BatchLinearAlgebra.cu

#include <ATen/Context.h>
#include <ATen/cuda/CUDAContext.h>
#include <ATen/Dispatch.h>
#include <ATen/NativeFunctions.h>
#include <ATen/cuda/PinnedMemoryAllocator.h>
#include <ATen/cuda/detail/IndexUtils.cuh>

#include <ATen/native/LinearAlgebraUtils.h>
#include <ATen/native/cuda/MiscUtils.h>
#include <ATen/native/Resize.h>
#include <ATen/native/BatchLinearAlgebra.h>
#include <ATen/native/cuda/BatchLinearAlgebraLib.h>
#include <ATen/native/cpu/zmath.h>

#include <THC/THC.h> // for USE_MAGMA

#ifdef USE_MAGMA
#include <magma_types.h>
#include <magma_v2.h>

const bool use_magma_ = true;
#else
const bool use_magma_ = false;

#endif

namespace at {
namespace native {

#ifdef USE_MAGMA
template<class scalar_t>
void magmaSolve(
    magma_int_t n, magma_int_t nrhs, scalar_t* dA, magma_int_t ldda,
    magma_int_t* ipiv, scalar_t* dB, magma_int_t lddb, magma_int_t* info);

template<class scalar_t>
void magmaSolveBatched(
    magma_int_t n, magma_int_t nrhs, scalar_t** dA_array, magma_int_t ldda,
    magma_int_t** dipiv_array, scalar_t** dB_array, magma_int_t lddb,
    magma_int_t* dinfo_array, magma_int_t batch_count, const MAGMAQueue& magma_queue);

template<class scalar_t>
void magmaLu(
    magma_int_t m, magma_int_t n, scalar_t* dA, magma_int_t ldda,
    magma_int_t* ipiv, magma_int_t* info);

template<class scalar_t>
void magmaLuBatched(
    magma_int_t m, magma_int_t n, scalar_t** dA_array, magma_int_t ldda,
    magma_int_t** ipiv_array, magma_int_t* info_array, magma_int_t batchsize,
    const MAGMAQueue& magma_queue);

template<class scalar_t>
void magmaLuNoPiv(
    magma_int_t m, magma_int_t n, scalar_t* dA, magma_int_t ldda,
    magma_int_t* info);

template<class scalar_t>
void magmaLuNoPivBatched(
    magma_int_t m, magma_int_t n, scalar_t** dA_array, magma_int_t ldda,
    magma_int_t* info_array, magma_int_t batchsize, const MAGMAQueue& magma_queue);

template<class scalar_t>
inline magma_int_t magmaGetriOptimalBlocksize(magma_int_t n);

template<class scalar_t>
void magmaGetri(
    magma_int_t n, scalar_t* dA, magma_int_t ldda, magma_int_t* ipiv, scalar_t* dwork,
    magma_int_t lwork, magma_int_t* info);

template<class scalar_t>
void magmaGetriBatched(
    magma_int_t n, scalar_t** dA_array, magma_int_t ldda,
    magma_int_t** ipiv_array, scalar_t** dinvA_array, magma_int_t lddia,
    magma_int_t* info_array, magma_int_t batchsize, const MAGMAQueue& magma_queue);

template<class scalar_t>
void magmaCholeskySolve(
    magma_uplo_t uplo, magma_int_t n, magma_int_t nrhs, scalar_t* dA, magma_int_t ldda,
    scalar_t* dB, magma_int_t lddb, magma_int_t* info);

template<class scalar_t>
void magmaCholeskySolveBatched(
    magma_uplo_t uplo, magma_int_t n, magma_int_t nrhs, scalar_t** dA_array, magma_int_t ldda,
    scalar_t** dB_array, magma_int_t lddb, magma_int_t& info, magma_int_t batchsize, const MAGMAQueue& magma_queue);

template<class scalar_t>
void magmaCholesky(
    magma_uplo_t uplo, magma_int_t n, scalar_t* dA,
    magma_int_t ldda, magma_int_t* info);

template<class scalar_t>
void magmaCholeskyBatched(
    magma_uplo_t uplo, magma_int_t n, scalar_t** dA_array, magma_int_t ldda,
    magma_int_t* info_array, magma_int_t batchsize, const MAGMAQueue& magma_queue);

template<class scalar_t>
void magmaTriangularSolveBatched(
    magma_uplo_t uplo, magma_trans_t trans, magma_diag_t diag, magma_int_t m, magma_int_t n,
    scalar_t** dA_array, magma_int_t ldda, scalar_t** dB_array, magma_int_t lddb, magma_int_t batchsize,
    const MAGMAQueue& magma_queue);

template<class scalar_t>
inline magma_int_t magmaGeqrfOptimalBlocksize(magma_int_t m, magma_int_t n);

template<class scalar_t>
void magmaGeqrf(
    magma_int_t m, magma_int_t n, scalar_t* dA, magma_int_t ldda,
    scalar_t* tau, scalar_t* dT, magma_int_t* info, bool is_v2);

template<class scalar_t>
void magmaOrgqr(
    magma_int_t m, magma_int_t n, magma_int_t k, scalar_t* dA,
    magma_int_t ldda, scalar_t* tau, scalar_t* dT, magma_int_t nb, magma_int_t* info);

template<class scalar_t, class value_t=scalar_t>
void magmaSyevd(
    magma_vec_t jobz, magma_uplo_t uplo, magma_int_t n, scalar_t* dA, magma_int_t ldda,
    value_t* w, scalar_t* wA, magma_int_t ldwa, scalar_t* work, magma_int_t lwork, value_t* rwork,
    magma_int_t lrwork, magma_int_t* iwork, magma_int_t liwork, magma_int_t* info);

template<class scalar_t, class value_t=scalar_t>
void magmaEig(
    magma_vec_t jobvl, magma_vec_t jobvr, magma_int_t n, scalar_t *A, magma_int_t lda,
    scalar_t *w, scalar_t *VL, magma_int_t ldvl,
    scalar_t *VR, magma_int_t ldvr, scalar_t *work, magma_int_t lwork,
    value_t *rwork,
    magma_int_t *info);

template<class scalar_t, class value_t=scalar_t>
void magmaSvd(
    magma_vec_t jobz, magma_int_t m, magma_int_t n, scalar_t* A,
    magma_int_t lda, value_t* s, scalar_t* U, magma_int_t ldu,
    scalar_t* VT, magma_int_t ldvt, scalar_t* work, magma_int_t lwork,
    value_t* rwork,
    magma_int_t* iwork, magma_int_t* info);

template<class scalar_t>
void magmaLuSolve(
    magma_int_t n, magma_int_t nrhs, scalar_t* dA, magma_int_t ldda, magma_int_t* ipiv,
    scalar_t* dB, magma_int_t lddb, magma_int_t* info);

template<class scalar_t>
void magmaLuSolveBatched(
    magma_int_t n, magma_int_t nrhs, scalar_t** dA_array, magma_int_t ldda, magma_int_t** dipiv_array,
    scalar_t** dB_array, magma_int_t lddb, magma_int_t& info,
    magma_int_t batchsize, const MAGMAQueue& magma_queue);

template<class scalar_t>
void magmaGels(
    magma_trans_t trans, magma_int_t m, magma_int_t n, magma_int_t nrhs,
    scalar_t* dA, magma_int_t ldda, scalar_t* dB, magma_int_t lddb,
    scalar_t* hwork, magma_int_t lwork, magma_int_t* info);

template<>
void magmaSolve<double>(
    magma_int_t n, magma_int_t nrhs, double* dA, magma_int_t ldda,
    magma_int_t* ipiv, double* dB, magma_int_t lddb, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_dgesv_gpu(n, nrhs, dA, ldda, ipiv, dB, lddb, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaSolve<float>(
    magma_int_t n, magma_int_t nrhs, float* dA, magma_int_t ldda,
    magma_int_t* ipiv, float* dB, magma_int_t lddb, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_sgesv_gpu(n, nrhs, dA, ldda, ipiv, dB, lddb, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaSolve<c10::complex<double>>(
    magma_int_t n, magma_int_t nrhs, c10::complex<double>* dA, magma_int_t ldda,
    magma_int_t* ipiv, c10::complex<double>* dB, magma_int_t lddb, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_zgesv_gpu(n, nrhs,
    reinterpret_cast<magmaDoubleComplex*>(dA), ldda, ipiv,
    reinterpret_cast<magmaDoubleComplex*>(dB), lddb, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaSolve<c10::complex<float>>(
    magma_int_t n, magma_int_t nrhs, c10::complex<float>* dA, magma_int_t ldda,
    magma_int_t* ipiv, c10::complex<float>* dB, magma_int_t lddb, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_cgesv_gpu(n, nrhs,
    reinterpret_cast<magmaFloatComplex*>(dA), ldda, ipiv,
    reinterpret_cast<magmaFloatComplex*>(dB), lddb, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaSolveBatched<double>(
    magma_int_t n, magma_int_t nrhs, double** dA_array, magma_int_t ldda,
    magma_int_t** dipiv_array, double** dB_array, magma_int_t lddb,
    magma_int_t* dinfo_array, magma_int_t batch_count, const MAGMAQueue& magma_queue) {
  magma_dgesv_batched(n, nrhs, dA_array, ldda, dipiv_array, dB_array, lddb, dinfo_array, batch_count, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaSolveBatched<float>(
    magma_int_t n, magma_int_t nrhs, float** dA_array, magma_int_t ldda,
    magma_int_t** dipiv_array, float** dB_array, magma_int_t lddb,
    magma_int_t* dinfo_array, magma_int_t batch_count, const MAGMAQueue& magma_queue) {
  magma_sgesv_batched(n, nrhs, dA_array, ldda, dipiv_array, dB_array, lddb, dinfo_array, batch_count, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaSolveBatched<c10::complex<double>>(
    magma_int_t n, magma_int_t nrhs, c10::complex<double>** dA_array, magma_int_t ldda,
    magma_int_t** dipiv_array, c10::complex<double>** dB_array, magma_int_t lddb,
    magma_int_t* dinfo_array, magma_int_t batch_count, const MAGMAQueue& magma_queue) {
  magma_zgesv_batched(n, nrhs,
    reinterpret_cast<magmaDoubleComplex**>(dA_array), ldda, dipiv_array,
    reinterpret_cast<magmaDoubleComplex**>(dB_array), lddb, dinfo_array, batch_count, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaSolveBatched<c10::complex<float>>(
    magma_int_t n, magma_int_t nrhs, c10::complex<float>** dA_array, magma_int_t ldda,
    magma_int_t** dipiv_array, c10::complex<float>** dB_array, magma_int_t lddb,
    magma_int_t* dinfo_array, magma_int_t batch_count, const MAGMAQueue& magma_queue) {
  magma_cgesv_batched(n, nrhs,
    reinterpret_cast<magmaFloatComplex**>(dA_array), ldda, dipiv_array,
    reinterpret_cast<magmaFloatComplex**>(dB_array), lddb, dinfo_array, batch_count, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLu<double>(
    magma_int_t m, magma_int_t n, double* dA, magma_int_t ldda,
    magma_int_t* ipiv, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_dgetrf_gpu(m, n, dA, ldda, ipiv, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLu<float>(
    magma_int_t m, magma_int_t n, float* dA, magma_int_t ldda,
    magma_int_t* ipiv, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_sgetrf_gpu(m, n, dA, ldda, ipiv, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLu<c10::complex<double>>(
    magma_int_t m, magma_int_t n, c10::complex<double>* dA, magma_int_t ldda,
    magma_int_t* ipiv, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_zgetrf_gpu(m, n, reinterpret_cast<magmaDoubleComplex*>(dA), ldda, ipiv, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLu<c10::complex<float>>(
    magma_int_t m, magma_int_t n, c10::complex<float>* dA, magma_int_t ldda,
    magma_int_t* ipiv, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_cgetrf_gpu(m, n, reinterpret_cast<magmaFloatComplex*>(dA), ldda, ipiv, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuBatched<double>(
    magma_int_t m, magma_int_t n, double** dA_array, magma_int_t ldda,
    magma_int_t** ipiv_array, magma_int_t* info_array, magma_int_t batchsize,
    const MAGMAQueue& magma_queue) {
  magma_dgetrf_batched(m, n, dA_array, ldda, ipiv_array, info_array, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuBatched<float>(
    magma_int_t m, magma_int_t n, float** dA_array, magma_int_t ldda,
    magma_int_t** ipiv_array, magma_int_t* info_array, magma_int_t batchsize,
    const MAGMAQueue& magma_queue) {
  magma_sgetrf_batched(m, n, dA_array, ldda, ipiv_array, info_array, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuBatched<c10::complex<double>>(
    magma_int_t m, magma_int_t n, c10::complex<double>** dA_array, magma_int_t ldda,
    magma_int_t** ipiv_array, magma_int_t* info_array, magma_int_t batchsize,
    const MAGMAQueue& magma_queue) {
  magma_zgetrf_batched(m, n, reinterpret_cast<magmaDoubleComplex**>(dA_array), ldda, ipiv_array, info_array, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuBatched<c10::complex<float>>(
    magma_int_t m, magma_int_t n, c10::complex<float>** dA_array, magma_int_t ldda,
    magma_int_t** ipiv_array, magma_int_t* info_array, magma_int_t batchsize,
    const MAGMAQueue& magma_queue) {
  magma_cgetrf_batched(m, n, reinterpret_cast<magmaFloatComplex**>(dA_array), ldda, ipiv_array, info_array, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuNoPiv<double>(
    magma_int_t m, magma_int_t n, double* dA, magma_int_t ldda,
    magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_dgetrf_nopiv_gpu(m, n, dA, ldda, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuNoPiv<float>(
    magma_int_t m, magma_int_t n, float* dA, magma_int_t ldda,
    magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_sgetrf_nopiv_gpu(m, n, dA, ldda, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuNoPiv<c10::complex<double>>(
    magma_int_t m, magma_int_t n, c10::complex<double>* dA, magma_int_t ldda,
    magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_zgetrf_nopiv_gpu(m, n, reinterpret_cast<magmaDoubleComplex*>(dA), ldda, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuNoPiv<c10::complex<float>>(
    magma_int_t m, magma_int_t n, c10::complex<float>* dA, magma_int_t ldda,
    magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_cgetrf_nopiv_gpu(m, n, reinterpret_cast<magmaFloatComplex*>(dA), ldda, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuNoPivBatched<double>(
    magma_int_t m, magma_int_t n, double** dA_array, magma_int_t ldda,
    magma_int_t* info_array, magma_int_t batchsize, const MAGMAQueue& magma_queue) {
  magma_dgetrf_nopiv_batched(m, n, dA_array, ldda, info_array, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuNoPivBatched<float>(
    magma_int_t m, magma_int_t n, float** dA_array, magma_int_t ldda,
    magma_int_t* info_array, magma_int_t batchsize, const MAGMAQueue& magma_queue) {
  magma_sgetrf_nopiv_batched(m, n, dA_array, ldda, info_array, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuNoPivBatched<c10::complex<double>>(
    magma_int_t m, magma_int_t n, c10::complex<double>** dA_array, magma_int_t ldda,
    magma_int_t* info_array, magma_int_t batchsize, const MAGMAQueue& magma_queue) {
  magma_zgetrf_nopiv_batched(m, n, reinterpret_cast<magmaDoubleComplex**>(dA_array), ldda, info_array, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuNoPivBatched<c10::complex<float>>(
    magma_int_t m, magma_int_t n, c10::complex<float>** dA_array, magma_int_t ldda,
    magma_int_t* info_array, magma_int_t batchsize, const MAGMAQueue& magma_queue) {
  magma_cgetrf_nopiv_batched(m, n, reinterpret_cast<magmaFloatComplex**>(dA_array), ldda, info_array, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
inline magma_int_t magmaGetriOptimalBlocksize<double>(magma_int_t n) {
  return magma_get_dgetri_nb(n);
}

template<>
inline magma_int_t magmaGetriOptimalBlocksize<float>(magma_int_t n) {
  return magma_get_sgetri_nb(n);
}

template <>
inline magma_int_t magmaGetriOptimalBlocksize<c10::complex<double>>(
    magma_int_t n) {
  return magma_get_zgetri_nb(n);
}

template <>
inline magma_int_t magmaGetriOptimalBlocksize<c10::complex<float>>(
    magma_int_t n) {
  return magma_get_cgetri_nb(n);
}

template<>
void magmaGetri<double>(
    magma_int_t n, double* dA, magma_int_t ldda, magma_int_t* ipiv, double* dwork,
    magma_int_t lwork, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_dgetri_gpu(n, dA, ldda, ipiv, dwork, lwork, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaGetri<float>(
    magma_int_t n, float* dA, magma_int_t ldda, magma_int_t* ipiv, float* dwork,
    magma_int_t lwork, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_sgetri_gpu(n, dA, ldda, ipiv, dwork, lwork, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template <>
void magmaGetri<c10::complex<double>>(
    magma_int_t n,
    c10::complex<double>* dA,
    magma_int_t ldda,
    magma_int_t* ipiv,
    c10::complex<double>* dwork,
    magma_int_t lwork,
    magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_zgetri_gpu(
      n,
      reinterpret_cast<magmaDoubleComplex*>(dA),
      ldda,
      ipiv,
      reinterpret_cast<magmaDoubleComplex*>(dwork),
      lwork,
      info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template <>
void magmaGetri<c10::complex<float>>(
    magma_int_t n,
    c10::complex<float>* dA,
    magma_int_t ldda,
    magma_int_t* ipiv,
    c10::complex<float>* dwork,
    magma_int_t lwork,
    magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_cgetri_gpu(
      n,
      reinterpret_cast<magmaFloatComplex*>(dA),
      ldda,
      ipiv,
      reinterpret_cast<magmaFloatComplex*>(dwork),
      lwork,
      info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaGetriBatched<double>(
    magma_int_t n, double** dA_array, magma_int_t ldda,
    magma_int_t** ipiv_array, double** dinvA_array, magma_int_t lddia,
    magma_int_t* info_array, magma_int_t batchsize, const MAGMAQueue& magma_queue) {
  magma_dgetri_outofplace_batched(n, dA_array, ldda, ipiv_array, dinvA_array, lddia, info_array, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaGetriBatched<float>(
    magma_int_t n, float** dA_array, magma_int_t ldda,
    magma_int_t** ipiv_array, float** dinvA_array, magma_int_t lddia,
    magma_int_t* info_array, magma_int_t batchsize, const MAGMAQueue& magma_queue) {
  magma_sgetri_outofplace_batched(n, dA_array, ldda, ipiv_array, dinvA_array, lddia, info_array, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template <>
void magmaGetriBatched<c10::complex<double>>(
    magma_int_t n,
    c10::complex<double>** dA_array,
    magma_int_t ldda,
    magma_int_t** ipiv_array,
    c10::complex<double>** dinvA_array,
    magma_int_t lddia,
    magma_int_t* info_array,
    magma_int_t batchsize,
    const MAGMAQueue& magma_queue) {
  magma_zgetri_outofplace_batched(
      n,
      reinterpret_cast<magmaDoubleComplex**>(dA_array),
      ldda,
      ipiv_array,
      reinterpret_cast<magmaDoubleComplex**>(dinvA_array),
      lddia,
      info_array,
      batchsize,
      magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template <>
void magmaGetriBatched<c10::complex<float>>(
    magma_int_t n,
    c10::complex<float>** dA_array,
    magma_int_t ldda,
    magma_int_t** ipiv_array,
    c10::complex<float>** dinvA_array,
    magma_int_t lddia,
    magma_int_t* info_array,
    magma_int_t batchsize,
    const MAGMAQueue& magma_queue) {
  magma_cgetri_outofplace_batched(
      n,
      reinterpret_cast<magmaFloatComplex**>(dA_array),
      ldda,
      ipiv_array,
      reinterpret_cast<magmaFloatComplex**>(dinvA_array),
      lddia,
      info_array,
      batchsize,
      magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaCholeskySolve<double>(
    magma_uplo_t uplo, magma_int_t n, magma_int_t nrhs, double* dA, magma_int_t ldda,
    double* dB, magma_int_t lddb, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_dpotrs_gpu(uplo, n, nrhs, dA, ldda, dB, lddb, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaCholeskySolve<float>(
    magma_uplo_t uplo, magma_int_t n, magma_int_t nrhs, float* dA, magma_int_t ldda,
    float* dB, magma_int_t lddb, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_spotrs_gpu(uplo, n, nrhs, dA, ldda, dB, lddb, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaCholeskySolve<c10::complex<double>>(
    magma_uplo_t uplo, magma_int_t n, magma_int_t nrhs, c10::complex<double>* dA, magma_int_t ldda,
    c10::complex<double>* dB, magma_int_t lddb, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_zpotrs_gpu(uplo, n, nrhs,
    reinterpret_cast<magmaDoubleComplex*>(dA), ldda,
    reinterpret_cast<magmaDoubleComplex*>(dB), lddb, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaCholeskySolve<c10::complex<float>>(
    magma_uplo_t uplo, magma_int_t n, magma_int_t nrhs, c10::complex<float>* dA, magma_int_t ldda,
    c10::complex<float>* dB, magma_int_t lddb, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_cpotrs_gpu(uplo, n, nrhs,
    reinterpret_cast<magmaFloatComplex*>(dA), ldda,
    reinterpret_cast<magmaFloatComplex*>(dB), lddb, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaCholeskySolveBatched<double>(
    magma_uplo_t uplo, magma_int_t n, magma_int_t nrhs, double** dA_array, magma_int_t ldda,
    double** dB_array, magma_int_t lddb, magma_int_t& info, magma_int_t batchsize, const MAGMAQueue& magma_queue) {
  info = magma_dpotrs_batched(uplo, n, nrhs, dA_array, ldda, dB_array, lddb, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaCholeskySolveBatched<float>(
    magma_uplo_t uplo, magma_int_t n, magma_int_t nrhs, float** dA_array, magma_int_t ldda,
    float** dB_array, magma_int_t lddb, magma_int_t& info, magma_int_t batchsize, const MAGMAQueue& magma_queue) {
  info = magma_spotrs_batched(uplo, n, nrhs, dA_array, ldda, dB_array, lddb, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaCholeskySolveBatched<c10::complex<double>>(
    magma_uplo_t uplo, magma_int_t n, magma_int_t nrhs, c10::complex<double>** dA_array, magma_int_t ldda,
    c10::complex<double>** dB_array, magma_int_t lddb, magma_int_t& info, magma_int_t batchsize, const MAGMAQueue& magma_queue) {
  info = magma_zpotrs_batched(uplo, n, nrhs,
    reinterpret_cast<magmaDoubleComplex**>(dA_array), ldda,
    reinterpret_cast<magmaDoubleComplex**>(dB_array), lddb, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaCholeskySolveBatched<c10::complex<float>>(
    magma_uplo_t uplo, magma_int_t n, magma_int_t nrhs, c10::complex<float>** dA_array, magma_int_t ldda,
    c10::complex<float>** dB_array, magma_int_t lddb, magma_int_t& info, magma_int_t batchsize, const MAGMAQueue& magma_queue) {
  info = magma_cpotrs_batched(uplo, n, nrhs,
    reinterpret_cast<magmaFloatComplex**>(dA_array), ldda,
    reinterpret_cast<magmaFloatComplex**>(dB_array), lddb, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaCholesky<double>(
    magma_uplo_t uplo, magma_int_t n, double* dA,
    magma_int_t ldda, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_dpotrf_gpu(uplo, n, dA, ldda, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaCholesky<float>(
    magma_uplo_t uplo, magma_int_t n, float* dA,
    magma_int_t ldda, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_spotrf_gpu(uplo, n, dA, ldda, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaCholesky<c10::complex<double>>(
    magma_uplo_t uplo, magma_int_t n, c10::complex<double>* dA,
    magma_int_t ldda, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_zpotrf_gpu(uplo, n, reinterpret_cast<magmaDoubleComplex*>(dA), ldda, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaCholesky<c10::complex<float>>(
    magma_uplo_t uplo, magma_int_t n, c10::complex<float>* dA,
    magma_int_t ldda, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_cpotrf_gpu(uplo, n, reinterpret_cast<magmaFloatComplex*>(dA), ldda, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaCholeskyBatched<double>(
    magma_uplo_t uplo, magma_int_t n, double** dA_array, magma_int_t ldda,
    magma_int_t* info_array, magma_int_t batchsize, const MAGMAQueue& magma_queue) {
  magma_dpotrf_batched(uplo, n, dA_array, ldda, info_array, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaCholeskyBatched<float>(
    magma_uplo_t uplo, magma_int_t n, float** dA_array, magma_int_t ldda,
    magma_int_t* info_array, magma_int_t batchsize, const MAGMAQueue& magma_queue) {
  magma_spotrf_batched(uplo, n, dA_array, ldda, info_array, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaCholeskyBatched<c10::complex<double>>(
    magma_uplo_t uplo, magma_int_t n, c10::complex<double>** dA_array, magma_int_t ldda,
    magma_int_t* info_array, magma_int_t batchsize, const MAGMAQueue& magma_queue) {
  magma_zpotrf_batched(uplo, n, reinterpret_cast<magmaDoubleComplex**>(dA_array), ldda, info_array, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaCholeskyBatched<c10::complex<float>>(
    magma_uplo_t uplo, magma_int_t n, c10::complex<float>** dA_array, magma_int_t ldda,
    magma_int_t* info_array, magma_int_t batchsize, const MAGMAQueue& magma_queue) {
  magma_cpotrf_batched(uplo, n, reinterpret_cast<magmaFloatComplex**>(dA_array), ldda, info_array, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaTriangularSolveBatched<double>(
    magma_uplo_t uplo, magma_trans_t trans, magma_diag_t diag, magma_int_t m, magma_int_t n,
    double** dA_array, magma_int_t ldda, double** dB_array, magma_int_t lddb, magma_int_t batchsize,
    const MAGMAQueue& magma_queue) {
  magmablas_dtrsm_batched(MagmaLeft, uplo, trans, diag, m, n, 1, dA_array, ldda, dB_array, lddb, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaTriangularSolveBatched<float>(
    magma_uplo_t uplo, magma_trans_t trans, magma_diag_t diag, magma_int_t m, magma_int_t n,
    float** dA_array, magma_int_t ldda, float** dB_array, magma_int_t lddb, magma_int_t batchsize,
    const MAGMAQueue& magma_queue) {
  magmablas_strsm_batched(MagmaLeft, uplo, trans, diag, m, n, 1, dA_array, ldda, dB_array, lddb, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaTriangularSolveBatched<c10::complex<double>>(
    magma_uplo_t uplo, magma_trans_t trans, magma_diag_t diag, magma_int_t m, magma_int_t n,
    c10::complex<double>** dA_array, magma_int_t ldda, c10::complex<double>** dB_array, magma_int_t lddb, magma_int_t batchsize,
    const MAGMAQueue& magma_queue) {
  magmaDoubleComplex alpha({1, 0});
  magmablas_ztrsm_batched(MagmaLeft, uplo, trans, diag, m, n, alpha,
    reinterpret_cast<magmaDoubleComplex**>(dA_array), ldda,
    reinterpret_cast<magmaDoubleComplex**>(dB_array), lddb, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaTriangularSolveBatched<c10::complex<float>>(
    magma_uplo_t uplo, magma_trans_t trans, magma_diag_t diag, magma_int_t m, magma_int_t n,
    c10::complex<float>** dA_array, magma_int_t ldda, c10::complex<float>** dB_array, magma_int_t lddb, magma_int_t batchsize,
    const MAGMAQueue& magma_queue) {
  magmaFloatComplex alpha({1, 0});
  magmablas_ctrsm_batched(MagmaLeft, uplo, trans, diag, m, n, alpha,
    reinterpret_cast<magmaFloatComplex**>(dA_array), ldda,
    reinterpret_cast<magmaFloatComplex**>(dB_array), lddb, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
inline magma_int_t magmaGeqrfOptimalBlocksize<double>(magma_int_t m, magma_int_t n) {
  return magma_get_dgeqrf_nb(m, n);
}

template<>
inline magma_int_t magmaGeqrfOptimalBlocksize<float>(magma_int_t m, magma_int_t n) {
  return magma_get_sgeqrf_nb(m, n);
}

template <>
inline magma_int_t magmaGeqrfOptimalBlocksize<c10::complex<double>>(
    magma_int_t m,
    magma_int_t n) {
  return magma_get_zgeqrf_nb(m, n);
}

template <>
inline magma_int_t magmaGeqrfOptimalBlocksize<c10::complex<float>>(
    magma_int_t m,
    magma_int_t n) {
  return magma_get_cgeqrf_nb(m, n);
}

template<>
void magmaGeqrf<double>(
    magma_int_t m, magma_int_t n, double* dA, magma_int_t ldda,
    double* tau, double* dT, magma_int_t* info, bool is_v2) {
  MagmaStreamSyncGuard guard;
  if (!is_v2) {
    magma_dgeqrf_gpu(m, n, dA, ldda, tau, dT, info);
  } else {
    magma_dgeqrf2_gpu(m, n, dA, ldda, tau, info);
  }
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaGeqrf<float>(
    magma_int_t m, magma_int_t n, float* dA, magma_int_t ldda,
    float* tau, float* dT, magma_int_t* info, bool is_v2) {
  MagmaStreamSyncGuard guard;
  if (!is_v2) {
    magma_sgeqrf_gpu(m, n, dA, ldda, tau, dT, info);
  } else {
    magma_sgeqrf2_gpu(m, n, dA, ldda, tau, info);
  }
  AT_CUDA_CHECK(cudaGetLastError());
}

template <>
void magmaGeqrf<c10::complex<double>>(
    magma_int_t m,
    magma_int_t n,
    c10::complex<double>* dA,
    magma_int_t ldda,
    c10::complex<double>* tau,
    c10::complex<double>* dT,
    magma_int_t* info,
    bool is_v2) {
  MagmaStreamSyncGuard guard;
  if (!is_v2) {
    magma_zgeqrf_gpu(
        m,
        n,
        reinterpret_cast<magmaDoubleComplex*>(dA),
        ldda,
        reinterpret_cast<magmaDoubleComplex*>(tau),
        reinterpret_cast<magmaDoubleComplex*>(dT),
        info);
  } else {
    magma_zgeqrf2_gpu(
        m,
        n,
        reinterpret_cast<magmaDoubleComplex*>(dA),
        ldda,
        reinterpret_cast<magmaDoubleComplex*>(tau),
        info);
  }
  AT_CUDA_CHECK(cudaGetLastError());
}

template <>
void magmaGeqrf<c10::complex<float>>(
    magma_int_t m,
    magma_int_t n,
    c10::complex<float>* dA,
    magma_int_t ldda,
    c10::complex<float>* tau,
    c10::complex<float>* dT,
    magma_int_t* info,
    bool is_v2) {
  MagmaStreamSyncGuard guard;
  if (!is_v2) {
    magma_cgeqrf_gpu(
        m,
        n,
        reinterpret_cast<magmaFloatComplex*>(dA),
        ldda,
        reinterpret_cast<magmaFloatComplex*>(tau),
        reinterpret_cast<magmaFloatComplex*>(dT),
        info);
  } else {
    magma_cgeqrf2_gpu(
        m,
        n,
        reinterpret_cast<magmaFloatComplex*>(dA),
        ldda,
        reinterpret_cast<magmaFloatComplex*>(tau),
        info);
  }
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaOrgqr<double>(
    magma_int_t m, magma_int_t n, magma_int_t k, double* dA, magma_int_t ldda,
    double* tau, double* dT, magma_int_t nb, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_dorgqr_gpu(m, n, k, dA, ldda, tau, dT, nb, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaOrgqr<float>(
    magma_int_t m, magma_int_t n, magma_int_t k, float* dA, magma_int_t ldda,
    float* tau, float* dT, magma_int_t nb, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_sorgqr_gpu(m, n, k, dA, ldda, tau, dT, nb, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template <>
void magmaOrgqr<c10::complex<double>>(
    magma_int_t m,
    magma_int_t n,
    magma_int_t k,
    c10::complex<double>* dA,
    magma_int_t ldda,
    c10::complex<double>* tau,
    c10::complex<double>* dT,
    magma_int_t nb,
    magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_zungqr_gpu(
      m,
      n,
      k,
      reinterpret_cast<magmaDoubleComplex*>(dA),
      ldda,
      reinterpret_cast<magmaDoubleComplex*>(tau),
      reinterpret_cast<magmaDoubleComplex*>(dT),
      nb,
      info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template <>
void magmaOrgqr<c10::complex<float>>(
    magma_int_t m,
    magma_int_t n,
    magma_int_t k,
    c10::complex<float>* dA,
    magma_int_t ldda,
    c10::complex<float>* tau,
    c10::complex<float>* dT,
    magma_int_t nb,
    magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_cungqr_gpu(
    m,
    n,
    k,
    reinterpret_cast<magmaFloatComplex*>(dA),
    ldda,
    reinterpret_cast<magmaFloatComplex*>(tau),
    reinterpret_cast<magmaFloatComplex*>(dT),
    nb,
    info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaSyevd<double>(
    magma_vec_t jobz, magma_uplo_t uplo, magma_int_t n, double* dA, magma_int_t ldda,
    double* w, double* wA, magma_int_t ldwa, double* work, magma_int_t lwork, double* rwork,
    magma_int_t lrwork, magma_int_t* iwork, magma_int_t liwork, magma_int_t* info) {
  (void)rwork;  // unused
  (void)lrwork;  // unused
  MagmaStreamSyncGuard guard;
  magma_dsyevd_gpu(jobz, uplo, n, dA, ldda, w, wA, ldwa, work, lwork, iwork, liwork, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaSyevd<float>(
    magma_vec_t jobz, magma_uplo_t uplo, magma_int_t n, float* dA, magma_int_t ldda,
    float* w, float* wA, magma_int_t ldwa, float* work, magma_int_t lwork, float* rwork,
    magma_int_t lrwork, magma_int_t* iwork, magma_int_t liwork, magma_int_t* info) {
  (void)rwork;  // unused
  (void)lrwork;  // unused
  MagmaStreamSyncGuard guard;
  magma_ssyevd_gpu(jobz, uplo, n, dA, ldda, w, wA, ldwa, work, lwork, iwork, liwork, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaSyevd<c10::complex<double>, double>(
    magma_vec_t jobz, magma_uplo_t uplo, magma_int_t n, c10::complex<double>* dA, magma_int_t ldda,
    double* w, c10::complex<double>* wA, magma_int_t ldwa, c10::complex<double>* work, magma_int_t lwork, double* rwork,
    magma_int_t lrwork, magma_int_t* iwork, magma_int_t liwork, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_zheevd_gpu(
      jobz, uplo, n, reinterpret_cast<magmaDoubleComplex*>(dA), ldda, w, reinterpret_cast<magmaDoubleComplex*>(wA),
      ldwa, reinterpret_cast<magmaDoubleComplex*>(work), lwork, rwork, lrwork, iwork, liwork, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaSyevd<c10::complex<float>, float>(
    magma_vec_t jobz, magma_uplo_t uplo, magma_int_t n, c10::complex<float>* dA, magma_int_t ldda,
    float* w, c10::complex<float>* wA, magma_int_t ldwa, c10::complex<float>* work, magma_int_t lwork, float* rwork,
    magma_int_t lrwork, magma_int_t* iwork, magma_int_t liwork, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_cheevd_gpu(
      jobz, uplo, n, reinterpret_cast<magmaFloatComplex*>(dA), ldda, w, reinterpret_cast<magmaFloatComplex*>(wA),
      ldwa, reinterpret_cast<magmaFloatComplex*>(work), lwork, rwork, lrwork, iwork, liwork, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaEig<double>(
    magma_vec_t jobvl, magma_vec_t jobvr, magma_int_t n,
    double *A, magma_int_t lda,
    double *w,
    double *VL, magma_int_t ldvl,
    double *VR, magma_int_t ldvr,
    double *work, magma_int_t lwork,
    double *rwork,
    magma_int_t *info) {
  MagmaStreamSyncGuard guard;
  // magma [sd]geev wants to separate output arrays: wr and wi for the real
  // and imaginary parts
  double *wr = w;
  double *wi = w + n;
  (void)rwork; // unused
  magma_dgeev(jobvl, jobvr, n, A, lda, wr, wi, VL, ldvl, VR, ldvr, work, lwork, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaEig<float>(
    magma_vec_t jobvl, magma_vec_t jobvr, magma_int_t n,
    float *A, magma_int_t lda,
    float *w,
    float *VL, magma_int_t ldvl,
    float *VR, magma_int_t ldvr,
    float *work, magma_int_t lwork,
    float *rwork,
    magma_int_t *info) {
  MagmaStreamSyncGuard guard;
  float *wr = w;
  float *wi = w + n;
  (void)rwork; // unused
  magma_sgeev(jobvl, jobvr, n, A, lda, wr, wi, VL, ldvl, VR, ldvr, work, lwork, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaEig<c10::complex<double>, double>(
    magma_vec_t jobvl, magma_vec_t jobvr, magma_int_t n,
    c10::complex<double> *A, magma_int_t lda,
    c10::complex<double> *w,
    c10::complex<double> *VL, magma_int_t ldvl,
    c10::complex<double> *VR, magma_int_t ldvr,
    c10::complex<double> *work, magma_int_t lwork,
    double *rwork,
    magma_int_t *info) {
  MagmaStreamSyncGuard guard;
  magma_zgeev(jobvl, jobvr, n,
         reinterpret_cast<magmaDoubleComplex*>(A), lda,
         reinterpret_cast<magmaDoubleComplex*>(w),
         reinterpret_cast<magmaDoubleComplex*>(VL), ldvl,
         reinterpret_cast<magmaDoubleComplex*>(VR), ldvr,
         reinterpret_cast<magmaDoubleComplex*>(work), lwork,
         rwork, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaEig<c10::complex<float>, float>(
    magma_vec_t jobvl, magma_vec_t jobvr, magma_int_t n,
    c10::complex<float> *A, magma_int_t lda,
    c10::complex<float> *w,
    c10::complex<float> *VL, magma_int_t ldvl,
    c10::complex<float> *VR, magma_int_t ldvr,
    c10::complex<float> *work, magma_int_t lwork,
    float *rwork,
    magma_int_t *info) {
  MagmaStreamSyncGuard guard;
  magma_cgeev(jobvl, jobvr, n,
         reinterpret_cast<magmaFloatComplex*>(A), lda,
         reinterpret_cast<magmaFloatComplex*>(w),
         reinterpret_cast<magmaFloatComplex*>(VL), ldvl,
         reinterpret_cast<magmaFloatComplex*>(VR), ldvr,
         reinterpret_cast<magmaFloatComplex*>(work), lwork,
         rwork, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaSvd<double>(
    magma_vec_t jobz, magma_int_t m, magma_int_t n, double* A,
    magma_int_t lda, double* s, double* U, magma_int_t ldu,
    double* VT, magma_int_t ldvt, double* work, magma_int_t lwork,
    double *rwork, magma_int_t* iwork, magma_int_t* info) {
  (void)rwork; // unused
  MagmaStreamSyncGuard guard;
  magma_dgesdd(jobz, m, n, A, lda, s, U, ldu, VT, ldvt, work, lwork, iwork, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaSvd<float>(
    magma_vec_t jobz, magma_int_t m, magma_int_t n, float* A,
    magma_int_t lda, float* s, float* U, magma_int_t ldu,
    float* VT, magma_int_t ldvt, float* work, magma_int_t lwork,
    float* rwork, magma_int_t* iwork, magma_int_t* info) {
  (void)rwork; // unused
  MagmaStreamSyncGuard guard;
  magma_sgesdd(jobz, m, n, A, lda, s, U, ldu, VT, ldvt, work, lwork, iwork, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaSvd<c10::complex<float>, float>(
    magma_vec_t jobz, magma_int_t m, magma_int_t n, c10::complex<float>* A,
    magma_int_t lda, float* s, c10::complex<float>* U, magma_int_t ldu,
    c10::complex<float>* VT, magma_int_t ldvt, c10::complex<float>* work, magma_int_t lwork,
    float *rwork, magma_int_t* iwork, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_cgesdd(jobz, m, n, reinterpret_cast<magmaFloatComplex*>(A), lda, s,
                reinterpret_cast<magmaFloatComplex*>(U), ldu,
                reinterpret_cast<magmaFloatComplex*>(VT), ldvt,
                reinterpret_cast<magmaFloatComplex*>(work), lwork,
                rwork, iwork, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaSvd<c10::complex<double>, double>(
    magma_vec_t jobz, magma_int_t m, magma_int_t n, c10::complex<double>* A,
    magma_int_t lda, double* s, c10::complex<double>* U, magma_int_t ldu,
    c10::complex<double>* VT, magma_int_t ldvt, c10::complex<double>* work, magma_int_t lwork,
    double *rwork, magma_int_t* iwork, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_zgesdd(jobz, m, n, reinterpret_cast<magmaDoubleComplex*>(A), lda, s,
                reinterpret_cast<magmaDoubleComplex*>(U), ldu,
                reinterpret_cast<magmaDoubleComplex*>(VT), ldvt,
                reinterpret_cast<magmaDoubleComplex*>(work), lwork,
                rwork, iwork, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuSolve<double>(
    magma_int_t n, magma_int_t nrhs, double* dA, magma_int_t ldda, magma_int_t* ipiv,
    double* dB, magma_int_t lddb, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_dgetrs_gpu(MagmaNoTrans, n, nrhs, dA, ldda, ipiv, dB, lddb, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuSolve<float>(
    magma_int_t n, magma_int_t nrhs, float* dA, magma_int_t ldda, magma_int_t* ipiv,
    float* dB, magma_int_t lddb, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_sgetrs_gpu(MagmaNoTrans, n, nrhs, dA, ldda, ipiv, dB, lddb, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuSolve<c10::complex<double>>(
    magma_int_t n, magma_int_t nrhs, c10::complex<double>* dA, magma_int_t ldda, magma_int_t* ipiv,
    c10::complex<double>* dB, magma_int_t lddb, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_zgetrs_gpu(MagmaNoTrans, n, nrhs, reinterpret_cast<magmaDoubleComplex*>(dA), ldda, ipiv, reinterpret_cast<magmaDoubleComplex*>(dB), lddb, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuSolve<c10::complex<float>>(
    magma_int_t n, magma_int_t nrhs, c10::complex<float>* dA, magma_int_t ldda, magma_int_t* ipiv,
    c10::complex<float>* dB, magma_int_t lddb, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_cgetrs_gpu(MagmaNoTrans, n, nrhs, reinterpret_cast<magmaFloatComplex*>(dA), ldda, ipiv, reinterpret_cast<magmaFloatComplex*>(dB), lddb, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuSolveBatched<double>(
    magma_int_t n, magma_int_t nrhs, double** dA_array, magma_int_t ldda, magma_int_t** dipiv_array,
    double** dB_array, magma_int_t lddb, magma_int_t& info,
    magma_int_t batchsize, const MAGMAQueue& magma_queue) {
  info = magma_dgetrs_batched(MagmaNoTrans, n, nrhs, dA_array, ldda, dipiv_array, dB_array, lddb, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuSolveBatched<float>(
    magma_int_t n, magma_int_t nrhs, float** dA_array, magma_int_t ldda, magma_int_t** dipiv_array,
    float** dB_array, magma_int_t lddb, magma_int_t& info,
    magma_int_t batchsize, const MAGMAQueue& magma_queue) {
 info = magma_sgetrs_batched(MagmaNoTrans, n, nrhs, dA_array, ldda, dipiv_array, dB_array, lddb, batchsize, magma_queue.get_queue());
 AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuSolveBatched<c10::complex<double>>(
    magma_int_t n, magma_int_t nrhs, c10::complex<double>** dA_array, magma_int_t ldda, magma_int_t** dipiv_array,
    c10::complex<double>** dB_array, magma_int_t lddb, magma_int_t& info,
    magma_int_t batchsize, const MAGMAQueue& magma_queue) {
  info = magma_zgetrs_batched(MagmaNoTrans, n, nrhs, reinterpret_cast<magmaDoubleComplex**>(dA_array), ldda, dipiv_array, reinterpret_cast<magmaDoubleComplex**>(dB_array), lddb, batchsize, magma_queue.get_queue());
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaLuSolveBatched<c10::complex<float>>(
    magma_int_t n, magma_int_t nrhs, c10::complex<float>** dA_array, magma_int_t ldda, magma_int_t** dipiv_array,
    c10::complex<float>** dB_array, magma_int_t lddb, magma_int_t& info,
    magma_int_t batchsize, const MAGMAQueue& magma_queue) {
 info = magma_cgetrs_batched(MagmaNoTrans, n, nrhs, reinterpret_cast<magmaFloatComplex**>(dA_array), ldda, dipiv_array, reinterpret_cast<magmaFloatComplex**>(dB_array), lddb, batchsize, magma_queue.get_queue());
 AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaGels<float>(
    magma_trans_t trans, magma_int_t m, magma_int_t n, magma_int_t nrhs,
    float* dA, magma_int_t ldda, float* dB, magma_int_t lddb,
    float* hwork, magma_int_t lwork, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_sgels_gpu(trans, m, n, nrhs,
      dA, ldda, dB, lddb,
      hwork, lwork, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaGels<double>(
    magma_trans_t trans, magma_int_t m, magma_int_t n, magma_int_t nrhs,
    double* dA, magma_int_t ldda, double* dB, magma_int_t lddb,
    double* hwork, magma_int_t lwork, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_dgels_gpu(trans, m, n, nrhs,
      dA, ldda, dB, lddb,
      hwork, lwork, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaGels<c10::complex<float>>(
    magma_trans_t trans, magma_int_t m, magma_int_t n, magma_int_t nrhs,
    c10::complex<float>* dA, magma_int_t ldda, c10::complex<float>* dB, magma_int_t lddb,
    c10::complex<float>* hwork, magma_int_t lwork, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_cgels_gpu(trans, m, n, nrhs,
      reinterpret_cast<magmaFloatComplex*>(dA), ldda,
      reinterpret_cast<magmaFloatComplex*>(dB), lddb,
      reinterpret_cast<magmaFloatComplex*>(hwork), lwork, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

template<>
void magmaGels<c10::complex<double>>(
    magma_trans_t trans, magma_int_t m, magma_int_t n, magma_int_t nrhs,
    c10::complex<double>* dA, magma_int_t ldda, c10::complex<double>* dB, magma_int_t lddb,
    c10::complex<double>* hwork, magma_int_t lwork, magma_int_t* info) {
  MagmaStreamSyncGuard guard;
  magma_zgels_gpu(trans, m, n, nrhs,
      reinterpret_cast<magmaDoubleComplex*>(dA), ldda,
      reinterpret_cast<magmaDoubleComplex*>(dB), lddb,
      reinterpret_cast<magmaDoubleComplex*>(hwork), lwork, info);
  AT_CUDA_CHECK(cudaGetLastError());
}

namespace {

/*
  MAGMA can return errors both as a return value and in the info argument.
  The return value and info should always be identical.
  In general, the meaning is as given in this table.
  Predefined error codes are large negative numbers. Using the symbolic
  constants below is preferred, but the numeric values can be found in
  include/magma_types.h.

  Info                       |  Description
  -----------                |  -----------
  info = 0 (MAGMA_SUCCESS)   |  Successful exit
  info < 0, but small        |  For info = -i, the i-th argument had an illegal value
  info > 0                   |  Function-specific error such as singular matrix
  MAGMA_ERR_DEVICE_ALLOC     |  Could not allocate GPU device memory
  MAGMA_ERR_HOST_ALLOC       |  Could not allocate CPU host memory
  MAGMA_ERR_ILLEGAL_VALUE    |  An argument had an illegal value (deprecated; instead it should return -i to say the i-th argument was bad)
  MAGMA_ERR_INVALID_PTR      |  Can't free pointer
  MAGMA_ERR_NOT_IMPLEMENTED  |  Function or option not implemented
  MAGMA_ERR_NOT_SUPPORTED    |  Function or option not supported on the current architecture
*/
void checkMagmaInternalError(magma_int_t info, const std::string& magma_function_name) {
  // if info > 0 the error is function-specific, do nothing in this case
  TORCH_CHECK(info >= 0,
      "MAGMA error: ",
      magma_strerror(info),
      ", info = ", info,
      ", when calling ", magma_function_name);
}

} // anonymous namespace

#endif // USE_MAGMA

#define ALLOCATE_ARRAY(name, type, size) \
  auto storage_##name = pin_memory<type>(size); \
  name = static_cast<type*>(storage_##name.data());

// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ solve ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

template <typename scalar_t>
static void apply_solve(Tensor& b, Tensor& A, Tensor& infos_out) {
#ifndef USE_MAGMA
AT_ERROR("solve: MAGMA library not found in "
    "compilation. Please rebuild with MAGMA.");
#else
  auto A_data = A.data_ptr<scalar_t>();
  auto b_data = b.data_ptr<scalar_t>();
  magma_int_t n = magma_int_cast(A.size(-2), "A.size(-2)");
  magma_int_t nrhs = magma_int_cast(b.size(-1), "b.size(-1)");
  magma_int_t lda = std::max(magma_int_t{1}, n);

  if (b.dim() == 2) {
    auto ipiv = at::empty({n}, at::kInt);
    // magmaSolve requires infos tensor to live on CPU
    Tensor infos = at::empty(infos_out.sizes(), infos_out.options().device(kCPU));
    magmaSolve<scalar_t>(n, nrhs, A_data, lda, ipiv.data_ptr<magma_int_t>(),
                        b_data, lda, infos.data_ptr<magma_int_t>());
    infos_out.copy_(infos);
  } else {
    auto infos_data = infos_out.data_ptr<magma_int_t>();
    auto A_mat_stride = matrixStride(A);
    auto b_mat_stride = matrixStride(b);
    magma_int_t batch_size = magma_int_cast(batchCount(A), "batchCount");

    magma_int_t* ipiv_data;
    magma_int_t** ipiv_array;
    scalar_t** A_array;
    scalar_t** b_array;

    ALLOCATE_ARRAY(ipiv_data, magma_int_t, batch_size * n);
    ALLOCATE_ARRAY(ipiv_array, magma_int_t*, batch_size);
    ALLOCATE_ARRAY(A_array, scalar_t*, batch_size);
    ALLOCATE_ARRAY(b_array, scalar_t*, batch_size);

    // Set up the created arrays
    for (int64_t i = 0; i < batch_size; i++) {
      A_array[i] = &A_data[i * A_mat_stride];
      b_array[i] = &b_data[i * b_mat_stride];
      ipiv_array[i] = &ipiv_data[i * n];
    }

    MAGMAQueue magma_queue(b.get_device());

    constexpr int64_t batch_limit = 65535;
    // Compute as many batches of 65535 possible
    // The number of "mini"-batches are floor(batch_size / batch_limit)
    // and these cover floor(batch_size / batch_limit) * batch_limit matrix solves
    int64_t mini_batches = batch_size / batch_limit, mini_idx;
    for (mini_idx = 0; mini_idx < mini_batches * batch_limit; mini_idx += batch_limit) {
      scalar_t** A_array_cur = &A_array[mini_idx];
      scalar_t** b_array_cur = &b_array[mini_idx];
      magma_int_t** ipiv_array_cur = &ipiv_array[mini_idx];
      magma_int_t* info_array_cur = &infos_data[mini_idx];

      magmaSolveBatched<scalar_t>(
          n, nrhs, A_array_cur, lda, ipiv_array_cur, b_array_cur, lda,
          info_array_cur, batch_limit, magma_queue);
    }

    // Compute whatever is left = batch_size - floor(batch_size / batch_limit) * batch_limit
    // which concisely is equal to batch_size % batch_limit
    if (batch_size % batch_limit != 0) {
      magmaSolveBatched<scalar_t>(
          n, nrhs, &A_array[mini_idx], lda, &ipiv_array[mini_idx], &b_array[mini_idx], lda,
          &infos_data[mini_idx], batch_size % batch_limit, magma_queue);
    }
  }
#endif
}

std::tuple<Tensor, Tensor> _solve_helper_cuda(const Tensor& self, const Tensor& A) {
  auto self_working_copy = cloneBatchedColumnMajor(self);
  auto A_working_copy = cloneBatchedColumnMajor(A);
  // infos might not get filled for empty inputs therefore at::zeros is used instead of at::empty
  auto infos = at::zeros({std::max<int64_t>(1, batchCount(self))}, self.options().dtype(kInt));
  AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(self.scalar_type(), "solve_cuda", [&]{
    apply_solve<scalar_t>(self_working_copy, A_working_copy, infos);
  });
  if (self.dim() > 2) {
    batchCheckErrors(infos, "solve_cuda");
  } else {
    singleCheckErrors(infos.item().toInt(), "solve_cuda");
  }
  return std::tuple<Tensor, Tensor>(self_working_copy, A_working_copy);
}

// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ inverse ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

/*
Computes the inverse of n-by-n matrix 'self', it is saved to 'self_inv'.
'infos' is an int Tensor containing error codes for each matrix in the batched input.
'infos_lu' is for holding magmaLU errors, and 'infos_getri' is for holding magmaGetri errors
For more information see MAGMA's documentation for GETRI and GETRF routines.
*/
template <typename scalar_t>
static void apply_batched_inverse(Tensor& self, Tensor& self_inv, Tensor& infos_lu, Tensor& infos_getri) {
#ifndef USE_MAGMA
AT_ERROR("inverse: MAGMA library not found in "
    "compilation. Please rebuild with MAGMA.");
#else
  auto self_data = self.data_ptr<scalar_t>();
  auto self_mat_stride = matrixStride(self);
  auto self_inv_data = self_inv.data_ptr<scalar_t>();
  auto self_inv_mat_stride = matrixStride(self_inv);

  auto infos_lu_data = infos_lu.data_ptr<magma_int_t>();
  auto infos_getri_data = infos_getri.data_ptr<magma_int_t>();

  magma_int_t batch_size = magma_int_cast(batchCount(self), "batchCount");
  // MAGMA does not work with batch_size == 0, let's return early in this case
  if (batch_size == 0) {
    return;
  }

  magma_int_t n = magma_int_cast(self.size(-2), "self.size(-2)");
  magma_int_t lda = std::max<magma_int_t>(1, n);

  magma_int_t* ipiv_data;
  magma_int_t** ipiv_array;
  scalar_t** self_array;
  scalar_t** self_inv_array;

  ALLOCATE_ARRAY(ipiv_data, magma_int_t, batch_size * lda);
  ALLOCATE_ARRAY(ipiv_array, magma_int_t*, batch_size);
  ALLOCATE_ARRAY(self_array, scalar_t*, batch_size);
  ALLOCATE_ARRAY(self_inv_array, scalar_t*, batch_size);

  // Set up the created arrays
  for (int64_t i = 0; i < batch_size; i++) {
    self_array[i] = &self_data[i * self_mat_stride];
    self_inv_array[i] = &self_inv_data[i * self_inv_mat_stride];
    ipiv_array[i] = &ipiv_data[i * n];
  }
  // magmaLuBatched leaves ipiv_data values unwritten for singular matrices.
  // Initialize to avoid memory access violations inside magma kernels (gh-51930).
  std::fill_n(ipiv_data, batch_size * n, 1);

  MAGMAQueue magma_queue(self.get_device());
  magmaLuBatched<scalar_t>(
    n, n, self_array, lda, ipiv_array, infos_lu_data,
    batch_size, magma_queue);

  constexpr int64_t batch_limit = 65535;
  // Compute as many batches of 65535 possible
  // The number of "mini"-batches are floor(batch_size / batch_limit)
  // and these cover floor(batch_size / batch_limit) * batch_limit matrix solves
  int64_t mini_batches = batch_size / batch_limit, mini_idx;
  for (mini_idx = 0; mini_idx < mini_batches * batch_limit; mini_idx += batch_limit) {
    scalar_t** self_array_cur = &self_array[mini_idx];
    scalar_t** self_inv_array_cur = &self_inv_array[mini_idx];
    magma_int_t** ipiv_array_cur = &ipiv_array[mini_idx];
    magma_int_t* info_array_cur_getri = &infos_getri_data[mini_idx];

    magmaGetriBatched<scalar_t>(
      n, self_array_cur, lda, ipiv_array_cur, self_inv_array_cur,
      lda, info_array_cur_getri, batch_limit, magma_queue);
  }

  // Compute whatever is left = batch_size - floor(batch_size / batch_limit) * batch_limit
  // which concisely is equal to batch_size % batch_limit
  if (batch_size % batch_limit != 0) {
    magmaGetriBatched<scalar_t>(
      n, &self_array[mini_idx], lda, &ipiv_array[mini_idx], &self_inv_array[mini_idx],
      lda, &infos_getri_data[mini_idx], batch_size % batch_limit, magma_queue);
  }
#endif
}

template <typename scalar_t>
static void apply_single_inverse(Tensor& self, Tensor& info_lu, Tensor& info_getri) {
#ifndef USE_MAGMA
AT_ERROR("inverse: MAGMA library not found in "
    "compilation. Please rebuild with MAGMA.");
#else
  auto self_data = self.data_ptr<scalar_t>();
  magma_int_t n = magma_int_cast(self.size(-2), "self.size(-2)");
  magma_int_t lda = std::max<magma_int_t>(1, n);
  magma_int_t lwork = n * magmaGetriOptimalBlocksize<scalar_t>(n);

  // magmaLu and magmaGetri requires info argument to live on CPU
  // but info_lu and info_getri tensors are on the same device as self
  magma_int_t info_lu_cpu = 0;
  magma_int_t info_getri_cpu = 0;

  Tensor ipiv = at::empty({lda}, at::kInt);
  Tensor dwork = at::empty({lwork}, self.options());
  magmaLu<scalar_t>(n, n, self_data, lda, ipiv.data_ptr<magma_int_t>(), &info_lu_cpu);
  magmaGetri<scalar_t>(
    n, self_data, lda, ipiv.data_ptr<magma_int_t>(), dwork.data_ptr<scalar_t>(), lwork, &info_getri_cpu);
  info_lu.fill_(info_lu_cpu);
  info_getri.fill_(info_getri_cpu);
#endif
}

Tensor _inverse_helper_cuda_legacy(const Tensor& self) {
  auto self_inv_working_copy = cloneBatchedColumnMajor(self);
  if (self.dim() > 2) {
    auto infos_lu = at::zeros({std::max<int64_t>(1, batchCount(self))}, self.options().dtype(kInt));
    auto infos_getri = at::zeros({std::max<int64_t>(1, batchCount(self))}, self.options().dtype(kInt));
    auto self_working_copy = cloneBatchedColumnMajor(self);
    AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(self.scalar_type(), "inverse_cuda", [&]{
      apply_batched_inverse<scalar_t>(
        self_working_copy, self_inv_working_copy, infos_lu, infos_getri);
    });
    batchCheckErrors(infos_lu, "inverse_cuda");
    batchCheckErrors(infos_getri, "inverse_cuda");
  } else {
    // magmaLu and magmaGetri requires infos tensor to live on CPU
    auto infos_lu = at::zeros({1}, self.options().dtype(kInt).device(kCPU));
    auto infos_getri = at::zeros({1}, self.options().dtype(kInt).device(kCPU));
    AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(self.scalar_type(), "inverse_cuda", [&]{
      apply_single_inverse<scalar_t>(self_inv_working_copy, infos_lu, infos_getri);
    });
    singleCheckErrors(infos_lu.item().toInt(), "inverse_cuda");
    singleCheckErrors(infos_getri.item().toInt(), "inverse_cuda");
  }
  return self_inv_working_copy;
}

Tensor _inverse_helper_cuda(const Tensor& self) {
#ifdef USE_CUSOLVER
  if ((self.dim() == 2) || (/* self.dim() > 2 && */ batchCount(self) <= 2) || !use_magma_) {
    return _inverse_helper_cuda_lib(self);    // cusolver or cublas
  } else {
    return _inverse_helper_cuda_legacy(self); // magma-cuda
  }
#else
  return _inverse_helper_cuda_legacy(self); // magma-cuda
#endif
}

// This is a type dispatching helper function for 'apply_batched_inverse' and 'singleCheckErrors'
Tensor& _linalg_inv_out_helper_cuda_legacy(Tensor& result, Tensor& infos_lu, Tensor& infos_getri) {
  // assuming result is in column major order and contains the matrices to invert
  if (result.dim() > 2) {
    auto input_working_copy = cloneBatchedColumnMajor(result);
    AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(result.scalar_type(), "linalg_inv_out_cuda", [&]{
      apply_batched_inverse<scalar_t>(
        input_working_copy, result, infos_lu, infos_getri);
    });
  } else {
    AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(result.scalar_type(), "linalg_inv_out_cuda", [&]{
      apply_single_inverse<scalar_t>(result, infos_lu, infos_getri);
    });
  }
  return result;
}

// This is a MAGMA/cuSOLVER dispatching helper function
Tensor& _linalg_inv_out_helper_cuda(Tensor &result, Tensor& infos_lu, Tensor& infos_getri) {
  // This function calculates the inverse matrix in-place
  // result should be in column major order and contain matrices to invert
#ifdef USE_CUSOLVER
  if ((result.dim() == 2) || (/* result.dim() > 2 && */ batchCount(result) <= 2) || !use_magma_) {
    return _linalg_inv_out_helper_cuda_lib(result, infos_lu, infos_getri);  // cusolver or cublas
  } else {
    return _linalg_inv_out_helper_cuda_legacy(result, infos_lu, infos_getri);  // magma-cuda
  }
#else
  return _linalg_inv_out_helper_cuda_legacy(result, infos_lu, infos_getri);  // magma-cuda
#endif
  return result;
}

// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ cholesky_solve ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

template <typename scalar_t>
static void apply_cholesky_solve(Tensor& b, Tensor& A, bool upper, int64_t& info) {
#ifndef USE_MAGMA
AT_ERROR("cholesky_solve: MAGMA library not found in "
    "compilation. Please rebuild with MAGMA.");
#else
  magma_uplo_t uplo = upper ? MagmaUpper : MagmaLower;

  auto A_data = A.data_ptr<scalar_t>();
  auto b_data = b.data_ptr<scalar_t>();
  magma_int_t n = magma_int_cast(A.size(-2), "A.size(-2)");
  magma_int_t lda = std::max<magma_int_t>(1, n);
  magma_int_t nrhs = magma_int_cast(b.size(-1), "b.size(-1)");

  int info_tmp = 0;
  if (b.dim() == 2) {
    magmaCholeskySolve<scalar_t>(uplo, n, nrhs, A_data, lda,
                                 b_data, lda, &info_tmp);
    info = info_tmp;
  } else {
    auto A_mat_stride = matrixStride(A);
    auto b_mat_stride = matrixStride(b);
    magma_int_t batch_size = magma_int_cast(batchCount(A), "batchCount");

    scalar_t** A_array;
    scalar_t** b_array;

    ALLOCATE_ARRAY(A_array, scalar_t*, batch_size);
    ALLOCATE_ARRAY(b_array, scalar_t*, batch_size);

    // Set up the created arrays
    for (int64_t i = 0; i < batch_size; i++) {
      A_array[i] = &A_data[i * A_mat_stride];
      b_array[i] = &b_data[i * b_mat_stride];
    }

    MAGMAQueue magma_queue(b.get_device());

    constexpr int64_t batch_limit = 65535;
    // Compute as many batches of 65535 possible
    // The number of "mini"-batches are floor(batch_size / batch_limit)
    // and these cover floor(batch_size / batch_limit) * batch_limit matrix solves
    int64_t mini_batches = batch_size / batch_limit, mini_idx;
    for (mini_idx = 0; mini_idx < mini_batches * batch_limit; mini_idx += batch_limit) {
      scalar_t** A_array_cur = &A_array[mini_idx];
      scalar_t** b_array_cur = &b_array[mini_idx];

      magmaCholeskySolveBatched<scalar_t>(
          uplo, n, nrhs, A_array_cur, lda, b_array_cur, lda,
          info_tmp, batch_limit, magma_queue);

      if (info_tmp != 0) {
        break;
      }
    }

    // Compute whatever is left = batch_size - floor(batch_size / batch_limit) * batch_limit
    // which concisely is equal to batch_size % batch_limit
    if (batch_size % batch_limit != 0 && info_tmp == 0) {
      magmaCholeskySolveBatched<scalar_t>(
          uplo, n, nrhs, &A_array[mini_idx], lda, &b_array[mini_idx], lda,
          info_tmp, batch_size % batch_limit, magma_queue);
    }

    info = info_tmp;
  }
#endif
}

Tensor _cholesky_solve_helper_cuda_magma(const Tensor& self, const Tensor& A, bool upper) {
  int64_t info = 0;
  auto self_working_copy = cloneBatchedColumnMajor(self);
  auto A_working_copy = cloneBatchedColumnMajor(A);
  AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(self.scalar_type(), "cholesky_solve_cuda", [&]{
    apply_cholesky_solve<scalar_t>(self_working_copy, A_working_copy, upper, info);
  });
  TORCH_CHECK(info == 0, "MAGMA cholesky_solve : invalid argument: ", -info);
  return self_working_copy;
}

// Todo: cusolverDn<T>potrsBatched only supports nrhs == 1 and does not have good performance.
//     Batched cholesky_solve is dispatched to magma.
Tensor _cholesky_solve_helper_cuda(const Tensor& self, const Tensor& A, bool upper) {
#ifdef USE_CUSOLVER
  if (batchCount(self) == 1 || !use_magma_) {
    return _cholesky_solve_helper_cuda_cusolver(self, A, upper);
  } else {
    return _cholesky_solve_helper_cuda_magma(self, A, upper);
  }
#else
  return _cholesky_solve_helper_cuda_magma(self, A, upper);
#endif
}

// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ cholesky ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

template <typename scalar_t>
static void apply_cholesky(const Tensor& self, bool upper, const Tensor& info) {
#ifndef USE_MAGMA
  TORCH_CHECK(
      false,
      "Calling torch.linalg.cholesky on a CUDA tensor requires compiling ",
      "PyTorch with MAGMA. Please use PyTorch built with MAGMA support.");
#else
  magma_uplo_t uplo = upper ? MagmaUpper : MagmaLower;

  auto self_data = self.data_ptr<scalar_t>();
  magma_int_t n = magma_int_cast(self.size(-2), "self.size(-2)");
  auto lda = std::max<magma_int_t>(1, n);

  if (self.dim() == 2) {
    // magmaCholesky requires info to be on CPU
    magma_int_t info_cpu = 0;
    magmaCholesky<scalar_t>(uplo, n, self_data, lda, &info_cpu);
    info.fill_(info_cpu);
  } else {
    TORCH_INTERNAL_ASSERT(info.is_cuda());
    auto info_data = info.data_ptr<magma_int_t>();

    // magmaCholeskyBatched supports only upper=false
    uplo = MagmaLower;

    auto self_mat_stride = matrixStride(self);
    magma_int_t batch_size = magma_int_cast(batchCount(self), "batchCount");

    scalar_t** self_array;

    ALLOCATE_ARRAY(self_array, scalar_t*, batch_size);

    // Set up the created arrays
    for (int64_t i = 0; i < batch_size; i++) {
      self_array[i] = &self_data[i * self_mat_stride];
    }

    MAGMAQueue magma_queue(self.get_device());

    // Compute as many batches of 262140 possible
    // 262140 is the size of the largest batch of matrices that can be run with
    // violating maximum kernel configuration
    // For complex input the batch limit is 65535 (determined experimentally, see https://github.com/pytorch/pytorch/pull/47047#discussion_r516086923 for more information)
    int64_t batch_limit = self.is_complex() ? 65535 : 262140;

    for (int64_t mini_idx = 0; mini_idx < batch_size; mini_idx += batch_limit) {
      int64_t nbatches = std::min(batch_limit, batch_size - mini_idx);
      scalar_t** self_array_cur = &self_array[mini_idx];
      magma_int_t* info_array_cur = &info_data[mini_idx];

      magmaCholeskyBatched<scalar_t>(
        uplo, n, self_array_cur, lda, info_array_cur, nbatches, magma_queue);
    }
  }
#endif
}

void cholesky_helper_magma(const Tensor& input, bool upper, const Tensor& info) {
  Tensor result = input;
  if (input.dim() > 2) {
    // MAGMA's batched cholesky operator has an off-by-one error causing IMA
    // (see https://github.com/pytorch/pytorch/issues/42666). This code is based
    // on the #cloneBatchedColumnMajor function however it pads the input with
    // one extra element utilizing the fact that the resize_as_ method preserves
    // the storage even if it's larger than the new sizes. This way if MAGMA
    // reads off bounds it will still be valid user memory.
    result = at::empty(input.numel() + 1, input.options());
    result.resize_as_(input).transpose_(-2, -1);
    TORCH_INTERNAL_ASSERT_DEBUG_ONLY(result.transpose(-2, -1).is_contiguous());

    // batched MAGMA doesn't support upper=true
    // we transpose and conjugate the input as a workaround
    result.copy_(upper ? input.conj().transpose(-2, -1) : input);
  }

  AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(
    input.scalar_type(), "cholesky_cuda", [&] {
      apply_cholesky<scalar_t>(result, upper, info);
    });

  if (input.dim() > 2) {
    // if upper=true we need to tranpose and conjugate the result tensor
    // because the cholesky decomposition is stored in the lower triangular part
    if (upper) {
      input.copy_(result.conj().transpose(-2, -1));
    } else {
      input.copy_(result);
    }
  }
}

static void cholesky_kernel(const Tensor& input, const Tensor& info, bool upper) {
#ifdef USE_CUSOLVER
  if (batchCount(input) == 1 || !use_magma_ || use_cusolver_potrf_batched_) {
    cholesky_helper_cusolver(input, upper, info);
  } else {
    cholesky_helper_magma(input, upper, info);
  }
#else
  cholesky_helper_magma(input, upper, info);
#endif // USE_CUSOLVER
}

REGISTER_DISPATCH(cholesky_stub, &cholesky_kernel)

// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ cholesky_inverse ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

/*
Computes the inverse of a symmetric (Hermitian) positive-definite matrix n-by-n matrix 'input' using the Cholesky solver
This is an in-place routine, content of 'input' is overwritten.
'infos' is an int Tensor containing error codes for each matrix in the batched input.
MAGMA requires 'infos' to reside in CPU memory.
For more information see MAGMA's documentation for POTRS routine.
*/
template <typename scalar_t>
static void apply_cholesky_inverse(Tensor& input, Tensor& infos, bool upper) {
#ifndef USE_MAGMA
  TORCH_CHECK(false, "cholesky_inverse: MAGMA library not found in compilation. Please rebuild with MAGMA.");
#else
  // magmaCholeskyInverse (magma_dpotri_gpu) is slow because internally
  // it transfers data several times between GPU and CPU and calls lapack routine on CPU
  // using magmaCholeskySolveBatched is a lot faster
  // note that magmaCholeskySolve is also slow

  // 'input' is modified in-place we need to clone it and replace with a diagonal matrix
  // for apply_cholesky_solve
  auto input_working_copy = cloneBatchedColumnMajor(input);

  // 'input' tensor has to be a batch of diagonal matrix
  input.fill_(0);
  input.diagonal(/*offset=*/0, /*dim1=*/-2, /*dim2=*/-1).fill_(1);

  Tensor result_u, input_u;
  if (input.dim() == 2) {
    // unsqueezing here so that the batched version is used
    result_u = input.unsqueeze(0);
    input_u = input_working_copy.unsqueeze(0);
  } else {
    result_u = input;
    input_u = input_working_copy;
  }

  // magma's potrs_batched doesn't take matrix-wise array of ints as an 'info' argument
  // it returns a single 'magma_int_t'
  // if info = 0 the operation is successful, if info = -i, the i-th parameter had an illegal value.
  int64_t info_tmp = 0;
  apply_cholesky_solve<scalar_t>(result_u, input_u, upper, info_tmp);
  infos.fill_(info_tmp);
#endif
}

// This is a type dispatching helper function for 'apply_cholesky_inverse'
Tensor& cholesky_inverse_kernel_impl_magma(Tensor &result, Tensor& infos, bool upper) {
  AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(result.scalar_type(), "cholesky_inverse_out_cuda", [&]{
    apply_cholesky_inverse<scalar_t>(result, infos, upper);
  });
  return result;
}

Tensor& cholesky_inverse_kernel_impl(Tensor &result, Tensor& infos, bool upper) {
  // This function calculates the inverse matrix in-place
  // result should be in column major order and contain matrices to invert
  // the content of result is overwritten by 'apply_cholesky_inverse'
#ifdef USE_CUSOLVER
  if (batchCount(result) == 1 || !use_magma_) {
    return cholesky_inverse_kernel_impl_cusolver(result, infos, upper);
  } else {
    return cholesky_inverse_kernel_impl_magma(result, infos, upper);
  }
#else
  return cholesky_inverse_kernel_impl_magma(result, infos, upper);
#endif

}

REGISTER_DISPATCH(cholesky_inverse_stub, &cholesky_inverse_kernel_impl);

// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ lu ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

/*
  Computes the LU decomposition of a m×n matrix or batch of matrices in 'input' tensor.
  This is an in-place routine, content of 'input', 'pivots', and 'infos' is overwritten.
  This is a "looped" variant for calling single input MAGMA function on batched input.

  Args:
  * `input` - [in] the input matrix for LU decomposition
              [out] the LU decomposition
  * `pivots` - [out] the pivot indices
  * `infos` - [out] error codes, positive values indicate singular matrices
  * `compute_pivots` - controls whether LU is computed with or without pivoting

  For further details, please see the MAGMA documentation for magma_dgetrf_gpu.
*/
template <typename scalar_t>
static void apply_lu_looped_magma(const Tensor& input, const Tensor& pivots, const Tensor& infos, bool compute_pivots) {
#ifndef USE_MAGMA
  TORCH_CHECK(
      false,
      "Calling torch.lu on a CUDA tensor requires compiling ",
      "PyTorch with MAGMA. lease rebuild with MAGMA.");
#else
  // magmaLu and magmaLuNoPiv require infos and pivots tensor to be on CPU
  // the data is later copied back to the appropriate output tensor
  Tensor infos_cpu = at::empty_like(infos, infos.options().device(kCPU).pinned_memory(true));

  auto input_data = input.data_ptr<scalar_t>();
  auto infos_data = infos_cpu.data_ptr<magma_int_t>();
  auto input_matrix_stride = matrixStride(input);
  auto pivots_stride = pivots.size(-1);
  auto batch_size = batchCount(input);
  magma_int_t m = magma_int_cast(input.size(-2), "m");
  magma_int_t n = magma_int_cast(input.size(-1), "n");
  auto leading_dimension = std::max<magma_int_t>(1, m);

  if (compute_pivots) {
    Tensor pivots_cpu = at::empty_like(pivots, pivots.options().device(kCPU).pinned_memory(true));
    auto pivots_data = pivots_cpu.data_ptr<magma_int_t>();
    for (decltype(batch_size) i = 0; i < batch_size; i++) {
      scalar_t* input_working_ptr = &input_data[i * input_matrix_stride];
      int* pivots_working_ptr = &pivots_data[i * pivots_stride];
      int* infos_working_ptr = &infos_data[i];
      magmaLu<scalar_t>(m, n, input_working_ptr, leading_dimension, pivots_working_ptr, infos_working_ptr);
    }
    pivots.copy_(pivots_cpu, /*non_blocking=*/true);
  } else {
    for (decltype(batch_size) i = 0; i < batch_size; i++) {
      scalar_t* input_working_ptr = &input_data[i * input_matrix_stride];
      int* infos_working_ptr = &infos_data[i];
      magmaLuNoPiv<scalar_t>(m, n, input_working_ptr, leading_dimension, infos_working_ptr);
    }

    // fill the pivots tensor with indices using 1-based (Fortran) indexing
    auto k = std::min(m, n);
    Tensor pivots_tmp = at::arange(1, k + 1, input.options().dtype(at::kInt)).expand_as(pivots);
    pivots.copy_(pivots_tmp);
  }
  infos.copy_(infos_cpu, /*non_blocking=*/true);
#endif
}

/*
  Computes the LU decomposition of a m×n matrix or batch of matrices in 'input' tensor.
  This is an in-place routine, content of 'input', 'pivots', and 'infos' is overwritten.
  This is a specialized batched variant, it is expected to be faster than the "looped" version only for small inputs.

  Args:
  * `input` - [in] the input matrix for LU decomposition
              [out] the LU decomposition
  * `pivots` - [out] the pivot indices
  * `infos` - [out] error codes, positive values indicate singular matrices
  * `compute_pivots` - controls whether LU is computed with or without pivoting

  For further details, please see the MAGMA documentation for magma_dgetrf_batched.
*/
template <typename scalar_t>
static void apply_lu_batched_magma(const Tensor& input, const Tensor& pivots, const Tensor& infos, bool compute_pivots) {
#ifndef USE_MAGMA
  TORCH_CHECK(
      false,
      "Calling torch.lu on a CUDA tensor requires compiling ",
      "PyTorch with MAGMA. lease rebuild with MAGMA.");
#else
  auto input_data = input.data_ptr<scalar_t>();
  auto infos_data = infos.data_ptr<magma_int_t>();
  auto input_matrix_stride = matrixStride(input);
  magma_int_t batch_size = magma_int_cast(batchCount(input), "batchCount");

  // magmaLuBatched doesn't work with zero batch dimensions
  // it gives CUDA error: invalid configuration argument
  if (batch_size == 0) {
    infos.fill_(0);
    return;
  }

  magma_int_t m = magma_int_cast(input.size(-2), "m");
  magma_int_t n = magma_int_cast(input.size(-1), "n");
  auto leading_dimension = std::max<magma_int_t>(1, m);

  scalar_t** input_array;
  ALLOCATE_ARRAY(input_array, scalar_t*, batch_size);

  // Set up array of pointers to matrices
  for (int64_t i = 0; i < batch_size; i++) {
    input_array[i] = &input_data[i * input_matrix_stride];
  }

  MAGMAQueue magma_queue(input.get_device());

  if (compute_pivots) {
    auto pivots_data = pivots.data_ptr<magma_int_t>();
    auto pivots_stride = pivots.size(-1);
    // fill pivots with ones to avoid memory access violations inside magma kernels
    // magmaLuBatched might not set the values for it
    // see https://github.com/pytorch/pytorch/pull/53064
    pivots.fill_(1);
    magma_int_t** pivots_array;
    ALLOCATE_ARRAY(pivots_array, magma_int_t*, batch_size);
    for (int64_t i = 0; i < batch_size; i++) {
      pivots_array[i] = &pivots_data[i * pivots_stride];
    }
    magmaLuBatched<scalar_t>(m, n, input_array, leading_dimension, pivots_array, infos_data, batch_size, magma_queue);
  } else {
    magmaLuNoPivBatched<scalar_t>(m, n, input_array, leading_dimension, infos_data, batch_size, magma_queue);

    // fill the pivots tensor with indices using 1-based (Fortran) indexing
    auto k = std::min(m, n);
    Tensor pivots_tmp = at::arange(1, k + 1, input.options().dtype(at::kInt)).expand_as(pivots);
    pivots.copy_(pivots_tmp);
  }

  // block CPU until all operations on the queue are finished
  // this explicit sync prevents garbage results from the subsequent magmaLuSolveBatched call from a different queue
  magma_queue_sync(magma_queue.get_queue());
#endif
}

static void lu_magma(const Tensor& input, const Tensor& pivots, const Tensor& infos, bool compute_pivots) {
  // TODO: compare performance and use the best performing option based on input's sizes
  if (input.dim() == 2) {
    AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(input.scalar_type(), "lu_magma", [&]{
      apply_lu_looped_magma<scalar_t>(input, pivots, infos, compute_pivots);
    });
  } else {
    AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(input.scalar_type(), "lu_magma", [&]{
      apply_lu_batched_magma<scalar_t>(input, pivots, infos, compute_pivots);
    });
  }
}

REGISTER_DISPATCH(lu_stub, &lu_magma);

// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ triangular_solve ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

template <typename scalar_t>
static void apply_triangular_solve_batched(Tensor& A, Tensor& b, bool upper, bool transpose, bool conjugate_transpose, bool unitriangular) {
#ifndef USE_MAGMA
AT_ERROR("triangular_solve: MAGMA library not found in "
         "compilation. Please rebuild with MAGMA.");
#else
  magma_uplo_t uplo = upper ? MagmaUpper : MagmaLower;
  magma_trans_t trans = transpose ? MagmaTrans : MagmaNoTrans;
  trans = conjugate_transpose ? MagmaConjTrans : trans;
  magma_diag_t diag = unitriangular ? MagmaUnit : MagmaNonUnit;

  auto A_data = A.data_ptr<scalar_t>();
  auto b_data = b.data_ptr<scalar_t>();
  magma_int_t m = magma_int_cast(A.size(-2), "A.size(-2)");
  magma_int_t n = magma_int_cast(A.size(-1), "A.size(-1)");
  magma_int_t nrhs = magma_int_cast(b.size(-1), "b.size(-1)");
  // magma returns early if m <= 0 || n <= 0 for magmaTriangularSolveBatched
  // magmaTriangularSolve is calling cuBLAS and it prints
  // ** On entry to DTRSM  parameter number 9 had an illegal value
  // so let's use proper lda parameter here
  magma_int_t lda = std::max<magma_int_t>(1, m);
  magma_int_t batch_size = magma_int_cast(batchCount(A), "batchCount");

  auto A_mat_stride = matrixStride(A);
  auto b_mat_stride = matrixStride(b);

  scalar_t** A_array;
  scalar_t** b_array;

  ALLOCATE_ARRAY(A_array, scalar_t*, batch_size);
  ALLOCATE_ARRAY(b_array, scalar_t*, batch_size);

  // Set up the created arrays
  for (int64_t i = 0; i < batch_size; i++) {
    A_array[i] = &A_data[i * A_mat_stride];
    b_array[i] = &b_data[i * b_mat_stride];
  }

  MAGMAQueue magma_queue(b.get_device());

  constexpr int64_t batch_limit = 65535;
  // Compute as many batches of 65535 possible
  // The number of "mini"-batches are floor(batch_size / batch_limit)
  // and these cover floor(batch_size / batch_limit) * batch_limit matrix solves
  int64_t mini_batches = batch_size / batch_limit;
  int64_t mini_idx; // this is outside the loop because it is used for the case batch_size % batch_limit != 0
  for (mini_idx = 0; mini_idx < mini_batches * batch_limit; mini_idx += batch_limit) {
    scalar_t** A_array_cur = &A_array[mini_idx];
    scalar_t** b_array_cur = &b_array[mini_idx];

    magmaTriangularSolveBatched<scalar_t>(
        uplo, trans, diag, n, nrhs, A_array_cur,
        lda, b_array_cur, lda, batch_limit, magma_queue);
  }

  // Compute whatever is left = batch_size - floor(batch_size / batch_limit) * batch_limit
  // which concisely is equal to batch_size % batch_limit
  if (batch_size % batch_limit != 0) {
    magmaTriangularSolveBatched<scalar_t>(
        uplo, trans, diag, n, nrhs, &A_array[mini_idx],
        lda, &b_array[mini_idx], lda, batch_size % batch_limit, magma_queue);
  }
#endif
}

void triangular_solve_batched_magma(Tensor& A, Tensor& B, Tensor& infos, bool upper, bool transpose, bool conjugate_transpose, bool unitriangular) {
  (void)infos; // unused
  AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(A.scalar_type(), "triangular_solve_cuda", [&]{
    apply_triangular_solve_batched<scalar_t>(A, B, upper, transpose, conjugate_transpose, unitriangular);
  });
}

void triangular_solve_kernel(Tensor& A, Tensor& B, Tensor& infos, bool upper, bool transpose, bool conjugate_transpose, bool unitriangular) {
  // For batches smaller than 8 and matrix sizes larger than 64x64 cuBLAS forloop is faster than batched version
  if (batchCount(A) <= 8 && A.size(-1) >= 64) {
    triangular_solve_cublas(A, B, infos, upper, transpose, conjugate_transpose, unitriangular);
  } else {
#ifndef USE_MAGMA
    triangular_solve_batched_cublas(A, B, infos, upper, transpose, conjugate_transpose, unitriangular);
#else
    // cuBLAS batched is faster than MAGMA batched up until 512x512, after that MAGMA is faster
    if (A.size(-1) <= 512) {
      triangular_solve_batched_cublas(A, B, infos, upper, transpose, conjugate_transpose, unitriangular);
    } else {
      triangular_solve_batched_magma(A, B, infos, upper, transpose, conjugate_transpose, unitriangular);
    }
#endif // USE_MAGMA
  }
}

REGISTER_DISPATCH(triangular_solve_stub, &triangular_solve_kernel);

// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ orgqr ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Tensor& orgqr_kernel_impl(Tensor& result, const Tensor& tau) {
  // TODO: It is possible to implement efficient batched orgqr for small tau (tau.size(-1) <= 32)
  // using MAGMA, however it fails on Windows because of some illegal memory reads inside MAGMA.
  // See discussions in https://github.com/pytorch/pytorch/pull/51348 for comparison of cuSOLVER-MAGMA
  // and Windows failure.
  // For reference here is the MAGMA-based implementation: https://gist.github.com/IvanYashchuk/2db50002c9d3c1462ff769e6410ad983
  #if defined(USE_CUSOLVER)
    return orgqr_helper_cusolver(result, tau); // cusolver
  #else
    TORCH_CHECK(false, "Calling torch.orgqr on a CUDA tensor requires compiling ",
      "PyTorch with cuSOLVER. Please use PyTorch built with cuSOLVER support.");
  #endif
}

REGISTER_DISPATCH(orgqr_stub, &orgqr_kernel_impl);

void ormqr_kernel(const Tensor& input, const Tensor& tau, const Tensor& other, bool left, bool transpose) {
#if defined(USE_CUSOLVER)
  ormqr_cusolver(input, tau, other, left, transpose);
#else
  TORCH_CHECK(false,
      "Calling torch.ormqr on a CUDA tensor requires compiling ",
      "PyTorch with cuSOLVER. Please use PyTorch built with cuSOLVER support.");
#endif
}

REGISTER_DISPATCH(ormqr_stub, &ormqr_kernel);

// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ qr ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

template <typename scalar_t>
static void apply_geqrf(const Tensor& input, const Tensor& tau) {
#ifndef USE_MAGMA
  TORCH_CHECK(
    false,
    "Calling torch.geqrf on a CUDA tensor requires compiling ",
    "PyTorch with MAGMA. Please use PyTorch built with MAGMA support.");
#else

  magma_int_t m = magma_int_cast(input.size(-2), "m");
  magma_int_t n = magma_int_cast(input.size(-1), "n");

  auto input_data = input.data_ptr<scalar_t>();
  auto input_matrix_stride = matrixStride(input);
  auto tau_stride = tau.size(-1);
  auto batch_size = batchCount(input);
  auto lda = std::max<int>(1, m);

  // magmaGeqrf uses a hybrid CPU-GPU algorithm to compute the elementary reflectors.
  // The driver routine geqrf2_gpu accepts a tensor on the CPU for elementary reflectors.
  Tensor tau_cpu = at::empty(tau.sizes(), tau.options().device(at::kCPU).pinned_memory(true));
  scalar_t* tau_data = tau_cpu.data_ptr<scalar_t>();
  scalar_t* work_data = nullptr; // workspace is not needed for geqrf2_gpu

  magma_int_t info = 0;
  for (int64_t i = 0; i < batch_size; i++) {
    scalar_t* input_working_ptr = &input_data[i * input_matrix_stride];
    scalar_t* tau_working_ptr = &tau_data[i * tau_stride];

    // now compute the actual QR and tau
    // MAGMA's geqrf2_gpu function is used, this version has LAPACK-complaint arguments.
    magmaGeqrf<scalar_t>(m, n, input_working_ptr, lda, tau_working_ptr, work_data, &info, /*is_v2=*/true);
    checkMagmaInternalError(info, "geqrf");
  }
  tau.copy_(tau_cpu, /*non_blocking=*/true);
#endif
}

// This is a type dispatching helper function for 'apply_geqrf'
void geqrf_magma(const Tensor& input, const Tensor& tau) {
  AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(input.scalar_type(), "geqrf_magma", [&]{
    apply_geqrf<scalar_t>(input, tau);
  });
}

// This is a backend library dispatching helper function for calling looped batch implementation
void geqrf_looped(const Tensor& input, const Tensor& tau) {
#if defined(USE_CUSOLVER)
  return geqrf_cusolver(input, tau);
#else
  return geqrf_magma(input, tau);
#endif
}

// This is a backend library dispatching helper function for calling specialized batched implementation
void geqrf_batched(const Tensor& input, const Tensor& tau) {
#ifdef CUDART_VERSION
  // if cuBLAS is available
  return geqrf_batched_cublas(input, tau);
#else
  // TODO: implement MAGMA-based path using magma_zgeqrf_expert_batched
  return geqrf_looped(input, tau);
#endif
}

void geqrf_kernel(const Tensor& input, const Tensor& tau) {
  // if number of rows is smaller than 32 batched is always faster for batch size > 1
  // for larger number of rows number of batches condition
  if (input.size(-2) <= 256 && batchCount(input) >= std::max<int64_t>(2, input.size(-2) / 16)) {
    return geqrf_batched(input, tau);
  } else {
    return geqrf_looped(input, tau);
  }
}

REGISTER_DISPATCH(geqrf_stub, &geqrf_kernel);

template <typename scalar_t>
static void apply_qr(Tensor& Q, Tensor& R, int64_t q_size_minus_2, int64_t r_size_minus_1, int64_t n_columns,
                     bool compute_q) {
#ifndef USE_MAGMA
AT_ERROR("qr: MAGMA library not found in "
    "compilation. Please rebuild with MAGMA.");
#else

  magma_int_t m = magma_int_cast(q_size_minus_2, "Q.size(-2)");
  magma_int_t n = magma_int_cast(r_size_minus_1, "R.size(-1)");

  auto r_data = R.data_ptr<scalar_t>();
  auto r_matrix_stride = matrixStride(R);
  magma_int_t k = m < n ? m : n;
  magma_int_t nb = magmaGeqrfOptimalBlocksize<scalar_t>(m, n);
  int64_t batch_size = batchCount(R);

  // magmaGeqrf uses a hybrid CPU-GPU algorithm to compute the elementary reflectors.
  // The driver routine magma_(d/s)geqrf2_gpu accepts a tensor on the CPU for elementary reflectors.
  Tensor tau = at::empty({k}, Q.options().device(at::kCPU));
  Tensor work = at::empty({(2 * k + magma_roundup(n, 32)) * nb}, R.options());
  scalar_t* tau_data = tau.data_ptr<scalar_t>();
  scalar_t* work_data = work.data_ptr<scalar_t>();

  // This phase computes R (the raw version)
  // This uses MAGMA's ?geqrf2_gpu function
  magma_int_t info = 0;
  for (int64_t i = 0; i < batch_size; i++) {
    scalar_t* r_working_ptr = &r_data[i * r_matrix_stride];
    magmaGeqrf<scalar_t>(m, n, r_working_ptr, m, tau_data, work_data, &info, /*is_v2=*/true);
    checkMagmaInternalError(info, "geqrf");
  }
  if (!compute_q) {
    // this is for mode='r'
    return;
  }

  // This phase computes Q (the raw version)
  // We require to perform ?geqrf_gpu again due to this bug in MAGMA:
  // - ?geqrf_gpu allows fast computation of Q via ?orgqr_gpu, but doesn't give R properly.
  // - ?geqrf2_gpu gives correct R, but doesn't allow computation of Q via ?orgqr_gpu
  // Refer to the below link for more details:
  // http://icl.cs.utk.edu/magma/forum/viewtopic.php?f=2&t=1015&p=2800&hilit=geqrf_gpu#p2800
  auto q_data = Q.data_ptr<scalar_t>();
  auto q_matrix_stride = matrixStride(Q);
  for (int64_t i = 0; i < batch_size; i++) {
    scalar_t* q_working_ptr = &q_data[i * q_matrix_stride];
    magmaGeqrf<scalar_t>(m, n, q_working_ptr, m, tau_data, work_data, &info, /*is_v2=*/false);
    checkMagmaInternalError(info, "geqrf");

    magmaOrgqr<scalar_t>(m, n_columns, k, q_working_ptr, m, tau_data, work_data, nb, &info);
    checkMagmaInternalError(info, "orgqr");
  }
#endif
}

std::tuple<Tensor, Tensor> linalg_qr_helper_magma(const Tensor& self, c10::string_view mode) {
  bool compute_q, reduced;
  std::tie(compute_q, reduced) = _parse_qr_mode(mode);

  // Setup input geometry and inputs for apply_qr
  std::vector<int64_t> q_sizes, q_strides;
  int64_t n_columns_q;
  std::tie(q_sizes, q_strides, n_columns_q) = _compute_geometry_for_Q(self, reduced);
  Tensor q_working_copy, r_working_copy;

  // If there are no elements, then we simply return a pair of tensors of required dimensions
  if (self.numel() == 0) {
    int64_t n = self.size(-1);
    auto r_shape = self.sizes().vec();
    r_shape.end()[-2] = n_columns_q;
    r_shape.end()[-1] = n;
    r_working_copy = at::empty(r_shape, self.options());
    if (compute_q) {
        auto q_shape = q_sizes;
        q_shape.end()[-1] = n_columns_q;
        q_working_copy = at::zeros(q_shape, self.options());
        q_working_copy.diagonal(/*offset=*/0, /*dim1=*/-2, /*dim2=*/-1).fill_(1);
    } else {
      q_working_copy = at::empty({0}, self.options());
    }
    return std::make_tuple(q_working_copy, r_working_copy);
  }

  if (compute_q) {
    q_working_copy = at::empty_strided(q_sizes, q_strides, self.options());
    q_working_copy.narrow(-1, 0, self.size(-1)).copy_(self);
  } else {
    q_working_copy = at::empty({0}, self.options());
  }
  r_working_copy = cloneBatchedColumnMajor(self);

  int64_t m = q_sizes[self.dim() - 2];
  int64_t n = r_working_copy.size(-1);

  AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(self.scalar_type(), "qr_cuda", [&]{
    apply_qr<scalar_t>(q_working_copy, r_working_copy, m, n, n_columns_q, compute_q);
  });

  if (compute_q) {
    q_working_copy = q_working_copy.narrow(-1, 0, n_columns_q);
  }
  r_working_copy = r_working_copy.narrow(-2, 0, n_columns_q).triu();
  return std::make_tuple(q_working_copy, r_working_copy);
}

std::tuple<Tensor, Tensor> _linalg_qr_helper_cuda(const Tensor& input, c10::string_view mode) {
#if defined(USE_CUSOLVER)
  // _linalg_qr_helper_default is a generic function that is implemented using
  // geqrf_stub and orgqr_stub. It dispatches to cuSOLVER for CUDA inputs if USE_CUSOLVER is defined
  return _linalg_qr_helper_default(input, mode);
#else
  return linalg_qr_helper_magma(input, mode);
#endif
}

// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ symeig ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

template <typename scalar_t>
static void apply_magma_eigh(Tensor& values, Tensor& vectors, Tensor& infos, bool upper, bool compute_eigenvectors) {
#ifndef USE_MAGMA
  TORCH_CHECK(
    false,
    "Calling torch.linalg.eigh/eigvalsh on a CUDA tensor requires compiling ",
    "PyTorch with MAGMA. Please use PyTorch built with MAGMA support.");
#else
  TORCH_INTERNAL_ASSERT_DEBUG_ONLY(values.device() == kCPU);
  TORCH_INTERNAL_ASSERT_DEBUG_ONLY(infos.device() == kCPU);

  using value_t = typename c10::scalar_value_type<scalar_t>::type;

  magma_uplo_t uplo = upper ? MagmaUpper : MagmaLower;
  magma_vec_t jobz = compute_eigenvectors ? MagmaVec : MagmaNoVec;

  magma_int_t n = magma_int_cast(vectors.size(-1), "n");
  auto lda = std::max<magma_int_t>(1, n);
  auto batch_size = batchCount(vectors);

  auto vectors_stride = matrixStride(vectors);
  auto values_stride = values.size(-1);

  auto vectors_data = vectors.data_ptr<scalar_t>();
  auto values_data = values.data_ptr<value_t>();
  auto infos_data = infos.data_ptr<magma_int_t>();

  scalar_t* wA;
  ALLOCATE_ARRAY(wA, scalar_t, lda * lda);

  // Run once, first to get the optimum work sizes.
  // Since we deal with batches of matrices with the same dimensions, doing this outside
  // the loop saves (batch_size - 1) workspace queries which would provide the same result
  // and (batch_size - 1) calls to allocate and deallocate workspace using at::empty()
  magma_int_t lwork = -1;
  scalar_t wkopt;
  magma_int_t liwork = -1;
  magma_int_t iwkopt;
  magma_int_t lrwork = -1;
  value_t rwkopt;
  magmaSyevd<scalar_t, value_t>(jobz, uplo, n, vectors_data, lda, values_data,
    wA, lda, &wkopt, lwork, &rwkopt, lrwork, &iwkopt, liwork, infos_data);

  scalar_t* work;
  magma_int_t* iwork;
  lwork = magma_int_cast(std::max<int64_t>(1, real_impl<scalar_t, value_t>(wkopt)), "work_size");
  liwork = magma_int_cast(std::max<int64_t>(1, iwkopt), "iwork_size");
  ALLOCATE_ARRAY(work, scalar_t, lwork);
  ALLOCATE_ARRAY(iwork, magma_int_t, liwork);

  value_t* rwork = nullptr;
  c10::Storage storage_rwork;
  if (vectors.is_complex()) {
    lrwork = magma_int_cast(std::max<int64_t>(1, rwkopt), "rwork_size");
    storage_rwork = pin_memory<value_t>(lrwork);
    rwork = static_cast<value_t*>(storage_rwork.data());
  }

  for (decltype(batch_size) i = 0; i < batch_size; i++) {
    scalar_t* vectors_working_ptr = &vectors_data[i * vectors_stride];
    value_t* values_working_ptr = &values_data[i * values_stride];
    magma_int_t* info_working_ptr = &infos_data[i];
    magmaSyevd<scalar_t, value_t>(jobz, uplo, n, vectors_working_ptr, lda, values_working_ptr,
      wA, lda, work, lwork, rwork, lrwork, iwork, liwork, info_working_ptr);
    // The current behaviour for Linear Algebra functions to raise an error if something goes wrong
    // or input doesn't satisfy some requirement
    // therefore return early since further computations will be wasted anyway
    if (*info_working_ptr != 0) {
      return;
    }
  }
#endif
}

std::tuple<Tensor, Tensor> _symeig_helper_cuda(const Tensor& self, bool eigenvectors, bool upper) {
  Tensor infos = at::zeros({std::max<int64_t>(1, batchCount(self))}, self.options().dtype(kInt).device(at::kCPU));

  auto eigvals_shape = IntArrayRef(self.sizes().data(), self.dim()-1);  // self.shape[:-1]
  ScalarType real_dtype = toValueType(self.scalar_type());

  // magmaSyevd uses a hybrid CPU-GPU algorithm to compute the eigenvalues and eigenvectors.
  // The driver routine magma_(d/s)syev_gpu accepts a tensor on the CPU for eigvalenvalues.
  // The data is later moved to the appropriate device.
  // In the case where self.numel() == 0, we just return an empty tensor of
  // dimensions on the CUDA (to avoid the unnecessary "to(at::kCUDA)")
  auto eigvals_working_copy = self.numel() == 0
                              ? at::empty(eigvals_shape, self.options().dtype(real_dtype))
                              : at::empty(eigvals_shape, self.options().dtype(real_dtype).device(at::kCPU));

  if (self.numel() == 0) {
    return std::tuple<Tensor, Tensor>(eigvals_working_copy, at::empty_like(self, LEGACY_CONTIGUOUS_MEMORY_FORMAT));
  }

  auto self_working_copy = cloneBatchedColumnMajor(self);
  AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(self.scalar_type(), "symeig_cuda", [&]{
    apply_magma_eigh<scalar_t>(eigvals_working_copy, self_working_copy, infos, upper, eigenvectors);
  });

  if (self.dim() > 2) {
    batchCheckErrors(infos, "symeig_cuda");
  } else {
    singleCheckErrors(infos.item().toInt(), "symeig_cuda");
  }
  if (eigenvectors) {
    return std::tuple<Tensor, Tensor>(eigvals_working_copy.to(self.device()), self_working_copy);
  } else {
    return std::tuple<Tensor, Tensor>(eigvals_working_copy.to(self.device()), at::empty({0}, self.options()));
  }
}

// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ linalg_eigh ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

// This is a type dispatch function for 'apply_magma_eigh'
// For small inputs result is computed on CPU
void linalg_eigh_magma(Tensor& eigenvalues, Tensor& eigenvectors, Tensor& infos, bool upper, bool compute_eigenvectors) {
  // MAGMA just calls LAPACK for eigenvectors.size(-1) <= 128
  // See https://bitbucket.org/icl/magma/src/e6fdca447bd402693e8b0b950a898b6879bbcc41/src/zheevd_gpu.cpp?at=master#lines-258
  // in addition lda is ignored breaking 0x0 inputs
  if (eigenvectors.size(-1) > 128) {
    // MAGMA requires eigenvalues and infos tensors to reside on CPU
    Tensor eigenvalues_cpu = eigenvalues.to(kCPU);
    infos = infos.to(kCPU);

    AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(
      eigenvectors.scalar_type(), "linalg_eigh_cpu", [&] {
        apply_magma_eigh<scalar_t>(
            eigenvalues_cpu, eigenvectors, infos, upper, compute_eigenvectors);
      });

    // Transfer computed by MAGMA results from CPU to GPU
    eigenvalues.copy_(eigenvalues_cpu);
  } else { // eigenvectors.size(-1) <= 128
    // transfer to CPU, compute the result and copy back to GPU
    // this is faster than going through MAGMA that does the same
    Tensor eigenvalues_cpu = at::empty_like(eigenvalues, eigenvalues.options().device(kCPU));
    if (compute_eigenvectors) {
      Tensor eigenvectors_cpu = at::empty_like(eigenvectors, eigenvectors.options().device(kCPU));
      at::linalg_eigh_out(eigenvalues_cpu, eigenvectors_cpu, eigenvectors.to(kCPU), upper ? "U" : "L");
      eigenvectors.copy_(eigenvectors_cpu);
    } else {
      at::linalg_eigvalsh_out(eigenvalues_cpu, eigenvectors.to(kCPU), upper ? "U" : "L");
    }
    eigenvalues.copy_(eigenvalues_cpu);
  }
}

void linalg_eigh_kernel(Tensor& eigenvalues, Tensor& eigenvectors, Tensor& infos, bool upper, bool compute_eigenvectors) {
#if defined(USE_CUSOLVER)
  linalg_eigh_cusolver(eigenvalues, eigenvectors, infos, upper, compute_eigenvectors);
#else
  linalg_eigh_magma(eigenvalues, eigenvectors, infos, upper, compute_eigenvectors);
#endif
}

REGISTER_DISPATCH(linalg_eigh_stub, &linalg_eigh_kernel);

// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ eig ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

// magmaEig uses a hybrid CPU-GPU algorithm, which takes and return CPU
// memory. So, we accept a GPU tensor, copy it to CPU memory, and later copy
// the returned values from CPU to GPU. See also magmaSymeig, which uses a
// similar approach.

template <typename scalar_t>
static void apply_eig(const Tensor& self, bool eigenvectors, Tensor& out_eigvals, Tensor& out_eigvecs,
                      int64_t *info_ptr) {
#ifndef USE_MAGMA
TORCH_CHECK(false, "Calling torch.eig on a CUDA tensor requires compiling PyTorch with MAGMA. "
                   "Either transfer the tensor to the CPU before calling torch.eig or recompile with MAGMA.");
#else
  TORCH_INTERNAL_ASSERT(self.device() == at::kCPU, "Internal error: apply_eig needs a CPU tensor");
  using value_t = typename c10::scalar_value_type<scalar_t>::type;
  magma_vec_t jobvr = eigenvectors ? MagmaVec : MagmaNoVec;
  magma_int_t n = magma_int_cast(self.size(-1), "n");
  auto self_data = self.data_ptr<scalar_t>();

  auto out_eigvals_data = out_eigvals.data_ptr<scalar_t>();
  scalar_t *wr = out_eigvals_data;

  scalar_t *vr_data = NULL;
  magma_int_t ldvr = 1;
  if (jobvr == MagmaVec)
  {
      vr_data = out_eigvecs.data_ptr<scalar_t>();
      ldvr = n;
  }

  value_t *rwork_data = nullptr;
  if (isComplexType(at::typeMetaToScalarType(self.dtype()))) {
    ALLOCATE_ARRAY(rwork_data, value_t, n*2);
  }

  if (n > 0) {
    // call magmaEig once to get the optimal size of work_data
    scalar_t wkopt;
    magma_int_t info;
    magmaEig<scalar_t, value_t>(MagmaNoVec, jobvr, n, self_data, n, wr, NULL, 1, vr_data, ldvr, &wkopt, -1, rwork_data, &info);
    magma_int_t lwork = static_cast<magma_int_t>(real_impl<scalar_t, value_t>(wkopt));

    // call it a 2nd time to to the actual work
    scalar_t *work_data = nullptr;
    ALLOCATE_ARRAY(work_data, scalar_t, lwork);
    magmaEig<scalar_t, value_t>(MagmaNoVec, jobvr, n, self_data, n, wr, NULL, 1, vr_data, ldvr, work_data, lwork, rwork_data, &info);
    *info_ptr = info;
  }
#endif
}

/*
 * Internal helper; like eig_cuda but:
 *   1. assume that self is a square matrix of side "n"
 *   2. return CPU tensors (because this is what magmaEig returns), which will be copied to GPU memory
 *      by the caller
 */
std::tuple<Tensor, Tensor> eig_kernel_impl(const Tensor& self, bool& eigenvectors) {
  int64_t n = self.size(-1);
  // copy self to pinned CPU memory
  auto self_working_copy = at::empty_strided(
      {n, n}, // square matrix
      {1, n}, // column-ordered, as magmaEig expects
      at::TensorOptions(at::kCPU).dtype(self.dtype()).pinned_memory(true));
  self_working_copy.copy_(self);

  // tensors holding the results. We use empty_strided to make them column-ordered
  auto options = self.options().device(at::kCPU).memory_format(LEGACY_CONTIGUOUS_MEMORY_FORMAT);
  Tensor out_eigvals;
  if (isComplexType(at::typeMetaToScalarType(self.dtype()))) {
      out_eigvals = at::empty({n}, options);
  } else {
      out_eigvals = at::empty_strided({n, 2}, {1, n}, options);
  }
  auto out_eigvecs = eigenvectors
                     ? at::empty_strided({n, n}, {1, n}, options)
                     : Tensor();

  int64_t info;
  AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(self.scalar_type(), "eig_cuda", [&]{
    apply_eig<scalar_t>(self_working_copy, eigenvectors, out_eigvals, out_eigvecs, &info);
  });
  singleCheckErrors(info, "eig_cuda");

  return std::tuple<Tensor, Tensor>(out_eigvals, out_eigvecs);
}

REGISTER_DISPATCH(eig_stub, &eig_kernel_impl);

// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ linalg_eig ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

/*
Computes the eigenvalues and eigenvectors of n-by-n matrix 'input'.
This is an in-place routine, content of 'input', 'values', 'vectors' is overwritten.
'infos' is an int Tensor containing error codes for each matrix in the batched input.
For more information see MAGMA's documentation for GEEV routine.
*/
template <typename scalar_t>
void apply_linalg_eig(Tensor& values, Tensor& vectors, Tensor& input, Tensor& infos, bool compute_eigenvectors) {
#ifndef USE_MAGMA
TORCH_CHECK(false, "Calling torch.linalg.eig on a CUDA tensor requires compiling PyTorch with MAGMA. "
                   "Either transfer the tensor to the CPU before calling torch.linalg.eig or recompile with MAGMA.");
#else
  TORCH_INTERNAL_ASSERT_DEBUG_ONLY(input.device() == at::kCPU);
  TORCH_INTERNAL_ASSERT_DEBUG_ONLY(values.device() == at::kCPU);
  TORCH_INTERNAL_ASSERT_DEBUG_ONLY(infos.device() == at::kCPU);
  if (compute_eigenvectors) {
    TORCH_INTERNAL_ASSERT_DEBUG_ONLY(vectors.device() == at::kCPU);
  }

  using value_t = typename c10::scalar_value_type<scalar_t>::type;

  magma_vec_t jobvr = compute_eigenvectors ? MagmaVec : MagmaNoVec;
  magma_vec_t jobvl = MagmaNoVec;  // only right eigenvectors are computed
  magma_int_t n = magma_int_cast(input.size(-1), "n");
  auto lda = std::max<magma_int_t>(1, n);
  auto batch_size = batchCount(input);
  auto input_matrix_stride = matrixStride(input);
  auto values_stride = values.size(-1);
  auto input_data = input.data_ptr<scalar_t>();
  auto values_data = values.data_ptr<scalar_t>();
  auto infos_data = infos.data_ptr<magma_int_t>();
  auto rvectors_data = compute_eigenvectors ? vectors.data_ptr<scalar_t>() : nullptr;
  scalar_t* lvectors_data = nullptr;  // only right eigenvectors are computed
  int64_t ldvr = compute_eigenvectors ? lda : 1;
  int64_t ldvl = 1;

  Tensor rwork;
  value_t* rwork_data = nullptr;
  if (input.is_complex()) {
    ScalarType real_dtype = toValueType(input.scalar_type());
    rwork = at::empty({lda * 2}, input.options().dtype(real_dtype));
    rwork_data = rwork.data_ptr<value_t>();
  }

  // call magmaEig once to get the optimal size of work_data
  scalar_t work_query;
  magmaEig<scalar_t, value_t>(jobvl, jobvr, n, input_data, lda, values_data,
    lvectors_data, ldvl, rvectors_data, ldvr, &work_query, -1, rwork_data, &infos_data[0]);

  magma_int_t lwork = std::max<magma_int_t>(1, static_cast<magma_int_t>(real_impl<scalar_t, value_t>(work_query)));
  Tensor work = at::empty({lwork}, input.dtype());
  auto work_data = work.data_ptr<scalar_t>();

  for (auto i = decltype(batch_size){0}; i < batch_size; i++) {
    scalar_t* input_working_ptr = &input_data[i * input_matrix_stride];
    scalar_t* values_working_ptr = &values_data[i * values_stride];
    scalar_t* rvectors_working_ptr = compute_eigenvectors ? &rvectors_data[i * input_matrix_stride] : nullptr;
    int* info_working_ptr = &infos_data[i];
    magmaEig<scalar_t, value_t>(jobvl, jobvr, n, input_working_ptr, lda, values_working_ptr,
      lvectors_data, ldvl, rvectors_working_ptr, ldvr, work_data, lwork, rwork_data, info_working_ptr);
  }
#endif
}

// This is a type dispatching helper function for 'apply_linalg_eig'
void linalg_eig_kernel(Tensor& eigenvalues, Tensor& eigenvectors, Tensor& infos, const Tensor& input, bool compute_eigenvectors) {
  // This function calculates the non-symmetric eigendecomposition in-place
  // tensors should be in batched column major memory format
  // the content of eigenvalues, eigenvectors and infos is overwritten by 'apply_linalg_eig'

  // apply_linalg_eig modifies the provided input matrix in-place, therefore we need a copy
  // MAGMA doesn't have GPU interface for the eigendecomposition and it forces us to transfer 'input' to CPU
  TORCH_INTERNAL_ASSERT_DEBUG_ONLY(input.is_cuda());
  Tensor input_working_copy = at::empty(input.sizes(), input.options().device(kCPU));
  input_working_copy.transpose_(-2, -1);  // make input_working_copy to have Fortran contiguous memory layout
  input_working_copy.copy_(input);

  AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(input.scalar_type(), "linalg_eig_out_cuda", [&]{
    apply_linalg_eig<scalar_t>(eigenvalues, eigenvectors, input_working_copy, infos, compute_eigenvectors);
  });
}

REGISTER_DISPATCH(linalg_eig_stub, &linalg_eig_kernel);

// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ svd ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

template<typename scalar_t>
static void apply_svd(Tensor& self, Tensor& U, Tensor& S, Tensor& VT,
                      char jobchar, std::vector<int64_t>& infos) {
#ifndef USE_MAGMA
AT_ERROR("svd: MAGMA library not found in "
    "compilation. Please rebuild with MAGMA.");
#else
  using value_t = typename c10::scalar_value_type<scalar_t>::type;
  auto self_data = self.data_ptr<scalar_t>();
  auto U_data = U.data_ptr<scalar_t>();
  auto S_data = S.data_ptr<value_t>();
  auto VT_data = VT.data_ptr<scalar_t>();
  auto self_stride = matrixStride(self);
  auto U_stride = matrixStride(U);
  auto S_stride = S.size(-1);
  auto VT_stride = matrixStride(VT);
  auto batchsize = batchCount(self);

  magma_vec_t jobz = jobchar == 'A' ? MagmaAllVec : (jobchar == 'S' ? MagmaSomeVec : MagmaNoVec);

  magma_int_t m = magma_int_cast(self.size(-2), "m");
  magma_int_t n = magma_int_cast(self.size(-1), "n");
  auto lda = std::max<magma_int_t>(1, m);
  auto ldvt = std::max<magma_int_t>(1, n);
  auto mn = std::min(m, n);

  c10::Storage storage_rwork;
  value_t* rwork = nullptr;

  magma_int_t* iwork;
  ALLOCATE_ARRAY(iwork, magma_int_t, 8 * mn);
  if (isComplexType(at::typeMetaToScalarType(self.dtype()))) {
    auto lrwork = computeLRWorkDim(jobchar, m, n);
    storage_rwork = pin_memory<value_t>(lrwork);
    rwork = static_cast<value_t*>(storage_rwork.data());
  }

  magma_int_t info = 0;
  // Run once, first to get the optimum work size.
  // Since we deal with batches of matrices with the same dimensions, doing this outside
  // the loop saves (batch_size - 1) workspace queries which would provide the same result
  // and (batch_size - 1) calls to allocate and deallocate workspace using at::empty()
  magma_int_t lwork = -1;
  scalar_t wkopt = 1; // MAGMA might not set the value for the optimal workspace therefore use 1 as the default value
  magmaSvd<scalar_t, value_t>(jobz, m, n, self_data, lda, S_data, U_data, lda, VT_data, ldvt, &wkopt, lwork, rwork, iwork, &info);
  lwork = magma_int_cast(real_impl<scalar_t, value_t>(wkopt), "work_size");
  scalar_t* work;
  ALLOCATE_ARRAY(work, scalar_t, lwork);

  for (int64_t i = 0; i < batchsize; i++) {
    scalar_t* self_working_ptr = &self_data[i * self_stride];
    value_t* S_working_ptr = &S_data[i * S_stride];
    scalar_t* U_working_ptr = &U_data[i * U_stride];
    scalar_t* VT_working_ptr = &VT_data[i * VT_stride];

    // Compute S, U (optionally), VT (optionally)
    magmaSvd<scalar_t, value_t>(jobz, m, n, self_working_ptr, lda,
                                S_working_ptr, U_working_ptr, lda, VT_working_ptr, ldvt, work, lwork, rwork, iwork, &info);
    infos[i] = info;
    if (info != 0) {
      return;
    }
  }
#endif
}

std::tuple<Tensor, Tensor, Tensor> _svd_helper_cuda_legacy(const Tensor& self, bool some, bool compute_uv) {
  std::vector<int64_t> infos(batchCount(self), 0);
  int64_t m = self.size(-2), n = self.size(-1);
  int64_t k = std::min(m, n);

  char jobchar = compute_uv ? (some ? 'S' : 'A') : 'N';

  Tensor U_working_copy, S_working_copy, VT_working_copy;
  std::tie(U_working_copy, S_working_copy, VT_working_copy) = _create_U_S_VT(self, some, compute_uv);

  // The input matrix, U, S and VT have to reside in pinned memory.
  // Additionally, the input and U have to be in column major format.
  // _create_U_S_VT takes care of a part of these requirements (for U, S and VT)
  // For the input matrix, this requirements are being taken care of below.
  // Specify strides
  auto self_col_major_strides = at::detail::defaultStrides(self.sizes());
  self_col_major_strides[self.dim() - 2] = 1;
  self_col_major_strides[self.dim() - 1] = m;
  // Create strided tensor in pinned memory
  auto self_working_copy = at::empty_strided(self.sizes(), self_col_major_strides,
                                              at::TensorOptions(at::kCPU).dtype(self.dtype()).pinned_memory(true));
  self_working_copy.copy_(self);

  AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(self.scalar_type(), "svd_cuda", [&] {
    apply_svd<scalar_t>(self_working_copy, U_working_copy, S_working_copy, VT_working_copy, jobchar, infos);
  });

  if (self.dim() > 2) {
    batchCheckErrors(infos, "svd_cuda");
  } else {
    singleCheckErrors(infos[0], "svd_cuda");
  }

  U_working_copy = same_stride_to(U_working_copy, self.options());
  S_working_copy = same_stride_to(S_working_copy, S_working_copy.options().device(self.device()));
  VT_working_copy = same_stride_to(VT_working_copy, self.options());

  if (!compute_uv) {
    VT_working_copy.zero_();
    U_working_copy.zero_();
  }

  if (some) {
    VT_working_copy = VT_working_copy.narrow(-2, 0, k);
  }

  // so far we have computed VT, but torch.svd returns V instead. Adjust accordingly.
  // Note that the 'apply_svd' routine returns VT = V^T (for real inputs) or VT = V^H (for complex inputs), not V.
  VT_working_copy = VT_working_copy.conj();
  VT_working_copy.transpose_(-2, -1);
  return std::make_tuple(U_working_copy, S_working_copy, VT_working_copy);
}

std::tuple<Tensor, Tensor, Tensor> _svd_helper_cuda(const Tensor& self, bool some, bool compute_uv) {
#ifdef USE_CUSOLVER
  return _svd_helper_cuda_lib(self, some, compute_uv);
#else
  return _svd_helper_cuda_legacy(self, some, compute_uv);
#endif
}

// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ lu_solve ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

/*
  Solves the matrix equation A X = B
  X and B are n-by-nrhs matrices, A is represented using the LU factorization.
  This is an in-place routine, content of `b` is overwritten.
  This is a "looped" variant for calling single input MAGMA function on batched input.

  Args:
  * `b` -  [in] the right hand side matrix B
           [out] the solution matrix X
  * `lu` - [in] the LU factorization of matrix A (see at::_lu_with_info)
  * `pivots` - [in] the pivot indices (see at::_lu_with_info)

  For further details, please see the MAGMA documentation for magma_dgetrs_gpu.
*/
template <typename scalar_t>
static void apply_lu_solve_looped_magma(const Tensor& b, const Tensor& lu, const Tensor& pivots) {
#ifndef USE_MAGMA
  TORCH_CHECK(
      false,
      "Calling torch.lu_solve on a CUDA tensor requires compiling ",
      "PyTorch with MAGMA. lease rebuild with MAGMA.");
#else
  auto b_data = b.data_ptr<scalar_t>();
  auto lu_data = lu.data_ptr<scalar_t>();

  // MAGMA requires pivots to be a CPU tensor
  Tensor pivots_cpu = pivots.cpu();
  auto pivots_data = pivots_cpu.data_ptr<magma_int_t>();

  auto b_stride = matrixStride(b);
  auto lu_stride = matrixStride(lu);
  auto pivots_stride = pivots_cpu.size(-1);
  auto batch_size = batchCount(b);

  magma_int_t n = magma_int_cast(lu.size(-2), "n");
  magma_int_t nrhs = magma_int_cast(b.size(-1), "nrhs");
  auto leading_dimension = std::max<magma_int_t>(1, n);

  int info = 0;
  for (decltype(batch_size) i = 0; i < batch_size; i++) {
    scalar_t* b_working_ptr = &b_data[i * b_stride];
    scalar_t* lu_working_ptr = &lu_data[i * lu_stride];
    int* pivots_working_ptr = &pivots_data[i * pivots_stride];

    magmaLuSolve<scalar_t>(n, nrhs, lu_working_ptr, leading_dimension, pivots_working_ptr, b_working_ptr, leading_dimension, &info);

    // info from magmaLuSolve only reports if the i-th parameter is wrong
    // so we don't need to check it all the time
    TORCH_INTERNAL_ASSERT_DEBUG_ONLY(info == 0);
  }
#endif
}

/*
  Solves the matrix equation A X = B
  X and B are n-by-nrhs matrices, A is represented using the LU factorization.
  This is an in-place routine, content of `b` is overwritten.
  This is a specialized batched variant, it is expected to be faster than the "looped" version only for small inputs.

  Args:
  * `b` -  [in] the right hand side matrix B
           [out] the solution matrix X
  * `lu` - [in] the LU factorization of matrix A (see at::_lu_with_info)
  * `pivots` - [in] the pivot indices (see at::_lu_with_info)

  For further details, please see the MAGMA documentation for magma_dgetrs_batched.
*/
template <typename scalar_t>
static void apply_lu_solve_batched_magma(const Tensor& b, const Tensor& lu, const Tensor& pivots) {
#ifndef USE_MAGMA
  TORCH_CHECK(
      false,
      "Calling torch.lu_solve on a CUDA tensor requires compiling ",
      "PyTorch with MAGMA. lease rebuild with MAGMA.");
#else
  auto b_data = b.data_ptr<scalar_t>();
  auto lu_data = lu.data_ptr<scalar_t>();

  magma_int_t n = magma_int_cast(lu.size(-2), "n");
  magma_int_t nrhs = magma_int_cast(b.size(-1), "nrhs");
  auto leading_dimension = std::max<magma_int_t>(1, n);

  auto pivots_data = pivots.data_ptr<magma_int_t>();

  auto b_stride = matrixStride(b);
  auto lu_stride = matrixStride(lu);
  auto pivots_stride = pivots.size(-1);
  magma_int_t batch_size = magma_int_cast(batchCount(b), "batchCount");

  magma_int_t** pivots_array;
  scalar_t** lu_array;
  scalar_t** b_array;

  ALLOCATE_ARRAY(pivots_array, magma_int_t*, batch_size);
  ALLOCATE_ARRAY(lu_array, scalar_t*, batch_size);
  ALLOCATE_ARRAY(b_array, scalar_t*, batch_size);

  for (int64_t i = 0; i < batch_size; i++) {
    pivots_array[i] = &pivots_data[i * pivots_stride];
    b_array[i] = &b_data[i * b_stride];
    lu_array[i] = &lu_data[i * lu_stride];
  }

  MAGMAQueue magma_queue(b.get_device());

  // Compute the result in batches of 65535
  // that is the maximum allowed number for batch_size in MAGMA
  constexpr int64_t batch_limit = 65535;

  for (int64_t mini_idx = 0; mini_idx < batch_size; mini_idx += batch_limit) {
    int64_t nbatches = std::min(batch_limit, batch_size - mini_idx);
    scalar_t** lu_array_cur = &lu_array[mini_idx];
    scalar_t** b_array_cur = &b_array[mini_idx];
    magma_int_t** pivots_array_cur = &pivots_array[mini_idx];

    int info;
    magmaLuSolveBatched<scalar_t>(
        n, nrhs, lu_array_cur, leading_dimension,
        pivots_array_cur, b_array_cur, leading_dimension,
        info, nbatches, magma_queue);

    // info from magmaLuSolveBatched only reports if the i-th parameter is wrong
    // so we don't need to check it all the time
    TORCH_INTERNAL_ASSERT_DEBUG_ONLY(info == 0);
  }
#endif
}

static void lu_solve_batched_magma(const Tensor& b, const Tensor& lu, const Tensor& pivots) {
  AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(b.scalar_type(), "lu_solve_batched_magma", [&]{
    apply_lu_solve_batched_magma<scalar_t>(b, lu, pivots);
  });
}

static void lu_solve_looped_magma(const Tensor& b, const Tensor& lu, const Tensor& pivots) {
  AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(b.scalar_type(), "lu_solve_looped_magma", [&]{
    apply_lu_solve_looped_magma<scalar_t>(b, lu, pivots);
  });
}

static void lu_solve_dispatch(const Tensor& b, const Tensor& lu, const Tensor& pivots) {
  auto batch_size = batchCount(lu);
  auto m = lu.size(-2);
#ifdef USE_CUSOLVER
  if (batch_size == 1 && m > 512) {
    lu_solve_looped_cusolver(b, lu, pivots);
  }
#else
  if (batch_size == 1) {
    lu_solve_looped_magma(b, lu, pivots);
  }
#endif // ifdef USE_CUSOLVER
#ifdef CUDART_VERSION
  else if (batch_size > 2 && m <= 128) {
    lu_solve_batched_cublas(b, lu, pivots);
  }
#endif // ifdef CUDART_VERSION
  else {
    lu_solve_batched_magma(b, lu, pivots);
  }
}

REGISTER_DISPATCH(lu_solve_stub, &lu_solve_dispatch);

// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ lstsq ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

template <typename scalar_t>
static void apply_gels(const Tensor& a, Tensor& b, Tensor& infos) {
#ifndef USE_MAGMA
  TORCH_CHECK(false, "torch.linalg.lstsq: MAGMA library not found in "
    "compilation. Please rebuild with MAGMA.");
#else
  auto trans = MagmaNoTrans;
  auto m = magma_int_cast(a.size(-2), "m");
  auto n = magma_int_cast(a.size(-1), "n");

  TORCH_CHECK(
    m >= n,
    "torch.linalg.lstsq: only overdetermined systems (input.size(-2) >= input.size(-1)) are allowed on CUDA");

  auto nrhs = magma_int_cast(b.size(-1), "nrhs");
  auto ldda = std::max<magma_int_t>(1, m);
  auto lddb = std::max<magma_int_t>(1, std::max(m, n));
  auto nb = magmaGeqrfOptimalBlocksize<scalar_t>(m, n);
  auto lwork = (m - n + nb) * (nrhs + nb) + nrhs * nb;
  Tensor hwork = at::empty({static_cast<int64_t>(lwork)}, a.scalar_type());
  auto* hwork_ptr = hwork.data_ptr<scalar_t>();

  // MAGMA requires infos tensor to live on CPU
  infos = infos.to(at::kCPU);
  auto infos_data = infos.data_ptr<magma_int_t>();

  batch_iterator_with_broadcasting<scalar_t>(a, b,
    [&](scalar_t* a_working_ptr, scalar_t* b_working_ptr,
      int64_t a_linear_batch_idx) {
      magma_int_t* infos_working_ptr = &infos_data[a_linear_batch_idx];
      magmaGels<scalar_t>(trans, m, n, nrhs,
        a_working_ptr, ldda, b_working_ptr, lddb,
        hwork_ptr, lwork, infos_working_ptr);
    }
  );
#endif
}

void gels_magma(const Tensor& a, Tensor& b, Tensor& infos) {
  AT_DISPATCH_FLOATING_AND_COMPLEX_TYPES(a.scalar_type(), "gels_magma", [&] {
    apply_gels<scalar_t>(a, b, infos);
  });
}

void linalg_lstsq_gels(const Tensor& A, const Tensor& B, const Tensor& infos) {
  // The steps for using the QR decomposition for solving least squares problems
  // are outlined here https://en.wikipedia.org/wiki/QR_decomposition#Using_for_solution_to_linear_inverse_problems
  auto m = A.size(-2);
  auto n = A.size(-1);
  auto mn = std::min(m, n);

  // explicitly broadcast the batch dimensions of A
  // TODO: revisit this later to use batch_iterator_with_broadcasting in triangular_solve
  IntArrayRef A_batch_sizes(A.sizes().data(), A.dim() - 2);
  IntArrayRef B_batch_sizes(B.sizes().data(), B.dim() - 2);
  std::vector<int64_t> expand_batch_portion = at::infer_size(A_batch_sizes, B_batch_sizes);

  auto tau_shape = A.sizes().vec();
  tau_shape.pop_back();
  tau_shape.back() = mn;
  Tensor tau = at::empty(tau_shape, A.options());

  if (m >= n) {
    // Step 1: compute QR factorization using geqrf
    geqrf_kernel(A, tau);

    // explicitly broadcast the batch dimensions of A
    // we do it after geqrf so that we don't do redundant computations for the same input
    auto A_expand_batch = expand_batch_portion;
    A_expand_batch.insert(A_expand_batch.end(), {A.size(-2), A.size(-1)});
    Tensor A_expanded = A.expand({A_expand_batch});
    bool is_fortran_contiguous = A_expanded.transpose(-2, -1).is_contiguous();
    Tensor A_broadcasted = is_fortran_contiguous ? A_expanded : cloneBatchedColumnMajor(A_expanded);
    auto tau_expand_batch = expand_batch_portion;
    tau_expand_batch.push_back(tau.size(-1));
    Tensor tau_broadcasted = tau.expand({tau_expand_batch}).contiguous();

    // Step 2: B <- Q^H B
    ormqr_kernel(A_broadcasted, tau_broadcasted, B, /*left=*/true, /*transpose=*/true);

    // Step 3: solve R X = B
    bool upper = true;
    bool transpose = false;
    bool conjugate_transpose = false;
    bool unitriangular = false;
    triangular_solve_kernel(
        const_cast<Tensor&>(A_broadcasted),
        const_cast<Tensor&>(B),
        const_cast<Tensor&>(infos),
        upper, transpose, conjugate_transpose, unitriangular);
  } else { // underdetermined case
    Tensor Ah = cloneBatchedColumnMajor(A.conj().transpose(-2, -1));

    // Step 1: compute QR factorization of conjugate transpose of A using geqrf
    geqrf_kernel(Ah, tau);

    // explicitly broadcast the batch dimensions of A
    // we do it after geqrf so that we don't do redundant computations for the same input
    auto A_expand_batch = expand_batch_portion;
    A_expand_batch.insert(A_expand_batch.end(), {Ah.size(-2), Ah.size(-1)});
    Tensor Ah_expanded = Ah.expand({A_expand_batch});
    bool is_fortran_contiguous = Ah_expanded.transpose(-2, -1).is_contiguous();
    Tensor Ah_broadcasted = is_fortran_contiguous ? Ah_expanded : cloneBatchedColumnMajor(Ah_expanded);

    // Step 2: R^H Z = B
    bool upper = true;
    bool transpose = true;
    bool conjugate_transpose = true;
    bool unitriangular = false;
    triangular_solve_kernel(
        const_cast<Tensor&>(Ah_broadcasted),
        const_cast<Tensor&>(B),
        const_cast<Tensor&>(infos),
        upper, transpose, conjugate_transpose, unitriangular);

    // B matrix has the size max(m, n) x nrhs
    // triangular_solve_kernel writes its output into the first m rows of B leaving the rest untouched
    // we need to set the rest of the rows to zero so that the multiplication from step 3 is correct
    B.narrow(-2, m, n - m).zero_();

    auto tau_expand_batch = expand_batch_portion;
    tau_expand_batch.push_back(tau.size(-1));
    Tensor tau_broadcasted = tau.expand({tau_expand_batch}).contiguous();

    // Step 3: X <- Q Z
    ormqr_kernel(Ah_broadcasted, tau_broadcasted, B, /*left=*/true, /*transpose=*/false);
  }
}

void gels_looped(const Tensor& a, Tensor& b, Tensor& infos) {
#if defined(USE_CUSOLVER)
  // linalg_lstsq_gels is a generic function that is implemented using
  // geqrf_stub, ormqr_stub, and triangular_solve_stub
  // It dispatches to cuSOLVER for CUDA inputs if USE_CUSOLVER is defined
  return linalg_lstsq_gels(a, b, infos);
#else
  return gels_magma(a, b, infos);
#endif
}

void lstsq_kernel(const Tensor& a, Tensor& b, Tensor& /*rank*/, Tensor& /*singular_values*/, Tensor& infos, double /*rcond*/, std::string /*driver_name*/)  {
  auto m = a.size(-2);
  auto n = a.size(-1);

  // first handle the underdetermined case (m < n)
  // this case is not supported by MAGMA or cuBLAS
  if (m < n) {
#if defined(USE_CUSOLVER)
    linalg_lstsq_gels(a, b, infos);
#else
    TORCH_CHECK(
        false,
        "torch.linalg.lstsq: only overdetermined systems (input.size(-2) >= input.size(-1)) are allowed on CUDA. ",
        "Please rebuild with cuSOLVER.");
#endif
  } else { // m >= n
#ifndef USE_MAGMA
    // MAGMA is not available we can either use cuBLAS or cuSOLVER here
    // the batched vs looped dispatch is implemented based on the following performance results
    // https://github.com/pytorch/pytorch/pull/54725#issuecomment-832234456
    if (m <= 256 && batchCount(b) >= std::max<int64_t>(2, m / 16)) {
      // if CUDART_VERSION is defined then cuBLAS is available
      #ifdef CUDART_VERSION
      gels_batched_cublas(a, b, infos);
      #else
      // this would either call cuSOLVER or MAGMA,
      // if MAGMA is called a runtime error is thrown about not finding MAGMA in compilation
      gels_looped(a, b, infos);
      #endif // CUDART_VERSION
    } else {
      gels_looped(a, b, infos);
    }
#else
    // if both MAGMA and cuSOLVER are available this would call cuSOLVER
    // MAGMA is called if cuSOLVER is not available
    gels_looped(a, b, infos);
#endif // USE_MAGMA
  }
}

REGISTER_DISPATCH(lstsq_stub, &lstsq_kernel);

}}  // namespace at::native

#undef ALLOCATE_ARRAY