|
| 1 | +//==---------- big_const_initializer.cpp - DPC++ ESIMD on-device test -----==// |
| 2 | +// |
| 3 | +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 4 | +// See https://llvm.org/LICENSE.txt for license information. |
| 5 | +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| 6 | +// |
| 7 | +//===----------------------------------------------------------------------===// |
| 8 | +// TODO enable on Windows |
| 9 | +// REQUIRES: linux && gpu |
| 10 | +// UNSUPPORTED: cuda |
| 11 | +// RUN: %clangxx-esimd -fsycl -I%S/.. %s -o %t.out |
| 12 | +// RUN: %ESIMD_RUN_PLACEHOLDER %t.out |
| 13 | + |
| 14 | +// This test checks that ESIMD program with big constant initializer list can |
| 15 | +// compile and run correctly. |
| 16 | + |
| 17 | +#include "esimd_test_utils.hpp" |
| 18 | + |
| 19 | +#include <CL/sycl.hpp> |
| 20 | +#include <CL/sycl/INTEL/esimd.hpp> |
| 21 | + |
| 22 | +#include <iostream> |
| 23 | + |
| 24 | +#define SIMD_WIDTH 16 |
| 25 | + |
| 26 | +#define N_SAMPLES (SIMD_WIDTH * 100) |
| 27 | +#define N_PRINT 20 |
| 28 | + |
| 29 | +#define VAL1 0x9E3779B9 |
| 30 | +#define VAL2 0xBB67AE85 |
| 31 | + |
| 32 | +inline void foo(sycl::INTEL::gpu::simd<std::uint32_t, 16> &k) { |
| 33 | + sycl::INTEL::gpu::simd<std::uint32_t, 16> k_add = { |
| 34 | + VAL1, VAL2, VAL1, VAL2, VAL1, VAL2, VAL1, VAL2, |
| 35 | + VAL1, VAL2, VAL1, VAL2, VAL1, VAL2, VAL1, VAL2}; |
| 36 | + k += k_add; |
| 37 | +} |
| 38 | + |
| 39 | +int main(int argc, char **argv) { |
| 40 | + size_t nsamples = N_SAMPLES; |
| 41 | + sycl::queue queue(esimd_test::ESIMDSelector{}, |
| 42 | + esimd_test::createExceptionHandler()); |
| 43 | + |
| 44 | + std::cout << "Running on " |
| 45 | + << queue.get_device().get_info<sycl::info::device::name>() |
| 46 | + << std::endl; |
| 47 | + std::cout << "Driver version " |
| 48 | + << queue.get_device().get_info<sycl::info::device::driver_version>() |
| 49 | + << std::endl; |
| 50 | + |
| 51 | + sycl::buffer<std::uint32_t, 1> r(sycl::range<1>{nsamples}); |
| 52 | + |
| 53 | + try { |
| 54 | + queue.submit([&](sycl::handler &cgh) { |
| 55 | + auto r_acc = r.template get_access<sycl::access::mode::write>(cgh); |
| 56 | + cgh.parallel_for<class Test>( |
| 57 | + sycl::range<1>{nsamples / SIMD_WIDTH}, |
| 58 | + [=](sycl::item<1> item) SYCL_ESIMD_KERNEL { |
| 59 | + size_t id = item.get_id(0); |
| 60 | + sycl::INTEL::gpu::simd<std::uint32_t, 16> key{ |
| 61 | + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}; |
| 62 | + foo(key); |
| 63 | + sycl::INTEL::gpu::block_store( |
| 64 | + r_acc, id * SIMD_WIDTH * sizeof(std::uint32_t), key); |
| 65 | + }); |
| 66 | + }); |
| 67 | + } catch (const sycl::exception &e) { |
| 68 | + std::cout << "*** EXCEPTION caught: " << e.what() << "\n"; |
| 69 | + return 1; |
| 70 | + } |
| 71 | + auto acc = r.template get_access<sycl::access::mode::read>(); |
| 72 | + for (int i = 0; i < N_PRINT; i++) { |
| 73 | + std::cout << acc[i] << " "; |
| 74 | + } |
| 75 | + std::cout << std::endl; |
| 76 | + |
| 77 | + int err_cnt = 0; |
| 78 | + |
| 79 | + for (unsigned i = 0; i < nsamples; ++i) { |
| 80 | + uint32_t gold = i % 2 == 0 ? VAL1 + 1 : VAL2 + 1; |
| 81 | + uint32_t test = acc[i]; |
| 82 | + if (test != gold) { |
| 83 | + if (++err_cnt < 10) { |
| 84 | + std::cout << "failed at index " << i << ", " << test << " != " << gold |
| 85 | + << " (expected)" |
| 86 | + << "\n"; |
| 87 | + } |
| 88 | + } |
| 89 | + } |
| 90 | + if (err_cnt > 0) { |
| 91 | + std::cout << " pass rate: " |
| 92 | + << ((float)(nsamples - err_cnt) / (float)nsamples) * 100.0f |
| 93 | + << "% (" << (nsamples - err_cnt) << "/" << nsamples << ")\n"; |
| 94 | + } |
| 95 | + |
| 96 | + std::cout << (err_cnt > 0 ? "FAILED\n" : "Passed\n"); |
| 97 | + return err_cnt > 0 ? 1 : 0; |
| 98 | +} |
0 commit comments