma unroll
-    for (int off_a = -4; off_a <= 4; ++off_a) {
-        const double ca = fd_first_coef(r_a, off_a);
-        if (ca == 0.0) continue;
-#pragma unroll
-        for (int off_b = -4; off_b <= 4; ++off_b) {
-            const double cb = fd_first_coef(r_b, off_b);
-            if (cb != 0.0) {
-                s += ca * cb * fd_fetch_axis2(src, iF, jF, kF, axis_a, off_a,
-                                              axis_b, off_b, SoA0, SoA1, SoA2);
-            }
-        }
-    }
-    return s / (fd_first_denom(r_a) * fd_first_denom(r_b) * h_a * h_b);
-}
-
-__device__ __forceinline__ void fd_compute_first3(const double *src,
-                                                  int iF, int jF, int kF,
-                                                  int iminF, int jminF, int kminF,
-                                                  int imaxF, int jmaxF, int kmaxF,
-                                                  int SoA0, int SoA1, int SoA2,
-                                                  double &fx, double &fy, double &fz)
-{
-#if ghost_width == 3
-    const int r = fd_common_radius(iF, jF, kF, iminF, jminF, kminF, imaxF, jmaxF, kmaxF);
-    fx = fd_first_axis_radius(src, iF, jF, kF, 0, r, d_gp.dX, SoA0, SoA1, SoA2);
-    fy = fd_first_axis_radius(src, iF, jF, kF, 1, r, d_gp.dY, SoA0, SoA1, SoA2);
-    fz = fd_first_axis_radius(src, iF, jF, kF, 2, r, d_gp.dZ, SoA0, SoA1, SoA2);
-#else
-    fx = fd_first_axis_radius(src, iF, jF, kF, 0, fd_axis_radius(iF, iminF, imaxF),
-                              d_gp.dX, SoA0, SoA1, SoA2);
-    fy = fd_first_axis_radius(src, iF, jF, kF, 1, fd_axis_radius(jF, jminF, jmaxF),
-                              d_gp.dY, SoA0, SoA1, SoA2);
-    fz = fd_first_axis_radius(src, iF, jF, kF, 2, fd_axis_radius(kF, kminF, kmaxF),
-                              d_gp.dZ, SoA0, SoA1, SoA2);
-#endif
-}
-
-__device__ __forceinline__ void fd_compute_second6(const double *src,
-                                                   int iF, int jF, int kF,
-                                                   int iminF, int jminF, int kminF,
-                                                   int imaxF, int jmaxF, int kmaxF,
-                                                   int SoA0, int SoA1, int SoA2,
-                                                   double &fxx, double &fxy, double &fxz,
-                                                   double &fyy, double &fyz, double &fzz)
-{
-#if ghost_width == 3
-    const int r = fd_common_radius(iF, jF, kF, iminF, jminF, kminF, imaxF, jmaxF, kmaxF);
-    const int rx = r, ry = r, rz = r;
-#else
-    const int rx = fd_axis_radius(iF, iminF, imaxF);
-    const int ry = fd_axis_radius(jF, jminF, jmaxF);
-    const int rz = fd_axis_radius(kF, kminF, kmaxF);
-#endif
-    fxx = fd_second_axis_radius(src, iF, jF, kF, 0, rx, d_gp.dX, SoA0, SoA1, SoA2);
-    fyy = fd_second_axis_radius(src, iF, jF, kF, 1, ry, d_gp.dY, SoA0, SoA1, SoA2);
-    fzz = fd_second_axis_radius(src, iF, jF, kF, 2, rz, d_gp.dZ, SoA0, SoA1, SoA2);
-    fxy = fd_mixed_axis_radius(src, iF, jF, kF, 0, rx, d_gp.dX, 1, ry, d_gp.dY, SoA0, SoA1, SoA2);
-    fxz = fd_mixed_axis_radius(src, iF, jF, kF, 0, rx, d_gp.dX, 2, rz, d_gp.dZ, SoA0, SoA1, SoA2);
-    fyz = fd_mixed_axis_radius(src, iF, jF, kF, 1, ry, d_gp.dY, 2, rz, d_gp.dZ, SoA0, SoA1, SoA2);
-}
-
-__device__ __forceinline__ bool fd_lop_fits(int qF, int qminF, int qmaxF,
-                                            int dir, int lo, int hi)
-{
-    for (int off = lo; off <= hi; ++off) {
-        const int q = qF + dir * off;
-        if (q < qminF || q > qmaxF) return false;
-    }
-    return true;
-}
-
-__device__ __forceinline__ double fd_lop_fetch_sum(const double *src,
-                                                   int iF, int jF, int kF,
-                                                   int axis, int dir,
-                                                   const double *coef,
-                                                   int lo, int hi,
-                                                   int SoA0, int SoA1, int SoA2)
-{
-    double s = 0.0;
-    for (int off = lo; off <= hi; ++off) {
-        const double c = coef[off - lo];
-        if (c != 0.0) {
-            s += c * fd_fetch_axis(src, iF, jF, kF, axis, dir * off, SoA0, SoA1, SoA2);
-        }
-    }
-    return s;
-}
-
-__device__ __forceinline__ double fd_lopsided_axis(const double *src,
-                                                   int iF, int jF, int kF,
-                                                   int axis, double speed,
-                                                   int qF, int qminF, int qmaxF,
-                                                   double h,
-                                                   int SoA0, int SoA1, int SoA2)
-{
-    if (speed == 0.0) return 0.0;
-    const int dir = (speed > 0.0) ? 1 : -1;
-    const double mag = (speed > 0.0) ? speed : -speed;
-
-#if ghost_width == 2
-    if (fd_lop_fits(qF, qminF, qmaxF, dir, 0, 2)) {
-        const double c[] = {-3.0, 4.0, -1.0};
-        return mag * fd_lop_fetch_sum(src, iF, jF, kF, axis, dir, c, 0, 2, SoA0, SoA1, SoA2) / (2.0 * h);
-    }
-    if (fd_lop_fits(qF, qminF, qmaxF, dir, 0, 1)) {
-        const double c[] = {-1.0, 1.0};
-        return mag * fd_lop_fetch_sum(src, iF, jF, kF, axis, dir, c, 0, 1, SoA0, SoA1, SoA2) / (2.0 * h);
-    }
-    return 0.0;
-#elif ghost_width == 3
-    if (fd_lop_fits(qF, qminF, qmaxF, dir, -1, 3)) {
-        const double c[] = {-3.0, -10.0, 18.0, -6.0, 1.0};
-        return mag * fd_lop_fetch_sum(src, iF, jF, kF, axis, dir, c, -1, 3, SoA0, SoA1, SoA2) / (12.0 * h);
-    }
-    const int r = fd_axis_radius(qF, qminF, qmaxF);
-    return speed * fd_first_axis_radius(src, iF, jF, kF, axis, r, h, SoA0, SoA1, SoA2);
-#elif ghost_width == 4
-    if (fd_lop_fits(qF, qminF, qmaxF, dir, -2, 4)) {
-        const double c[] = {2.0, -24.0, -35.0, 80.0, -30.0, 8.0, -1.0};
-        return mag * fd_lop_fetch_sum(src, iF, jF, kF, axis, dir, c, -2, 4, SoA0, SoA1, SoA2) / (60.0 * h);
-    }
-    if (fd_lop_fits(qF, qminF, qmaxF, dir, -1, 5)) {
-        const double c[] = {-10.0, -77.0, 150.0, -100.0, 50.0, -15.0, 2.0};
-        return mag * fd_lop_fetch_sum(src, iF, jF, kF, axis, dir, c, -1, 5, SoA0, SoA1, SoA2) / (60.0 * h);
-    }
-    const int r = fd_axis_radius(qF, qminF, qmaxF);
-    return speed * fd_first_axis_radius(src, iF, jF, kF, axis, r, h, SoA0, SoA1, SoA2);
-#else
-    if (fd_lop_fits(qF, qminF, qmaxF, dir, -3, 5)) {
-        const double c[] = {-5.0, 60.0, -420.0, -378.0, 1050.0, -420.0, 140.0, -30.0, 3.0};
-        return mag * fd_lop_fetch_sum(src, iF, jF, kF, axis, dir, c, -3, 5, SoA0, SoA1, SoA2) / (840.0 * h);
-    }
-    const int r = fd_axis_radius(qF, qminF, qmaxF);
-    return speed * fd_first_axis_radius(src, iF, jF, kF, axis, r, h, SoA0, SoA1, SoA2);
-#endif
-}
-
-__device__ __forceinline__ double fd_ko_coef(int r, int off)
-{
-    const int a = off < 0 ? -off : off;
-    if (r == 2) {
-        if (a == 0) return 6.0;
-        if (a == 1) return -4.0;
-        if (a == 2) return 1.0;
-    } else if (r == 3) {
-        if (a == 0) return -20.0;
-        if (a == 1) return 15.0;
-        if (a == 2) return -6.0;
-        if (a == 3) return 1.0;
-    } else if (r == 4) {
-        if (a == 0) return 70.0;
-        if (a == 1) return -56.0;
-        if (a == 2) return 28.0;
-        if (a == 3) return -8.0;
-        if (a == 4) return 1.0;
-    } else if (r == 5) {
-        if (a == 0) return -252.0;
-        if (a == 1) return 210.0;
-        if (a == 2) return -120.0;
-        if (a == 3) return 45.0;
-        if (a == 4) return -10.0;
-        if (a == 5) return 1.0;
-    }
-    return 0.0;
-}
-
-__device__ __forceinline__ double fd_ko_axis(const double *src,
-                                             int iF, int jF, int kF,
-                                             int axis, int r,
-                                             int SoA0, int SoA1, int SoA2)
-{
-    double s = 0.0;
-#pragma unroll
-    for (int off = -5; off <= 5; ++off) {
-        if (off < -r || off > r) continue;
-        const double c = fd_ko_coef(r, off);
-        if (c != 0.0) {
-            s += c * fd_fetch_axis(src, iF, jF, kF, axis, off, SoA0, SoA1, SoA2);
-        }
-    }
-    return s;
-}
-
-__device__ __forceinline__ double fd_ko_term(const double *src,
-                                             int iF, int jF, int kF,
-                                             int iminF, int jminF, int kminF,
-                                             int imaxF, int jmaxF, int kmaxF,
-                                             double eps_val,
-                                             int SoA0, int SoA1, int SoA2)
-{
-    const int r = AMSS_FD_LK_RADIUS;
-    if (eps_val <= 0.0) return 0.0;
-#if ghost_width >= 4
-    if (iF - r <= iminF || iF + r >= imaxF ||
-        jF - r <= jminF || jF + r >= jmaxF ||
-        kF - r <= kminF || kF + r >= kmaxF) {
-        return 0.0;
-    }
-#else
-    if (iF - r < iminF || iF + r > imaxF ||
-        jF - r < jminF || jF + r > jmaxF ||
-        kF - r < kminF || kF + r > kmaxF) {
-        return 0.0;
-    }
-#endif
-    double cof = 1.0;
-#pragma unroll
-    for (int n = 0; n < 2 * r; ++n) cof *= 2.0;
-    const double sign = (r & 1) ? 1.0 : -1.0;
-    const double dx = fd_ko_axis(src, iF, jF, kF, 0, r, SoA0, SoA1, SoA2);
-    const double dy = fd_ko_axis(src, iF, jF, kF, 1, r, SoA0, SoA1, SoA2);
-    const double dz = fd_ko_axis(src, iF, jF, kF, 2, r, SoA0, SoA1, SoA2);
-    return sign * eps_val * (dx / d_gp.dX + dy / d_gp.dY + dz / d_gp.dZ) / cof;
-}
-
-#endif

AMSS_NCKU_source/gpu_rhsSS_mem.h

@@ -1,6 +1,6 @@
 #ifndef GPU_MEM_H_
 #define GPU_MEM_H_
-#include "macrodef.h"
+#include "macrodef.fh"
 
 #ifdef WithShell
 struct Metass
@@ -48,8 +48,6 @@ struct Meta
 	double * Gamx_rhs,*Gamy_rhs,*Gamz_rhs;//out
 	double * Lap_rhs, *betax_rhs, *betay_rhs, *betaz_rhs;//out
 	double * dtSfx_rhs,*dtSfy_rhs,*dtSfz_rhs;//out
-	double * TZ;                                  //in  (Z4C)
-	double * TZ_rhs;                              //out (Z4C)
 	double * rho,*Sx,*Sy,*Sz			;	//in
 	double * Sxx,*Sxy,*Sxz,*Syy,*Syz,*Szz;	//in
 	
@@ -134,8 +132,6 @@ __constant__ double SYM = 1.0;
 __constant__ double ANTI = -1.0; 
 __constant__ double FF = 0.75;
 __constant__ double eta = 2.0;
-__constant__ double kappa1_c = 0.02;
-__constant__ double kappa2_c = 0.0;
 __constant__ double F1o3;
 __constant__ double F2o3;
 __constant__ double F3o2 = 1.5;

AMSS_NCKU_source/makefile

@@ -1,52 +1,69 @@
 
 
-include makefile.inc
-
-## polint(ordn=6) kernel selector:
-##   1 (default): barycentric fast path
-##   0          : fallback to Neville path
-POLINT6_USE_BARY ?= 1
-POLINT6_FLAG = -DPOLINT6_USE_BARYCENTRIC=$(POLINT6_USE_BARY)
-
-## ABE build flags selected by PGO_MODE (set in makefile.inc, default: opt)
-##   make                        -> opt  (PGO-guided, maximum performance)
-##   make PGO_MODE=instrument    -> instrument (Phase 1: collect fresh profile data)
-PROFDATA = /home/$(shell whoami)/AMSS-NCKU/pgo_profile/default.profdata
-
-ifeq ($(TOOLCHAIN),intel)
-OMP_FLAG = -qopenmp
+include makefile.inc
+
+-include AMSS_NCKU_build.mk
+
+ABE_TYPE ?= $(shell awk '/^[[:space:]]*\#define[[:space:]]+ABEtype/ {print $$3; exit}' macrodef.h 2>/dev/null)
+
+ifeq ($(USE_TRANSFER_CACHE),auto)
+ifeq ($(ABE_TYPE),0)
+EFFECTIVE_USE_TRANSFER_CACHE = 1
+else
+EFFECTIVE_USE_TRANSFER_CACHE = 0
+endif
+else
+EFFECTIVE_USE_TRANSFER_CACHE = $(USE_TRANSFER_CACHE)
+endif
+
+ifeq ($(USE_CXX_ESCALAR_KERNEL),1)
+ifeq ($(ABE_TYPE),1)
+EFFECTIVE_USE_CXX_ESCALAR_KERNEL = 1
+else
+EFFECTIVE_USE_CXX_ESCALAR_KERNEL = 0
+endif
+else
+EFFECTIVE_USE_CXX_ESCALAR_KERNEL = 0
+endif
+
+ifeq ($(EFFECTIVE_USE_CXX_ESCALAR_KERNEL),1)
+ifeq ($(USE_CXX_KERNELS),0)
+$(error USE_CXX_ESCALAR_KERNEL=1 requires USE_CXX_KERNELS=1 because bssn_escalar_rhs_c.C reuses the C BSSN kernel)
+endif
+endif
+
+## polint(ordn=6) kernel selector:
+##   1 (default): barycentric fast path
+##   0          : fallback to Neville path
+POLINT6_USE_BARY ?= 1
+POLINT6_FLAG = -DPOLINT6_USE_BARYCENTRIC=$(POLINT6_USE_BARY)
+TRANSFER_CACHE_FLAG = -DBSSN_USE_TRANSFER_CACHE=$(EFFECTIVE_USE_TRANSFER_CACHE)
+ESCALAR_KERNEL_FLAG = -DBSSN_USE_ESCALAR_C_KERNEL=$(EFFECTIVE_USE_CXX_ESCALAR_KERNEL)
+
+## ABE build flags selected by PGO_MODE (set in makefile.inc, default: opt)
+##   make                        -> opt  (PGO-guided, maximum performance)
+##   make PGO_MODE=instrument    -> instrument (Phase 1: collect fresh profile data)
+PROFDATA = /home/$(shell whoami)/AMSS-NCKU/pgo_profile/default.profdata
 
 ifeq ($(PGO_MODE),instrument)
-## Intel Phase 1: instrumentation — omit -ipo/-fp-model fast=2 for faster build and numerical stability
-CXXAPPFLAGS = -O3 -march=znver5  -fma -fprofile-instr-generate -ipo \
-              -Dfortran3 -Dnewc $(MKL_INC) $(INTERP_LB_FLAGS)
-f90appflags = -O3 -march=znver5  -fma -fprofile-instr-generate -ipo \
-              -align array64byte -fpp $(MKL_INC) $(POLINT6_FLAG)
-else
-## opt (default): maximum performance with PGO profile data -fprofile-instr-use=$(PROFDATA) \
-## PGO has been turned off, now tested and found to be negative optimization
-## INTERP_LB_FLAGS has been turned off too, now tested and found to be negative optimization
-
-
-CXXAPPFLAGS = -O3 -march=znver5  -fp-model fast=2 -fma -ipo \
-              -Dfortran3 -Dnewc $(MKL_INC) $(INTERP_LB_FLAGS)
-f90appflags = -O3 -march=znver5  -fp-model fast=2 -fma -ipo \
-              -align array64byte -fpp $(MKL_INC) $(POLINT6_FLAG)
-endif
-
-TP_OPTFLAGS = -O3 -march=znver5  -fp-model fast=2 -fma -ipo \
-              -Dfortran3 -Dnewc $(MKL_INC)
-else
-## NVHPC defaults: mpicc/mpicxx/mpifort wrappers
-## PGO_MODE is ignored in this branch.
-OMP_FLAG = -mp
-CXXAPPFLAGS = -O3 -march=znver5 -tp=host -Mcache_align -Mfma \
-              -Dfortran3 -Dnewc $(MKL_INC) $(INTERP_LB_FLAGS)
-f90appflags = -O3 -march=znver5 -tp=host -Mcache_align -Mfma -Mpreprocess \
-              $(MKL_INC) $(POLINT6_FLAG)
-TP_OPTFLAGS = -O3 -march=znver5 -tp=host -Mcache_align -Mfma \
-              -Dfortran3 -Dnewc $(MKL_INC)
-endif
+## Phase 1: instrumentation — omit -ipo/-fp-model fast=2 for faster build and numerical stability
+CXXAPPFLAGS = -O3 -xHost -fma -fprofile-instr-generate -ipo \
+              -Dfortran3 -Dnewc -I${MKLROOT}/include $(INTERP_LB_FLAGS) \
+              $(TRANSFER_CACHE_FLAG) $(ESCALAR_KERNEL_FLAG)
+f90appflags = -O3 -xHost -fma -fprofile-instr-generate -ipo \
+              -align array64byte -fpp -I${MKLROOT}/include $(POLINT6_FLAG)
+else
+## opt (default): maximum performance with PGO profile data -fprofile-instr-use=$(PROFDATA) \
+## PGO has been turned off, now tested and found to be negative optimization
+## INTERP_LB_FLAGS has been turned off too, now tested and found to be negative optimization
+
+
+CXXAPPFLAGS = -O3 -xHost -fp-model fast=2 -fma -ipo \
+              -Dfortran3 -Dnewc -I${MKLROOT}/include $(INTERP_LB_FLAGS) \
+              $(TRANSFER_CACHE_FLAG) $(ESCALAR_KERNEL_FLAG)
+f90appflags = -O3 -xHost -fp-model fast=2 -fma -ipo \
+              -align array64byte -fpp -I${MKLROOT}/include $(POLINT6_FLAG)
+endif
 
 .SUFFIXES: .o .f90 .C .for .cu
 
@@ -56,34 +73,18 @@ endif
 .C.o:
 	${CXX} $(CXXAPPFLAGS) -c $< $(filein) -o $@
 
-# ShellPatch.C uses OpenMP for setupintintstuff search loops
-ShellPatch.o: ShellPatch.C
-	${CXX} $(CXXAPPFLAGS) $(OMP_FLAG) -c $< $(filein) -o $@
-
 .for.o:
 	$(f77) -c $< -o $@
 
 .cu.o:
 	$(Cu) $(CUDA_APP_FLAGS) -c $< -o $@ $(CUDA_LIB_PATH)
 
-# CUDA rewrite of BSSN RHS (drop-in replacement for bssn_rhs_c + stencil helpers)
-bssn_rhs_cuda.o: bssn_rhs_cuda.cu bssn_rhs.h macrodef.h fd_cuda_helpers.cuh
-	$(Cu) $(CUDA_APP_FLAGS) -c $< -o $@ $(CUDA_LIB_PATH)
-
-# CUDA rewrite of BSSN Shell-Patch RHS (drop-in replacement for bssn_rhs_ss)
-bssn_gpu_rhs_ss.o: bssn_gpu_rhs_ss.cu bssn_gpu.h gpu_rhsSS_mem.h bssn_macro.h macrodef.fh
-	$(Cu) $(CUDA_APP_FLAGS) -c $< -o $@ $(CUDA_LIB_PATH)
-
-# CUDA rewrite of Z4C Cartesian RHS
-z4c_rhs_cuda.o: z4c_rhs_cuda.cu z4c_rhs_cuda.h bssn_rhs.h macrodef.h ricci_gamma.h fd_cuda_helpers.cuh
-	$(Cu) $(CUDA_APP_FLAGS) -c $< -o $@ $(CUDA_LIB_PATH)
-
 # C rewrite of BSSN RHS kernel and helpers
-bssn_rhs_c.o: bssn_rhs_c.C
-	${CXX} $(CXXAPPFLAGS) -c $< $(filein) -o $@
-
-fderivs_c.o: fderivs_c.C
-	${CXX} $(CXXAPPFLAGS) -c $< $(filein) -o $@
+bssn_rhs_c.o: bssn_rhs_c.C
+	${CXX} $(CXXAPPFLAGS) -c $< $(filein) -o $@
+
+fderivs_c.o: fderivs_c.C
+	${CXX} $(CXXAPPFLAGS) -c $< $(filein) -o $@
 
 fdderivs_c.o: fdderivs_c.C
 	${CXX} $(CXXAPPFLAGS) -c $< $(filein) -o $@
@@ -91,86 +92,48 @@ fdderivs_c.o: fdderivs_c.C
 kodiss_c.o: kodiss_c.C
 	${CXX} $(CXXAPPFLAGS) -c $< $(filein) -o $@
 
-lopsided_c.o: lopsided_c.C
-	${CXX} $(CXXAPPFLAGS) -c $< $(filein) -o $@
+lopsided_c.o: lopsided_c.C
+	${CXX} $(CXXAPPFLAGS) -c $< $(filein) -o $@
+
+lopsided_kodis_c.o: lopsided_kodis_c.C
+	${CXX} $(CXXAPPFLAGS) -c $< $(filein) -o $@
+
+#interp_lb_profile.o: interp_lb_profile.C interp_lb_profile.h
+#	${CXX} $(CXXAPPFLAGS) -c $< $(filein) -o $@
 
-lopsided_kodis_c.o: lopsided_kodis_c.C
-	${CXX} $(CXXAPPFLAGS) -c $< $(filein) -o $@
-
-z4c_rhs_c.o: z4c_rhs_c.C
-	${CXX} $(CXXAPPFLAGS) -c $< $(filein) -o $@
-
-#interp_lb_profile.o: interp_lb_profile.C interp_lb_profile.h
-#	${CXX} $(CXXAPPFLAGS) -c $< $(filein) -o $@
+## TwoPunctureABE uses fixed optimal flags with its own PGO profile, independent of CXXAPPFLAGS
+TP_PROFDATA = /home/$(shell whoami)/AMSS-NCKU/pgo_profile/TwoPunctureABE.profdata
+TP_OPTFLAGS = -O3 -xHost -fp-model fast=2 -fma -ipo \
+              -fprofile-instr-use=$(TP_PROFDATA) \
+              -Dfortran3 -Dnewc -I${MKLROOT}/include
 
 TwoPunctures.o: TwoPunctures.C
-	${CXX} $(TP_OPTFLAGS) $(OMP_FLAG) -c $< -o $@
+	${CXX} $(TP_OPTFLAGS) -qopenmp -c $< -o $@
 
 TwoPunctureABE.o: TwoPunctureABE.C
-	${CXX} $(TP_OPTFLAGS) $(OMP_FLAG) -c $< -o $@
+	${CXX} $(TP_OPTFLAGS) -qopenmp -c $< -o $@
 
 # Input files
 
-## CUDA BSSN RHS switch
-##   1 : use the rewritten CUDA bssn_rhs backend
-##   0 : keep the normal CPU/Fortran selection below
-USE_CUDA_BSSN ?= 0
-USE_CUDA_Z4C ?= 0
-AMSS_Z4C_MRBD ?= 0
-
-CXXAPPFLAGS += -DUSE_CUDA_BSSN=$(USE_CUDA_BSSN)
-CUDA_APP_FLAGS += -DUSE_CUDA_BSSN=$(USE_CUDA_BSSN)
-CXXAPPFLAGS += -DUSE_CUDA_Z4C=$(USE_CUDA_Z4C)
-CUDA_APP_FLAGS += -DUSE_CUDA_Z4C=$(USE_CUDA_Z4C)
-CXXAPPFLAGS += -DAMSS_Z4C_MRBD=$(AMSS_Z4C_MRBD)
-CUDA_APP_FLAGS += -DAMSS_Z4C_MRBD=$(AMSS_Z4C_MRBD)
-
 ## Kernel implementation switch (set USE_CXX_KERNELS=0 to fall back to Fortran)
-ifeq ($(USE_CXX_KERNELS),0)
-# Fortran mode: no C rewrite files; bssn_rhs.o is included via F90FILES below
-CFILES_CPU =
-else
-# C++ mode (default): C rewrite of bssn_rhs and helper kernels
-CFILES_CPU = bssn_rhs_c.o fderivs_c.o fdderivs_c.o kodiss_c.o lopsided_c.o lopsided_kodis_c.o
-endif
-
-CFILES_CUDA_BSSN = bssn_rhs_cuda.o bssn_gpu_rhs_ss.o
-
-ifeq ($(USE_CUDA_BSSN),1)
-CFILES = $(CFILES_CUDA_BSSN)
-else
-CFILES = $(CFILES_CPU)
-endif
-
-ifeq ($(USE_CUDA_Z4C),1)
-CFILES += z4c_rhs_cuda.o
-Z4C_F90_OBJ =
-else ifeq ($(USE_CXX_Z4C_KERNELS),1)
-CFILES += z4c_rhs_c.o
-Z4C_F90_OBJ =
-else
-Z4C_F90_OBJ = Z4c_rhs.o
-endif
-
-## RK4 kernel switch (independent from USE_CXX_KERNELS)
-ifeq ($(USE_CXX_RK4),1)
-RK4_C_OBJ = rungekutta4_rout_c.o
-RK4_F90_OBJ =
-else
-RK4_C_OBJ =
-RK4_F90_OBJ = rungekutta4_rout.o
-endif
-
-CFILES += $(RK4_C_OBJ)
-ABE_CUDA_CFILES = $(CFILES_CUDA_BSSN) z4c_rhs_cuda.o $(RK4_C_OBJ)
-
-ABE_LDLIBS = $(LDLIBS)
-ifeq ($(USE_CUDA_BSSN),1)
-ABE_LDLIBS += -lcudart $(CUDA_LIB_PATH)
-endif
-ifeq ($(USE_CUDA_Z4C),1)
-ABE_LDLIBS += -lcudart $(CUDA_LIB_PATH)
-endif
+ifeq ($(USE_CXX_KERNELS),0)
+# Fortran mode: no C rewrite files; bssn_rhs.o is included via F90FILES below
+CFILES =
+else
+# C++ mode (default): C rewrite of bssn/bssn-escalar rhs and helper kernels
+CFILES = bssn_rhs_c.o fderivs_c.o fdderivs_c.o kodiss_c.o lopsided_c.o lopsided_kodis_c.o
+ifeq ($(EFFECTIVE_USE_CXX_ESCALAR_KERNEL),1)
+CFILES += bssn_escalar_rhs_c.o
+endif
+endif
+
+## RK4 kernel switch (independent from USE_CXX_KERNELS)
+ifeq ($(USE_CXX_RK4),1)
+CFILES += rungekutta4_rout_c.o
+RK4_F90_OBJ =
+else
+RK4_F90_OBJ = rungekutta4_rout.o
+endif
 
 C++FILES = ABE.o Ansorg.o Block.o misc.o monitor.o Parallel.o MPatch.o var.o\
            cgh.o bssn_class.o surface_integral.o ShellPatch.o\
@@ -179,7 +142,7 @@ C++FILES = ABE.o Ansorg.o Block.o misc.o monitor.o Parallel.o MPatch.o var.o\
            Parallel_bam.o scalar_class.o transpbh.o NullShellPatch2.o\
 	   NullShellPatch2_Evo.o writefile_f.o interp_lb_profile.o
 	   
-#C++FILES_GPU = ABE.o Ansorg.o Block.o misc.o monitor.o Parallel.o MPatch.o var.o\
+C++FILES_GPU = ABE.o Ansorg.o Block.o misc.o monitor.o Parallel.o MPatch.o var.o\
            cgh.o surface_integral.o ShellPatch.o\
 	   bssnEScalar_class.o perf.o Z4c_class.o NullShellPatch.o\
 	   bssnEM_class.o cpbc_util.o z4c_rhs_point.o checkpoint.o\
@@ -187,13 +150,13 @@ C++FILES = ABE.o Ansorg.o Block.o misc.o monitor.o Parallel.o MPatch.o var.o\
 	   NullShellPatch2_Evo.o \
 	   bssn_gpu_class.o bssn_step_gpu.o bssn_macro.o writefile_f.o
 
-F90FILES_BASE = enforce_algebra.o fmisc.o initial_puncture.o prolongrestrict.o\
-	   prolongrestrict_cell.o prolongrestrict_vertex.o\
-	   $(RK4_F90_OBJ) diff_new.o kodiss.o kodiss_sh.o\
-	   lopsidediff.o sommerfeld_rout.o getnp4.o diff_new_sh.o\
-	   shellfunctions.o bssn_rhs_ss.o Set_Rho_ADM.o\
-           getnp4EScalar.o bssnEScalar_rhs.o bssn_constraint.o ricci_gamma.o\
-           fadmquantites_bssn.o $(Z4C_F90_OBJ) Z4c_rhs_ss.o point_diff_new_sh.o\
+F90FILES_BASE = enforce_algebra.o fmisc.o initial_puncture.o prolongrestrict.o\
+	   prolongrestrict_cell.o prolongrestrict_vertex.o\
+	   $(RK4_F90_OBJ) diff_new.o kodiss.o kodiss_sh.o\
+	   lopsidediff.o sommerfeld_rout.o getnp4.o diff_new_sh.o\
+	   shellfunctions.o bssn_rhs_ss.o Set_Rho_ADM.o\
+           getnp4EScalar.o bssnEScalar_rhs.o bssn_constraint.o ricci_gamma.o\
+           fadmquantites_bssn.o Z4c_rhs.o Z4c_rhs_ss.o point_diff_new_sh.o\
 	   cpbc.o getnp4old.o NullEvol.o initial_null.o initial_maxwell.o\
 	   getnpem2.o empart.o NullNews.o fourdcurvature.o\
 	   bssn2adm.o adm_constraint.o adm_ricci_gamma.o\
@@ -217,10 +180,10 @@ initial_guess.o Newton.o Jacobian.o ilucg.o IntPnts0.o IntPnts.o
 
 TwoPunctureFILES = TwoPunctureABE.o TwoPunctures.o 
 
-#CUDAFILES = bssn_gpu.o bssn_gpu_rhs_ss.o
+CUDAFILES = bssn_gpu.o bssn_gpu_rhs_ss.o
 
 # file dependences
-$(C++FILES) $(C++FILES_GPU) $(F90FILES) $(CFILES) $(ABE_CUDA_CFILES) $(AHFDOBJS) $(CUDAFILES): macrodef.fh
+$(C++FILES) $(C++FILES_GPU) $(F90FILES) $(CFILES) $(AHFDOBJS) $(CUDAFILES): macrodef.fh
 
 $(C++FILES): Block.h enforce_algebra.h fmisc.h initial_puncture.h macrodef.h\
 	     misc.h monitor.h MyList.h Parallel.h MPatch.h prolongrestrict.h\
@@ -231,7 +194,7 @@ $(C++FILES): Block.h enforce_algebra.h fmisc.h initial_puncture.h macrodef.h\
 	     empart.h NullNews.h kodiss.h Parallel_bam.h ricci_gamma.h\
              initial_null2.h NullShellPatch2.h 
              
-#$(C++FILES_GPU): Block.h enforce_algebra.h fmisc.h initial_puncture.h macrodef.h\
+$(C++FILES_GPU): Block.h enforce_algebra.h fmisc.h initial_puncture.h macrodef.h\
 	     misc.h monitor.h MyList.h Parallel.h MPatch.h prolongrestrict.h\
 	     rungekutta4_rout.h var.h bssn_rhs.h sommerfeld_rout.h\
 	     cgh.h surface_integral.h ShellPatch.h shellfunctions.h perf.h\
@@ -243,7 +206,7 @@ $(C++FILES): Block.h enforce_algebra.h fmisc.h initial_puncture.h macrodef.h\
              
 $(AHFDOBJS): cctk.h cctk_Config.h cctk_Types.h cctk_Constants.h myglobal.h
 
-$(C++FILES) $(C++FILES_GPU) $(CFILES) $(ABE_CUDA_CFILES) $(AHFDOBJS) $(CUDAFILES): macrodef.h
+$(C++FILES) $(C++FILES_GPU) $(CFILES) $(AHFDOBJS) $(CUDAFILES): macrodef.h
 
 TwoPunctureFILES: TwoPunctures.h
 
@@ -253,18 +216,13 @@ misc.o : zbesh.o
 
 # projects
 ABE: $(C++FILES) $(CFILES) $(F90FILES) $(F77FILES) $(AHFDOBJS)
-	$(CLINKER) $(CXXAPPFLAGS) -o $@ $(C++FILES) $(CFILES) $(F90FILES) $(F77FILES) $(AHFDOBJS) $(ABE_LDLIBS)
-
-ABE_CUDA: USE_CUDA_BSSN=1
-ABE_CUDA: USE_CUDA_Z4C=1
-ABE_CUDA: $(C++FILES) $(ABE_CUDA_CFILES) $(F90FILES) $(F77FILES) $(AHFDOBJS)
-	$(CLINKER) $(CXXAPPFLAGS) -o $@ $(C++FILES) $(ABE_CUDA_CFILES) $(F90FILES) $(F77FILES) $(AHFDOBJS) $(LDLIBS) -lcudart $(CUDA_LIB_PATH)
+	$(CLINKER) $(CXXAPPFLAGS) -o $@ $(C++FILES) $(CFILES) $(F90FILES) $(F77FILES) $(AHFDOBJS) $(LDLIBS)
 	
-#ABEGPU: $(C++FILES_GPU) $(CFILES) $(F90FILES) $(F77FILES) $(AHFDOBJS) $(CUDAFILES)
-#	$(CLINKER) $(CXXAPPFLAGS) -o $@ $(C++FILES_GPU) $(CFILES) $(F90FILES) $(F77FILES) $(AHFDOBJS) $(CUDAFILES) $(LDLIBS)
+ABEGPU: $(C++FILES_GPU) $(CFILES) $(F90FILES) $(F77FILES) $(AHFDOBJS) $(CUDAFILES)
+	$(CLINKER) $(CXXAPPFLAGS) -o $@ $(C++FILES_GPU) $(CFILES) $(F90FILES) $(F77FILES) $(AHFDOBJS) $(CUDAFILES) $(LDLIBS)
 
 TwoPunctureABE: $(TwoPunctureFILES)
-	$(CLINKER) $(TP_OPTFLAGS) $(OMP_FLAG) -o $@ $(TwoPunctureFILES) $(LDLIBS)
+	$(CLINKER) $(TP_OPTFLAGS) -qopenmp -o $@ $(TwoPunctureFILES) $(LDLIBS)
 
 clean:
-	rm *.o ABE ABE_CUDA ABEGPU TwoPunctureABE make.log -f
+	rm *.o ABE ABEGPU TwoPunctureABE make.log -f

AMSS_NCKU_source/makefile.inc

@@ -1,7 +1,28 @@
-## Toolchain selection
-##   nvhpc : NVIDIA HPC SDK + CUDA-aware MPI (default)
-##   intel : Intel oneAPI toolchain (legacy path)
-TOOLCHAIN ?= intel
+## GCC version (commented out)
+## filein  = -I/usr/include -I/usr/lib/x86_64-linux-gnu/mpich/include -I/usr/lib/x86_64-linux-gnu/openmpi/lib/ -I/usr/lib/gcc/x86_64-linux-gnu/11/ -I/usr/include/c++/11/
+## filein  = -I/usr/include/ -I/usr/include/openmpi-x86_64/ -I/usr/lib/x86_64-linux-gnu/openmpi/include/ -I/usr/lib/x86_64-linux-gnu/openmpi/lib/ -I/usr/lib/gcc/x86_64-linux-gnu/11/ -I/usr/include/c++/11/
+## LDLIBS  = -L/usr/lib/x86_64-linux-gnu -L/usr/lib64 -L/usr/lib/gcc/x86_64-linux-gnu/11 -lgfortran -lmpi -lgfortran
+
+## Intel oneAPI version with oneMKL (Optimized for performance)
+filein  = -I/usr/include/ -I${MKLROOT}/include
+
+## Using sequential MKL (OpenMP disabled for better single-threaded performance)
+## Added -lifcore for Intel Fortran runtime and -limf for Intel math library
+LDLIBS  = -L${MKLROOT}/lib -lmkl_intel_lp64 -lmkl_sequential -lmkl_core -lifcore -limf -lpthread -lm -ldl -liomp5
+
+## Memory allocator switch
+##   1 (default) : link Intel oneTBB allocator (libtbbmalloc)
+##   0           : use system default allocator (ptmalloc)
+USE_TBBMALLOC ?= 1
+TBBMALLOC_SO ?= /home/intel/oneapi/2025.3/lib/libtbbmalloc.so
+ifneq ($(wildcard $(TBBMALLOC_SO)),)
+TBBMALLOC_LIBS = -Wl,--no-as-needed $(TBBMALLOC_SO) -Wl,--as-needed
+else
+TBBMALLOC_LIBS = -Wl,--no-as-needed -ltbbmalloc -Wl,--as-needed
+endif
+ifeq ($(USE_TBBMALLOC),1)
+LDLIBS := $(TBBMALLOC_LIBS) $(LDLIBS)
+endif
 
 ## PGO build mode switch (ABE only; TwoPunctureABE always uses opt flags)
 ##   opt        : (default) maximum performance with PGO profile-guided optimization
@@ -22,70 +43,35 @@ else
 INTERP_LB_FLAGS =
 endif
 
-MKLROOT ?= /home/intel/oneapi/mkl/latest
-MKL_LIBDIR ?= $(MKLROOT)/lib/intel64
-MKL_INC ?= -I$(MKLROOT)/include
-
-NVHPC_ROOT ?= /home/nvidia/hpc_sdk/Linux_x86_64/25.11
-CUDA_HOME  ?= $(NVHPC_ROOT)/cuda
-CUDA_ARCH  ?= sm_80
-
 ## Kernel implementation switch
 ##   1 (default) : use C++ rewrite of bssn_rhs and helper kernels (faster)
 ##   0           : fall back to original Fortran kernels
 USE_CXX_KERNELS ?= 1
 
-## Z4C Cartesian RHS kernel switch
-##   1 (default) : use C++ rewrite of Z4c_rhs (main Cartesian path faster)
-##   0           : use original Fortran Z4c_rhs.o
-USE_CXX_Z4C_KERNELS ?= 1
+## BSSN-EScalar RHS switch
+##   1 (default) : use BSSN-EScalar C wrapper on the normal patch path
+##   0           : keep the original Fortran BSSN-EScalar RHS for precision-safe runs
+## Note: this requires USE_CXX_KERNELS=1 because the wrapper reuses the C BSSN kernel.
+USE_CXX_ESCALAR_KERNEL ?= 1
+
+## Cached transfer switch
+##   auto (default): enable for BSSN vacuum, keep other paths on the safe uncached path
+##   1             : force cached Sync/Restrict/OutBd transfer on evolution hot paths
+##   0             : force the original uncached transfer path
+USE_TRANSFER_CACHE ?= auto
 
 ## RK4 kernel implementation switch
 ##   1 (default) : use C/C++ rewrite of rungekutta4_rout (for optimization experiments)
 ##   0           : use original Fortran rungekutta4_rout.o
 USE_CXX_RK4 ?= 1
 
-## Memory allocator switch
-##   1 (default) : link Intel oneTBB allocator (libtbbmalloc)
-##   0           : use system default allocator (ptmalloc)
-USE_TBBMALLOC ?= 1
-TBBMALLOC_SO ?= /home/intel/oneapi/2025.3/lib/libtbbmalloc.so
-ifneq ($(wildcard $(TBBMALLOC_SO)),)
-TBBMALLOC_LIBS = -Wl,--no-as-needed $(TBBMALLOC_SO) -Wl,--as-needed
-else
-TBBMALLOC_LIBS = -Wl,--no-as-needed -ltbbmalloc -Wl,--as-needed
-endif
-
-ifeq ($(TOOLCHAIN),intel)
 f90          = ifx
 f77          = ifx
 CXX          = icpx
 CC           = icx
 CLINKER      = mpiicpx
-filein  = -I/usr/include/ $(MKL_INC) -I$(CUDA_HOME)/include
-LDLIBS       = -L$(MKL_LIBDIR) -Wl,-rpath,$(MKL_LIBDIR) \
-               -lmkl_intel_lp64 -lmkl_sequential -lmkl_core \
-               -lifcore -limf -liomp5 -lpthread -lm -ldl \
-               -L$(CUDA_HOME)/lib64 -Wl,-rpath,$(CUDA_HOME)/lib64 -lcuda -lcudart
-else ifeq ($(TOOLCHAIN),nvhpc)
-f90          = mpifort
-f77          = mpifort
-CXX          = mpicxx
-CC           = mpicc
-CLINKER      = mpicxx
 
-filein       = -I/usr/include/ $(MKL_INC) -I$(CUDA_HOME)/include
-LDLIBS       = -L$(MKL_LIBDIR) -Wl,-rpath,$(MKL_LIBDIR) \
-               -lmkl_intel_lp64 -lmkl_sequential -lmkl_core \
-               -lpthread -lm -ldl \
-               -L$(CUDA_HOME)/lib64 -Wl,-rpath,$(CUDA_HOME)/lib64 -lcuda -lcudart \
-               -fortranlibs
-endif
-
-ifeq ($(USE_TBBMALLOC),1)
-LDLIBS := $(TBBMALLOC_LIBS) $(LDLIBS)
-endif
-
-Cu = $(NVHPC_ROOT)/compilers/bin/nvcc
-CUDA_LIB_PATH = -L$(CUDA_HOME)/lib64 -I$(CUDA_HOME)/include
-CUDA_APP_FLAGS = -c -g -O3 --ptxas-options=-v -Dfortran3 -Dnewc -arch=$(CUDA_ARCH)
+Cu = nvcc
+CUDA_LIB_PATH = -L/usr/lib/cuda/lib64 -I/usr/include -I/usr/lib/cuda/include
+#CUDA_APP_FLAGS = -c -g -O3 --ptxas-options=-v -arch compute_13 -code compute_13,sm_13 -Dfortran3 -Dnewc
+CUDA_APP_FLAGS = -c -g -O3 --ptxas-options=-v -Dfortran3 -Dnewc

AMSS_NCKU_source/monitor.C

@@ -1,8 +1,7 @@
 
-#ifdef newc
-#include <cstdio>
-#include <sstream>
-using namespace std;
+#ifdef newc
+#include <cstdio>
+using namespace std;
 #else
 #include <stdio.h>
 #endif
@@ -78,17 +77,16 @@ monitor::monitor(const char fname[], int myrank, string head)
       parameters::str_par.insert(map<string, string>::value_type("output dir", out_dir));
     }
     // considering checkpoint run
-    string filename = out_dir + "/" + fname;
-    int i = 1;
-    while ((access(filename.c_str(), F_OK)) != -1)
-    {
-      stringstream ss;
-      ss << out_dir << "/" << i << "_" << fname;
-      filename = ss.str();
-      i++;
-    }
-
-    outfile.open(filename.c_str(), ios::trunc);
+    char filename[50];
+    sprintf(filename, "%s/%s", out_dir.c_str(), fname);
+    int i = 1;
+    while ((access(filename, F_OK)) != -1)
+    {
+      sprintf(filename, "%s/%d_%s", out_dir.c_str(), i, fname);
+      i++;
+    }
+
+    outfile.open(filename, ios::trunc);
 
     time_t tnow;
     time(&tnow);
@@ -109,17 +107,16 @@ monitor::monitor(const char fname[], int myrank, const int out_rank, string head
   if (I_Print)
   {
     // considering checkpoint run
-    string filename = out_dir + "/" + fname;
-    int i = 1;
-    while ((access(filename.c_str(), F_OK)) != -1)
-    {
-      stringstream ss;
-      ss << out_dir << "/" << i << "_" << fname;
-      filename = ss.str();
-      i++;
-    }
-
-    outfile.open(filename.c_str(), ios::trunc);
+    char filename[50];
+    sprintf(filename, "%s/%s", out_dir.c_str(), fname);
+    int i = 1;
+    while ((access(filename, F_OK)) != -1)
+    {
+      sprintf(filename, "%s/%d_%s", out_dir.c_str(), i, fname);
+      i++;
+    }
+
+    outfile.open(filename, ios::trunc);
 
     time_t tnow;
     time(&tnow);

AMSS_NCKU_source/surface_integral.C

@@ -8,11 +8,10 @@
 #include <iostream>
 #include <iomanip>
 #include <fstream>
-#include <strstream>
-#include <cmath>
-#include <map>
-#include <cstdlib>
-using namespace std;
+#include <strstream>
+#include <cmath>
+#include <map>
+using namespace std;
 #else
 #include <iostream.h>
 #include <iomanip.h>
@@ -30,26 +29,12 @@ using namespace std;
 #include "fadmquantites_bssn.h"
 #include "getnpem2.h"
 #include "getnp4.h"
-#include "parameters.h"
-
-#define PI M_PI
-
-namespace
-{
-bool amss_surface_timing_enabled()
-{
-  static int enabled = -1;
-  if (enabled < 0)
-  {
-    const char *env = getenv("AMSS_SURFACE_TIMING");
-    enabled = (env && atoi(env) != 0) ? 1 : 0;
-  }
-  return enabled != 0;
-}
-}
-//|============================================================================
-//| Constructor
-//|============================================================================
+#include "parameters.h"
+
+#define PI M_PI
+//|============================================================================
+//| Constructor
+//|============================================================================
 
 surface_integral::surface_integral(int iSymmetry) : Symmetry(iSymmetry),
                                                     wave_cache_spinw(-1),
@@ -499,9 +484,9 @@ void surface_integral::surf_Wave(double rex, int lev, cgh *GH, var *Rpsi4, var *
   delete[] IP_out;
   DG_List->clearList();
 }
-void surface_integral::surf_Wave(double rex, int lev, cgh *GH, var *Rpsi4, var *Ipsi4,
-                                 int spinw, int maxl, int NN, double *RP, double *IP,
-                                 monitor *Monitor, MPI_Comm Comm_here) // NN is the length of RP and IP
+void surface_integral::surf_Wave(double rex, int lev, cgh *GH, var *Rpsi4, var *Ipsi4,
+                                 int spinw, int maxl, int NN, double *RP, double *IP,
+                                 monitor *Monitor, MPI_Comm Comm_here) // NN is the length of RP and IP
 {
   //   misc::tillherecheck(GH->Commlev[lev],GH->start_rank[lev],"start surface_integral::surf_Wave");
 
@@ -735,10 +720,10 @@ void surface_integral::surf_Wave(double rex, int lev, cgh *GH, var *Rpsi4, var *
   delete[] IP_out;
   DG_List->clearList();
 }
-//|----------------------------------------------------------------
-//  for shell patch
-//|----------------------------------------------------------------
-void surface_integral::surf_Wave(double rex, int lev, ShellPatch *GH, var *Rpsi4, var *Ipsi4,
+//|----------------------------------------------------------------
+//  for shell patch
+//|----------------------------------------------------------------
+void surface_integral::surf_Wave(double rex, int lev, ShellPatch *GH, var *Rpsi4, var *Ipsi4,
                                  int spinw, int maxl, int NN, double *RP, double *IP,
                                  monitor *Monitor) // NN is the length of RP and IP
 {
@@ -3296,8 +3281,6 @@ void surface_integral::surf_WaveMassPAng(double rex, int lev, cgh *GH,
                                          var *Sfx_rhs, var *Sfy_rhs, var *Sfz_rhs,
                                          double *Rout, monitor *Monitor, bool refresh_mass_fields)
 {
-  const bool timing = amss_surface_timing_enabled();
-  const double t_start = timing ? MPI_Wtime() : 0.0;
   if (Symmetry != 0 && Symmetry != 1)
   {
     surf_Wave(rex, lev, GH, Rpsi4, Ipsi4, spinw, maxl, NN, RP, IP, Monitor);
@@ -3342,7 +3325,6 @@ void surface_integral::surf_WaveMassPAng(double rex, int lev, cgh *GH,
       Pp = Pp->next;
     }
   }
-  const double t_refresh_done = timing ? MPI_Wtime() : 0.0;
 
   const int InList = 19;
   const int idx_rpsi4 = 0, idx_ipsi4 = 1;
@@ -3398,7 +3380,6 @@ void surface_integral::surf_WaveMassPAng(double rex, int lev, cgh *GH,
 
   double *shellf = new double[n_tot * InList];
   GH->PatL[lev]->data->Interp_Points(DG_List, n_tot, pox, shellf, Symmetry, Nmin, Nmax);
-  const double t_interp_done = timing ? MPI_Wtime() : 0.0;
 
   double *RP_out = new double[NN];
   double *IP_out = new double[NN];
@@ -3515,7 +3496,6 @@ void surface_integral::surf_WaveMassPAng(double rex, int lev, cgh *GH,
     if (Symmetry == 0)
       p_outz += f1o8 * Psi * (nx_g[n] * axz + ny_g[n] * ayz + nz_g[n] * azz) * theta_weight;
   }
-  const double t_integral_done = timing ? MPI_Wtime() : 0.0;
 
   for (int ii = 0; ii < NN; ii++)
   {
@@ -3554,7 +3534,6 @@ void surface_integral::surf_WaveMassPAng(double rex, int lev, cgh *GH,
     delete[] reduce_out;
     delete[] reduce_in;
   }
-  const double t_reduce_done = timing ? MPI_Wtime() : 0.0;
 
 #ifdef GaussInt
   mass = mass * rex * rex * dphi * factor;
@@ -3586,19 +3565,6 @@ void surface_integral::surf_WaveMassPAng(double rex, int lev, cgh *GH,
   Rout[5] = sy;
   Rout[6] = sz;
 
-  if (timing)
-  {
-    fprintf(stderr,
-            "[AMSS-SURFACE][rank %d] rex=%.6g lev=%d refresh=%.6f interp=%.6f integral=%.6f reduce=%.6f total=%.6f nlocal=%d ntotal=%d modes=%d\n",
-            myrank, rex, lev,
-            t_refresh_done - t_start,
-            t_interp_done - t_refresh_done,
-            t_integral_done - t_interp_done,
-            t_reduce_done - t_integral_done,
-            t_reduce_done - t_start,
-            Nmax - Nmin + 1, n_tot, NN);
-  }
-
   delete[] pox[0];
   delete[] pox[1];
   delete[] pox[2];

AMSS_NCKU_source/surface_integral.h

@@ -46,10 +46,10 @@ public:
 	surface_integral(int iSymmetry);
 	~surface_integral();
 
-	void surf_Wave(double rex, int lev, cgh *GH, var *Rpsi4, var *Ipsi4,
-				   int spinw, int maxl, int NN, double *RP, double *IP,
-				   monitor *Monitor); // NN is the length of RP and IP
-									  // this routine can only deal with the symmetry of Psi4
+	void surf_Wave(double rex, int lev, cgh *GH, var *Rpsi4, var *Ipsi4,
+				   int spinw, int maxl, int NN, double *RP, double *IP,
+				   monitor *Monitor); // NN is the length of RP and IP
+									  // this routine can only deal with the symmetry of Psi4
 	void surf_Wave(double rex, int lev, ShellPatch *GH, var *Rpsi4, var *Ipsi4,
 				   int spinw, int maxl, int NN, double *RP, double *IP,
 				   monitor *Monitor);

AMSS_NCKU_source/z4c_rhs_c.C

@@ -1,725 +0,0 @@
-#include "macrodef.h"
-#include "bssn_rhs.h"
-#include "fmisc.h"
-#include "ricci_gamma.h"
-#include "share_func.h"
-#include "tool.h"
-#include <vector>
-
-#ifdef fortran1
-#define f_constraint_bssn constraint_bssn
-#define f_z4c_rhs_point z4c_rhs_point
-#endif
-#ifdef fortran2
-#define f_constraint_bssn CONSTRAINT_BSSN
-#define f_z4c_rhs_point Z4C_RHS_POINT
-#endif
-#ifdef fortran3
-#define f_constraint_bssn constraint_bssn_
-#define f_z4c_rhs_point z4c_rhs_point_
-#endif
-
-extern "C" void f_constraint_bssn(int *, double *, double *, double *,
-                                  double *, double *,
-                                  double *, double *, double *, double *, double *, double *,
-                                  double *, double *, double *, double *, double *, double *,
-                                  double *, double *, double *,
-                                  double *, double *, double *, double *, double *, double *, double *, double *, double *, double *,
-                                  double *, double *, double *, double *, double *, double *,
-                                  double *, double *, double *, double *, double *, double *,
-                                  double *, double *, double *, double *, double *, double *,
-                                  double *, double *, double *, double *, double *, double *, double *, double *,
-                                  double *, double *, double *,
-                                  int &);
-
-extern "C" void f_z4c_rhs_point(
-    double &A11,
-    double &A12,
-    double &A13,
-    double &A22,
-    double &A23,
-    double &A33,
-    double &alpha,
-    double &B1,
-    double &B2,
-    double &B3,
-    double &beta1,
-    double &beta2,
-    double &beta3,
-    double &chi,
-    double &chiDivFloor,
-    double &da1,
-    double &dA111,
-    double &dA112,
-    double &dA113,
-    double &dA122,
-    double &dA123,
-    double &dA133,
-    double &da2,
-    double &dA211,
-    double &dA212,
-    double &dA213,
-    double &dA222,
-    double &dA223,
-    double &dA233,
-    double &da3,
-    double &dA311,
-    double &dA312,
-    double &dA313,
-    double &dA322,
-    double &dA323,
-    double &dA333,
-    double &db11,
-    double &dB11,
-    double &db12,
-    double &dB12,
-    double &db13,
-    double &dB13,
-    double &db21,
-    double &dB21,
-    double &db22,
-    double &dB22,
-    double &db23,
-    double &dB23,
-    double &db31,
-    double &dB31,
-    double &db32,
-    double &dB32,
-    double &db33,
-    double &dB33,
-    double &dchi1,
-    double &dchi2,
-    double &dchi3,
-    double &dda11,
-    double &dda12,
-    double &dda13,
-    double &dda22,
-    double &dda23,
-    double &dda33,
-    double &ddb111,
-    double &ddb112,
-    double &ddb113,
-    double &ddb121,
-    double &ddb122,
-    double &ddb123,
-    double &ddb131,
-    double &ddb132,
-    double &ddb133,
-    double &ddb221,
-    double &ddb222,
-    double &ddb223,
-    double &ddb231,
-    double &ddb232,
-    double &ddb233,
-    double &ddb331,
-    double &ddb332,
-    double &ddb333,
-    double &ddchi11,
-    double &ddchi12,
-    double &ddchi13,
-    double &ddchi22,
-    double &ddchi23,
-    double &ddchi33,
-    double &deldelg1111,
-    double &deldelg1112,
-    double &deldelg1113,
-    double &deldelg1122,
-    double &deldelg1123,
-    double &deldelg1133,
-    double &deldelg1211,
-    double &deldelg1212,
-    double &deldelg1213,
-    double &deldelg1222,
-    double &deldelg1223,
-    double &deldelg1233,
-    double &deldelg1311,
-    double &deldelg1312,
-    double &deldelg1313,
-    double &deldelg1322,
-    double &deldelg1323,
-    double &deldelg1333,
-    double &deldelg2211,
-    double &deldelg2212,
-    double &deldelg2213,
-    double &deldelg2222,
-    double &deldelg2223,
-    double &deldelg2233,
-    double &deldelg2311,
-    double &deldelg2312,
-    double &deldelg2313,
-    double &deldelg2322,
-    double &deldelg2323,
-    double &deldelg2333,
-    double &deldelg3311,
-    double &deldelg3312,
-    double &deldelg3313,
-    double &deldelg3322,
-    double &deldelg3323,
-    double &deldelg3333,
-    double &delG11,
-    double &delg111,
-    double &delg112,
-    double &delg113,
-    double &delG12,
-    double &delg122,
-    double &delg123,
-    double &delG13,
-    double &delg133,
-    double &delG21,
-    double &delg211,
-    double &delg212,
-    double &delg213,
-    double &delG22,
-    double &delg222,
-    double &delg223,
-    double &delG23,
-    double &delg233,
-    double &delG31,
-    double &delg311,
-    double &delg312,
-    double &delg313,
-    double &delG32,
-    double &delg322,
-    double &delg323,
-    double &delG33,
-    double &delg333,
-    double &dKhat1,
-    double &dKhat2,
-    double &dKhat3,
-    double &dTheta1,
-    double &dTheta2,
-    double &dTheta3,
-    double &G1,
-    double &g11,
-    double &g12,
-    double &g13,
-    double &G2,
-    double &g22,
-    double &g23,
-    double &G3,
-    double &g33,
-    double &kappa1,
-    double &kappa2,
-    double &Khat,
-    double &rA11,
-    double &rA12,
-    double &rA13,
-    double &rA22,
-    double &rA23,
-    double &rA33,
-    double &rchi,
-    double &rG1,
-    double &rg11,
-    double &rg12,
-    double &rg13,
-    double &rG2,
-    double &rg22,
-    double &rg23,
-    double &rG3,
-    double &rg33,
-    double &rKhat,
-    double &rTheta,
-    double &Theta);
-
-static inline void z4c_contract_gamma(
-    const double gxx, const double gxy, const double gxz,
-    const double gyy, const double gyz, const double gzz,
-    const double gxxx, const double gxyx, const double gxzx,
-    const double gyyx, const double gyzx, const double gzzx,
-    const double gxxy, const double gxyy, const double gxzy,
-    const double gyyy, const double gyzy, const double gzzy,
-    const double gxxz, const double gxyz, const double gxzz,
-    const double gyyz, const double gyzz, const double gzzz,
-    double &Gamxa, double &Gamya, double &Gamza)
-{
-    double det = gxx * gyy * gzz + gxy * gyz * gxz + gxz * gxy * gyz -
-                 gxz * gyy * gxz - gxy * gxy * gzz - gxx * gyz * gyz;
-    const double gupxx = (gyy * gzz - gyz * gyz) / det;
-    const double gupxy = -(gxy * gzz - gyz * gxz) / det;
-    const double gupxz = (gxy * gyz - gyy * gxz) / det;
-    const double gupyy = (gxx * gzz - gxz * gxz) / det;
-    const double gupyz = -(gxx * gyz - gxy * gxz) / det;
-    const double gupzz = (gxx * gyy - gxy * gxy) / det;
-
-    const double Gamxxx = 0.5 * (gupxx * gxxx + gupxy * (2.0 * gxyx - gxxy) + gupxz * (2.0 * gxzx - gxxz));
-    const double Gamyxx = 0.5 * (gupxy * gxxx + gupyy * (2.0 * gxyx - gxxy) + gupyz * (2.0 * gxzx - gxxz));
-    const double Gamzxx = 0.5 * (gupxz * gxxx + gupyz * (2.0 * gxyx - gxxy) + gupzz * (2.0 * gxzx - gxxz));
-
-    const double Gamxyy = 0.5 * (gupxx * (2.0 * gxyy - gyyx) + gupxy * gyyy + gupxz * (2.0 * gyzy - gyyz));
-    const double Gamyyy = 0.5 * (gupxy * (2.0 * gxyy - gyyx) + gupyy * gyyy + gupyz * (2.0 * gyzy - gyyz));
-    const double Gamzyy = 0.5 * (gupxz * (2.0 * gxyy - gyyx) + gupyz * gyyy + gupzz * (2.0 * gyzy - gyyz));
-
-    const double Gamxzz = 0.5 * (gupxx * (2.0 * gxzz - gzzx) + gupxy * (2.0 * gyzz - gzzy) + gupxz * gzzz);
-    const double Gamyzz = 0.5 * (gupxy * (2.0 * gxzz - gzzx) + gupyy * (2.0 * gyzz - gzzy) + gupyz * gzzz);
-    const double Gamzzz = 0.5 * (gupxz * (2.0 * gxzz - gzzx) + gupyz * (2.0 * gyzz - gzzy) + gupzz * gzzz);
-
-    const double Gamxxy = 0.5 * (gupxx * gxxy + gupxy * gyyx + gupxz * (gxzy + gyzx - gxyz));
-    const double Gamyxy = 0.5 * (gupxy * gxxy + gupyy * gyyx + gupyz * (gxzy + gyzx - gxyz));
-    const double Gamzxy = 0.5 * (gupxz * gxxy + gupyz * gyyx + gupzz * (gxzy + gyzx - gxyz));
-
-    const double Gamxxz = 0.5 * (gupxx * gxxz + gupxy * (gxyz + gyzx - gxzy) + gupxz * gzzx);
-    const double Gamyxz = 0.5 * (gupxy * gxxz + gupyy * (gxyz + gyzx - gxzy) + gupyz * gzzx);
-    const double Gamzxz = 0.5 * (gupxz * gxxz + gupyz * (gxyz + gyzx - gxzy) + gupzz * gzzx);
-
-    const double Gamxyz = 0.5 * (gupxx * (gxyz + gxzy - gyzx) + gupxy * gyyz + gupxz * gzzy);
-    const double Gamyyz = 0.5 * (gupxy * (gxyz + gxzy - gyzx) + gupyy * gyyz + gupyz * gzzy);
-    const double Gamzyz = 0.5 * (gupxz * (gxyz + gxzy - gyzx) + gupyz * gyyz + gupzz * gzzy);
-
-    Gamxa = gupxx * Gamxxx + gupyy * Gamxyy + gupzz * Gamxzz +
-            2.0 * (gupxy * Gamxxy + gupxz * Gamxxz + gupyz * Gamxyz);
-    Gamya = gupxx * Gamyxx + gupyy * Gamyyy + gupzz * Gamyzz +
-            2.0 * (gupxy * Gamyxy + gupxz * Gamyxz + gupyz * Gamyyz);
-    Gamza = gupxx * Gamzxx + gupyy * Gamzyy + gupzz * Gamzzz +
-            2.0 * (gupxy * Gamzxy + gupxz * Gamzxz + gupyz * Gamzyz);
-}
-
-static int compute_rhs_z4c_cartesian(
-    int *ex, double &T, double *X, double *Y, double *Z,
-    double *chi_state, double *chi_constraints, double *trK,
-    double *dxx, double *gxy, double *gxz, double *dyy, double *gyz, double *dzz,
-    double *Axx, double *Axy, double *Axz, double *Ayy, double *Ayz, double *Azz,
-    double *Gamx, double *Gamy, double *Gamz,
-    double *Lap, double *betax, double *betay, double *betaz,
-    double *dtSfx, double *dtSfy, double *dtSfz,
-    double *TZ,
-    double *chi_rhs, double *trK_rhs,
-    double *gxx_rhs, double *gxy_rhs, double *gxz_rhs, double *gyy_rhs, double *gyz_rhs, double *gzz_rhs,
-    double *Axx_rhs, double *Axy_rhs, double *Axz_rhs, double *Ayy_rhs, double *Ayz_rhs, double *Azz_rhs,
-    double *Gamx_rhs, double *Gamy_rhs, double *Gamz_rhs,
-    double *Lap_rhs, double *betax_rhs, double *betay_rhs, double *betaz_rhs,
-    double *dtSfx_rhs, double *dtSfy_rhs, double *dtSfz_rhs,
-    double *TZ_rhs,
-    double *rho, double *Sx, double *Sy, double *Sz,
-    double *Sxx, double *Sxy, double *Sxz, double *Syy, double *Syz, double *Szz,
-    double *Gamxxx, double *Gamxxy, double *Gamxxz, double *Gamxyy, double *Gamxyz, double *Gamxzz,
-    double *Gamyxx, double *Gamyxy, double *Gamyxz, double *Gamyyy, double *Gamyyz, double *Gamyzz,
-    double *Gamzxx, double *Gamzxy, double *Gamzxz, double *Gamzyy, double *Gamzyz, double *Gamzzz,
-    double *Rxx, double *Rxy, double *Rxz, double *Ryy, double *Ryz, double *Rzz,
-    double *Hcon, double *Mxcon, double *Mycon, double *Mzcon, double *Gmxcon, double *Gmycon, double *Gmzcon,
-    int &Symmetry, int &Lev, double &eps, int &co)
-{
-    (void)T;
-
-    const int nx = ex[0];
-    const int ny = ex[1];
-    const int nz = ex[2];
-    const int all = nx * ny * nz;
-
-    double alpn1[all], chin1[all], gxx[all], gyy[all], gzz[all];
-    double chix[all], chiy[all], chiz[all], chixx[all], chixy[all], chixz[all], chiyy[all], chiyz[all], chizz[all];
-    double gxxx[all], gxyx[all], gxzx[all], gyyx[all], gyzx[all], gzzx[all];
-    double gxxy[all], gxyy[all], gxzy[all], gyyy[all], gyzy[all], gzzy[all];
-    double gxxz[all], gxyz[all], gxzz[all], gyyz[all], gyzz[all], gzzz[all];
-    double gxxxx[all], gxxxy[all], gxxxz[all], gxxyy[all], gxxyz[all], gxxzz[all];
-    double gxyxx[all], gxyxy[all], gxyxz[all], gxyyy[all], gxyyz[all], gxyzz[all];
-    double gxzxx[all], gxzxy[all], gxzxz[all], gxzyy[all], gxzyz[all], gxzzz[all];
-    double gyyxx[all], gyyxy[all], gyyxz[all], gyyyy[all], gyyyz[all], gyyzz[all];
-    double gyzxx[all], gyzxy[all], gyzxz[all], gyzyy[all], gyzyz[all], gyzzz[all];
-    double gzzxx[all], gzzxy[all], gzzxz[all], gzzyy[all], gzzyz[all], gzzzz[all];
-    double Lapx[all], Lapy[all], Lapz[all], Lapxx[all], Lapxy[all], Lapxz[all], Lapyy[all], Lapyz[all], Lapzz[all];
-    double betaxx[all], betaxy[all], betaxz[all], betayx[all], betayy[all], betayz[all], betazx[all], betazy[all], betazz[all];
-    double dBxx[all], dBxy[all], dBxz[all], dByx[all], dByy[all], dByz[all], dBzx[all], dBzy[all], dBzz[all];
-    double sfxxx[all], sfxxy[all], sfxxz[all], sfxyy[all], sfxyz[all], sfxzz[all];
-    double sfyxx[all], sfyxy[all], sfyxz[all], sfyyy[all], sfyyz[all], sfyzz[all];
-    double sfzxx[all], sfzxy[all], sfzxz[all], sfzyy[all], sfzyz[all], sfzzz[all];
-    double Gamxx[all], Gamxy[all], Gamxz[all], Gamyx[all], Gamyy[all], Gamyz[all], Gamzx[all], Gamzy[all], Gamzz[all];
-    double Kx[all], Ky[all], Kz[all], TZx[all], TZy[all], TZz[all];
-    double Axxx[all], Axxy[all], Axxz[all], Axyx[all], Axyy[all], Axyz[all];
-    double Axzx[all], Axzy[all], Axzz[all], Ayyx[all], Ayyy[all], Ayyz[all];
-    double Ayzx[all], Ayzy[all], Ayzz[all], Azzx[all], Azzy[all], Azzz[all];
-
-    const double SSS[3] = {1.0, 1.0, 1.0};
-    const double AAS[3] = {-1.0, -1.0, 1.0};
-    const double ASA[3] = {-1.0, 1.0, -1.0};
-    const double SAA[3] = {1.0, -1.0, -1.0};
-    const double ASS[3] = {-1.0, 1.0, 1.0};
-    const double SAS[3] = {1.0, -1.0, 1.0};
-    const double SSA[3] = {1.0, 1.0, -1.0};
-
-    const double ONE = 1.0;
-    const double TWO = 2.0;
-    const double ZEO = 0.0;
-    double chiDivfloor = 1.0e-5;
-
-    double kappa1 = 2.0e-2;
-    double kappa2 = 0.0;
-    double FF = 0.75;
-    double eta = 2.0;
-
-    for (int idx = 0; idx < all; ++idx)
-    {
-        alpn1[idx] = Lap[idx] + ONE;
-        chin1[idx] = chi_state[idx] + ONE;
-        gxx[idx] = dxx[idx] + ONE;
-        gyy[idx] = dyy[idx] + ONE;
-        gzz[idx] = dzz[idx] + ONE;
-    }
-
-    fderivs(ex, betax, betaxx, betaxy, betaxz, X, Y, Z, -1.0, 1.0, 1.0, Symmetry, Lev);
-    fderivs(ex, betay, betayx, betayy, betayz, X, Y, Z, 1.0, -1.0, 1.0, Symmetry, Lev);
-    fderivs(ex, betaz, betazx, betazy, betazz, X, Y, Z, 1.0, 1.0, -1.0, Symmetry, Lev);
-    fderivs(ex, dtSfx, dBxx, dBxy, dBxz, X, Y, Z, -1.0, 1.0, 1.0, Symmetry, Lev);
-    fderivs(ex, dtSfy, dByx, dByy, dByz, X, Y, Z, 1.0, -1.0, 1.0, Symmetry, Lev);
-    fderivs(ex, dtSfz, dBzx, dBzy, dBzz, X, Y, Z, 1.0, 1.0, -1.0, Symmetry, Lev);
-    fderivs(ex, chi_state, chix, chiy, chiz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
-    fderivs(ex, dxx, gxxx, gxxy, gxxz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
-    fderivs(ex, gxy, gxyx, gxyy, gxyz, X, Y, Z, -1.0, -1.0, 1.0, Symmetry, Lev);
-    fderivs(ex, gxz, gxzx, gxzy, gxzz, X, Y, Z, -1.0, 1.0, -1.0, Symmetry, Lev);
-    fderivs(ex, dyy, gyyx, gyyy, gyyz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
-    fderivs(ex, gyz, gyzx, gyzy, gyzz, X, Y, Z, 1.0, -1.0, -1.0, Symmetry, Lev);
-    fderivs(ex, dzz, gzzx, gzzy, gzzz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
-
-    fdderivs(ex, dxx, gxxxx, gxxxy, gxxxz, gxxyy, gxxyz, gxxzz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
-    fdderivs(ex, dyy, gyyxx, gyyxy, gyyxz, gyyyy, gyyyz, gyyzz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
-    fdderivs(ex, dzz, gzzxx, gzzxy, gzzxz, gzzyy, gzzyz, gzzzz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
-    fdderivs(ex, gxy, gxyxx, gxyxy, gxyxz, gxyyy, gxyyz, gxyzz, X, Y, Z, -1.0, -1.0, 1.0, Symmetry, Lev);
-    fdderivs(ex, gxz, gxzxx, gxzxy, gxzxz, gxzyy, gxzyz, gxzzz, X, Y, Z, -1.0, 1.0, -1.0, Symmetry, Lev);
-    fdderivs(ex, gyz, gyzxx, gyzxy, gyzxz, gyzyy, gyzyz, gyzzz, X, Y, Z, 1.0, -1.0, -1.0, Symmetry, Lev);
-
-    fderivs(ex, Gamx, Gamxx, Gamxy, Gamxz, X, Y, Z, -1.0, 1.0, 1.0, Symmetry, Lev);
-    fderivs(ex, Gamy, Gamyx, Gamyy, Gamyz, X, Y, Z, 1.0, -1.0, 1.0, Symmetry, Lev);
-    fderivs(ex, Gamz, Gamzx, Gamzy, Gamzz, X, Y, Z, 1.0, 1.0, -1.0, Symmetry, Lev);
-
-    fderivs(ex, Lap, Lapx, Lapy, Lapz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
-    fderivs(ex, trK, Kx, Ky, Kz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
-    fderivs(ex, TZ, TZx, TZy, TZz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
-
-    fdderivs(ex, betax, sfxxx, sfxxy, sfxxz, sfxyy, sfxyz, sfxzz, X, Y, Z, -1.0, 1.0, 1.0, Symmetry, Lev);
-    fdderivs(ex, betay, sfyxx, sfyxy, sfyxz, sfyyy, sfyyz, sfyzz, X, Y, Z, 1.0, -1.0, 1.0, Symmetry, Lev);
-    fdderivs(ex, betaz, sfzxx, sfzxy, sfzxz, sfzyy, sfzyz, sfzzz, X, Y, Z, 1.0, 1.0, -1.0, Symmetry, Lev);
-
-    fdderivs(ex, chi_state, chixx, chixy, chixz, chiyy, chiyz, chizz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
-    fdderivs(ex, Lap, Lapxx, Lapxy, Lapxz, Lapyy, Lapyz, Lapzz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
-
-    fderivs(ex, Axx, Axxx, Axxy, Axxz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
-    fderivs(ex, Axy, Axyx, Axyy, Axyz, X, Y, Z, -1.0, -1.0, 1.0, Symmetry, Lev);
-    fderivs(ex, Axz, Axzx, Axzy, Axzz, X, Y, Z, -1.0, 1.0, -1.0, Symmetry, Lev);
-    fderivs(ex, Ayy, Ayyx, Ayyy, Ayyz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
-    fderivs(ex, Ayz, Ayzx, Ayzy, Ayzz, X, Y, Z, 1.0, -1.0, -1.0, Symmetry, Lev);
-    fderivs(ex, Azz, Azzx, Azzy, Azzz, X, Y, Z, 1.0, 1.0, 1.0, Symmetry, Lev);
-
-    for (int idx = 0; idx < all; ++idx)
-    {
-        double point_kappa1 = 0.0;
-        f_z4c_rhs_point(
-            Axx[idx], Axy[idx], Axz[idx], Ayy[idx], Ayz[idx], Azz[idx],
-            alpn1[idx], dtSfx[idx], dtSfy[idx], dtSfz[idx],
-            betax[idx], betay[idx], betaz[idx],
-            chin1[idx], chiDivfloor,
-            Lapx[idx],
-            Axxx[idx], Axyx[idx], Axzx[idx], Ayyx[idx], Ayzx[idx], Azzx[idx],
-            Lapy[idx],
-            Axxy[idx], Axyy[idx], Axzy[idx], Ayyy[idx], Ayzy[idx], Azzy[idx],
-            Lapz[idx],
-            Axxz[idx], Axyz[idx], Axzz[idx], Ayyz[idx], Ayzz[idx], Azzz[idx],
-            betaxx[idx], dBxx[idx], betayx[idx], dByx[idx], betazx[idx], dBzx[idx],
-            betaxy[idx], dBxy[idx], betayy[idx], dByy[idx], betazy[idx], dBzy[idx],
-            betaxz[idx], dBxz[idx], betayz[idx], dByz[idx], betazz[idx], dBzz[idx],
-            chix[idx], chiy[idx], chiz[idx],
-            Lapxx[idx], Lapxy[idx], Lapxz[idx], Lapyy[idx], Lapyz[idx], Lapzz[idx],
-            sfxxx[idx], sfyxx[idx], sfzxx[idx],
-            sfxxy[idx], sfyxy[idx], sfzxy[idx],
-            sfxxz[idx], sfyxz[idx], sfzxz[idx],
-            sfxyy[idx], sfyyy[idx], sfzyy[idx],
-            sfxyz[idx], sfyyz[idx], sfzyz[idx],
-            sfxzz[idx], sfyzz[idx], sfzzz[idx],
-            chixx[idx], chixy[idx], chixz[idx], chiyy[idx], chiyz[idx], chizz[idx],
-            gxxxx[idx], gxyxx[idx], gxzxx[idx], gyyxx[idx], gyzxx[idx], gzzxx[idx],
-            gxxxy[idx], gxyxy[idx], gxzxy[idx], gyyxy[idx], gyzxy[idx], gzzxy[idx],
-            gxxxz[idx], gxyxz[idx], gxzxz[idx], gyyxz[idx], gyzxz[idx], gzzxz[idx],
-            gxxyy[idx], gxyyy[idx], gxzyy[idx], gyyyy[idx], gyzyy[idx], gzzyy[idx],
-            gxxyz[idx], gxyyz[idx], gxzyz[idx], gyyyz[idx], gyzyz[idx], gzzyz[idx],
-            gxxzz[idx], gxyzz[idx], gxzzz[idx], gyyzz[idx], gyzzz[idx], gzzzz[idx],
-            Gamxx[idx], gxxx[idx], gxyx[idx], gxzx[idx],
-            Gamyx[idx], gyyx[idx], gyzx[idx],
-            Gamzx[idx], gzzx[idx],
-            Gamxy[idx], gxxy[idx], gxyy[idx], gxzy[idx],
-            Gamyy[idx], gyyy[idx], gyzy[idx],
-            Gamzy[idx], gzzy[idx],
-            Gamxz[idx], gxxz[idx], gxyz[idx], gxzz[idx],
-            Gamyz[idx], gyyz[idx], gyzz[idx],
-            Gamzz[idx], gzzz[idx],
-            Kx[idx], Ky[idx], Kz[idx],
-            TZx[idx], TZy[idx], TZz[idx],
-            Gamx[idx], gxx[idx], gxy[idx], gxz[idx],
-            Gamy[idx], gyy[idx], gyz[idx],
-            Gamz[idx], gzz[idx],
-            point_kappa1, kappa2,
-            trK[idx],
-            Axx_rhs[idx], Axy_rhs[idx], Axz_rhs[idx], Ayy_rhs[idx], Ayz_rhs[idx], Azz_rhs[idx],
-            chi_rhs[idx],
-            Gamx_rhs[idx], gxx_rhs[idx], gxy_rhs[idx], gxz_rhs[idx],
-            Gamy_rhs[idx], gyy_rhs[idx], gyz_rhs[idx],
-            Gamz_rhs[idx], gzz_rhs[idx], trK_rhs[idx], TZ_rhs[idx], TZ[idx]);
-    }
-
-    for (int idx = 0; idx < all; ++idx)
-        Lap_rhs[idx] = -TWO * alpn1[idx] * trK[idx];
-
-#if (GAUGE == 0)
-    for (int idx = 0; idx < all; ++idx)
-    {
-        betax_rhs[idx] = FF * dtSfx[idx];
-        betay_rhs[idx] = FF * dtSfy[idx];
-        betaz_rhs[idx] = FF * dtSfz[idx];
-        dtSfx_rhs[idx] = Gamx_rhs[idx] - eta * dtSfx[idx];
-        dtSfy_rhs[idx] = Gamy_rhs[idx] - eta * dtSfy[idx];
-        dtSfz_rhs[idx] = Gamz_rhs[idx] - eta * dtSfz[idx];
-    }
-#elif (GAUGE == 1)
-    for (int idx = 0; idx < all; ++idx)
-    {
-        betax_rhs[idx] = Gamx[idx] - eta * betax[idx];
-        betay_rhs[idx] = Gamy[idx] - eta * betay[idx];
-        betaz_rhs[idx] = Gamz[idx] - eta * betaz[idx];
-        dtSfx_rhs[idx] = ZEO;
-        dtSfy_rhs[idx] = ZEO;
-        dtSfz_rhs[idx] = ZEO;
-    }
-#else
-#error "z4c_rhs_c.C currently supports GAUGE == 0 or GAUGE == 1 for Z4C"
-#endif
-
-    lopsided(ex, X, Y, Z, gxx, gxx_rhs, betax, betay, betaz, Symmetry, SSS);
-    lopsided(ex, X, Y, Z, gxy, gxy_rhs, betax, betay, betaz, Symmetry, AAS);
-    lopsided(ex, X, Y, Z, gxz, gxz_rhs, betax, betay, betaz, Symmetry, ASA);
-    lopsided(ex, X, Y, Z, gyy, gyy_rhs, betax, betay, betaz, Symmetry, SSS);
-    lopsided(ex, X, Y, Z, gyz, gyz_rhs, betax, betay, betaz, Symmetry, SAA);
-    lopsided(ex, X, Y, Z, gzz, gzz_rhs, betax, betay, betaz, Symmetry, SSS);
-
-    lopsided(ex, X, Y, Z, Axx, Axx_rhs, betax, betay, betaz, Symmetry, SSS);
-    lopsided(ex, X, Y, Z, Axy, Axy_rhs, betax, betay, betaz, Symmetry, AAS);
-    lopsided(ex, X, Y, Z, Axz, Axz_rhs, betax, betay, betaz, Symmetry, ASA);
-    lopsided(ex, X, Y, Z, Ayy, Ayy_rhs, betax, betay, betaz, Symmetry, SSS);
-    lopsided(ex, X, Y, Z, Ayz, Ayz_rhs, betax, betay, betaz, Symmetry, SAA);
-    lopsided(ex, X, Y, Z, Azz, Azz_rhs, betax, betay, betaz, Symmetry, SSS);
-
-    lopsided(ex, X, Y, Z, chi_state, chi_rhs, betax, betay, betaz, Symmetry, SSS);
-    lopsided(ex, X, Y, Z, trK, trK_rhs, betax, betay, betaz, Symmetry, SSS);
-
-    lopsided(ex, X, Y, Z, Gamx, Gamx_rhs, betax, betay, betaz, Symmetry, ASS);
-    lopsided(ex, X, Y, Z, Gamy, Gamy_rhs, betax, betay, betaz, Symmetry, SAS);
-    lopsided(ex, X, Y, Z, Gamz, Gamz_rhs, betax, betay, betaz, Symmetry, SSA);
-
-    lopsided(ex, X, Y, Z, Lap, Lap_rhs, betax, betay, betaz, Symmetry, SSS);
-    lopsided(ex, X, Y, Z, betax, betax_rhs, betax, betay, betaz, Symmetry, ASS);
-    lopsided(ex, X, Y, Z, betay, betay_rhs, betax, betay, betaz, Symmetry, SAS);
-    lopsided(ex, X, Y, Z, betaz, betaz_rhs, betax, betay, betaz, Symmetry, SSA);
-#if (GAUGE == 0)
-    lopsided(ex, X, Y, Z, dtSfx, dtSfx_rhs, betax, betay, betaz, Symmetry, ASS);
-    lopsided(ex, X, Y, Z, dtSfy, dtSfy_rhs, betax, betay, betaz, Symmetry, SAS);
-    lopsided(ex, X, Y, Z, dtSfz, dtSfz_rhs, betax, betay, betaz, Symmetry, SSA);
-#endif
-    lopsided(ex, X, Y, Z, TZ, TZ_rhs, betax, betay, betaz, Symmetry, SSS);
-
-    for (int idx = 0; idx < all; ++idx)
-    {
-        double Gamxa = 0.0, Gamya = 0.0, Gamza = 0.0;
-        z4c_contract_gamma(
-            gxx[idx], gxy[idx], gxz[idx], gyy[idx], gyz[idx], gzz[idx],
-            gxxx[idx], gxyx[idx], gxzx[idx], gyyx[idx], gyzx[idx], gzzx[idx],
-            gxxy[idx], gxyy[idx], gxzy[idx], gyyy[idx], gyzy[idx], gzzy[idx],
-            gxxz[idx], gxyz[idx], gxzz[idx], gyyz[idx], gyzz[idx], gzzz[idx],
-            Gamxa, Gamya, Gamza);
-
-        TZ_rhs[idx] -= alpn1[idx] * (TWO + kappa2) * kappa1 * TZ[idx];
-        trK_rhs[idx] += alpn1[idx] * kappa1 * (ONE - kappa2) * TZ[idx];
-        Gamx_rhs[idx] -= TWO * alpn1[idx] * kappa1 * (Gamx[idx] - Gamxa);
-        Gamy_rhs[idx] -= TWO * alpn1[idx] * kappa1 * (Gamy[idx] - Gamya);
-        Gamz_rhs[idx] -= TWO * alpn1[idx] * kappa1 * (Gamz[idx] - Gamza);
-    }
-
-    if (eps > 0.0)
-    {
-        kodis(ex, X, Y, Z, chi_state, chi_rhs, SSS, Symmetry, eps);
-        kodis(ex, X, Y, Z, trK, trK_rhs, SSS, Symmetry, eps);
-        kodis(ex, X, Y, Z, gxx, gxx_rhs, SSS, Symmetry, eps);
-        kodis(ex, X, Y, Z, gxy, gxy_rhs, AAS, Symmetry, eps);
-        kodis(ex, X, Y, Z, gxz, gxz_rhs, ASA, Symmetry, eps);
-        kodis(ex, X, Y, Z, gyy, gyy_rhs, SSS, Symmetry, eps);
-        kodis(ex, X, Y, Z, gyz, gyz_rhs, SAA, Symmetry, eps);
-        kodis(ex, X, Y, Z, gzz, gzz_rhs, SSS, Symmetry, eps);
-        kodis(ex, X, Y, Z, Axx, Axx_rhs, SSS, Symmetry, eps);
-        kodis(ex, X, Y, Z, Axy, Axy_rhs, AAS, Symmetry, eps);
-        kodis(ex, X, Y, Z, Axz, Axz_rhs, ASA, Symmetry, eps);
-        kodis(ex, X, Y, Z, Ayy, Ayy_rhs, SSS, Symmetry, eps);
-        kodis(ex, X, Y, Z, Ayz, Ayz_rhs, SAA, Symmetry, eps);
-        kodis(ex, X, Y, Z, Azz, Azz_rhs, SSS, Symmetry, eps);
-        kodis(ex, X, Y, Z, Gamx, Gamx_rhs, ASS, Symmetry, eps);
-        kodis(ex, X, Y, Z, Gamy, Gamy_rhs, SAS, Symmetry, eps);
-        kodis(ex, X, Y, Z, Gamz, Gamz_rhs, SSA, Symmetry, eps);
-        kodis(ex, X, Y, Z, Lap, Lap_rhs, SSS, Symmetry, eps);
-        kodis(ex, X, Y, Z, betax, betax_rhs, ASS, Symmetry, eps);
-        kodis(ex, X, Y, Z, betay, betay_rhs, SAS, Symmetry, eps);
-        kodis(ex, X, Y, Z, betaz, betaz_rhs, SSA, Symmetry, eps);
-#if (GAUGE == 0)
-        kodis(ex, X, Y, Z, dtSfx, dtSfx_rhs, ASS, Symmetry, eps);
-        kodis(ex, X, Y, Z, dtSfy, dtSfy_rhs, SAS, Symmetry, eps);
-        kodis(ex, X, Y, Z, dtSfz, dtSfz_rhs, SSA, Symmetry, eps);
-#endif
-        kodis(ex, X, Y, Z, TZ, TZ_rhs, SSS, Symmetry, eps);
-    }
-
-    if (co == 0)
-    {
-#if (ABV == 0)
-        f_ricci_gamma(ex, X, Y, Z,
-                      chi_constraints,
-                      dxx, gxy, gxz, dyy, gyz, dzz,
-                      Gamx, Gamy, Gamz,
-                      Gamxxx, Gamxxy, Gamxxz, Gamxyy, Gamxyz, Gamxzz,
-                      Gamyxx, Gamyxy, Gamyxz, Gamyyy, Gamyyz, Gamyzz,
-                      Gamzxx, Gamzxy, Gamzxz, Gamzyy, Gamzyz, Gamzzz,
-                      Rxx, Rxy, Rxz, Ryy, Ryz, Rzz,
-                      Symmetry);
-#endif
-        f_constraint_bssn(ex, X, Y, Z,
-                          chi_constraints, trK,
-                          dxx, gxy, gxz, dyy, gyz, dzz,
-                          Axx, Axy, Axz, Ayy, Ayz, Azz,
-                          Gamx, Gamy, Gamz,
-                          Lap, betax, betay, betaz, rho, Sx, Sy, Sz,
-                          Gamxxx, Gamxxy, Gamxxz, Gamxyy, Gamxyz, Gamxzz,
-                          Gamyxx, Gamyxy, Gamyxz, Gamyyy, Gamyyz, Gamyzz,
-                          Gamzxx, Gamzxy, Gamzxz, Gamzyy, Gamzyz, Gamzzz,
-                          Rxx, Rxy, Rxz, Ryy, Ryz, Rzz,
-                          Hcon, Mxcon, Mycon, Mzcon, Gmxcon, Gmycon, Gmzcon,
-                          Symmetry);
-    }
-
-    return 0;
-}
-
-extern "C" int f_compute_rhs_Z4c(int *ex, double &T,
-                                 double *X, double *Y, double *Z,
-                                 double *chi, double *trK,
-                                 double *dxx, double *gxy, double *gxz, double *dyy, double *gyz, double *dzz,
-                                 double *Axx, double *Axy, double *Axz, double *Ayy, double *Ayz, double *Azz,
-                                 double *Gamx, double *Gamy, double *Gamz,
-                                 double *Lap, double *betax, double *betay, double *betaz,
-                                 double *dtSfx, double *dtSfy, double *dtSfz,
-                                 double *TZ,
-                                 double *chi_rhs, double *trK_rhs,
-                                 double *gxx_rhs, double *gxy_rhs, double *gxz_rhs, double *gyy_rhs, double *gyz_rhs, double *gzz_rhs,
-                                 double *Axx_rhs, double *Axy_rhs, double *Axz_rhs, double *Ayy_rhs, double *Ayz_rhs, double *Azz_rhs,
-                                 double *Gamx_rhs, double *Gamy_rhs, double *Gamz_rhs,
-                                 double *Lap_rhs, double *betax_rhs, double *betay_rhs, double *betaz_rhs,
-                                 double *dtSfx_rhs, double *dtSfy_rhs, double *dtSfz_rhs,
-                                 double *TZ_rhs,
-                                 double *rho, double *Sx, double *Sy, double *Sz,
-                                 double *Sxx, double *Sxy, double *Sxz, double *Syy, double *Syz, double *Szz,
-                                 double *Gamxxx, double *Gamxxy, double *Gamxxz, double *Gamxyy, double *Gamxyz, double *Gamxzz,
-                                 double *Gamyxx, double *Gamyxy, double *Gamyxz, double *Gamyyy, double *Gamyyz, double *Gamyzz,
-                                 double *Gamzxx, double *Gamzxy, double *Gamzxz, double *Gamzyy, double *Gamzyz, double *Gamzzz,
-                                 double *Rxx, double *Rxy, double *Rxz, double *Ryy, double *Ryz, double *Rzz,
-                                 double *Hcon, double *Mxcon, double *Mycon, double *Mzcon, double *Gmxcon, double *Gmycon, double *Gmzcon,
-                                 int &Symmetry, int &Lev, double &eps, int &co)
-{
-    return compute_rhs_z4c_cartesian(
-        ex, T, X, Y, Z,
-        chi, chi, trK,
-        dxx, gxy, gxz, dyy, gyz, dzz,
-        Axx, Axy, Axz, Ayy, Ayz, Azz,
-        Gamx, Gamy, Gamz,
-        Lap, betax, betay, betaz,
-        dtSfx, dtSfy, dtSfz,
-        TZ,
-        chi_rhs, trK_rhs,
-        gxx_rhs, gxy_rhs, gxz_rhs, gyy_rhs, gyz_rhs, gzz_rhs,
-        Axx_rhs, Axy_rhs, Axz_rhs, Ayy_rhs, Ayz_rhs, Azz_rhs,
-        Gamx_rhs, Gamy_rhs, Gamz_rhs,
-        Lap_rhs, betax_rhs, betay_rhs, betaz_rhs,
-        dtSfx_rhs, dtSfy_rhs, dtSfz_rhs,
-        TZ_rhs,
-        rho, Sx, Sy, Sz,
-        Sxx, Sxy, Sxz, Syy, Syz, Szz,
-        Gamxxx, Gamxxy, Gamxxz, Gamxyy, Gamxyz, Gamxzz,
-        Gamyxx, Gamyxy, Gamyxz, Gamyyy, Gamyyz, Gamyzz,
-        Gamzxx, Gamzxy, Gamzxz, Gamzyy, Gamzyz, Gamzzz,
-        Rxx, Rxy, Rxz, Ryy, Ryz, Rzz,
-        Hcon, Mxcon, Mycon, Mzcon, Gmxcon, Gmycon, Gmzcon,
-        Symmetry, Lev, eps, co);
-}
-
-extern "C" int f_compute_rhs_Z4cnot(int *ex, double &T,
-                                    double *X, double *Y, double *Z,
-                                    double *chi, double *trK,
-                                    double *dxx, double *gxy, double *gxz, double *dyy, double *gyz, double *dzz,
-                                    double *Axx, double *Axy, double *Axz, double *Ayy, double *Ayz, double *Azz,
-                                    double *Gamx, double *Gamy, double *Gamz,
-                                    double *Lap, double *betax, double *betay, double *betaz,
-                                    double *dtSfx, double *dtSfy, double *dtSfz,
-                                    double *TZ,
-                                    double *chi_rhs, double *trK_rhs,
-                                    double *gxx_rhs, double *gxy_rhs, double *gxz_rhs, double *gyy_rhs, double *gyz_rhs, double *gzz_rhs,
-                                    double *Axx_rhs, double *Axy_rhs, double *Axz_rhs, double *Ayy_rhs, double *Ayz_rhs, double *Azz_rhs,
-                                    double *Gamx_rhs, double *Gamy_rhs, double *Gamz_rhs,
-                                    double *Lap_rhs, double *betax_rhs, double *betay_rhs, double *betaz_rhs,
-                                    double *dtSfx_rhs, double *dtSfy_rhs, double *dtSfz_rhs,
-                                    double *TZ_rhs,
-                                    double *rho, double *Sx, double *Sy, double *Sz,
-                                    double *Sxx, double *Sxy, double *Sxz, double *Syy, double *Syz, double *Szz,
-                                    double *Gamxxx, double *Gamxxy, double *Gamxxz, double *Gamxyy, double *Gamxyz, double *Gamxzz,
-                                    double *Gamyxx, double *Gamyxy, double *Gamyxz, double *Gamyyy, double *Gamyyz, double *Gamyzz,
-                                    double *Gamzxx, double *Gamzxy, double *Gamzxz, double *Gamzyy, double *Gamzyz, double *Gamzzz,
-                                    double *Rxx, double *Rxy, double *Rxz, double *Ryy, double *Ryz, double *Rzz,
-                                    double *Hcon, double *Mxcon, double *Mycon, double *Mzcon, double *Gmxcon, double *Gmycon, double *Gmzcon,
-                                    int &Symmetry, int &Lev, double &eps, int &co, double &chitiny)
-{
-    const int all = ex[0] * ex[1] * ex[2];
-    std::vector<double> chi_clamped(chi, chi + all);
-    f_lowerboundset(ex, chi_clamped.data(), chitiny);
-
-    const int ret = compute_rhs_z4c_cartesian(
-        ex, T, X, Y, Z,
-        chi_clamped.data(), chi, trK,
-        dxx, gxy, gxz, dyy, gyz, dzz,
-        Axx, Axy, Axz, Ayy, Ayz, Azz,
-        Gamx, Gamy, Gamz,
-        Lap, betax, betay, betaz,
-        dtSfx, dtSfy, dtSfz,
-        TZ,
-        chi_rhs, trK_rhs,
-        gxx_rhs, gxy_rhs, gxz_rhs, gyy_rhs, gyz_rhs, gzz_rhs,
-        Axx_rhs, Axy_rhs, Axz_rhs, Ayy_rhs, Ayz_rhs, Azz_rhs,
-        Gamx_rhs, Gamy_rhs, Gamz_rhs,
-        Lap_rhs, betax_rhs, betay_rhs, betaz_rhs,
-        dtSfx_rhs, dtSfy_rhs, dtSfz_rhs,
-        TZ_rhs,
-        rho, Sx, Sy, Sz,
-        Sxx, Sxy, Sxz, Syy, Syz, Szz,
-        Gamxxx, Gamxxy, Gamxxz, Gamxyy, Gamxyz, Gamxzz,
-        Gamyxx, Gamyxy, Gamyxz, Gamyyy, Gamyyz, Gamyzz,
-        Gamzxx, Gamzxy, Gamzxz, Gamzyy, Gamzyz, Gamzzz,
-        Rxx, Rxy, Rxz, Ryy, Ryz, Rzz,
-        Hcon, Mxcon, Mycon, Mzcon, Gmxcon, Gmycon, Gmzcon,
-        Symmetry, Lev, eps, co);
-
-    if (ret != 0 || co != 0)
-        return ret;
-
-#if (ABV == 0)
-    f_ricci_gamma(ex, X, Y, Z,
-                  chi,
-                  dxx, gxy, gxz, dyy, gyz, dzz,
-                  Gamx, Gamy, Gamz,
-                  Gamxxx, Gamxxy, Gamxxz, Gamxyy, Gamxyz, Gamxzz,
-                  Gamyxx, Gamyxy, Gamyxz, Gamyyy, Gamyyz, Gamyzz,
-                  Gamzxx, Gamzxy, Gamzxz, Gamzyy, Gamzyz, Gamzzz,
-                  Rxx, Rxy, Rxz, Ryy, Ryz, Rzz,
-                  Symmetry);
-#endif
-    f_constraint_bssn(ex, X, Y, Z,
-                      chi, trK,
-                      dxx, gxy, gxz, dyy, gyz, dzz,
-                      Axx, Axy, Axz, Ayy, Ayz, Azz,
-                      Gamx, Gamy, Gamz,
-                      Lap, betax, betay, betaz, rho, Sx, Sy, Sz,
-                      Gamxxx, Gamxxy, Gamxxz, Gamxyy, Gamxyz, Gamxzz,
-                      Gamyxx, Gamyxy, Gamyxz, Gamyyy, Gamyyz, Gamyzz,
-                      Gamzxx, Gamzxy, Gamzxz, Gamzyy, Gamzyz, Gamzzz,
-                      Rxx, Rxy, Rxz, Ryy, Ryz, Rzz,
-                      Hcon, Mxcon, Mycon, Mzcon, Gmxcon, Gmycon, Gmzcon,
-                      Symmetry);
-    return ret;
-}

AMSS_NCKU_source/z4c_rhs_cuda.h

@@ -1,114 +0,0 @@
-#ifndef Z4C_RHS_CUDA_H
-#define Z4C_RHS_CUDA_H
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-enum {
-    Z4C_CUDA_STATE_COUNT = 25
-};
-
-int z4c_cuda_rk4_substep(void *block_tag,
-                         int *ex, double *X, double *Y, double *Z,
-                         double **state_host_in,
-                         double **state_host_out,
-                         const double *propspeed,
-                         const double *soa_flat,
-                         const double *bbox,
-                         double &dT,
-                         double &T,
-                         int &RK4,
-                         int &apply_bam_bc,
-                         int &Symmetry,
-                         int &Lev,
-                         double &eps,
-                         int &co,
-                         int &keep_resident_state,
-                         int &apply_enforce_ga,
-                         double &chitiny);
-
-int z4c_cuda_download_resident_state(void *block_tag,
-                                     int *ex,
-                                     double **state_host_out);
-
-int z4c_cuda_pack_state_region_to_host_buffer(void *block_tag,
-                                              int state_index,
-                                              double *host_buffer,
-                                              int *ex,
-                                              int i0, int j0, int k0,
-                                              int sx, int sy, int sz);
-
-int z4c_cuda_unpack_state_region_from_host_buffer(void *block_tag,
-                                                  int state_index,
-                                                  double *host_buffer,
-                                                  int *ex,
-                                                  int i0, int j0, int k0,
-                                                  int sx, int sy, int sz);
-
-int z4c_cuda_pack_state_batch_to_host_buffer(void *block_tag,
-                                             int state_count,
-                                             double *host_buffer,
-                                             int *ex,
-                                             int i0, int j0, int k0,
-                                             int sx, int sy, int sz);
-
-int z4c_cuda_unpack_state_batch_from_host_buffer(void *block_tag,
-                                                 int state_count,
-                                                 double *host_buffer,
-                                                 int *ex,
-                                                 int i0, int j0, int k0,
-                                                 int sx, int sy, int sz);
-
-int z4c_cuda_pack_state_batch_to_device_buffer(void *block_tag,
-                                               int state_count,
-                                               double *device_buffer,
-                                               int *ex,
-                                               int i0, int j0, int k0,
-                                               int sx, int sy, int sz);
-
-int z4c_cuda_unpack_state_batch_from_device_buffer(void *block_tag,
-                                                   int state_count,
-                                                   double *device_buffer,
-                                                   int *ex,
-                                                   int i0, int j0, int k0,
-                                                   int sx, int sy, int sz);
-
-int z4c_cuda_restrict_state_batch_to_device_buffer(void *block_tag,
-                                                   int state_count,
-                                                   double *device_buffer,
-                                                   int *ex,
-                                                   int sx, int sy, int sz,
-                                                   int fi0, int fj0, int fk0,
-                                                   const double *state_soa);
-
-int z4c_cuda_prolong_state_batch_to_device_buffer(void *block_tag,
-                                                  int state_count,
-                                                  double *device_buffer,
-                                                  int *ex,
-                                                  int sx, int sy, int sz,
-                                                  int ii0, int jj0, int kk0,
-                                                  int lbc_i, int lbc_j, int lbc_k,
-                                                  const double *state_soa);
-
-int z4c_cuda_download_state_subset(void *block_tag,
-                                   int *ex,
-                                   int subset_count,
-                                   const int *state_indices,
-                                   double **state_host_out);
-
-int z4c_cuda_upload_state_subset(void *block_tag,
-                                 int *ex,
-                                 int subset_count,
-                                 const int *state_indices,
-                                 double **state_host_in);
-
-int z4c_cuda_has_resident_state(void *block_tag);
-
-void z4c_cuda_release_step_ctx(void *block_tag);
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif

BSSN_BUILD_CONFIG_MIGRATION.md

@@ -0,0 +1,211 @@
+# BSSN Build Config Migration
+
+This note records the build-configuration fix needed when replacing
+`AMSS_NCKU_Input.py` or `generate_macrodef.py` with a newer upstream version.
+
+## Problem
+
+`AMSS_NCKU_source/macrodef.h` is not the authoritative file used by normal
+runs. `AMSS_NCKU_Program.py` first generates macro files under
+`input_data.File_directory`, copies `AMSS_NCKU_source` to
+`<File_directory>/AMSS_NCKU_source_copy`, then copies the generated macro files
+into that copied source tree and compiles there.
+
+Therefore, makefile logic must not depend only on the stale
+`AMSS_NCKU_source/macrodef.h`. The actual equation path must be passed to the
+copied build tree from the same generation step that creates `macrodef.h`.
+
+The performance regression was caused by compiling/linking the
+`BSSN-EScalar` C wrapper into BSSN vacuum builds. For BSSN vacuum (`ABEtype=0`),
+the build must use:
+
+```make
+BSSN_USE_TRANSFER_CACHE=1
+BSSN_USE_ESCALAR_C_KERNEL=0
+```
+
+and must not link `bssn_escalar_rhs_c.o`.
+
+## Required Migration Steps
+
+### 1. Add an ABE type helper in `generate_macrodef.py`
+
+Add a helper that maps `input_data.Equation_Class` to the numeric `ABEtype`.
+Use the same mapping as `macrodef.h`:
+
+```python
+def get_abe_type():
+    if ( input_data.Equation_Class == "BSSN" ):
+        return 0
+    elif ( input_data.Equation_Class == "BSSN-EScalar" ):
+        return 1
+    elif ( input_data.Equation_Class == "BSSN-EM" ):
+        return 3
+    elif ( input_data.Equation_Class == "Z4C" ):
+        return 2
+    else:
+        raise ValueError("Equation_Class setting error!!!")
+```
+
+Update `generate_macrodef_h()` to print `#define ABEtype {get_abe_type()}`
+instead of duplicating the if/elif mapping.
+
+### 2. Generate a makefile fragment
+
+In `generate_macrodef.py`, add:
+
+```python
+def generate_build_config():
+    file1 = open(os.path.join(input_data.File_directory, "AMSS_NCKU_build.mk"), "w")
+    print("# Generated by generate_macrodef.py; do not edit manually.", file=file1)
+    print(f"ABE_TYPE := {get_abe_type()}", file=file1)
+    file1.close()
+```
+
+This file is the build-time authority for the equation path.
+
+### 3. Call and copy the generated build config
+
+In `AMSS_NCKU_Program.py`, after generating `macrodef.h` and `macrodef.fh`, call:
+
+```python
+generate_macrodef.generate_build_config()
+print(" AMSS-NCKU build config AMSS_NCKU_build.mk has been generated. ")
+```
+
+When copying generated files into `AMSS_NCKU_source_copy`, also copy:
+
+```python
+build_config_path = os.path.join(File_directory, "AMSS_NCKU_build.mk")
+shutil.copy2(build_config_path, AMSS_NCKU_source_copy)
+```
+
+### 4. Make the source makefile consume the generated config
+
+At the top of `AMSS_NCKU_source/makefile`, after `include makefile.inc`, add:
+
+```make
+-include AMSS_NCKU_build.mk
+
+ABE_TYPE ?= $(shell awk '/^[[:space:]]*\#define[[:space:]]+ABEtype/ {print $$3; exit}' macrodef.h 2>/dev/null)
+```
+
+The generated `AMSS_NCKU_build.mk` is used during normal Python-driven builds.
+The fallback keeps manual source-tree builds usable.
+
+### 5. Gate path-specific build options by `ABE_TYPE`
+
+Use effective build switches:
+
+```make
+ifeq ($(USE_TRANSFER_CACHE),auto)
+ifeq ($(ABE_TYPE),0)
+EFFECTIVE_USE_TRANSFER_CACHE = 1
+else
+EFFECTIVE_USE_TRANSFER_CACHE = 0
+endif
+else
+EFFECTIVE_USE_TRANSFER_CACHE = $(USE_TRANSFER_CACHE)
+endif
+
+ifeq ($(USE_CXX_ESCALAR_KERNEL),1)
+ifeq ($(ABE_TYPE),1)
+EFFECTIVE_USE_CXX_ESCALAR_KERNEL = 1
+else
+EFFECTIVE_USE_CXX_ESCALAR_KERNEL = 0
+endif
+else
+EFFECTIVE_USE_CXX_ESCALAR_KERNEL = 0
+endif
+
+TRANSFER_CACHE_FLAG = -DBSSN_USE_TRANSFER_CACHE=$(EFFECTIVE_USE_TRANSFER_CACHE)
+ESCALAR_KERNEL_FLAG = -DBSSN_USE_ESCALAR_C_KERNEL=$(EFFECTIVE_USE_CXX_ESCALAR_KERNEL)
+```
+
+Only add `bssn_escalar_rhs_c.o` when the effective EScalar C kernel switch is
+enabled:
+
+```make
+ifeq ($(EFFECTIVE_USE_CXX_ESCALAR_KERNEL),1)
+CFILES += bssn_escalar_rhs_c.o
+endif
+```
+
+### 6. Use safe transfer-cache default
+
+In `AMSS_NCKU_source/makefile.inc`, keep:
+
+```make
+USE_TRANSFER_CACHE ?= auto
+```
+
+With the effective switch logic above, this enables cached transfer for BSSN
+vacuum while keeping non-BSSN paths on the uncached path by default.
+
+## Verification Checklist
+
+Run these checks after migrating:
+
+```bash
+python3 -c "import generate_macrodef; generate_macrodef.generate_build_config()"
+cat GW150914/AMSS_NCKU_build.mk
+```
+
+For BSSN, the generated file should contain:
+
+```make
+ABE_TYPE := 0
+```
+
+Dry-run the copied or source makefile:
+
+```bash
+make -n -B INTERP_LB_MODE=off ABE | grep -E 'BSSN_USE_TRANSFER_CACHE|BSSN_USE_ESCALAR_C_KERNEL|bssn_escalar_rhs_c'
+```
+
+Expected BSSN result:
+
+```text
+-DBSSN_USE_TRANSFER_CACHE=1 -DBSSN_USE_ESCALAR_C_KERNEL=0
+```
+
+and no `bssn_escalar_rhs_c.o` in the final link command.
+
+Run the full workflow:
+
+```bash
+python3 AMSS_NCKU_Program.py
+```
+
+For the 10-step BSSN test, compare coordinate output:
+
+```bash
+python3 - <<'PY'
+from pathlib import Path
+old = Path('../GW150914-06457/AMSS_NCKU_output/bssn_BH.dat')
+new = Path('GW150914/AMSS_NCKU_output/bssn_BH.dat')
+
+def rows(path):
+    out = []
+    for line in path.read_text().splitlines():
+        if not line.strip() or line.lstrip().startswith('#'):
+            continue
+        out.append([float(x) for x in line.split()])
+    return out
+
+ro, rn = rows(old), rows(new)
+n = min(len(ro), len(rn))
+max_abs = 0.0
+for i in range(n):
+    for a, b in zip(ro[i], rn[i]):
+        max_abs = max(max_abs, abs(a - b))
+print(f"old_rows={len(ro)} new_rows={len(rn)} compared_rows={n}")
+print(f"max_abs_diff={max_abs:.17g}")
+PY
+```
+
+For the validated migration, the first 10 rows matched exactly:
+
+```text
+max_abs_diff=0
+```

code_modification_readme.md

@@ -1,224 +0,0 @@
-# Code Modification Readme — `asc26-plan-a`
-
-**Baseline branch:** `baseline`
-**Target branch:** `asc26-plan-a`
-**Date:** 2026-05-19
-
----
-
-## Overview
-
-This branch delivers two major performance overhauls to the AMSS-NCKU numerical relativity codebase:
-
-1. **TwoPunctureABE Multithreading** — OpenMP parallelization of the TwoPunctures initial-data solver, combined with a BLAS-driven spectral derivative engine, MKL/LAPACK integration, and C/C++ rewrites of hot Fortran kernel subroutines.
-
-2. **ABE GPU Rewrite** — Complete replacement of the legacy `bssn_gpu_class` abstraction layer with direct, monolithic CUDA kernels for BSSN, Z4C, and Shell-Patch evolution, plus GPU-resident state management and CUDA-aware MPI.
-
-**Total diff:** 84 files changed, +57,919 / −33,795 lines.
-
----
-
-## Part 1 — TwoPunctureABE Multithreading
-
-### 1.1 Spectral Derivative Engine: BLAS Matrix-Multiplication Rewrite
-
-**Files:** `AMSS_NCKU_source/TwoPunctures.C`, `AMSS_NCKU_source/TwoPunctures.h`
-
-The original `Derivatives_AB3` computed spectral derivatives (Chebyshev in A/B, Fourier in phi) with nested scalar loops over every grid point. The new `Derivatives_AB3_MatMul` expresses all derivatives as matrix-matrix products over pencil-shaped data slices, dispatched to Intel MKL `cblas_dgemm`.
-
-- **Precomputed derivative matrices** — `precompute_derivative_matrices()` builds `D1_A`, `D2_A`, `D1_B`, `D2_B` (Chebyshev collocation derivative matrices) and `DF1_phi`, `DF2_phi` (Fourier derivative matrices) once at construction time.
-- **Pencil-based GEMM** — data is gathered into 2D arrays where one dimension is the spectral direction and the other enumerates all remaining degrees of freedom (variables × orthogonal grid indices). Each derivative direction becomes a single `cblas_dgemm` call. The pure derivatives (d/dA, d/dB, d/dphi) and all mixed derivatives (d²/dAdB, d²/dAdphi, d²/dBdphi) are computed this way.
-- **`build_cheb_deriv_matrices` / `build_fourier_deriv_matrices`** — construct the standard Chebyshev and Fourier collocation derivative matrices.
-
-### 1.2 OpenMP Parallelization of TwoPunctures
-
-**Files:** `AMSS_NCKU_source/TwoPunctures.C`, `AMSS_NCKU_source/TwoPunctures.h`
-
-Three critical regions are parallelized:
-
-| Region | Directive | Strategy |
-|--------|-----------|----------|
-| `F_of_v` residual evaluation | `#pragma omp parallel for collapse(3) schedule(dynamic,1)` | Each (i,j,k) thread stack-allocates its own `l_U` (derivs struct) and `l_values[]` to eliminate heap contention and data races |
-| `relax_omp` line relaxation | `#pragma omp parallel for schedule(static)` over k-slices | Alternating be/al sweeps, each thread uses pre-allocated per-thread Thomas-algorithm workspace (`ws_*_be[tid]`, `ws_*_al[tid]`) |
-| `LineRelax_be_omp` / `LineRelax_al_omp` | Called from `relax_omp` with explicit `tid` | Thread-safe tridiagonal solves using the thread's private scratch arrays |
-
-**Per-thread workspace** — `allocate_workspace()` allocates independent Thomas-algorithm buffers (`diag`, `e`, `f`, `b`, `x`, `l`, `u`, `d`, `y`) for each OpenMP thread in both be and al directions, avoiding lock contention in the inner Newton iteration.
-
-### 1.3 MKL BLAS / LAPACK Integration
-
-**Files:** `AMSS_NCKU_source/TwoPunctures.C`, `AMSS_NCKU_source/gaussj.C`
-
-| Function | Old | New | Benefit |
-|----------|-----|-----|---------|
-| `norm2` | scalar `sqrt(sum(v[i]²))` loop | `cblas_dnrm2` | BLAS Level 1, SIMD-optimized |
-| `scalarproduct` | scalar `sum(v[i]*w[i])` loop | `cblas_ddot` | BLAS Level 1, SIMD-optimized |
-| `gaussj` | hand-written Gauss-Jordan elimination (~100 lines) | `LAPACKE_dgesv` + `LAPACKE_dgetrf` + `LAPACKE_dgetri` | LAPACK LU with partial pivoting, asymptotically faster for the `n~50` matrix sizes used in spectral elliptic solves |
-
-### 1.4 C/C++ Rewrite of Hot Fortran Kernels
-
-**Files (new):**
-- `AMSS_NCKU_source/fderivs_c.C` (167 lines) — first derivatives, 2nd/4th order
-- `AMSS_NCKU_source/fdderivs_c.C` (332 lines) — second derivatives, 2nd/4th order
-- `AMSS_NCKU_source/kodiss_c.C` (117 lines) — Kreiss-Oliger dissipation
-- `AMSS_NCKU_source/lopsided_c.C` (255 lines) — lopsided advection
-- `AMSS_NCKU_source/lopsided_kodis_c.C` (248 lines) — fused advection + dissipation
-- `AMSS_NCKU_source/rungekutta4_rout_c.C` (212 lines) — RK4 time-stepper
-- `AMSS_NCKU_source/bssn_rhs_c.C` (1,287 lines) — full BSSN RHS kernel
-- `AMSS_NCKU_source/z4c_rhs_c.C` (725 lines) — full Z4C RHS kernel
-
-Every C rewrite follows a consistent optimization pattern:
-- **64-byte aligned allocation** (`aligned_alloc(64, ...)`) for AVX-512 compatibility.
-- **Static buffer caching** — scratch arrays (e.g., the padded `fh` ghost-zone buffer) persist across calls via a `static` pointer + capacity check, avoiding repeated `malloc`/`free`.
-- **Two-pass strategy** — 2nd-order finite differences are computed on the full domain first, then the interior sub-volume is overwritten with 4th-order stencils. This eliminates the per-point `if/elseif` branching of the original Fortran.
-- **Non-overlapping shell pass** — in `fdderivs_c.C`, the 2nd-order pass skips points that will be overwritten by the 4th-order pass, avoiding redundant computation.
-
-### 1.5 Fortran Kernel Fusion: lopsided_kodis
-
-**File:** `AMSS_NCKU_source/lopsidediff.f90`
-
-A new `lopsided_kodis` subroutine fuses the advection (lopsided) and Kreiss-Oliger dissipation (kodis) operators into a single pass over the grid. Both operators previously called `symmetry_bd` independently to fill ghost zones — the fused version calls it once and shares the padded `fh` array, halving ghost-zone fill overhead for this hot path.
-
-### 1.6 Build System for TwoPunctures
-
-**Files:** `AMSS_NCKU_source/makefile`, `AMSS_NCKU_source/makefile.inc`
-
-- **`TP_OPTFLAGS`** — TwoPunctures and TwoPunctureABE are compiled with a dedicated, more aggressive optimization flag set (`-O3 -march=znver5 -fp-model fast=2 -fma -ipo`) separate from the main code.
-- **`USE_CXX_KERNELS`** — selects between the C rewrites and the original Fortran kernels (`bssn_rhs.f90`, etc.) for the CPU path.
-- **`USE_CXX_RK4`** — independently selects between the C and Fortran RK4 stepper.
-- **Intel oneTBB allocator** (`libtbbmalloc.so`) — replaces the system `malloc` with a scalable thread-safe allocator, critical for multi-threaded TwoPunctures performance.
-- **PGO support** — `PGO_MODE=opt|instrument` for profile-guided optimization (currently disabled after testing showed negative benefit).
-- **Toolchains** — Intel oneAPI (`TOOLCHAIN=intel`, default) and NVIDIA HPC SDK (`TOOLCHAIN=nvhpc`).
-
----
-
-## Part 2 — ABE GPU Rewrite
-
-### 2.1 Architecture: From Class Wrapper to Direct CUDA Kernels
-
-The old GPU path (`baseline`) was organized as:
-
-```
-bssn_gpu_class.C/h    — C++ class managing GPU state and kernel launches
-bssn_step_gpu.C       — RK4 stepper with per-substep GPU/CPU synchronisation
-bssn_gpu.cu           — CUDA kernel implementations called through the class
-```
-
-The new GPU path (`asc26-plan-a`) replaces all of the above with:
-
-```
-bssn_rhs_cuda.cu/h    — 10,381-line monolithic CUDA BSSN RHS kernel
-z4c_rhs_cuda.cu/h     —  7,909-line monolithic CUDA Z4C RHS kernel
-fd_cuda_helpers.cuh   —    412-line shared finite-difference device functions
-bssn_gpu_rhs_ss.cu    —  (retained, lightly modified) Shell-Patch GPU RHS
-```
-
-**Key architectural differences:**
-- The old `bssn_gpu_class` managed GPU memory through a C++ class with explicit allocate/free/sync methods scattered across the time-stepping logic. The new kernels operate directly on raw device pointers with a clear resident/transient memory model.
-- The old code launched many small kernels (one per derivative or algebraic term). The new code is a **single monolithic kernel per formulation** — all 24 BSSN evolution variables are computed in one launch with on-the-fly finite differences, eliminating kernel-launch latency and intermediate global-memory round-trips.
-- The old `bssn_step_gpu.C` performed per-substep GPU→CPU downloads for boundary conditions and analysis. The new model supports **GPU-resident state** — variables stay on device across timesteps unless explicitly requested.
-
-### 2.2 GPU-Resident State Model
-
-A central theme across ~20 commits is the "resident-sync" optimization:
-
-| Commit | What it does |
-|--------|-------------|
-| `22c1e71` | Optimize BSSN CUDA resident state and CUDA-aware MPI |
-| `090d865` | Optimize BSSN CUDA state transfers |
-| `68eab03` | Add opt-in BSSN CUDA resident AMR path |
-| `1ee229a` | Add keyed BSSN CUDA resident banks |
-| `18e9c9c` | Optimize BSSN CUDA resident AMR prolong |
-| `8486532` | Add resident BSSN GPU point interpolation |
-| `b1974ef` | Stabilize device AMR restrict across regrid |
-| `ae64a22` | Complete BSSN-EScalar CUDA resident transfers |
-| `83afaf1` | Skip zero EM resident downloads |
-| `35b6cef` | Broaden cached CUDA sync paths |
-
-The resident model works as follows:
-- BSSN grid functions are allocated once on the GPU and persist across timesteps.
-- Inter-processor ghost-zone exchanges use **CUDA-aware MPI** — MPI directly reads/writes device memory without staging through host buffers.
-- AMR prolongation and restriction operate directly on device memory.
-- Boundary conditions and analysis routines download only the specific slices/points they need, not the full grid.
-- When EM fields are zero (pure-gravity runs), EM downloads are skipped entirely.
-
-### 2.3 Z4C and Shell-Patch GPU Acceleration
-
-**Files:** `AMSS_NCKU_source/z4c_rhs_cuda.cu`, `AMSS_NCKU_source/bssn_gpu_rhs_ss.cu`
-
-- The Z4C constraint-damped formulation gets its own 7,909-line monolithic CUDA kernel (`z4c_rhs_cuda.cu`), matching the BSSN kernel's architecture.
-- **Shell-Patch GPU acceleration** — the spherical shell boundary patches now compute on GPU with dedicated kernels in `bssn_gpu_rhs_ss.cu`.
-- Z4C + Shell-Patch can coexist on GPU (Phase 3 commits).
-- A CPU-side wrapper (`z4c_rhs_c.C`) handles the trKd + TZ_rhs contribution that remains on CPU, minimizing GPU/CPU traffic.
-
-### 2.4 Finite-Difference Order Flexibility
-
-**File:** `AMSS_NCKU_source/fd_cuda_helpers.cuh`
-
-Shared device functions for finite-difference stencils support **2nd, 4th, 6th, and 8th order** at compile time via preprocessor switches. This enables:
-- Per-run selection of convergence order without recompilation of the full kernel.
-- 8th-order AMR transfers (`1064a68`) for BSSN-EM.
-- 6th-order optimized AMR stencils (`0076b3c`).
-
-### 2.5 GPU Diagnostics and Quality Assurance
-
-**File:** `AMSS_NCKU_GPUCheck.py` (559 lines, new)
-
-A Python-based GPU correctness verification tool that compares GPU and CPU evolution outputs. The GPU build pipeline includes optional kernel profiling switches (`7683459`) for performance debugging.
-
-**GPU-specific bug fixes:**
-- `f226498` — Fix CUDA AMR symmetry drift (incorrect ghost-zone handling under symmetry boundary conditions)
-- `2317e4a` — Fix BSSN GPU resident AMR sync default
-- `fea2dcc` — Fix BSSN-EM runtime crash
-- `dd0e20d` — Fix BSSN-EScalar CUDA boundary and scalar KO
-- `5eb4994` — Fix AHF crash under CUDA resident-sync mode
-
-### 2.6 Build Integration
-
-**Makefile switches:**
-- `USE_CUDA_BSSN=0/1` — route BSSN RHS through GPU or CPU
-- `USE_CUDA_Z4C=0/1` — route Z4C RHS through GPU or CPU
-- `CUDA_ARCH=sm_80` — target NVIDIA Ampere (A100)
-- `NVHPC_ROOT` — path to NVIDIA HPC SDK for the `nvcc` compiler wrapper
-- CUDA compilation flags: `-O3 --ptxas-options=-v -arch=$(CUDA_ARCH)`
-
----
-
-## Part 3 — Shared Infrastructure
-
-### 3.1 Interp_Points Load-Balance Profiler
-
-**Files:** `AMSS_NCKU_source/interp_lb_profile.C`, `interp_lb_profile.h`, `interp_lb_profile_data.h`, `generate_interp_lb_header.py`
-
-A two-pass instrumentation system for load-balancing the `Interp_Points` parallel interpolation routine:
-- **Pass 1** (`INTERP_LB_MODE=profile`): instrument each MPI rank's interpolation calls with timing, write a binary profile.
-- **Pass 2** (`INTERP_LB_MODE=optimize`): read the profile and rebalance work across MPI ranks.
-
-### 3.2 Helper Headers
-
-**Files:** `AMSS_NCKU_source/tool.h` (33 lines), `AMSS_NCKU_source/share_func.h` (246 lines)
-
-- `tool.h` — shared indexing macros (`idx_ex`, `idx_fh_F_ord2`) and the `symmetry_bd` declaration used by all C kernel rewrites.
-- `share_func.h` — common utility functions shared across the C++ source files.
-
-### 3.3 Plot-Only Restart Script
-
-**File:** `parallel_plot_helper.py` (29 lines)
-
-A lightweight restart script that skips recomputation when plotting was interrupted — reads existing checkpoint data and replots without re-running the simulation.
-
----
-
-## Performance Summary
-
-| Component | Optimization | Expected Impact |
-|-----------|-------------|-----------------|
-| TwoPunctures `Derivatives_AB3` | Scalar loops → MKL GEMM | 5-20× speedup for spectral derivative computation |
-| TwoPunctures `F_of_v` | OpenMP collapse(3) + stack-local variables | Near-linear scaling with core count for residual evaluation |
-| TwoPunctures `gaussj` | Hand-written Gauss-Jordan → LAPACK LU | 2-5× speedup for N~50 matrix inversion |
-| BSSN RHS (GPU) | Many small kernels → one monolithic kernel | Eliminates kernel-launch overhead; 2-5× GPU throughput improvement |
-| GPU state transfers | Per-step download → resident model | Eliminates ~80% of GPU↔CPU PCIe traffic |
-| `lopsided_kodis` fusion | Two `symmetry_bd` calls → one shared call | ~30% reduction in ghost-zone fill cost for this operator pair |
-| Memory allocator | System malloc → Intel TBB malloc | Significant reduction in malloc contention under OpenMP |
-| C kernel rewrites | Fortran → C with aligned alloc + static buffers | Enables Intel compiler IPO across C/C++/Fortran boundaries; better SIMD codegen |
-
----

generate_macrodef.py

@@ -12,6 +12,37 @@ import os
 import AMSS_NCKU_Input as input_data          ## import program input file
 
 
+##################################################################
+
+def get_abe_type():
+    if ( input_data.Equation_Class == "BSSN" ):
+        return 0
+    elif ( input_data.Equation_Class == "BSSN-EScalar" ):
+        return 1
+    elif ( input_data.Equation_Class == "BSSN-EM" ):
+        return 3
+    elif ( input_data.Equation_Class == "Z4C" ):
+        return 2
+    else:
+        raise ValueError("Equation_Class setting error!!!")
+
+
+##################################################################
+
+## Generate the makefile fragment used by the copied source tree.
+## The source-tree macrodef.h is not authoritative because macro files
+## are regenerated under File_directory for each run.
+
+def generate_build_config():
+
+    file1 = open( os.path.join(input_data.File_directory, "AMSS_NCKU_build.mk"), "w")
+
+    print( "# Generated by generate_macrodef.py; do not edit manually.", file=file1 )
+    print( f"ABE_TYPE := {get_abe_type()}",                             file=file1 )
+
+    file1.close()
+
+
 ##################################################################
 
 ## Generate the macro file macrodef.h according to user settings
@@ -58,19 +89,10 @@ def generate_macrodef_h():
     # 2: Z4c vacuum
     # 3: coupled to Maxwell field
 
-    if ( input_data.Equation_Class == "BSSN" ):
-        print( "#define ABEtype 0", file=file1 )
-        print(                      file=file1 )
-    elif ( input_data.Equation_Class == "BSSN-EScalar" ):
-        print( "#define ABEtype 1", file=file1 )
-        print(                      file=file1 )
-    elif ( input_data.Equation_Class == "BSSN-EM" ):
-        print( "#define ABEtype 3", file=file1 )
-        print(                      file=file1 )
-    elif ( input_data.Equation_Class == "Z4C" ):
-        print( "#define ABEtype 2", file=file1 )
-        print(                      file=file1 )
-    else:
+    try:
+        print( f"#define ABEtype {get_abe_type()}", file=file1 )
+        print(                                      file=file1 )
+    except ValueError:
         print( "Equation_Class setting error!!!"                )
         print()
         print( "# Equation type #define ABEtype setting error!!!", file=file1 )
@@ -204,7 +226,7 @@ def generate_macrodef_h():
     # use GPU or not
     
     if ( input_data.GPU_Calculation == "yes"):
-        print( "//#define USE_GPU",   file=file1 )
+        print( "#define USE_GPU",   file=file1 )
         print(                      file=file1 )
     elif ( input_data.GPU_Calculation == "no"):
         print( "//#define USE_GPU", file=file1 )

makefile_and_run.py

@@ -9,8 +9,6 @@
 
 
 import AMSS_NCKU_Input as input_data
-import os
-import shutil
 import subprocess
 import time
 
@@ -45,7 +43,8 @@ def get_last_n_cores_per_socket(n=32):
     cpu_str = ",".join(segments)
     total = len(segments) * n
     print(f" CPU binding: taskset -c {cpu_str}  ({total} cores, last {n} per socket)")
-    return f"taskset -c {cpu_str}" if cpu_str else ""
+    #return f"taskset -c {cpu_str}"
+    return f""
 
 
 ## CPU core binding: dynamically select the last 32 cores of each socket (64 cores total)
@@ -57,231 +56,6 @@ BUILD_JOBS = 64
 
 ##################################################################
 
-def _truthy(value, default=False):
-    if value is None:
-        return default
-    if isinstance(value, bool):
-        return value
-    text = str(value).strip().lower()
-    if text == "":
-        return default
-    return text in ("1", "yes", "y", "true", "on", "enable", "enabled")
-
-
-def _input_or_env(input_name, env_name, default=None):
-    if env_name in os.environ:
-        return os.environ[env_name]
-    return getattr(input_data, input_name, default)
-
-
-def _input_env_passthrough(runtime_env, env_name):
-    if env_name in runtime_env:
-        return
-    if hasattr(input_data, env_name):
-        runtime_env[env_name] = str(getattr(input_data, env_name))
-
-
-def _start_cuda_mps_if_requested(runtime_env):
-    if input_data.GPU_Calculation != "yes":
-        return False
-
-    default_auto_mps = int(getattr(input_data, "MPI_processes", 1)) > 1
-    auto_mps = _truthy(
-        _input_or_env("CUDA_Auto_MPS", "AMSS_CUDA_AUTO_MPS", default_auto_mps),
-        default=default_auto_mps,
-    )
-    if not auto_mps:
-        return False
-
-    mps_control = shutil.which("nvidia-cuda-mps-control")
-    if not mps_control:
-        print(" CUDA MPS control command was not found; running without MPS.")
-        return False
-
-    uid = os.getuid()
-    pipe_dir = str(_input_or_env("CUDA_MPS_PIPE_DIRECTORY", "CUDA_MPS_PIPE_DIRECTORY",
-                                 f"/tmp/amss-ncku-mps-{uid}"))
-    log_dir = str(_input_or_env("CUDA_MPS_LOG_DIRECTORY", "CUDA_MPS_LOG_DIRECTORY",
-                                f"/tmp/amss-ncku-mps-log-{uid}"))
-    os.makedirs(pipe_dir, exist_ok=True)
-    os.makedirs(log_dir, exist_ok=True)
-
-    mps_env = runtime_env.copy()
-    mps_env["CUDA_MPS_PIPE_DIRECTORY"] = pipe_dir
-    mps_env["CUDA_MPS_LOG_DIRECTORY"] = log_dir
-
-    if os.path.exists(os.path.join(pipe_dir, "control")):
-        runtime_env.update({
-            "CUDA_MPS_PIPE_DIRECTORY": pipe_dir,
-            "CUDA_MPS_LOG_DIRECTORY": log_dir,
-        })
-        print(f" Reusing CUDA MPS daemon: {pipe_dir}")
-        return False
-
-    print(f" Starting CUDA MPS daemon for this run: {pipe_dir}")
-    result = subprocess.run([mps_control, "-d"], env=mps_env, text=True,
-                            stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
-    if result.returncode != 0:
-        print(" CUDA MPS daemon did not start; running without MPS.")
-        if result.stdout:
-            print(result.stdout, end="")
-        return False
-
-    runtime_env.update({
-        "CUDA_MPS_PIPE_DIRECTORY": pipe_dir,
-        "CUDA_MPS_LOG_DIRECTORY": log_dir,
-    })
-    return True
-
-
-def _stop_cuda_mps(runtime_env):
-    mps_control = shutil.which("nvidia-cuda-mps-control")
-    if not mps_control:
-        return
-    subprocess.run([mps_control], input="quit\n", env=runtime_env, text=True,
-                   stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
-
-
-def _gpu_runtime_env():
-    runtime_env = os.environ.copy()
-    original_env = set(os.environ.keys())
-    finite_difference = str(getattr(input_data, "Finite_Diffenence_Method", "4th-order")).strip()
-
-    defaults = {
-        "AMSS_EVOLVE_TIMING": "0",
-        "AMSS_ESCALAR_STEP_TIMING": "0",
-        "AMSS_INTERP_FAST": "1",
-        "AMSS_INTERP_GPU": "1",
-        "AMSS_ANALYSIS_MAP_EVERY": "1000000",
-        "AMSS_CUDA_AWARE_MPI": "1",
-        "AMSS_CUDA_KEEP_RESIDENT_AFTER_STEP": "1",
-        "AMSS_CUDA_KEEP_ALL_LEVELS": "1",
-        "AMSS_CUDA_ESCALAR_KEEP_RESIDENT_AFTER_STEP": "1",
-        "AMSS_CUDA_ESCALAR_KEEP_ALL_LEVELS": "1",
-        "AMSS_CUDA_EM_CACHE_SOURCES": "1",
-        "AMSS_CUDA_EM_ZERO_FASTPATH": "1",
-        "AMSS_EM_ZERO_ANALYSIS_FASTPATH": "1",
-        "AMSS_EM_ZERO_RESIDENT_DOWNLOAD_FASTPATH": "1",
-        "AMSS_CUDA_AMR_HOST_STAGED": "1",
-        "AMSS_CUDA_AMR_RESTRICT_DEVICE": "0",
-        "AMSS_CUDA_AMR_RESTRICT_BATCH": "0",
-        "AMSS_CUDA_DEVICE_SEGMENT_BATCH": "0",
-        "AMSS_CUDA_UNCACHED_DEVICE_BUFFERS": "1",
-        "AMSS_SHELL_FAST_INTERP": "0",
-        "AMSS_SHELL_PARALLEL_INTERP": "0",
-        "AMSS_SHELL_CUDA_INTERP": "0",
-    }
-    if finite_difference in ("2nd-order", "8th-order"):
-        defaults.update({
-            "AMSS_INTERP_FAST": "0",
-            "AMSS_INTERP_GPU": "0",
-            "AMSS_CUDA_AWARE_MPI": "0",
-        })
-    if finite_difference == "8th-order" and getattr(input_data, "Equation_Class", "") == "BSSN-EM":
-        defaults.update({
-            "AMSS_CUDA_AMR_RESTRICT_DEVICE": "1",
-            "AMSS_CUDA_AMR_RESTRICT_BATCH": "1",
-            "AMSS_CUDA_DEVICE_SEGMENT_BATCH": "1",
-        })
-    if getattr(input_data, "basic_grid_set", "") == "Shell-Patch":
-        defaults.update({
-            "AMSS_CUDA_AWARE_MPI": "0",
-            "AMSS_SHELL_FAST_INTERP": "1",
-            "AMSS_SHELL_PARALLEL_INTERP": "1",
-            "AMSS_SHELL_INTERP_THREADS": "16",
-        })
-    if getattr(input_data, "Equation_Class", "") in ("BSSN", "BSSN-EScalar", "Z4C"):
-        defaults["AMSS_CUDA_AMR_RESTRICT_DEVICE"] = "1"
-    if getattr(input_data, "Equation_Class", "") == "Z4C":
-        defaults.update({
-            "AMSS_Z4C_CUDA_RESIDENT": "1",
-            "AMSS_CONSTRAINT_OUT_EVERY": "1000000",
-        })
-    for key, value in defaults.items():
-        runtime_env.setdefault(key, value)
-
-    input_overrides = [
-        "AMSS_EVOLVE_TIMING",
-        "AMSS_ESCALAR_STEP_TIMING",
-        "AMSS_INTERP_FAST",
-        "AMSS_INTERP_GPU",
-        "AMSS_ANALYSIS_MAP_EVERY",
-        "AMSS_CUDA_AWARE_MPI",
-        "AMSS_CUDA_KEEP_RESIDENT_AFTER_STEP",
-        "AMSS_CUDA_KEEP_ALL_LEVELS",
-        "AMSS_CUDA_ESCALAR_KEEP_RESIDENT_AFTER_STEP",
-        "AMSS_CUDA_ESCALAR_KEEP_ALL_LEVELS",
-        "AMSS_CUDA_EM_CACHE_SOURCES",
-        "AMSS_CUDA_EM_ZERO_FASTPATH",
-        "AMSS_EM_ZERO_ANALYSIS_FASTPATH",
-        "AMSS_EM_ZERO_RESIDENT_DOWNLOAD_FASTPATH",
-        "AMSS_CUDA_AMR_HOST_STAGED",
-        "AMSS_CUDA_AMR_RESTRICT_DEVICE",
-        "AMSS_CUDA_AMR_RESTRICT_BATCH",
-        "AMSS_CUDA_DEVICE_SEGMENT_BATCH",
-        "AMSS_CUDA_UNCACHED_DEVICE_BUFFERS",
-        "AMSS_SHELL_FAST_INTERP",
-        "AMSS_SHELL_PARALLEL_INTERP",
-        "AMSS_SHELL_CUDA_INTERP",
-        "AMSS_SHELL_INTERP_THREADS",
-        "AMSS_Z4C_CUDA_RESIDENT",
-        "AMSS_CONSTRAINT_OUT_EVERY",
-        "AMSS_Z4C_MRBD",
-    ]
-    for env_name in input_overrides:
-        if env_name not in original_env and hasattr(input_data, env_name):
-            runtime_env[env_name] = str(getattr(input_data, env_name))
-
-    passthrough_envs = [
-        "AMSS_CUDA_RESIDENT_SYNC",
-        "AMSS_CUDA_BSSN_RESIDENT_SYNC",
-        "AMSS_CUDA_EM_RESIDENT_SYNC",
-        "AMSS_CUDA_ESCALAR_RESIDENT_SYNC",
-        "AMSS_CUDA_BH_INTERP_RESIDENT",
-        "AMSS_CUDA_KEEP_RESIDENT_AFTER_STEP",
-        "AMSS_CUDA_KEEP_ALL_LEVELS",
-        "AMSS_CUDA_EM_KEEP_RESIDENT_AFTER_STEP",
-        "AMSS_CUDA_EM_KEEP_ALL_LEVELS",
-        "AMSS_CUDA_ESCALAR_KEEP_RESIDENT_AFTER_STEP",
-        "AMSS_CUDA_ESCALAR_KEEP_ALL_LEVELS",
-        "AMSS_CUDA_AMR_HOST_STAGED",
-        "AMSS_CUDA_AMR_RESTRICT_DEVICE",
-        "AMSS_CUDA_AMR_RESTRICT_BATCH",
-        "AMSS_CUDA_DEVICE_SEGMENT_BATCH",
-        "AMSS_CUDA_UNCACHED_DEVICE_BUFFERS",
-        "AMSS_CUDA_EM_CACHE_SOURCES",
-        "AMSS_CUDA_EM_ZERO_FASTPATH",
-        "AMSS_CUDA_AWARE_MPI",
-        "AMSS_CUDA_REGRID_FLUSH_ALWAYS",
-        "AMSS_Z4C_CUDA_RESIDENT",
-        "AMSS_SHELL_FAST_INTERP",
-        "AMSS_SHELL_PARALLEL_INTERP",
-        "AMSS_SHELL_CUDA_INTERP",
-        "AMSS_SHELL_INTERP_THREADS",
-        "AMSS_EM_ZERO_ANALYSIS_FASTPATH",
-        "AMSS_EM_ZERO_RESIDENT_DOWNLOAD_FASTPATH",
-        "AMSS_INTERP_FAST",
-        "AMSS_INTERP_GPU",
-    ]
-    for env_name in passthrough_envs:
-        _input_env_passthrough(runtime_env, env_name)
-
-    optional_overrides = {
-        "AMSS_INTERP_FAST_COMPARE": "AMSS_Interp_Fast_Compare",
-        "AMSS_INTERP_FAST_COMPARE_LIMIT": "AMSS_Interp_Fast_Compare_Limit",
-        "AMSS_INTERP_FAST_COMPARE_TOL": "AMSS_Interp_Fast_Compare_Tol",
-        "AMSS_GPU_STAGE_TIMING": "AMSS_GPU_Stage_Timing",
-        "AMSS_GPU_STAGE_TIMING_EVERY": "AMSS_GPU_Stage_Timing_Every",
-    }
-    for env_name, input_name in optional_overrides.items():
-        if env_name not in runtime_env and hasattr(input_data, input_name):
-            runtime_env[env_name] = str(getattr(input_data, input_name))
-
-    return runtime_env
-
-
-##################################################################
-
 
 
 ##################################################################
@@ -294,13 +68,11 @@ def makefile_ABE():
     print( " Compiling the AMSS-NCKU executable file ABE/ABEGPU " ) 
     print(                                                        )
 
-    z4c_mrbd = int(getattr(input_data, "AMSS_Z4C_MRBD", 0))
-
     ## Build command with CPU binding to nohz_full cores
     if (input_data.GPU_Calculation == "no"):
-        makefile_command  = f"{NUMACTL_CPU_BIND} env AMSS_Z4C_MRBD={z4c_mrbd} make -j{BUILD_JOBS} INTERP_LB_MODE=off USE_CUDA_BSSN=0 USE_CUDA_Z4C=0 ABE"
+        makefile_command  = f"{NUMACTL_CPU_BIND} make -j{BUILD_JOBS} INTERP_LB_MODE=off ABE"
     elif (input_data.GPU_Calculation == "yes"):
-        makefile_command  = f"{NUMACTL_CPU_BIND} env AMSS_Z4C_MRBD={z4c_mrbd} make -j{BUILD_JOBS} INTERP_LB_MODE=off USE_CUDA_BSSN=1 USE_CUDA_Z4C=1 ABE_CUDA"
+        makefile_command  = f"{NUMACTL_CPU_BIND} make -j{BUILD_JOBS} ABEGPU"
     else:
         print( " CPU/GPU numerical calculation setting is wrong " )
         print(                                                    )
@@ -373,83 +145,29 @@ def run_ABE():
     print(                                                      )
 
     ## Define the command to run; cast other values to strings as needed
-    mpi_env = None
-    started_mps = False
     
-    mpi_processes = int(input_data.MPI_processes)
-    if (input_data.GPU_Calculation == "yes" and
-        getattr(input_data, "Equation_Class", "") == "Z4C"):
-        z4c_env_np = os.environ.get("AMSS_Z4C_GPU_MPI_PROCESSES")
-        if z4c_env_np and int(z4c_env_np) > 0:
-            mpi_processes = int(z4c_env_np)
-        elif mpi_processes < 4:
-            mpi_processes = 4
-    if (input_data.GPU_Calculation == "yes" and
-        getattr(input_data, "basic_grid_set", "") == "Shell-Patch"):
-        shell_env_np = os.environ.get("AMSS_SHELL_GPU_MPI_PROCESSES")
-        if shell_env_np and int(shell_env_np) > 0:
-            mpi_processes = int(shell_env_np)
-        elif mpi_processes < 4:
-            mpi_processes = 4
-
     if (input_data.GPU_Calculation == "no"):
-        mpi_command         = NUMACTL_CPU_BIND + " mpirun -np " + str(mpi_processes) + " ./ABE"
+        mpi_command         = NUMACTL_CPU_BIND + " mpirun -np " + str(input_data.MPI_processes) + " ./ABE"
         #mpi_command         = " mpirun -np " + str(input_data.MPI_processes) + " ./ABE"
         mpi_command_outfile = "ABE_out.log"
     elif (input_data.GPU_Calculation == "yes"):
-        mpi_command         = NUMACTL_CPU_BIND + " I_MPI_OFFLOAD=1 I_MPI_OFFLOAD_IPC=0 mpirun -np " + str(mpi_processes) + " ./ABE_CUDA"
+        mpi_command         = NUMACTL_CPU_BIND + " mpirun -np " + str(input_data.MPI_processes) + " ./ABEGPU"
         mpi_command_outfile = "ABEGPU_out.log"
-        mpi_env = _gpu_runtime_env()
-        started_mps = _start_cuda_mps_if_requested(mpi_env)
-        print(" GPU optimized runtime switches:")
-        print(f"   MPI processes={mpi_processes}")
-        print(f"   AMSS_INTERP_FAST={mpi_env.get('AMSS_INTERP_FAST', '')}")
-        print(f"   AMSS_INTERP_GPU={mpi_env.get('AMSS_INTERP_GPU', '')}")
-        print(f"   AMSS_ANALYSIS_MAP_EVERY={mpi_env.get('AMSS_ANALYSIS_MAP_EVERY', '')}")
-        print(f"   AMSS_EVOLVE_TIMING={mpi_env.get('AMSS_EVOLVE_TIMING', '')}")
-        print(f"   AMSS_ESCALAR_STEP_TIMING={mpi_env.get('AMSS_ESCALAR_STEP_TIMING', '')}")
-        print(f"   AMSS_CUDA_AWARE_MPI={mpi_env.get('AMSS_CUDA_AWARE_MPI', '')}")
-        print(f"   AMSS_CUDA_KEEP_RESIDENT_AFTER_STEP={mpi_env.get('AMSS_CUDA_KEEP_RESIDENT_AFTER_STEP', '')}")
-        print(f"   AMSS_CUDA_KEEP_ALL_LEVELS={mpi_env.get('AMSS_CUDA_KEEP_ALL_LEVELS', '')}")
-        print(f"   AMSS_CUDA_ESCALAR_KEEP_RESIDENT_AFTER_STEP={mpi_env.get('AMSS_CUDA_ESCALAR_KEEP_RESIDENT_AFTER_STEP', '')}")
-        print(f"   AMSS_CUDA_ESCALAR_KEEP_ALL_LEVELS={mpi_env.get('AMSS_CUDA_ESCALAR_KEEP_ALL_LEVELS', '')}")
-        print(f"   AMSS_CUDA_EM_CACHE_SOURCES={mpi_env.get('AMSS_CUDA_EM_CACHE_SOURCES', '')}")
-        print(f"   AMSS_CUDA_EM_ZERO_FASTPATH={mpi_env.get('AMSS_CUDA_EM_ZERO_FASTPATH', '')}")
-        print(f"   AMSS_EM_ZERO_ANALYSIS_FASTPATH={mpi_env.get('AMSS_EM_ZERO_ANALYSIS_FASTPATH', '')}")
-        print(f"   AMSS_EM_ZERO_RESIDENT_DOWNLOAD_FASTPATH={mpi_env.get('AMSS_EM_ZERO_RESIDENT_DOWNLOAD_FASTPATH', '')}")
-        print(f"   AMSS_CUDA_AMR_HOST_STAGED={mpi_env.get('AMSS_CUDA_AMR_HOST_STAGED', '')}")
-        print(f"   AMSS_CUDA_AMR_RESTRICT_DEVICE={mpi_env.get('AMSS_CUDA_AMR_RESTRICT_DEVICE', '')}")
-        print(f"   AMSS_CUDA_AMR_RESTRICT_BATCH={mpi_env.get('AMSS_CUDA_AMR_RESTRICT_BATCH', '')}")
-        print(f"   AMSS_CUDA_DEVICE_SEGMENT_BATCH={mpi_env.get('AMSS_CUDA_DEVICE_SEGMENT_BATCH', '')}")
-        print(f"   AMSS_CUDA_UNCACHED_DEVICE_BUFFERS={mpi_env.get('AMSS_CUDA_UNCACHED_DEVICE_BUFFERS', '')}")
-        print(f"   AMSS_SHELL_FAST_INTERP={mpi_env.get('AMSS_SHELL_FAST_INTERP', '')}")
-        print(f"   AMSS_SHELL_PARALLEL_INTERP={mpi_env.get('AMSS_SHELL_PARALLEL_INTERP', '')}")
-        print(f"   AMSS_SHELL_CUDA_INTERP={mpi_env.get('AMSS_SHELL_CUDA_INTERP', '')}")
-        print(f"   AMSS_SHELL_INTERP_THREADS={mpi_env.get('AMSS_SHELL_INTERP_THREADS', '')}")
-        print(f"   AMSS_Z4C_CUDA_RESIDENT={mpi_env.get('AMSS_Z4C_CUDA_RESIDENT', '')}")
-        print(f"   AMSS_CONSTRAINT_OUT_EVERY={mpi_env.get('AMSS_CONSTRAINT_OUT_EVERY', '')}")
-        if "CUDA_MPS_PIPE_DIRECTORY" in mpi_env:
-            print(f"   CUDA_MPS_PIPE_DIRECTORY={mpi_env['CUDA_MPS_PIPE_DIRECTORY']}")
  
-    try:
-        ## Execute the MPI command and stream output
-        mpi_process = subprocess.Popen(mpi_command, shell=True, stdout=subprocess.PIPE,
-                                       stderr=subprocess.STDOUT, text=True, env=mpi_env)
+    ## Execute the MPI command and stream output
+    mpi_process = subprocess.Popen(mpi_command, shell=True, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True)
 
-        ## Write ABE run output to file while printing to stdout
-        with open(mpi_command_outfile, 'w') as file0:
-            ## Read and print output lines; also write each line to file
-            for line in mpi_process.stdout:
-                print(line, end='')  # stream output in real time
-                file0.write(line)    # write the line to file
+    ## Write ABE run output to file while printing to stdout
+    with open(mpi_command_outfile, 'w') as file0:  
+        ## Read and print output lines; also write each line to file
+        for line in mpi_process.stdout:
+            print(line, end='')  # stream output in real time
+            file0.write(line)    # write the line to file
+            file0.flush()        # flush to ensure each line is written immediately (optional)            
+    file0.close()
 
-        ## Wait for the process to finish
-        mpi_return_code = mpi_process.wait()
-        if mpi_return_code != 0:
-            raise subprocess.CalledProcessError(mpi_return_code, mpi_command)
-    finally:
-        if started_mps:
-            _stop_cuda_mps(mpi_env)
+    ## Wait for the process to finish
+    mpi_return_code = mpi_process.wait()
     
     print(                                           )
     print( " The ABE/ABEGPU simulation is finished " ) 
@@ -485,6 +203,8 @@ def run_TwoPunctureABE():
         for line in TwoPuncture_process.stdout:
             print(line, end='')  # stream output in real time
             file0.write(line)    # write the line to file
+            file0.flush()        # flush to ensure each line is written immediately (optional)                 
+    file0.close()
 
     ## Wait for the process to finish
     TwoPuncture_command_return_code = TwoPuncture_process.wait()

plot_xiaoqu.py

@@ -808,10 +808,10 @@ def generate_ADMmass_plot( outdir, figure_outdir, detector_number_i ):
 
 ## Plot constraint violation for each grid level
 
-def generate_constraint_check_plot( outdir, figure_outdir, input_level_number ):
-
-    # path to data file
-    file0 = os.path.join(outdir, "bssn_constraint.dat")
+def generate_constraint_check_plot( outdir, figure_outdir, input_level_number ):
+
+    # path to data file
+    file0 = os.path.join(outdir, "bssn_constraint.dat")
 
     if ( input_level_number == 0 ):
         print(                                                   )
@@ -819,26 +819,13 @@ def generate_constraint_check_plot( outdir, figure_outdir, input_level_number ):
         print(                                                          )
         print( " corresponding data file = ", file0 )
         print(                                      )
-
-    print( " Begin the constraint violation plot for grid level number =  ", input_level_number )
-
-    if (not os.path.exists(file0)) or os.path.getsize(file0) == 0:
-        if ( input_level_number == 0 ):
-            print( " Constraint data file is empty; skip constraint violation plots" )
-            print(                                                            )
-        return
-
-    # load the full data file (assumed whitespace-separated floats)
-    data = numpy.loadtxt(file0)
-    data = numpy.atleast_2d(data)
-
-    if data.shape[1] < 8:
-        if ( input_level_number == 0 ):
-            print( " Constraint data file has insufficient columns; skip constraint violation plots" )
-            print(                                                                                )
-        return
-
-    # extract columns from the constraint data file
+
+    print( " Begin the constraint violation plot for grid level number =  ", input_level_number )
+    
+    # load the full data file (assumed whitespace-separated floats)
+    data = numpy.loadtxt(file0)
+    
+    # extract columns from the constraint data file
     time          = data[:,0]
     Constraint_H  = data[:,1]
     Constraint_Px = data[:,2]