pyvpmr 250523__tar.gz → 250527__tar.gz

{pyvpmr-250523 → pyvpmr-250527}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pyvpmr
-Version: 250523
+Version: 250527
 Summary: The VPMR Algorithm
 Author-email: Theodore Chang <tlcfem@gmail.com>
 Maintainer-email: Theodore Chang <tlcfem@gmail.com>
@@ -813,7 +813,7 @@ Options:
     -c, --max-exponent          <int>     maximum exponent (default: 4)
     -d, --precision-bits        <int>     number of precision bits (default: 512)
     -q, --quadrature-order      <int>     quadrature order (default: 500)
-    -m, --precision-multiplier  <float>   precision multiplier (default: 1.5)
+    -m, --precision-multiplier  <float>   precision multiplier (default: 1.05)
     -e, --tolerance             <float>   tolerance (default: 1E-8)
     -k, --kernel                <string>  file name of kernel function (default uses: exp(-t^2/4))
     -s, --singular-values                 print singular values
@@ -842,7 +842,7 @@ Using the following parameters:
     exponent = 4.
    precision = 355.
  quad. order = 500.
-  multiplier = 1.5000e+00.
+  multiplier = 1.0500e+00.
    tolerance = 1.0000e-08.
       kernel = exp(-t*t/4).
 
@@ -854,7 +854,7 @@ Using the following parameters:
 [6/6] Done.
 
 M = 
-+1.1745193571738943e+01+6.4089561283054790e-107j
++1.1745193571738943e+01+6.4196555242161141e-106j
 -5.5143304351134397e+00+5.7204056791636839e+00j
 -5.5143304351134397e+00-5.7204056791636839e+00j
 -1.6161617424833762e-02+2.3459542440459513e+00j
@@ -874,7 +874,7 @@ S =
 +1.7655956664692953e+00-2.7555720406099038e+00j
 +1.7655956664692953e+00+2.7555720406099038e+00j
 
-Running time: 3112 ms.
+Running time: 2044 ms.
 ```
 
 ![exp(-t^2/4)](https://raw.githubusercontent.com/TLCFEM/vpmr/master/resource/example.png)
@@ -906,8 +906,8 @@ echo "exp(-t*t/10)" > kernel.txt
 
 ## Performance
 
-The computation of weights, that involves integrals, and SVD are parallelised.
-A typical profiling would yield something similar to the following.
+The majority of the algorithm is parallelised to extract the maximum performance.
+The following is a typical performance profile on a i7-10750H platform using the `./vpmr -n 80`.
 
 ![profiling](https://raw.githubusercontent.com/TLCFEM/vpmr/master/resource/profile.png)
 

{pyvpmr-250523 → pyvpmr-250527}/README.md

@@ -106,7 +106,7 @@ Options:
     -c, --max-exponent          <int>     maximum exponent (default: 4)
     -d, --precision-bits        <int>     number of precision bits (default: 512)
     -q, --quadrature-order      <int>     quadrature order (default: 500)
-    -m, --precision-multiplier  <float>   precision multiplier (default: 1.5)
+    -m, --precision-multiplier  <float>   precision multiplier (default: 1.05)
     -e, --tolerance             <float>   tolerance (default: 1E-8)
     -k, --kernel                <string>  file name of kernel function (default uses: exp(-t^2/4))
     -s, --singular-values                 print singular values
@@ -135,7 +135,7 @@ Using the following parameters:
     exponent = 4.
    precision = 355.
  quad. order = 500.
-  multiplier = 1.5000e+00.
+  multiplier = 1.0500e+00.
    tolerance = 1.0000e-08.
       kernel = exp(-t*t/4).
 
@@ -147,7 +147,7 @@ Using the following parameters:
 [6/6] Done.
 
 M = 
-+1.1745193571738943e+01+6.4089561283054790e-107j
++1.1745193571738943e+01+6.4196555242161141e-106j
 -5.5143304351134397e+00+5.7204056791636839e+00j
 -5.5143304351134397e+00-5.7204056791636839e+00j
 -1.6161617424833762e-02+2.3459542440459513e+00j
@@ -167,7 +167,7 @@ S =
 +1.7655956664692953e+00-2.7555720406099038e+00j
 +1.7655956664692953e+00+2.7555720406099038e+00j
 
-Running time: 3112 ms.
+Running time: 2044 ms.
 ```
 
 ![exp(-t^2/4)](resource/example.png)
@@ -199,8 +199,8 @@ echo "exp(-t*t/10)" > kernel.txt
 
 ## Performance
 
-The computation of weights, that involves integrals, and SVD are parallelised.
-A typical profiling would yield something similar to the following.
+The majority of the algorithm is parallelised to extract the maximum performance.
+The following is a typical performance profile on a i7-10750H platform using the `./vpmr -n 80`.
 
 ![profiling](resource/profile.png)
 

{pyvpmr-250523 → pyvpmr-250527}/eigen/Eigen/src/Core/AssignEvaluator.h

@@ -558,11 +558,10 @@ struct dense_assignment_loop_impl<Kernel, InnerVectorizedTraversal, InnerUnrolli
 /***********************
 *** Linear traversal ***
 ***********************/
-static tbb::affinity_partitioner ap;
+
 template <typename Kernel>
 struct dense_assignment_loop_impl<Kernel, LinearTraversal, NoUnrolling> {
   EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE constexpr void run(Kernel& kernel) {
-    return tbb::parallel_for(Index(0), kernel.size(), [&](const Index l) { kernel.assignCoeff(l); }, ap);
     const Index size = kernel.size();
     for (Index i = 0; i < size; ++i) kernel.assignCoeff(i);
   }

{pyvpmr-250523 → pyvpmr-250527}/eigen/Eigen/src/Core/products/GeneralBlockPanelKernel.h

@@ -2108,7 +2108,8 @@ EIGEN_DONT_INLINE void gebp_kernel<LhsScalar, RhsScalar, Index, DataMapper, mr,
         }
       }
 #endif
-      for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
+      tbb::parallel_for(packet_cols8 / 4, packet_cols4 / 4, [&](Index j2) {
+        j2 *= 4;
         for (Index i = i1; i < actual_panel_end; i += 2 * LhsProgress) {
           // We selected a 2*Traits::LhsProgress x nr micro block of res which is entirely
           // stored into 2 x nr registers.
@@ -2227,7 +2228,7 @@ EIGEN_DONT_INLINE void gebp_kernel<LhsScalar, RhsScalar, Index, DataMapper, mr,
           r3.storePacket(0 * Traits::ResPacketSize, R2);
           r3.storePacket(1 * Traits::ResPacketSize, R3);
         }
-      }
+      });
 
       // Deal with remaining columns of the rhs
       for (Index j2 = packet_cols4; j2 < cols; j2++) {

{pyvpmr-250523 → pyvpmr-250527}/eigen/Eigen/src/Core/products/GeneralMatrixVector.h

@@ -13,8 +13,6 @@
 // IWYU pragma: private
 #include "../InternalHeaderCheck.h"
 
-#include <tbb/parallel_reduce.h>
-
 namespace Eigen {
 
 namespace internal {

{pyvpmr-250523 → pyvpmr-250527}/eigen/Eigen/src/SVD/BDCSVD.h

@@ -30,8 +30,6 @@
 // IWYU pragma: private
 #include "./InternalHeaderCheck.h"
 
-#include <tbb/parallel_reduce.h>
-
 #ifdef EIGEN_BDCSVD_DEBUG_VERBOSE
 #include <iostream>
 #endif
@@ -826,7 +824,7 @@ typename BDCSVD<MatrixType, Options>::RealScalar BDCSVD<MatrixType, Options>::se
       tbb::blocked_range<Index>(Index(0), perm.size()), RealScalar(1),
       [&](const tbb::blocked_range<Index>& r, RealScalar running_total) {
         for (auto i = r.begin(); i < r.end(); ++i) {
-          Index j = perm(i);
+          const Index j = perm(i);
           running_total += (col0(j) / (diagShifted(j) - mu)) * (col0(j) / (diag(j) + shift + mu));
         }
         return running_total;
@@ -1187,7 +1185,7 @@ void BDCSVD<MatrixType, Options>::computeSingVecs(const ArrayRef& zhat, const Ar
       tbb::parallel_for(
           Index(0), m,
           [&](const Index l) {
-            Index i = perm(l);
+            const Index i = perm(l);
             U(i, k) = zhat(i) / (((diag(i) - shifts(k)) - mus(k))) / ((diag(i) + singVals[k]));
           },
           ap);

{pyvpmr-250523 → pyvpmr-250527}/patch_size.patch

@@ -1,34 +1,31 @@
-diff --git a/Eigen/src/Core/AssignEvaluator.h b/Eigen/src/Core/AssignEvaluator.h
-index 36f0a9d74..13b31110c 100644
---- a/Eigen/src/Core/AssignEvaluator.h
-+++ b/Eigen/src/Core/AssignEvaluator.h
-@@ -558,10 +558,11 @@ struct dense_assignment_loop_impl<Kernel, InnerVectorizedTraversal, InnerUnrolli
- /***********************
- *** Linear traversal ***
- ***********************/
--
-+static tbb::affinity_partitioner ap;
- template <typename Kernel>
- struct dense_assignment_loop_impl<Kernel, LinearTraversal, NoUnrolling> {
-   EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE constexpr void run(Kernel& kernel) {
-+    return tbb::parallel_for(Index(0), kernel.size(), [&](const Index l) { kernel.assignCoeff(l); }, ap);
-     const Index size = kernel.size();
-     for (Index i = 0; i < size; ++i) kernel.assignCoeff(i);
-   }
+diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h
+index e72c6b48e..4f918cd20 100644
+--- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h
++++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h
+@@ -2108,7 +2108,8 @@ EIGEN_DONT_INLINE void gebp_kernel<LhsScalar, RhsScalar, Index, DataMapper, mr,
+         }
+       }
+ #endif
+-      for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
++      tbb::parallel_for(packet_cols8 / 4, packet_cols4 / 4, [&](Index j2) {
++        j2 *= 4;
+         for (Index i = i1; i < actual_panel_end; i += 2 * LhsProgress) {
+           // We selected a 2*Traits::LhsProgress x nr micro block of res which is entirely
+           // stored into 2 x nr registers.
+@@ -2227,7 +2228,7 @@ EIGEN_DONT_INLINE void gebp_kernel<LhsScalar, RhsScalar, Index, DataMapper, mr,
+           r3.storePacket(0 * Traits::ResPacketSize, R2);
+           r3.storePacket(1 * Traits::ResPacketSize, R3);
+         }
+-      }
++      });
+ 
+       // Deal with remaining columns of the rhs
+       for (Index j2 = packet_cols4; j2 < cols; j2++) {
 diff --git a/Eigen/src/Core/products/GeneralMatrixVector.h b/Eigen/src/Core/products/GeneralMatrixVector.h
-index ba72a8a4f..51775ba5f 100644
+index ba72a8a4f..4ce9f99ab 100644
 --- a/Eigen/src/Core/products/GeneralMatrixVector.h
 +++ b/Eigen/src/Core/products/GeneralMatrixVector.h
-@@ -13,6 +13,8 @@
- // IWYU pragma: private
- #include "../InternalHeaderCheck.h"
- 
-+#include <tbb/parallel_reduce.h>
-+
- namespace Eigen {
- 
- namespace internal {
-@@ -150,30 +152,20 @@ general_matrix_vector_product<Index, LhsScalar, LhsMapper, ColMajor, ConjugateLh
+@@ -150,30 +150,20 @@ general_matrix_vector_product<Index, LhsScalar, LhsMapper, ColMajor, ConjugateLh
      Index jend = numext::mini(j2 + block_cols, cols);
      Index i = 0;
      for (; i < n8; i += ResPacketSize * 8) {
@@ -73,19 +70,10 @@ index ba72a8a4f..51775ba5f 100644
      if (i < n4) {
        ResPacket c0 = pset1<ResPacket>(ResScalar(0)), c1 = pset1<ResPacket>(ResScalar(0)),
 diff --git a/Eigen/src/SVD/BDCSVD.h b/Eigen/src/SVD/BDCSVD.h
-index 6fab905e5..98fd68cdb 100644
+index 6fab905e5..9ae75c579 100644
 --- a/Eigen/src/SVD/BDCSVD.h
 +++ b/Eigen/src/SVD/BDCSVD.h
-@@ -30,6 +30,8 @@
- // IWYU pragma: private
- #include "./InternalHeaderCheck.h"
- 
-+#include <tbb/parallel_reduce.h>
-+
- #ifdef EIGEN_BDCSVD_DEBUG_VERBOSE
- #include <iostream>
- #endif
-@@ -820,6 +822,16 @@ template <typename MatrixType, int Options>
+@@ -820,6 +820,16 @@ template <typename MatrixType, int Options>
  typename BDCSVD<MatrixType, Options>::RealScalar BDCSVD<MatrixType, Options>::secularEq(
      RealScalar mu, const ArrayRef& col0, const ArrayRef& diag, const IndicesRef& perm, const ArrayRef& diagShifted,
      RealScalar shift) {
@@ -93,7 +81,7 @@ index 6fab905e5..98fd68cdb 100644
 +      tbb::blocked_range<Index>(Index(0), perm.size()), RealScalar(1),
 +      [&](const tbb::blocked_range<Index>& r, RealScalar running_total) {
 +        for (auto i = r.begin(); i < r.end(); ++i) {
-+          Index j = perm(i);
++          const Index j = perm(i);
 +          running_total += (col0(j) / (diagShifted(j) - mu)) * (col0(j) / (diag(j) + shift + mu));
 +        }
 +        return running_total;
@@ -102,7 +90,7 @@ index 6fab905e5..98fd68cdb 100644
    Index m = perm.size();
    RealScalar res = Literal(1);
    for (Index i = 0; i < m; ++i) {
-@@ -1161,6 +1173,8 @@ template <typename MatrixType, int Options>
+@@ -1161,6 +1171,8 @@ template <typename MatrixType, int Options>
  void BDCSVD<MatrixType, Options>::computeSingVecs(const ArrayRef& zhat, const ArrayRef& diag, const IndicesRef& perm,
                                                    const VectorType& singVals, const ArrayRef& shifts,
                                                    const ArrayRef& mus, MatrixXr& U, MatrixXr& V) {
@@ -111,7 +99,7 @@ index 6fab905e5..98fd68cdb 100644
    Index n = zhat.size();
    Index m = perm.size();
  
-@@ -1170,10 +1184,13 @@ void BDCSVD<MatrixType, Options>::computeSingVecs(const ArrayRef& zhat, const Ar
+@@ -1170,10 +1182,13 @@ void BDCSVD<MatrixType, Options>::computeSingVecs(const ArrayRef& zhat, const Ar
        if (m_compV) V.col(k) = VectorType::Unit(n, k);
      } else {
        U.col(k).setZero();
@@ -122,7 +110,7 @@ index 6fab905e5..98fd68cdb 100644
 +      tbb::parallel_for(
 +          Index(0), m,
 +          [&](const Index l) {
-+            Index i = perm(l);
++            const Index i = perm(l);
 +            U(i, k) = zhat(i) / (((diag(i) - shifts(k)) - mus(k))) / ((diag(i) + singVals[k]));
 +          },
 +          ap);

{pyvpmr-250523 → pyvpmr-250527}/pybind11/CMakePresets.json

@@ -12,10 +12,12 @@
       },
       "cacheVariables": {
         "CMAKE_BUILD_TYPE": "Debug",
-        "PYBIND11_FINDPYTHON": "NEW",
-        "PYBIND11_WERROR": true,
+        "CMAKE_EXPORT_COMPILE_COMMANDS": true,
         "DOWNLOAD_CATCH": true,
-        "DOWNLOAD_EIGEN": true
+        "DOWNLOAD_EIGEN": true,
+        "PYBIND11_FINDPYTHON": "NEW",
+        "PYBIND11_WERROR": true
+
       }
     },
     {
@@ -23,13 +25,18 @@
       "displayName": "Venv",
       "inherits": "default",
       "cacheVariables": {
-        "CMAKE_BUILD_TYPE": "Debug",
         "PYBIND11_CREATE_WITH_UV": "python3",
-        "Python_ROOT_DIR": ".venv",
-        "PYBIND11_WERROR": true,
-        "PYBIND11_FINDPYTHON": "NEW",
-        "DOWNLOAD_CATCH": true,
-        "DOWNLOAD_EIGEN": true
+        "Python_ROOT_DIR": ".venv"
+      }
+    },
+    {
+      "name": "tidy",
+      "displayName": "Clang-tidy",
+      "inherits": "default",
+      "binaryDir": "build-tidy",
+      "cacheVariables": {
+        "CMAKE_CXX_CLANG_TIDY": "clang-tidy;--use-color;--warnings-as-errors=*",
+        "CMAKE_CXX_STANDARD": "17"
       }
     }
   ],
@@ -44,6 +51,12 @@
       "displayName": "Venv Build",
       "configurePreset": "venv"
     },
+    {
+      "name": "tidy",
+      "displayName": "Clang-tidy Build",
+      "configurePreset": "tidy",
+      "nativeToolOptions": ["-k0"]
+    },
     {
       "name": "tests",
       "displayName": "Tests (for workflow)",

{pyvpmr-250523 → pyvpmr-250527}/pybind11/include/pybind11/cast.h

@@ -817,7 +817,9 @@ protected:
     cast_impl(T &&src, return_value_policy policy, handle parent, index_sequence<Is...>) {
         PYBIND11_WORKAROUND_INCORRECT_MSVC_C4100(src, policy, parent);
         PYBIND11_WORKAROUND_INCORRECT_GCC_UNUSED_BUT_SET_PARAMETER(policy, parent);
+
         std::array<object, size> entries{{reinterpret_steal<object>(
+            // NOLINTNEXTLINE(bugprone-use-after-move)
             make_caster<Ts>::cast(std::get<Is>(std::forward<T>(src)), policy, parent))...}};
         for (const auto &entry : entries) {
             if (!entry) {

{pyvpmr-250523 → pyvpmr-250527}/pybind11/include/pybind11/chrono.h

@@ -63,6 +63,9 @@ public:
     get_duration(const std::chrono::duration<rep, period> &src) {
         return src;
     }
+    static const std::chrono::duration<rep, period> &
+    get_duration(const std::chrono::duration<rep, period> &&)
+        = delete;
 
     // If this is a time_point get the time_since_epoch
     template <typename Clock>

pyvpmr-250527/pybind11/include/pybind11/critical_section.h

@@ -0,0 +1,50 @@
+// Copyright (c) 2016-2025 The Pybind Development Team.
+// All rights reserved. Use of this source code is governed by a
+// BSD-style license that can be found in the LICENSE file.
+
+#pragma once
+
+#include "pytypes.h"
+
+PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
+
+/// This does not do anything if there's a GIL. On free-threaded Python,
+/// it locks an object. This uses the CPython API, which has limits
+class scoped_critical_section {
+public:
+#ifdef Py_GIL_DISABLED
+    explicit scoped_critical_section(handle obj) : has2(false) {
+        PyCriticalSection_Begin(&section, obj.ptr());
+    }
+
+    scoped_critical_section(handle obj1, handle obj2) : has2(true) {
+        PyCriticalSection2_Begin(&section2, obj1.ptr(), obj2.ptr());
+    }
+
+    ~scoped_critical_section() {
+        if (has2) {
+            PyCriticalSection2_End(&section2);
+        } else {
+            PyCriticalSection_End(&section);
+        }
+    }
+#else
+    explicit scoped_critical_section(handle) {};
+    scoped_critical_section(handle, handle) {};
+    ~scoped_critical_section() = default;
+#endif
+
+    scoped_critical_section(const scoped_critical_section &) = delete;
+    scoped_critical_section &operator=(const scoped_critical_section &) = delete;
+
+private:
+#ifdef Py_GIL_DISABLED
+    bool has2;
+    union {
+        PyCriticalSection section;
+        PyCriticalSection2 section2;
+    };
+#endif
+};
+
+PYBIND11_NAMESPACE_END(PYBIND11_NAMESPACE)

{pyvpmr-250523 → pyvpmr-250527}/pybind11/include/pybind11/detail/common.h

@@ -14,13 +14,35 @@
 #    error "PYTHON < 3.8 IS UNSUPPORTED. pybind11 v2.13 was the last to support Python 3.7."
 #endif
 
+// Similar to Python's convention: https://docs.python.org/3/c-api/apiabiversion.html
+// See also: https://github.com/python/cpython/blob/HEAD/Include/patchlevel.h
+/* -- start version constants -- */
 #define PYBIND11_VERSION_MAJOR 3
 #define PYBIND11_VERSION_MINOR 0
+#define PYBIND11_VERSION_MICRO 0
+// ALPHA = 0xA, BETA = 0xB, GAMMA = 0xC (release candidate), FINAL = 0xF (stable release)
+// - The release level is set to "alpha" for development versions.
+//   Use 0xA0 (LEVEL=0xA, SERIAL=0) for development versions.
+// - For stable releases, set the serial to 0.
+#define PYBIND11_VERSION_RELEASE_LEVEL PY_RELEASE_LEVEL_GAMMA
+#define PYBIND11_VERSION_RELEASE_SERIAL 1
+// String version of (micro, release level, release serial), e.g.: 0a0, 0b1, 0rc1, 0
 #define PYBIND11_VERSION_PATCH 0rc1
+/* -- end version constants -- */
 
-// Similar to Python's convention: https://docs.python.org/3/c-api/apiabiversion.html
-// Use 0xA0 for dev
-#define PYBIND11_VERSION_HEX 0x030000C1
+#if !defined(Py_PACK_FULL_VERSION)
+// Stable API since Python 3.14.0a4
+#    define Py_PACK_FULL_VERSION(X, Y, Z, LEVEL, SERIAL)                                          \
+        ((((X) & 0xff) << 24) | (((Y) & 0xff) << 16) | (((Z) & 0xff) << 8)                        \
+         | (((LEVEL) & 0xf) << 4) | (((SERIAL) & 0xf) << 0))
+#endif
+// Version as a single 4-byte hex number, e.g. 0x030C04B5 == 3.12.4b5.
+#define PYBIND11_VERSION_HEX                                                                      \
+    Py_PACK_FULL_VERSION(PYBIND11_VERSION_MAJOR,                                                  \
+                         PYBIND11_VERSION_MINOR,                                                  \
+                         PYBIND11_VERSION_MICRO,                                                  \
+                         PYBIND11_VERSION_RELEASE_LEVEL,                                          \
+                         PYBIND11_VERSION_RELEASE_SERIAL)
 
 #include "pybind11_namespace_macros.h"
 
@@ -363,35 +385,39 @@
     }                                                                                             \
     PyObject *pybind11_init()
 
+// this push is for the next several macros
 PYBIND11_WARNING_PUSH
 PYBIND11_WARNING_DISABLE_CLANG("-Wgnu-zero-variadic-macro-arguments")
+
+/**
+Create a PyInit_ function for this module.
+
+Note that this is run once for each (sub-)interpreter the module is imported into, including
+possibly concurrently.  The PyModuleDef is allowed to be static, but the PyObject* resulting from
+PyModuleDef_Init should be treated like any other PyObject (so not shared across interpreters).
+ */
 #define PYBIND11_MODULE_PYINIT(name, pre_init, ...)                                               \
-    static ::pybind11::module_::module_def PYBIND11_CONCAT(pybind11_module_def_, name);           \
-    static ::pybind11::module_::slots_array PYBIND11_CONCAT(pybind11_module_slots_, name);        \
     static int PYBIND11_CONCAT(pybind11_exec_, name)(PyObject *);                                 \
-    static void PYBIND11_CONCAT(pybind11_init_, name)(::pybind11::module_ &);                     \
     PYBIND11_PLUGIN_IMPL(name) {                                                                  \
         PYBIND11_CHECK_PYTHON_VERSION                                                             \
         pre_init;                                                                                 \
         PYBIND11_ENSURE_INTERNALS_READY                                                           \
-        static auto result = []() {                                                               \
-            auto &slots = PYBIND11_CONCAT(pybind11_module_slots_, name);                          \
-            slots[0] = {Py_mod_exec,                                                              \
-                        reinterpret_cast<void *>(&PYBIND11_CONCAT(pybind11_exec_, name))};        \
-            slots[1] = {0, nullptr};                                                              \
-            return ::pybind11::module_::initialize_multiphase_module_def(                         \
-                PYBIND11_TOSTRING(name),                                                          \
-                nullptr,                                                                          \
-                &PYBIND11_CONCAT(pybind11_module_def_, name),                                     \
-                slots,                                                                            \
-                ##__VA_ARGS__);                                                                   \
-        }();                                                                                      \
-        return result.ptr();                                                                      \
+        static ::pybind11::detail::slots_array slots = ::pybind11::detail::init_slots(            \
+            &PYBIND11_CONCAT(pybind11_exec_, name), ##__VA_ARGS__);                               \
+        static PyModuleDef def{/* m_base */ PyModuleDef_HEAD_INIT,                                \
+                               /* m_name */ PYBIND11_TOSTRING(name),                              \
+                               /* m_doc */ nullptr,                                               \
+                               /* m_size */ 0,                                                    \
+                               /* m_methods */ nullptr,                                           \
+                               /* m_slots */ slots.data(),                                        \
+                               /* m_traverse */ nullptr,                                          \
+                               /* m_clear */ nullptr,                                             \
+                               /* m_free */ nullptr};                                             \
+        return PyModuleDef_Init(&def);                                                            \
     }
 
-PYBIND11_WARNING_POP
-
 #define PYBIND11_MODULE_EXEC(name, variable)                                                      \
+    static void PYBIND11_CONCAT(pybind11_init_, name)(::pybind11::module_ &);                     \
     int PYBIND11_CONCAT(pybind11_exec_, name)(PyObject * pm) {                                    \
         try {                                                                                     \
             auto m = pybind11::reinterpret_borrow<::pybind11::module_>(pm);                       \
@@ -442,12 +468,12 @@ PYBIND11_WARNING_POP
         }
 
 \endrst */
-PYBIND11_WARNING_PUSH
-PYBIND11_WARNING_DISABLE_CLANG("-Wgnu-zero-variadic-macro-arguments")
 #define PYBIND11_MODULE(name, variable, ...)                                                      \
     PYBIND11_MODULE_PYINIT(                                                                       \
         name, (pybind11::detail::get_num_interpreters_seen() += 1), ##__VA_ARGS__)                \
     PYBIND11_MODULE_EXEC(name, variable)
+
+// pop gnu-zero-variadic-macro-arguments
 PYBIND11_WARNING_POP
 
 PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)

{pyvpmr-250523 → pyvpmr-250527}/pybind11/include/pybind11/detail/init.h

@@ -273,7 +273,8 @@ struct constructor {
     static void execute(Class &cl, const Extra &...extra) {
         cl.def(
             "__init__",
-            [](value_and_holder &v_h, Args... args) {
+            [](value_and_holder &v_h,
+               Args... args) { // NOLINT(performance-unnecessary-value-param)
                 v_h.value_ptr() = construct_or_initialize<Cpp<Class>>(std::forward<Args>(args)...);
             },
             is_new_style_constructor(),

{pyvpmr-250523 → pyvpmr-250527}/pybind11/include/pybind11/detail/struct_smart_holder.h

@@ -340,7 +340,7 @@ struct smart_holder {
     }
 
     template <typename T>
-    static smart_holder from_shared_ptr(std::shared_ptr<T> shd_ptr) {
+    static smart_holder from_shared_ptr(const std::shared_ptr<T> &shd_ptr) {
         smart_holder hld;
         hld.vptr = std::static_pointer_cast<void>(shd_ptr);
         hld.vptr_is_external_shared_ptr = true;

{pyvpmr-250523 → pyvpmr-250527}/pybind11/include/pybind11/functional.h

@@ -48,7 +48,7 @@ struct func_wrapper_base {
 template <typename Return, typename... Args>
 struct func_wrapper : func_wrapper_base {
     using func_wrapper_base::func_wrapper_base;
-    Return operator()(Args... args) const {
+    Return operator()(Args... args) const { // NOLINT(performance-unnecessary-value-param)
         gil_scoped_acquire acq;
         // casts the returned object as a rvalue to the return type
         return hfunc.f(std::forward<Args>(args)...).template cast<Return>();