* Copyright (c) 2026 Huawei Technologies Co., Ltd.
* This program is free software, you can redistribute it and/or modify it under the terms and conditions of
* CANN Open Software License Agreement Version 2.0 (the "License").
* Please refer to the License for details. You may not use this file except in compliance with the License.
* THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
* INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
* See LICENSE in the root of the software repository for the full text of the License.
*/
#include <cstdint>
#include <cstring>
#include "gtest/gtest.h"
#include "tikicpulib.h"
#include "../../../../random_common/op_kernel/arch35/random_unified_tiling_data_arch35.h"
extern "C" __global__ __aicore__ void stateless_random(
GM_ADDR shape, GM_ADDR seed, GM_ADDR offset, GM_ADDR from, GM_ADDR to, GM_ADDR y, GM_ADDR workspace,
GM_ADDR tiling);
namespace {
constexpr uint32_t kNumBlocks = 1;
constexpr uint64_t kTilingKey = 100;
constexpr int64_t kElementCount = 256;
inline size_t Align32(size_t size)
{
return (size + 31U) / 32U * 32U;
}
}
class StatelessRandomKernelTest : public testing::Test {};
TEST_F(StatelessRandomKernelTest, smoke_int32_no_range)
{
auto* shape = static_cast<uint8_t*>(AscendC::GmAlloc(Align32(2 * sizeof(int32_t))));
auto* seed = static_cast<uint8_t*>(AscendC::GmAlloc(Align32(sizeof(int64_t))));
auto* offset = static_cast<uint8_t*>(AscendC::GmAlloc(Align32(sizeof(int64_t))));
auto* from = static_cast<uint8_t*>(AscendC::GmAlloc(Align32(sizeof(int64_t))));
auto* to = static_cast<uint8_t*>(AscendC::GmAlloc(Align32(sizeof(int64_t))));
auto* y = static_cast<uint8_t*>(AscendC::GmAlloc(Align32(kElementCount * sizeof(int32_t))));
auto* workspace = static_cast<uint8_t*>(AscendC::GmAlloc(Align32(16 * 1024 * 1024)));
auto* tiling = static_cast<uint8_t*>(AscendC::GmAlloc(Align32(sizeof(RandomUnifiedSimtTilingDataStruct))));
std::memset(y, 0, kElementCount * sizeof(int32_t));
std::memset(tiling, 0, sizeof(RandomUnifiedSimtTilingDataStruct));
reinterpret_cast<int32_t*>(shape)[0] = 16;
reinterpret_cast<int32_t*>(shape)[1] = 16;
*reinterpret_cast<int64_t*>(seed) = 42;
*reinterpret_cast<int64_t*>(offset) = 0;
*reinterpret_cast<int64_t*>(from) = 0;
*reinterpret_cast<int64_t*>(to) = 100;
auto* tilingData = reinterpret_cast<RandomUnifiedSimtTilingDataStruct*>(tiling);
tilingData->usedCoreNum = kNumBlocks;
tilingData->outputSize = kElementCount;
tilingData->seed = 42;
tilingData->offset = 0;
tilingData->ubSize = 196608;
tilingData->extraInt64Param1 = 4;
tilingData->from = 0;
tilingData->range = 100;
tilingData->splitBlockCount = 1;
tilingData->splitBlocks[0].numel = kElementCount;
tilingData->splitBlocks[0].gmOffset = 0;
tilingData->splitBlocks[0].grid = 1;
tilingData->splitBlocks[0].totalThreads = 512;
tilingData->splitBlocks[0].kernelOffset = 0;
AscendC::SetKernelMode(KernelMode::AIV_MODE);
ICPU_SET_TILING_KEY(kTilingKey);
ICPU_RUN_KF(stateless_random, kNumBlocks, shape, seed, offset, from, to, y, workspace, tiling);
auto* yData = reinterpret_cast<int32_t*>(y);
for (int64_t i = 0; i < kElementCount; ++i) {
EXPECT_GE(yData[i], 0);
EXPECT_LT(yData[i], 100);
}
AscendC::GmFree(shape);
AscendC::GmFree(seed);
AscendC::GmFree(offset);
AscendC::GmFree(from);
AscendC::GmFree(to);
AscendC::GmFree(y);
AscendC::GmFree(workspace);
AscendC::GmFree(tiling);
}
TEST_F(StatelessRandomKernelTest, smoke_int32_with_range)
{
auto* shape = static_cast<uint8_t*>(AscendC::GmAlloc(Align32(2 * sizeof(int32_t))));
auto* seed = static_cast<uint8_t*>(AscendC::GmAlloc(Align32(sizeof(int64_t))));
auto* offset = static_cast<uint8_t*>(AscendC::GmAlloc(Align32(sizeof(int64_t))));
auto* from = static_cast<uint8_t*>(AscendC::GmAlloc(Align32(sizeof(int64_t))));
auto* to = static_cast<uint8_t*>(AscendC::GmAlloc(Align32(sizeof(int64_t))));
auto* y = static_cast<uint8_t*>(AscendC::GmAlloc(Align32(kElementCount * sizeof(int32_t))));
auto* workspace = static_cast<uint8_t*>(AscendC::GmAlloc(Align32(16 * 1024 * 1024)));
auto* tiling = static_cast<uint8_t*>(AscendC::GmAlloc(Align32(sizeof(RandomUnifiedSimtTilingDataStruct))));
std::memset(y, 0, kElementCount * sizeof(int32_t));
std::memset(tiling, 0, sizeof(RandomUnifiedSimtTilingDataStruct));
reinterpret_cast<int32_t*>(shape)[0] = 16;
reinterpret_cast<int32_t*>(shape)[1] = 16;
*reinterpret_cast<int64_t*>(seed) = 42;
*reinterpret_cast<int64_t*>(offset) = 0;
*reinterpret_cast<int64_t*>(from) = 5;
*reinterpret_cast<int64_t*>(to) = 15;
auto* tilingData = reinterpret_cast<RandomUnifiedSimtTilingDataStruct*>(tiling);
tilingData->usedCoreNum = kNumBlocks;
tilingData->outputSize = kElementCount;
tilingData->seed = 42;
tilingData->offset = 0;
tilingData->ubSize = 196608;
tilingData->extraInt64Param1 = 4;
tilingData->from = 5;
tilingData->range = 10;
tilingData->splitBlockCount = 1;
tilingData->splitBlocks[0].numel = kElementCount;
tilingData->splitBlocks[0].gmOffset = 0;
tilingData->splitBlocks[0].grid = 1;
tilingData->splitBlocks[0].totalThreads = 512;
tilingData->splitBlocks[0].kernelOffset = 0;
AscendC::SetKernelMode(KernelMode::AIV_MODE);
ICPU_SET_TILING_KEY(kTilingKey);
ICPU_RUN_KF(stateless_random, kNumBlocks, shape, seed, offset, from, to, y, workspace, tiling);
auto* yData = reinterpret_cast<int32_t*>(y);
for (int64_t i = 0; i < kElementCount; ++i) {
EXPECT_GE(yData[i], 5);
EXPECT_LT(yData[i], 15);
}
AscendC::GmFree(shape);
AscendC::GmFree(seed);
AscendC::GmFree(offset);
AscendC::GmFree(from);
AscendC::GmFree(to);
AscendC::GmFree(y);
AscendC::GmFree(workspace);
AscendC::GmFree(tiling);
}
