c api test wrappers

eth-cscs · Jun 17, 2024 · 09ae24c · 09ae24c
1 parent 4c5ebe1
commit 09ae24c
Show file tree

Hide file tree

Showing 3 changed files with 210 additions and 0 deletions.
diff --git a/include/dlaf_c/eigensolver/gen_eigensolver.h b/include/dlaf_c/eigensolver/gen_eigensolver.h
@@ -58,6 +58,52 @@ DLAF_EXTERN_C int dlaf_hermitian_generalized_eigensolver_z(
     dlaf_complex_z* b, const struct DLAF_descriptor dlaf_descb, double* w, dlaf_complex_z* z,
     const struct DLAF_descriptor dlaf_descz) DLAF_NOEXCEPT;
 
+/// Generalized eigensolver
+///
+/// @pre The matrices \f$\mathbf{A}\f$, \f$\mathbf{B}\f$,  and \f$\mathbf{Z}\f$ are assumed to be
+/// distributed and in host memory. The vector of eigenvalues \f$\mathbf{w}\f$ is assumed to be local
+/// (non-distributed) and in host memory. Moving to and from GPU memory is handled internally.
+///
+/// @pre The matrix \f$\mathbf{B}\f$ is assumed to be factorized; it is the result of a Cholesky
+/// factorization
+///
+/// @post The pika runtime is resumed when this function is called and suspended when the call
+/// terminates.
+///
+/// @param dlaf_context context associated to the DLA-Future grid created with @ref dlaf_create_grid
+/// @param uplo indicates whether the upper ('U') or lower ('L') triangular part of the global submatrix
+/// \f$\mathbf{A}\f$ is referenced
+/// @param a Local part of the global matrix \f$\mathbf{A}\f$
+/// @param dlaf_desca DLA-Future descriptor of the global matrix \f$\mathbf{A}\f$
+/// @param b Local part of the Cholesky factorization of the global matrix \f$\mathbf{B}\f$
+/// @param dlaf_descb DLA-Future descriptor of the global matrix \f$\mathbf{B}\f$
+/// @param w Local vector of eigenvalues (non-distributed)
+/// @param z Local part of the global matrix \f$\mathbf{Z}\f$
+/// @param dlaf_descz DLA-Future descriptor of the global matrix \f$\mathbf{Z}\f$
+/// @return 0 if the eigensolver completed normally
+DLAF_EXTERN_C int dlaf_symmetric_generalized_eigensolver_factorized_s(
+    const int dlaf_context, const char uplo, float* a, const struct DLAF_descriptor dlaf_desca, float* b,
+    const struct DLAF_descriptor dlaf_descb, float* w, float* z,
+    const struct DLAF_descriptor dlaf_descz) DLAF_NOEXCEPT;
+
+/// @copydoc dlaf_symmetric_generalized_eigensolver_factorized_s
+DLAF_EXTERN_C int dlaf_symmetric_generalized_eigensolver_factorized_d(
+    const int dlaf_context, const char uplo, double* a, const struct DLAF_descriptor dlaf_desca,
+    double* b, const struct DLAF_descriptor dlaf_descb, double* w, double* z,
+    const struct DLAF_descriptor dlaf_descz) DLAF_NOEXCEPT;
+
+/// @copydoc dlaf_symmetric_generalized_eigensolver_factorized_s
+DLAF_EXTERN_C int dlaf_hermitian_generalized_eigensolver_factorized_c(
+    const int dlaf_context, const char uplo, dlaf_complex_c* a, const struct DLAF_descriptor dlaf_desca,
+    dlaf_complex_c* b, const struct DLAF_descriptor dlaf_descb, float* w, dlaf_complex_c* z,
+    const struct DLAF_descriptor dlaf_descz) DLAF_NOEXCEPT;
+
+/// @copydoc dlaf_symmetric_generalized_eigensolver_factorized_s
+DLAF_EXTERN_C int dlaf_hermitian_generalized_eigensolver_factorized_z(
+    const int dlaf_context, const char uplo, dlaf_complex_z* a, const struct DLAF_descriptor dlaf_desca,
+    dlaf_complex_z* b, const struct DLAF_descriptor dlaf_descb, double* w, dlaf_complex_z* z,
+    const struct DLAF_descriptor dlaf_descz) DLAF_NOEXCEPT;
+
 #ifdef DLAF_WITH_SCALAPACK
 
 /// Generalized eigensolver
@@ -121,4 +167,72 @@ DLAF_EXTERN_C void dlaf_pzhegvx(const char uplo, const int n, dlaf_complex_z* a,
                                 const int jb, const int descb[9], double* w, dlaf_complex_z* z,
                                 const int iz, const int jz, const int descz[9], int* info) DLAF_NOEXCEPT;
 
+/// Generalized eigensolver
+///
+/// @remark This function is only available when DLAF_WITH_SCALAPACK=ON.
+///
+/// @pre The matrices \f$\mathbf{A}\f$, \f$\mathbf{B}\f$, and \f$\mathbf{Z}\f$ are assumed to be
+/// distributed and in host memory. The vector of eigenvalues \f$\mathbf{w}\f$ is assumed to be local
+/// (non-distributed) and in host memory. Moving to and from GPU memory is handled internally.
+///
+/// @pre The matrix \f$\mathbf{B}\f$ is assumed to be factorized; it is the result of a Cholesky
+/// factorization
+///
+/// @pre Submatrices are currently not supported, so @p n is the size of the full matrices
+/// \f$\mathbf{A}\f$, \f$\mathbf{B}\f$, and \f$\mathbf{Z}\f$ and @p ia, @p ja, @p ib, @p jb, @p iz, and
+/// @p jz need to be 1.
+///
+/// @post The pika runtime is resumed when this function is called and suspended when the call
+/// terminates.
+///
+/// @param uplo indicates whether the upper ('U') or lower ('L') triangular part of the global submatrix
+/// \f$\mathbf{A}\f$ is referenced
+/// @param n order of the sumbatrix \f$\mathbf{A}\f$ used in the computation
+/// @param a Local part of the global matrix \f$\mathbf{A}\f$
+/// @param ia row index of the global matrix \f$\mathbf{A}\f$ identifying the first row of the submatrix
+/// $A$, has to be 1
+/// @param ja column index of the global matrix \f$\mathbf{A}\f$ identifying the first column of the
+/// submatrix \f$\mathbf{A}\f$, has to be 1
+/// @param desca ScaLAPACK array descriptor of the global matrix \f$\mathbf{A}\f$
+/// @param b Local part of the Cholesky factorization of the global matrix \f$\mathbf{B}\f$
+/// @param ib row index of the global matrix \f$\mathbf{B}\f$ identifying the first row of the submatrix
+/// \f$\mathbf{B}\f$, has to be 1
+/// @param jb column index of the global matrix \f$\mathbf{A}\f$ identifying the first column of the
+/// submatrix \f$\mathbf{B}\f$, has to be 1
+/// @param descb ScaLAPACK array descriptor of the global matrix \f$\mathbf{B}\f$
+/// @param w Local vector of eigenvalues (non-distributed)
+/// @param z Local part of the global matrix \f$\mathbf{Z}\f$
+/// @param iz row index of the global matrix \f$\mathbf{Z}\f$ identifying the first row of the submatrix
+/// \f$\mathbf{Z}\f$, has to be 1
+/// @param jz column index of the global matrix \f$\mathbf{A}\f$ identifying the first column of the
+/// submatrix \f$\mathbf{A}\f$, has to be 1
+/// @param descz ScaLAPACK array descriptor of the global matrix \f$\mathbf{Z}\f$
+/// @param[out] info 0 if the eigensolver completed normally
+DLAF_EXTERN_C void dlaf_pssygvx_factorized(const char uplo, const int n, float* a, const int ia,
+                                           const int ja, const int desca[9], float* b, const int ib,
+                                           const int jb, const int descb[9], float* w, float* z,
+                                           const int iz, const int jz, const int descz[9],
+                                           int* info) DLAF_NOEXCEPT;
+
+/// @copydoc dlaf_pssygvx_factorized
+DLAF_EXTERN_C void dlaf_pdsygvx_factorized(const char uplo, const int n, double* a, const int ia,
+                                           const int ja, const int desca[9], double* b, const int ib,
+                                           const int jb, const int descb[9], double* w, double* z,
+                                           const int iz, const int jz, const int descz[9],
+                                           int* info) DLAF_NOEXCEPT;
+
+/// @copydoc dlaf_pssygvx_factorized
+DLAF_EXTERN_C void dlaf_pchegvx_factorized(const char uplo, const int n, dlaf_complex_c* a, const int ia,
+                                           const int ja, const int desca[9], dlaf_complex_c* b,
+                                           const int ib, const int jb, const int descb[9], float* w,
+                                           dlaf_complex_c* z, const int iz, const int jz,
+                                           const int descz[9], int* info) DLAF_NOEXCEPT;
+
+/// @copydoc dlaf_pssygvx_factorized
+DLAF_EXTERN_C void dlaf_pzhegvx_factorized(const char uplo, const int n, dlaf_complex_z* a, const int ia,
+                                           const int ja, const int desca[9], dlaf_complex_z* b,
+                                           const int ib, const int jb, const int descb[9], double* w,
+                                           dlaf_complex_z* z, const int iz, const int jz,
+                                           const int descz[9], int* info) DLAF_NOEXCEPT;
+
 #endif
diff --git a/test/unit/c_api/eigensolver/test_gen_eigensolver_c_api_wrapper.c b/test/unit/c_api/eigensolver/test_gen_eigensolver_c_api_wrapper.c
@@ -44,6 +44,36 @@ int C_dlaf_hermitian_generalized_eigensolver_z(const int dlaf_context, const cha
   return dlaf_hermitian_generalized_eigensolver_z(dlaf_context, uplo, a, desca, b, descb, w, z, descz);
 }
 
+int C_dlaf_symmetric_generalized_eigensolver_factorized_s(const int dlaf_context, const char uplo, float* a,
+                                               const struct DLAF_descriptor desca, float* b,
+                                               const struct DLAF_descriptor descb, float* w, float* z,
+                                               const struct DLAF_descriptor descz) {
+  return dlaf_symmetric_generalized_eigensolver_factorized_s(dlaf_context, uplo, a, desca, b, descb, w, z, descz);
+}
+
+int C_dlaf_symmetric_generalized_eigensolver_factorized_d(const int dlaf_context, const char uplo, double* a,
+                                               const struct DLAF_descriptor desca, double* b,
+                                               const struct DLAF_descriptor descb, double* w, double* z,
+                                               const struct DLAF_descriptor descz) {
+  return dlaf_symmetric_generalized_eigensolver_factorized_d(dlaf_context, uplo, a, desca, b, descb, w, z, descz);
+}
+
+int C_dlaf_hermitian_generalized_eigensolver_factorized_c(const int dlaf_context, const char uplo,
+                                               dlaf_complex_c* a, const struct DLAF_descriptor desca,
+                                               dlaf_complex_c* b, const struct DLAF_descriptor descb,
+                                               float* w, dlaf_complex_c* z,
+                                               const struct DLAF_descriptor descz) {
+  return dlaf_hermitian_generalized_eigensolver_factorized_c(dlaf_context, uplo, a, desca, b, descb, w, z, descz);
+}
+
+int C_dlaf_hermitian_generalized_eigensolver_factorized_z(const int dlaf_context, const char uplo,
+                                               dlaf_complex_z* a, const struct DLAF_descriptor desca,
+                                               dlaf_complex_z* b, const struct DLAF_descriptor descb,
+                                               double* w, dlaf_complex_z* z,
+                                               const struct DLAF_descriptor descz) {
+  return dlaf_hermitian_generalized_eigensolver_factorized_z(dlaf_context, uplo, a, desca, b, descb, w, z, descz);
+}
+
 #ifdef DLAF_WITH_SCALAPACK
 
 void C_dlaf_pssygvx(char uplo, const int m, float* a, const int ia, const int ja, const int desca[9],
@@ -72,4 +102,30 @@ void C_dlaf_pzhegvx(const char uplo, const int m, dlaf_complex_z* a, const int i
   dlaf_pzhegvx(uplo, m, a, ia, ja, desca, b, ib, jb, descb, w, z, iz, jz, descz, info);
 }
 
+void C_dlaf_pssygvx_factorized(char uplo, const int m, float* a, const int ia, const int ja, const int desca[9],
+                    float* b, const int ib, const int jb, const int descb[9], float* w, float* z,
+                    const int iz, const int jz, const int descz[9], int* info) {
+  dlaf_pssygvx(uplo, m, a, ia, ja, desca, b, ib, jb, descb, w, z, iz, jz, descz, info);
+}
+
+void C_dlaf_pdsygvx_factorized(const char uplo, const int m, double* a, const int ia, const int ja,
+                    const int desca[9], double* b, const int ib, const int jb, const int descb[9],
+                    double* w, double* z, const int iz, const int jz, const int descz[9], int* info) {
+  dlaf_pdsygvx_factorized(uplo, m, a, ia, ja, desca, b, ib, jb, descb, w, z, iz, jz, descz, info);
+}
+
+void C_dlaf_pchegvx_factorized(const char uplo, const int m, dlaf_complex_c* a, const int ia, const int ja,
+                    const int desca[9], dlaf_complex_c* b, const int ib, const int jb,
+                    const int descb[9], float* w, dlaf_complex_c* z, const int iz, const int jz,
+                    const int descz[9], int* info) {
+  dlaf_pchegvx_factorized(uplo, m, a, ia, ja, desca, b, ib, jb, descb, w, z, iz, jz, descz, info);
+}
+
+void C_dlaf_pzhegvx_factorized(const char uplo, const int m, dlaf_complex_z* a, const int ia, const int ja,
+                    const int desca[9], dlaf_complex_z* b, const int ib, const int jb,
+                    const int descb[9], double* w, dlaf_complex_z* z, const int iz, const int jz,
+                    const int descz[9], int* info) {
+  dlaf_pzhegvx_factorized(uplo, m, a, ia, ja, desca, b, ib, jb, descb, w, z, iz, jz, descz, info);
+}
+
 #endif
diff --git a/test/unit/c_api/eigensolver/test_gen_eigensolver_c_api_wrapper.h b/test/unit/c_api/eigensolver/test_gen_eigensolver_c_api_wrapper.h
@@ -34,6 +34,24 @@ DLAF_EXTERN_C int C_dlaf_hermitian_generalized_eigensolver_z(
     dlaf_complex_z* b, const struct DLAF_descriptor descb, double* w, dlaf_complex_z* z,
     const struct DLAF_descriptor descz);
 
+DLAF_EXTERN_C int C_dlaf_symmetric_generalized_eigensolver_factorized_s(
+    const int dlaf_context, const char uplo, float* a, const struct DLAF_descriptor desca, float* b,
+    const struct DLAF_descriptor descb, float* w, float* z, const struct DLAF_descriptor descz);
+
+DLAF_EXTERN_C int C_dlaf_symmetric_generalized_eigensolver_factorized_d(
+    const int dlaf_context, const char uplo, double* a, const struct DLAF_descriptor desca, double* b,
+    const struct DLAF_descriptor descb, double* w, double* z, const struct DLAF_descriptor descz);
+
+DLAF_EXTERN_C int C_dlaf_hermitian_generalized_eigensolver_factorized_c(
+    const int dlaf_context, const char uplo, dlaf_complex_c* a, const struct DLAF_descriptor desca,
+    dlaf_complex_c* b, const struct DLAF_descriptor descb, float* w, dlaf_complex_c* z,
+    const struct DLAF_descriptor descz);
+
+DLAF_EXTERN_C int C_dlaf_hermitian_generalized_eigensolver_factorized_z(
+    const int dlaf_context, const char uplo, dlaf_complex_z* a, const struct DLAF_descriptor desca,
+    dlaf_complex_z* b, const struct DLAF_descriptor descb, double* w, dlaf_complex_z* z,
+    const struct DLAF_descriptor descz);
+
 #ifdef DLAF_WITH_SCALAPACK
 DLAF_EXTERN_C void C_dlaf_pssygvx(char uplo, const int m, float* a, const int ia, const int ja,
                                   const int desca[9], float* b, const int ib, const int jb,
@@ -54,4 +72,26 @@ DLAF_EXTERN_C void C_dlaf_pzhegvx(const char uplo, const int m, dlaf_complex_z*
                                   const int ja, const int desca[9], dlaf_complex_z* b, const int ib,
                                   const int jb, const int descb[9], double* w, dlaf_complex_z* z,
                                   const int iz, const int jz, const int descz[9], int* info);
+
+DLAF_EXTERN_C void C_dlaf_pssygvx_factorized(char uplo, const int m, float* a, const int ia,
+                                             const int ja, const int desca[9], float* b, const int ib,
+                                             const int jb, const int descb[9], float* w, float* z,
+                                             const int iz, const int jz, const int descz[9], int* info);
+
+DLAF_EXTERN_C void C_dlaf_pdsygvx_factorized(const char uplo, const int m, double* a, const int ia,
+                                             const int ja, const int desca[9], double* b, const int ib,
+                                             const int jb, const int descb[9], double* w, double* z,
+                                             const int iz, const int jz, const int descz[9], int* info);
+
+DLAF_EXTERN_C void C_dlaf_pchegvx_factorized(const char uplo, const int m, dlaf_complex_c* a,
+                                             const int ia, const int ja, const int desca[9],
+                                             dlaf_complex_c* b, const int ib, const int jb,
+                                             const int descb[9], float* w, dlaf_complex_c* z,
+                                             const int iz, const int jz, const int descz[9], int* info);
+
+DLAF_EXTERN_C void C_dlaf_pzhegvx_factorized(const char uplo, const int m, dlaf_complex_z* a,
+                                             const int ia, const int ja, const int desca[9],
+                                             dlaf_complex_z* b, const int ib, const int jb,
+                                             const int descb[9], double* w, dlaf_complex_z* z,
+                                             const int iz, const int jz, const int descz[9], int* info);
 #endif