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
695 lines
26 KiB
C++
695 lines
26 KiB
C++
/* **************************************************************************
|
|
* Copyright (C) 2020-2025 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/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 <typename T, typename S>
|
|
void bdsqr_checkBadArgs(const rocblas_handle handle,
|
|
const rocblas_fill uplo,
|
|
const rocblas_int n,
|
|
const rocblas_int nv,
|
|
const rocblas_int nu,
|
|
const rocblas_int nc,
|
|
S dD,
|
|
S dE,
|
|
T dV,
|
|
const rocblas_int ldv,
|
|
T dU,
|
|
const rocblas_int ldu,
|
|
T dC,
|
|
const rocblas_int ldc,
|
|
rocblas_int* dinfo)
|
|
{
|
|
// handle
|
|
EXPECT_ROCBLAS_STATUS(
|
|
rocsolver_bdsqr(nullptr, uplo, n, nv, nu, nc, dD, dE, dV, ldv, dU, ldu, dC, ldc, dinfo),
|
|
rocblas_status_invalid_handle);
|
|
|
|
// values
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_bdsqr(handle, rocblas_fill_full, n, nv, nu, nc, dD, dE, dV, ldv,
|
|
dU, ldu, dC, ldc, dinfo),
|
|
rocblas_status_invalid_value);
|
|
|
|
// pointers
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_bdsqr(handle, uplo, n, nv, nu, nc, (S) nullptr, dE, dV, ldv, dU,
|
|
ldu, dC, ldc, dinfo),
|
|
rocblas_status_invalid_pointer);
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_bdsqr(handle, uplo, n, nv, nu, nc, dD, (S) nullptr, dV, ldv, dU,
|
|
ldu, dC, ldc, dinfo),
|
|
rocblas_status_invalid_pointer);
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_bdsqr(handle, uplo, n, nv, nu, nc, dD, dE, (T) nullptr, ldv, dU,
|
|
ldu, dC, ldc, dinfo),
|
|
rocblas_status_invalid_pointer);
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_bdsqr(handle, uplo, n, nv, nu, nc, dD, dE, dV, ldv, (T) nullptr,
|
|
ldu, dC, ldc, dinfo),
|
|
rocblas_status_invalid_pointer);
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_bdsqr(handle, uplo, n, nv, nu, nc, dD, dE, dV, ldv, dU, ldu,
|
|
(T) nullptr, ldc, dinfo),
|
|
rocblas_status_invalid_pointer);
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_bdsqr(handle, uplo, n, nv, nu, nc, dD, dE, dV, ldv, dU, ldu, dC,
|
|
ldc, (rocblas_int*)nullptr),
|
|
rocblas_status_invalid_pointer);
|
|
|
|
// quick return with invalid pointers
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_bdsqr(handle, uplo, 0, nv, nu, nc, (S) nullptr, (S) nullptr,
|
|
(T) nullptr, ldv, (T) nullptr, ldu, (T) nullptr, ldc, dinfo),
|
|
rocblas_status_success);
|
|
}
|
|
|
|
template <typename T>
|
|
void testing_bdsqr_bad_arg()
|
|
{
|
|
using S = decltype(std::real(T{}));
|
|
|
|
// safe arguments
|
|
rocblas_local_handle handle;
|
|
rocblas_fill uplo = rocblas_fill_upper;
|
|
rocblas_int n = 2;
|
|
rocblas_int nv = 2;
|
|
rocblas_int nu = 2;
|
|
rocblas_int nc = 2;
|
|
rocblas_int ldv = 2;
|
|
rocblas_int ldu = 2;
|
|
rocblas_int ldc = 2;
|
|
|
|
// memory allocations
|
|
device_strided_batch_vector<S> dD(1, 1, 1, 1);
|
|
device_strided_batch_vector<S> dE(1, 1, 1, 1);
|
|
device_strided_batch_vector<T> dV(1, 1, 1, 1);
|
|
device_strided_batch_vector<T> dU(1, 1, 1, 1);
|
|
device_strided_batch_vector<T> dC(1, 1, 1, 1);
|
|
device_strided_batch_vector<rocblas_int> dinfo(1, 1, 1, 1);
|
|
CHECK_HIP_ERROR(dD.memcheck());
|
|
CHECK_HIP_ERROR(dE.memcheck());
|
|
CHECK_HIP_ERROR(dV.memcheck());
|
|
CHECK_HIP_ERROR(dU.memcheck());
|
|
CHECK_HIP_ERROR(dC.memcheck());
|
|
CHECK_HIP_ERROR(dinfo.memcheck());
|
|
|
|
// check bad arguments
|
|
bdsqr_checkBadArgs(handle, uplo, n, nv, nu, nc, dD.data(), dE.data(), dV.data(), ldv, dU.data(),
|
|
ldu, dC.data(), ldc, dinfo.data());
|
|
}
|
|
|
|
template <bool CPU, bool GPU, typename T, typename S, typename Sd, typename Td, typename Ud, typename Sh, typename Th, typename Uh>
|
|
void bdsqr_initData(const rocblas_handle handle,
|
|
const rocblas_fill uplo,
|
|
const rocblas_int n,
|
|
const rocblas_int nv,
|
|
const rocblas_int nu,
|
|
const rocblas_int nc,
|
|
Sd& dD,
|
|
Sd& dE,
|
|
Td& dV,
|
|
const rocblas_int ldv,
|
|
Td& dU,
|
|
const rocblas_int ldu,
|
|
Td& dC,
|
|
const rocblas_int ldc,
|
|
Ud& dInfo,
|
|
Sh& hD,
|
|
Sh& hE,
|
|
Th& hV,
|
|
Th& hU,
|
|
Th& hC,
|
|
Uh& hInfo,
|
|
std::vector<S>& D,
|
|
std::vector<S>& E,
|
|
const bool test)
|
|
{
|
|
if(CPU)
|
|
{
|
|
rocblas_init<S>(hD, true);
|
|
rocblas_init<S>(hE, false);
|
|
|
|
// Adding possible gaps to fully test the algorithm.
|
|
for(rocblas_int i = 0; i < n - 1; ++i)
|
|
{
|
|
hE[0][i] -= 5;
|
|
hD[0][i] -= 4;
|
|
}
|
|
hD[0][n - 1] -= 4;
|
|
|
|
// (Forcing non-convergence expecting lapack and rocsolver to give
|
|
// the same orthogonal equivalent matrix is not possible. Testing
|
|
// implicitly the equivalent matrix is very complicated and it boils
|
|
// down to essentially run the algorithm again and until convergence is achieved).
|
|
|
|
// make copy of original data to test vectors if required
|
|
if(test && (nv || nu || nc))
|
|
{
|
|
for(rocblas_int i = 0; i < n - 1; ++i)
|
|
{
|
|
E[i] = hE[0][i];
|
|
D[i] = hD[0][i];
|
|
}
|
|
D[n - 1] = hD[0][n - 1];
|
|
}
|
|
|
|
// make V,U and C identities so that results are actually singular vectors
|
|
// of B
|
|
if(nv > 0)
|
|
{
|
|
memset(hV[0], 0, ldv * nv * sizeof(T));
|
|
for(rocblas_int i = 0; i < std::min(n, nv); ++i)
|
|
hV[0][i + i * ldv] = T(1.0);
|
|
}
|
|
if(nu > 0)
|
|
{
|
|
memset(hU[0], 0, ldu * n * sizeof(T));
|
|
for(rocblas_int i = 0; i < std::min(n, nu); ++i)
|
|
hU[0][i + i * ldu] = T(1.0);
|
|
}
|
|
if(nc > 0)
|
|
{
|
|
memset(hC[0], 0, ldc * nc * sizeof(T));
|
|
for(rocblas_int i = 0; i < std::min(n, nc); ++i)
|
|
hC[0][i + i * ldc] = T(1.0);
|
|
}
|
|
}
|
|
|
|
if(GPU)
|
|
{
|
|
// now copy to the GPU
|
|
CHECK_HIP_ERROR(dD.transfer_from(hD));
|
|
CHECK_HIP_ERROR(dE.transfer_from(hE));
|
|
if(nv > 0)
|
|
CHECK_HIP_ERROR(dV.transfer_from(hV));
|
|
if(nu > 0)
|
|
CHECK_HIP_ERROR(dU.transfer_from(hU));
|
|
if(nc > 0)
|
|
CHECK_HIP_ERROR(dC.transfer_from(hC));
|
|
}
|
|
}
|
|
|
|
template <typename T, typename Sd, typename Td, typename Ud, typename Sh, typename Th, typename Uh>
|
|
void bdsqr_getError(const rocblas_handle handle,
|
|
const rocblas_fill uplo,
|
|
const rocblas_int n,
|
|
const rocblas_int nv,
|
|
const rocblas_int nu,
|
|
const rocblas_int nc,
|
|
Sd& dD,
|
|
Sd& dE,
|
|
Td& dV,
|
|
const rocblas_int ldv,
|
|
Td& dU,
|
|
const rocblas_int ldu,
|
|
Td& dC,
|
|
const rocblas_int ldc,
|
|
Ud& dInfo,
|
|
const rocblas_int nvT,
|
|
const rocblas_int nuT,
|
|
const rocblas_int nvRes,
|
|
const rocblas_int nuRes,
|
|
Sh& hD,
|
|
Sh& hDRes,
|
|
Sh& hE,
|
|
Sh& hERes,
|
|
Th& hV,
|
|
Th& hU,
|
|
Th& hC,
|
|
Uh& hInfo,
|
|
Uh& hInfoRes,
|
|
double* max_err,
|
|
double* max_errv)
|
|
{
|
|
using S = decltype(std::real(T{}));
|
|
std::vector<S> hW(4 * n);
|
|
std::vector<S> D(n);
|
|
std::vector<S> E(n);
|
|
|
|
// input data initialization
|
|
bdsqr_initData<true, false, T>(handle, uplo, n, nvRes, nuRes, nc, dD, dE, dV, ldv, dU, ldu, dC,
|
|
ldc, dInfo, hD, hE, hV, hU, hC, hInfo, D, E, true);
|
|
|
|
// execute computations:
|
|
// complementary execution to compute all singular vectors if needed
|
|
if(nvT > 0 || nuT > 0)
|
|
{
|
|
// send data to GPU
|
|
bdsqr_initData<false, true, T>(handle, uplo, n, nvT, nuT, 0, dD, dE, dV, ldv, dU, ldu, dC,
|
|
ldc, dInfo, hD, hE, hV, hU, hC, hInfo, D, E, false);
|
|
|
|
CHECK_ROCBLAS_ERROR(rocsolver_bdsqr(handle, uplo, n, nvT, nuT, 0, dD.data(), dE.data(),
|
|
dV.data(), ldv, dU.data(), ldu, dC.data(), ldc,
|
|
dInfo.data()));
|
|
}
|
|
|
|
// send data to GPU
|
|
bdsqr_initData<false, true, T>(handle, uplo, n, nv, nu, nc, dD, dE, dV, ldv, dU, ldu, dC, ldc,
|
|
dInfo, hD, hE, hV, hU, hC, hInfo, D, E, false);
|
|
|
|
// execute computations
|
|
// CPU lapack
|
|
cpu_bdsqr(uplo, n, 0, 0, 0, hD[0], hE[0], hV[0], ldv, hU[0], ldu, hC[0], ldc, hW.data(),
|
|
hInfo[0]);
|
|
|
|
// GPU lapack
|
|
CHECK_ROCBLAS_ERROR(rocsolver_bdsqr(handle, uplo, n, nv, nu, nc, dD.data(), dE.data(), dV.data(),
|
|
ldv, dU.data(), ldu, dC.data(), ldc, dInfo.data()));
|
|
CHECK_HIP_ERROR(hDRes.transfer_from(dD));
|
|
CHECK_HIP_ERROR(hERes.transfer_from(dE));
|
|
CHECK_HIP_ERROR(hInfoRes.transfer_from(dInfo));
|
|
if(nvRes > 0)
|
|
CHECK_HIP_ERROR(hV.transfer_from(dV));
|
|
if(nuRes > 0)
|
|
CHECK_HIP_ERROR(hU.transfer_from(dU));
|
|
if(nc > 0)
|
|
CHECK_HIP_ERROR(hC.transfer_from(dC));
|
|
|
|
// Check info for non-convergence
|
|
*max_err = 0;
|
|
EXPECT_EQ(hInfo[0][0], hInfoRes[0][0]);
|
|
if(hInfo[0][0] != hInfoRes[0][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).
|
|
|
|
// error is ||hD - hDRes||
|
|
// (THIS DOES NOT ACCOUNT FOR NUMERICAL REPRODUCIBILITY ISSUES.
|
|
// IT MIGHT BE REVISITED IN THE FUTURE)
|
|
double err;
|
|
T tmp;
|
|
*max_errv = 0;
|
|
err = norm_error('F', 1, n, 1, hD[0], hDRes[0]);
|
|
*max_err = err > *max_err ? err : *max_err;
|
|
|
|
// Check the singular vectors if required
|
|
if(hInfo[0][0] == 0 && (nv || nu))
|
|
{
|
|
err = 0;
|
|
rocblas_int n_comp = std::min(n, nvRes);
|
|
|
|
if(uplo == rocblas_fill_upper)
|
|
{
|
|
// check singular vectors implicitly (A'*u_i = s_i*v_i)
|
|
for(rocblas_int i = 0; i < n_comp; ++i)
|
|
{
|
|
for(rocblas_int j = 0; j < n; ++j)
|
|
{
|
|
if(i > 0)
|
|
tmp = D[i] * hU[0][i + j * ldu] + E[i - 1] * hU[0][(i - 1) + j * ldu]
|
|
- hDRes[0][j] * hV[0][j + i * ldv];
|
|
else
|
|
tmp = D[i] * hU[0][i + j * ldu] - hDRes[0][j] * hV[0][j + i * ldv];
|
|
err += std::abs(tmp) * std::abs(tmp);
|
|
}
|
|
}
|
|
}
|
|
else
|
|
{
|
|
// check singular vectors implicitly (A*v_i = s_i*u_i)
|
|
for(rocblas_int i = 0; i < n_comp; ++i)
|
|
{
|
|
for(rocblas_int j = 0; j < n; ++j)
|
|
{
|
|
if(i > 0)
|
|
tmp = D[i] * hV[0][j + i * ldv] + E[i - 1] * hV[0][j + (i - 1) * ldv]
|
|
- hDRes[0][j] * hU[0][i + j * ldu];
|
|
else
|
|
tmp = D[i] * hV[0][j + i * ldv] - hDRes[0][j] * hU[0][i + j * ldu];
|
|
err += std::abs(tmp) * std::abs(tmp);
|
|
}
|
|
}
|
|
}
|
|
|
|
double normD = double(snorm('F', 1, n, D.data(), 1));
|
|
double normE = double(snorm('F', 1, n - 1, E.data(), 1));
|
|
err = std::sqrt(err) / std::sqrt(normD * normD + normE * normE);
|
|
*max_errv = err > *max_errv ? err : *max_errv;
|
|
}
|
|
|
|
// C should be the transpose of U
|
|
if(hInfo[0][0] == 0 && nc)
|
|
{
|
|
err = 0;
|
|
for(rocblas_int i = 0; i < nc; ++i)
|
|
{
|
|
for(rocblas_int j = 0; j < n; ++j)
|
|
{
|
|
tmp = hC[0][j + i * ldc] - hU[0][i + j * ldu];
|
|
err += std::abs(tmp) * std::abs(tmp);
|
|
}
|
|
}
|
|
err = std::sqrt(err);
|
|
*max_errv = err > *max_errv ? err : *max_errv;
|
|
}
|
|
}
|
|
|
|
template <typename T, typename Sd, typename Td, typename Ud, typename Sh, typename Th, typename Uh>
|
|
void bdsqr_getPerfData(const rocblas_handle handle,
|
|
const rocblas_fill uplo,
|
|
const rocblas_int n,
|
|
const rocblas_int nv,
|
|
const rocblas_int nu,
|
|
const rocblas_int nc,
|
|
Sd& dD,
|
|
Sd& dE,
|
|
Td& dV,
|
|
const rocblas_int ldv,
|
|
Td& dU,
|
|
const rocblas_int ldu,
|
|
Td& dC,
|
|
const rocblas_int ldc,
|
|
Ud& dInfo,
|
|
Sh& hD,
|
|
Sh& hE,
|
|
Th& hV,
|
|
Th& hU,
|
|
Th& hC,
|
|
Uh& hInfo,
|
|
double* gpu_time_used,
|
|
double* cpu_time_used,
|
|
const rocblas_int hot_calls,
|
|
const int profile,
|
|
const bool profile_kernels,
|
|
const bool perf)
|
|
{
|
|
using S = decltype(std::real(T{}));
|
|
std::vector<S> hW(4 * n);
|
|
std::vector<S> D;
|
|
std::vector<S> E;
|
|
|
|
if(!perf)
|
|
{
|
|
bdsqr_initData<true, false, T>(handle, uplo, n, nv, nu, nc, dD, dE, dV, ldv, dU, ldu, dC,
|
|
ldc, dInfo, hD, hE, hV, hU, hC, hInfo, D, E, false);
|
|
|
|
// cpu-lapack performance (only if not in perf mode)
|
|
*cpu_time_used = get_time_us_no_sync();
|
|
cpu_bdsqr(uplo, n, nv, nu, nc, hD[0], hE[0], hV[0], ldv, hU[0], ldu, hC[0], ldc, hW.data(),
|
|
hInfo[0]);
|
|
*cpu_time_used = get_time_us_no_sync() - *cpu_time_used;
|
|
}
|
|
|
|
bdsqr_initData<true, false, T>(handle, uplo, n, nv, nu, nc, dD, dE, dV, ldv, dU, ldu, dC, ldc,
|
|
dInfo, hD, hE, hV, hU, hC, hInfo, D, E, false);
|
|
|
|
// cold calls
|
|
for(int iter = 0; iter < 2; iter++)
|
|
{
|
|
bdsqr_initData<false, true, T>(handle, uplo, n, nv, nu, nc, dD, dE, dV, ldv, dU, ldu, dC,
|
|
ldc, dInfo, hD, hE, hV, hU, hC, hInfo, D, E, false);
|
|
|
|
CHECK_ROCBLAS_ERROR(rocsolver_bdsqr(handle, uplo, n, nv, nu, nc, dD.data(), dE.data(),
|
|
dV.data(), ldv, dU.data(), ldu, dC.data(), ldc,
|
|
dInfo.data()));
|
|
}
|
|
|
|
// 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++)
|
|
{
|
|
bdsqr_initData<false, true, T>(handle, uplo, n, nv, nu, nc, dD, dE, dV, ldv, dU, ldu, dC,
|
|
ldc, dInfo, hD, hE, hV, hU, hC, hInfo, D, E, false);
|
|
|
|
timer.start(stream);
|
|
rocsolver_bdsqr(handle, uplo, n, nv, nu, nc, dD.data(), dE.data(), dV.data(), ldv,
|
|
dU.data(), ldu, dC.data(), ldc, dInfo.data());
|
|
timer.end(stream);
|
|
}
|
|
*gpu_time_used = timer.get_combined();
|
|
}
|
|
|
|
template <typename T>
|
|
void testing_bdsqr(Arguments& argus)
|
|
{
|
|
using S = decltype(std::real(T{}));
|
|
|
|
// get arguments
|
|
rocblas_local_handle handle;
|
|
char uploC = argus.get<char>("uplo");
|
|
rocblas_int n = argus.get<rocblas_int>("n");
|
|
rocblas_int nv = argus.get<rocblas_int>("nv", 0);
|
|
rocblas_int nu = argus.get<rocblas_int>("nu", 0);
|
|
rocblas_int nc = argus.get<rocblas_int>("nc", 0);
|
|
rocblas_int ldv = argus.get<rocblas_int>("ldv", nv > 0 ? n : 1);
|
|
rocblas_int ldu = argus.get<rocblas_int>("ldu", nu > 0 ? nu : 1);
|
|
rocblas_int ldc = argus.get<rocblas_int>("ldc", nc > 0 ? n : 1);
|
|
|
|
rocblas_fill uplo = char2rocblas_fill(uploC);
|
|
rocblas_int hot_calls = argus.iters;
|
|
|
|
if(argus.alg_mode == 1)
|
|
{
|
|
EXPECT_ROCBLAS_STATUS(
|
|
rocsolver_set_alg_mode(handle, rocsolver_function_bdsqr, rocsolver_alg_mode_hybrid),
|
|
rocblas_status_success);
|
|
|
|
rocsolver_alg_mode alg_mode;
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_get_alg_mode(handle, rocsolver_function_bdsqr, &alg_mode),
|
|
rocblas_status_success);
|
|
|
|
EXPECT_EQ(alg_mode, rocsolver_alg_mode_hybrid);
|
|
}
|
|
|
|
// check non-supported values
|
|
if(uplo != rocblas_fill_upper && uplo != rocblas_fill_lower)
|
|
{
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_bdsqr(handle, uplo, n, nv, nu, nc, (S*)nullptr, (S*)nullptr,
|
|
(T*)nullptr, ldv, (T*)nullptr, ldu, (T*)nullptr, ldc,
|
|
(rocblas_int*)nullptr),
|
|
rocblas_status_invalid_value);
|
|
|
|
if(argus.timing)
|
|
rocsolver_bench_inform(inform_invalid_args);
|
|
|
|
return;
|
|
}
|
|
|
|
// sizes for testing singular vectors
|
|
// TESTING OF SINGULAR VECTORS IS DONE IMPLICITLY (NOT EXPLICITLY COMPARING
|
|
// WITH LAPACK), SO WE ALWAYS NEED TO COMPUTE THE SAME NUMBER OF ELEMENTS OF
|
|
// THE RIGHT AND LEFT VECTORS
|
|
rocblas_int nvA = nv, nuA = nu;
|
|
rocblas_int nvT = 0, nuT = 0;
|
|
rocblas_int nvRes = nv, nuRes = nu;
|
|
rocblas_int ldvRes = ldv, lduRes = ldu;
|
|
|
|
if(nv && nu)
|
|
{
|
|
nvRes = nvA = std::max(nv, nu);
|
|
nuRes = nuA = std::max(nvRes, nc);
|
|
lduRes = std::max(lduRes, nuRes);
|
|
}
|
|
else if(nu)
|
|
{
|
|
nvRes = nvT = nu;
|
|
nuRes = nuA = std::max(nu, nc);
|
|
ldvRes = n;
|
|
lduRes = std::max(lduRes, nuRes);
|
|
}
|
|
else if(nv || nc)
|
|
{
|
|
nuRes = nuT = std::max(nv, nc);
|
|
lduRes = nuRes;
|
|
}
|
|
|
|
// E, V, U, and C could have size zero in cases that are not quick-return or
|
|
// invalid cases setting the size to one to avoid possible memory-access
|
|
// errors in the rest of the unit test
|
|
size_t size_D = size_t(n);
|
|
size_t size_E = n > 1 ? size_t(n - 1) : 1;
|
|
size_t size_V = std::max(size_t(ldv) * nv, size_t(1));
|
|
size_t size_U = std::max(size_t(ldu) * n, size_t(1));
|
|
size_t size_C = std::max(size_t(ldc) * nc, size_t(1));
|
|
size_t size_VRes = std::max(size_t(ldvRes) * nvRes, size_t(1));
|
|
size_t size_URes = std::max(size_t(lduRes) * n, size_t(1));
|
|
double max_error = 0, gpu_time_used = 0, cpu_time_used = 0, max_errorv = 0;
|
|
|
|
// check invalid sizes
|
|
bool invalid_size = (n < 0 || nv < 0 || nu < 0 || nc < 0 || ldu < nu || ldv < 1 || ldc < 1)
|
|
|| (nv > 0 && ldv < n) || (nc > 0 && ldc < n);
|
|
if(invalid_size)
|
|
{
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_bdsqr(handle, uplo, n, nv, nu, nc, (S*)nullptr, (S*)nullptr,
|
|
(T*)nullptr, ldv, (T*)nullptr, ldu, (T*)nullptr, ldc,
|
|
(rocblas_int*)nullptr),
|
|
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));
|
|
CHECK_ALLOC_QUERY(rocsolver_bdsqr(handle, uplo, n, nv, nu, nc, (S*)nullptr, (S*)nullptr,
|
|
(T*)nullptr, ldv, (T*)nullptr, ldu, (T*)nullptr, ldc,
|
|
(rocblas_int*)nullptr));
|
|
|
|
size_t size;
|
|
CHECK_ROCBLAS_ERROR(rocblas_stop_device_memory_size_query(handle, &size));
|
|
|
|
rocsolver_bench_inform(inform_mem_query, size);
|
|
return;
|
|
}
|
|
|
|
// memory allocations
|
|
host_strided_batch_vector<S> hD(size_D, 1, size_D, 1);
|
|
host_strided_batch_vector<S> hE(size_E, 1, size_E, 1);
|
|
host_strided_batch_vector<T> hC(size_C, 1, size_C, 1);
|
|
host_strided_batch_vector<rocblas_int> hInfo(1, 1, 1, 1);
|
|
device_strided_batch_vector<S> dD(size_D, 1, size_D, 1);
|
|
device_strided_batch_vector<S> dE(size_E, 1, size_E, 1);
|
|
device_strided_batch_vector<T> dC(size_C, 1, size_C, 1);
|
|
device_strided_batch_vector<rocblas_int> dInfo(1, 1, 1, 1);
|
|
if(size_D)
|
|
CHECK_HIP_ERROR(dD.memcheck());
|
|
if(size_E)
|
|
CHECK_HIP_ERROR(dE.memcheck());
|
|
if(size_C)
|
|
CHECK_HIP_ERROR(dC.memcheck());
|
|
CHECK_HIP_ERROR(dInfo.memcheck());
|
|
|
|
// check computations
|
|
if(argus.unit_check || argus.norm_check)
|
|
{
|
|
host_strided_batch_vector<S> hDRes(size_D, 1, size_D, 1);
|
|
host_strided_batch_vector<S> hERes(size_E, 1, size_E, 1);
|
|
host_strided_batch_vector<T> hV(size_VRes, 1, size_VRes, 1);
|
|
host_strided_batch_vector<T> hU(size_URes, 1, size_URes, 1);
|
|
host_strided_batch_vector<rocblas_int> hInfoRes(1, 1, 1, 1);
|
|
device_strided_batch_vector<T> dV(size_VRes, 1, size_VRes, 1);
|
|
device_strided_batch_vector<T> dU(size_URes, 1, size_URes, 1);
|
|
if(size_VRes)
|
|
CHECK_HIP_ERROR(dV.memcheck());
|
|
if(size_URes)
|
|
CHECK_HIP_ERROR(dU.memcheck());
|
|
|
|
// check quick return
|
|
if(n == 0)
|
|
{
|
|
EXPECT_ROCBLAS_STATUS(rocsolver_bdsqr(handle, uplo, n, nv, nu, nc, dD.data(), dE.data(),
|
|
dV.data(), ldv, dU.data(), ldu, dC.data(), ldc,
|
|
dInfo.data()),
|
|
rocblas_status_success);
|
|
if(argus.timing)
|
|
rocsolver_bench_inform(inform_quick_return);
|
|
|
|
return;
|
|
}
|
|
|
|
bdsqr_getError<T>(handle, uplo, n, nvA, nuA, nc, dD, dE, dV, ldvRes, dU, lduRes, dC, ldc,
|
|
dInfo, nvT, nuT, nvRes, nuRes, hD, hDRes, hE, hERes, hV, hU, hC, hInfo,
|
|
hInfoRes, &max_error, &max_errorv);
|
|
}
|
|
|
|
// collect performance data
|
|
if(argus.timing && hot_calls > 0)
|
|
{
|
|
host_strided_batch_vector<T> hV(size_V, 1, size_V, 1);
|
|
host_strided_batch_vector<T> hU(size_U, 1, size_U, 1);
|
|
device_strided_batch_vector<T> dV(size_V, 1, size_V, 1);
|
|
device_strided_batch_vector<T> dU(size_U, 1, size_U, 1);
|
|
if(size_V)
|
|
CHECK_HIP_ERROR(dV.memcheck());
|
|
if(size_U)
|
|
CHECK_HIP_ERROR(dU.memcheck());
|
|
|
|
bdsqr_getPerfData<T>(handle, uplo, n, nv, nu, nc, dD, dE, dV, ldv, dU, ldu, dC, ldc, dInfo,
|
|
hD, hE, hV, hU, hC, hInfo, &gpu_time_used, &cpu_time_used, hot_calls,
|
|
argus.profile, argus.profile_kernels, argus.perf);
|
|
}
|
|
|
|
// validate results for rocsolver-test
|
|
// using 2 * n * machine_precision as tolerance
|
|
if(argus.unit_check)
|
|
{
|
|
ROCSOLVER_TEST_CHECK(T, max_error, 2 * n);
|
|
if(nv || nu || nc)
|
|
ROCSOLVER_TEST_CHECK(T, max_errorv, 2 * n);
|
|
}
|
|
|
|
// output results for rocsolver-bench
|
|
if(argus.timing)
|
|
{
|
|
if(!argus.perf)
|
|
{
|
|
if(nv || nu || nc)
|
|
max_error = (max_error >= max_errorv) ? max_error : max_errorv;
|
|
rocsolver_bench_header("Arguments:");
|
|
rocsolver_bench_output("uplo", "n", "nv", "nu", "nc", "ldv", "ldu", "ldc");
|
|
rocsolver_bench_output(uploC, n, nv, nu, nc, ldv, ldu, ldc);
|
|
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_BDSQR(...) extern template void testing_bdsqr<__VA_ARGS__>(Arguments&);
|
|
|
|
INSTANTIATE(EXTERN_TESTING_BDSQR, FOREACH_SCALAR_TYPE, APPLY_STAMP)
|