Refactor: rename dngvd files (#6543)

* Rename dngvd to hegvd, add comment & docs

* Update hegvd_op docs

* Rename source files

* Change dngv/dnev to hegv/heev in source

* Rename pyabacus dngvd

Author: Cstandardlib

python/pyabacus/src/hsolver/CMakeLists.txt

@@ -9,7 +9,7 @@ list(APPEND _diago
     ${HSOLVER_PATH}/diago_pxxxgvx.cpp
 
 
-    ${HSOLVER_PATH}/kernels/dngvd_op.cpp
+    ${HSOLVER_PATH}/kernels/hegvd_op.cpp
     ${HSOLVER_PATH}/kernels/bpcg_kernel_op.cpp
     # dependency
     ${BASE_PATH}/kernels/math_kernel_op.cpp

source/CMakeLists.txt

@@ -35,7 +35,7 @@ list(APPEND device_srcs
   source_pw/module_pwdft/kernels/meta_op.cpp
   source_pw/module_stodft/kernels/hpsi_norm_op.cpp
   source_basis/module_pw/kernels/pw_op.cpp
-  source_hsolver/kernels/dngvd_op.cpp
+  source_hsolver/kernels/hegvd_op.cpp
   source_hsolver/kernels/bpcg_kernel_op.cpp
   source_estate/kernels/elecstate_op.cpp
 
@@ -70,7 +70,7 @@ if(USE_CUDA)
     source_pw/module_stodft/kernels/cuda/hpsi_norm_op.cu
     source_pw/module_pwdft/kernels/cuda/onsite_op.cu
     source_basis/module_pw/kernels/cuda/pw_op.cu
-    source_hsolver/kernels/cuda/dngvd_op.cu
+    source_hsolver/kernels/cuda/hegvd_op.cu
     source_hsolver/kernels/cuda/bpcg_kernel_op.cu
     source_estate/kernels/cuda/elecstate_op.cu
 
@@ -101,7 +101,7 @@ if(USE_ROCM)
     source_pw/module_pwdft/kernels/rocm/onsite_op.hip.cu
     source_pw/module_stodft/kernels/rocm/hpsi_norm_op.hip.cu
     source_basis/module_pw/kernels/rocm/pw_op.hip.cu
-    source_hsolver/kernels/rocm/dngvd_op.hip.cu
+    source_hsolver/kernels/rocm/hegvd_op.hip.cu
     source_hsolver/kernels/rocm/bpcg_kernel_op.hip.cu
     source_estate/kernels/rocm/elecstate_op.hip.cu
 

source/Makefile.Objects

@@ -395,7 +395,7 @@ OBJS_HSOLVER=diago_cg.o\
     hsolver_lcaopw.o\
     hsolver_pw_sdft.o\
     diago_iter_assist.o\
-    dngvd_op.o\
+    hegvd_op.o\
     bpcg_kernel_op.o\
     diag_const_nums.o\
     diag_hs_para.o\

source/source_base/module_container/ATen/kernels/cuda/lapack.cu

@@ -88,7 +88,7 @@ struct lapack_potrf<T, DEVICE_GPU> {
 };
 
 template <typename T>
-struct lapack_dnevd<T, DEVICE_GPU> {
+struct lapack_heevd<T, DEVICE_GPU> {
     using Real = typename GetTypeReal<T>::type;
     void operator()(
         const char& jobz,
@@ -97,12 +97,12 @@ struct lapack_dnevd<T, DEVICE_GPU> {
         const int& dim,
         Real* eigen_val)
     {
-        cuSolverConnector::dnevd(cusolver_handle, jobz, uplo, dim, Mat, dim, eigen_val);
+        cuSolverConnector::heevd(cusolver_handle, jobz, uplo, dim, Mat, dim, eigen_val);
     }
 };
 
 template <typename T>
-struct lapack_dngvd<T, DEVICE_GPU> {
+struct lapack_hegvd<T, DEVICE_GPU> {
     using Real = typename GetTypeReal<T>::type;
     void operator()(
         const int& itype,
@@ -113,7 +113,7 @@ struct lapack_dngvd<T, DEVICE_GPU> {
         const int& dim,
         Real* eigen_val)
     {
-        cuSolverConnector::dngvd(cusolver_handle, itype, jobz, uplo, dim, Mat_A, dim, Mat_B, dim, eigen_val);
+        cuSolverConnector::hegvd(cusolver_handle, itype, jobz, uplo, dim, Mat_A, dim, Mat_B, dim, eigen_val);
     }
 };
 
@@ -175,15 +175,15 @@ template struct lapack_potrf<double, DEVICE_GPU>;
 template struct lapack_potrf<std::complex<float>,  DEVICE_GPU>;
 template struct lapack_potrf<std::complex<double>, DEVICE_GPU>;
 
-template struct lapack_dnevd<float,  DEVICE_GPU>;
-template struct lapack_dnevd<double, DEVICE_GPU>;
-template struct lapack_dnevd<std::complex<float>,  DEVICE_GPU>;
-template struct lapack_dnevd<std::complex<double>, DEVICE_GPU>;
+template struct lapack_heevd<float,  DEVICE_GPU>;
+template struct lapack_heevd<double, DEVICE_GPU>;
+template struct lapack_heevd<std::complex<float>,  DEVICE_GPU>;
+template struct lapack_heevd<std::complex<double>, DEVICE_GPU>;
 
-template struct lapack_dngvd<float,  DEVICE_GPU>;
-template struct lapack_dngvd<double, DEVICE_GPU>;
-template struct lapack_dngvd<std::complex<float>,  DEVICE_GPU>;
-template struct lapack_dngvd<std::complex<double>, DEVICE_GPU>;
+template struct lapack_hegvd<float,  DEVICE_GPU>;
+template struct lapack_hegvd<double, DEVICE_GPU>;
+template struct lapack_hegvd<std::complex<float>,  DEVICE_GPU>;
+template struct lapack_hegvd<std::complex<double>, DEVICE_GPU>;
 
 template struct lapack_getrf<float,  DEVICE_GPU>;
 template struct lapack_getrf<double, DEVICE_GPU>;

source/source_base/module_container/ATen/kernels/lapack.cpp

@@ -63,7 +63,7 @@ struct lapack_potrf<T, DEVICE_CPU> {
 };
 
 template <typename T>
-struct lapack_dnevd<T, DEVICE_CPU> {
+struct lapack_heevd<T, DEVICE_CPU> {
     using Real = typename GetTypeReal<T>::type;
     void operator()(
         const char& jobz,
@@ -85,15 +85,15 @@ struct lapack_dnevd<T, DEVICE_CPU> {
         Tensor iwork(DataTypeToEnum<int>::value, DeviceType::CpuDevice, {liwork});
         iwork.zero();
 
-        lapackConnector::dnevd(jobz, uplo, dim, Mat, dim, eigen_val,  work.data<T>(), lwork, rwork.data<Real>(), lrwork, iwork.data<int>(), liwork, info);
+        lapackConnector::heevd(jobz, uplo, dim, Mat, dim, eigen_val,  work.data<T>(), lwork, rwork.data<Real>(), lrwork, iwork.data<int>(), liwork, info);
         if (info != 0) {
-            throw std::runtime_error("dnevd failed with info = " + std::to_string(info));
+            throw std::runtime_error("heevd failed with info = " + std::to_string(info));
         }
     }
 };
 
 template <typename T>
-struct lapack_dngvd<T, DEVICE_CPU> {
+struct lapack_hegvd<T, DEVICE_CPU> {
     using Real = typename GetTypeReal<T>::type;
     void operator()(
         const int& itype,
@@ -117,9 +117,9 @@ struct lapack_dngvd<T, DEVICE_CPU> {
         Tensor iwork(DataType::DT_INT, DeviceType::CpuDevice, {liwork});
         iwork.zero();
 
-        lapackConnector::dngvd(itype, jobz, uplo, dim, Mat_A, dim, Mat_B, dim, eigen_val, work.data<T>(), lwork, rwork.data<Real>(), lrwork, iwork.data<int>(), liwork, info);
+        lapackConnector::hegvd(itype, jobz, uplo, dim, Mat_A, dim, Mat_B, dim, eigen_val, work.data<T>(), lwork, rwork.data<Real>(), lrwork, iwork.data<int>(), liwork, info);
         if (info != 0) {
-            throw std::runtime_error("dngvd failed with info = " + std::to_string(info));
+            throw std::runtime_error("hegvd failed with info = " + std::to_string(info));
         }
     }
 };
@@ -194,15 +194,15 @@ template struct lapack_trtri<double, DEVICE_CPU>;
 template struct lapack_trtri<std::complex<float>,  DEVICE_CPU>;
 template struct lapack_trtri<std::complex<double>, DEVICE_CPU>;
 
-template struct lapack_dnevd<float,  DEVICE_CPU>;
-template struct lapack_dnevd<double, DEVICE_CPU>;
-template struct lapack_dnevd<std::complex<float>,  DEVICE_CPU>;
-template struct lapack_dnevd<std::complex<double>, DEVICE_CPU>;
+template struct lapack_heevd<float,  DEVICE_CPU>;
+template struct lapack_heevd<double, DEVICE_CPU>;
+template struct lapack_heevd<std::complex<float>,  DEVICE_CPU>;
+template struct lapack_heevd<std::complex<double>, DEVICE_CPU>;
 
-template struct lapack_dngvd<float,  DEVICE_CPU>;
-template struct lapack_dngvd<double, DEVICE_CPU>;
-template struct lapack_dngvd<std::complex<float>,  DEVICE_CPU>;
-template struct lapack_dngvd<std::complex<double>, DEVICE_CPU>;
+template struct lapack_hegvd<float,  DEVICE_CPU>;
+template struct lapack_hegvd<double, DEVICE_CPU>;
+template struct lapack_hegvd<std::complex<float>,  DEVICE_CPU>;
+template struct lapack_hegvd<std::complex<double>, DEVICE_CPU>;
 
 template struct lapack_getrf<float,  DEVICE_CPU>;
 template struct lapack_getrf<double, DEVICE_CPU>;

source/source_base/module_container/ATen/kernels/lapack.h

@@ -41,7 +41,7 @@ struct lapack_potrf {
 
 
 template <typename T, typename Device>
-struct lapack_dnevd {
+struct lapack_heevd {
     using Real = typename GetTypeReal<T>::type;
     void operator()(
         const char& jobz,
@@ -53,7 +53,7 @@ struct lapack_dnevd {
 
 
 template <typename T, typename Device>
-struct lapack_dngvd {
+struct lapack_hegvd {
     using Real = typename GetTypeReal<T>::type;
     void operator()(
         const int& itype,

source/source_base/module_container/ATen/kernels/rocm/lapack.hip.cu

@@ -94,7 +94,7 @@ struct lapack_potrf<T, DEVICE_GPU> {
 };
 
 template <typename T>
-struct lapack_dnevd<T, DEVICE_GPU> {
+struct lapack_heevd<T, DEVICE_GPU> {
     using Real = typename GetTypeReal<T>::type;
     void operator()(
         const char& jobz,
@@ -103,19 +103,19 @@ struct lapack_dnevd<T, DEVICE_GPU> {
         const int& dim,
         Real* eigen_val)
     {
-        // hipSolverConnector::dnevd(hipsolver_handle, jobz, uplo, dim, Mat, dim, eigen_val);
+        // hipSolverConnector::heevd(hipsolver_handle, jobz, uplo, dim, Mat, dim, eigen_val);
         std::vector<T> H_Mat(dim * dim, static_cast<T>(0.0));
         std::vector<Real> H_eigen_val(dim, static_cast<Real>(0.0));
         hipMemcpy(H_Mat.data(), Mat, sizeof(T) * H_Mat.size(), hipMemcpyDeviceToHost);
         hipMemcpy(H_eigen_val.data(), eigen_val, sizeof(Real) * H_eigen_val.size(), hipMemcpyDeviceToHost);
-        lapack_dnevd<T, DEVICE_CPU>()(jobz, uplo, H_Mat.data(), dim, H_eigen_val.data());
+        lapack_heevd<T, DEVICE_CPU>()(jobz, uplo, H_Mat.data(), dim, H_eigen_val.data());
         hipMemcpy(Mat, H_Mat.data(), sizeof(T) * H_Mat.size(), hipMemcpyHostToDevice);
         hipMemcpy(eigen_val, H_eigen_val.data(), sizeof(Real) * H_eigen_val.size(), hipMemcpyHostToDevice);
     }
 };
 
 template <typename T>
-struct lapack_dngvd<T, DEVICE_GPU> {
+struct lapack_hegvd<T, DEVICE_GPU> {
     using Real = typename GetTypeReal<T>::type;
     void operator()(
         const int& itype,
@@ -126,7 +126,7 @@ struct lapack_dngvd<T, DEVICE_GPU> {
         const int& dim,
         Real* eigen_val)
     {
-        hipSolverConnector::dngvd(hipsolver_handle, itype, jobz, uplo, dim, Mat_A, dim, Mat_B, dim, eigen_val);
+        hipSolverConnector::hegvd(hipsolver_handle, itype, jobz, uplo, dim, Mat_A, dim, Mat_B, dim, eigen_val);
     }
 };
 
@@ -145,15 +145,15 @@ template struct lapack_potrf<double, DEVICE_GPU>;
 template struct lapack_potrf<std::complex<float>,  DEVICE_GPU>;
 template struct lapack_potrf<std::complex<double>, DEVICE_GPU>;
 
-template struct lapack_dnevd<float,  DEVICE_GPU>;
-template struct lapack_dnevd<double, DEVICE_GPU>;
-template struct lapack_dnevd<std::complex<float>,  DEVICE_GPU>;
-template struct lapack_dnevd<std::complex<double>, DEVICE_GPU>;
+template struct lapack_heevd<float,  DEVICE_GPU>;
+template struct lapack_heevd<double, DEVICE_GPU>;
+template struct lapack_heevd<std::complex<float>,  DEVICE_GPU>;
+template struct lapack_heevd<std::complex<double>, DEVICE_GPU>;
 
-template struct lapack_dngvd<float,  DEVICE_GPU>;
-template struct lapack_dngvd<double, DEVICE_GPU>;
-template struct lapack_dngvd<std::complex<float>,  DEVICE_GPU>;
-template struct lapack_dngvd<std::complex<double>, DEVICE_GPU>;
+template struct lapack_hegvd<float,  DEVICE_GPU>;
+template struct lapack_hegvd<double, DEVICE_GPU>;
+template struct lapack_hegvd<std::complex<float>,  DEVICE_GPU>;
+template struct lapack_hegvd<std::complex<double>, DEVICE_GPU>;
 
 } // namespace kernels
 } // namespace container

source/source_base/module_container/ATen/kernels/test/lapack_test.cpp

@@ -92,14 +92,14 @@ TYPED_TEST(LapackTest, Potrf) {
     EXPECT_EQ(A, C);
 }
 
-TYPED_TEST(LapackTest, dnevd) {
+TYPED_TEST(LapackTest, heevd) {
     using Type = typename std::tuple_element<0, decltype(TypeParam())>::type;
     using Real = typename GetTypeReal<Type>::type;
     using Device = typename std::tuple_element<1, decltype(TypeParam())>::type;
 
     blas_gemm<Type, Device> gemmCalculator;
     blas_axpy<Type, Device> axpyCalculator;
-    lapack_dnevd<Type, Device> dnevdCalculator;
+    lapack_heevd<Type, Device> heevdCalculator;
 
     const int dim = 3;
     Tensor A = std::move(Tensor({static_cast<Type>(4.0), static_cast<Type>(1.0), static_cast<Type>(1.0),
@@ -121,7 +121,7 @@ TYPED_TEST(LapackTest, dnevd) {
     const Type beta  = static_cast<Type>(0.0);
     // Note all blas and lapack operators within container are column major!
     // For this reason, we should employ 'L' instead of 'U' in the subsequent line.
-    dnevdCalculator('V', 'U', B.data<Type>(), dim, E.data<Real>());
+    heevdCalculator('V', 'U', B.data<Type>(), dim, E.data<Real>());
 
     E = E.to_device<DEVICE_CPU>();
     const Tensor Alpha = std::move(Tensor({
@@ -139,14 +139,14 @@ TYPED_TEST(LapackTest, dnevd) {
 }
 
 
-TYPED_TEST(LapackTest, dngvd) {
+TYPED_TEST(LapackTest, hegvd) {
     using Type = typename std::tuple_element<0, decltype(TypeParam())>::type;
     using Real = typename GetTypeReal<Type>::type;
     using Device = typename std::tuple_element<1, decltype(TypeParam())>::type;
 
     blas_gemm<Type, Device> gemmCalculator;
     blas_axpy<Type, Device> axpyCalculator;
-    lapack_dngvd<Type, Device> dngvdCalculator;
+    lapack_hegvd<Type, Device> hegvdCalculator;
 
     const int dim = 3;
     Tensor A = std::move(Tensor({static_cast<Type>(4.0), static_cast<Type>(1.0), static_cast<Type>(1.0),
@@ -172,7 +172,7 @@ TYPED_TEST(LapackTest, dngvd) {
     const Type beta  = static_cast<Type>(0.0);
     // Note al<l blas and lapack operators within container are column major!
     // For this reason, we should employ 'L' instead of 'U' in the subsequent line.
-    dngvdCalculator(1, 'V', 'U', B.data<Type>(), I.data<Type>(), dim, E.data<Real>());
+    hegvdCalculator(1, 'V', 'U', B.data<Type>(), I.data<Type>(), dim, E.data<Real>());
 
     E = E.to_device<DEVICE_CPU>();
     const Tensor Alpha = std::move(Tensor({