AI/rocm-systems
2
0
Garpu 0
Kode Masalah Tarik Permintaan Actions Packages Projects Rilis Wiki Kegiatan

Merge branch 'master' into amd-master-next

Telusuri Sumber 
Conflicts:
	CMakeLists.txt
	tests/src/texture/simpleTexture2DLayered.cpp
	tests/src/texture/simpleTexture3D.cpp

Change-Id: I4aa4754d391b5f37ddf15fa0bcfc84d9da020119
This commit is contained in:
Lad, Aditya
2020-03-06 14:10:04 -05:00
orang tua 4b312caca4 5c036520b1
melakukan d80edf9541
42 mengubah file dengan 1410 tambahan dan 640 penghapusan
Download Patch File Download Diff File Expand all files Collapse all files
tests/src/kernel/hipShflTests.cpp
+130
Melihat File
@@ -0,0 +1,130 @@
/*
Copyright (c) 2015-present 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.
*/
/* HIT_START
* BUILD: %t %s ../test_common.cpp
* TEST: %t
* HIT_END
*/
#include <iostream>
#include <hip/hip_runtime.h>
#include "test_common.h"
#define WIDTH 4
#define NUM (WIDTH * WIDTH)
#define THREADS_PER_BLOCK_X 4
#define THREADS_PER_BLOCK_Y 4
#define THREADS_PER_BLOCK_Z 1
// Device (Kernel) function, it must be void
template <typename T>
__global__ void matrixTranspose(T* out, T* in, const int width) {
int x = hipBlockDim_x * hipBlockIdx_x + hipThreadIdx_x;
T val = in[x];
for (int i = 0; i < width; i++) {
for (int j = 0; j < width; j++) out[i * width + j] = __shfl(val, j * width + i);
}
}
// CPU implementation of matrix transpose
template <typename T>
void matrixTransposeCPUReference(T* output, T* input, const unsigned int width) {
for (unsigned int j = 0; j < width; j++) {
for (unsigned int i = 0; i < width; i++) {
output[i * width + j] = input[j * width + i];
}
}
}
template<typename T>
void runTest() {
T* Matrix;
T* TransposeMatrix;
T* cpuTransposeMatrix;
T* gpuMatrix;
T* gpuTransposeMatrix;
hipDeviceProp_t devProp;
hipGetDeviceProperties(&devProp, 0);
int i;
int errors;
Matrix = (T*)malloc(NUM * sizeof(T));
TransposeMatrix = (T*)malloc(NUM * sizeof(T));
cpuTransposeMatrix = (T*)malloc(NUM * sizeof(T));
// initialize the input data
for (i = 0; i < NUM; i++) {
Matrix[i] = (T)i * 10l;
}
// allocate the memory on the device side
hipMalloc((void**)&gpuMatrix, NUM * sizeof(T));
hipMalloc((void**)&gpuTransposeMatrix, NUM * sizeof(T));
// Memory transfer from host to device
hipMemcpy(gpuMatrix, Matrix, NUM * sizeof(T), hipMemcpyHostToDevice);
// Lauching kernel from host
hipLaunchKernelGGL(matrixTranspose<T>, dim3(1), dim3(THREADS_PER_BLOCK_X * THREADS_PER_BLOCK_Y), 0, 0,
gpuTransposeMatrix, gpuMatrix, WIDTH);
// Memory transfer from device to host
hipMemcpy(TransposeMatrix, gpuTransposeMatrix, NUM * sizeof(T), hipMemcpyDeviceToHost);
// CPU MatrixTranspose computation
matrixTransposeCPUReference(cpuTransposeMatrix, Matrix, WIDTH);
// verify the results
errors = 0;
double eps = 1.0E-6;
for (i = 0; i < NUM; i++) {
if (TransposeMatrix[i] != cpuTransposeMatrix[i]) {
errors++;
}
}
// free the resources on device side
hipFree(gpuMatrix);
hipFree(gpuTransposeMatrix);
// free the resources on host side
free(Matrix);
free(TransposeMatrix);
free(cpuTransposeMatrix);
if (errors != 0) {
failed("Mismatch present");
}
}
int main() {
runTest<int>();
runTest<float>();
runTest<long>();
runTest<long long>();
passed();
}
tests/src/kernel/hipShflUpDownTest.cpp
+102
Melihat File
@@ -0,0 +1,102 @@
/*
Copyright (c) 2015-present 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.
*/
/* HIT_START
* BUILD: %t %s ../test_common.cpp
* TEST: %t
* HIT_END
*/
#include <iostream>
#include <hip/hip_runtime.h>
#include "test_common.h"
template <typename T>
__global__ void shflDownSum(T* a, int size) {
T val = a[threadIdx.x];
for (int i = size / 2; i > 0; i /= 2) {
val += __shfl_down(val, i, size);
}
a[threadIdx.x] = val;
}
template <typename T>
__global__ void shflUpSum(T* a, int size) {
T val = a[threadIdx.x];
for (int i = size / 2; i > 0; i /= 2) {
val += __shfl_up(val, i, size);
}
a[threadIdx.x] = val;
}
template <typename T>
void runTestShflUp() {
const int size = 32;
T a[size];
T cpuSum = 0;
for (int i = 0; i < size; i++) {
a[i] = i;
cpuSum += a[i];
}
T* d_a;
hipMalloc(&d_a, sizeof(T) * size);
hipMemcpy(d_a, &a, sizeof(T) * size, hipMemcpyDefault);
hipLaunchKernelGGL(shflUpSum<T>, 1, size, 0, 0, d_a, size);
hipMemcpy(&a, d_a, sizeof(T) * size, hipMemcpyDefault);
if (a[size - 1] != cpuSum) {
hipFree(d_a);
failed("Shfl Up Sum did not match.");
}
hipFree(d_a);
}
template <typename T>
void runTestShflDown() {
const int size = 32;
T a[size];
T cpuSum = 0;
for (int i = 0; i < size; i++) {
a[i] = i;
cpuSum += a[i];
}
T* d_a;
hipMalloc(&d_a, sizeof(T) * size);
hipMemcpy(d_a, &a, sizeof(T) * size, hipMemcpyDefault);
hipLaunchKernelGGL(shflDownSum<T>, 1, size, 0, 0, d_a, size);
hipMemcpy(&a, d_a, sizeof(T) * size, hipMemcpyDefault);
if (a[0] != cpuSum) {
hipFree(d_a);
failed("Shfl Up Sum did not match.");
}
hipFree(d_a);
}
int main() {
runTestShflUp<int>();
runTestShflUp<float>();
runTestShflUp<long>();
runTestShflUp<long long>();
runTestShflDown<int>();
runTestShflDown<float>();
runTestShflDown<long>();
runTestShflDown<long long>();
passed();
}