derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
1034 lines
38 KiB
C++
1034 lines
38 KiB
C++
/* **************************************************************************
|
|
* Copyright (C) 2021-2026 Advanced Micro Devices, Inc. All rights reserved.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions
|
|
* are met:
|
|
*
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer.
|
|
*
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
|
* notice, this list of conditions and the following disclaimer in the
|
|
* documentation and/or other materials provided with the distribution.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
|
|
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
|
|
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
|
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
|
|
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
|
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
|
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
|
|
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
|
* SUCH DAMAGE.
|
|
* *************************************************************************/
|
|
|
|
#pragma once
|
|
|
|
#include "common/matrix_utils/matrix_utils.hpp"
|
|
#include "common/misc/client_util.hpp"
|
|
#include "common/misc/clientcommon.hpp"
|
|
#include "common/misc/lapack_host_reference.hpp"
|
|
#include "common/misc/norm.hpp"
|
|
#include "common/misc/rocsolver.hpp"
|
|
#include "common/misc/rocsolver_arguments.hpp"
|
|
#include "common/misc/rocsolver_test.hpp"
|
|
#include "common/misc/rocsolver_timer.hpp"
|
|
|
|
template <bool STRIDED, typename I, typename T, typename S, typename U>
|
|
void syevd_heevd_checkBadArgs(const rocblas_handle handle,
|
|
const rocblas_evect evect,
|
|
const rocblas_fill uplo,
|
|
const I n,
|
|
T dA,
|
|
const I lda,
|
|
const rocblas_stride stA,
|
|
S dD,
|
|
const rocblas_stride stD,
|
|
S dE,
|
|
const rocblas_stride stE,
|
|
U dinfo,
|
|
const I bc)
|
|
{
|
|
// handle
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, nullptr, evect, uplo, n, dA, lda, stA, dD,
|
|
stD, dE, stE, dinfo, bc),
|
|
rocblas_status_invalid_handle);
|
|
|
|
// values
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, handle, rocblas_evect(0), uplo, n, dA, lda,
|
|
stA, dD, stD, dE, stE, dinfo, bc),
|
|
rocblas_status_invalid_value);
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, handle, evect, rocblas_fill_full, n, dA,
|
|
lda, stA, dD, stD, dE, stE, dinfo, bc),
|
|
rocblas_status_invalid_value);
|
|
|
|
// sizes (only check batch_count if applicable)
|
|
if(STRIDED)
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n, dA, lda, stA,
|
|
dD, stD, dE, stE, dinfo, I(-1)),
|
|
rocblas_status_invalid_size);
|
|
|
|
// pointers
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n, (T) nullptr, lda,
|
|
stA, dD, stD, dE, stE, dinfo, bc),
|
|
rocblas_status_invalid_pointer);
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n, dA, lda, stA,
|
|
(S) nullptr, stD, dE, stE, dinfo, bc),
|
|
rocblas_status_invalid_pointer);
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n, dA, lda, stA, dD,
|
|
stD, (S) nullptr, stE, dinfo, bc),
|
|
rocblas_status_invalid_pointer);
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n, dA, lda, stA, dD,
|
|
stD, dE, stE, (U) nullptr, bc),
|
|
rocblas_status_invalid_pointer);
|
|
|
|
// quick return with invalid pointers
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, I(0), (T) nullptr, lda,
|
|
stA, (S) nullptr, stD, (S) nullptr, stE, dinfo, bc),
|
|
rocblas_status_success);
|
|
|
|
// quick return with zero batch_count if applicable
|
|
if(STRIDED)
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n, dA, lda, stA,
|
|
dD, stD, dE, stE, (U) nullptr, I(0)),
|
|
rocblas_status_success);
|
|
}
|
|
|
|
template <bool BATCHED, bool STRIDED, typename T, typename I>
|
|
void testing_syevd_heevd_bad_arg()
|
|
{
|
|
using S = decltype(std::real(T{}));
|
|
|
|
// safe arguments
|
|
rocblas_local_handle handle;
|
|
rocblas_evect evect = rocblas_evect_none;
|
|
rocblas_fill uplo = rocblas_fill_lower;
|
|
I n = 1;
|
|
I lda = 1;
|
|
rocblas_stride stA = 1;
|
|
rocblas_stride stD = 1;
|
|
rocblas_stride stE = 1;
|
|
I bc = 1;
|
|
|
|
#ifndef ROCSOLVER_ENABLE_EIGENSOLVERS_64
|
|
// 64-bit API disabled: entry points must report not_implemented (see note in testing_syevd_heevd).
|
|
if constexpr(std::is_same<I, int64_t>::value)
|
|
{
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n, (T*)nullptr,
|
|
lda, stA, (S*)nullptr, stD, (S*)nullptr, stE,
|
|
(I*)nullptr, bc),
|
|
rocblas_status_not_implemented);
|
|
return;
|
|
}
|
|
#endif
|
|
|
|
if(BATCHED)
|
|
{
|
|
// memory allocations
|
|
device_batch_vector<T> dA(1, 1, 1);
|
|
device_strided_batch_vector<S> dD(1, 1, 1, 1);
|
|
device_strided_batch_vector<S> dE(1, 1, 1, 1);
|
|
device_strided_batch_vector<rocblas_int> dinfo(1, 1, 1, 1);
|
|
CHECK_HIP_ERROR(dA.memcheck());
|
|
CHECK_HIP_ERROR(dD.memcheck());
|
|
CHECK_HIP_ERROR(dE.memcheck());
|
|
CHECK_HIP_ERROR(dinfo.memcheck());
|
|
|
|
// check bad arguments
|
|
syevd_heevd_checkBadArgs<STRIDED, I>(handle, evect, uplo, n, dA.data(), lda, stA, dD.data(),
|
|
stD, dE.data(), stE, dinfo.data(), bc);
|
|
}
|
|
else
|
|
{
|
|
// memory allocations
|
|
device_strided_batch_vector<T> dA(1, 1, 1, 1);
|
|
device_strided_batch_vector<S> dD(1, 1, 1, 1);
|
|
device_strided_batch_vector<S> dE(1, 1, 1, 1);
|
|
device_strided_batch_vector<rocblas_int> dinfo(1, 1, 1, 1);
|
|
CHECK_HIP_ERROR(dA.memcheck());
|
|
CHECK_HIP_ERROR(dD.memcheck());
|
|
CHECK_HIP_ERROR(dE.memcheck());
|
|
CHECK_HIP_ERROR(dinfo.memcheck());
|
|
|
|
// check bad arguments
|
|
syevd_heevd_checkBadArgs<STRIDED, I>(handle, evect, uplo, n, dA.data(), lda, stA, dD.data(),
|
|
stD, dE.data(), stE, dinfo.data(), bc);
|
|
}
|
|
}
|
|
|
|
template <bool CPU, bool GPU, typename T, typename I, typename Td, typename Th>
|
|
void syevd_heevd_default_initData(const rocblas_handle handle,
|
|
const rocblas_evect evect,
|
|
const I n,
|
|
Td& dA,
|
|
const I lda,
|
|
const I bc,
|
|
Th& hA,
|
|
std::vector<T>& A,
|
|
bool test = true)
|
|
{
|
|
if(CPU)
|
|
{
|
|
rocblas_init<T>(hA, true);
|
|
|
|
// scale A to avoid singularities
|
|
for(I b = 0; b < bc; ++b)
|
|
{
|
|
for(I i = 0; i < n; i++)
|
|
{
|
|
for(I j = i; j < n; j++)
|
|
{
|
|
if(i == j)
|
|
hA[b][i + j * lda] = std::real(hA[b][i + j * lda]) + 400;
|
|
else
|
|
{
|
|
hA[b][i + j * lda] -= 4;
|
|
hA[b][j + i * lda] = sconj(hA[b][i + j * lda]);
|
|
}
|
|
}
|
|
}
|
|
|
|
// make copy of original data to test vectors if required
|
|
if(test && evect == rocblas_evect_original)
|
|
{
|
|
for(I i = 0; i < n; i++)
|
|
{
|
|
for(I j = 0; j < n; j++)
|
|
A[b * lda * n + i + j * lda] = hA[b][i + j * lda];
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
if(GPU)
|
|
{
|
|
// now copy to the GPU
|
|
CHECK_HIP_ERROR(dA.transfer_from(hA));
|
|
}
|
|
}
|
|
|
|
// Creates an `n` by `n` matrix `A` with the following eigenvalues:
|
|
//
|
|
// spectrum(A) = { l_i = ulp * i, for 1 <= i <= n - 1; l_n = 1 }
|
|
//
|
|
// where `ulp` is the smallest floating point number such that `1 + ulp > 1`.
|
|
//
|
|
template <bool CPU, bool GPU, typename T, typename I, typename Td, typename Th>
|
|
void syevd_heevd_eig7_initData(const rocblas_handle handle,
|
|
const rocblas_evect evect,
|
|
const I n,
|
|
Td& dA,
|
|
const I lda,
|
|
const I bc,
|
|
Th& hA,
|
|
std::vector<T>& A,
|
|
bool test = true)
|
|
{
|
|
using S = decltype(std::real(T{}));
|
|
|
|
if(CPU)
|
|
{
|
|
rocblas_init<T>(hA, true);
|
|
|
|
for(I b = 0; b < bc; ++b)
|
|
{
|
|
// New matrix initialization
|
|
using HMat = HostMatrix<T, I>;
|
|
using BDesc = typename HMat::BlockDescriptor;
|
|
|
|
auto hAw = HMat::Wrap(hA[b], lda, n);
|
|
if(hAw) // update matrix hA if n >= 1
|
|
{
|
|
auto eigs = std::numeric_limits<S>::epsilon() * HMat::FromRange(1, n - 1, n - 1);
|
|
eigs = cat(eigs, HMat::Ones(1, 1));
|
|
auto [Q, _] = qr((*hAw).block(BDesc().nrows(n).ncols(n)));
|
|
hAw->set_to_zero();
|
|
|
|
hAw->copy_data_from(Q * HMat::Zeros(n).diag(eigs) * adjoint(Q));
|
|
}
|
|
|
|
// make copy of original data to test vectors if required
|
|
if(test && evect == rocblas_evect_original)
|
|
{
|
|
for(I i = 0; i < n; i++)
|
|
{
|
|
for(I j = 0; j < n; j++)
|
|
A[b * lda * n + i + j * lda] = hA[b][i + j * lda];
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
if(GPU)
|
|
{
|
|
// now copy to the GPU
|
|
CHECK_HIP_ERROR(dA.transfer_from(hA));
|
|
}
|
|
}
|
|
|
|
// Creates an `n` by `n` tridiagonal, Wilkinson matrix, which is formed as follows:
|
|
//
|
|
// 1. If `n` is even:
|
|
// ( 1 1 1 1 )
|
|
// W_{2m + 1} = tridiag ( m (m - 1) ... 0.5 0.5 ... (m - 1) m )
|
|
// ( 1 1 1 1 )
|
|
//
|
|
// 2. If `n` is odd:
|
|
// ( 1 1 1 1 )
|
|
// W_{2m + 1} = tridiag ( m (m - 1) ... 1 0 1 ... (m - 1) m )
|
|
// ( 1 1 1 1 )
|
|
//
|
|
// where `n = 2m + 1`.
|
|
//
|
|
template <bool CPU, bool GPU, typename T, typename I, typename Td, typename Th>
|
|
void syevd_heevd_wilkinson_initData(const rocblas_handle handle,
|
|
const rocblas_evect evect,
|
|
const I n,
|
|
Td& dA,
|
|
const I lda,
|
|
const I bc,
|
|
Th& hA,
|
|
std::vector<T>& A,
|
|
bool test = true)
|
|
{
|
|
using S = decltype(std::real(T{}));
|
|
|
|
if(CPU)
|
|
{
|
|
rocblas_init<T>(hA, true);
|
|
|
|
// scale A to avoid singularities
|
|
for(I b = 0; b < bc; ++b)
|
|
{
|
|
// New matrix initialization
|
|
using HMat = HostMatrix<T, I>;
|
|
using BDesc = typename HMat::BlockDescriptor;
|
|
|
|
auto hAw = HMat::Wrap(hA[b], lda, n);
|
|
if(hAw) // update matrix hA if n >= 1
|
|
{
|
|
S m = (n - 1) / S(2);
|
|
auto A = HMat::Zeros(n, n);
|
|
auto E = HMat::Ones(n - 1, 1);
|
|
auto D = HMat::Zeros(n, 1);
|
|
|
|
for(I i = 0; i < n / 2; ++i)
|
|
{
|
|
D[i] = m - i;
|
|
D[n - 1 - i] = m - i;
|
|
}
|
|
|
|
A.diag(D);
|
|
A.sup_diag(E);
|
|
A.sub_diag(E);
|
|
|
|
hAw->set_to_zero();
|
|
hAw->copy_data_from(A);
|
|
}
|
|
|
|
// make copy of original data to test vectors if required
|
|
if(test && evect == rocblas_evect_original)
|
|
{
|
|
for(I i = 0; i < n; i++)
|
|
{
|
|
for(I j = 0; j < n; j++)
|
|
A[b * lda * n + i + j * lda] = hA[b][i + j * lda];
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
if(GPU)
|
|
{
|
|
// now copy to the GPU
|
|
CHECK_HIP_ERROR(dA.transfer_from(hA));
|
|
}
|
|
}
|
|
|
|
// Creates an `n` by `n` tridiagonal, Toeplitz matrix T of the following form:
|
|
//
|
|
// ( 1 1 1 )
|
|
// T = tridiag ( 2 2 ... 2 2 ... 2 2 )
|
|
// ( 1 1 1 )
|
|
//
|
|
template <bool CPU, bool GPU, typename T, typename I, typename Td, typename Th>
|
|
void syevd_heevd_toeplitz_initData(const rocblas_handle handle,
|
|
const rocblas_evect evect,
|
|
const I n,
|
|
Td& dA,
|
|
const I lda,
|
|
const I bc,
|
|
Th& hA,
|
|
std::vector<T>& A,
|
|
bool test = true)
|
|
{
|
|
using S = decltype(std::real(T{}));
|
|
|
|
if(CPU)
|
|
{
|
|
rocblas_init<T>(hA, true);
|
|
|
|
// scale A to avoid singularities
|
|
for(I b = 0; b < bc; ++b)
|
|
{
|
|
// New matrix initialization
|
|
using HMat = HostMatrix<T, I>;
|
|
using BDesc = typename HMat::BlockDescriptor;
|
|
|
|
auto hAw = HMat::Wrap(hA[b], lda, n);
|
|
if(hAw) // update matrix hA if n >= 1
|
|
{
|
|
auto A = HMat::Zeros(n, n);
|
|
auto E = HMat::Ones(n - 1, 1);
|
|
auto D = 2 * HMat::Ones(n, 1);
|
|
|
|
A.diag(D);
|
|
A.sup_diag(E);
|
|
A.sub_diag(E);
|
|
|
|
hAw->set_to_zero();
|
|
hAw->copy_data_from(A);
|
|
}
|
|
|
|
// make copy of original data to test vectors if required
|
|
if(test && evect == rocblas_evect_original)
|
|
{
|
|
for(I i = 0; i < n; i++)
|
|
{
|
|
for(I j = 0; j < n; j++)
|
|
A[b * lda * n + i + j * lda] = hA[b][i + j * lda];
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
if(GPU)
|
|
{
|
|
// now copy to the GPU
|
|
CHECK_HIP_ERROR(dA.transfer_from(hA));
|
|
}
|
|
}
|
|
|
|
// For `n > 1`, creates a symmetrized `n` by `n` tridiagonal, Clement matrix T of the following form:
|
|
//
|
|
// ( sqrt(n - 1) sqrt(2(n - 2)) sqrt((n - 2)2) sqrt(n - 1) )
|
|
// T = tridiag ( 0 0 ... 0 0 )
|
|
// ( sqrt(n - 1) sqrt(2(n - 2)) sqrt((n - 2)2) sqrt(n - 1) )
|
|
//
|
|
// were the `i-th` off-diagonal entry is sqrt(i(n - i)), 1 <= i < n.
|
|
//
|
|
template <bool CPU, bool GPU, typename T, typename I, typename Td, typename Th>
|
|
void syevd_heevd_clement_initData(const rocblas_handle handle,
|
|
const rocblas_evect evect,
|
|
const I n,
|
|
Td& dA,
|
|
const I lda,
|
|
const I bc,
|
|
Th& hA,
|
|
std::vector<T>& A,
|
|
bool test = true)
|
|
{
|
|
using S = decltype(std::real(T{}));
|
|
|
|
if(CPU)
|
|
{
|
|
rocblas_init<T>(hA, true);
|
|
|
|
// scale A to avoid singularities
|
|
for(I b = 0; b < bc; ++b)
|
|
{
|
|
// New matrix initialization
|
|
using HMat = HostMatrix<T, I>;
|
|
using BDesc = typename HMat::BlockDescriptor;
|
|
|
|
auto hAw = HMat::Wrap(hA[b], lda, n);
|
|
if(hAw) // update matrix hA if n >= 1
|
|
{
|
|
auto A = HMat::Zeros(n, n);
|
|
auto E = HMat::Ones(n - 1, 1);
|
|
auto D = HMat::Zeros(n, 1);
|
|
|
|
for(I i = 1; i < n; ++i)
|
|
{
|
|
E[i - 1] = std::sqrt(i * (n - i));
|
|
}
|
|
|
|
A.diag(D);
|
|
A.sup_diag(E);
|
|
A.sub_diag(E);
|
|
|
|
hAw->set_to_zero();
|
|
hAw->copy_data_from(A);
|
|
}
|
|
|
|
// make copy of original data to test vectors if required
|
|
if(test && evect == rocblas_evect_original)
|
|
{
|
|
for(I i = 0; i < n; i++)
|
|
{
|
|
for(I j = 0; j < n; j++)
|
|
A[b * lda * n + i + j * lda] = hA[b][i + j * lda];
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
if(GPU)
|
|
{
|
|
// now copy to the GPU
|
|
CHECK_HIP_ERROR(dA.transfer_from(hA));
|
|
}
|
|
}
|
|
|
|
template <bool CPU, bool GPU, typename T, typename I, typename Td, typename Th>
|
|
void syevd_heevd_initData(const rocblas_handle handle,
|
|
const rocblas_evect evect,
|
|
const I n,
|
|
Td& dA,
|
|
const I lda,
|
|
const I bc,
|
|
Th& hA,
|
|
std::vector<T>& A,
|
|
bool test = true)
|
|
{
|
|
if((std::getenv("TEST_EIG7") != nullptr) || (std::getenv("SYEVD_TEST_EIG7") != nullptr))
|
|
{
|
|
syevd_heevd_eig7_initData<CPU, GPU, T, I>(handle, evect, n, dA, lda, bc, hA, A, test);
|
|
}
|
|
else if((std::getenv("TEST_WILKINSON") != nullptr)
|
|
|| (std::getenv("SYEVD_TEST_WILKINSON") != nullptr))
|
|
{
|
|
syevd_heevd_wilkinson_initData<CPU, GPU, T, I>(handle, evect, n, dA, lda, bc, hA, A, test);
|
|
}
|
|
else if((std::getenv("TEST_CLEMENT") != nullptr)
|
|
|| (std::getenv("SYEVD_TEST_CLEMENT") != nullptr))
|
|
{
|
|
syevd_heevd_clement_initData<CPU, GPU, T, I>(handle, evect, n, dA, lda, bc, hA, A, test);
|
|
}
|
|
else if((std::getenv("TEST_TOEPLITZ") != nullptr)
|
|
|| (std::getenv("SYEVD_TEST_TOEPLITZ") != nullptr))
|
|
{
|
|
syevd_heevd_toeplitz_initData<CPU, GPU, T, I>(handle, evect, n, dA, lda, bc, hA, A, test);
|
|
}
|
|
else
|
|
{
|
|
syevd_heevd_default_initData<CPU, GPU, T, I>(handle, evect, n, dA, lda, bc, hA, A, test);
|
|
}
|
|
|
|
return;
|
|
}
|
|
|
|
template <bool STRIDED, typename T, typename I, typename Sd, typename Td, typename Id, typename Sh, typename Th, typename Ih>
|
|
void syevd_heevd_getError(const rocblas_handle handle,
|
|
const rocblas_evect evect,
|
|
const rocblas_fill uplo,
|
|
const I n,
|
|
Td& dA,
|
|
const I lda,
|
|
const rocblas_stride stA,
|
|
Sd& dD,
|
|
const rocblas_stride stD,
|
|
Sd& dE,
|
|
const rocblas_stride stE,
|
|
Id& dinfo,
|
|
const I bc,
|
|
Th& hA,
|
|
Th& hAres,
|
|
Sh& hD,
|
|
Sh& hDres,
|
|
Ih& hinfo,
|
|
Ih& hinfoRes,
|
|
double* max_err,
|
|
double* max_errv)
|
|
{
|
|
constexpr bool COMPLEX = rocblas_is_complex<T>;
|
|
using S = decltype(std::real(T{}));
|
|
|
|
using HMat = HostMatrix<T, I>;
|
|
using BDesc = typename HMat::BlockDescriptor;
|
|
|
|
int lgn = floor(log(n - 1) / log(2)) + 1;
|
|
int sizeE, lwork;
|
|
if(!COMPLEX)
|
|
{
|
|
sizeE = (evect == rocblas_evect_none ? 2 * n + 1 : 1 + 6 * n + 2 * n * n);
|
|
lwork = 0;
|
|
}
|
|
else
|
|
{
|
|
sizeE = (evect == rocblas_evect_none ? n : 1 + 5 * n + 2 * n * n);
|
|
lwork = (evect == rocblas_evect_none ? n + 1 : 2 * n + n * n);
|
|
}
|
|
int liwork = (evect == rocblas_evect_none ? 1 : 3 + 5 * n);
|
|
|
|
std::vector<T> work(lwork);
|
|
std::vector<S> hE(sizeE);
|
|
std::vector<int> iwork(liwork);
|
|
std::vector<T> A(lda * n * bc);
|
|
|
|
// input data initialization
|
|
syevd_heevd_initData<true, true, T, I>(handle, evect, n, dA, lda, bc, hA, A);
|
|
|
|
// execute computations
|
|
// GPU lapack
|
|
CHECK_ROCBLAS_ERROR(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n, dA.data(), lda, stA,
|
|
dD.data(), stD, dE.data(), stE, dinfo.data(), bc));
|
|
|
|
CHECK_HIP_ERROR(hDres.transfer_from(dD));
|
|
CHECK_HIP_ERROR(hinfoRes.transfer_from(dinfo));
|
|
if(evect == rocblas_evect_original)
|
|
CHECK_HIP_ERROR(hAres.transfer_from(dA));
|
|
|
|
// CPU lapack
|
|
for(I b = 0; b < bc; ++b)
|
|
{
|
|
rocblas_int info_tmp = 0;
|
|
cpu_syevd_heevd(evect, uplo, n, hA[b], lda, hD[b], work.data(), lwork, hE.data(), sizeE,
|
|
iwork.data(), liwork, &info_tmp);
|
|
hinfo[b][0] = info_tmp;
|
|
}
|
|
|
|
// Check info for non-convergence
|
|
*max_err = 0;
|
|
for(I b = 0; b < bc; ++b)
|
|
{
|
|
EXPECT_EQ(hinfo[b][0], hinfoRes[b][0]) << "where b = " << b;
|
|
if(hinfo[b][0] != hinfoRes[b][0])
|
|
*max_err += 1;
|
|
}
|
|
|
|
// (We expect the used input matrices to always converge. Testing
|
|
// implicitly the equivalent non-converged matrix is very complicated and it boils
|
|
// down to essentially run the algorithm again and until convergence is achieved).
|
|
|
|
double err = 0;
|
|
|
|
for(I b = 0; b < bc; ++b)
|
|
{
|
|
if(evect != rocblas_evect_original)
|
|
{
|
|
// only eigenvalues needed; compare with LAPACK
|
|
|
|
// error is ||hD - hDRes|| / ||hD||
|
|
// using frobenius norm
|
|
if(hinfo[b][0] == 0)
|
|
err = norm_error('F', 1, n, 1, hD[b], hDres[b]);
|
|
*max_err = err > *max_err ? err : *max_err;
|
|
}
|
|
else
|
|
{
|
|
// both eigenvalues and eigenvectors needed; compare with input
|
|
// matrix
|
|
if((hinfo[b][0] == 0) && (n > 0))
|
|
{
|
|
// Input matrix
|
|
auto M = HMat::Wrap(A.data() + b * lda * n, lda, n)->block(BDesc().nrows(n).ncols(n));
|
|
|
|
// Computed eigenvectors
|
|
auto U = HMat::Wrap(hAres[b], lda, n)->block(BDesc().nrows(n).ncols(n));
|
|
// Computed eigenvalues
|
|
auto d = HMat::Convert(hDres[b], 1, n)->block(BDesc().nrows(1).ncols(n));
|
|
// Diagonal matrix of size n by n with computed eigenvalues
|
|
auto D = HMat::Zeros(n, n).diag(d);
|
|
|
|
// Orthogonal error
|
|
auto OE = U * adjoint(U) - HMat::Eye(n, n);
|
|
err = OE.max_col_norm();
|
|
*max_errv = err > *max_err ? err : *max_err;
|
|
|
|
// Residual error
|
|
auto RE = M - U * D * adjoint(U);
|
|
err = RE.norm() / M.norm();
|
|
*max_err = err > *max_err ? err : *max_err;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
template <bool STRIDED, typename T, typename I, typename Sd, typename Td, typename Id, typename Sh, typename Th, typename Ih>
|
|
void syevd_heevd_getPerfData(const rocblas_handle handle,
|
|
const rocblas_evect evect,
|
|
const rocblas_fill uplo,
|
|
const I n,
|
|
Td& dA,
|
|
const I lda,
|
|
const rocblas_stride stA,
|
|
Sd& dD,
|
|
const rocblas_stride stD,
|
|
Sd& dE,
|
|
const rocblas_stride stE,
|
|
Id& dinfo,
|
|
const I bc,
|
|
Th& hA,
|
|
Sh& hD,
|
|
Ih& hinfo,
|
|
double* gpu_time_used,
|
|
double* cpu_time_used,
|
|
const rocblas_int hot_calls,
|
|
const int profile,
|
|
const bool profile_kernels,
|
|
const bool perf)
|
|
{
|
|
constexpr bool COMPLEX = rocblas_is_complex<T>;
|
|
using S = decltype(std::real(T{}));
|
|
|
|
int sizeE, lwork;
|
|
if(!COMPLEX)
|
|
{
|
|
sizeE = (evect == rocblas_evect_none ? 2 * n + 1 : 1 + 6 * n + 2 * n * n);
|
|
lwork = 0;
|
|
}
|
|
else
|
|
{
|
|
sizeE = (evect == rocblas_evect_none ? n : 1 + 5 * n + 2 * n * n);
|
|
lwork = (evect == rocblas_evect_none ? n + 1 : 2 * n + n * n);
|
|
}
|
|
int liwork = (evect == rocblas_evect_none ? 1 : 3 + 5 * n);
|
|
|
|
std::vector<T> work(lwork);
|
|
std::vector<S> hE(sizeE);
|
|
std::vector<int> iwork(liwork);
|
|
std::vector<T> A;
|
|
|
|
if(!perf)
|
|
{
|
|
syevd_heevd_initData<true, false, T, I>(handle, evect, n, dA, lda, bc, hA, A, 0);
|
|
|
|
// cpu-lapack performance (only if not in perf mode)
|
|
*cpu_time_used = get_time_us_no_sync();
|
|
for(I b = 0; b < bc; ++b)
|
|
{
|
|
rocblas_int info_tmp = 0;
|
|
cpu_syevd_heevd(evect, uplo, n, hA[b], lda, hD[b], work.data(), lwork, hE.data(), sizeE,
|
|
iwork.data(), liwork, &info_tmp);
|
|
hinfo[b][0] = info_tmp;
|
|
}
|
|
*cpu_time_used = get_time_us_no_sync() - *cpu_time_used;
|
|
}
|
|
|
|
syevd_heevd_initData<true, false, T, I>(handle, evect, n, dA, lda, bc, hA, A, 0);
|
|
|
|
// cold calls
|
|
for(int iter = 0; iter < 2; iter++)
|
|
{
|
|
syevd_heevd_initData<false, true, T, I>(handle, evect, n, dA, lda, bc, hA, A, 0);
|
|
|
|
CHECK_ROCBLAS_ERROR(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n, dA.data(), lda, stA,
|
|
dD.data(), stD, dE.data(), stE, dinfo.data(), bc));
|
|
}
|
|
|
|
// gpu-lapack performance
|
|
hipStream_t stream;
|
|
CHECK_ROCBLAS_ERROR(rocblas_get_stream(handle, &stream));
|
|
rocsolver_timer timer;
|
|
|
|
if(profile > 0)
|
|
{
|
|
if(profile_kernels)
|
|
rocsolver_log_set_layer_mode(rocblas_layer_mode_log_profile
|
|
| rocblas_layer_mode_ex_log_kernel);
|
|
else
|
|
rocsolver_log_set_layer_mode(rocblas_layer_mode_log_profile);
|
|
rocsolver_log_set_max_levels(profile);
|
|
}
|
|
|
|
for(rocblas_int iter = 0; iter < hot_calls; iter++)
|
|
{
|
|
syevd_heevd_initData<false, true, T, I>(handle, evect, n, dA, lda, bc, hA, A, 0);
|
|
|
|
timer.start(stream);
|
|
rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n, dA.data(), lda, stA, dD.data(), stD,
|
|
dE.data(), stE, dinfo.data(), bc);
|
|
timer.end(stream);
|
|
}
|
|
*gpu_time_used = timer.get_combined();
|
|
}
|
|
|
|
template <bool BATCHED, bool STRIDED, typename T, typename I>
|
|
void testing_syevd_heevd(Arguments& argus)
|
|
{
|
|
using S = decltype(std::real(T{}));
|
|
|
|
// get arguments
|
|
rocblas_local_handle handle;
|
|
char evectC = argus.get<char>("evect");
|
|
char uploC = argus.get<char>("uplo");
|
|
rocblas_int n = argus.get<rocblas_int>("n");
|
|
rocblas_int lda = argus.get<rocblas_int>("lda", n);
|
|
rocblas_stride stA = argus.get<rocblas_stride>("strideA", lda * n);
|
|
rocblas_stride stD = argus.get<rocblas_stride>("strideD", n);
|
|
rocblas_stride stE = argus.get<rocblas_stride>("strideE", n);
|
|
|
|
rocblas_evect evect = char2rocblas_evect(evectC);
|
|
rocblas_fill uplo = char2rocblas_fill(uploC);
|
|
I bc = argus.batch_count;
|
|
rocblas_int hot_calls = argus.iters;
|
|
|
|
#ifndef ROCSOLVER_ENABLE_EIGENSOLVERS_64
|
|
// The 64-bit eigensolver API is gated behind the ROCSOLVER_ENABLE_EIGENSOLVERS_64
|
|
// feature flag; when the flag is off the _64 entry points must report
|
|
// rocblas_status_not_implemented instead of running.
|
|
if constexpr(std::is_same<I, int64_t>::value)
|
|
{
|
|
if(BATCHED)
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n,
|
|
(T* const*)nullptr, lda, stA, (S*)nullptr,
|
|
stD, (S*)nullptr, stE, (I*)nullptr, bc),
|
|
rocblas_status_not_implemented);
|
|
else
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n,
|
|
(T*)nullptr, lda, stA, (S*)nullptr, stD,
|
|
(S*)nullptr, stE, (I*)nullptr, bc),
|
|
rocblas_status_not_implemented);
|
|
return;
|
|
}
|
|
#endif
|
|
|
|
if(argus.alg_mode == 1)
|
|
{
|
|
EXPECT_ROCBLAS_STATUS(
|
|
rocsolver_set_alg_mode(handle, rocsolver_function_sterf, rocsolver_alg_mode_hybrid),
|
|
rocblas_status_success);
|
|
|
|
rocsolver_alg_mode alg_mode;
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_get_alg_mode(handle, rocsolver_function_sterf, &alg_mode),
|
|
rocblas_status_success);
|
|
|
|
EXPECT_EQ(alg_mode, rocsolver_alg_mode_hybrid);
|
|
}
|
|
|
|
// check non-supported values
|
|
if(uplo == rocblas_fill_full || evect == rocblas_evect_tridiagonal)
|
|
{
|
|
if(BATCHED)
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n,
|
|
(T* const*)nullptr, lda, stA, (S*)nullptr,
|
|
stD, (S*)nullptr, stE, (I*)nullptr, bc),
|
|
rocblas_status_invalid_value);
|
|
else
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n,
|
|
(T*)nullptr, lda, stA, (S*)nullptr, stD,
|
|
(S*)nullptr, stE, (I*)nullptr, bc),
|
|
rocblas_status_invalid_value);
|
|
|
|
if(argus.timing)
|
|
rocsolver_bench_inform(inform_invalid_args);
|
|
|
|
return;
|
|
}
|
|
|
|
// determine sizes
|
|
size_t size_A = size_t(lda) * n;
|
|
size_t size_D = n;
|
|
size_t size_E = size_D;
|
|
size_t size_Ares = (argus.unit_check || argus.norm_check) ? size_A : 0;
|
|
size_t size_Dres = (argus.unit_check || argus.norm_check) ? size_D : 0;
|
|
|
|
double max_error = 0, max_ortho_error = 0, gpu_time_used = 0, cpu_time_used = 0;
|
|
|
|
// check invalid sizes
|
|
bool invalid_size = (n < 0 || lda < n || bc < 0);
|
|
if(invalid_size)
|
|
{
|
|
if(BATCHED)
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n,
|
|
(T* const*)nullptr, lda, stA, (S*)nullptr,
|
|
stD, (S*)nullptr, stE, (I*)nullptr, bc),
|
|
rocblas_status_invalid_size);
|
|
else
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n,
|
|
(T*)nullptr, lda, stA, (S*)nullptr, stD,
|
|
(S*)nullptr, stE, (I*)nullptr, bc),
|
|
rocblas_status_invalid_size);
|
|
|
|
if(argus.timing)
|
|
rocsolver_bench_inform(inform_invalid_size);
|
|
|
|
return;
|
|
}
|
|
|
|
// memory size query is necessary
|
|
if(argus.mem_query)
|
|
{
|
|
CHECK_ROCBLAS_ERROR(rocblas_start_device_memory_size_query(handle));
|
|
if(BATCHED)
|
|
CHECK_ALLOC_QUERY(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n,
|
|
(T* const*)nullptr, lda, stA, (S*)nullptr, stD,
|
|
(S*)nullptr, stE, (I*)nullptr, bc));
|
|
else
|
|
CHECK_ALLOC_QUERY(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n, (T*)nullptr,
|
|
lda, stA, (S*)nullptr, stD, (S*)nullptr, stE,
|
|
(I*)nullptr, bc));
|
|
|
|
size_t size;
|
|
CHECK_ROCBLAS_ERROR(rocblas_stop_device_memory_size_query(handle, &size));
|
|
|
|
rocsolver_bench_inform(inform_mem_query, size);
|
|
return;
|
|
}
|
|
|
|
// memory allocations (all cases)
|
|
// host
|
|
host_strided_batch_vector<S> hD(size_D, 1, stD, bc);
|
|
host_strided_batch_vector<I> hinfo(1, 1, 1, bc);
|
|
host_strided_batch_vector<I> hinfoRes(1, 1, 1, bc);
|
|
host_strided_batch_vector<S> hDres(size_Dres, 1, stD, bc);
|
|
// device
|
|
device_strided_batch_vector<S> dE(size_E, 1, stE, bc);
|
|
device_strided_batch_vector<S> dD(size_D, 1, stD, bc);
|
|
device_strided_batch_vector<I> dinfo(1, 1, 1, bc);
|
|
if(size_E)
|
|
CHECK_HIP_ERROR(dE.memcheck());
|
|
if(size_D)
|
|
CHECK_HIP_ERROR(dD.memcheck());
|
|
CHECK_HIP_ERROR(dinfo.memcheck());
|
|
|
|
if(BATCHED)
|
|
{
|
|
// memory allocations
|
|
host_batch_vector<T> hA(size_A, 1, bc);
|
|
host_batch_vector<T> hAres(size_Ares, 1, bc);
|
|
device_batch_vector<T> dA(size_A, 1, bc);
|
|
if(size_A)
|
|
CHECK_HIP_ERROR(dA.memcheck());
|
|
|
|
// check quick return
|
|
if(n == 0 || bc == 0)
|
|
{
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n, dA.data(),
|
|
lda, stA, dD.data(), stD, dE.data(), stE,
|
|
dinfo.data(), bc),
|
|
rocblas_status_success);
|
|
if(argus.timing)
|
|
rocsolver_bench_inform(inform_quick_return);
|
|
|
|
return;
|
|
}
|
|
|
|
// check computations
|
|
if(argus.unit_check || argus.norm_check)
|
|
{
|
|
syevd_heevd_getError<STRIDED, T, I>(handle, evect, uplo, n, dA, lda, stA, dD, stD, dE,
|
|
stE, dinfo, bc, hA, hAres, hD, hDres, hinfo,
|
|
hinfoRes, &max_error, &max_ortho_error);
|
|
}
|
|
|
|
// collect performance data
|
|
if(argus.timing && hot_calls > 0)
|
|
{
|
|
syevd_heevd_getPerfData<STRIDED, T, I>(handle, evect, uplo, n, dA, lda, stA, dD, stD,
|
|
dE, stE, dinfo, bc, hA, hD, hinfo,
|
|
&gpu_time_used, &cpu_time_used, hot_calls,
|
|
argus.profile, argus.profile_kernels, argus.perf);
|
|
}
|
|
}
|
|
|
|
else
|
|
{
|
|
// memory allocations
|
|
host_strided_batch_vector<T> hA(size_A, 1, stA, bc);
|
|
host_strided_batch_vector<T> hAres(size_Ares, 1, stA, bc);
|
|
device_strided_batch_vector<T> dA(size_A, 1, stA, bc);
|
|
if(size_A)
|
|
CHECK_HIP_ERROR(dA.memcheck());
|
|
|
|
// check quick return
|
|
if(n == 0 || bc == 0)
|
|
{
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_syevd_heevd(STRIDED, handle, evect, uplo, n, dA.data(),
|
|
lda, stA, dD.data(), stD, dE.data(), stE,
|
|
dinfo.data(), bc),
|
|
rocblas_status_success);
|
|
if(argus.timing)
|
|
rocsolver_bench_inform(inform_quick_return);
|
|
|
|
return;
|
|
}
|
|
|
|
// check computations
|
|
if(argus.unit_check || argus.norm_check)
|
|
{
|
|
syevd_heevd_getError<STRIDED, T, I>(handle, evect, uplo, n, dA, lda, stA, dD, stD, dE,
|
|
stE, dinfo, bc, hA, hAres, hD, hDres, hinfo,
|
|
hinfoRes, &max_error, &max_ortho_error);
|
|
}
|
|
|
|
// collect performance data
|
|
if(argus.timing && hot_calls > 0)
|
|
{
|
|
syevd_heevd_getPerfData<STRIDED, T, I>(handle, evect, uplo, n, dA, lda, stA, dD, stD,
|
|
dE, stE, dinfo, bc, hA, hD, hinfo,
|
|
&gpu_time_used, &cpu_time_used, hot_calls,
|
|
argus.profile, argus.profile_kernels, argus.perf);
|
|
}
|
|
}
|
|
|
|
// validate results for rocsolver-test
|
|
// using 10 * n * machine_precision as tolerance
|
|
if(argus.unit_check)
|
|
{
|
|
ROCSOLVER_TEST_CHECK(T, max_error, 10 * n);
|
|
if(evect != rocblas_evect_none)
|
|
ROCSOLVER_TEST_CHECK(T, max_ortho_error, 10 * n);
|
|
}
|
|
|
|
// output results for rocsolver-bench
|
|
if(argus.timing)
|
|
{
|
|
if(!argus.perf)
|
|
{
|
|
rocsolver_bench_header("Arguments:");
|
|
if(BATCHED)
|
|
{
|
|
rocsolver_bench_output("evect", "uplo", "n", "lda", "strideD", "strideE", "batch_c");
|
|
rocsolver_bench_output(evectC, uploC, n, lda, stD, stE, bc);
|
|
}
|
|
else if(STRIDED)
|
|
{
|
|
rocsolver_bench_output("evect", "uplo", "n", "lda", "strideA", "strideD", "strideE",
|
|
"batch_c");
|
|
rocsolver_bench_output(evectC, uploC, n, lda, stA, stD, stE, bc);
|
|
}
|
|
else
|
|
{
|
|
rocsolver_bench_output("evect", "uplo", "n", "lda");
|
|
rocsolver_bench_output(evectC, uploC, n, lda);
|
|
}
|
|
rocsolver_bench_header("Results:");
|
|
if(argus.norm_check)
|
|
{
|
|
rocsolver_bench_output("cpu_time_us", "gpu_time_us", "error");
|
|
rocsolver_bench_output(cpu_time_used, gpu_time_used, max_error);
|
|
}
|
|
else
|
|
{
|
|
rocsolver_bench_output("cpu_time_us", "gpu_time_us");
|
|
rocsolver_bench_output(cpu_time_used, gpu_time_used);
|
|
}
|
|
rocsolver_bench_endl();
|
|
}
|
|
else
|
|
{
|
|
if(argus.norm_check)
|
|
rocsolver_bench_output(gpu_time_used, max_error);
|
|
else
|
|
rocsolver_bench_output(gpu_time_used);
|
|
}
|
|
}
|
|
|
|
// ensure all arguments were consumed
|
|
argus.validate_consumed();
|
|
}
|
|
|
|
#define EXTERN_TESTING_SYEVD_HEEVD(...) \
|
|
extern template void testing_syevd_heevd<__VA_ARGS__>(Arguments&);
|
|
|
|
INSTANTIATE(EXTERN_TESTING_SYEVD_HEEVD,
|
|
FOREACH_MATRIX_DATA_LAYOUT,
|
|
FOREACH_SCALAR_TYPE,
|
|
FOREACH_INT_TYPE,
|
|
APPLY_STAMP)
|