* Copyright (c) 2025 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 <gtest/gtest.h>
#include "kernel_operator.h"
#include "stub_def.h"
#include "stub_fun.h"
using namespace std;
using namespace AscendC;
class TestScalarInstr : public testing::Test {
protected:
void SetUp() {}
void TearDown() {}
};
TEST_F(TestScalarInstr, ScalarInstrCase)
{
uint64_t u64XnVal = 28;
int64_t s64XnVal = 28;
EXPECT_EQ(ScalarGetCountOfValue<0>(u64XnVal), 61);
EXPECT_EQ(ScalarGetCountOfValue<1>(u64XnVal), 3);
EXPECT_EQ(ScalarCountLeadingZero(u64XnVal), 59);
EXPECT_EQ(CountBitsCntSameAsSignBit(s64XnVal), 58);
EXPECT_EQ(ScalarGetSFFValue<0>(u64XnVal), 0);
EXPECT_EQ(ScalarGetSFFValue<1>(u64XnVal), 2);
float convXnVal = 2.5;
half f16XdVal = convXnVal;
float maxOverFlowXnVal = 2147483648;
float minOverFlowXnVal = -__FLT_MAX__;
float underFlowXnVal = 0.0000000061;
uint16_t maxValue = 0x7c00;
half f16MaxVal = *reinterpret_cast<half *>(&maxValue);
uint16_t minValue = 0xfc00;
half f16MinVal = *reinterpret_cast<half *>(&minValue);
constexpr int64_t s32Max = 2147483647;
constexpr int64_t s32Min = -2147483648;
set_ctrl(sbitset0(get_ctrl(), 59));
half f16Res = ScalarCast<float, half, RoundMode::CAST_ODD>(convXnVal);
EXPECT_EQ(f16Res, f16XdVal);
int32_t s32Res = ScalarCast<float, int32_t, RoundMode::CAST_ROUND>(convXnVal);
EXPECT_EQ(s32Res, 3);
s32Res = ScalarCast<float, int32_t, RoundMode::CAST_CEIL>(convXnVal);
EXPECT_EQ(s32Res, 3);
s32Res = ScalarCast<float, int32_t, RoundMode::CAST_FLOOR>(convXnVal);
EXPECT_EQ(s32Res, 2);
s32Res = ScalarCast<float, int32_t, RoundMode::CAST_RINT>(convXnVal);
EXPECT_EQ(s32Res, 2);
f16Res = ScalarCast<float, half, RoundMode::CAST_ODD>(maxOverFlowXnVal);
EXPECT_EQ(f16Res, f16MaxVal);
s32Res = ScalarCast<float, int32_t, RoundMode::CAST_ROUND>(maxOverFlowXnVal);
EXPECT_EQ(s32Res, s32Max);
s32Res = ScalarCast<float, int32_t, RoundMode::CAST_CEIL>(maxOverFlowXnVal);
EXPECT_EQ(s32Res, s32Max);
s32Res = ScalarCast<float, int32_t, RoundMode::CAST_FLOOR>(maxOverFlowXnVal);
EXPECT_EQ(s32Res, s32Max);
s32Res = ScalarCast<float, int32_t, RoundMode::CAST_RINT>(maxOverFlowXnVal);
EXPECT_EQ(s32Res, s32Max);
f16Res = ScalarCast<float, half, RoundMode::CAST_ODD>(minOverFlowXnVal);
EXPECT_EQ(f16Res, f16MinVal);
s32Res = ScalarCast<float, int32_t, RoundMode::CAST_ROUND>(minOverFlowXnVal);
EXPECT_EQ(s32Res, s32Min);
s32Res = ScalarCast<float, int32_t, RoundMode::CAST_CEIL>(minOverFlowXnVal);
EXPECT_EQ(s32Res, s32Min);
s32Res = ScalarCast<float, int32_t, RoundMode::CAST_FLOOR>(minOverFlowXnVal);
EXPECT_EQ(s32Res, s32Min);
s32Res = ScalarCast<float, int32_t, RoundMode::CAST_RINT>(minOverFlowXnVal);
EXPECT_EQ(s32Res, s32Min);
set_ctrl(sbitset1(get_ctrl(), 59));
s32Res = ScalarCast<float, int32_t, RoundMode::CAST_ROUND>(minOverFlowXnVal);
EXPECT_EQ(s32Res, s32Min);
s32Res = ScalarCast<float, int32_t, RoundMode::CAST_CEIL>(minOverFlowXnVal);
EXPECT_EQ(s32Res, s32Min);
s32Res = ScalarCast<float, int32_t, RoundMode::CAST_FLOOR>(minOverFlowXnVal);
EXPECT_EQ(s32Res, s32Min);
s32Res = ScalarCast<float, int32_t, RoundMode::CAST_RINT>(minOverFlowXnVal);
EXPECT_EQ(s32Res, s32Min);
f16Res = ScalarCast<float, half, RoundMode::CAST_ODD>(underFlowXnVal);
EXPECT_EQ(f16Res, static_cast<half>(0));
half hScale = 1.0;
SetDeqScale(hScale);
float fScale = 1.0;
int16_t offset = 0;
bool signMode = true;
SetDeqScale(fScale, offset, signMode);
TPipe tpipe;
TBuf<TPosition::VECCALC> tbuf;
tpipe.InitBuffer(tbuf, 64 * sizeof(uint64_t));
LocalTensor<uint64_t> tmpBuffer = tbuf.Get<uint64_t>();
float vdeqScale[16] = {0};
int16_t vdeqOffset[16] = {0};
bool vdeqSignMode[16] = {0};
for (int i = 0; i < 16; i++) {
vdeqScale[i] = 1.0;
vdeqOffset[i] = 0;
vdeqSignMode[i] = true;
}
VdeqInfo vdeqInfo(vdeqScale, vdeqOffset, vdeqSignMode);
SetDeqScale<uint64_t>(tmpBuffer, vdeqInfo);
set_cond(u64XnVal);
EXPECT_EQ(get_cond(), 0);
set_condition_flag(u64XnVal);
EXPECT_EQ(get_condition_flag(), 0);
}
TEST_F(TestScalarInstr, ScalarInstrCastCase)
{
uint64_t u64XnVal = 28;
int64_t s64XnVal = 28;
EXPECT_EQ(ScalarGetCountOfValue<0>(u64XnVal), 61);
EXPECT_EQ(ScalarGetCountOfValue<1>(u64XnVal), 3);
EXPECT_EQ(ScalarCountLeadingZero(u64XnVal), 59);
EXPECT_EQ(CountBitsCntSameAsSignBit(s64XnVal), 58);
EXPECT_EQ(ScalarGetSFFValue<0>(u64XnVal), 0);
EXPECT_EQ(ScalarGetSFFValue<1>(u64XnVal), 2);
float convXnVal = 2.5;
half f16XdVal = convXnVal;
float maxOverFlowXnVal = 2147483648;
float minOverFlowXnVal = -__FLT_MAX__;
float underFlowXnVal = 0.0000000061;
uint16_t maxValue = 0x7c00;
half f16MaxVal = *reinterpret_cast<half *>(&maxValue);
uint16_t minValue = 0xfc00;
half f16MinVal = *reinterpret_cast<half *>(&minValue);
constexpr int64_t s32Max = 2147483647;
constexpr int64_t s32Min = -2147483648;
set_ctrl(sbitset0(get_ctrl(), 59));
half f16Res = Cast<float, half, RoundMode::CAST_ODD>(convXnVal);
EXPECT_EQ(f16Res, f16XdVal);
int32_t s32Res = Cast<float, int32_t, RoundMode::CAST_ROUND>(convXnVal);
EXPECT_EQ(s32Res, 3);
s32Res = Cast<float, int32_t, RoundMode::CAST_CEIL>(convXnVal);
EXPECT_EQ(s32Res, 3);
s32Res = Cast<float, int32_t, RoundMode::CAST_FLOOR>(convXnVal);
EXPECT_EQ(s32Res, 2);
s32Res = Cast<float, int32_t, RoundMode::CAST_RINT>(convXnVal);
EXPECT_EQ(s32Res, 2);
f16Res = Cast<float, half, RoundMode::CAST_ODD>(maxOverFlowXnVal);
EXPECT_EQ(f16Res, f16MaxVal);
s32Res = Cast<float, int32_t, RoundMode::CAST_ROUND>(maxOverFlowXnVal);
EXPECT_EQ(s32Res, s32Max);
s32Res = Cast<float, int32_t, RoundMode::CAST_CEIL>(maxOverFlowXnVal);
EXPECT_EQ(s32Res, s32Max);
s32Res = Cast<float, int32_t, RoundMode::CAST_FLOOR>(maxOverFlowXnVal);
EXPECT_EQ(s32Res, s32Max);
s32Res = Cast<float, int32_t, RoundMode::CAST_RINT>(maxOverFlowXnVal);
EXPECT_EQ(s32Res, s32Max);
f16Res = Cast<float, half, RoundMode::CAST_ODD>(minOverFlowXnVal);
EXPECT_EQ(f16Res, f16MinVal);
s32Res = Cast<float, int32_t, RoundMode::CAST_ROUND>(minOverFlowXnVal);
EXPECT_EQ(s32Res, s32Min);
s32Res = Cast<float, int32_t, RoundMode::CAST_CEIL>(minOverFlowXnVal);
EXPECT_EQ(s32Res, s32Min);
s32Res = Cast<float, int32_t, RoundMode::CAST_FLOOR>(minOverFlowXnVal);
EXPECT_EQ(s32Res, s32Min);
s32Res = Cast<float, int32_t, RoundMode::CAST_RINT>(minOverFlowXnVal);
EXPECT_EQ(s32Res, s32Min);
set_ctrl(sbitset1(get_ctrl(), 59));
s32Res = Cast<float, int32_t, RoundMode::CAST_ROUND>(minOverFlowXnVal);
EXPECT_EQ(s32Res, s32Min);
s32Res = Cast<float, int32_t, RoundMode::CAST_CEIL>(minOverFlowXnVal);
EXPECT_EQ(s32Res, s32Min);
s32Res = Cast<float, int32_t, RoundMode::CAST_FLOOR>(minOverFlowXnVal);
EXPECT_EQ(s32Res, s32Min);
s32Res = Cast<float, int32_t, RoundMode::CAST_RINT>(minOverFlowXnVal);
EXPECT_EQ(s32Res, s32Min);
f16Res = Cast<float, half, RoundMode::CAST_ODD>(underFlowXnVal);
EXPECT_EQ(f16Res, static_cast<half>(0));
half hScale = 1.0;
SetDeqScale(hScale);
float fScale = 1.0;
int16_t offset = 0;
bool signMode = true;
SetDeqScale(fScale, offset, signMode);
TPipe tpipe;
TBuf<TPosition::VECCALC> tbuf;
tpipe.InitBuffer(tbuf, 64 * sizeof(uint64_t));
LocalTensor<uint64_t> tmpBuffer = tbuf.Get<uint64_t>();
float vdeqScale[16] = {0};
int16_t vdeqOffset[16] = {0};
bool vdeqSignMode[16] = {0};
for (int i = 0; i < 16; i++) {
vdeqScale[i] = 1.0;
vdeqOffset[i] = 0;
vdeqSignMode[i] = true;
}
VdeqInfo vdeqInfo(vdeqScale, vdeqOffset, vdeqSignMode);
SetDeqScale<uint64_t>(tmpBuffer, vdeqInfo);
set_cond(u64XnVal);
EXPECT_EQ(get_cond(), 0);
set_condition_flag(u64XnVal);
EXPECT_EQ(get_condition_flag(), 0);
}
