derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
451 lines
17 KiB
C++
451 lines
17 KiB
C++
/* ************************************************************************
|
|
* Copyright (C) 2016-2025 Advanced Micro Devices, Inc. All rights reserved.
|
|
*
|
|
* Permission is hereby granted, free of charge, to any person obtaining a copy
|
|
* of this software and associated documentation files (the "Software"), to deal
|
|
* in the Software without restriction, including without limitation the rights
|
|
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
|
* copies of the Software, and to permit persons to whom the Software is
|
|
* furnished to do so, subject to the following conditions:
|
|
*
|
|
* The above copyright notice and this permission notice shall be included in
|
|
* all copies or substantial portions of the Software.
|
|
*
|
|
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
|
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
|
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
|
* SOFTWARE.
|
|
*
|
|
* ************************************************************************ */
|
|
|
|
#include "unit.h"
|
|
#include "hipblas.h"
|
|
#include "host_vector.hpp"
|
|
#include "utility.h"
|
|
|
|
/* ========================================Gtest Unit Check
|
|
* ==================================================== */
|
|
|
|
/*! \brief Template: gtest unit compare two matrices float/double/complex */
|
|
// This returns from the current function if an error occurs
|
|
|
|
#ifndef GOOGLE_TEST
|
|
|
|
#define UNIT_CHECK(M, N, batch_count, lda, strideA, hCPU, hGPU, UNIT_ASSERT_EQ)
|
|
#define UNIT_CHECK_B(M, N, batch_count, lda, hCPU, hGPU, UNIT_ASSERT_EQ)
|
|
|
|
#else // GOOGLE_TEST
|
|
|
|
#define UNIT_CHECK(M, N, batch_count, lda, strideA, hCPU, hGPU, UNIT_ASSERT_EQ) \
|
|
do \
|
|
{ \
|
|
for(size_t k = 0; k < batch_count; k++) \
|
|
for(size_t j = 0; j < N; j++) \
|
|
for(size_t i = 0; i < M; i++) \
|
|
if(hipblas_isnan(hCPU[i + j * lda + k * strideA])) \
|
|
{ \
|
|
ASSERT_TRUE(hipblas_isnan(hGPU[i + j * lda + k * strideA])); \
|
|
} \
|
|
else \
|
|
{ \
|
|
UNIT_ASSERT_EQ(hCPU[i + j * lda + k * strideA], \
|
|
hGPU[i + j * lda + k * strideA]); \
|
|
} \
|
|
} while(0)
|
|
|
|
#define UNIT_CHECK_B(M, N, batch_count, lda, hCPU, hGPU, UNIT_ASSERT_EQ) \
|
|
do \
|
|
{ \
|
|
for(size_t k = 0; k < batch_count; k++) \
|
|
for(size_t j = 0; j < N; j++) \
|
|
for(size_t i = 0; i < M; i++) \
|
|
if(hipblas_isnan(hCPU[k][i + j * lda])) \
|
|
{ \
|
|
ASSERT_TRUE(hipblas_isnan(hGPU[k][i + j * lda])); \
|
|
} \
|
|
else \
|
|
{ \
|
|
UNIT_ASSERT_EQ(hCPU[k][i + j * lda], hGPU[k][i + j * lda]); \
|
|
} \
|
|
} while(0)
|
|
|
|
#endif // GOOGLE_TEST
|
|
|
|
#define ASSERT_HALF_EQ(a, b) ASSERT_FLOAT_EQ(half_to_float(a), half_to_float(b))
|
|
#define ASSERT_BF16_EQ(a, b) ASSERT_FLOAT_EQ(bfloat16_to_float(a), bfloat16_to_float(b))
|
|
|
|
// Compare float to rocblas_bfloat16
|
|
// Allow the rocblas_bfloat16 to match the rounded or truncated value of float
|
|
// Only call ASSERT_FLOAT_EQ with the rounded value if the truncated value does not match
|
|
#include <gtest/internal/gtest-internal.h>
|
|
#define ASSERT_FLOAT_BF16_EQ(a, b) \
|
|
do \
|
|
{ \
|
|
using testing::internal::FloatingPoint; \
|
|
if(!FloatingPoint<float>(b).AlmostEquals(FloatingPoint<float>(hipblasBfloat16(a)))) \
|
|
ASSERT_FLOAT_EQ(b, hipblasBfloat16(a)); \
|
|
} while(0)
|
|
|
|
#define ASSERT_FLOAT_COMPLEX_EQ(a, b) \
|
|
do \
|
|
{ \
|
|
ASSERT_FLOAT_EQ(a.real(), b.real()); \
|
|
ASSERT_FLOAT_EQ(a.imag(), b.imag()); \
|
|
} while(0)
|
|
|
|
#define ASSERT_DOUBLE_COMPLEX_EQ(a, b) \
|
|
do \
|
|
{ \
|
|
ASSERT_DOUBLE_EQ(a.real(), b.real()); \
|
|
ASSERT_DOUBLE_EQ(a.imag(), b.imag()); \
|
|
} while(0)
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M, int64_t N, int64_t lda, hipblasHalf* hCPU, hipblasHalf* hGPU)
|
|
{
|
|
UNIT_CHECK(M, N, 1, lda, 0, hCPU, hGPU, ASSERT_HALF_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(
|
|
int64_t M, int64_t N, int64_t lda, hipblasBfloat16* hCPU, hipblasBfloat16* hGPU)
|
|
{
|
|
UNIT_CHECK(M, N, 1, lda, 0, hCPU, hGPU, ASSERT_BF16_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M, int64_t N, int64_t lda, float* hCPU, float* hGPU)
|
|
{
|
|
UNIT_CHECK(M, N, 1, lda, 0, hCPU, hGPU, ASSERT_FLOAT_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M, int64_t N, int64_t lda, double* hCPU, double* hGPU)
|
|
{
|
|
UNIT_CHECK(M, N, 1, lda, 0, hCPU, hGPU, ASSERT_DOUBLE_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(
|
|
int64_t M, int64_t N, int64_t lda, std::complex<float>* hCPU, std::complex<float>* hGPU)
|
|
{
|
|
#ifdef GOOGLE_TEST
|
|
for(int64_t j = 0; j < N; j++)
|
|
for(int64_t i = 0; i < M; i++)
|
|
{
|
|
ASSERT_FLOAT_EQ(hCPU[i + j * lda].real(), hGPU[i + j * lda].real());
|
|
ASSERT_FLOAT_EQ(hCPU[i + j * lda].imag(), hGPU[i + j * lda].imag());
|
|
}
|
|
#endif
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(
|
|
int64_t M, int64_t N, int64_t lda, std::complex<double>* hCPU, std::complex<double>* hGPU)
|
|
{
|
|
#ifdef GOOGLE_TEST
|
|
for(int64_t j = 0; j < N; j++)
|
|
for(int64_t i = 0; i < M; i++)
|
|
{
|
|
ASSERT_DOUBLE_EQ(hCPU[i + j * lda].real(), hGPU[i + j * lda].real());
|
|
ASSERT_DOUBLE_EQ(hCPU[i + j * lda].imag(), hGPU[i + j * lda].imag());
|
|
}
|
|
#endif
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M, int64_t N, int64_t lda, int* hCPU, int* hGPU)
|
|
{
|
|
UNIT_CHECK(M, N, 1, lda, 0, hCPU, hGPU, ASSERT_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M, int64_t N, int64_t lda, int64_t* hCPU, int64_t* hGPU)
|
|
{
|
|
UNIT_CHECK(M, N, 1, lda, 0, hCPU, hGPU, ASSERT_EQ);
|
|
}
|
|
|
|
// batched checks
|
|
template <>
|
|
void unit_check_general(
|
|
int64_t M, int64_t N, int64_t batch_count, int64_t lda, hipblasHalf** hCPU, hipblasHalf** hGPU)
|
|
{
|
|
UNIT_CHECK_B(M, N, batch_count, lda, hCPU, hGPU, ASSERT_HALF_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M,
|
|
int64_t N,
|
|
int64_t batch_count,
|
|
int64_t lda,
|
|
hipblasBfloat16** hCPU,
|
|
hipblasBfloat16** hGPU)
|
|
{
|
|
UNIT_CHECK_B(M, N, batch_count, lda, hCPU, hGPU, ASSERT_BF16_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(
|
|
int64_t M, int64_t N, int64_t batch_count, int64_t lda, float** hCPU, float** hGPU)
|
|
{
|
|
UNIT_CHECK_B(M, N, batch_count, lda, hCPU, hGPU, ASSERT_FLOAT_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(
|
|
int64_t M, int64_t N, int64_t batch_count, int64_t lda, double** hCPU, double** hGPU)
|
|
{
|
|
UNIT_CHECK_B(M, N, batch_count, lda, hCPU, hGPU, ASSERT_DOUBLE_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(
|
|
int64_t M, int64_t N, int64_t batch_count, int64_t lda, int** hCPU, int** hGPU)
|
|
{
|
|
UNIT_CHECK_B(M, N, batch_count, lda, hCPU, hGPU, ASSERT_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(
|
|
int64_t M, int64_t N, int64_t batch_count, int64_t lda, int64_t** hCPU, int64_t** hGPU)
|
|
{
|
|
UNIT_CHECK_B(M, N, batch_count, lda, hCPU, hGPU, ASSERT_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M,
|
|
int64_t N,
|
|
int64_t batch_count,
|
|
int64_t lda,
|
|
std::complex<float>** hCPU,
|
|
std::complex<float>** hGPU)
|
|
{
|
|
UNIT_CHECK_B(M, N, batch_count, lda, hCPU, hGPU, ASSERT_FLOAT_COMPLEX_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M,
|
|
int64_t N,
|
|
int64_t batch_count,
|
|
int64_t lda,
|
|
std::complex<double>** hCPU,
|
|
std::complex<double>** hGPU)
|
|
{
|
|
UNIT_CHECK_B(M, N, batch_count, lda, hCPU, hGPU, ASSERT_DOUBLE_COMPLEX_EQ);
|
|
}
|
|
|
|
// batched checks for host_vector[]s
|
|
template <>
|
|
void unit_check_general(int64_t M,
|
|
int64_t N,
|
|
int64_t batch_count,
|
|
int64_t lda,
|
|
host_vector<hipblasHalf> hCPU[],
|
|
host_vector<hipblasHalf> hGPU[])
|
|
{
|
|
UNIT_CHECK_B(M, N, batch_count, lda, hCPU, hGPU, ASSERT_HALF_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M,
|
|
int64_t N,
|
|
int64_t batch_count,
|
|
int64_t lda,
|
|
host_vector<hipblasBfloat16> hCPU[],
|
|
host_vector<hipblasBfloat16> hGPU[])
|
|
{
|
|
UNIT_CHECK_B(M, N, batch_count, lda, hCPU, hGPU, ASSERT_BF16_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M,
|
|
int64_t N,
|
|
int64_t batch_count,
|
|
int64_t lda,
|
|
host_vector<int> hCPU[],
|
|
host_vector<int> hGPU[])
|
|
{
|
|
UNIT_CHECK_B(M, N, batch_count, lda, hCPU, hGPU, ASSERT_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M,
|
|
int64_t N,
|
|
int64_t batch_count,
|
|
int64_t lda,
|
|
host_vector<float> hCPU[],
|
|
host_vector<float> hGPU[])
|
|
{
|
|
UNIT_CHECK_B(M, N, batch_count, lda, hCPU, hGPU, ASSERT_FLOAT_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M,
|
|
int64_t N,
|
|
int64_t batch_count,
|
|
int64_t lda,
|
|
host_vector<double> hCPU[],
|
|
host_vector<double> hGPU[])
|
|
{
|
|
UNIT_CHECK_B(M, N, batch_count, lda, hCPU, hGPU, ASSERT_DOUBLE_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M,
|
|
int64_t N,
|
|
int64_t batch_count,
|
|
int64_t lda,
|
|
host_vector<std::complex<float>> hCPU[],
|
|
host_vector<std::complex<float>> hGPU[])
|
|
{
|
|
UNIT_CHECK_B(M, N, batch_count, lda, hCPU, hGPU, ASSERT_FLOAT_COMPLEX_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M,
|
|
int64_t N,
|
|
int64_t batch_count,
|
|
int64_t lda,
|
|
host_vector<std::complex<double>> hCPU[],
|
|
host_vector<std::complex<double>> hGPU[])
|
|
{
|
|
UNIT_CHECK_B(M, N, batch_count, lda, hCPU, hGPU, ASSERT_DOUBLE_COMPLEX_EQ);
|
|
}
|
|
|
|
// strided_batched checks
|
|
template <>
|
|
void unit_check_general(int64_t M,
|
|
int64_t N,
|
|
int64_t batch_count,
|
|
int64_t lda,
|
|
hipblasStride strideA,
|
|
hipblasHalf* hCPU,
|
|
hipblasHalf* hGPU)
|
|
{
|
|
UNIT_CHECK(M, N, batch_count, lda, strideA, hCPU, hGPU, ASSERT_HALF_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M,
|
|
int64_t N,
|
|
int64_t batch_count,
|
|
int64_t lda,
|
|
hipblasStride strideA,
|
|
hipblasBfloat16* hCPU,
|
|
hipblasBfloat16* hGPU)
|
|
{
|
|
UNIT_CHECK(M, N, batch_count, lda, strideA, hCPU, hGPU, ASSERT_BF16_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M,
|
|
int64_t N,
|
|
int64_t batch_count,
|
|
int64_t lda,
|
|
hipblasStride strideA,
|
|
float* hCPU,
|
|
float* hGPU)
|
|
{
|
|
UNIT_CHECK(M, N, batch_count, lda, strideA, hCPU, hGPU, ASSERT_FLOAT_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M,
|
|
int64_t N,
|
|
int64_t batch_count,
|
|
int64_t lda,
|
|
hipblasStride strideA,
|
|
double* hCPU,
|
|
double* hGPU)
|
|
{
|
|
UNIT_CHECK(M, N, batch_count, lda, strideA, hCPU, hGPU, ASSERT_DOUBLE_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M,
|
|
int64_t N,
|
|
int64_t batch_count,
|
|
int64_t lda,
|
|
hipblasStride strideA,
|
|
std::complex<float>* hCPU,
|
|
std::complex<float>* hGPU)
|
|
{
|
|
UNIT_CHECK(M, N, batch_count, lda, strideA, hCPU, hGPU, ASSERT_FLOAT_COMPLEX_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M,
|
|
int64_t N,
|
|
int64_t batch_count,
|
|
int64_t lda,
|
|
hipblasStride strideA,
|
|
std::complex<double>* hCPU,
|
|
std::complex<double>* hGPU)
|
|
{
|
|
UNIT_CHECK(M, N, batch_count, lda, strideA, hCPU, hGPU, ASSERT_DOUBLE_COMPLEX_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M,
|
|
int64_t N,
|
|
int64_t batch_count,
|
|
int64_t lda,
|
|
hipblasStride strideA,
|
|
int* hCPU,
|
|
int* hGPU)
|
|
{
|
|
UNIT_CHECK(M, N, batch_count, lda, strideA, hCPU, hGPU, ASSERT_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_general(int64_t M,
|
|
int64_t N,
|
|
int64_t batch_count,
|
|
int64_t lda,
|
|
hipblasStride strideA,
|
|
int64_t* hCPU,
|
|
int64_t* hGPU)
|
|
{
|
|
UNIT_CHECK(M, N, batch_count, lda, strideA, hCPU, hGPU, ASSERT_EQ);
|
|
}
|
|
|
|
// mixed
|
|
|
|
template <typename T, typename T_hpa>
|
|
void unit_check_mixed(
|
|
int64_t M, int64_t N, int64_t lda, const std::remove_cv_t<T_hpa>* hCPU, const T* hGPU)
|
|
{
|
|
UNIT_CHECK(M, N, 1, lda, 0, hCPU, hGPU, ASSERT_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_mixed(
|
|
int64_t M, int64_t N, int64_t lda, const hipblasBfloat16* hCPU, const hipblasBfloat16* hGPU)
|
|
{
|
|
UNIT_CHECK(M, N, 1, lda, 0, hCPU, hGPU, ASSERT_BF16_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_mixed(
|
|
int64_t M, int64_t N, int64_t lda, const hipblasHalf* hCPU, const hipblasHalf* hGPU)
|
|
{
|
|
UNIT_CHECK(M, N, 1, lda, 0, hCPU, hGPU, ASSERT_HALF_EQ);
|
|
}
|
|
|
|
template <>
|
|
void unit_check_mixed<hipblasBfloat16, float>(
|
|
int64_t M, int64_t N, int64_t lda, const float* hCPU, const hipblasBfloat16* hGPU)
|
|
{
|
|
UNIT_CHECK(M, N, 1, lda, 0, hCPU, hGPU, ASSERT_FLOAT_BF16_EQ);
|
|
}
|
|
|
|
#define INST(Ta_, Tb_) \
|
|
template void unit_check_mixed<Ta_, Tb_>( \
|
|
int64_t M, int64_t N, int64_t lda, const Ta_* hCPU, const Tb_* hGPU);
|
|
|
|
INST(float, float)
|
|
INST(double, double)
|