/* ************************************************************************ * Copyright (C) 2020-2024 Advanced Micro Devices, Inc. All rights reserved. * * Permission is hereby granted, free of charge, to any person obtaining a copy * of this software and associated documentation files (the "Software"), to deal * in the Software without restriction, including without limitation the rights * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell cop- * ies of the Software, and to permit persons to whom the Software is furnished * to do so, subject to the following conditions: * * The above copyright notice and this permission notice shall be included in all * copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IM- * PLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS * FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR * COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER * IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNE- * CTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * * * ************************************************************************ */ #pragma once #include "clientcommon.hpp" template void gebrd_checkBadArgs(const hipsolverHandle_t handle, const int m, const int n, T dA, const int lda, const int stA, S dD, const int stD, S dE, const int stE, U dTauq, const int stQ, U dTaup, const int stP, U dWork, const int lwork, V dInfo, const int bc) { // handle EXPECT_ROCBLAS_STATUS(hipsolver_gebrd(API, nullptr, m, n, dA, lda, stA, dD, stD, dE, stE, dTauq, stQ, dTaup, stP, dWork, lwork, dInfo, bc), HIPSOLVER_STATUS_NOT_INITIALIZED); // values // N/A #if defined(__HIP_PLATFORM_HCC__) || defined(__HIP_PLATFORM_AMD__) // pointers EXPECT_ROCBLAS_STATUS(hipsolver_gebrd(API, handle, m, n, (T) nullptr, lda, stA, dD, stD, dE, stE, dTauq, stQ, dTaup, stP, dWork, lwork, dInfo, bc), HIPSOLVER_STATUS_INVALID_VALUE); EXPECT_ROCBLAS_STATUS(hipsolver_gebrd(API, handle, m, n, dA, lda, stA, (S) nullptr, stD, dE, stE, dTauq, stQ, dTaup, stP, dWork, lwork, dInfo, bc), HIPSOLVER_STATUS_INVALID_VALUE); EXPECT_ROCBLAS_STATUS(hipsolver_gebrd(API, handle, m, n, dA, lda, stA, dD, stD, (S) nullptr, stE, dTauq, stQ, dTaup, stP, dWork, lwork, dInfo, bc), HIPSOLVER_STATUS_INVALID_VALUE); EXPECT_ROCBLAS_STATUS(hipsolver_gebrd(API, handle, m, n, dA, lda, stA, dD, stD, dE, stE, (U) nullptr, stQ, dTaup, stP, dWork, lwork, dInfo, bc), HIPSOLVER_STATUS_INVALID_VALUE); EXPECT_ROCBLAS_STATUS(hipsolver_gebrd(API, handle, m, n, dA, lda, stA, dD, stD, dE, stE, dTauq, stQ, (U) nullptr, stP, dWork, lwork, dInfo, bc), HIPSOLVER_STATUS_INVALID_VALUE); EXPECT_ROCBLAS_STATUS(hipsolver_gebrd(API, handle, m, n, dA, lda, stA, dD, stD, dE, stE, dTauq, stQ, dTaup, stP, dWork, lwork, (V) nullptr, bc), HIPSOLVER_STATUS_INVALID_VALUE); #endif } template void testing_gebrd_bad_arg() { using S = decltype(std::real(T{})); // safe arguments hipsolver_local_handle handle; int m = 2; int n = 2; int lda = 2; int stA = 1; int stD = 1; int stE = 1; int stQ = 1; int stP = 1; int bc = 1; if(BATCHED) { // // memory allocations // device_batch_vector dA(1, 1, 1); // device_strided_batch_vector dD(1, 1, 1, 1); // device_strided_batch_vector dE(1, 1, 1, 1); // device_strided_batch_vector dTauq(1, 1, 1, 1); // device_strided_batch_vector dTaup(1, 1, 1, 1); // device_strided_batch_vector dInfo(1, 1, 1, 1); // CHECK_HIP_ERROR(dA.memcheck()); // CHECK_HIP_ERROR(dD.memcheck()); // CHECK_HIP_ERROR(dE.memcheck()); // CHECK_HIP_ERROR(dTauq.memcheck()); // CHECK_HIP_ERROR(dTaup.memcheck()); // CHECK_HIP_ERROR(dInfo.memcheck()); // int size_W; // hipsolver_gebrd_bufferSize(API, handle, m, n, dA.data(), lda, &size_W); // device_strided_batch_vector dWork(size_W, 1, size_W, 1); // if(size_W) // CHECK_HIP_ERROR(dWork.memcheck()); // // check bad arguments // gebrd_checkBadArgs(handle, // m, // n, // dA.data(), // lda, // stA, // dD.data(), // stD, // dE.data(), // stE, // dTauq.data(), // stQ, // dTaup.data(), // stP, // dWork.data(), // size_W, // dInfo.data(), // bc); } else { // memory allocations device_strided_batch_vector dA(1, 1, 1, 1); device_strided_batch_vector dD(1, 1, 1, 1); device_strided_batch_vector dE(1, 1, 1, 1); device_strided_batch_vector dTauq(1, 1, 1, 1); device_strided_batch_vector dTaup(1, 1, 1, 1); device_strided_batch_vector dInfo(1, 1, 1, 1); CHECK_HIP_ERROR(dA.memcheck()); CHECK_HIP_ERROR(dD.memcheck()); CHECK_HIP_ERROR(dE.memcheck()); CHECK_HIP_ERROR(dTauq.memcheck()); CHECK_HIP_ERROR(dTaup.memcheck()); CHECK_HIP_ERROR(dInfo.memcheck()); int size_W; hipsolver_gebrd_bufferSize(API, handle, m, n, dA.data(), lda, &size_W); device_strided_batch_vector dWork(size_W, 1, size_W, 1); if(size_W) CHECK_HIP_ERROR(dWork.memcheck()); // check bad arguments gebrd_checkBadArgs(handle, m, n, dA.data(), lda, stA, dD.data(), stD, dE.data(), stE, dTauq.data(), stQ, dTaup.data(), stP, dWork.data(), size_W, dInfo.data(), bc); } } template void gebrd_initData(const hipsolverHandle_t handle, const int m, const int n, Td& dA, const int lda, const int stA, Sd& dD, const int stD, Sd& dE, const int stE, Ud& dTauq, const int stQ, Ud& dTaup, const int stP, const int bc, Th& hA, Sh& hD, Sh& hE, Uh& hTauq, Uh& hTaup) { if(CPU) { rocblas_init(hA, true); // scale A to avoid singularities for(int b = 0; b < bc; ++b) { for(int i = 0; i < m; i++) { for(int j = 0; j < n; j++) { if(i == j || (m >= n && j == i + 1) || (m < n && i == j + 1)) hA[b][i + j * lda] += 400; else hA[b][i + j * lda] -= 4; } } } } if(GPU) { // now copy to the GPU CHECK_HIP_ERROR(dA.transfer_from(hA)); } } template void gebrd_getError(const hipsolverHandle_t handle, const int m, const int n, Td& dA, const int lda, const int stA, Sd& dD, const int stD, Sd& dE, const int stE, Ud& dTauq, const int stQ, Ud& dTaup, const int stP, Ud& dWork, const int lwork, Vd& dInfo, const int bc, Th& hA, Th& hARes, Sh& hD, Sh& hE, Uh& hTauq, Uh& hTaup, Vh& hInfo, Vh& hInfoRes, double* max_err) { constexpr bool COMPLEX = is_complex; constexpr bool VERIFY_IMPLICIT_TEST = false; std::vector hW(max(m, n)); // input data initialization gebrd_initData(handle, m, n, dA, lda, stA, dD, stD, dE, stE, dTauq, stQ, dTaup, stP, bc, hA, hD, hE, hTauq, hTaup); // execute computations // use verify_implicit_test to check correctness of the implicit test using // CPU lapack if(!VERIFY_IMPLICIT_TEST) { // GPU lapack CHECK_ROCBLAS_ERROR(hipsolver_gebrd(API, handle, m, n, dA.data(), lda, stA, dD.data(), stD, dE.data(), stE, dTauq.data(), stQ, dTaup.data(), stP, dWork.data(), lwork, dInfo.data(), bc)); CHECK_HIP_ERROR(hARes.transfer_from(dA)); CHECK_HIP_ERROR(hTauq.transfer_from(dTauq)); CHECK_HIP_ERROR(hTaup.transfer_from(dTaup)); CHECK_HIP_ERROR(hInfoRes.transfer_from(dInfo)); } else { // CPU lapack for(int b = 0; b < bc; ++b) { memcpy(hARes[b], hA[b], lda * n * sizeof(T)); cpu_gebrd(m, n, hARes[b], lda, hD[b], hE[b], hTauq[b], hTaup[b], hW.data(), max(m, n), hInfoRes[b]); } } // reconstruct A from the factorization for implicit testing std::vector vec(max(m, n)); vec[0] = 1; for(int b = 0; b < bc; ++b) { T* a = hARes[b]; T* tauq = hTauq[b]; T* taup = hTaup[b]; if(m >= n) { for(int j = n - 1; j >= 0; j--) { if(j < n - 1) { if(COMPLEX) { cpu_lacgv(1, taup + j, 1); cpu_lacgv(n - j - 1, a + j + (j + 1) * lda, lda); } for(int i = 1; i < n - j - 1; i++) { vec[i] = a[j + (j + i + 1) * lda]; a[j + (j + i + 1) * lda] = 0; } cpu_larf(HIPSOLVER_SIDE_RIGHT, m - j, n - j - 1, vec.data(), 1, taup + j, a + j + (j + 1) * lda, lda, hW.data()); if(COMPLEX) cpu_lacgv(1, taup + j, 1); } for(int i = 1; i < m - j; i++) { vec[i] = a[(j + i) + j * lda]; a[(j + i) + j * lda] = 0; } cpu_larf(HIPSOLVER_SIDE_LEFT, m - j, n - j, vec.data(), 1, tauq + j, a + j + j * lda, lda, hW.data()); } } else { for(int j = m - 1; j >= 0; j--) { if(j < m - 1) { for(int i = 1; i < m - j - 1; i++) { vec[i] = a[(j + i + 1) + j * lda]; a[(j + i + 1) + j * lda] = 0; } cpu_larf(HIPSOLVER_SIDE_LEFT, m - j - 1, n - j, vec.data(), 1, tauq + j, a + (j + 1) + j * lda, lda, hW.data()); } if(COMPLEX) { cpu_lacgv(1, taup + j, 1); cpu_lacgv(n - j, a + j + j * lda, lda); } for(int i = 1; i < n - j; i++) { vec[i] = a[j + (j + i) * lda]; a[j + (j + i) * lda] = 0; } cpu_larf(HIPSOLVER_SIDE_RIGHT, m - j, n - j, vec.data(), 1, taup + j, a + j + j * lda, lda, hW.data()); if(COMPLEX) cpu_lacgv(1, taup + j, 1); } } } // error is ||hA - hARes|| / ||hA|| // using frobenius norm double err; *max_err = 0; for(int b = 0; b < bc; ++b) { err = norm_error('F', m, n, lda, hA[b], hARes[b]); *max_err = err > *max_err ? err : *max_err; } // check info err = 0; for(int b = 0; b < bc; ++b) { EXPECT_EQ(hInfoRes[b][0], 0) << "where b = " << b; if(hInfoRes[b][0] != 0) err++; } *max_err += err; } template void gebrd_getPerfData(const hipsolverHandle_t handle, const int m, const int n, Td& dA, const int lda, const int stA, Sd& dD, const int stD, Sd& dE, const int stE, Ud& dTauq, const int stQ, Ud& dTaup, const int stP, Ud& dWork, const int lwork, Vd& dInfo, const int bc, Th& hA, Sh& hD, Sh& hE, Uh& hTauq, Uh& hTaup, Vh& hInfo, double* gpu_time_used, double* cpu_time_used, const int hot_calls, const bool perf) { std::vector hW(max(m, n)); if(!perf) { gebrd_initData(handle, m, n, dA, lda, stA, dD, stD, dE, stE, dTauq, stQ, dTaup, stP, bc, hA, hD, hE, hTauq, hTaup); // cpu-lapack performance (only if not in perf mode) *cpu_time_used = get_time_us_no_sync(); for(int b = 0; b < bc; ++b) cpu_gebrd( m, n, hA[b], lda, hD[b], hE[b], hTauq[b], hTaup[b], hW.data(), max(m, n), hInfo[b]); *cpu_time_used = get_time_us_no_sync() - *cpu_time_used; } gebrd_initData(handle, m, n, dA, lda, stA, dD, stD, dE, stE, dTauq, stQ, dTaup, stP, bc, hA, hD, hE, hTauq, hTaup); // cold calls for(int iter = 0; iter < 2; iter++) { gebrd_initData(handle, m, n, dA, lda, stA, dD, stD, dE, stE, dTauq, stQ, dTaup, stP, bc, hA, hD, hE, hTauq, hTaup); CHECK_ROCBLAS_ERROR(hipsolver_gebrd(API, handle, m, n, dA.data(), lda, stA, dD.data(), stD, dE.data(), stE, dTauq.data(), stQ, dTaup.data(), stP, dWork.data(), lwork, dInfo.data(), bc)); } // gpu-lapack performance hipStream_t stream; CHECK_ROCBLAS_ERROR(hipsolverGetStream(handle, &stream)); double start; for(int iter = 0; iter < hot_calls; iter++) { gebrd_initData(handle, m, n, dA, lda, stA, dD, stD, dE, stE, dTauq, stQ, dTaup, stP, bc, hA, hD, hE, hTauq, hTaup); start = get_time_us_sync(stream); hipsolver_gebrd(API, handle, m, n, dA.data(), lda, stA, dD.data(), stD, dE.data(), stE, dTauq.data(), stQ, dTaup.data(), stP, dWork.data(), lwork, dInfo.data(), bc); *gpu_time_used += get_time_us_sync(stream) - start; } *gpu_time_used /= hot_calls; } template void testing_gebrd(Arguments& argus) { using S = decltype(std::real(T{})); // get arguments hipsolver_local_handle handle; int m = argus.get("m"); int n = argus.get("n", m); int lda = argus.get("lda", m); int stA = argus.get("strideA", lda * n); int stD = argus.get("strideD", min(m, n)); int stE = argus.get("strideE", min(m, n) - 1); int stQ = argus.get("strideQ", min(m, n)); int stP = argus.get("strideP", min(m, n)); int bc = argus.batch_count; int hot_calls = argus.iters; int stARes = (argus.unit_check || argus.norm_check) ? stA : 0; // check non-supported values // N/A // determine sizes size_t size_A = lda * n; size_t size_D = min(m, n); size_t size_E = min(m, n) - 1; size_t size_Q = min(m, n); size_t size_P = min(m, n); double max_error = 0, gpu_time_used = 0, cpu_time_used = 0; size_t size_ARes = (argus.unit_check || argus.norm_check) ? size_A : 0; // check invalid sizes bool invalid_size = (m < 0 || n < 0 || lda < m || bc < 0); if(invalid_size) { if(BATCHED) { // EXPECT_ROCBLAS_STATUS(hipsolver_gebrd(API, // handle, // m, // n, // (T* const*)nullptr, // lda, // stA, // (S*)nullptr, // stD, // (S*)nullptr, // stE, // (T*)nullptr, // stQ, // (T*)nullptr, // stP, // (T*)nullptr, // 0, // (int*)nullptr, // bc), // HIPSOLVER_STATUS_INVALID_VALUE); } else { EXPECT_ROCBLAS_STATUS(hipsolver_gebrd(API, handle, m, n, (T*)nullptr, lda, stA, (S*)nullptr, stD, (S*)nullptr, stE, (T*)nullptr, stQ, (T*)nullptr, stP, (T*)nullptr, 0, (int*)nullptr, bc), HIPSOLVER_STATUS_INVALID_VALUE); } if(argus.timing) rocsolver_bench_inform(inform_invalid_size); return; } // memory size query is necessary int size_W; hipsolver_gebrd_bufferSize(API, handle, m, n, (T*)nullptr, lda, &size_W); if(argus.mem_query) { rocsolver_bench_inform(inform_mem_query, size_W); return; } if(BATCHED) { // // memory allocations // host_batch_vector hA(size_A, 1, bc); // host_batch_vector hARes(size_ARes, 1, bc); // host_strided_batch_vector hD(size_D, 1, stD, bc); // host_strided_batch_vector hE(size_E, 1, stE, bc); // host_strided_batch_vector hTaup(size_P, 1, stP, bc); // host_strided_batch_vector hTauq(size_Q, 1, stQ, bc); // host_strided_batch_vector hInfo(1, 1, 1, bc); // host_strided_batch_vector hInfoRes(1, 1, 1, bc); // device_batch_vector dA(size_A, 1, bc); // device_strided_batch_vector dD(size_D, 1, stD, bc); // device_strided_batch_vector dE(size_E, 1, stE, bc); // device_strided_batch_vector dTauq(size_Q, 1, stQ, bc); // device_strided_batch_vector dTaup(size_P, 1, stP, bc); // device_strided_batch_vector dInfo(1, 1, 1, bc); // device_strided_batch_vector dWork(size_W, 1, size_W, 1); // size_W accounts for bc // if(size_A) // CHECK_HIP_ERROR(dA.memcheck()); // if(size_D) // CHECK_HIP_ERROR(dD.memcheck()); // if(size_E) // CHECK_HIP_ERROR(dE.memcheck()); // if(size_Q) // CHECK_HIP_ERROR(dTauq.memcheck()); // if(size_P) // CHECK_HIP_ERROR(dTaup.memcheck()); // CHECK_HIP_ERROR(dInfo.memcheck()); // if(size_W) // CHECK_HIP_ERROR(dWork.memcheck()); // // check computations // if(argus.unit_check || argus.norm_check) // gebrd_getError(handle, // m, // n, // dA, // lda, // stA, // dD, // stD, // dE, // stE, // dTauq, // stQ, // dTaup, // stP, // dWork, // size_W, // dInfo, // bc, // hA, // hARes, // hD, // hE, // hTauq, // hTaup, // hInfo, // hInfoRes, // &max_error); // // collect performance data // if(argus.timing) // gebrd_getPerfData(handle, // m, // n, // dA, // lda, // stA, // dD, // stD, // dE, // stE, // dTauq, // stQ, // dTaup, // stP, // dWork, // size_W, // dInfo, // bc, // hA, // hD, // hE, // hTauq, // hTaup, // hInfo, // &gpu_time_used, // &cpu_time_used, // hot_calls, // argus.perf); } else { // memory allocations host_strided_batch_vector hA(size_A, 1, stA, bc); host_strided_batch_vector hARes(size_ARes, 1, stARes, bc); host_strided_batch_vector hD(size_D, 1, stD, bc); host_strided_batch_vector hE(size_E, 1, stE, bc); host_strided_batch_vector hTaup(size_P, 1, stP, bc); host_strided_batch_vector hTauq(size_Q, 1, stQ, bc); host_strided_batch_vector hInfo(1, 1, 1, bc); host_strided_batch_vector hInfoRes(1, 1, 1, bc); device_strided_batch_vector dA(size_A, 1, stA, bc); device_strided_batch_vector dD(size_D, 1, stD, bc); device_strided_batch_vector dE(size_E, 1, stE, bc); device_strided_batch_vector dTauq(size_Q, 1, stQ, bc); device_strided_batch_vector dTaup(size_P, 1, stP, bc); device_strided_batch_vector dInfo(1, 1, 1, bc); device_strided_batch_vector dWork(size_W, 1, size_W, 1); // size_W accounts for bc if(size_A) CHECK_HIP_ERROR(dA.memcheck()); if(size_D) CHECK_HIP_ERROR(dD.memcheck()); if(size_E) CHECK_HIP_ERROR(dE.memcheck()); if(size_Q) CHECK_HIP_ERROR(dTauq.memcheck()); if(size_P) CHECK_HIP_ERROR(dTaup.memcheck()); CHECK_HIP_ERROR(dInfo.memcheck()); if(size_W) CHECK_HIP_ERROR(dWork.memcheck()); // check computations if(argus.unit_check || argus.norm_check) gebrd_getError(handle, m, n, dA, lda, stA, dD, stD, dE, stE, dTauq, stQ, dTaup, stP, dWork, size_W, dInfo, bc, hA, hARes, hD, hE, hTauq, hTaup, hInfo, hInfoRes, &max_error); // collect performance data if(argus.timing) gebrd_getPerfData(handle, m, n, dA, lda, stA, dD, stD, dE, stE, dTauq, stQ, dTaup, stP, dWork, size_W, dInfo, bc, hA, hD, hE, hTauq, hTaup, hInfo, &gpu_time_used, &cpu_time_used, hot_calls, argus.perf); } // validate results for rocsolver-test // using m*n * machine_precision as tolerance if(argus.unit_check) ROCSOLVER_TEST_CHECK(T, max_error, m * n); // output results for rocsolver-bench if(argus.timing) { if(!argus.perf) { std::cerr << "\n============================================\n"; std::cerr << "Arguments:\n"; std::cerr << "============================================\n"; if(BATCHED) { rocsolver_bench_output("m", "n", "lda", "strideP", "batch_c"); rocsolver_bench_output(m, n, lda, stP, bc); } else if(STRIDED) { rocsolver_bench_output("m", "n", "lda", "strideA", "strideP", "batch_c"); rocsolver_bench_output(m, n, lda, stA, stP, bc); } else { rocsolver_bench_output("m", "n", "lda"); rocsolver_bench_output(m, n, lda); } std::cerr << "\n============================================\n"; std::cerr << "Results:\n"; std::cerr << "============================================\n"; if(argus.norm_check) { rocsolver_bench_output("cpu_time", "gpu_time", "error"); rocsolver_bench_output(cpu_time_used, gpu_time_used, max_error); } else { rocsolver_bench_output("cpu_time", "gpu_time"); rocsolver_bench_output(cpu_time_used, gpu_time_used); } std::cerr << std::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(); }