swapLast2Axes【免费下载链接】sip本项目是CANN提供的一款高效、可靠的高性能信号处理算子加速库基于华为Ascend AI处理器专门为信号处理领域而设计。项目地址: https://gitcode.com/cann/sip产品支持情况产品是否支持Atlas 200I/500 A2 推理产品×Atlas 推理系列产品×Atlas 训练系列产品×Atlas A3 训练系列产品/Atlas A3 推理系列产品√Atlas A2 训练系列产品/Atlas A2 推理系列产品√Ascend 950PR/Ascend 950DT×功能说明接口功能swapLast2AxesGetWorkspaceSize计算swapLast2Axes算子所需的workspace大小。swapLast2Axes交换Tensor的最后两维。计算公式$$ outTensor_{bij} inTensor_{bji}\ $$ 其中b为数据的批次号i为输入数据的行号 j为输入数据的列号。 示例示例一 输入“inTensor”为[[[1.0.j, 2.0.j, 3.0.j]]]调用swapLast2Axes算子后输出“outTensor”为[[[1.0.j], [2.0.j], [3.0.j]]]示例二 输入“inTensor”为[[[ 0.0.j, 1.0.j, 2.0.j],[ 3.0.j, 4.0.j, 5.0.j]],[[ 6.0.j, 7.0.j, 8.0.j],[ 9.0.j, 10.0.j, 11.0.j]]]调用swapLast2Axes算子后输出“outTensor”为[[[ 0.0.j, 3.0.j],[ 1.0.j, 4.0.j],[ 2.0.j, 5.0.j]],[[ 6.0.j, 9.0.j],[ 7.0.j, 10.0.j],[ 8.0.j, 11.0.j]]]函数原型若需使用“swapLast2Axes”算子必须先调用“swapLast2AxesGetWorkspaceSize”接口获取入参并根据计算流程计算所需workspace大小再调用“swapLast2Axes”接口执行计算。AsdSip::AspbStatus swapLast2AxesGetWorkspaceSize( size_t *size)AsdSip::AspbStatus swapLast2Axes( const aclTensor * inTensor, aclTensor * outTensor, void * stream, void * workspace nullptr)返回值返回状态码具体参见SiP返回码。swapLast2AxesGetWorkspaceSize参数说明参数名输入/输出描述sizesize_t *输入/输出swapLast2Axes算子所需要的workspace。返回值返回状态码具体参见SiP返回码。swapLast2Axes参数说明参数名输入/输出描述inTensoraclTensor *输入表示输入的张量数据对应公式中的inTensor。输入的最大元素数为3600000000 ([60000, 60000]以内)。数据类型仅支持COMPLEX64数据格式支持ND。输入dim限制为2或3。outTensoraclTensor *输出表示输出的张量数据对应公式中的outTensor。数据类型仅支持COMPLEX64数据类型需要与inTensor的数据类型一致。如果inTensor的shape为[kxy]outTensor的shape为[kyx]。数据格式支持ND。workspacevoid *输入swapLast2Axes算子所需要的workspace。stream(void *)输入npu执行流。返回值返回状态码具体参见SiP返回码。约束说明算子实际计算时不支持ND高维度运算不支持维度3的运算。调用示例示例代码如下该样例旨在提供快速上手、开发和调试算子的最小化实现其核心目标是使用最精简的代码展示算子的核心功能而非提供生产级的安全保障。不推荐用户直接将示例代码作为业务代码若用户将示例代码应用在自身的真实业务场景中且发生了安全问题则需用户自行承担。#include iostream #include vector #include asdsip.h #include acl/acl.h #include acl_meta.h using namespace AsdSip; #define ASD_STATUS_CHECK(err) \ do { \ AsdSip::AspbStatus err_ (err); \ if (err_ ! AsdSip::ErrorType::ACL_SUCCESS) { \ std::cout Execute failed. std::endl; \ exit(-1); \ } \ } while (0) #define CHECK_RET(cond, return_expr) \ do { \ if (!(cond)) { \ return_expr; \ } \ } while (0) #define LOG_PRINT(message, ...) \ do { \ printf(message, ##__VA_ARGS__); \ } while (0) int64_t GetShapeSize(const std::vectorint64_t shape) { int64_t shapeSize 1; for (auto i : shape) { shapeSize * i; } return shapeSize; } int Init(int32_t deviceId, aclrtStream *stream) { // 固定写法acl初始化 auto ret aclInit(nullptr); CHECK_RET(ret ::ACL_SUCCESS, LOG_PRINT(aclInit failed. ERROR: %d\n, ret); return ret); ret aclrtSetDevice(deviceId); CHECK_RET(ret ::ACL_SUCCESS, LOG_PRINT(aclrtSetDevice failed. ERROR: %d\n, ret); return ret); ret aclrtCreateStream(stream); CHECK_RET(ret ::ACL_SUCCESS, LOG_PRINT(aclrtCreateStream failed. ERROR: %d\n, ret); return ret); return 0; } template typename T int CreateAclTensor(const std::vectorT hostData, const std::vectorint64_t shape, void **deviceAddr, aclDataType dataType, aclTensor **tensor) { auto size GetShapeSize(shape) * sizeof(T) * 2; // 调用aclrtMalloc申请device侧内存 auto ret aclrtMalloc(deviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST); CHECK_RET(ret ::ACL_SUCCESS, LOG_PRINT(aclrtMalloc failed. ERROR: %d\n, ret); return ret); // 调用aclrtMemcpy将host侧数据复制到device侧内存上 ret aclrtMemcpy(*deviceAddr, size, hostData.data(), size, ACL_MEMCPY_HOST_TO_DEVICE); CHECK_RET(ret ::ACL_SUCCESS, LOG_PRINT(aclrtMemcpy failed. ERROR: %d\n, ret); return ret); // 计算连续tensor的strides std::vectorint64_t strides(shape.size(), 1); for (int64_t i shape.size() - 2; i 0; i--) { strides[i] shape[i 1] * strides[i 1]; } // 调用aclCreateTensor接口创建aclTensor *tensor aclCreateTensor(shape.data(), shape.size(), dataType, strides.data(), 0, aclFormat::ACL_FORMAT_ND, shape.data(), shape.size(), *deviceAddr); return 0; } void printTensor(const float *tensorData, size_t row, size_t col) { for (size_t r 0; r row; r) { for (size_t c 0; c col; c) { size_t index (r * col c) * 2; std::cout ( int(tensorData[index]) , int(tensorData[index 1]) ) ; } std::cout \n; } } int main(int argc, char **argv) { int deviceId 0; aclrtStream stream; auto ret Init(deviceId, stream); CHECK_RET(ret ::ACL_SUCCESS, LOG_PRINT(Init acl failed. ERROR: %d\n, ret); return ret); int64_t row 3; int64_t col 2; const int64_t tensorSize row * col * 2; std::vectorfloat tensorInData; tensorInData.reserve(tensorSize); for (int64_t i 0; i tensorSize; i) { tensorInData[i] 0.0 i; } std::vectorfloat tensorOutData; tensorOutData.reserve(tensorSize); std::vectorint64_t inShape {row, col}; std::vectorint64_t outShape {col, row}; aclTensor *input nullptr; aclTensor *output nullptr; void *inputDeviceAddr nullptr; void *outputDeviceAddr nullptr; ret CreateAclTensor(tensorInData, inShape, inputDeviceAddr, aclDataType::ACL_COMPLEX64, input); CHECK_RET(ret ::ACL_SUCCESS, return ret); ret CreateAclTensor(tensorOutData, outShape, outputDeviceAddr, aclDataType::ACL_COMPLEX64, output); CHECK_RET(ret ::ACL_SUCCESS, return ret); void *workspace nullptr; size_t lwork 0; swapLast2AxesGetWorkspaceSize(lwork); std::cout lwork lwork std::endl; if (lwork 0) { ret aclrtMalloc(workspace, static_castint64_t(lwork), ACL_MEM_MALLOC_HUGE_FIRST); CHECK_RET(ret ::ACL_SUCCESS, LOG_PRINT(allocate workspace failed. ERROR: %d\n, ret); return ret); } ASD_STATUS_CHECK(swapLast2Axes(input, output, stream, workspace)); ret aclrtSynchronizeStream(stream); CHECK_RET(ret ::ACL_SUCCESS, LOG_PRINT(aclrtSynchronizeStream failed. ERROR: %d\n, ret); return ret); ret aclrtMemcpy(tensorOutData.data(), tensorSize * sizeof(float), outputDeviceAddr, tensorSize * sizeof(float), ACL_MEMCPY_DEVICE_TO_HOST); CHECK_RET(ret ::ACL_SUCCESS, LOG_PRINT(copy output tensor from device to host failed. ERROR: %d\n, ret); return ret); std::cout row row , col col std::endl; std::cout ------- Input ------- std::endl; printTensor(tensorInData.data(), row, col); std::cout ------- Output ------- std::endl; printTensor(tensorOutData.data(), col, row); std::cout Execute successfully. std::endl; aclrtFree(inputDeviceAddr); aclrtFree(outputDeviceAddr); aclDestroyTensor(input); aclDestroyTensor(output); if (lwork 0) { aclrtFree(workspace); } aclrtDestroyStream(stream); aclrtResetDevice(deviceId); aclFinalize(); return 0; }【免费下载链接】sip本项目是CANN提供的一款高效、可靠的高性能信号处理算子加速库基于华为Ascend AI处理器专门为信号处理领域而设计。项目地址: https://gitcode.com/cann/sip创作声明：本文部分内容由AI辅助生成（AIGC），仅供参考