/* ************************************************************************ * Copyright 2020-2022 Advanced Micro Devices, Inc. * ************************************************************************ */ #pragma once #include "clientcommon.hpp" template void gesv_checkBadArgs(const hipsolverHandle_t handle, const int n, const int nrhs, T dA, const int lda, const int stA, U dIpiv, const int stP, T dB, const int ldb, const int stB, T dX, const int ldx, const int stX, T dWork, const size_t lwork, U niters, U dInfo, const int bc) { #if defined(__HIP_PLATFORM_HCC__) || defined(__HIP_PLATFORM_AMD__) // handle EXPECT_ROCBLAS_STATUS(hipsolver_gesv(API, false, nullptr, n, nrhs, dA, lda, stA, dIpiv, stP, dB, ldb, stB, dX, ldx, stX, dWork, lwork, niters, dInfo, bc), HIPSOLVER_STATUS_NOT_INITIALIZED); // values // N/A // pointers EXPECT_ROCBLAS_STATUS(hipsolver_gesv(API, false, handle, n, nrhs, (T) nullptr, lda, stA, dIpiv, stP, dB, ldb, stB, dX, ldx, stX, dWork, lwork, niters, dInfo, bc), HIPSOLVER_STATUS_INVALID_VALUE); EXPECT_ROCBLAS_STATUS(hipsolver_gesv(API, false, handle, n, nrhs, dA, lda, stA, (U) nullptr, stP, dB, ldb, stB, dX, ldx, stX, dWork, lwork, niters, dInfo, bc), HIPSOLVER_STATUS_INVALID_VALUE); EXPECT_ROCBLAS_STATUS(hipsolver_gesv(API, false, handle, n, nrhs, dA, lda, stA, dIpiv, stP, (T) nullptr, ldb, stB, dX, ldx, stX, dWork, lwork, niters, dInfo, bc), HIPSOLVER_STATUS_INVALID_VALUE); EXPECT_ROCBLAS_STATUS(hipsolver_gesv(API, false, handle, n, nrhs, dA, lda, stA, dIpiv, stP, dB, ldb, stB, (T) nullptr, ldx, stX, dWork, lwork, niters, dInfo, bc), HIPSOLVER_STATUS_INVALID_VALUE); EXPECT_ROCBLAS_STATUS(hipsolver_gesv(API, false, handle, n, nrhs, dA, lda, stA, dIpiv, stP, dB, ldb, stB, dX, ldx, stX, dWork, lwork, niters, (U) nullptr, bc), HIPSOLVER_STATUS_INVALID_VALUE); #endif } template void testing_gesv_bad_arg() { // safe arguments hipsolver_local_handle handle; int n = 1; int nrhs = 1; int lda = 1; int ldb = 1; int ldx = 1; int stA = 1; int stP = 1; int stB = 1; int stX = 1; int bc = 1; if(BATCHED) { // // memory allocations // host_strided_batch_vector hNiters(1, 1, 1, 1); // device_batch_vector dA(1, 1, 1); // device_batch_vector dB(1, 1, 1); // device_batch_vector dX(1, 1, 1); // device_strided_batch_vector dIpiv(1, 1, 1, 1); // device_strided_batch_vector dInfo(1, 1, 1, 1); // CHECK_HIP_ERROR(dA.memcheck()); // CHECK_HIP_ERROR(dB.memcheck()); // CHECK_HIP_ERROR(dX.memcheck()); // CHECK_HIP_ERROR(dIpiv.memcheck()); // CHECK_HIP_ERROR(dInfo.memcheck()); // size_t size_W; // hipsolver_gesv_bufferSize(API, // handle, // n, // nrhs, // dA.data(), // lda, // dIpiv.data(), // dB.data(), // ldb, // dX.data(), // ldx, // &size_W); // device_strided_batch_vector dWork(size_W, 1, size_W, bc); // if(size_W) // CHECK_HIP_ERROR(dWork.memcheck()); // // check bad arguments // gesv_checkBadArgs(handle, // n, // nrhs, // dA.data(), // lda, // stA, // dIpiv.data(), // stP, // dB.data(), // ldb, // stB, // dX.data(), // ldx, // stX, // dWork.data(), // size_W, // hNiters.data(), // dInfo.data(), // bc); } else { // memory allocations host_strided_batch_vector hNiters(1, 1, 1, 1); device_strided_batch_vector dA(1, 1, 1, 1); device_strided_batch_vector dB(1, 1, 1, 1); device_strided_batch_vector dX(1, 1, 1, 1); device_strided_batch_vector dIpiv(1, 1, 1, 1); device_strided_batch_vector dInfo(1, 1, 1, 1); CHECK_HIP_ERROR(dA.memcheck()); CHECK_HIP_ERROR(dB.memcheck()); CHECK_HIP_ERROR(dX.memcheck()); CHECK_HIP_ERROR(dIpiv.memcheck()); CHECK_HIP_ERROR(dInfo.memcheck()); size_t size_W; hipsolver_gesv_bufferSize(API, handle, n, nrhs, dA.data(), lda, dIpiv.data(), dB.data(), ldb, dX.data(), ldx, &size_W); device_strided_batch_vector dWork(size_W, 1, size_W, bc); if(size_W) CHECK_HIP_ERROR(dWork.memcheck()); // check bad arguments gesv_checkBadArgs(handle, n, nrhs, dA.data(), lda, stA, dIpiv.data(), stP, dB.data(), ldb, stB, dX.data(), ldx, stX, dWork.data(), size_W, hNiters.data(), dInfo.data(), bc); } } template void gesv_initData(const hipsolverHandle_t handle, const int n, const int nrhs, Td& dA, const int lda, const int stA, Ud& dIpiv, const int stP, Td& dB, const int ldb, const int stB, const int bc, Th& hA, Uh& hIpiv, Th& hB) { if(CPU) { rocblas_init(hA, true); rocblas_init(hB, true); // scale A to avoid singularities for(int b = 0; b < bc; ++b) { for(int i = 0; i < n; i++) { for(int j = 0; j < n; j++) { if(i == j) hA[b][i + j * lda] += 400; else hA[b][i + j * lda] -= 4; } } } } if(GPU) { // now copy pivoting indices and matrices to the GPU CHECK_HIP_ERROR(dA.transfer_from(hA)); CHECK_HIP_ERROR(dB.transfer_from(hB)); CHECK_HIP_ERROR(dIpiv.transfer_from(hIpiv)); } } template void gesv_getError(const hipsolverHandle_t handle, const int n, const int nrhs, Td& dA, const int lda, const int stA, Ud& dIpiv, const int stP, Td& dB, const int ldb, const int stB, Td& dX, const int ldx, const int stX, Td& dWork, const size_t lwork, Ud& dInfo, const int bc, Th& hA, Uh& hIpiv, Th& hB, Th& hBRes, Uh& hNiters, Uh& hInfo, Uh& hInfoRes, double* max_err) { // input data initialization gesv_initData( handle, n, nrhs, dA, lda, stA, dIpiv, stP, dB, ldb, stB, bc, hA, hIpiv, hB); // execute computations // GPU lapack CHECK_ROCBLAS_ERROR(hipsolver_gesv(API, INPLACE, handle, n, nrhs, dA.data(), lda, stA, dIpiv.data(), stP, dB.data(), ldb, stB, dX.data(), ldx, stX, dWork.data(), lwork, hNiters.data(), dInfo.data(), bc)); if(!INPLACE) CHECK_HIP_ERROR(hBRes.transfer_from(dX)); else CHECK_HIP_ERROR(hBRes.transfer_from(dB)); CHECK_HIP_ERROR(hInfoRes.transfer_from(dInfo)); // CPU lapack for(int b = 0; b < bc; ++b) { cblas_gesv(n, nrhs, hA[b], lda, hIpiv[b], hB[b], ldb, hInfo[b]); } // error is ||hB - hBRes|| / ||hB|| // (THIS DOES NOT ACCOUNT FOR NUMERICAL REPRODUCIBILITY ISSUES. // IT MIGHT BE REVISITED IN THE FUTURE) // using vector-induced infinity norm double err; *max_err = 0; for(int b = 0; b < bc; ++b) { if(hInfoRes[b][0] == 0) { err = norm_error('I', n, nrhs, ldb, hB[b], hBRes[b], (!INPLACE ? ldx : ldb)); *max_err = err > *max_err ? err : *max_err; } } // also check info for singularities err = 0; for(int b = 0; b < bc; ++b) if(hInfo[b][0] != hInfoRes[b][0]) err++; *max_err += err; } template void gesv_getPerfData(const hipsolverHandle_t handle, const int n, const int nrhs, Td& dA, const int lda, const int stA, Ud& dIpiv, const int stP, Td& dB, const int ldb, const int stB, Td& dX, const int ldx, const int stX, Td& dWork, const size_t lwork, Ud& dInfo, const int bc, Th& hA, Uh& hIpiv, Th& hB, Uh& hNiters, Uh& hInfo, double* gpu_time_used, double* cpu_time_used, const int hot_calls, const bool perf) { if(!perf) { gesv_initData( handle, n, nrhs, dA, lda, stA, dIpiv, stP, dB, ldb, stB, bc, hA, hIpiv, hB); // cpu-lapack performance (only if not in perf mode) *cpu_time_used = get_time_us_no_sync(); for(int b = 0; b < bc; ++b) { cblas_gesv(n, nrhs, hA[b], lda, hIpiv[b], hB[b], ldb, hInfo[b]); } *cpu_time_used = get_time_us_no_sync() - *cpu_time_used; } gesv_initData( handle, n, nrhs, dA, lda, stA, dIpiv, stP, dB, ldb, stB, bc, hA, hIpiv, hB); // cold calls for(int iter = 0; iter < 2; iter++) { gesv_initData( handle, n, nrhs, dA, lda, stA, dIpiv, stP, dB, ldb, stB, bc, hA, hIpiv, hB); CHECK_ROCBLAS_ERROR(hipsolver_gesv(API, INPLACE, handle, n, nrhs, dA.data(), lda, stA, dIpiv.data(), stP, dB.data(), ldb, stB, dX.data(), ldx, stX, dWork.data(), lwork, hNiters.data(), 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++) { gesv_initData( handle, n, nrhs, dA, lda, stA, dIpiv, stP, dB, ldb, stB, bc, hA, hIpiv, hB); start = get_time_us_sync(stream); hipsolver_gesv(API, INPLACE, handle, n, nrhs, dA.data(), lda, stA, dIpiv.data(), stP, dB.data(), ldb, stB, dX.data(), ldx, stX, dWork.data(), lwork, hNiters.data(), dInfo.data(), bc); *gpu_time_used += get_time_us_sync(stream) - start; } *gpu_time_used /= hot_calls; } template void testing_gesv(Arguments& argus) { // get arguments hipsolver_local_handle handle; int n = argus.get("n"); int nrhs = argus.get("nrhs", n); int lda = argus.get("lda", n); int ldb = argus.get("ldb", n); int ldx = argus.get("ldx", n); int stA = argus.get("strideA", lda * n); int stP = argus.get("strideP", n); int stB = argus.get("strideB", ldb * nrhs); int stX = argus.get("strideX", ldx * nrhs); int bc = argus.batch_count; int hot_calls = argus.iters; int stBRes = (argus.unit_check || argus.norm_check) ? stX : 0; // check non-supported values // N/A // determine sizes size_t size_A = size_t(lda) * n; size_t size_B = size_t(ldb) * nrhs; size_t size_X = size_t(ldx) * nrhs; size_t size_P = size_t(n); double max_error = 0, gpu_time_used = 0, cpu_time_used = 0; size_t size_BRes = (argus.unit_check || argus.norm_check) ? (!INPLACE ? size_X : size_B) : 0; // check invalid sizes bool invalid_size = (n < 0 || nrhs < 0 || lda < n || ldb < n || ldx < n || bc < 0); if(invalid_size) { if(BATCHED) { // EXPECT_ROCBLAS_STATUS(hipsolver_gesv(API, // INPLACE, // handle, // n, // nrhs, // (T* const*)nullptr, // lda, // stA, // (int*)nullptr, // stP, // (T* const*)nullptr, // ldb, // stB, // (T* const*)nullptr, // ldx, // stX, // (T*)nullptr, // 0, // (int*)nullptr, // (int*)nullptr, // bc), // HIPSOLVER_STATUS_INVALID_VALUE); } else { EXPECT_ROCBLAS_STATUS(hipsolver_gesv(API, INPLACE, handle, n, nrhs, (T*)nullptr, lda, stA, (int*)nullptr, stP, (T*)nullptr, ldb, stB, (T*)nullptr, ldx, stX, (T*)nullptr, 0, (int*)nullptr, (int*)nullptr, bc), HIPSOLVER_STATUS_INVALID_VALUE); } if(argus.timing) ROCSOLVER_BENCH_INFORM(1); return; } if(BATCHED) { // // memory allocations // host_batch_vector hA(size_A, 1, bc); // host_batch_vector hB(size_B, 1, bc); // host_batch_vector hBRes(size_BRes, 1, bc); // host_strided_batch_vector hIpiv(size_P, 1, stP, bc); // host_strided_batch_vector hNiters(1, 1, 1, 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_batch_vector dB(size_B, 1, bc); // device_batch_vector dX(size_X, 1, bc); // device_strided_batch_vector dIpiv(size_P, 1, stP, bc); // device_strided_batch_vector dInfo(1, 1, 1, bc); // if(size_A) // CHECK_HIP_ERROR(dA.memcheck()); // if(size_B) // CHECK_HIP_ERROR(dB.memcheck()); // if(size_X) // CHECK_HIP_ERROR(dX.memcheck()); // if(size_P) // CHECK_HIP_ERROR(dIpiv.memcheck()); // CHECK_HIP_ERROR(dInfo.memcheck()); // size_t size_W; // hipsolver_gesv_bufferSize(API, // handle, // n, // nrhs, // dA.data(), // lda, // dIpiv.data(), // dB.data(), // ldb, // dX.data(), // ldx, // &size_W); // device_strided_batch_vector dWork(size_W, 1, size_W, bc); // if(size_W) // CHECK_HIP_ERROR(dWork.memcheck()); // // check computations // if(argus.unit_check || argus.norm_check) // gesv_getError(handle, // n, // nrhs, // dA, // lda, // stA, // dIpiv, // stP, // dB, // ldb, // stB, // dX, // ldx, // stX, // dWork, // size_W, // dInfo, // bc, // hA, // hIpiv, // hB, // hBRes, // hNiters, // hInfo, // hInfoRes, // &max_error); // // collect performance data // if(argus.timing) // gesv_getPerfData(handle, // n, // nrhs, // dA, // lda, // stA, // dIpiv, // stP, // dB, // ldb, // stB, // dX, // ldx, // stX, // dWork, // size_W, // dInfo, // bc, // hA, // hIpiv, // hB, // hNiters, // 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 hB(size_B, 1, stB, bc); host_strided_batch_vector hBRes(size_BRes, 1, stBRes, bc); host_strided_batch_vector hIpiv(size_P, 1, stP, bc); host_strided_batch_vector hNiters(1, 1, 1, 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 dB(size_B, 1, stB, bc); device_strided_batch_vector dX(size_X, 1, stX, bc); device_strided_batch_vector dIpiv(size_P, 1, stP, bc); device_strided_batch_vector dInfo(1, 1, 1, bc); if(size_A) CHECK_HIP_ERROR(dA.memcheck()); if(size_B) CHECK_HIP_ERROR(dB.memcheck()); if(size_X) CHECK_HIP_ERROR(dX.memcheck()); if(size_P) CHECK_HIP_ERROR(dIpiv.memcheck()); CHECK_HIP_ERROR(dInfo.memcheck()); size_t size_W; hipsolver_gesv_bufferSize(API, handle, n, nrhs, dA.data(), lda, dIpiv.data(), dB.data(), ldb, dX.data(), ldx, &size_W); device_strided_batch_vector dWork(size_W, 1, size_W, bc); if(size_W) CHECK_HIP_ERROR(dWork.memcheck()); // check computations if(argus.unit_check || argus.norm_check) gesv_getError(handle, n, nrhs, dA, lda, stA, dIpiv, stP, dB, ldb, stB, dX, ldx, stX, dWork, size_W, dInfo, bc, hA, hIpiv, hB, hBRes, hNiters, hInfo, hInfoRes, &max_error); // collect performance data if(argus.timing) gesv_getPerfData(handle, n, nrhs, dA, lda, stA, dIpiv, stP, dB, ldb, stB, dX, ldx, stX, dWork, size_W, dInfo, bc, hA, hIpiv, hB, hNiters, hInfo, &gpu_time_used, &cpu_time_used, hot_calls, argus.perf); } // validate results for rocsolver-test // using n * machine_precision as tolerance if(argus.unit_check) ROCSOLVER_TEST_CHECK(T, max_error, 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("n", "nrhs", "lda", "ldb", "ldx", "strideP", "batch_c"); rocsolver_bench_output(n, nrhs, lda, ldb, ldx, stP, bc); } else if(STRIDED) { rocsolver_bench_output("n", "nrhs", "lda", "ldb", "ldx", "strideA", "strideP", "strideB", "strideX", "batch_c"); rocsolver_bench_output(n, nrhs, lda, ldb, ldx, stA, stP, stB, stX, bc); } else { rocsolver_bench_output("n", "nrhs", "lda", "ldb", "ldx"); rocsolver_bench_output(n, nrhs, lda, ldb, ldx); } 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(); }