/* ************************************************************************ * Copyright (C) 2020-2025 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 ormtr_unmtr_checkBadArgs(const hipsolverHandle_t handle, const hipsolverSideMode_t side, const hipsolverFillMode_t uplo, const hipsolverOperation_t trans, const int m, const int n, T dA, const int lda, T dIpiv, T dC, const int ldc, T dWork, const int lwork, U dInfo) { // handle EXPECT_ROCBLAS_STATUS( hipsolver_ormtr_unmtr( API, nullptr, side, uplo, trans, m, n, dA, lda, dIpiv, dC, ldc, dWork, lwork, dInfo), HIPSOLVER_STATUS_NOT_INITIALIZED); // values EXPECT_ROCBLAS_STATUS(hipsolver_ormtr_unmtr(API, handle, hipsolverSideMode_t(-1), uplo, trans, m, n, dA, lda, dIpiv, dC, ldc, dWork, lwork, dInfo), HIPSOLVER_STATUS_INVALID_ENUM); EXPECT_ROCBLAS_STATUS(hipsolver_ormtr_unmtr(API, handle, side, hipsolverFillMode_t(-1), trans, m, n, dA, lda, dIpiv, dC, ldc, dWork, lwork, dInfo), HIPSOLVER_STATUS_INVALID_ENUM); EXPECT_ROCBLAS_STATUS(hipsolver_ormtr_unmtr(API, handle, side, uplo, hipsolverOperation_t(-1), m, n, dA, lda, dIpiv, dC, ldc, dWork, lwork, dInfo), HIPSOLVER_STATUS_INVALID_ENUM); if(COMPLEX) EXPECT_ROCBLAS_STATUS(hipsolver_ormtr_unmtr(API, handle, side, uplo, HIPSOLVER_OP_T, m, n, dA, lda, dIpiv, dC, ldc, dWork, lwork, dInfo), HIPSOLVER_STATUS_INVALID_VALUE); else EXPECT_ROCBLAS_STATUS(hipsolver_ormtr_unmtr(API, handle, side, uplo, HIPSOLVER_OP_C, m, n, dA, lda, dIpiv, dC, ldc, dWork, lwork, dInfo), HIPSOLVER_STATUS_INVALID_VALUE); #if defined(__HIP_PLATFORM_HCC__) || defined(__HIP_PLATFORM_AMD__) // pointers EXPECT_ROCBLAS_STATUS(hipsolver_ormtr_unmtr(API, handle, side, uplo, trans, m, n, (T) nullptr, lda, dIpiv, dC, ldc, dWork, lwork, dInfo), HIPSOLVER_STATUS_INVALID_VALUE); EXPECT_ROCBLAS_STATUS(hipsolver_ormtr_unmtr(API, handle, side, uplo, trans, m, n, dA, lda, (T) nullptr, dC, ldc, dWork, lwork, dInfo), HIPSOLVER_STATUS_INVALID_VALUE); EXPECT_ROCBLAS_STATUS(hipsolver_ormtr_unmtr(API, handle, side, uplo, trans, m, n, dA, lda, dIpiv, (T) nullptr, ldc, dWork, lwork, dInfo), HIPSOLVER_STATUS_INVALID_VALUE); EXPECT_ROCBLAS_STATUS(hipsolver_ormtr_unmtr(API, handle, side, uplo, trans, m, n, dA, lda, dIpiv, dC, ldc, dWork, lwork, (U) nullptr), HIPSOLVER_STATUS_INVALID_VALUE); #endif } template > void testing_ormtr_unmtr_bad_arg() { // safe arguments hipsolver_local_handle handle; hipsolverSideMode_t side = HIPSOLVER_SIDE_LEFT; hipsolverFillMode_t uplo = HIPSOLVER_FILL_MODE_UPPER; hipsolverOperation_t trans = HIPSOLVER_OP_N; int m = 2; int n = 2; int lda = 2; int ldc = 2; // memory allocation device_strided_batch_vector dA(1, 1, 1, 1); device_strided_batch_vector dIpiv(1, 1, 1, 1); device_strided_batch_vector dC(1, 1, 1, 1); device_strided_batch_vector dInfo(1, 1, 1, 1); CHECK_HIP_ERROR(dA.memcheck()); CHECK_HIP_ERROR(dIpiv.memcheck()); CHECK_HIP_ERROR(dC.memcheck()); CHECK_HIP_ERROR(dInfo.memcheck()); int size_W; hipsolver_ormtr_unmtr_bufferSize(API, handle, side, uplo, trans, m, n, dA.data(), lda, dIpiv.data(), dC.data(), ldc, &size_W); size_t bytes_W = std::getenv("HIPSOLVER_BUFFERSIZE_RETURN_BYTES") != nullptr ? size_W : sizeof(T) * size_W; device_strided_batch_vector dWork(bytes_W, 1, bytes_W, 1); if(size_W) CHECK_HIP_ERROR(dWork.memcheck()); // check bad arguments ormtr_unmtr_checkBadArgs(handle, side, uplo, trans, m, n, dA.data(), lda, dIpiv.data(), dC.data(), ldc, dWork.data(), size_W, dInfo.data()); } template void ormtr_unmtr_initData(const hipsolverHandle_t handle, const hipsolverSideMode_t side, const hipsolverFillMode_t uplo, const hipsolverOperation_t trans, const int m, const int n, Td& dA, const int lda, Td& dIpiv, Td& dC, const int ldc, Th& hA, Th& hIpiv, Th& hC, std::vector& hW, size_t size_W) { if(CPU) { using S = decltype(std::real(T{})); int nq = (side == HIPSOLVER_SIDE_LEFT) ? m : n; std::vector E(nq - 1); std::vector D(nq); rocblas_init(hA, true); rocblas_init(hIpiv, true); rocblas_init(hC, true); // scale to avoid singularities for(int i = 0; i < nq; ++i) { for(int j = 0; j < nq; ++j) { if(i == j) hA[0][i + j * lda] += 400; else hA[0][i + j * lda] -= 4; } } // compute sytrd/hetrd cpu_sytrd_hetrd(uplo, nq, hA[0], lda, D.data(), E.data(), hIpiv[0], hW.data(), size_W); } if(GPU) { // copy data from CPU to device CHECK_HIP_ERROR(dA.transfer_from(hA)); CHECK_HIP_ERROR(dIpiv.transfer_from(hIpiv)); CHECK_HIP_ERROR(dC.transfer_from(hC)); } } template void ormtr_unmtr_getError(const hipsolverHandle_t handle, const hipsolverSideMode_t side, const hipsolverFillMode_t uplo, const hipsolverOperation_t trans, const int m, const int n, Td& dA, const int lda, Td& dIpiv, Td& dC, const int ldc, Td& dWork, const int lwork, Ud& dInfo, Th& hA, Th& hIpiv, Th& hC, Th& hCRes, Uh& hInfo, Uh& hInfoRes, double* max_err) { size_t size_W = (side == HIPSOLVER_SIDE_LEFT ? m : n) * 32; std::vector hW(size_W); // initialize data ormtr_unmtr_initData( handle, side, uplo, trans, m, n, dA, lda, dIpiv, dC, ldc, hA, hIpiv, hC, hW, size_W); // execute computations // GPU lapack CHECK_ROCBLAS_ERROR(hipsolver_ormtr_unmtr(API, handle, side, uplo, trans, m, n, dA.data(), lda, dIpiv.data(), dC.data(), ldc, dWork.data(), lwork, dInfo.data())); CHECK_HIP_ERROR(hCRes.transfer_from(dC)); CHECK_HIP_ERROR(hInfoRes.transfer_from(dInfo)); // CPU lapack cpu_ormtr_unmtr( side, uplo, trans, m, n, hA[0], lda, hIpiv[0], hC[0], ldc, hW.data(), size_W, hInfo[0]); // error is ||hC - hCr|| / ||hC|| // (THIS DOES NOT ACCOUNT FOR NUMERICAL REPRODUCIBILITY ISSUES. // IT MIGHT BE REVISITED IN THE FUTURE) // using frobenius norm *max_err = norm_error('F', m, n, ldc, hC[0], hCRes[0]); // check info EXPECT_EQ(hInfo[0][0], hInfoRes[0][0]); if(hInfo[0][0] != hInfoRes[0][0]) *max_err += 1; } template void ormtr_unmtr_getPerfData(const hipsolverHandle_t handle, const hipsolverSideMode_t side, const hipsolverFillMode_t uplo, const hipsolverOperation_t trans, const int m, const int n, Td& dA, const int lda, Td& dIpiv, Td& dC, const int ldc, Td& dWork, const int lwork, Ud& dInfo, Th& hA, Th& hIpiv, Th& hC, Uh& hInfo, double* gpu_time_used, double* cpu_time_used, const int hot_calls, const bool perf) { size_t size_W = (side == HIPSOLVER_SIDE_LEFT ? m : n) * 32; std::vector hW(size_W); if(!perf) { ormtr_unmtr_initData( handle, side, uplo, trans, m, n, dA, lda, dIpiv, dC, ldc, hA, hIpiv, hC, hW, size_W); // cpu-lapack performance (only if not in perf mode) *cpu_time_used = get_time_us_no_sync(); cpu_ormtr_unmtr( side, uplo, trans, m, n, hA[0], lda, hIpiv[0], hC[0], ldc, hW.data(), size_W, hInfo[0]); *cpu_time_used = get_time_us_no_sync() - *cpu_time_used; } ormtr_unmtr_initData( handle, side, uplo, trans, m, n, dA, lda, dIpiv, dC, ldc, hA, hIpiv, hC, hW, size_W); // cold calls for(int iter = 0; iter < 2; iter++) { ormtr_unmtr_initData( handle, side, uplo, trans, m, n, dA, lda, dIpiv, dC, ldc, hA, hIpiv, hC, hW, size_W); CHECK_ROCBLAS_ERROR(hipsolver_ormtr_unmtr(API, handle, side, uplo, trans, m, n, dA.data(), lda, dIpiv.data(), dC.data(), ldc, dWork.data(), lwork, dInfo.data())); } // gpu-lapack performance hipStream_t stream; CHECK_ROCBLAS_ERROR(hipsolverGetStream(handle, &stream)); double start; for(int iter = 0; iter < hot_calls; iter++) { ormtr_unmtr_initData( handle, side, uplo, trans, m, n, dA, lda, dIpiv, dC, ldc, hA, hIpiv, hC, hW, size_W); start = get_time_us_sync(stream); hipsolver_ormtr_unmtr(API, handle, side, uplo, trans, m, n, dA.data(), lda, dIpiv.data(), dC.data(), ldc, dWork.data(), lwork, dInfo.data()); *gpu_time_used += get_time_us_sync(stream) - start; } *gpu_time_used /= hot_calls; } template > void testing_ormtr_unmtr(Arguments& argus) { // get arguments hipsolver_local_handle handle; char sideC = argus.get("side"); char uploC = argus.get("uplo"); char transC = argus.get("trans"); int m, n; if(sideC == 'L') { m = argus.get("m"); n = argus.get("n", m); } else { n = argus.get("n"); m = argus.get("m", n); } int nq = (sideC == 'L' ? m : n); int lda = argus.get("lda", nq); int ldc = argus.get("ldc", m); hipsolverSideMode_t side = char2hipsolver_side(sideC); hipsolverFillMode_t uplo = char2hipsolver_fill(uploC); hipsolverOperation_t trans = char2hipsolver_operation(transC); int hot_calls = argus.iters; // check non-supported values bool invalid_value = ((COMPLEX && trans == HIPSOLVER_OP_T) || (!COMPLEX && trans == HIPSOLVER_OP_C)); if(invalid_value) { EXPECT_ROCBLAS_STATUS(hipsolver_ormtr_unmtr(API, handle, side, uplo, trans, m, n, (T*)nullptr, lda, (T*)nullptr, (T*)nullptr, ldc, (T*)nullptr, 0, (int*)nullptr), HIPSOLVER_STATUS_INVALID_VALUE); if(argus.timing) rocsolver_bench_inform(inform_invalid_args); return; } // determine sizes bool left = (side == HIPSOLVER_SIDE_LEFT); size_t size_P = size_t(nq); size_t size_C = size_t(ldc) * n; size_t size_A = size_t(lda) * nq; double max_error = 0, gpu_time_used = 0, cpu_time_used = 0; size_t size_CRes = (argus.unit_check || argus.norm_check) ? size_C : 0; // check invalid sizes bool invalid_size = (m < 0 || n < 0 || ldc < m || lda < nq); if(invalid_size) { EXPECT_ROCBLAS_STATUS(hipsolver_ormtr_unmtr(API, handle, side, uplo, trans, m, n, (T*)nullptr, lda, (T*)nullptr, (T*)nullptr, ldc, (T*)nullptr, 0, (int*)nullptr), HIPSOLVER_STATUS_INVALID_VALUE); if(argus.timing) rocsolver_bench_inform(inform_invalid_size); return; } // memory size query is necessary int size_W; hipsolver_ormtr_unmtr_bufferSize(API, handle, side, uplo, trans, m, n, (T*)nullptr, lda, (T*)nullptr, (T*)nullptr, ldc, &size_W); size_t bytes_W = std::getenv("HIPSOLVER_BUFFERSIZE_RETURN_BYTES") != nullptr ? size_W : sizeof(T) * size_W; if(argus.mem_query) { rocsolver_bench_inform(inform_mem_query, bytes_W); return; } // memory allocations host_strided_batch_vector hC(size_C, 1, size_C, 1); host_strided_batch_vector hCRes(size_CRes, 1, size_CRes, 1); host_strided_batch_vector hIpiv(size_P, 1, size_P, 1); host_strided_batch_vector hA(size_A, 1, size_A, 1); host_strided_batch_vector hInfo(1, 1, 1, 1); host_strided_batch_vector hInfoRes(1, 1, 1, 1); device_strided_batch_vector dC(size_C, 1, size_C, 1); device_strided_batch_vector dIpiv(size_P, 1, size_P, 1); device_strided_batch_vector dA(size_A, 1, size_A, 1); device_strided_batch_vector dInfo(1, 1, 1, 1); device_strided_batch_vector dWork(bytes_W, 1, bytes_W, 1); if(size_A) CHECK_HIP_ERROR(dA.memcheck()); if(size_P) CHECK_HIP_ERROR(dIpiv.memcheck()); if(size_C) CHECK_HIP_ERROR(dC.memcheck()); CHECK_HIP_ERROR(dInfo.memcheck()); if(size_W) CHECK_HIP_ERROR(dWork.memcheck()); // check computations if(argus.unit_check || argus.norm_check) ormtr_unmtr_getError(handle, side, uplo, trans, m, n, dA, lda, dIpiv, dC, ldc, dWork, size_W, dInfo, hA, hIpiv, hC, hCRes, hInfo, hInfoRes, &max_error); // collect performance data if(argus.timing) ormtr_unmtr_getPerfData(handle, side, uplo, trans, m, n, dA, lda, dIpiv, dC, ldc, dWork, size_W, dInfo, hA, hIpiv, hC, hInfo, &gpu_time_used, &cpu_time_used, hot_calls, argus.perf); // validate results for rocsolver-test // using s * machine_precision as tolerance int s = left ? m : n; if(argus.unit_check) ROCSOLVER_TEST_CHECK(T, max_error, s); // output results for rocsolver-bench if(argus.timing) { if(!argus.perf) { std::cerr << "\n============================================\n"; std::cerr << "Arguments:\n"; std::cerr << "============================================\n"; rocsolver_bench_output("side", "uplo", "trans", "m", "n", "lda", "ldc"); rocsolver_bench_output(sideC, uploC, transC, m, n, lda, ldc); 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(); }