Reduce redundant GPU allocations (#2393)

# Introduction - Do not allocate and fill storage for resetting ion concentrations if these are never changed. Reasoning: If concentrations are never changed, we do not reset them and thus do not need to store the values. - Remove a duplicate of the CV area in solver and shared state and pass a reference during solving - Remove redundant `solution` from GPU solver - Do not store / allocate diffusion values if not needed. This saves **per CV** - `1 x 8B` for `cv_area` unconditionally - `1 x 8B` for `Xd` for each ion with no diffusion is in use (majority of cases) - `2 x 8B` for `Xi` for each ion (`reset` and `init`) if not written (reasonably often) - `2 x 8B` for `Xo` for each ion (`reset` and `init`) if not written (majority of cases) - `1 x 8B` for `eX` reset for each ion if not read (majority) - `1 x 8B` for `eX` for each ion if not read (rarely) In my standard benchmark, `busyring` with complex cells, this saves about 18% of the total GPU allocation for the cell data (`shared_state`). This has become a mixed bag, fixing a few additional things that came up during testing this: - a bug in event handling on GPU - pybind11 stub generation --------- Co-authored-by: Jannik Luboeinski <[email protected]> Co-authored-by: boeschf <[email protected]>
arbor-sim · Oct 23, 2024 · b8b768d · b8b768d
1 parent fd2718c
commit b8b768d
Show file tree

Hide file tree

Showing 17 changed files with 184 additions and 185 deletions.
diff --git a/.github/workflows/codeql.yml b/.github/workflows/codeql.yml
@@ -66,7 +66,7 @@ jobs:
       run: |
            mkdir build
            cd build
-           cmake .. -DCMAKE_CXX_COMPILER=g++-12 -DCMAKE_C_COMPILER=gcc-12 -DARB_WITH_PYTHON=ON -DARB_VECTORIZE=ON -DPython3_EXECUTABLE=`which python` -DARB_WITH_MPI=OFF -DARB_WITH_ASSERTIONS=ON -DARB_WITH_PROFILING=ON
+           cmake .. -DCMAKE_CXX_COMPILER=g++-12 -DCMAKE_C_COMPILER=gcc-12 -DARB_WITH_PYTHON=ON -DARB_VECTORIZE=ON -DPython3_EXECUTABLE=`which python` -DARB_WITH_MPI=OFF -DARB_WITH_ASSERTIONS=ON -DARB_WITH_PROFILING=ON -DARB_BUILD_PYTHON_STUBS=OFF
            make -j4 tests examples pyarb
            cd -
 

diff --git a/.github/workflows/sanitize.yml b/.github/workflows/sanitize.yml
@@ -62,7 +62,7 @@ jobs:
           mkdir build
           cd build
           export SAN="-fsanitize=${{ matrix.sanitizer }} -fno-omit-frame-pointer"
-          cmake .. -GNinja -DCMAKE_BUILD_TYPE=debug -DCMAKE_CXX_FLAGS="$SAN" -DCMAKE_C_FLAGS="$SAN" -DCMAKE_EXE_LINKER_FLAGS="$SAN" -DCMAKE_MODULE_LINKER_FLAGS="$SAN" -DARB_WITH_ASSERTIONS=ON -DCMAKE_CXX_COMPILER=$CXX -DCMAKE_C_COMPILER=$CC -DARB_VECTORIZE=${{ matrix.simd }} -DARB_WITH_MPI=OFF -DARB_WITH_PYTHON=ON -DPython3_EXECUTABLE=`which python`
+          cmake .. -GNinja -DCMAKE_BUILD_TYPE=debug -DCMAKE_CXX_FLAGS="$SAN" -DCMAKE_C_FLAGS="$SAN" -DCMAKE_EXE_LINKER_FLAGS="$SAN" -DCMAKE_MODULE_LINKER_FLAGS="$SAN" -DARB_BUILD_PYTHON_STUBS=OFF -DARB_WITH_ASSERTIONS=ON -DCMAKE_CXX_COMPILER=$CXX -DCMAKE_C_COMPILER=$CC -DARB_VECTORIZE=${{ matrix.simd }} -DARB_WITH_MPI=OFF -DARB_WITH_PYTHON=ON -DPython3_EXECUTABLE=`which python`
           ninja -j4 -v tests examples pyarb
           cd -
       - name: Run unit tests

diff --git a/.github/workflows/test-matrix.yml b/.github/workflows/test-matrix.yml
@@ -116,7 +116,7 @@ jobs:
       - name: Update pip
         run: python -m pip install --upgrade pip
       - name: Install Python packages
-        run:  pip install numpy sphinx svgwrite sphinx-rtd-theme mpi4py pandas seaborn
+        run:  pip install numpy sphinx svgwrite sphinx-rtd-theme mpi4py pandas seaborn pybind11-stubgen
       - name: Clone w/ submodules
         uses: actions/checkout@v4
         with:

diff --git a/.github/workflows/test-spack.yml b/.github/workflows/test-spack.yml
@@ -74,9 +74,9 @@ jobs:
       - name: build and install arbor through spack dev-build
         run: |
           source spack/share/spack/setup-env.sh
-          spack install --no-check-signature --only dependencies arbor@develop target=x86_64_v2 +python ^python@${{ matrix.python-version }}
+          spack install --no-check-signature --only dependencies arbor@develop target=x86_64_v2 +python ~pystubs ^python@${{ matrix.python-version }}
           cd arbor
-          spack dev-build arbor@develop target=x86_64_v2 +python ^python@${{ matrix.python-version }}
+          spack dev-build arbor@develop target=x86_64_v2 +python ~pystubs ^python@${{ matrix.python-version }}
 
       - name: load arbor and verify installation, python examples.
         run: |

diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -7,7 +7,6 @@ set(CMAKE_SUPPRESS_DEVELOPER_WARNINGS ON CACHE INTERNAL "" FORCE)
 
 # Make CUDA support throw errors if architectures remain unclear
 cmake_policy(SET CMP0104 NEW)
-
 # Ensure CMake is aware of the policies for modern RPATH behavior
 cmake_policy(SET CMP0072 NEW)
 
@@ -46,6 +45,9 @@ if(NOT DEFINED CMAKE_INTERPROCEDURAL_OPTIMIZATION)
   endif()
 endif()
 
+# Use pybind11-stubgen to make type stubs.
+cmake_dependent_option(ARB_BUILD_PYTHON_STUBS "Use pybind11-stubgen to build type stubs." ON "ARB_WITH_PYTHON" OFF)
+
 # Turn on this option to force the compilers to produce color output when output is
 # redirected from the terminal (e.g. when using ninja or a pager).
 

diff --git a/arbor/backends/common_types.hpp b/arbor/backends/common_types.hpp
@@ -2,6 +2,7 @@
 
 #include <arbor/common_types.hpp>
 
+#include "fvm_layout.hpp"
 #include "util/range.hpp"
 #include "backends/threshold_crossing.hpp"
 #include "execution_context.hpp"
@@ -21,4 +22,44 @@ struct fvm_detector_info {
     execution_context ctx;
 };
 
+struct ion_data_flags {
+    // flags for resetting ion states after W access
+    bool write_eX_:1 = false; // is eX written?
+    bool write_Xo_:1 = false; // is Xo written?
+    bool write_Xi_:1 = false; // is Xi written?
+    bool write_Xd_:1 = false; // is Xd written?
+    bool read_eX_:1  = false; // is eX read?
+    bool read_Xo_:1  = false; // is Xo read?
+    bool read_Xi_:1  = false; // is Xi read?
+    bool read_Xd_:1  = false; // is Xd read?
+
+    ion_data_flags(const fvm_ion_config& config):
+        write_eX_(config.revpot_written),
+        write_Xo_(config.econc_written),
+        write_Xi_(config.iconc_written),
+        write_Xd_(config.is_diffusive),
+        read_eX_(config.revpot_read),
+        read_Xo_(config.econc_read),
+        read_Xi_(config.iconc_read),
+        read_Xd_(config.is_diffusive)
+    {}
+
+    ion_data_flags() = default;
+    ion_data_flags(const ion_data_flags&) = default;
+    ion_data_flags& operator=(const ion_data_flags&) = default;
+
+    bool xi() const { return read_Xi_ || write_Xi_; }
+    bool reset_xi() const { return write_Xi_; }
+
+    bool xo() const { return read_Xo_ || write_Xo_; }
+    bool reset_xo() const { return write_Xo_; }
+
+    bool xd() const { return read_Xd_ || write_Xd_; }
+    bool reset_xd() const { return write_Xd_; }
+
+    bool ex() const { return read_eX_ || write_eX_; }
+    bool reset_ex() const { return write_eX_; }
+};
+
+
 }
diff --git a/arbor/backends/gpu/event_stream.hpp b/arbor/backends/gpu/event_stream.hpp
@@ -19,7 +19,7 @@ struct event_stream : BaseEventStream {
 
   protected:
     void init() override final {
-        resize(this->device_ev_data_, this->device_ev_data_.size());
+        resize(this->device_ev_data_, this->ev_data_.size());
         memory::copy_async(this->ev_data_, this->device_ev_data_);
         this->base_ptr_ = this->device_ev_data_.data();
     }

diff --git a/arbor/backends/gpu/shared_state.cpp b/arbor/backends/gpu/shared_state.cpp
@@ -46,51 +46,31 @@ std::pair<arb_value_type, arb_value_type> minmax_value_impl(arb_size_type n, con
 ion_state::ion_state(const fvm_ion_config& ion_data,
                      unsigned, // alignment/padding ignored.
                      solver_ptr ptr):
-    write_eX_(ion_data.revpot_written),
-    write_Xo_(ion_data.econc_written),
-    write_Xi_(ion_data.iconc_written),
-    write_Xd_(ion_data.is_diffusive),
-    read_Xo_(ion_data.econc_written || ion_data.econc_read), // ensure that if we have W access, also R access is flagged
-    read_Xi_(ion_data.iconc_written || ion_data.iconc_read),
+    flags_(ion_data),
     node_index_(make_const_view(ion_data.cv)),
     iX_(ion_data.cv.size(), NAN),
-    eX_(ion_data.init_revpot.begin(), ion_data.init_revpot.end()),
     gX_(ion_data.cv.size(), NAN),
     charge(1u, static_cast<arb_value_type>(ion_data.charge)),
     solver(std::move(ptr)) {
-    // We don't need to allocate these if we never use them...
-    if (read_Xi_) {
-        Xi_ = make_const_view(ion_data.init_iconc);
-    }
-    if (read_Xo_) {
-        Xo_ = make_const_view(ion_data.init_econc);
-    }
-    if (write_Xi_ || write_Xd_) {
-        // ... but this is used by Xd and Xi!
-        reset_Xi_ = make_const_view(ion_data.reset_iconc);
-    }
-    if (write_Xi_) {
-        init_Xi_ = make_const_view(ion_data.init_iconc);
-        arb_assert(node_index_.size()==init_Xi_.size());
-    }
-    if (write_Xo_) {
-        init_Xo_ = make_const_view(ion_data.init_econc);
-        reset_Xo_ = make_const_view(ion_data.reset_econc);
-        arb_assert(node_index_.size()==init_Xo_.size());
-    }
-    if (write_eX_) {
-        init_eX_ = make_const_view(ion_data.init_revpot);
-        arb_assert(node_index_.size()==init_eX_.size());
-    }
-    if (write_Xd_) {
-        Xd_ = array(ion_data.cv.size(), NAN);
-    }
+    if (flags_.reset_xi()
+      ||flags_.reset_xd()) reset_Xi_ = make_const_view(ion_data.reset_iconc);
+    if (flags_.reset_xi()) init_Xi_  = make_const_view(ion_data.init_iconc);
+    if (flags_.xi())       Xi_       = make_const_view(ion_data.init_iconc);
+
+    if (flags_.reset_xo()) reset_Xo_ = make_const_view(ion_data.reset_econc);
+    if (flags_.reset_xo()) init_Xo_  = make_const_view(ion_data.init_econc);
+    if (flags_.xo())       Xo_       = make_const_view(ion_data.init_econc);
+
+    if (flags_.reset_ex()) init_eX_  = make_const_view(ion_data.init_revpot);
+    if (flags_.ex())       eX_       = make_const_view(ion_data.init_revpot);
+
+    if (flags_.xd())       Xd_       = make_const_view(ion_data.reset_iconc);
 }
 
 void ion_state::init_concentration() {
     // NB. not resetting Xd here, it's controlled via the solver.
-    if (write_Xi_) memory::copy(init_Xi_, Xi_);
-    if (write_Xo_) memory::copy(init_Xo_, Xo_);
+    if (flags_.reset_xi()) memory::copy(init_Xi_, Xi_);
+    if (flags_.reset_xo()) memory::copy(init_Xo_, Xo_);
 }
 
 void ion_state::zero_current() {
@@ -100,13 +80,10 @@ void ion_state::zero_current() {
 
 void ion_state::reset() {
     zero_current();
-    if (write_Xi_) memory::copy(reset_Xi_, Xi_);
-    if (write_Xo_) memory::copy(reset_Xo_, Xo_);
-    if (write_eX_) memory::copy(init_eX_, eX_);
-    // This goes _last_ or at least after Xi since we might have removed reset_Xi
-    // when Xi is constant. Thus conditionally resetting Xi first and then copying
-    // Xi -> Xd is save in all cases.
-    if (write_Xd_) memory::copy(reset_Xi_, Xd_);
+    if (flags_.reset_xi()) memory::copy(reset_Xi_, Xi_);
+    if (flags_.reset_xo()) memory::copy(reset_Xo_, Xo_);
+    if (flags_.reset_ex()) memory::copy(init_eX_, eX_);
+    if (flags_.reset_xd()) memory::copy(reset_Xi_, Xd_);
 }
 
 // istim_state methods:

diff --git a/arbor/backends/gpu/shared_state.hpp b/arbor/backends/gpu/shared_state.hpp
@@ -24,7 +24,6 @@
 
 namespace arb {
 namespace gpu {
-
 /*
  * Ion state fields correspond to NMODL ion variables, where X
  * is replaced with the name of the ion. E.g. for calcium 'ca':
@@ -40,30 +39,23 @@ struct ARB_ARBOR_API ion_state {
     using solver_type = arb::gpu::diffusion_state<arb_value_type, arb_index_type>;
     using solver_ptr  = std::unique_ptr<solver_type>;
 
-    bool write_eX_:1;          // is eX written?
-    bool write_Xo_:1;          // is Xo written?
-    bool write_Xi_:1;          // is Xi written?
-    bool write_Xd_:1;          // is Xd written?
-    bool read_eX_:1;           // is eX read?
-    bool read_Xo_:1;           // is Xo read?
-    bool read_Xi_:1;           // is Xi read?
-    bool read_Xd_:1;           // is Xd read?
-
-    iarray node_index_; // Instance to CV map.
-    array iX_;          // (A/m²) current density
-    array eX_;          // (mV) reversal potential
-    array Xi_;          // (mM) internal concentration
-    array Xd_;          // (mM) diffusive concentration
-    array Xo_;          // (mM) external concentration
-    array gX_;          // (kS/m²) per-species conductivity
-
-    array init_Xi_;     // (mM) area-weighted initial internal concentration
-    array init_Xo_;     // (mM) area-weighted initial external concentration
-    array reset_Xi_;    // (mM) area-weighted user-set internal concentration
-    array reset_Xo_;    // (mM) area-weighted user-set internal concentration
-    array init_eX_;     // (mM) initial reversal potential
-
-    array charge;       // charge of ionic species (global, length 1)
+    ion_data_flags flags_;    // Track what and when to reset / allocate
+
+    iarray node_index_;       // Instance to CV map.
+    array iX_;                // (A/m²) current density
+    array eX_;                // (mV) reversal potential
+    array Xi_;                // (mM) internal concentration
+    array Xd_;                // (mM) diffusive concentration
+    array Xo_;                // (mM) external concentration
+    array gX_;                // (kS/m²) per-species conductivity
+
+    array init_Xi_;           // (mM) area-weighted initial internal concentration
+    array init_Xo_;           // (mM) area-weighted initial external concentration
+    array reset_Xi_;          // (mM) area-weighted user-set internal concentration
+    array reset_Xo_;          // (mM) area-weighted user-set internal concentration
+    array init_eX_;           // (mM) initial reversal potential
+
+    array charge;             // charge of ionic species (global, length 1)
 
     solver_ptr solver = nullptr;
 

diff --git a/arbor/backends/multicore/shared_state.cpp b/arbor/backends/multicore/shared_state.cpp
@@ -54,52 +54,31 @@ ion_state::ion_state(const fvm_ion_config& ion_data,
                      unsigned align,
                      solver_ptr ptr):
     alignment(min_alignment(align)),
-    write_eX_(ion_data.revpot_written),
-    write_Xo_(ion_data.econc_written),
-    write_Xi_(ion_data.iconc_written),
-    write_Xd_(ion_data.is_diffusive),
-    read_Xo_(ion_data.econc_written || ion_data.econc_read), // ensure that if we have W access, also R access is flagged
-    read_Xi_(ion_data.iconc_written || ion_data.iconc_read),
-    read_Xd_(ion_data.is_diffusive),
+    flags_{ion_data},
     node_index_(ion_data.cv.begin(), ion_data.cv.end(), pad(alignment)),
     iX_(ion_data.cv.size(), NAN, pad(alignment)),
-    eX_(ion_data.init_revpot.begin(), ion_data.init_revpot.end(), pad(alignment)),
     gX_(ion_data.cv.size(), NAN, pad(alignment)),
     charge(1u, ion_data.charge, pad(alignment)),
     solver(std::move(ptr)) {
-    // We don't need to allocate these if we never use them...
-    if (read_Xi_) {
-        Xi_ = {ion_data.init_iconc.begin(), ion_data.init_iconc.end(), pad(alignment)};
-    }
-    if (read_Xo_) {
-        Xo_ = {ion_data.init_econc.begin(), ion_data.init_econc.end(), pad(alignment)};
-    }
-    if (write_Xi_ || write_Xd_) {
-        // ... but this is used by Xd and Xi!
-        reset_Xi_ = {ion_data.reset_iconc.begin(), ion_data.reset_iconc.end(), pad(alignment)};
-    }
-    if (write_Xi_) {
-        init_Xi_  = {ion_data.init_iconc.begin(), ion_data.init_iconc.end(), pad(alignment)};
-        arb_assert(node_index_.size()==init_Xi_.size());
-    }
-    if (write_Xo_) {
-        init_Xo_  = {ion_data.init_econc.begin(), ion_data.init_econc.end(), pad(alignment)};
-        reset_Xo_ = {ion_data.reset_econc.begin(), ion_data.reset_econc.end(), pad(alignment)};
-        arb_assert(node_index_.size()==init_Xo_.size());
-    }
-    if (write_eX_) {
-        init_eX_ = {ion_data.init_revpot.begin(), ion_data.init_revpot.end(), pad(alignment)};
-        arb_assert(node_index_.size()==init_eX_.size());
-    }
-    if (read_Xd_) {
-        Xd_ = {ion_data.reset_iconc.begin(), ion_data.reset_iconc.end(), pad(alignment)};
-    }
+    if (flags_.reset_xi()
+      ||flags_.reset_xd()) reset_Xi_ = {ion_data.reset_iconc.begin(), ion_data.reset_iconc.end(), pad(alignment)};
+    if (flags_.reset_xi()) init_Xi_  = {ion_data.init_iconc.begin(),  ion_data.init_iconc.end(),  pad(alignment)};
+    if (flags_.xi())       Xi_       = {ion_data.init_iconc.begin(),  ion_data.init_iconc.end(),  pad(alignment)};
+
+    if (flags_.reset_xo()) reset_Xo_ = {ion_data.reset_econc.begin(), ion_data.reset_econc.end(), pad(alignment)};
+    if (flags_.reset_xo()) init_Xo_  = {ion_data.init_econc.begin(),  ion_data.init_econc.end(),  pad(alignment)};
+    if (flags_.xo())       Xo_       = {ion_data.init_econc.begin(),  ion_data.init_econc.end(),  pad(alignment)};
+
+    if (flags_.reset_ex()) init_eX_ = {ion_data.init_revpot.begin(), ion_data.init_revpot.end(),  pad(alignment)};
+    if (flags_.ex())       eX_      = {ion_data.init_revpot.begin(), ion_data.init_revpot.end(),  pad(alignment)};
+
+    if (flags_.xd())       Xd_      = {ion_data.reset_iconc.begin(), ion_data.reset_iconc.end(),  pad(alignment)};
 }
 
 void ion_state::init_concentration() {
     // NB. not resetting Xd here, it's controlled via the solver.
-    if (write_Xi_) std::copy(init_Xi_.begin(), init_Xi_.end(), Xi_.begin());
-    if (write_Xo_) std::copy(init_Xo_.begin(), init_Xo_.end(), Xo_.begin());
+    if (flags_.reset_xi()) std::copy(init_Xi_.begin(), init_Xi_.end(), Xi_.begin());
+    if (flags_.reset_xo()) std::copy(init_Xo_.begin(), init_Xo_.end(), Xo_.begin());
 }
 
 void ion_state::zero_current() {
@@ -109,13 +88,10 @@ void ion_state::zero_current() {
 
 void ion_state::reset() {
     zero_current();
-    if (write_Xi_) std::copy(reset_Xi_.begin(), reset_Xi_.end(), Xi_.begin());
-    if (write_Xo_) std::copy(reset_Xo_.begin(), reset_Xo_.end(), Xo_.begin());
-    if (write_eX_) std::copy(init_eX_.begin(), init_eX_.end(), eX_.begin());
-    // This goes _last_ or at least after Xi since we might have removed reset_Xi
-    // when Xi is constant. Thus conditionally resetting Xi first and then copying
-    // Xi -> Xd is safe in all cases.
-    if (write_Xd_) std::copy(Xi_.begin(), Xi_.end(), Xd_.begin());
+    if (flags_.reset_xi()) std::copy(reset_Xi_.begin(), reset_Xi_.end(), Xi_.begin());
+    if (flags_.reset_xo()) std::copy(reset_Xo_.begin(), reset_Xo_.end(), Xo_.begin());
+    if (flags_.reset_ex()) std::copy(init_eX_.begin(), init_eX_.end(), eX_.begin());
+    if (flags_.reset_xd()) std::copy(reset_Xi_.begin(), reset_Xi_.end(), Xd_.begin());
 }
 
 // istim_state methods:

diff --git a/arbor/backends/multicore/shared_state.hpp b/arbor/backends/multicore/shared_state.hpp
@@ -48,32 +48,25 @@ struct ARB_ARBOR_API ion_state {
     using solver_type = diffusion_solver;
     using solver_ptr  = std::unique_ptr<solver_type>;
 
-    unsigned alignment = 1; // Alignment and padding multiple.
+    unsigned alignment = 1;   // Alignment and padding multiple.
+
+    ion_data_flags flags_;    // Track what and when to reset / allocate
+
+    iarray node_index_;       // Instance to CV map.
+    array iX_;                // (A/m²)  current density
+    array eX_;                // (mV)    reversal potential
+    array Xi_;                // (mM)    internal concentration
+    array Xd_;                // (mM)    diffusive internal concentration
+    array Xo_;                // (mM)    external concentration
+    array gX_;                // (kS/m²) per-species conductivity
+
+    array init_Xi_;           // (mM) area-weighted initial internal concentration
+    array init_Xo_;           // (mM) area-weighted initial external concentration
+    array reset_Xi_;          // (mM) area-weighted user-set internal concentration
+    array reset_Xo_;          // (mM) area-weighted user-set internal concentration
+    array init_eX_;           // (mV) initial reversal potential
 
-    bool write_eX_:1;          // is eX written?
-    bool write_Xo_:1;          // is Xo written?
-    bool write_Xi_:1;          // is Xi written?
-    bool write_Xd_:1;          // is Xd written?
-    bool read_eX_:1;           // is eX read?
-    bool read_Xo_:1;           // is Xo read?
-    bool read_Xi_:1;           // is Xi read?
-    bool read_Xd_:1;           // is Xd read?
-
-    iarray node_index_;     // Instance to CV map.
-    array iX_;              // (A/m²)  current density
-    array eX_;              // (mV)    reversal potential
-    array Xi_;              // (mM)    internal concentration
-    array Xd_;              // (mM)    diffusive internal concentration
-    array Xo_;              // (mM)    external concentration
-    array gX_;              // (kS/m²) per-species conductivity
-
-    array init_Xi_;         // (mM) area-weighted initial internal concentration
-    array init_Xo_;         // (mM) area-weighted initial external concentration
-    array reset_Xi_;        // (mM) area-weighted user-set internal concentration
-    array reset_Xo_;        // (mM) area-weighted user-set internal concentration
-    array init_eX_;         // (mV) initial reversal potential
-
-    array charge;           // charge of ionic species (global value, length 1)
+    array charge;             // charge of ionic species (global value, length 1)
 
     solver_ptr solver = nullptr;