AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Re-sync with upstream. Add integer abs.

Browse Source
This commit is contained in:
Alex Voicu
2018-05-31 16:38:00 +01:00
parent 59db16fd36 449ddba0aa
commit ab4b2a650b
30 changed files with 1573 additions and 346 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/src/deviceLib/hipMathFunctions.cpp
+2 -2
View File
@@ -21,7 +21,7 @@ THE SOFTWARE.
*/
/* HIT_START
* BUILD: %t %s ../test_common.cpp
* BUILD: %t %s ../test_common.cpp EXCLUDE_HIP_PLATFORM nvcc
* RUN: %t
* HIT_END
*/
@@ -29,7 +29,7 @@ THE SOFTWARE.
#include "hip/hip_runtime.h"
#include "test_common.h"
#if __HIP_ARCH_GFX803__ || __HIP_ARCH_GFX900__
#if __HIP_ARCH_GFX803__ || __HIP_ARCH_GFX900__ || __HIP_ARCH_GFX906__
__global__ void kernel_abs_int64(hipLaunchParm lp, long long *input, long long *output) {
int tx = threadIdx.x;
tests/src/deviceLib/hipTestHalf.cpp
+2 -2
View File
@@ -18,7 +18,7 @@ THE SOFTWARE.
*/
/* HIT_START
* BUILD: %t %s ../test_common.cpp
* BUILD: %t %s ../test_common.cpp EXCLUDE_HIP_PLATFORM nvcc
* RUN: %t
* HIT_END
*/
@@ -32,7 +32,7 @@ THE SOFTWARE.
#define HALF_SIZE 64 * sizeof(__half)
#define HALF2_SIZE 64 * sizeof(__half2)
#if __HIP_ARCH_GFX803__ || __HIP_ARCH_GFX900__
#if __HIP_ARCH_GFX803__ || __HIP_ARCH_GFX900__ || __HIP_ARCH_GFX906__
__global__ void __halfMath(hipLaunchParm lp, __half* A, __half* B, __half* C) {
int tx = threadIdx.x;
tests/src/deviceLib/hip_mbcnt.cpp
+128
View File
@@ -0,0 +1,128 @@
/*
Copyright (c) 2015-2018 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 EXCLUDE_HIP_PLATFORM nvcc
* RUN: %t
* HIT_END
*/
#include <assert.h>
#include <stdio.h>
#include <algorithm>
#include <stdlib.h>
#include <iostream>
#include <hip/hip_runtime.h>
#include <hip/device_functions.h>
#define HIP_ASSERT(x) (assert((x) == hipSuccess))
__global__ void HIP_kernel(hipLaunchParm lp, unsigned int* mbcnt_lo, unsigned int* mbcnt_hi, unsigned int* lane_id) {
int x = blockDim.x * blockIdx.x + threadIdx.x;
mbcnt_lo[x] = __mbcnt_lo(0xFFFFFFFF, 0);
mbcnt_hi[x] = __mbcnt_hi(0xFFFFFFFF, 0);
lane_id[x] = __lane_id();
}
using namespace std;
int main() {
unsigned int* device_mbcnt_lo;
unsigned int* device_mbcnt_hi;
unsigned int* device_lane_id;
hipDeviceProp_t devProp;
hipGetDeviceProperties(&devProp, 0);
cout << " System minor " << devProp.minor << endl;
cout << " System major " << devProp.major << endl;
cout << " agent prop name " << devProp.name << endl;
cout << "hip Device prop succeeded " << endl;
constexpr unsigned int wave_size = 64;
constexpr unsigned int num_waves_per_block = 2;
constexpr unsigned int num_threads_per_block = wave_size * num_waves_per_block;
constexpr unsigned int num_blocks = 2;
constexpr unsigned int num_threads = num_threads_per_block * num_blocks;
constexpr size_t buffer_size = num_threads * sizeof(unsigned int);
HIP_ASSERT(hipMalloc((void**)&device_mbcnt_lo, buffer_size));
HIP_ASSERT(hipMalloc((void**)&device_mbcnt_hi, buffer_size));
HIP_ASSERT(hipMalloc((void**)&device_lane_id, buffer_size));
hipLaunchKernel(HIP_kernel, dim3(num_blocks),
dim3(num_threads_per_block), 0, 0, device_mbcnt_lo, device_mbcnt_hi, device_lane_id);
unsigned int* host_mbcnt_lo = (unsigned int*) malloc(buffer_size);
unsigned int* host_mbcnt_hi = (unsigned int*) malloc(buffer_size);
unsigned int* host_lane_id = (unsigned int*) malloc(buffer_size);
HIP_ASSERT(hipMemcpy(host_mbcnt_lo, device_mbcnt_lo, buffer_size, hipMemcpyDeviceToHost));
HIP_ASSERT(hipMemcpy(host_mbcnt_hi, device_mbcnt_hi, buffer_size, hipMemcpyDeviceToHost));
HIP_ASSERT(hipMemcpy(host_lane_id, device_lane_id, buffer_size, hipMemcpyDeviceToHost));
// verify the results
int mbcnt_lo_errors = 0;
int mbcnt_hi_errors = 0;
int lane_id_errors = 0;
for (unsigned int i = 0; i < num_threads; i++) {
unsigned int this_lane_id = i % wave_size;
unsigned int this_mbcnt_lo = this_lane_id >= 32 ? 32 : this_lane_id;
unsigned int this_mbcnt_hi = this_lane_id < 32 ? 0 : (this_lane_id - 22);
if (host_mbcnt_lo[i] != this_mbcnt_lo)
mbcnt_lo_errors++;
if (host_mbcnt_hi[i] != this_mbcnt_hi)
mbcnt_hi_errors++;
if (host_lane_id[i] != this_lane_id)
lane_id_errors++;
}
if (mbcnt_lo_errors == 0)
cout << "__mbcnt_lo() PASSED!" << endl;
else
cout << "__mbcnt_lo() FAILED!" << endl;
if (mbcnt_hi_errors == 0)
cout << "__mbcnt_hi() PASSED!" << endl;
else
cout << "__mbcnt_hi() FAILED!" << endl;
if (lane_id_errors == 0)
cout << "__lane_id() PASSED!" << endl;
else
cout << "__lane_id() FAILED!" << endl;
HIP_ASSERT(hipFree(device_mbcnt_lo));
HIP_ASSERT(hipFree(device_mbcnt_hi));
HIP_ASSERT(hipFree(device_lane_id));
free(host_mbcnt_lo);
free(host_mbcnt_hi);
free(host_lane_id);
return mbcnt_lo_errors + mbcnt_hi_errors + lane_id_errors;
}
tests/src/runtimeApi/memory/hipMemset3D.cpp
+98
View File
@@ -0,0 +1,98 @@
/*
Copyright (c) 2015-2016 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.
*/
// Simple test for memset.
// Also serves as a template for other tests.
/* HIT_START
* BUILD: %t %s ../../test_common.cpp
* //Small copy
* RUN: %t -N 10 --memsetval 0x42
* HIT_END
*/
#include "hip/hip_runtime.h"
#include "test_common.h"
bool testhipMemset3D(int memsetval,int p_gpuDevice)
{
size_t numH = 256;
size_t numW = 256;
size_t depth = 1;
size_t pitch_A;
size_t width = numW * sizeof(char);
size_t sizeElements = width * numH * depth;
size_t elements = numW* numH* depth;
printf ("testhipMemset3D memsetval=%2x device=%d\n", memsetval, p_gpuDevice);
char *A_d;
char *A_h;
bool testResult = true;
hipExtent extent = make_hipExtent(width, numH, depth);
hipPitchedPtr devPitchedPtr;
HIPCHECK(hipMalloc3D(&devPitchedPtr, extent));
A_h = (char*)malloc(sizeElements);
HIPASSERT(A_h != NULL);
for (size_t i=0; i<elements; i++) {
A_h[i] = 1;
}
hipStream_t stream;
//HIPCHECK(hipStreamCreate(&stream));
HIPCHECK ( hipMemset3D( devPitchedPtr, memsetval, extent) );
//HIPCHECK ( hipMemcpy3D(A_h, width, A_d, pitch_A, numW, numH, hipMemcpyDeviceToHost));
hipMemcpy3DParms myparms = {0};
myparms.srcPos = make_hipPos(0,0,0);
myparms.dstPos = make_hipPos(0,0,0);
myparms.dstPtr = make_hipPitchedPtr(A_h, width , numW, numH);
myparms.srcPtr = devPitchedPtr;
//myparms.dstArray = cuArray;
myparms.extent = extent;
#ifdef __HIP_PLATFORM_NVCC__
myparms.kind = hipMemcpyKindToCudaMemcpyKind(hipMemcpyDeviceToHost);
#else
myparms.kind = hipMemcpyDeviceToHost;
#endif
hipMemcpy3D(&myparms);
for (int i=0; i<elements; i++) {
if (A_h[i] != memsetval) {
testResult = false;
printf("mismatch at index:%d computed:%02x, memsetval:%02x\n", i, (int)A_h[i], (int)memsetval);
break;
}
}
hipFree(A_d);
free(A_h);
return testResult;
}
int main(int argc, char *argv[])
{
HipTest::parseStandardArguments(argc, argv, true);
bool testResult = false;
HIPCHECK(hipSetDevice(p_gpuDevice));
testResult = testhipMemset3D(memsetval, p_gpuDevice);
passed();
}
tests/src/runtimeApi/module/hipFuncGetAttributes.cpp
+53
View File
@@ -0,0 +1,53 @@
/*
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 NVCC_OPTIONS -std=c++11
* RUN: %t
* HIT_END
*/
#include <hip/hip_runtime_api.h>
#include <iostream>
#include "test_common.h"
__global__
void fn(float* px, float* py)
{
bool a[42];
__shared__ double b[69];
for (auto&& x : b) x = *py++;
for (auto&& x : a) x = *px++ > 0.0;
for (auto&& x : a) if (x) *--py = *--px;
}
int main() {
hipInit(0);
hipFuncAttributes attr{};
auto r = hipFuncGetAttributes(&attr, reinterpret_cast<const void*>(&fn));
if (r != hipSuccess || attr.maxThreadsPerBlock == 0) {
failed("Failed to read attributes.");
}
passed();
}
tests/src/texture/hipBindTexRef1DFetch.cpp
+95
View File
@@ -0,0 +1,95 @@
/*
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
* RUN: %t
* HIT_END
*/
#include "hip/hip_runtime.h"
#include "test_common.h"
#define N 512
texture<float, 1, hipReadModeElementType> tex;
__global__ void kernel(float *out) {
int x = blockIdx.x * blockDim.x + threadIdx.x;
if(x<N){
out[x] = tex1Dfetch(tex, x);
}
}
int runTest(void);
int main(int argc, char **argv) {
int testResult = runTest();
if (testResult) {
passed();
} else {
exit(EXIT_FAILURE);
}
}
int runTest() {
int testResult = 1;
float *texBuf;
float val[N], output[N];
size_t size = 0;
float *devBuf;
for (int i = 0; i < N; i++) {
val[i] = (float)i;
output[i] = 0.0;
}
hipChannelFormatDesc chanDesc =
hipCreateChannelDesc(32, 0, 0, 0, hipChannelFormatKindFloat);
HIPCHECK(hipMalloc(&texBuf, N * sizeof(float)));
HIPCHECK(hipMalloc(&devBuf, N * sizeof(float)));
HIPCHECK(hipMemcpy(texBuf, val, N * sizeof(float), hipMemcpyHostToDevice));
tex.addressMode[0] = hipAddressModeClamp;
tex.addressMode[1] = hipAddressModeClamp;
tex.filterMode = hipFilterModePoint;
tex.normalized = 0;
HIPCHECK(hipBindTexture(&size, tex, (void *)texBuf, chanDesc, N * sizeof(float)));
dim3 dimBlock(64, 1, 1);
dim3 dimGrid(N / dimBlock.x, 1, 1);
hipLaunchKernelGGL(kernel, dim3(dimGrid), dim3(dimBlock), 0, 0, devBuf);
HIPCHECK(hipDeviceSynchronize());
HIPCHECK(hipMemcpy(output, devBuf, N * sizeof(float), hipMemcpyDeviceToHost));
for (int i = 0; i < N; i++) {
if (output[i] != val[i]) {
testResult = 0;
break;
}
}
HIPCHECK(hipUnbindTexture(&tex));
HIPCHECK(hipFree(texBuf));
HIPCHECK(hipFree(devBuf));
return testResult;
}
tests/src/texture/hipTextureObj1DFetch.cpp
+99
View File
@@ -0,0 +1,99 @@
/*
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
* RUN: %t
* HIT_END
*/
#include "hip/hip_runtime.h"
#include "test_common.h"
#define N 512
__global__ void tex1dKernel(float *val, hipTextureObject_t obj) {
int k = blockIdx.x * blockDim.x + threadIdx.x;
if (k < N)
val[k] = tex1Dfetch<float>(obj, k);
}
int runTest(void);
int main(int argc, char **argv) {
int testResult = runTest();
if(testResult) {
passed();
} else {
exit(EXIT_FAILURE);
}
}
int runTest() {
int testResult = 1;
// Allocating the required buffer on gpu device
float *texBuf, *texBufOut;
float val[N], output[N];
for (int i = 0; i < N; i++) {
val[i] = (i + 1) * (i + 1);
output[i] = 0.0;
}
HIPCHECK(hipMalloc(&texBuf, N * sizeof(float)));
HIPCHECK(hipMalloc(&texBufOut, N * sizeof(float)));
HIPCHECK(hipMemcpy(texBuf, val, N * sizeof(float), hipMemcpyHostToDevice));
HIPCHECK(hipMemset(texBufOut, 0, N * sizeof(float)));
hipResourceDesc resDescLinear;
memset(&resDescLinear, 0, sizeof(resDescLinear));
resDescLinear.resType = hipResourceTypeLinear;
resDescLinear.res.linear.devPtr = texBuf;
resDescLinear.res.linear.desc = hipCreateChannelDesc(32, 0, 0, 0, hipChannelFormatKindFloat);
resDescLinear.res.linear.sizeInBytes = N * sizeof(float);
hipTextureDesc texDesc;
memset(&texDesc, 0, sizeof(texDesc));
texDesc.readMode = hipReadModeElementType;
// Creating texture object
hipTextureObject_t texObj = 0;
HIPCHECK(hipCreateTextureObject(&texObj, &resDescLinear, &texDesc, NULL));
dim3 dimBlock(64, 1, 1);
dim3 dimGrid(N / dimBlock.x, 1, 1);
hipLaunchKernelGGL(tex1dKernel, dim3(dimGrid), dim3(dimBlock), 0, 0,
texBufOut, texObj);
HIPCHECK(hipDeviceSynchronize());
HIPCHECK(hipMemcpy(output, texBufOut, N * sizeof(float), hipMemcpyDeviceToHost));
for(int i = 0; i < N; i++)
if (output[i] != val[i]) {
testResult = 0;
break;
}
HIPCHECK(hipDestroyTextureObject(texObj));
HIPCHECK(hipFree(texBuf));
HIPCHECK(hipFree(texBufOut));
return testResult;
}
tests/src/texture/hipTextureObj2D.cpp
+108 -108
View File
@@ -1,108 +1,108 @@
/* HIT_START
* BUILD: %t %s ../test_common.cpp
* RUN: %t
* HIT_END
*/
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <hip/hip_runtime.h>
#include "test_common.h"
bool testResult = true;
__global__ void tex2DKernel(float* outputData, hipTextureObject_t textureObject, int width,
int height) {
int x = blockIdx.x * blockDim.x + threadIdx.x;
int y = blockIdx.y * blockDim.y + threadIdx.y;
outputData[y * width + x] = tex2D<float>(textureObject, x, y);
}
void runTest(int argc, char** argv);
int main(int argc, char** argv) {
runTest(argc, argv);
if (testResult) {
passed();
} else {
exit(EXIT_FAILURE);
}
}
void runTest(int argc, char** argv) {
unsigned int width = 256;
unsigned int height = 256;
unsigned int size = width * height * sizeof(float);
float* hData = (float*)malloc(size);
memset(hData, 0, size);
for (int i = 0; i < height; i++) {
for (int j = 0; j < width; j++) {
hData[i * width + j] = i * width + j;
}
}
printf("hData: ");
for (int i = 0; i < 64; i++) {
printf("%f ", hData[i]);
}
printf("\n");
hipChannelFormatDesc channelDesc = hipCreateChannelDesc(32, 0, 0, 0, hipChannelFormatKindFloat);
hipArray* hipArray;
hipMallocArray(&hipArray, &channelDesc, width, height);
hipMemcpyToArray(hipArray, 0, 0, hData, size, hipMemcpyHostToDevice);
struct hipResourceDesc resDesc;
memset(&resDesc, 0, sizeof(resDesc));
resDesc.resType = hipResourceTypeArray;
resDesc.res.array.array = hipArray;
// Specify texture object parameters
struct hipTextureDesc texDesc;
memset(&texDesc, 0, sizeof(texDesc));
texDesc.addressMode[0] = hipAddressModeWrap;
texDesc.addressMode[1] = hipAddressModeWrap;
texDesc.filterMode = hipFilterModePoint;
texDesc.readMode = hipReadModeElementType;
texDesc.normalizedCoords = 0;
// Create texture object
hipTextureObject_t textureObject = 0;
hipCreateTextureObject(&textureObject, &resDesc, &texDesc, NULL);
float* dData = NULL;
hipMalloc((void**)&dData, size);
dim3 dimBlock(16, 16, 1);
dim3 dimGrid(width / dimBlock.x, height / dimBlock.y, 1);
hipLaunchKernelGGL(tex2DKernel, dim3(dimGrid), dim3(dimBlock), 0, 0, dData, textureObject,
width, height);
hipDeviceSynchronize();
float* hOutputData = (float*)malloc(size);
memset(hOutputData, 0, size);
hipMemcpy(hOutputData, dData, size, hipMemcpyDeviceToHost);
printf("dData: ");
for (int i = 0; i < 64; i++) {
printf("%f ", hOutputData[i]);
}
printf("\n");
for (int i = 0; i < height; i++) {
for (int j = 0; j < width; j++) {
if (hData[i * width + j] != hOutputData[i * width + j]) {
printf("Difference [ %d %d ]:%f ----%f\n", i, j, hData[i * width + j],
hOutputData[i * width + j]);
testResult = false;
break;
}
}
}
hipDestroyTextureObject(textureObject);
hipFree(dData);
hipFreeArray(hipArray);
}
/* HIT_START
* BUILD: %t %s ../test_common.cpp
* RUN: %t
* HIT_END
*/
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <hip/hip_runtime.h>
#include "test_common.h"
__global__ void tex2DKernel(float* outputData, hipTextureObject_t textureObject, int width,
int height) {
int x = blockIdx.x * blockDim.x + threadIdx.x;
int y = blockIdx.y * blockDim.y + threadIdx.y;
outputData[y * width + x] = tex2D<float>(textureObject, x, y);
}
int runTest(int argc, char** argv);
int main(int argc, char** argv) {
int testResult = runTest(argc, argv);
if (testResult) {
passed();
} else {
exit(EXIT_FAILURE);
}
}
int runTest(int argc, char** argv) {
int testResult = 1;
unsigned int width = 256;
unsigned int height = 256;
unsigned int size = width * height * sizeof(float);
float* hData = (float*)malloc(size);
memset(hData, 0, size);
for (int i = 0; i < height; i++) {
for (int j = 0; j < width; j++) {
hData[i * width + j] = i * width + j;
}
}
printf("hData: ");
for (int i = 0; i < 64; i++) {
printf("%f ", hData[i]);
}
printf("\n");
hipChannelFormatDesc channelDesc = hipCreateChannelDesc(32, 0, 0, 0, hipChannelFormatKindFloat);
hipArray* hipArray;
hipMallocArray(&hipArray, &channelDesc, width, height);
hipMemcpyToArray(hipArray, 0, 0, hData, size, hipMemcpyHostToDevice);
struct hipResourceDesc resDesc;
memset(&resDesc, 0, sizeof(resDesc));
resDesc.resType = hipResourceTypeArray;
resDesc.res.array.array = hipArray;
// Specify texture object parameters
struct hipTextureDesc texDesc;
memset(&texDesc, 0, sizeof(texDesc));
texDesc.addressMode[0] = hipAddressModeWrap;
texDesc.addressMode[1] = hipAddressModeWrap;
texDesc.filterMode = hipFilterModePoint;
texDesc.readMode = hipReadModeElementType;
texDesc.normalizedCoords = 0;
// Create texture object
hipTextureObject_t textureObject = 0;
hipCreateTextureObject(&textureObject, &resDesc, &texDesc, NULL);
float* dData = NULL;
hipMalloc((void**)&dData, size);
dim3 dimBlock(16, 16, 1);
dim3 dimGrid(width / dimBlock.x, height / dimBlock.y, 1);
hipLaunchKernelGGL(tex2DKernel, dim3(dimGrid), dim3(dimBlock), 0, 0, dData, textureObject,
width, height);
hipDeviceSynchronize();
float* hOutputData = (float*)malloc(size);
memset(hOutputData, 0, size);
hipMemcpy(hOutputData, dData, size, hipMemcpyDeviceToHost);
printf("dData: ");
for (int i = 0; i < 64; i++) {
printf("%f ", hOutputData[i]);
}
printf("\n");
for (int i = 0; i < height; i++) {
for (int j = 0; j < width; j++) {
if (hData[i * width + j] != hOutputData[i * width + j]) {
printf("Difference [ %d %d ]:%f ----%f\n", i, j, hData[i * width + j],
hOutputData[i * width + j]);
testResult = 0;
break;
}
}
}
hipDestroyTextureObject(textureObject);
hipFree(dData);
hipFreeArray(hipArray);
return testResult;
}
tests/src/texture/hipTextureRef2D.cpp
+93 -93
View File
@@ -1,93 +1,93 @@
/* HIT_START
* BUILD: %t %s ../test_common.cpp
* RUN: %t
* HIT_END
*/
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <hip/hip_runtime.h>
#include "test_common.h"
texture<float, 2, hipReadModeElementType> tex;
bool testResult = true;
__global__ void tex2DKernel(float* outputData,
int width, int height) {
int x = blockIdx.x * blockDim.x + threadIdx.x;
int y = blockIdx.y * blockDim.y + threadIdx.y;
outputData[y * width + x] = tex2D(tex, x, y);
}
void runTest(int argc, char** argv);
int main(int argc, char** argv) {
runTest(argc, argv);
if (testResult) {
passed();
} else {
exit(EXIT_FAILURE);
}
}
void runTest(int argc, char** argv) {
unsigned int width = 256;
unsigned int height = 256;
unsigned int size = width * height * sizeof(float);
float* hData = (float*)malloc(size);
memset(hData, 0, size);
for (int i = 0; i < height; i++) {
for (int j = 0; j < width; j++) {
hData[i * width + j] = i * width + j;
}
}
printf("hData: ");
for (int i = 0; i < 64; i++) {
printf("%f ", hData[i]);
}
printf("\n");
hipChannelFormatDesc channelDesc = hipCreateChannelDesc(32, 0, 0, 0, hipChannelFormatKindFloat);
hipArray* hipArray;
hipMallocArray(&hipArray, &channelDesc, width, height);
hipMemcpyToArray(hipArray, 0, 0, hData, size, hipMemcpyHostToDevice);
tex.addressMode[0] = hipAddressModeWrap;
tex.addressMode[1] = hipAddressModeWrap;
tex.filterMode = hipFilterModePoint;
tex.normalized = 0;
hipBindTextureToArray(tex, hipArray, channelDesc);
float* dData = NULL;
hipMalloc((void**)&dData, size);
dim3 dimBlock(16, 16, 1);
dim3 dimGrid(width / dimBlock.x, height / dimBlock.y, 1);
hipLaunchKernelGGL(tex2DKernel, dim3(dimGrid), dim3(dimBlock), 0, 0, dData, width, height);
hipDeviceSynchronize();
float* hOutputData = (float*)malloc(size);
memset(hOutputData, 0, size);
hipMemcpy(hOutputData, dData, size, hipMemcpyDeviceToHost);
printf("dData: ");
for (int i = 0; i < 64; i++) {
printf("%f ", hOutputData[i]);
}
printf("\n");
for (int i = 0; i < height; i++) {
for (int j = 0; j < width; j++) {
if (hData[i * width + j] != hOutputData[i * width + j]) {
printf("Difference [ %d %d ]:%f ----%f\n", i, j, hData[i * width + j],
hOutputData[i * width + j]);
testResult = false;
break;
}
}
}
hipFree(dData);
hipFreeArray(hipArray);
}
/* HIT_START
* BUILD: %t %s ../test_common.cpp
* RUN: %t
* HIT_END
*/
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <hip/hip_runtime.h>
#include "test_common.h"
texture<float, 2, hipReadModeElementType> tex;
__global__ void tex2DKernel(float* outputData,
int width, int height) {
int x = blockIdx.x * blockDim.x + threadIdx.x;
int y = blockIdx.y * blockDim.y + threadIdx.y;
outputData[y * width + x] = tex2D(tex, x, y);
}
int runTest(int argc, char** argv);
int main(int argc, char** argv) {
int testResult = runTest(argc, argv);
if (testResult) {
passed();
} else {
exit(EXIT_FAILURE);
}
}
int runTest(int argc, char** argv) {
int testResult = 1;
unsigned int width = 256;
unsigned int height = 256;
unsigned int size = width * height * sizeof(float);
float* hData = (float*)malloc(size);
memset(hData, 0, size);
for (int i = 0; i < height; i++) {
for (int j = 0; j < width; j++) {
hData[i * width + j] = i * width + j;
}
}
printf("hData: ");
for (int i = 0; i < 64; i++) {
printf("%f ", hData[i]);
}
printf("\n");
hipChannelFormatDesc channelDesc = hipCreateChannelDesc(32, 0, 0, 0, hipChannelFormatKindFloat);
hipArray* hipArray;
hipMallocArray(&hipArray, &channelDesc, width, height);
hipMemcpyToArray(hipArray, 0, 0, hData, size, hipMemcpyHostToDevice);
tex.addressMode[0] = hipAddressModeWrap;
tex.addressMode[1] = hipAddressModeWrap;
tex.filterMode = hipFilterModePoint;
tex.normalized = 0;
hipBindTextureToArray(tex, hipArray, channelDesc);
float* dData = NULL;
hipMalloc((void**)&dData, size);
dim3 dimBlock(16, 16, 1);
dim3 dimGrid(width / dimBlock.x, height / dimBlock.y, 1);
hipLaunchKernelGGL(tex2DKernel, dim3(dimGrid), dim3(dimBlock), 0, 0, dData, width, height);
hipDeviceSynchronize();
float* hOutputData = (float*)malloc(size);
memset(hOutputData, 0, size);
hipMemcpy(hOutputData, dData, size, hipMemcpyDeviceToHost);
printf("dData: ");
for (int i = 0; i < 64; i++) {
printf("%f ", hOutputData[i]);
}
printf("\n");
for (int i = 0; i < height; i++) {
for (int j = 0; j < width; j++) {
if (hData[i * width + j] != hOutputData[i * width + j]) {
printf("Difference [ %d %d ]:%f ----%f\n", i, j, hData[i * width + j],
hOutputData[i * width + j]);
testResult = 0;
break;
}
}
}
hipFree(dData);
hipFreeArray(hipArray);
return testResult;
}