400 řádky
10 KiB
C++
400 řádky
10 KiB
C++
/*
|
|
Copyright (c) 2023 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 WARRANNTY OF ANY KIND, EXPRESS OR
|
|
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
FITNNESS 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 INN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
|
OUT OF OR INN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
|
|
THE SOFTWARE.
|
|
*/
|
|
#include <hip_test_common.hh>
|
|
|
|
#define N 512
|
|
|
|
__device__ bool check_erfcinvf() {
|
|
uint32_t len = 4;
|
|
float Val[] = {0.1, 1.2, 1, 0.9};
|
|
float Out[] = {1.16309, -0.179144, 0, 0.0889};
|
|
for (int i = 0; i < len; i++) {
|
|
if ((Out[i] - erfcinvf(Val[i])) > 0.0001) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_erfcxf() {
|
|
uint32_t len = 4;
|
|
float Val[] = {-0.5, 15, 3.2, 1};
|
|
float Out[] = {1.9524, 0.0375, 0.1687, 0.4276};
|
|
for (int i = 0; i < len; i++) {
|
|
if (Out[i] - erfcxf(Val[i]) > 0.0001) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_erfinvf() {
|
|
uint32_t len = 4;
|
|
float Val[] = {0, -0.5, 0.9, -0.2};
|
|
float Out[] = {0, -0.4769, 1.1631, -0.1791};
|
|
for (int i = 0; i < len; i++) {
|
|
if (Out[i] - erfinvf(Val[i]) > 0.0001) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_fdividef() {
|
|
uint32_t len = 4;
|
|
float Val[] = {0, -0.5, 0.9, -0.2};
|
|
float Out[] = {1, -0.4769, 1.1631, -0.1791};
|
|
for (int i = 0; i < len; i++) {
|
|
if (Val[i] / Out[i] - fdividef(Val[i], Out[i]) > 0.0001) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_erfcinv() {
|
|
uint32_t len = 4;
|
|
double Val[] = {0.1, 1.2, 1, 0.9};
|
|
double Out[] = {1.16309, -0.179144, 0, 0.0889};
|
|
for (int i = 0; i < len; i++) {
|
|
if (Out[i] - erfcinv(Val[i]) > 0.0001) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_erfcx() {
|
|
uint32_t len = 4;
|
|
double Val[] = {-0.5, 15, 3.2, 1};
|
|
double Out[] = {1.9524, 0.0375, 0.1687, 0.4276};
|
|
for (int i = 0; i < len; i++) {
|
|
if (Out[i] - erfcx(Val[i]) > 0.0001) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_erfinv() {
|
|
uint32_t len = 4;
|
|
double Val[] = {0, -0.5, 0.9, -0.2};
|
|
double Out[] = {0, -0.4769, 1.1631, -0.1791};
|
|
for (int i = 0; i < len; i++) {
|
|
if (Out[i] - erfinv(Val[i]) > 0.0001) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_fdivide() {
|
|
uint32_t len = 4;
|
|
double Val[] = {0, -0.5, 0.9, -0.2};
|
|
double Out[] = {1, -0.4769, 1.1631, -0.1791};
|
|
for (int i = 0; i < len; i++) {
|
|
if (Val[i] / Out[i] - fdividef(Val[i], Out[i]) > 0.0001) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_modff() {
|
|
uint32_t len = 4;
|
|
float Val[] = {0, -0.5, 0.9, -0.2};
|
|
float iPtr[] = {0, 0, 0, 0};
|
|
float frac[] = {0, -0.5, 0.9, -0.2};
|
|
float Out[] = {1, 1, 1, 1};
|
|
for (int i = 0; i < len; i++) {
|
|
if (frac[i] - modff(Val[i], Out + i) > 0.0001 && iPtr[i] == Out[i]) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_modf() {
|
|
uint32_t len = 4;
|
|
double Val[] = {0, -0.5, 0.9, -0.2};
|
|
double iPtr[] = {0, 0, 0, 0};
|
|
double frac[] = {0, -0.5, 0.9, -0.2};
|
|
double Out[] = {1, 1, 1, 1};
|
|
for (int i = 0; i < len; i++) {
|
|
if (frac[i] - modf(Val[i], Out + i) > 0.0001 && iPtr[i] == Out[i]) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_nextafterf() {
|
|
uint32_t len = 4;
|
|
float Val[] = {0, -0.5, 0.9, -0.2};
|
|
for (int i = 0; i < len; i++) {
|
|
if (nextafterf(Val[i], 1) - Val[i] > 0.0001) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_nextafter() {
|
|
uint32_t len = 4;
|
|
double Val[] = {0, -0.5, 0.9, -0.2};
|
|
for (int i = 0; i < len; i++) {
|
|
if (nextafter(Val[i], 1) - Val[i] > 0.0001) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_norm3df(float* A) {
|
|
float f = norm3df(A[0], A[1], A[2]);
|
|
float out = sqrt(A[0] * A[0] + A[1] * A[1] + A[2] * A[2]);
|
|
if (f - out > 0.0001) {
|
|
return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_norm3d(double* A) {
|
|
double f = norm3d(A[0], A[1], A[2]);
|
|
double out = sqrt(A[0] * A[0] + A[1] * A[1] + A[2] * A[2]);
|
|
if (f - out > 0.0001) {
|
|
return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_norm4df(float* A) {
|
|
float f = norm4df(A[0], A[1], A[2], A[3]);
|
|
float out = sqrt(A[0] * A[0] + A[1] * A[1] + A[2] * A[2] + A[3] * A[3]);
|
|
if (f - out > 0.0001) {
|
|
return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_norm4d(double* A) {
|
|
double f = norm4d(A[0], A[1], A[2], A[3]);
|
|
double out = sqrt(A[0] * A[0] + A[1] * A[1] + A[2] * A[2] + A[3] * A[3]);
|
|
if (f - out > 0.0001) {
|
|
return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_normcdff() {
|
|
uint32_t len = 2;
|
|
float Val[] = {0, 1};
|
|
float Out[] = {0.5, 0.8413};
|
|
for (int i = 0; i < len; i++) {
|
|
if (Out[i] - normcdff(Val[i]) > 0.0001) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_normcdf() {
|
|
uint32_t len = 2;
|
|
float Val[] = {0, 1};
|
|
float Out[] = {0.5, 0.8413};
|
|
for (int i = 0; i < len; i++) {
|
|
if (Out[i] - normcdf(Val[i]) > 0.0001) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_normcdfinvf() {
|
|
uint32_t len = 2;
|
|
double Val[] = {0.5, 0.8413};
|
|
for (int i = 0; i < len; i++) {
|
|
if (Val[i] - normcdfinvf(normcdff(Val[i])) > 0.0001) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_normcdfinv() {
|
|
uint32_t len = 2;
|
|
double Val[] = {0.5, 0.8413};
|
|
for (int i = 0; i < len; i++) {
|
|
if (Val[i] - normcdfinv(normcdf(Val[i])) > 0.0001) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_rcbrtf() {
|
|
float f = 1.0f;
|
|
if (rcbrtf(f) != 1.0f) {
|
|
return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_rcbrt() {
|
|
double f = 1.0;
|
|
if (rcbrt(f) != 1.0) {
|
|
return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_rhypotf() {
|
|
float f = 1.0f;
|
|
float g = 2.0f;
|
|
float val = rhypotf(f, g);
|
|
float sq = f * f + g * g;
|
|
if (1 / (val * val) - sq > 0.0001) {
|
|
return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_rhypot() {
|
|
double f = 1.0f;
|
|
double g = 2.0f;
|
|
double val = rhypot(f, g);
|
|
double sq = f * f + g * g;
|
|
if (1 / (val * val) - sq > 0.0001) {
|
|
return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_rnorm3df(float* A) {
|
|
float f = rnorm3df(A[0], A[1], A[2]);
|
|
float out = sqrt(A[0] * A[0] + A[1] * A[1] + A[2] * A[2]);
|
|
if (f - 1 / out > 0.0001) {
|
|
return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_rnorm3d(double* A) {
|
|
double f = rnorm3d(A[0], A[1], A[2]);
|
|
double out = sqrt(A[0] * A[0] + A[1] * A[1] + A[2] * A[2]);
|
|
if (f - 1 / out > 0.0001) {
|
|
return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_rnorm4df(float* A) {
|
|
float f = rnorm4df(A[0], A[1], A[2], A[3]);
|
|
float out = sqrt(A[0] * A[0] + A[1] * A[1] + A[2] * A[2] + A[3] * A[3]);
|
|
if (f - 1 / out > 0.0001) {
|
|
return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_rnorm4d(double* A) {
|
|
double f = rnorm4d(A[0], A[1], A[2], A[3]);
|
|
double out = sqrt(A[0] * A[0] + A[1] * A[1] + A[2] * A[2] + A[3] * A[3]);
|
|
if (f - 1 / out > 0.0001) {
|
|
return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
__device__ bool check_rnormf(float* A) {
|
|
return (rnorm3df(A[0], A[1], A[2]) - rnormf(3, A) < 0.0001) &&
|
|
(rnorm4df(A[0], A[1], A[2], A[3]) - rnormf(4, A) < 0.0001);
|
|
}
|
|
|
|
__device__ bool check_rnorm(double* A) {
|
|
return (rnorm3d(A[0], A[1], A[2]) - rnorm(3, A) < 0.0001) &&
|
|
(rnorm4d(A[0], A[1], A[2], A[3]) - rnorm(4, A) < 0.0001);
|
|
}
|
|
|
|
__device__ bool check_sincospif() {
|
|
float s1, c1, s2, c2;
|
|
float in1 = 1, in2 = 0.5;
|
|
sincospif(in1, &s1, &c1);
|
|
sincospif(in2, &s2, &c2);
|
|
if ((s1 - 0 < 0.00001) && (s2 - 1 < 0.00001) && (c1 + 1 < 0.00001) && (c2 - 0 < 0.00001)) {
|
|
return true;
|
|
}
|
|
return false;
|
|
}
|
|
|
|
__device__ bool check_sincospi() {
|
|
double s1, c1, s2, c2;
|
|
double in1 = 1, in2 = 0.5;
|
|
sincospi(in1, &s1, &c1);
|
|
sincospi(in2, &s2, &c2);
|
|
if ((s1 - 0 < 0.00001) && (s2 - 1 < 0.00001) && (c1 + 1 < 0.00001) && (c2 - 0 < 0.00001)) {
|
|
return true;
|
|
}
|
|
return false;
|
|
}
|
|
__global__ void testFunctions(bool* result, float* Af, double* A) {
|
|
result[0] &= check_erfcinvf() && check_erfcxf() && check_erfcinvf() && check_erfcinv() &&
|
|
check_erfcx() && check_erfcinv() && check_fdividef() && check_fdivide() &&
|
|
check_modff() && check_modf() && check_nextafterf() && check_norm3df(Af) &&
|
|
check_norm3d(A) && check_norm4df(Af) && check_norm4d(A) && check_normcdff() &&
|
|
check_normcdf() && check_normcdfinvf() && check_normcdfinv() && check_rcbrtf() &&
|
|
check_rcbrt() && check_rhypotf() && check_rhypot() && check_rnorm3df(Af) &&
|
|
check_rnorm3d(A) && check_rnorm4df(Af) && check_rnorm4d(A) && check_rnormf(Af) &&
|
|
check_rnorm(A) && check_sincospif() && check_sincospi() && check_nextafter();
|
|
}
|
|
|
|
TEST_CASE("Unit_TestDevice_DoublePrecisionMathFunc") {
|
|
float* Af = new float[N];
|
|
double* A = new double[N];
|
|
for (int i = 0; i < N; i++) {
|
|
Af[i] = i * 1.0f;
|
|
A[i] = i * 1.0;
|
|
}
|
|
float* Afd;
|
|
double* Ad;
|
|
bool *srcPtr, *devicePtr;
|
|
srcPtr = new bool;
|
|
srcPtr[0] = true;
|
|
// Device pointers
|
|
HIP_CHECK(hipMalloc(reinterpret_cast<void**>(&devicePtr), sizeof(bool)));
|
|
HIP_CHECK(hipMalloc(reinterpret_cast<void**>(&Afd), sizeof(float) * N));
|
|
HIP_CHECK(hipMalloc(reinterpret_cast<void**>(&Ad), sizeof(double) * N));
|
|
// MemCpy
|
|
HIP_CHECK(hipMemcpy(devicePtr, srcPtr, sizeof(bool), hipMemcpyHostToDevice));
|
|
HIP_CHECK(hipMemcpy(Afd, Af, sizeof(float) * N, hipMemcpyHostToDevice));
|
|
HIP_CHECK(hipMemcpy(Ad, A, sizeof(double) * N, hipMemcpyHostToDevice));
|
|
// Kernel Launch
|
|
hipLaunchKernelGGL(testFunctions, dim3(1), dim3(1), 0, 0, devicePtr, Afd, Ad);
|
|
HIP_CHECK(hipMemcpy(srcPtr, devicePtr, sizeof(bool), hipMemcpyDeviceToHost));
|
|
// Validation
|
|
REQUIRE(srcPtr[0] == true);
|
|
|
|
HIP_CHECK(hipFree(devicePtr));
|
|
HIP_CHECK(hipFree(Afd));
|
|
HIP_CHECK(hipFree(Ad));
|
|
delete srcPtr;
|
|
delete[] Af;
|
|
delete[] A;
|
|
}
|